Week 11 Discussion - Withdrawal Designs/ Single Case

gclaudia

SingleSubjectResearch-ApplicationsinEducationalClinicalSettings-StephenB_Richards.pdf

Home >Applied Sciences homework help >Week 11 Discussion - Withdrawal Designs/ Single Case

2ND EDITION

Single Subject Research Applications in Educational and Clinical Settings

Stephen B. Richards Associate Professor Department of Teacher Education University of Dayton Dayton, Ohio

Ronald L. Taylor Professor Exceptional Student Education Florida Atlantic University Boca Raton, Florida

Rangasamy Ramasamy Professor Exceptional Student Education Florida Atlantic University Boca Raton, Florida

Australia • Brazil • Japan • Korea • Mexico • Singapore • Spain • United Kingdom • United States

Copyright 2012 Cengage Learning. All Rights Reserved. May not be copied, scanned, or duplicated, in whole or in part. Due to electronic rights, some third party content may be suppressed from the eBook and/or eChapter(s). Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. Cengage Learning reserves the right to remove additional content at any time if subsequent rights restrictions require it.

This is an electronic version of the print textbook. Due to electronic rights restrictions, some third party content may be suppressed. Editorial review has deemed that any suppressed content does not materially affect the overall learning experience. The publisher reserves the right to remove content from this title at any time if subsequent rights restrictions require it. For valuable information on pricing, previous editions, changes to current editions, and alternate formats, please visit www.cengage.com/highered to search by ISBN#, author, title, or keyword for materials in your areas of interest.

Single Subject Research: Applications in Educational and Clinical Settings, Second Edition Stephen B. Richards, Ronald L. Taylor and Rangasamy Ramasamy

Editor-in-Chief: Linda Ganster

Executive Editor: Mark Kerr

Developmental Editor: Julie Martinez

Assistant Editor: Joshua Taylor

Editorial Assistant: Greta Lindquist

Senior Media Editor: Ashley Cronin

Executive Brand Manager: Melissa Larmon

Market Development Manager: Kara Kindstrom

Art and Cover Direction, Production Services, and Composition: PreMediaGlobal

Manufacturing Planner: Doug Bertke

Rights Acquisitions Specialist: Dean Dauphinais

Cover Image: Alex Slobodkin/iStockphoto

ALL RIGHTS RESERVED. No part of this work covered by the copyright herein may be reproduced, transmitted, stored, or used in any form or by any means graphic, electronic, or mechanical, including but not limited to photocopying, recording, scanning, digitizing, taping, Web distribution, information networks, or information storage and retrieval systems, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without the prior written permission of the publisher.

For product information and technology assistance, contact us at Cengage Learning Customer & Sales Support, 1-800-354-9706.

For permission to use material from this text or product, submit all requests online at www.cengage.com/permissions. Further permissions questions can be e-mailed to

[email protected].

Library of Congress Control Number: 2012946793

ISBN-13: 978-1-133-96317-2

ISBN-10: 1-133-96317-X

Wadsworth 20 Davis Drive, Belmont, CA 94002-3098 USA

Cengage Learning is a leading provider of customized learning solu- tions with office locations around the globe, including Singapore, the United Kingdom, Australia, Mexico, Brazil, and Japan. Locate your local office at www.cengage.com/global.

Cengage Learning products are represented in Canada by Nelson Education, Ltd.

To learn more about Wadsworth, visit www.cengage.com/wadsworth

Purchase any of our products at your local college store or at our preferred online store www.cengagebrain.com.

Printed in the United States of America 1 2 3 4 5 6 7 16 15 14 13 12

Contents

Foreword xv

PART 1 Conducting Single Subject Research: Issues and Procedures 1

CHAPTER 1 Historical Perspectives and Important Concepts in Single Subject Research 3 Historical Perspectives in Single Subject Research 4

Basic Concepts and Definitions of Terms 7

Independent, Dependent, and Extraneous Variables 9

Baseline, Intervention, and Follow-Up Phases 11

Notations 14

The X-Y or Line Graph 16

The Dependent Variable 16

The X-Axis 18

The Independent Variable and Phase Change Lines 19

Data Paths 19

The Legend 20

Summary 21

Key Concepts/Terms 21

Possible Answers to Check It Out 22

CHAPTER 2 Methods for Changing Target Behaviors 25 Methods to Increase Behavior 26

Positive Reinforcement 29

Negative Reinforcement 30

Premack Principle 32

Shaping 33

Methods to Maintain Behavior 33

Reinforcer Menus 33

Satiation 34

Primary, Secondary, and Generalized Reinforcers 35

Quality of Reinforcers 36

iii

Reinforcement Schedules 37

Generalization of Target Behaviors 40

Methods to Decrease Behavior 41

Positive Punishment 42

Negative Punishment 42

Extinction 43

Differential Reinforcement 44

Other Methods to Decrease Behavior 47

Summary 48

Key Concepts/Terms 49

Possible Answers to Check It Out 51

CHAPTER 3 Methods for Recording Behaviors 53 The Importance of Observable, Measurable Behavior 54

Interobserver Agreement 54

Choosing a Recording Procedure 57

Event-Based Methods for Recording and Reporting Behavior 57

Frequency Recording 58

Rate 60

Interval Recording 61

Evaluating Permanent Products 66

Time-Based Methods for Recording and Reporting Behavior 69

Duration Recording 70

Latency Recording 71

Recording Procedures for Specific Purposes 72

Trials to Criterion Recording 73

Cumulative Recording 73

A Few Words about These Methods 74

Choosing the Appropriate Recording Procedure 75

Summary 76

Key Concepts/Terms 76

Possible Answers to Check It Out 77

CHAPTER 4 Issues in Single Subject Research 91 The Concepts of Prediction, Verification, and Replication 92

Prediction 93

Verification 94

Replication 94

iv CONTENTS

Reliability and Validity 96

Reliability 97

Validity 100

Ethics 107

Key Concepts/Terms 111

Possible Answers to Check It Out 112

PART 2 Overview and Application of Single Subject Designs 117

CHAPTER 5 Overview of Withdrawal Designs 119 A and B Designs 121

The A-B Design 122

The A-B Design and Action Research 123

Mechanics of the Withdrawal Designs 125

The A-B-A Design 125

The A-B-A-B Design 128

Prediction, Verification, and Replication 129

Advantages of the Withdrawal Design 130

Disadvantages of the Withdrawal Design 131

Adaptations of the Typical Withdrawal Design 132

The B-A-B Design (No Initial Baseline) 132

The A-B-A-B-A-B Design (Repeated Withdrawals) 134

Key Concepts/Terms 135

Possible Answers to Check It Out 136

CHAPTER 6 Application of Withdrawal Designs 139 The A-B Design 140

Purpose of the Study 140

Subject 140

Setting 141

Dependent Variable 141

Independent Variable 141

The Design 141

Intervention 141

Obtaining the Data and Plotting the Results 142

Results 142

Why Use an A-B Design? 143

Limitations of the Study 143

Summary 144

CONTENTS v

The A-B-A Design 144

Purpose of the Study 144

Subject 144

Setting 145

Dependent Variables 145

Independent Variable 145

The Design 145

The Intervention 145

Obtaining the Data and Plotting the Results 145

Results 146

Why Use an A-B-A Design? 147

Limitations of the Study 147

Summary 147

The A-B-A-B Design 148

Purpose of the Study 148

Subjects 148

Setting 148

Dependent Variables 148

Independent Variable 149

The Design 149

The Intervention 149

Obtaining the Data and Plotting the Results 149

Results 150

Why Use an A-B-A-B Design? 151

Limitations of the Study 151

Summary 152

The B-A-B Design 152

Purpose of the Study 152

Subjects 152

Setting 152

Dependent Variables 152

Independent Variable 153

Design 153

The Intervention 153

Obtaining the Data and Plotting the Results 154

Results 155

Why Use B-A-B Design? 155

vi CONTENTS

Limitations of the Study 155

Summary 156

Application Practice: Brian and Brittany 157

Brian and Brittany: The Questions 158

Brian and Brittany: The Answers 160

Purpose of the Study 160

Subjects 160

Setting 160

Dependent Variables 160

Independent Variable 160

The Design 161

The Intervention 161

Obtaining the Data and Plotting the Results 161

Results 161

Why Use an A-B-A-B Design? 162

Limitations of the Study 162

Summary 163

CHAPTER 7 Overview of Changing Conditions and Changing Criterion Designs 165 The Changing Conditions Designs 166

The A-B-C Design 166

The A-B-A-C (Multiple Treatment) Design 168

Prediction, Verification, and Replication 170

Advantages and Disadvantages of the Changing Conditions Design 171

Changing Criterion Designs 172

Issues Related to Changing Criterion Designs 175

Prediction, Verification, and Replication 180

Advantages of the Changing Criterion Design 180

Disadvantages of the Changing Criterion Design 181

Key Concepts/Terms 182

Possible Answers to Check It Out 183

CHAPTER 8 Application of Changing Conditions and Changing Criterion Design 185 The A-B-A-C Design 186

Purpose of the Study 186

Subjects 186

Setting 186

CONTENTS vii

Dependent Variable 186

Independent Variables 187

The Design 187

The Intervention 187

Obtaining the Data and Plotting the Results 187

Results 188

Why Use an A-B-A-C Design? 188

Limitations of the Study 189

Summary 189

Application Practice: Kumar 189

Kumar: The Questions 191

Kumar: The Answers 191

Purpose of the Study 191

Subject 191

Setting 191

Dependent Variable 191

Independent Variables 192

The Design 192

The Intervention 192

Obtaining the Data and Plotting the Results 192

Results 192

Why Use an A-B-C Design? 193

Limitations of the Study 194

Summary 194

Changing Criterion Design with a Return to Baseline Phase 195

Purpose of the Study 195

Subjects 195

Setting 195

Dependent Variable 195

Independent Variable 195

The Design 195

The Intervention 196

Obtaining the Data and Plotting the Results 196

Results 196

Why Use Changing Criterion Design with a Return to Baseline? 198

Limitations of the Study 198

Summary 198

viii CONTENTS

Application Practice: Jerry 198

Jerry: The Questions 200

Jerry: The Answers 201

Purpose of the Study 201

Subject 201

Setting 201

Dependent Variable 202

Independent Variable 202

The Design 202

The Intervention 202

Obtaining the Data and Plotting the Results 202

Results 202

Why Use the Changing Criterion Design? 202

Limitations of the Study 203

Summary 203

CHAPTER 9 Overview of Multiple Baseline Designs 205 The Basic Multiple Baseline Design 206

Mechanics of the Multiple Baseline Design 207

Prediction, Verification, and Replication 210

Covariance Among Dependent Variables 211

Advantages of the Multiple Baseline Design 215

Disadvantages of the Multiple Baseline Design 215

The Different Multiple Baseline Designs 216

Multiple Baseline Across Behaviors 216

Multiple Baseline Across Settings 218

Multiple Baseline Across Subjects 221

Adaptations of the Multiple Baseline Design 224

Multiple Probe Design 224

Delayed Multiple Baseline Design 228

Summary 231

Key Concepts/Terms 232

Possible Answers to Check It Out 233

CHAPTER 10 Application of Multiple Baseline Designs 235 Multiple Baseline Across Behaviors 236

Purpose of the Study 236

Subjects 236

Setting 236

CONTENTS ix

Dependent Variables 237

Independent Variable 237

The Design 237

Intervention 237

Obtaining the Data and Plotting the Results 238

Results 238

Why Use a Multiple Baseline Across Behaviors Design? 238

Limitations of the study 242

Summary 243

Multiple Baseline Across Settings 243

Purpose of the Study 243

Subjects 244

Setting 244

Dependent Variable 244

Independent Variable 244

Design 245

Intervention 245

Obtaining the Data and Plotting the Results 245

Results 245

Why Use a Multiple Baseline Across Settings Design? 246

Limitations of the Study 247

Summary 248

Multiple Baseline Across Subjects 248

Purpose of the Study 248

Subjects 248

Setting 248

Dependent Variable 248

Independent variable 249

The Design 249

Intervention 249

Obtaining the Data and Plotting the Results 249

Results 250

Why Use a Multiple Baseline Across Subjects Design? 250

Limitations of the Study 250

Summary 251

Multiple Probe Design 252

Purpose of the Study 252

Subjects 252

x CONTENTS

Setting 252

Dependent Variables 252

Independent Variable 253

The Design 253

Intervention 253

Obtaining the Data and Plotting the Results 253

Results 254

Why Use a Multiple Probe Design? 254

Limitations of the Study 254

Summary 254

Application Practice: Donald 256

Donald: The Questions 258

Donald: The Answers 258

Purpose of the Study 258

Subject 258

Setting 258

Dependent Variables 258

Independent Variable 259

The Design 259

The Intervention 259

Obtaining the Data and Plotting the Results 259

Results 259

Why Use a Multiple Baseline Across Settings Design? 259

Limitations of the Study 261

Summary 261

CHAPTER 11 Overview of Alternating Treatments Designs 263 Alternating Treatments Design with No Baseline 265

Interpreting Data from Alternating Treatments Designs 266

Alternating Treatments Design with a Baseline 269

Alternating Treatments Design with a Baseline and a Final Treatment Phase 271

Prediction, Verification, and Replication 271

Advantages of the Alternating Treatments Design 273

Disadvantages of the Alternating Treatments Design 274

Adaptations of the Alternating Treatments Design 277

Multielement Design 277

Simultaneous Treatments Design 278

Adapted Alternating Treatments Design 279

CONTENTS xi

Key Concepts/Terms 280

Possible Answers to Check It Out 281

CHAPTER 12 Application of Alternating Treatments Designs 283 Alternating Treatments Design with No Baseline (without a no-treatment

condition) 284

Purpose of the Study 284

Subjects 284

Setting 284

Dependent Variables 284

Independent Variables 285

Design 285

Intervention 285

Obtaining the Data and Plotting the Results 285

Results 285

Why Use an Alternating Treatments Design with No Baseline (without a no-treatment design)? 286

Limitations 288

Summary 288

Alternating Treatments Design with No Baseline (with a no-treatments phase) 288

Purpose of the study 289

Subjects 289

Setting 289

Dependent Variables 289

Independent Variables 289

The Design 290

The Intervention 290

Obtaining the Data and Plotting the Results 290

Results 290

Why Use an Alternating Treatments with No Baseline (with a no-treatment phase) Design? 290

Limitations of the Study 292

Summary 292

Alternating Treatments Design with a Baseline 293

Purpose of the Study 293

Subjects 293

Setting 293

Dependent Variables 293

xii CONTENTS

Independent Variables 293

The Design 294

The Intervention 294

Obtaining the Data and Plotting the Results 294

Results 295

Why Use an Alternating Treatments Design with a Baseline? 296

Limitations of the Study 296

Summary 297

Alternating Treatments Design with a Baseline and a Final Treatment Phase 297

Purpose of the Study 297

Subjects 297

Setting 298

Dependent Variable 298

Independent Variables 298

The Design 298

The Intervention 298

Obtaining the Data and Plotting the Results 299

Results 299

Why Use an Alternating Treatments Design with a Baseline and a Final Treatment Phase? 300

Limitations of the Study 300

Summary 300

Application Practice: Latisha 301

Latisha: The Questions 302

Latisha: The Answers 303

Purpose of the Study 303

Subjects 303

Setting 303

Dependent Variable 303

Independent Variables 303

The Design 303

The Intervention 304

Obtaining the Data and Plotting the Results 304

Results 304

Why Use an Alternating Treatments Design with a Baseline and Final Treatment Phase? 304

Limitations of the Study 304

Summary 305

CONTENTS xiii

PART 3 Analyzing Results from Single Subject Studies 307

CHAPTER 13 Methods for Analyzing Data 309 Visual Analysis 310

When to Use Visual Analysis 311

Applying Visual Analysis within Phases 312

Applying Visual Analysis Across Phases 317

Advantages of Visual Analysis 321

Limitations to Visual Analysis 321

Statistical Analysis 322

When to Use Statistical Analysis 322

How to Use Statistical Analysis 323

Statistical Procedures 324

Qualitative Analysis 329

When to Use Qualitative Analysis 330

How to Use Qualitative Analysis 331

Limitations of Qualitative Methods 334

Key Concepts/Terms 335

Possible Answers to Check It Out 337

Index 341

xiv CONTENTS

Foreword

In this second edition, the reader will note changes from the first edition.First, the changing conditions and changing criterion chapters have beencombined into a single chapter. Second, the withdrawal designs chapter is now a stand-alone one. Third, references have been updated in many instances, but foundational studies and research have been retained. Graphs have been updated and enhanced. Examples provided throughout chapters to illustrate various concepts have been updated to better reflect the fields of psychology, social work, and counseling. Educational interventions remain as a major focus in these examples.

This new edition includes some additional features and content. These are:

• Examples have been added to all chapters. These examples provide the reader with illustrative scenarios that emphasize major concepts pre- sented in chapters.

• In the design application chapters, vignettes serve the additional purpose of allowing the reader to check her/his understanding of the major con- cepts by first thinking about the vignette and then checking those answers against those provided by the authors.

• Check It Out features have been added in all but the design application chapters. The Check It Out features present information/questions to the reader who can then consider her/his answers. The authors provide sug- gested answers at the end of chapters for the reader to check against her/ his own thinking.

• There is an accompanying website with several useful features for stu- dents and instructors.

The accompanying website includes:

• A test bank with multiple choice and short answer questions for all but the design application chapters. Correct answers for multiple choice questions and sample answer material for short answer questions are also provided.

• Powerpoint presentations for all chapters.

• Potentially useful web resources.

We would also like to address two issues related to terminology. Many authors use the term single-case studies (as a case may represent a group of individuals treated as a single-case). Many authors now use the term partic- ipant rather than subject. We have retained the term “single-subject” for its historical significance and recognition. We acknowledge the use of other terms and do, in fact, frequently use the term participant in this text.

We trust the reader will find this new edition of the text retains its read- ability and accessibility to the novice or consumer of single subject research. It is also our hope that the changes will present new challenges to the readers’ thinking in a way that allows instructors and students to gauge their under- standing of the concepts presented.

Acknowledgements The authors wish to acknowledge the invaluable work of Emily Hendricks, graduate school psychology student at the University of Dayton, in the prep- aration of figures for this text. Her attention to detail and dedication to excellence are much appreciated.

We appreciate the assistance offered in the revision of this text by the following reviewers:

Tammie Bolling, Tennessee Technology Center at Jacksboro

Lisa Jackson, Schoolcraft College

Todd Haydon, University of Cincinnati

John Begeny, North Carolina State University

Mary Streit, Kaplan University

xvi FOREWORD

P A R T 1 Conducting Single Subject Research: Issues and Procedures

CHAPTER 1 Historical Perspectives and Important Concepts in Single Subject Research 3

CHAPTER 2 Methods for Changing Target Behaviors 25

CHAPTER 3 Methods for Recording Behaviors 53

CHAPTER 4 Issues in Single Subject Research 91

CHAPTER

1 Historical Perspectives and Important Concepts in Single Subject Research

IMPORTANT CONCEPTS TO KNOW HISTORICAL PERSPECTIVES IN SINGLE SUBJECT RESEARCH

Mark’s Vignette

BASIC CONCEPTS AND DEFINITIONS OF TERMS Independent, Dependent, and Extraneous Variables

Mark Revisited #1

Baseline, Intervention, and Follow-Up Phases Mark Revisited #2

NOTATIONS

THE X-Y OR LINE GRAPH The Dependent Variable The X-Axis The Independent Variable and Phase Change Lines Data Paths The Legend Summary

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

Historical Perspectives in Single Subject Research In the early 20th century, J. B. Watson argued that psychologists should focus efforts on observable behavior. Watson declared that the theoretical goal of behaviorists was the prediction and control of behavior (Watson, 1913). Watson stressed the need for obtaining data that were of scientific merit and not dependent on subjective interpretations. Watson argued that behaviorists should focus on the relationships between stimuli in the environment and the subsequent responses of individuals (the stimulus-response paradigm for human behavior). By the 1930s, B. F. Skinner had published The Behavior of Organisms, a landmark book, which extended the knowledge and prevail- ing views of behaviorism. Skinner distinguished between respondent (reflex- ive) behavior and operant (or voluntary) behavior. Skinner asserted that operant behavior was largely influenced by events that succeeded, rather than just those that preceded, the behavior. Skinner and colleagues conducted many experiments over subsequent decades outlining the principles that have become important in applied behavior analysis (e.g., reinforcement and pun- ishment paradigms). Initially, Skinner’s work focused on animal research, but he later devoted attention to the potential impact of applying behavioral prin- ciples to human endeavors including education (e.g., see Walden Two and Science and Human Behavior). Skinner acknowledged the importance of events that go unseen (cognitive and emotional processes) that influence human beings, but also stressed that the overt responses to those events may still be observable and addressed by behavior analysts.

By the 1950s and 1960s, behavior analysts were publishing studies that examined the effects of the application of behavioral principles to both normally developing and atypically developing individuals. By 1968, the Journal of Applied Behavior Analysis (JABA) had begun publication. Therein, many single subject studies were (and continue to be) reviewed and published. New designs and variations on existing designs were investi- gated. In the initial issue of JABA, Baer, Wolf, and Risley (1968) discussed major dimensions of applied behavior analysis.

Baer et al. (1968) noted that research that focuses on socially important behaviors may be applied, behavioral, and analytic. We summarize their dis- cussion on these concepts in the following passages. Applied refers to the inter- est displayed by society in the problems being studied. In applied research, there is typically a close relationship between the stimuli and behavior being studied, and the individual whose behavior is changing. Society and the indi- vidual have a vested interest in the behavior change, and therefore, the behav- ior change has social validity (Holman, 1977; Wolf, 1978). Behavioral refers to the pragmatic nature of the study. First, emphasis is placed on what a person can do rather than what she or he can say (except when the behavior under

4 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

study is some verbal response). Second, reliable quantifying of behavior must be achieved and that measurement often involves systematic observation by other people. Third, it is equally important to question not only what behavior was changed during a study, but also whose behavior has changed. Therefore, the explicit measurement of the observations to establish the reliability of the data collected (as we discuss in Chapter 3) is necessary. The analytic aspect of applied behavioral research refers to a believable demonstration that events controlled by the researcher account for the presence or absence of the behav- ior in question. Baer et al. (1968) noted that two single subject designs, the reversal and multiple baseline designs, were viable methods for generating a favorable judgment that the analytic goal has been met. Further, applied behavior analysts seek to encourage “valuable” behavior. That is, behaviors are targeted that will be reinforced in the individual’s natural environments, thereby maintaining changes when the experimental conditions are withdrawn. Good technical descriptions of procedures are also necessary for the analytic goal to be met. Good technical descriptions refers to the complete identification and description of any techniques applied to encourage behavior change. This bears directly on the need for replication, a process required not only to meet the analytic goal, but also to generalize results and demonstrate the robustness of any behavior change procedures. Considerable detail must be given in applied research reports to ensure the reader could reasonably replicate the events described. Baer et al. (1968) also stressed that applied behavior analysts must describe their procedures within a conceptual system that allows under- standing and expansion of technologies. For example, describing the exact method by which a child is taught to discriminate letters of the alphabet is good; describing the process in terms of fading antecedent stimuli and differen- tial reinforcement is better. The latter allows professionals to discuss procedures in a more universal sense and to recognize commonalities and differences among various experimental methods. Another area of concern for these authors included the effectiveness of the techniques used; Baer et al. noted that if behavior changes did not produce practical value, then the study has failed. Similarly, behavior changes ideally should be maintained over time, be general- ized to new settings and new situations, and result in the emergence of new but related behaviors. In summary, Baer et al. stated that

an applied behavior analysis will make obvious the importance of the behavior changed, its quantitative characteristics, the experimental manipulations which analyze with clarity what was responsible for the change, the technologically exact description of all procedures contribut- ing to that change, the effectiveness of those procedures in making suffi- cient change for value, and the generality of that change. (p. 97) —Source: Baer, D. M., Wolf, M. W., & Risley, T. R. (1968). Some current dimensions

of applied behavior analysis. Journal of Applied Behavior Analysis, 1, 91–97. © Society for the Experimental Analysis of Behavior, Inc. used by permission.

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 5

Since Baer et al.’s landmark piece, many developments have occurred in the field of applied behavior analysis and single subject research. New designs have been developed as well as variations of existing ones. The field continues to expand, at least in part because of the ever-growing num- ber of researchers and practitioners who have used this approach and main- tained the conceptual system that Baer et al. advocated. Single subject designs are used by educators, psychologists, social workers, speech and language pathologists, and other professionals. The remainder of this chap- ter is devoted to basic concepts in single subject research. These concepts are important to understanding the fundamental aspects of the designs and methods used by applied behavior analysts.

Certain conventions used in single subject research are similar to those used in other research designs. These include the concepts of independent, dependent, and extraneous (or confounding) variables. Other concepts are more unique to single subject designs. These include the concepts of base- line, intervention, and follow-up phases; the notations used; the basic x-y graph and the use of each axis in relation to the independent and dependent variables; the plotting of data collected; and the legend for a figure that includes an x-y graph. Some of these concepts receive additional attention (e.g., extraneous variables) in other chapters, particularly Chapter 4, “Issues in Single Subject Research.” However, this discussion should provide the reader with a developing understanding that will better enable her or him to comprehend immediately succeeding chapters.

Mark’s Vignette will provide the reader with a situation and information for illustration of various discussed concepts throughout this chapter. Other vign- ettes will appear in other chapters as well. Mark’s Vignette is intended to illus- trate some very good practices as well as some that may confound or complicate the situation. We will provide appropriate comments throughout the chapter to illustrate how Mark’s Vignette applies to the concepts being discussed.

Mark’s Vignette Mark is an 8-year-old student early in his third grade year of school. Mark’s parents have reported to his school district that Mark has atten- tion deficit/hyperactivity disorder. He was diagnosed when his parents were concerned about reports of Mark’s behaviors that were interfering with his learning and the learning of other students (getting out of his seat frequently and without permission, speaking out of turn when in large-group or small-group settings, making frequent, careless errors in his work, being easily distracted, difficulty sustaining attention suffi- ciently to complete independent work). Mark began receiving medication to control the symptoms of his condition, but his pediatrician also stressed that Mark would need to learn to independently control his

6 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

own impulsive behaviors to the extent possible. The school personnel were very much “on board” with the pediatrician’s recommendations. Prior to Mark’s diagnosis, it was observed by his teacher that daily, he did not com- plete the 20 computational math problems assigned for independent prac- tice within the allotted 10-minute time period. Additionally, she noted that of the problems he did complete, his average accuracy was only 50% of all completed problems correct. It was decided Mark’s learning objective would be, “When presented with independent math work, Mark will com- plete 20 problems within the allotted time and with 80% or better accuracy for five consecutive assignments.” Mark was to be rewarded for meeting his learning objective by placing a star on a chart on his desk that would be used to record his progress. With each star achieved, he would be verbally praised by the teacher and receive 5 minutes of free time at the end of the day to engage in an appropriate activity of his choice. When he obtained five consecutive stars, he would also be rewarded with an additional rein- forcer of selecting an appropriate class “fun” activity of up to 30 min- utes duration. The team, comprised of Mark’s parents, Mark, and school personnel, agreed that if Mark would be able to achieve this level of competence for five consecutive assignments, he would have learned to complete his work accurately and would no longer need the additional rewards of the star chart and free time. Instead, the teacher and aide would continue to occasionally praise him for maintaining his independent and accurate work. The team also considered that if the behavior intervention plan to improve Mark’s math work was success- ful, the same reward system might also be applied to other behaviors of Mark, such as completing his daily journal entries within the allotted time period and with 80% correct or better spelling. Later, after the team meeting, Mark’s parents told Mark that if he completed his math learning objective successfully, they would take him for a day trip to Cedar Point Amusement Park on Lake Erie, his favorite family outing. Eventually, Mark was successful in completing his math learning objec- tive and the same reinforcement system was implemented to improve his spelling on daily journal entries. Figure 1-1 includes a chart indicating Mark’s progress on his math learning objective and implementation of the reinforcement system with his journal entry writing.

Basic Concepts and Definitions of Terms As Baer et al. (1968) noted, there should be consistent and technologically clear descriptions of procedures. In the vignette of Mark, we have provided information as one might typically expect to read and receive in the every- day settings and language of educators and other school personnel. We have

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 7

Reinforcement System

Criterion of 80% of words correctly spelled

Baseline Journal Writing

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Daily Observations

% C

or re

ct ly

S p

el le

d W

or d

s W

h en

J ou

rn al

En tr

y Is

C om

p le

te d

i n

A llo

tt ed

T im

Follow-Up

Criterion of 80% (16 correctly answered problems of 20)

Reinforcement System

Baseline

Math

2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Daily Observations

# of

M at

h P

ro b

le m

s Co

rr ec

tl y

Co m

p le

te d

o f

20 w

it h

in 1

0 M

in ut

Criterion of 5 consecutive

assignments met

FIGURE 1-1 Mark’s Journal Writing Data

20 14

8 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

also emphasized that single subject research designs use conventions that are also commonly used in other applied research designs. The basic con- cepts and definitions of the terms associated with those conventions follow along with more technologically exact descriptions of Mark’s behavior intervention plan in relation to single subject research.

As the title of this text suggests, the term subjects is the historical term for individuals (e.g., Mark in the vignette) in a single subject study who are involved in changing their own behaviors. More recently, the term participants has come into common use. Participants is preferred by many researchers as it suggests a willing individual, who has given informed con- sent to the procedures in the study, and who is actively involved and vested in the study outcomes, rather than a person being “subjected” to a treat- ment/intervention. We will use the terms interchangeably, and the reader should be aware that older studies in particular are likely to use the term subjects, while more recent studies may use the term participant.

Independent, Dependent, and Extraneous Variables As in other types of research, the terms independent and dependent variables are used to describe the elements in a study that are related to the demonstration of a functional relationship. The independent variable is the intervention(s) used to encourage change in human behavior in single subject research. In essence, the independent variable is the treatment or intervention that the researcher controls in order to influence changes in the dependent variable. It is worth noting that only the individual partici- pant may change his or her behavior. The educator or therapist may create conditions (e.g., through the manipulation of antecedents and consequences) that encourage change, but clearly one cannot change another person’s behavior. For the sake of convenience, however, we will refer to changing the behavior of individual subjects with this understanding.

The dependent variable is used to measure changes (or the lack thereof) that demonstrate that the desired outcomes of the study are or are not being achieved. The term dependent is useful, because if a functional relationship exists in the study, the dependent variable should change in accordance with (or dependent on) the presence or absence of changes in the independent variable. For our purposes, we will equate the dependent variable with the term target behavior (see later discussion). Changes in the target behavior are used to determine whether the treatment or intervention is having the desired effect. For example, if positive reinforcement is being applied for progressively better handwriting, the systematic application of positive rein- forcement should have a direct influence on the dependent variable or target behavior (legible handwriting).

We acknowledge that some experts may assert that “technically” the target behavior and dependent variable are not always the same. For

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 9

example, the target behavior could be speaking louder. The dependent vari- able could be the decibel measurement of the actual loudness of the indivi- dual’s voice. Again, to avoid confusion (or perhaps the tedium associated with repeatedly making this possible distinction) we will interchangeably use target behavior and dependent variable.

Extraneous (or confounding) variables are any elements of a study that may confuse or obscure the believability that the independent variable and dependent variable share a functional relationship. Examples could include who is delivering the intervention, where the intervention is being delivered, the emotional or physical maturation of the individual subject, the influence of concerned parents or significant others, and so on. The possibilities are, unfortunately, endless. Extraneous variables are discussed at greater length in Chapter 4.

Mark Revisited #1 The independent variable in Mark’s Vignette would be the reinforcement system including the star chart, praise, and the free time awarded for every successful assignment as well as the ultimate 30-minute activity reinforcer for completion of his objective if he reaches the required level of problems completed with appropriate accuracy for five consecutive assignments.

The dependent variable or target behavior in Mark’s Vignette is com- pleting math work in the specified time period and with at least 80% accu- racy for the completed problems. Note this target behavior includes three dimensions: the number of problems to be completed, the specified time period in which to complete them, and the level of accuracy required (at least 16 of 20 problems or 80% completed correctly). To simplify data col- lection, the teacher decides to record and graph the number of problems completed correctly within the allotted time. Dependent variables are not always multidimensional.

There is at least one possible extraneous (or confounding) variable. Mark was successful, but the team can be less sure if it was actually the reinforcement system (star chart, praise, and free time) that encouraged the behavior change, or was it the trip to Cedar Point, or was it the com- bination of the two? Also, if successful, it might also be confounding as to whether the star chart, praise, 5-minute individual activity, or the 30-minute class activity might have had more or less impact on Mark’s performance. That might seem irrelevant if Mark has achieved success, but from a research perspective, it would be difficult for a researcher to make a firm assertion it was the school-based reinforcement system (and specific components of that system) that really accounted for the change in Mark’s behavior. If Mark is successful in also changing his spelling in his journal writing with only the school-based reinforcement system in

10 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

effect (no trip to Cedar Point this time!), then the functional relationship between the independent (reinforcement system) and dependent variables (the math target behavior and the journal-writing target behavior) might be considered stronger. However, it could still be confounding as to which aspect(s) of the reinforcement system may have had the greater impact on Mark changing his behavior.

For the reader’s practice in identifying independent, dependent, and confounding variables, consider the following example.

Baseline, Intervention, and Follow-Up Phases In most single subject research designs, the baseline phase is the first stage. Baseline data in an applied study are data collected when the independent variable is not being implemented. This does not mean, however, that dur- ing the baseline phase the individuals in the study setting do nothing. The status quo is maintained unless clearly harmful to the individual or others. For example, the teacher/researcher begins collecting data on accuracy in solving independent math problems. The environment currently includes the grading of papers as the consequence of the target behavior perfor- mance. Rather than alter the current conditions, the researcher would collect data concerning the target behavior under these existing conditions. The purpose of the baseline measure is at least twofold. First, the researcher should gain a standard of current performance on the dependent variable

4 C H E C K I T O U T # 1 A school psychologist works with an adolescent who has just begun to have issues with making threatening gestures and remarks to other students, although the adolescent has never acted on any of those threats nor do the other students appear to take her threats seriously. The psychologist works with the adolescent, her parents, and other school team members to design a program to eliminate this threatening behavior. The team arrives at a “contract” that states, “If Emily threatens another student (gesturally or verbally) at a zero frequency (no threats at all), she will be praised that day and provided with an opportunity to read a book of her choice for 15 minutes at the end of the day. After 5 consecutive school days of zero threats, she will be reinforced with lunch from McDonald’s on Fridays for her and a friend of her choice.”

What might be considered the independent variable in Emily’s learning pro- gram? What might be considered the dependent variable?

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 11

by which future changes may be compared. Second, the baseline provides additional opportunity to glean information that may reveal important aspects about the dependent variable performance and the environment. For example, the researcher may begin collecting data and soon discover that per- formance is very uneven. That is, the dependent variable measure seems to fluctuate from one measurement to the next. This should suggest to the researcher that some variables that may not have been identified as yet are influencing performance. The baseline phase should end when there is stabil- ity in performance on the dependent variable. Stability is typically assumed when there is limited variability in the baseline data (e.g., the level of the data does not dramatically differ from one observation to the next and there is no clear increasing or decreasing trend). It is not unusual for individuals to suggest that baseline should represent three measurements or some similar number of observations, although there is no specific number of observations that the teacher/researcher should use. In some instances, the baseline may be shortened or skipped altogether when ethical treatment demands so. For example, if an individual is harmful to himself or others, to prolong a baseline phase (or even to implement a baseline phase) to gain acceptable stability could be dangerous. In other situations, the dependent variable is unlikely to change until the independent variable is introduced. An individual who is acquiring new communication skills may be very unlikely to acquire those skills unless there is some direct intervention designed to improve those skills. In such situations, the baseline may be extremely abbreviated (i.e., a sufficient number of measures to demonstrate the skills are currently not in the indivi- dual’s repertoire). Once stability or an acceptable number of measurements is obtained based on ethical treatment considerations, the intervention phase(s) should be implemented (although, as the reader will discover in the chapters devoted to the specific designs, there are variations to this procedure).

The intervention phase(s) is implemented following the baseline phase or, when justified, immediately and without a baseline phase. We will refer to the intervention phase in the singular, but the reader should be aware that there may be multiple intervention phases. The researcher systemati- cally implements the independent variable during the intervention phase. Typically, some level of performance (criterion) on the dependent variable is identified a priori to determine when the desired outcome has been achieved. The researcher continues to measure the dependent variable to determine the effectiveness of the independent variable. Should progress not be achieved, the researcher may alter the independent variable or even iden- tify a new one altogether. When this contingency arises, there may be a need to introduce another baseline phase to once again establish stability in the dependent variable. Also, reintroducing the baseline may serve to avoid mul- tiple treatment interference of having one intervention immediately following another (see Chapter 4). Once the desired outcomes have been achieved

12 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

during the intervention phase, a follow-up phase is advisable, although not always implemented in research studies. In some designs, there may even be a return to a baseline phase following intervention phase (e.g., in withdrawal designs). Again, the reader should understand the basic concept of baseline phase followed by intervention phase, which is typical and is most useful at this point. This is also the typical design in the everyday practice of teachers and other personnel working with individuals in encouraging behavior change.

The follow-up phase is intended to measure the effect of the independent variable on the dependent variable over time. This phase is implemented after the successful intervention phase. The status of the independent vari- able may be that it is no longer in effect, or it may be at a reduced level of intensity, or it may be maintained at a previously successful level. Generally, the overall goal during follow-up is to remove the independent variable, par- ticularly if it requires a rather intrusive effort, to demonstrate that its effects will continue long after its withdrawal. In other words, the researcher wishes to demonstrate that the changes in the dependent variable are rela- tively permanent and will be maintained even in the absence of the indepen- dent variable. In teaching or clinical practice, dependent variable changes would be comparable to maintenance, independent practice, or possibly generalization of the target behavior. Independent variables that are clearly an enhancement to the environment (e.g., verbal praise) may be maintained by those who work with the individual even if the independent variable is officially withdrawn. Perhaps the most important point to remember in this discussion is that a follow-up phase strengthens a study by demonstrat- ing the effectiveness of the intervention over time and thereby strengthens the social and ecological validities of the study.

Mark Revisited #2 Review Figure 1-1 that includes the data and visual depiction of Mark’s progress. Note that there is a baseline phase for Mark’s math target behavior and a baseline phase for Mark’s journal-writing target behavior. There is the implementation of an intervention phase (implementation of the independent variable—the reinforcement system) with Mark’s math target behavior. Meanwhile, the baseline phase is maintained for Mark’s journal-writing target behavior. When Mark achieves his math objective, the intervention phase is begun with his journal-writing target behavior. The follow-up phase is implemented for his math target behavior. That is, the reinforcement system for his math work is withdrawn, and we can see that Mark is maintaining his level of achievement. Mark’s journal-writing target behavior shows improvement once the reinforcement system is implemented (and not before it is implemented for journal writing). When changes in the dependent variable (target behavior) occur in direct relation

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 13

to the implementation of the independent variable (intervention), a func- tional relationship between the two variables is more likely demonstrated.

Finally, it would be understandable if a member of Mark’s educa- tional team would question if it was ethical to withhold the intervention with the journal writing for an extended period if the team believes it would encourage a desirable change in Mark’s behavior. Such ethical considerations may arise in conducting single subject research.

Notations In single subject designs, letters are used to denote certain phases or aspects of a design. In most (but not all) instances, the first phase of a single subject design is the baseline phase. It is denoted by the letter A. Each subsequent letter (B–Z) is used to denote a particular independent variable (or interven- tion phase). Occasionally, a combination of letters (e.g., BC) is used to denote that a combination of treatments or independent variables is being simultaneously applied to the dependent variable or target behavior. If a phase is repeated during a study (e.g., see Chapters 5 and 6 for withdrawal designs), the same letter is used to denote the phase. For example, a researcher implements a baseline phase (A), followed by an intervention phase (B), followed by a return to a baseline phase (A), followed by the implementation of a new independent variable (C). This study would be denoted as an A-B-A-C design. See Figure 1-2 for a sample depiction of this notation. The researcher is telling others that both A phases involved the same conditions and that the B and C phases differed significantly from the A phases and from each other. A hyphen typically is inserted between the letters to denote the separate phases. Some researchers prefer to assign a number as well when phases are repeated (e.g., A1-B1-A2-B2). Occasionally, a package intervention is used that combines more than one independent variable (e.g., positive reinforcement along with response cost). When this occurs, the phase is denoted by multiple letters (e.g., BC, with B representing positive reinforcement and C representing response cost). Thus, a notation might be A-BC-A-B. The notation in this example would reveal that during a final intervention phase, only positive reinforce- ment was used (B) as an intervention and not response cost (C). The combi- nations of letters denoting phases are virtually infinite. However, the researcher should be aware that when multiple interventions are implemen- ted in single or multiple phases (e.g., BC-D-E-BE phases), the effects of one independent variable may begin to have an effect (multiple treatment inter- ference) on either simultaneous or subsequent ones. That is, the researcher’s ability to determine which independent variable is effective may be seriously diminished. In our example of the combined use of positive reinforcement

14 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

21 22

23 24

A o

r A 1

Ba se

lin e

Ph as

A o

r A 2

Ba se

lin e

Ph as

B or

B 1

Ba se

lin e

Ph as

C o

r C 1

D i!

er en

t In

te rv

en tio

n th

an B

15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

1 2

3 4

5 6

7 8

10 O

b se

rv at

io n

Dependent Variable

9 11

12 13

14 15

16 17

18 19

FI GU

RE 1- 2

Ty pi ca lN ot at io n Us ed

in Si ng le Su bj ec t

De sig n X– Y.

©CengageLearning2014

and response cost, it may be impossible to tell if either is more effective or if they must be used in combination to be effective. Although the desired out- come (the change in the target behavior) may be achieved in such a case, which is fortunate, the researcher may have trouble explaining why it was achieved. When a follow-up phase is implemented, it may be noted with the words “Follow-Up” or some notation that is used to make clear this is a different phase. We would not recommend it be denoted with a letter that suggests it is an intervention phase.

The X-Y or Line Graph The x-y graph, or line graph, is the typical but by no means only manner in which the results of a study are depicted. We will discuss the line graph because of its widespread use. The reader may wish to examine other possi- bilities (e.g., bar graphs, histograms, contingency tables) if the data and var- ious components of a study are more easily understood by using some other method. The line graph has a number of components that are commonly found in research reports. These include the dependent variable, the depic- tion of the passage of time or the various measures of data, the data path and breaks in that path, the independent variable and its phase change lines, and the legend.

The Dependent Variable In Chapter 3, you will read a more detailed discussion concerning recording of target behaviors and the manners in which different types of data may be reported. For example, one may record how frequently a target behavior occurs, how long it occurs, or how long it might take an individual to begin emitting a target behavior. The dependent variable is graphed along the y- (or vertical) axis (see Figure 1-3). Tick marks, or lines that indicate

4 C H E C K I T O U T # 2 Consider the earlier example of Emily, who was being encouraged to reduce verbal and gestural threats to zero rates. In such a study, there could be a baseline phase, the implementation of an intervention (positive reinforcement with daily praise and free time for reading), followed by a return to baseline, followed by the implemen- tation of a new intervention phase (daily praise and McDonald’s lunch if successful each day that week up to lunch time on Friday). What would be the notations for such a single subject design?

16 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Ba se

lin e

Ti ck

M ar

D at

a Pa

th Br

ea k

D at

a Pa

th Pa

rt ia

l Ph

as e

C ha

ng e

B Ba

se lin

e C

Fi xe

d Ra

tio -5

Fi xe

d Ra

tio -1

0 Fo

llo w

-U p

O b

se rv

at io

n s

Dependent Variable

Ph as

e C

ha ng

e Li

ne s

FI GU

RE 1- 3

Ty pi ca lX – Y Lin e

Gr ap h Ex am pl e

©CengageLearning1999

the units of measure (e.g., number of occurrences, number of seconds or minutes, or degrees of intensity), are evenly spaced along that axis. The tick marks should be displayed so that the visual picture created when changes occur in the dependent variable does not mislead or confuse the viewer. The spaces between and the number of marks should serve to accu- rately display changes in the dependent variable. For example, a researcher is measuring the rate of an individual’s behavior and that rate is expressed in terms of how many responses are made per minute (e.g., 1.00 response or 0.50 response per minute). Changes in the dependent variable may be quite small from measurement to measurement, perhaps changing by tenths of a response per minute. If the researcher chose to allow each tick mark to rep- resent 0.10 response per minute and to space those marks more widely, then even a small change might result in a visual picture that suggested a rather marked change. In some cases, where relatively small changes in the depen- dent variable represent significant change (e.g., changes in self-abusive behavior) based on social and ecological standards of validity, such a prac- tice may be warranted. However, in most instances, the y-axis should be scaled in such a way that small changes appear small when plotted and big changes appear significant. The researcher is in the best position to deter- mine how data may be depicted through the scaling of the y-axis and should be careful to not create literally a false picture for the viewer.

The X-Axis The x-axis is used to display the various measurements made of the depen- dent variable. Usually, the x-axis depicts the passage of time. As with the y-axis, the researcher includes tick marks, or lines appropriately scaled. Each tick mark represents an observation and measurement of the depen- dent variable (see Figure 1-3). A tick mark might represent an observation across 5 minutes, across 1 hour, a morning, or a whole day, or longer (though usually not more than 1 day). In some instances, the tick mark might represent an average or summary of observations (e.g., average dura- tion of time spent “out of seat” per day; total duration spent “out of seat” per day). The length of the observation may in fact vary from tick mark to tick mark (see Chapter 3 for an explanation of varying observation periods, but this usually refers to an average rate of response for the target behavior or similar measure), but the results of the observation are always expressed in the same terms (e.g., frequency or duration or rate) as determined by the dependent variable graphed along the y-axis. The tick marks along the x-axis serve to create a visual picture of how the dependent variable has changed as the number of observations increases and, more importantly, in conjunction with changes in the baseline, intervention, and follow-up phases of a study.

18 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

The Independent Variable and Phase Change Lines As these various phases are implemented, solid vertical lines drawn parallel to the y-axis and perpendicular to the x-axis are used to depict each phase change (see Figure 1-3). The phase itself typically is written at the top of the graph and may be expressed by the letter notations discussed earlier or by wording that describes the phase (e.g., Baseline, Positive Reinforcement, Baseline, Response Cost, or A-B-A-C). Occasionally, a dotted or broken line may be drawn similar to the phase change line to depict a change within a phase but not a complete change in the independent variable (see Figure 1-3). For example, positive reinforcement is being delivered to a stu- dent on a ratio of one reinforcer to 5 correct responses (called a fixed-ratio schedule, discussed in Chapter 2). The researcher moves from this fixed ratio to a ratio of one reinforcer for every 10 correct responses. Such a change is not a complete change from the use of the independent variable of positive reinforcement, but may represent a change worth noting to the viewer. A broken line with appropriate headings (e.g., Fixed Ratio-5, Fixed Ratio-10) would help to explain the change that occurred.

Data Paths Data paths are lines that connect each data point plotted along the x-y axes. Individual data points are plotted. For example, following a researcher’s first observation of the dependent variable, the researcher would track verti- cally up from the x-axis from the tick mark denoting the first observation. The researcher would continue to track horizontally from the appropriate tick mark on the y-axis that corresponds with the data obtained on the first observation. Where the two tracks intersect, a point would be plotted. This process is repeated for the data gathered from each observation session (remembering a session could vary considerably in length of time and could include a number of individual observations that are summarized and com- piled across the overall session). As they are plotted, the points are usually connected by a solid line. This solid line is not typically drawn across phase change lines. Also, the zero level of the dependent variable on the y-axis may be marked just above the intersection with the x-axis to avoid clutter- ing the x-axis should zero-level responding occur. This may not be possible using a graphing software program, and plotting on the x-axis itself is acceptable. Should the collection of data be interrupted due to unforeseen circumstances (e.g., the individual becomes sick and is unable to be observed), the data path is also interrupted and there should be some indi- cation where the actual break in data collection occurred (see Figure 1-3). The x-axis should also be amended as needed to reflect this break in data collection, and a special notation to the viewer may be included along the

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 19

x-axis or at the break in the data path. Generally, if the individual partici- pant is a student and no data are collected on weekends, it is unnecessary to treat such natural weekend breaks as breaks in the data path. To include such breaks would not be wrong; we merely wish to point out that this type of break has been omitted when reporting data from school environments.

There are circumstances in which more than one data path is plotted on the same line graph (e.g., performance by two or more individuals in the same study, the performance of the same individual under different condi- tions). In this case, a different path is assigned for each dependent variable path (see Mark’s Math Data graph for an example of plotting across a phase change line) (e.g., a solid line for one, a dotted line for another), and often the actual data points making up each path may be assigned different shapes (e.g., solid circles, squares, triangles). Color coding may also assist in this process. The key to plotting multiple data paths is to use a graph suffi- ciently large and clear that the viewer can easily discern each path and the differences in performance for each dependent variable (see Chapters 11 and 12 on alternating treatments designs). Also, within this text, we have pre- sented variations of x-y graphs.

The Legend The more complicated a line graph becomes (e.g., multiple data paths, use of several independent variables depicted), the more important the legend becomes. A graph can become very cluttered and confusing, if not unread- able, if a researcher includes too much information. The legend allows the

4 C H E C K I T O U T # 3 Once again, consider the example of Emily who is being encouraged to eliminate verbal or gestural threats. The school psychologist/researcher knows she must record the number of threats (verbal or gestural) each day. During a very brief baseline, Emily has used threats at a frequency of 3, 5, and 4 occurrences on 3 con- secutive days (two team members questioned if any baseline should occur due to the nature of the target behavior but agreed a very brief baseline would be accept- able as Emily had never actually acted on any threat and other students did not appear to take her threats seriously). The independent variable (a reinforcement system for threatening at a zero rate per day) has been implemented following this third day. Can you create an x-y graph that includes the y-axis with tick marks indicating the dependent variable, an x-axis indicating the 3 days of baseline data, a phase change line indicating the implementation of the independent variable, fol- lowed by a second baseline phase and a new independent variable including only McDonald’s lunch on Fridays as a reinforcer?

20 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

researcher to abbreviate and to use alphabetic notations instead of whole words, which can be explained in the legend. Figure 1-3 includes an ade- quate legend.

Summary In this chapter, we have provided a brief historical perspective on single sub- ject research and applied behavior analysis and a discussion of basic con- cepts. This chapter, particularly the basic concepts portion, sets the stage for the following discussions. It is not until Chapter 5 that we discuss any specific designs; Chapters 2 and 3 provide the reader with a more detailed background on applied behavior analysis, while Chapter 4 provides a thor- ough examination of principles and procedures found in virtually all single subject research designs.

Key Concepts/Terms Applied—The interest of society in the problem being studied. Behavioral—The pragmatic nature of the study; emphasis is on what the

person can do rather than what she or he can say. Analytic—A believable demonstration that events controlled by the

researcher account for the presence or absence of the behavior in question.

Replication—A process required not only to meet the analytic goal, but also to generalize results and demonstrate the robustness of any behavior change procedures

Subjects/participants—Subjects is the term used for the individuals who are changing their behaviors in a single subject study. Participants is a term more recently used. These terms are used interchangeably in this text.

Independent variable—The intervention(s) or treatment(s) used to encour- age (or maintain) change in behavior; the independent and dependent variables should share a functional relationship.

Dependent variable—The target behavior that is measured to determine the effects of the independent variable; changes in the target behavior should be dependent on changes in the independent variable (a func- tional relationship).

Extraneous variable (or confounding variables)—Almost any element in the study that confuses or obscures the functional relationship.

Baseline phase—Generally, the first phase in a study, during which initial performance on the target behavior is measured before implementation of the intervention (independent variable).

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 21

Intervention phase—A phase when the intervention has been introduced and data are collected to determine effects on the target behavior (dependent variable).

Follow-up phase—Typically, a final phase of a study when the researcher continues to measure performance on the target behavior, although the inde- pendent variable may have been withdrawn following successful intervention.

Notations—The system of letters used to identify the type of design used; A refers to a baseline phase; B and all subsequent letters to intervention phases. Each letter (e.g., B, C) refers to a different intervention; combi- nations (e.g., BC) refer to package, or combinations of, interventions; numbers may be attached (e.g., A1, B1) to denote a first baseline phase or a first intervention phase when there are subsequent baseline or inter- vention phases (e.g., A2, B2).

X-Y or line graph—The typical line graph used to depict the quantitative data collected in single subject research; data are plotted at the appropri- ate intersects along the x- and y-axes.

Dependent variable on x-y graph—The target behavior performance is plotted along the y-axis; the y-axis must be calibrated (with tick marks used to denote units of measurement) so that changes in the dependent variable are appropriately depicted.

X-axis—The x-axis is used to depict observations across time (using tick marks to denote which observation is being plotted along the y-axis).

Independent variable and phase change lines—Implementation of and changes in the independent variable are depicted through lines drawn parallel to the y-axis; broken vertical lines indicate a change in the inde- pendent variable but not a complete phase change (e.g., from a fixed to a variable ratio reinforcement schedule).

Data path—Each data point that is plotted on the graph is connected by a line; the data path line is not typically drawn across phase change lines; breaks in the data path indicate an interruption in the observation or measurement of the dependent variable; multiple data paths are some- times required (e.g., for alternating treatments designs).

Legend—A guide to the x-y graph that allows the researcher to abbreviate on the graph and clarify (e.g., FR-5 on the graph could be further explained as a fixed ratio-5 response schedule of reinforcement in the legend).

4 Possible Answers to Check It Out

¶ The independent variable would be the reinforcement system of praise,free time for reading, and the lunch from McDonald’s when she suc- cessfully meets the 5-day requirement. The dependent variable would be the frequency of threats (verbal or gestural).

22 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

· The notation could be A-B-A-C or A1-B-A2-C. One might consider thenotation as A-B-A-B or A1-B1-A2-B2 if one were to argue the variation in the reinforcement system (to praise only and 5 days without threats to receive McDonald’s lunch) does not represent a substantive change in the independent variable.

¸ The graph should appear reasonably similar to the one shown below.

References Baer, D. M., Wolf, M. W., & Risley, T. R. (1968). Some current dimensions of

applied behavior analysis. Journal of Applied Behavior Analysis, 1, 91–97. Holman, J. (1977). The moral risk and high cost of ecological concern in applied

behavior analysis. Journal of Teacher Education, 37, 27–34. Watson, J. B. (1913). Psychology as the behaviorist views it. Psychological Review,

20, 158–177. Wolf, M. (1978). Social validity: The case for subjective measurement or how

applied behavior analysis is finding its heart. Journal of Applied Behavior Analysis, 11, 203–214.

Baseline1 (A1)

Baseline2 A2

Reinforcement System

(B)

Friday Lunch Reinforcement

9 8 7 6 5 4 3 2 1 0

1 2 3 4 5 6 7 8 10 Observations

Fr eq

ue n

cy o

f Th

re at

9 11 12 13 14 15 16 17 18 19 20

Sample of How Emily’s Graph Might

20 14

CHAPTER 1 HISTORICAL PERSPECTIVES AND IMPORTANT CONCEPTS IN SINGLE SUBJECT RESEARCH 23

CHAPTER

2 Methods for Changing Target Behaviors

IMPORTANT CONCEPTS TO KNOW METHODS TO INCREASE BEHAVIOR Everett’s Vignette

Positive Reinforcement Negative Reinforcement Premack Principle

Everett Revisited #1 Shaping

METHODS TO MAINTAIN BEHAVIOR Reinforcer Menus Satiation Primary, Secondary, and Generalized Reinforcers Quality of Reinforcers

Everett Revisited #2 Reinforcement Schedules Generalization of Target Behaviors

METHODS TO DECREASE BEHAVIOR Positive Punishment Negative Punishment Extinction Differential Reinforcement

Everett Revisited #3 Other Methods to Decrease Behavior Summary

KEY CONCEPTS/TERMS POSSIBLE ANSWERS TO CHECK IT OUT

Methods to Increase Behavior As noted in Chapter 1, the independent variable may be thought of roughly as the intervention used in the study. More specifically, it is that variable (or those variables) manipulated by the practitioner or researcher that are intended to encourage a change in or maintenance of the level of the depen- dent variable or target behavior. Within the context of applied behavior analysis, a variety of interventions are available for use as independent vari- ables. These include methods for increasing or maintaining behavior and methods for decreasing or eliminating behavior. The methods discussed are not mutually exclusive. That is, the interventions may be used alone or in combination and may be used in combination with other independent vari- ables not discussed in this chapter. For example, social workers and psy- chologists may use various therapeutic strategies such as counseling in their practices. Similarly, speech and language pathologists and other related ser- vices professionals may use strategies specific to their discipline and practice.

When a practitioner wishes to encourage a decrease in a maladaptive tar- get behavior, it is important ethically to encourage a corresponding increase in an adaptive response. Currently, many advocates for individuals with disabil- ities (particularly those with severe disabilities) stress that methods for decreas- ing behavior that focus on punishment are to be discouraged if not ruled out altogether as treatment options. Certainly, the use of more aversive techniques (e.g., electric shock, foul-smelling or foul-tasting substances) to decrease a mal- adaptive target behavior has come under scrutiny for a number of reasons. These include the efforts of advocacy groups, oversight by professional and institutional review boards, and a general reluctance by professionals them- selves to employ pain-inducing or other potentially unpleasant or unusual pro- cedures (see Chapter 4 for a discussion of ethics in single subject research). Still, it is necessary in a text such as this to discuss available intervention options; therefore, punishment techniques will be presented. The use of any intervention should be justified through ethical considerations and on an indi- vidual basis with informed consent obtained. Further, the right of the individ- ual subject to terminate an intervention is also advisable in instances when there are unpleasant events involved in the delivery of the intervention.

As noted previously, we will use the term intervention as synonymous with independent variable, and target behavior or response as synonymous with dependent variable. It should be stressed that these are convenient terms related to applied behavior analysis and that the concepts of indepen- dent and dependent variables may encompass a very broad spectrum of actual applications (e.g., medication/other medical interventions; responses on a behavior-rating scale or to interview questions, respectively). Also, it is a convenience to refer to changing another person’s behavior through

26 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

intervention. Remember, however, that only the individual participant may actually change his or her behavior. One of the goals of single subject research is to demonstrate a functional relationship between the intervention and behavior change (independent and dependent variables). The researcher or practitioner may create conditions (e.g., through the manipulation of antecedents and consequences) that encourage change, but clearly one can- not change another person’s behavior.

It is necessary to discuss first the basic principles of operant condition- ing. The essential paradigm for operant behavior includes an antecedent (or antecedents) that occurs before the target behavior and may potentially influence the occurrence of the target behavior, followed by the behavior itself, which in turn is followed by a consequence (or consequences) that may potentially affect the future occurrence of the behavior under the same or similar antecedent conditions (see Figure 2-1). Professionals who are encouraging change in another individual’s behavior are, in fact, using ante- cedent conditions and delivering consequences in the belief the future occur- rence of the target behavior will be altered when the individual finds her- or himself in the presence of the same antecedent conditions.

It is the consistent pairing of antecedent with behavior, behavior with consequence, and consequence with antecedent that encourages a change in or maintenance of the target behavior. This allows the practitioner or researcher to predict the impact of the independent variable on the target behavior. Generally, experts suggest that consequences (whether they are reinforcing or punishing consequences) be delivered immediately after the target behavior to increase the likelihood that the pairing of the two is more obvious to the individual.

Antecedents Behavior Consequence

What is occuring prior to emission

of behavior?

What is occuring that is both observable

and measurable?

What occurs immediately following the emission of

behavior?

Do the same antecendents appear

to elicit the same behavior?

Is the same behavior occuring?

Does the delivery of the same consequence appear to in!uence the

next behavioral response?

Data are collected to answer questions.

FIGURE 2-1 Paradigm for Operant

Conditioning

19 99

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 27

When referring to interventions that may be used to increase or main- tain behavior, we are referring to positive and negative reinforcement. Rein- forcement occurs when the probability that a behavior will occur in the future under the same or similar antecedent conditions is increased (or maintained) by the delivery of a consequence following the behavior. There are several critical issues to understand about reinforcement. First, reinforce- ment should not be perceived as meaning “good.” It is a phenomenon, an occurrence, not a quality. It is perfectly possible for desirable responses (e.g., wiping one’s mouth with a napkin or correctly sounding “th”) as well as undesirable responses (e.g., throwing food or engaging in disruptive behav- ior) to be reinforced. Second, the notion that we are affecting the future probability of behavior is important. The behavior that is reinforced (i.e., to which the consequence is applied) has already occurred; that particular response cannot be altered because it has already happened. Changes in the occurrence of behavior in the future (e.g., more frequent, with greater intensity, of longer duration) determines whether or not reinforcement has actually occurred. Third, collection of data is necessary to ensure reinforce- ment has occurred. Although we would like the reinforcement to be system- atic, it need not be so. Of course, the aim in this text is to provide examples of systematic reinforcement. Still, it should be noted that reinforcement may occur accidentally (e.g., a child’s tantrums may be reinforced through atten- tion). Fourth, the individual being reinforced determines whether a conse- quence is actually reinforcing or not. That is, the practitioner or researcher may believe a consequence has a reinforcing quality, but it is the individual’s future behavioral responses that reveal whether reinforcement has occurred. Every individual has his or her own unique history of reinforcement that will influence the degree to which a consequence serves as a reinforcer. What the researcher may believe is a reinforcer may hold no such quality to the individual participant. Finally, the antecedents and reinforcing conse- quences of behavior are not always observable. Because human beings are complicated organisms living in an equally complicated world, the variables affecting behavior can be numerous, and intrinsic to the individual as well as extrinsic. Therefore, the antecedents and consequences that influence the occurrence of any particular behavior are not always easily identifiable. For example, a researcher may be working with an individual to eliminate aggressive responses. The researcher has been quite successful in identifying antecedents and consequences that appear to influence the occurrence of aggression in the study setting. Yet, the target behavior remains largely unchanged. The individual has developed cognitive processes through which he justifies his aggression toward others because he believes “they are out to get him.” The individual has an older brother who served as a high-status model for the individual and is currently on the run because of aggressive criminal acts. This modeling affects the individual’s perceptions

28 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

about the appropriateness of aggression. Each of these influences has an effect on the occurrence of aggressive acts and may not have been identified by the researcher. In single subject research, we are, however, concerned with verifiable relationships, so we tend to focus on those aspects that are both observable and measurable.

Everett’s Vignette Everett is a young man with a history of mental illness requiring medication, who is currently hospitalized under a court order. His social worker, psychologist, his attending physician, and Everett himself have agreed that if he continues to refuse or “skip” taking his medication, he will likely experience serious health and behavioral consequences including possible incarceration. This team decides that if Everett will take his medication three times daily as prescribed, he should be rewarded weekly with an out- ing to a local business, event, or activity of his choosing. Everett is to self- record when he takes his medication. The nurse assigned to Everett will also monitor his medication intake. After the start of the program, Everett fails to obtain his weekly activity for the first 3 weeks, asking each weekend when he is going out. His social worker goes over his plan and explains to Everett each instance in which he did not take his medication. Over time, the opportunity for his weekly outings leads to Everett’s consistent use of medi- cation. Everett himself reports in his team meeting after three months, “I finally got it that I have to do these things if I want to get better. And, I am feeling much better now that I am taking my medication as prescribed.”

Positive Reinforcement Perhaps one of the more misunderstood notions related to applied behavior analysis is that positive reinforcement means that something good is happen- ing. In fact, positive here refers to the type of consequence delivered, not to the quality of the reinforcing event. Table 2-1 illustrates the relationship between type of consequence delivered and the different types of reinforcement and punishment. As depicted, when positive reinforcement occurs, a consequence is delivered that involves adding to the environment. Examples might be the delivery of praise, the awarding of a token, or allowing an individual to engage in a preferred activity, as in the case of Everett. In each example, the goal is to increase the probability that the target behavior will occur again in the same or similar conditions. Giving attention for inappropriate remarks made in class or therapy might also increase the probability of remarks being made in the future when in the same class or therapy session. This is not nec- essarily a “positive” event in terms of desirable outcomes, but it does represent positive reinforcement. The key to determining if reinforcement has occurred is

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 29

in evaluating whether behavior has strengthened or increased. With Everett, it was necessary to record and validate the occurrences of his target behavior to verify when he should or should not receive his weekly chosen activity. This key concept holds true also for negative reinforcement.

As noted previously, when operant conditioning occurs, there are ante- cedents to behavior, a behavior, and consequences. In operant conditioning, the consequences of behavior are paramount in encouraging the individual to make desirable changes in his or her life. Because we are concerned with systematic and predictable behavior change that is socially valid, we “tar- get” behaviors that are in the individual’s best interests and endeavor to use consequences that are acceptable, especially to the individual who is changing his/her behavior and/or are acceptable to that individual’s advo- cates (e.g., family members).

Negative Reinforcement When negative reinforcement occurs, behavior also is increased. That defines the occurrence of reinforcement in the most fundamental sense. Neg- ative reinforcement appears to be a frequently misunderstood concept in our experiences with university students and practitioners alike. We have wit- nessed students using the term synonymously with punishment. We have also seen it used to refer to a situation when an undesirable behavior is

TABLE 2-1 Characteristics of

Reinforcement and Punishment

ANTECEDENT BEHAVIOR CONSEQUENCE

Positive reinforcement

No specific requirement

Future probability strengthens/increases under same/similar antecedent conditions

Something is introduced/added to individual’s world

Negative reinforcement

Must be aversive to the individual

Future probability strengthens/increases under same/similar antecedent conditions

Aversive antecedent is removed from individual’s world

Positive punishment

No specific requirement

Future probability weakens/decreases under same/similar antecedent conditions

Something is introduced/added to individual’s world

Negative punishment

Preferred stimulus available

Future probability weakens/decreases under same/similar antecedent conditions

Stimulus is removed from individual’s world

30 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

reinforced (i.e., a “negative” outcome from the practitioner’s perspective). Both of these are wrong. The term negative does not connote either a desir- able or undesirable outcome, but rather the particular use of consequences to increase or strengthen behavior. Recall the positive in positive reinforcement refers to adding something to the environment as a consequence of a behav- ior. The negative in negative reinforcement refers to the removal of a stimulus as a consequence of emission of a behavior (see Table 2-1). The result is an increase in the probability of the response occurring once again. Negative reinforcement also involves the presentation of a special antecedent stimulus as well. That antecedent is aversive to the individual; its removal as a conse- quence of the emission of the target behavior is what reinforces that response. The red lights and buzzers that are activated when a person starts a car are intended to be aversive antecedents that in turn are shut off or removed or avoided altogether when the seatbelt is buckled. The likelihood that you will buckle your seatbelt when being exposed to these same stimuli should increase if negative reinforcement has occurred. In other words, you will buckle up before the car is started or shortly afterwards to avoid or escape the buzzers and lights. Negative reinforcement relies on the desire of people to avoid or escape undesirable stimuli. The seatbelt situation serves as an example of a desirable response being negatively reinforced.

As with positive reinforcement, undesirable responses may also be nega- tively reinforced. When a practitioner notices that a student is escaping or avoiding some situation (or person or environment), this should be a clue that negative reinforcement may be occurring. For example, a student who repeatedly disrupts a class only to be removed from the class may be receiving negative reinforcement (even though the practitioner may believe punishment is occurring!). If the class includes some stimulus that is aversive to the student, and disruptive behavior results in removal from the presence of that aversive stimulus, it may lead to an increase in disruptive behavior. For example, a student who doesn’t read very well is disruptive to avoid being called on to read aloud and continues to disrupt each time reading aloud is expected and he succeeds in being removed from the reading situation. This represents an example of an undesirable behavior being negatively reinforced.

There are several other key concepts and terms associated with rein- forcement, particularly positive reinforcement. These include the Premack Principle; reinforcer menus; satiation; primary, secondary, and generalized reinforcers; quality of reinforcers; and reinforcement schedules.

4 C H E C K I T O U T # 1 Why is negative reinforcement not punishment? You might peek ahead to the section on punishment to assist you in answering this question.

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 31

Premack Principle The Premack Principle (Premack, 1959) is applied liberally in educa- tional, therapeutic, and home environments. It refers to engaging in a highly preferred activity as a consequence for performing a less preferred activity. Determination of what is a highly preferred activity should be based on the individual’s actual behavior, not assumptions made by the researcher or practitioner. Some individuals might choose to play video games in their free time, whereas others might watch television or read. When the identified high preference activity is presented contingently, the likelihood that the less preferred activity will be performed in the future is increased. Sometimes referred to as “grandma’s rule,” this operant conditioning strategy is as old as parenting and teaching. Common situa- tions that exemplify the Premack Principle are “finish your homework and then you may watch television,” “complete your voice exercises and then you may listen to a CD,” and “finish your math work and then you may play with your friend.” The Premack Principle is frequently used because the reinforcing consequence (activity) often is available in home, school, and clinical settings. It is important to remember that access to that reinforcing activity must be controllable as well. If the individual can gain access to that highly preferred activity noncontin- gently, then attempted use of the Premack Principle with that activity may be futile.

Everett Revisited #1 In Everett’s Vignette, it is clear Everett has already acquired the target behavior of “taking his medication three times daily.” The issue that influences the selection of an independent variable is Everett’s lack of motivation, or at least his inconsistency in taking his medication. In such a situation, teams and researchers often select positively reinforcing consequences to provide incentives for individuals to engage in target behaviors that are in the individual’s best interests. However, selecting a reinforcer can be challenging. Allowing Everett to self-select his reinfor- cing event was a good idea to make it more likely he will take his medi- cation in order to obtain the reinforcer he has selected. The team has applied the Premack Principle in Everett’s Vignette. He is being given access to a high preference activity (his chosen outing) as a result of per- forming his lower preference activity (taking his medication). Research- ers, behavior analysts, teachers, other professionals, and parents use the Premack Principle frequently, and as with Everett, hope that such a con- tingency will lead the individual to realize that the low preference activity is worth engaging in even in the absence of the high preference activity.

32 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Shaping Shaping refers to providing positive reinforcement for responses of the tar- get behavior that are closer and closer to the performance criterion. By cri- terion, we mean the expected level of performance of the target behavior that is deemed necessary for the behavior to be functional and performed successfully. For example, signing one’s name should be accomplished so that it is legible (one criterion) and within a reasonable period of time (another criterion). In shaping, the individual is reinforced each time she or he emits a response that is closer to or meets the performance criterion. For example, an individual could be systematically reinforced for continued improvement in a career skill like keyboarding, where words typed per minute increased and the number of errors were decreasing. One potential difficulty in shaping is how the closer approximation to the performance criterion is being determined. For example, in learning musical instruments, the judgment of the teacher is usually the determinant of whether the indi- vidual has improved her or his musical performance; the less subjective the determination, the better.

Methods to Maintain Behavior Once a target behavior has been learned, it is important from a therapeutic perspective that it be maintained. From a research perspective, demonstrat- ing a functional relationship between the independent and dependent vari- able is critical. However, it would be a poor researcher indeed who would not make efforts to ensure the beneficial change in behavior is likely to be maintained over time. A researcher should be concerned that when the inde- pendent variable is withdrawn because the functional relationship has been demonstrated, the withdrawal of the intervention does not result in a corre- sponding deterioration in the target behavior.

Reinforcer Menus As we stated previously, each individual has his or her own history of rein- forcement (and punishment). What is reinforcing to one individual may or may not be reinforcing to another. We may make assumptions about poten- tially reinforcing stimuli or consequences based on a variety of variables (e.g., individuals of the same age, ethnicity, and socioeconomic background may share some similar reinforcers), but we may not be certain without careful examination on an individual basis. For example, we may assume that candy has a reinforcing quality for children in general, but it may not for any particular child. Teacher attention in the form of lavish praise may be considered potentially reinforcing for younger children, but may be less

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 33

so for adolescents. It may be necessary before conducting a study with one or more individuals to develop a reinforcer menu if positive reinforcement is to serve as an independent variable. A reinforcer menu may also be needed so that the individual can continue to obtain reinforcement (even completely internalized reinforcement such as self-satisfaction with one’s behavioral change) to maintain success. A reinforcer menu is a compilation of various stimuli (e.g., activities, things to eat, types of praise or statements, objects) that possess a reinforcing quality for the individual. There are a variety of strategies for determining the reinforcer menu for an individual. Hall and Hall (1980) outlined nine steps to assist in selecting reinforcers: (1) consider the age, interests, and appetites of the individual; (2) consider the behavior to be reinforced and the degree and quality of reinforcement that would be comparable to the behavioral effort (see the “Quality of Reinforcers” sec- tion in this chapter); (3) list potential reinforcers based on this information; (4) identify reinforcing activities that may be applied using the Premack Principle; (5) interview or ask the person about what he or she likes and dis- likes; also ask others what they observe about the individual’s preferences; (6) consider using consequences that will be new to the individual; (7) use reinforcers that are readily available and occur naturally in the environment; (8) select the reinforcers that you will make available based on the previous steps; and (9) record data to ensure the applied consequences are actually producing a reinforcing effect. Additional means for identifying reinforcers may include observing the individual in natural settings to determine prefer- ences and exposing the individual to activities/stimuli noncontingently to determine preferences and possible new reinforcers. For example, young children may need to be given access to various activities noncontingently so that parents, teachers, and others can identify activities that the child enjoys and that may serve as reinforcers. The researcher is attempting to obtain as large a variety of potentially reinforcing stimuli as possible. By varying the actual reinforcing consequence regularly, the researcher may avoid the occurrence of satiation. And, in the case of alternating treatments designs (see Chapters 11 and 12), the determination of which consequence has the greatest reinforcing quality may be critical to the study itself.

Satiation Satiation occurs when a previously reinforcing stimulus no longer possesses its reinforcing quality. Generally, the individual has been overexposed to the stimulus so that it is no longer effective. In short, the individual may no lon- ger choose to emit the target behavior in order to obtain the consequent stimulus. This occurs frequently when food or drink is used as a reinforcer, but may occur with objects, tokens, and activities as well. For example, a student who receives stickers only for completing work in class might, after

34 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

some time, no longer care about receiving stickers as she has acquired so many that they have lost their reinforcing quality for her. For a consequent stimulus to possess a reinforcing quality, a state of deprivation should exist. By deprivation, we do not mean a state of neglect or abuse, but rather that access to the stimulus should be controlled. Subsequently, obtaining it is important to the individual. In other words, the consequent stimulus has sufficient value for the individual to merit the behavioral effort required to obtain it. For example, a student who is promised some time on the com- puter to engage in an appropriate academic game might not find this rein- forcing if she has her own computer at home that she can freely access. The use of various reinforcers (reinforcer menu) may assist in providing a num- ber of quality or valued consequent stimuli. However, the researcher must also be wary about varying the type of reinforcer because it may also con- found the results of an experiment. For example, a student may perform a target behavior exceedingly well on days when the opportunity to play a card game is available, and perform less well when only a choice of reading materials is available. Thus, the researcher may also need to control what type of reinforcer is delivered. This can be a key element in an alternating treatments design. In addition to considering satiation by varying reinfor- cers, the practitioner or researcher must also consider the type of reinforcer delivered.

Primary, Secondary, and Generalized Reinforcers Reinforcing stimuli may be categorized into three groups: primary, second- ary, and generalized reinforcers. Primary reinforcers are those stimuli with which an individual requires no prior experience in order for the stimuli to possess reinforcing qualities (Kazdin, 1975). Food, beverages, and affection are examples of primary reinforcers. Generally, we can think of these as being life sustainers, although some take exception (e.g., some might argue that affection is a primary reinforcer that is clearly not necessary to sustain life, but also requires no prior experience). As a rule, primary reinforcers are used exclusively only when other types of reinforcers cannot be identified. For example, exclusive use of food as a reinforcer would be unwarranted unless no other reinforcer could be identified. Secondary reinforcers require some experience and derive their reinforcing properties from being paired with primary reinforcers (Kazdin, 1975) or existing secondary reinforcers. For example, the use of praise or a pat on the back that has been paired with affection generally begins to acquire a reinforcing property all its own. Generalized reinforcers are a special type of secondary reinforcer. A general- ized reinforcer is one that may be exchanged for any one of a variety of pri- mary or secondary reinforcers (backup reinforcers), and through that pairing the generalized reinforcer obtains its quality or value (Kazdin, 1975). Money

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 35

is the most obvious example of a generalized reinforcer. Money is so prized because it can be exchanged for many possible reinforcers and as such, it can be a reinforcer for a great many people as well. Tokens and point systems are more commonly used in school and clinical settings.

Generalized reinforcers have a number of advantages in that they may be delivered easily, be saved, and provide for individualization of the backup reinforcers; and at times, they may be taken away to punish inap- propriate behaviors. Secondary reinforcers, when they involve attention and praise, are also generally easily delivered and have the advantage of being immediate and readily available. Primary reinforcers are not easily delivered, may interfere with diet restrictions, and may be subject to rapid satiation. Whenever primary reinforcers are used, they should be paired with secondary or generalized reinforcers with an aim toward using the latter two.

Quality of Reinforcers The quality of reinforcers is yet another consideration for the researcher as she or he examines independent variable options. The type, degree, number, and so forth of reinforcer delivered should be reasonably commensurate with the achievement of the target behavior (or its reduction). For example, awarding a student $10 for completing a single seat assignment might be viewed by most educators and clinicians as rather excessive. Awarding $10 for achieving a 90% mastery level on a 9-week math test might be a more commensurate reward for the achievement. It should be noted that the history of the individ- ual participant and the magnitude of the changes sought in the behavior change project will greatly influence this decision. Generally, the more difficult it is to emit the target behavior or ultimately achieve the performance criterion, the greater might be the quality or quantity of the reinforcement.

Everett Revisited #2 Everett’s team is using a reinforcer menu in that Everett can select from a variety of activities for his weekend outing. Also, because Everett himself is selecting his reinforcer, it is less likely he will satiate on it (he can choose from a variety of activities). The reinforcing activity should also have value as a result of the self-selection as well. If in time, Everett’s

4 C H E C K I T O U T # 2 Do you think it would be appropriate to take away previously earned reinforcers from a child or adolescent for engaging in an inappropriate activity? For example, is it appropriate for a teacher to take tokens away from a child who arrives at class late?

36 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

performance of the target behavior was not improving or was deterio- rating, the team might decide to use generalized reinforcers (e.g., points awarded each day he takes his medication three times as prescribed) that would allow a more versatile reinforcer menu than activities alone.

Reinforcement Schedules When reinforcers are delivered systematically as in a research project, a schedule is typically predetermined. Schedules may be categorized as either ratio, interval (Ferster & Skinner, 1957; Skinner, 1953), or duration (Alberto & Troutman, 2013).

Ratio schedules are employed in projects when the target behavior (dependent variable) is easily observed each time it occurs. For example, the number of times or frequency of speaking complete sentences can be counted. Ratio schedules may be further categorized into continuous, fixed, and variable ratio schedules.

A variable ratio is achieved by varying the number of correct responses required so that when all correct responses are divided by the number of reinforcers delivered that average number is obtained. By using a VR sched- ule, the individual is less likely to anticipate accurately when reinforcement will be delivered and should emit the target behavior in a more consistent or steady manner. Consider that when praise is delivered in classrooms for stu- dents who are generally being successful learners, the teacher is probably using a VR schedule albeit not necessarily systematically. She occasionally praises students for completing work, following rules, making a substantial effort, attempting new tasks, performing well on an assignment, and so forth. The occasional praise is intermittent, not delivered in a predictable ratio, and is intended to ensure students continue to be successful without a continuing and specific reinforcement schedule.

Continuous Schedules In continuous schedules, every correct response is reinforced. Continuous schedules are used typically when the target behavior is being newly acquired. Despite what might be a commonsense notion that this schedule would result in great strength (the continued emission of the target behavior without delivery of a reinforcing consequence), the opposite is actually true. For example, inserting coins in a vending machine is very close to a contin- uous schedule of reinforcement. You insert your money and receive your selection. However, if the machine malfunctions and fails to deliver any selection (or your money back), you are not likely to continue inserting money in the anticipation that reinforcement will be forthcoming. For that reason, continuous schedules are used to establish correct responding and then are typically leaned or thinned toward intermittent schedules that

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 37

require more than one correct response to regularly obtain reinforcement (Skinner, 1953). Continuous schedules of reinforcement may be abbreviated as CR or CRF schedules in the research literature. They are typically used when a target behavior is being acquired and gradually modified to ensure maintenance of the target behavior.

Fixed Ratio Schedules These schedules require that the individual perform a set number of responses of the target behavior before the delivery of reinforcement. For example, a student may have to write 4 complete sentences before a reinforcer is delivered. The number of responses required may be gradually increased as responding improves (e.g., from 4 to 6 to 10 responses required to receive reinforcement). Fixed ratio schedules may be abbreviated as FR schedules (e.g., FR-4 would mean a fixed ratio of 4 correct responses required before the delivery of reinforcement). Fixed ratio schedules help to build resistance to extinction because the target behavior must occur without reinforcement being available, making it more difficult to anticipate when reinforcement will be delivered than with a continuous schedule.

Variable Ratio Schedules These schedules require the individual to emit an average number of correct responses to obtain reinforcement. Variable ratio schedules may be abbrevi- ated as VR schedules (e.g., a VR-10 schedule would require an average of 10 correct responses to obtain reinforcement). The variable ratio schedule makes delivery of reinforcement even less predictable and builds greater resistance to extinction than a continuous or fixed ratio schedule. Also vari- able ratio schedules are better for weaning participants off a systematic rein- forcement system and to a naturally occurring schedule.

A word of caution is necessary. If the ratio of correct responses is increased too quickly or becomes too high in number, ratio strain may occur (Alberto & Troutman, 2013)). Ratio strain occurs when the demands placed on the indi- vidual become too great or the individual perceives that reinforcement is not coming and decreases or ceases responding. Again in a classroom, the teacher who gradually ceases to praise her students’ work and effort may find that their performance also gradually deteriorates unless she ensures she continues to provide occasional praise. When ratio strain occurs, the practitioner or researcher should return to a previously successful schedule and then slow down the thinning process. This is especially important in research designs where positive reinforcement is the primary or only intervention.

Interval Schedules Interval schedules are used when the target behavior occurs so frequently that measuring each occurrence would be problematic. When using interval schedules (not to be confused with an interval recording system as described

38 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

in Chapter 3, “Methods for Recording Behavior”), the first correct response following the elapse of a predetermined time period is reinforced. For exam- ple, if the target behavior is complimenting others, the first correct response following a 1-minute (or 30-second, or 10-minute, etc.) interval would be reinforced. Like ratio schedules, interval schedules may be fixed or variable. That is, reinforcement may be delivered for the first correct response after a fixed interval of time (e.g., after 1 minute following the last delivery of rein- forcement) or after an average or variable interval of time (e.g., after an aver- age of 10 minutes). For example, speaking in complete sentences could be so frequent that attempting to reinforce each complete sentence used, or every third response, or every fifth correct response on average would be nearly impossible and might even prove disruptive to the speech itself. An interval schedule would allow a teacher or speech and language pathologist to identify the first correct usage following an interval of time much more easily than monitoring the number of correct responses. Interval schedules may be thinned just as ratio schedules and may be abbreviated as FI (fixed interval) or VI (variable interval) schedules.

Individuals may begin to anticipate when reinforcement is likely to be delivered, particularly with a fixed interval schedule. Also, the individual may slow responding following reinforcement due to an unspoken understanding that reinforcement will not be delivered again for some period of time. Variable interval schedules help to reduce these problems. The researcher may use a lim- ited hold contingency as well. A limited hold involves specifying a period of time in which reinforcement is available following the elapse of the interval, which encourages a steadier or faster pace of responding (Alberto & Troutman, 2013). It is important to remember that even though reinforcement is being governed to some extent by the passage of time, interval schedules of rein- forcement are used with target behaviors that are counted rather than time based. When we are concerned primarily with how long a behavior is emit- ted, we use response duration schedules of reinforcement.

Response Duration Schedules Response duration schedules may be either fixed or variable (Alberto & Troutman, 2013). The same thinning procedures as those discussed with other intermittent schedules may be applied. In duration schedules, the researcher is delivering reinforcement based on how long a behavior is con- tinuously emitted. With a fixed response duration schedule, the researcher reinforces the individual after the target behavior has been continuously emit- ted for a fixed period of time (e.g., after every 2 minutes). With a variable response duration schedule, the individual is reinforced following an average length of continuous emission of the target behavior (e.g., VD-5 min). The averaging is achieved in the same manner as with variable ratio schedules, substituting time periods for the number of responses required. With fixed

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 39

response duration schedules, after being reinforced, the individual may stop emitting the target behavior following reinforcement as he or she gains under- standing that reinforcement will not be delivered for some period of time or because the time required to earn reinforcement is too great (Alberto & Troutman, 2013). Variable schedules help to reduce this problem.

Natural Reinforcement Alberto and Troutman (2013) suggested that reinforcement should be related to naturally occurring reinforcing events such as praise, activities, and tangibles such as star charts, certificates, and good grades/evaluations. Interventions using specific reinforcers that may be less natural (e.g., provid- ing very frequent food reinforcers for correct classroom responding) should eventually be replaced by naturally occurring reinforcers and therefore, nat- ural schedules of reinforcement as well. It is advisable that researchers select reinforcers that are naturally linked to the individual’s life and the target behavior as well. If this is accomplished, the target behavior is more likely to be maintained in the absence of a specific reinforcement intervention (Alberto & Troutman, 2013). Such specific reinforcement interventions are often costly in terms of time and effort and their withdrawal to naturally occurring and natural schedules of reinforcement is sometimes reflected in a final follow-up phase in a single subject design that occurs after the inter- vention phase(s).

Generalization of Target Behaviors Practitioners and researchers are concerned with maintaining behaviors through natural reinforcement. They are also concerned with generalization of target behaviors (emitting the target behavior in a new situation or setting, emitting similar but new responses). Generalization is often a concern when the target behavior is learned in only one situation or environment. For example, an individual who learns to praise herself for eating properly and drinking sufficient water in a hospital setting after having been homeless for some time should also continue to eat and drink appropriately when she returns to everyday life. Similarly, a student might develop a new study skill as a result of having enjoyed success in learning one that was directly taught him, even though the new skill was not directly taught. Generalization is sometimes systematically programmed in single subject research. For example, an individual who has learned to use complete sentences while speaking in school might also have the intervention used extended to less-structured school settings (e.g., in the cafeteria), at home, and in the community (e.g., while at church) by his family and close friends following success of the inter- vention in the controlled classroom environment. Target behaviors often should be generalized such that they occur in natural environments and improve the overall functioning and well being of the individual participant.

40 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Methods to Decrease Behavior As noted previously, there is debate as to whether certain methods to decrease behavior should ever be employed. Corporal punishment is an obvious example. Less obvious examples (use of foul-smelling odors, water sprays to the face) have also been debated. Our position is that the methods to decrease behavior should be discussed, as they are available and poten- tially useful. Clearly, the use of any method with potentially adverse side effects must be rigorously reviewed. Although punishment is certainly a method that may be used to decrease behavior, other methods are available that are generally perceived to be less restrictive and are more acceptable by the individual participant and his/her advocates.

Arguments against the use of punishment include the following: (a) the individual learns what not to do, not what to do; (b) it may create a model of aggression and physical control to be emulated by the individual; (c) it may inflict pain or hardship on the individual; (d) at times, it is used with individuals who are unable to express an unwillingness to participate in such measures (e.g., a nonverbal individual with profound disabilities); (e) it merely suppresses behavior and may not eliminate the undesired response, particularly in other settings or situations; and (f) individuals may avoid or escape from environments where punishment is employed. This list is by no means exhaustive.

Although most professionals could hardly be characterized as propo- nents of punishment, arguments may be made in favor of its use in special circumstances. One such circumstance might be when an individual exhibits a behavior that is clearly dangerous to oneself or to others and must be immediately suppressed (e.g., attempting to harm others with scissors, run- ning into traffic, or arson). When aversive stimuli are presented to suppress (i.e., punish) behavior, Wolery, Bailey, and Sugai (1988) outlined the fol- lowing considerations: (a) aversiveness must be individually determined; (b) it may be necessary to use more intense levels of aversive stimuli to achieve a desired reduction in behavior; (c) side effects should be expected; (d) the maintenance of the behavior reduction may be variable; (e) the aver- sive stimuli should be delivered consistently and immediately; and (f) the use of aversives should be restricted and carefully monitored. Each individual researcher or practitioner along with a team must consider many variables when designing a single subject study. However, the use of punishment may require more rigorous examination by other appropriate boards (human treatment or human subjects review boards) before approval of use. The side effects are numerous. The use of severe or unusual forms of punishment is generally to be avoided for the sake of one and all. We will discuss punishment methods followed by a discussion of less aversive techniques advocated by opponents to punishment.

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 41

Positive Punishment As with positive reinforcement, this term should not be construed to mean that punishment is a good thing. The positive here refers to the consequence for behavior rather than a quality of the phenomenon. Again, an antecedent occurs, the behavior occurs, and the consequence of the behavior is an addi- tion to the environment (e.g., a verbal reprimand), and the result is that the probability of future occurrence under the same or similar circumstances is decreased. The future reduction or weakening of behavior essentially defines punishment. As with positive reinforcement, it is important to remember that either desirable or undesirable behavior may be punished. For example, an adolescent who is given lavish praise for an oral response in class becomes less inclined to volunteer answers in the future. Although the teacher may have thought she was positively reinforcing the student, the consequence was that the student became less likely to emit the behavior under similar antecedent conditions. Thus, she actually was using positive punishment. Intentional positive punishment in the schools and clinical environments is probably most often in the form of verbal reprimands, although there are many other possibilities.

Each student has an individual history of punishment, just as for rein- forcement. Therefore, what is punishing to one individual may not be to another. The individual’s response to the consequence is what determines whether or not punishment has occurred. For example, one student may find the teacher’s reprimand punishing, while another may actually enjoy the attention itself and be reinforced by the reprimand rather than punished. Additionally, punishing consequences may be primary or secondary. Sche- dules of punishment are not generally discussed in the literature, as typically the researcher or practitioner would wish to address each and every instance of the targeted behavior. Negative punishment may also be used, although, unlike with negative reinforcement, an aversive antecedent is not necessary.

Negative Punishment Negative punishment also results in a decrease in the probability of the occurrence of the target behavior. In negative punishment, an antecedent occurs, the behavior is emitted, and the consequence is the removal of some- thing from the environment. The result is that the probability that the behavior will occur again under the same or similar antecedent conditions is reduced. For example, a practitioner asks a group of students a question, one blurts out an answer, the practitioner takes away one point from the student’s accrued total, and the student does not blurt out answers during the remainder of the session. Removal of privileges for misbehavior is another example of negative punishment. Negative punishment does require

42 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

the presence of some desired or preferred stimuli that can be removed as a consequence of the target behavior. It is important to remember that remov- ing something from an individual’s world is not always easily accomplished and may lead to undesirable reactions and side effects such as verbal or even physical outbursts.

Response cost is a form of negative punishment. In response cost, the practitioner or researcher assigns a fine for one or more specified target behaviors. When the target behavior occurs, the fine is levied. Generally, response cost is used in conjunction with token economies or point systems that focus more on what to do to earn reinforcement than what not to do (i.e., there are behaviors targeted for reinforcement as well). It should be noted, some professionals and parents insist that once something is given to a child, it should not be taken away.

Extinction Although not typically defined as punishment, extinction bears a resem- blance to negative punishment. With extinction, the reinforcer(s) of a behav- ior is identified and removed or withheld when the behavior occurs (Alberto & Troutman, 2013). Extinction relies on the premise that once reinforcement is withheld, the behavior eventually will decrease or be elimi- nated. As an example, an individual may use gestures to communicate when speech is more appropriate and within the individual’s repertoire. The prac- titioner or researcher could withhold any reinforcement that occurs as a result of gesturing (e.g., responding to a communicative request). In time, the gesturing should decrease if no reinforcer is available to maintain its strength. However, extinction tends to work best when the reinforcer is tangible and can be easily withheld. Two preschoolers may fight over who is going to play with a toy. Removing the toy altogether may eliminate the fighting which is reinforced by occasionally “winning” access to the toy.

A few words of caution should be given regarding the use of extinction. First, when it is implemented, the targeted behavior may continue to occur for some time (Alberto & Troutman, 2013). In the example given, the pre- schoolers may become even more disruptive when they realize the toy is not available. This realization can take a period of seconds or weeks. Generally, one may assume that the longer a behavior has been reinforced the more resistant it will be to extinction. That is, it will take a longer period for

4 C H E C K I T O U T # 3 Can you distinguish between negative reinforcement and negative punishment? How are they alike? How are they different?

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 43

extinction to occur. It is not always easy or even possible to hold out and not reinforce during the burst of increased responses that sometimes accom- panies the implementation of an extinction intervention. For example, if the behavior to be extinguished is calling-out by a student during class discus- sions, the increase in the behavior may be so disruptive that it cannot be ignored. Ironically, if the teacher does respond and teacher attention serves as a positive reinforcer, the calling-out behavior may be strengthened beyond the level that existed before extinction was introduced (because the teacher has accidentally used a variable ratio schedule of reinforcement). For this reason, the nature of the target behavior must be carefully considered. Second, it is not easy and sometimes is impossible to identify or control the reinforcer for a given response. Reinforcement may be available in so many environments and from so many sources that withholding it is impractical. Reinforcement may have been internalized and therefore the reinforcing consequence is not observable and not controllable by the researcher. Third, the withholding of reinforcement may result in intense reactions from the individual. Finally, some behaviors simply should not be treated through extinction. For example, a child may run away for the reinforce- ment of being chased. However, one cannot simply allow children to run away and forget about them. It is important to note that spontaneous recovery also may occur. Spontaneous recovery refers to circumstances when, for no apparent reason, an individual emits a behavior that had been extin- guished (Alberto & Troutman, 2013). Generally, the intervention may be reinstituted and the behavior eliminated once again and in less time than the original procedure. Extinction may be best employed when practitioner attention is reinforcing a relatively mild maladaptive response that has not been subjected to powerful and extensive reinforcement (e.g., ignoring con- versational interrupting). This may reduce the strength of the extinction burst if it does occur and may lead to a quick and complete extinction of the response. Extinction probably also is best used in conjunction with dif- ferential reinforcement so that the individual may learn new behaviors for which reinforcement is available.

We will discuss less aversive methods of reducing behavior through the use of differential reinforcement. Response interruption, response satiation, and overcorrection are methods that also may be used to weaken target behaviors.

Differential Reinforcement Differential reinforcement may be the preferred method for decreasing behavior because it incorporates behavior change strategies that assist in avoiding some of the more common adverse effects of punishment. With dif- ferential reinforcement, the target behavior to be weakened or eliminated is

44 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

identified. More desirable behaviors are identified that may replace the tar- get behavior (Dietz & Repp, 1983). These desired behaviors are reinforced while the target behavior is also being weakened. The strategies are varied for reducing the target behavior (e.g., positive or negative punishment, extinction, and response interruption). Differential reinforcement has dis- tinct advantages over programs designed only to decrease behaviors in that (a) the individual is also learning what to do as well as what not to do; (b) the individual’s overall level of reinforcement is less likely to be reduced; (c) the individual and her or his advocates (as well as human subjects review teams) may find such a program preferable to one that focuses on punish- ment; (d) the individual observes a model for encouraging behavior change that also stresses reinforcement; and (e) the individual may be less likely to wish to escape or avoid the environment if reinforcement is available.

Differential reinforcement takes several forms. These include differential reinforcement of other (or omitted or zero rates of) behavior (DRO), differ- ential reinforcement of incompatible behavior (DRI), differential reinforce- ment of alternative behavior (DRA), and differential reinforcement of low rates of behavior (DRL).

DRO DRO involves the rewarding of the absence of the targeted behavior for a specified period of time. Although it is not absolutely necessary that the individual understand the absence of the targeted behavior, DRO may be more effective when this is the circumstance. That is, the individual may be more likely to change his or her behavior in the desired direction if he or she understands not to perform the target behavior. The individual is rein- forced if the target behavior is not emitted during the specified time period. For example, a student may agree that she emits highly undesirable verbal comments (e.g., cursing, verbal threats, or uncomplimentary remarks to peers). The researcher or practitioner then rewards the student for not emit- ting the verbal comments for a period of time (e.g., 10 minutes). Theoretically, the student should be rewarded for behaviors other than the target behavior. In practice, researchers may find it necessary as well to target other undesir- able behaviors as well (e.g., other rule violations such as hitting, throwing things, or disrupting the class in some other fashion; see multiple baseline designs chapters). The researcher must also develop a contingency if the target behavior is emitted. Generally, the timer is reset and the student is informed as to what was done. For example, if the student did emit an inappropriate remark at 7 minutes into the time period, the teacher might inform the student of what she said, that it was inappropriate, and the timer would be started over requiring her to go the full 10 minutes in order to obtain rein- forcement. Other measures may include response interruption (see later dis- cussion) or response cost. The time required to not emit the behavior is

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 45

initially set at a realistically achievable level (e.g., one half the average time between emissions during baseline). The student must be able to achieve the reinforcement in order to expect him or her to change behavior. As progress is achieved, the length of time required to obtain reinforcement may be lengthened. Eventually, the student with the inappropriate comments may be required to go a day or even a school week without such an emission in order to obtain reinforcement.

DRI/DRA Differential reinforcement of incompatible behavior (DRI) and differential rein- forcement of alternative behavior (DRA) are similar to one another and some- what different from DRO. With DRI/DRA, specific adaptive responses are identified to replace the maladaptive responses. With DRI, the targeted adap- tive response is one that is physically incompatible with the targeted maladap- tive response (e.g., if a child is in her seat, she cannot be out of her seat at the same time). With DRA, the individual is rewarded for a more desirable response but one that is not physically incompatible with the target behavior (e.g., one may give compliments rather than insulting others). With DRI/DRA, the individual is reinforced for the desired response. There must also be a con- tingency plan for when the targeted undesirable response is emitted as well.

DRL Differential reinforcement of low rates of behavior is used when the targeted behavior is an appropriate response that occurs at an inappropriate level or is a behavior in need of elimination. In the first instance, going to the bath- room clearly is appropriate behavior, but going several times in a class period is not (assuming no physical reasons for this need). In the second instance, smoking cigarettes serves as a common example. With DRL, the researcher systematically rewards the individual for emitting fewer and fewer responses until the behavior is occurring at acceptable levels or has ceased altogether (see changing criterion designs chapters). DRL is sometimes

4 C H E C K I T O U T # 4 Assume a student has two undesirable behaviors being addressed by her team. One behavior is she frequently disrupts classes in her school by blurting out answers to teacher questions before anyone else has an opportunity to respond or even raise their hand to be recognized by the teacher. This same student also makes rude com- ments to other students during class when they give an incorrect response to a teacher question or ask the teacher a question. Which behavior might be better addressed through DRI/DRA and which might be better addressed by DRO?

46 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

considered a shaping procedure because the intervention is designed to grad- ually alter the response until the desired criterion level is reached.

Everett Revisited #3 Everett’s Vignette serves to illustrate a common occurrence in the use of applied behavior analysis. Everett does not receive his reinforcer if he does not meet the criterion for his target behavior (taking his medication three times each day as prescribed for the week). This is not punishment, but his team would be wise to consider that Everett, or any individual, may react to not receiving reinforcement as if punishment has occurred. Everett could become agitated or angry when he does not receive his rein- forcement. The team did anticipate this possibility and made sure the social worker carefully explained to Everett why he did not receive his reinforcer and how he could obtain it by his future performance.

Other Methods to Decrease Behavior Other methods are available for decreasing behavior, including response inter- ruption and overcorrection. These methods are not typically considered pun- ishment procedures, but may nevertheless elicit some of the same side effects as they may involve physical contact with students and removal of reinforcers. In each of these procedures, the teaching of alternative behaviors that may be reinforced is very desirable as an element of practice or a research study.

Response Interruption Response interruption is exactly as it sounds; the practitioner or researcher literally interrupts the emission of the target behavior (which is to be reduced or eliminated in strength). This may involve a verbal or physical interruption. For example, if a student has had some inappropriate speech targeted for reduction, the practitioner may verbally interrupt the student to stop the targeted behavior. Typically at this point, the individual would be provided with an opportunity to emit more appropriate speech and be reinforced for that. For another example, out of seat behavior may have been targeted for reduction. The intervention may involve physically guiding the student back to her seat whenever she gets up. As before, the student typically would also be reinforced for being in her seat. Response interruption may be used in conjunction with differential reinforcement.

Overcorrection Overcorrection is a procedure that is intended to teach the individual an alter- native or desirable behavior that corrects the effects of the behavior targeted for reduction (Azrin & Foxx, 1971). The corrective response is repeated over and over to enhance the likelihood that that response will be learned, or an exaggerated form of an adaptive behavior is performed (Foxx & Azrin,

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 47

1973), hence, the use of the term overcorrection. Overcorrection includes two basic intervention procedures: restitutional and positive practice overcorrection.

Simple restitution occurs frequently in clinics, classrooms, homes, and workplaces throughout each day. Simple restitution (which is not overcorrec- tion) involves restoring an environment to its previous condition. For example, a student spills some paint during an art activity and is required to clean up the mess. In everyday life, this typically serves the desired purpose to reduce the likelihood that the undesired response will occur in the same or similar circum- stances. The cleanup also provides learning in what must be done when spilling occurs and the student is likely to be more careful in the future.

Restitutional Overcorrection. Restitutional overcorrection involves restor- ing the environment to a better than previous condition. When the behavior targeted for reduction occurs, the individual is required to perform exagger- ated responses that not only correct the consequences of the target behavior, but actually improve the environment. For example, an individual has van- dalized a school setting by writing on walls. As an overcorrection procedure, the student is required to restore not only the areas of walls on which he wrote, but other areas of the wall as well.

Positive Practice Overcorrection. Positive practice overcorrection involves the repeated practicing of an alternative to the target behavior. For example, if an individual slams doors each time he leaves or enters a room, the indi- vidual could be required to repeatedly practice closing a door quietly each time he slams a door. Positive practice involves teaching the individual a more adaptive behavior than the target behavior and typically involves repeated or massed practice in a relatively short period of time.

Overcorrection may involve physically manipulating an individual to perform the correction procedures. Also, the procedure may require individ- ual supervision and be time consuming as well. Overcorrection may not be easily implemented in natural environments and may lead to undesirable reactions if physical contact is needed.

Summary The procedures we have discussed may involve various combinations and variations as reported in the research literature. This is by no means an exhaustive discussion, and the reader is encouraged to more thoroughly examine literature related to applied behavior analysis and other interven- tion procedures before selecting an independent variable for any particular project. This discussion should provide general guidelines and knowledge for reflection on options available, and some of the considerations needed in selecting an independent variable, or in understanding procedures often found in the research literature.

48 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Key Concepts/Terms Positive reinforcement—A consequence is delivered following the emission

of a behavior, and the probability that the behavior will occur again under the same or similar circumstances (antecedent conditions) is increased or strengthened; positive refers to the consequence being the addition of something to the individual’s environment, not a quality such as good.

Negative reinforcement—An aversive antecedent stimulus is introduced, a behavior occurs, and the consequence is that the aversive stimulus is removed; probability of the behavior occurring again under the same or similar circumstances is increased; negative refers to the removal of something from the individual’s environment as a consequence, not a quality such as bad; when behavior results in escape or avoidance of some unpleasant situation or event, negative reinforcement may be occurring.

Premack Principle—The use of a highly preferred activity as a reinforcing consequence for performing a lower-preference activity (target behav- ior); access to the high-preference activity must be controlled; commonly used in homes, schools, and clinics.

Shaping—Reinforcement of closer and closer approximations to a criterion level of performance of the target behavior; used to teach new behaviors.

Reinforcer menus—Individually determined lists of known or possible rein- forcing consequences.

Satiation—Occurs when repeated exposure to a reinforcing consequence results in loss of the reinforcing quality (e.g., too much candy as a rein- forcer results in little effort to obtain more candy).

Primary reinforcers—Reinforcers that require no previous exposure to possess reinforcing qualities; often thought of as life sustaining or funda- mental to existence (e.g., food, water, warmth).

Secondary reinforcers—Reinforcers that have obtained their reinforcing quality through pairing with primary reinforcers or existing secondary reinforcers (e.g. praise, good grades); generally preferred for use over primary reinforcers.

Generalized reinforcers—Reinforcers (e.g., tokens, points, money) that are delivered in lieu of either primary or secondary reinforcers, but can be exchanged later for other reinforcers.

Quality of reinforcers—The type, number, and degree of reinforcement should be reasonably commensurate with the target behavior.

Reinforcement schedules—Reinforcement may be delivered through a vari- ety of schedules; typically, reinforcement should be moved from more frequent and predictable to less frequent and less predictable.

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 49

Continuous schedule—Every correct response of the target behavior is rein- forced; used to establish new behaviors; not resistant to extinction (i.e., continuance of the behavior is unlikely in the absence of reinforcement).

Fixed ratio and variable ratio—Delivery of reinforcement after a specific number of correct responses or a variable number, respectively; variable ratio schedule should result in more consistent or steadier responding.

Ratio strain—Occurs when the demands placed on the individual become too great or the individual perceives that reinforcement is not coming and decreases or ceases responding.

Fixed interval and variable interval—Reinforcement is delivered for the first correct response following a specified or variable period of time; variable interval schedule should result in more consistent responding.

Fixed response duration and variable response duration—Reinforcement is delivered for the continuous occurrence of a target behavior for a specified or variable length of time (duration); variable duration sched- ule should result in more consistent responding.

Positive punishment—Following a behavior, something is added to the individual’s environment, and the probability of occurrence of the behavior under the same or similar conditions (antecedents) is decreased or weakened; each person has an individual history that determines which consequences will possess a punishing quality; punishment is to be avoided due to its many side effects.

Negative punishment—Following a behavior, something is removed from the individual’s environment, and the probability of occurrence of the behavior under the same or similar conditions (antecedents) is decreased.

Response cost—Typically a fine is levied for occurrence of a target behavior (negative punishment); frequently used with token, point, or monetary systems.

Extinction—The withdrawal or withholding of reinforcement following a response; behavior may actually increase before it begins to decrease.

Spontaneous recovery—The unexpected occurrence of a target behavior that had been previously extinguished.

Differential reinforcement—Use of methods of reinforcement to decrease behavior; often preferred because these methods are less intrusive than punishment.

Differential reinforcement of other (or omitted or zero rates of) behavior (DRO)—The individual is reinforced for not emitting the target behavior for a period of time.

Differential reinforcement of incompatible/alternative behavior (DRI/ DRA)—The individual is reinforced for an adaptive response that is intended to replace the maladaptive target behavior.

50 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Differential reinforcement of low rates of behavior (DRL)—The individual may be reinforced for reducing the level of responding of a behavior to an appropriate level (e.g., number of times going to the bathroom), or the reinforcement of a target behavior until a zero level of responding is achieved (e.g., smoking cigarettes).

Response interruption—The individual is interrupted when the target behavior is emitted; typically, the individual is redirected toward emit- ting a more adaptive response.

Overcorrection—Having an individual repeatedly perform a more adaptive behavior or performing an exaggerated adaptive response when the tar- get behavior occurs; overcorrection is intended to teach the individual what to do, not just what not to do; may involve physical contact and often requires one-on-one attention to implement.

Simple restitution—This is not true overcorrection; refers to restoring an environment to its original condition (e.g., mopping the floor after throwing a liquid on it).

Restitutional overcorrection—Restoring the environment to a better than previous condition (in the above example, mopping several floors in addition to the one where liquid was spilled).

Positive practice overcorrection—Repeatedly practicing an adaptive response (e.g., repeatedly practicing quietly closing a door following slamming a door).

4 Possible Answers to Check It Out

¶ By definition, negative reinforcement leads to an increase in behavior,the opposite of punishment. The term negative, however, leads to confusion such that many individuals mistakenly assume the term means a negative outcome, negative approach, or reinforcing a negative or undesir- able target behavior. Negative refers to the consequence in the contingency (withdrawal of an aversive antecedent as a result of performing the target behavior).

· The questions are commonly asked in school settings. Some profes-sionals and parents believe it is wrong to take something away from a child that you have given to them. Others believe it is perfectly appropriate to “fine” a student for inappropriate behavior because the tokens are above and beyond what is commonly given to students and therefore, taking tokens away is justifiable. There is no “right” answer, and the procedures need to be individualized to the participants and the setting/situation. When you have read the section on response cost, you may or may not change your answer.

CHAPTER 2 METHODS FOR CHANGING TARGET BEHAVIORS 51

¸ Negative reinforcement and negative punishment are alike in that theconsequence in each procedure is the removal of something from the individual’s world. Specifically in negative reinforcement, an aversive ante- cedent stimulus is removed. However, negative reinforcement results in an increase in the target behavior. Negative punishment results in a decrease in the target behavior.

! The blurting out of answers might be better addressed through DRI/DRA. The student clearly needs to learn to raise her hand and wait to be recognized before answering the teacher’s questions. This might be con- sidered more DRA because she could both raise her hand and still blurt out answers; again, the important aspect of the contingency is that she is work- ing towards learning a more acceptable target behavior to replace the unac- ceptable one. The rude comments might be better addressed through DRO. The student could be positively reinforced for not making rude comments. One might also consider that rude comment behavior might also be addressed through DRI/DRA if the student was taught to make more complimentary or encouraging statements and was reinforced accordingly for doing so.

References Alberto, P. A., & Troutman, A. C. (2013). Applied behavior analysis for teachers

(9th ed.). Upper Saddle River, NJ: Pearson. Azrin, N. H., & Foxx, R. M. (1971). A rapid method of toilet training the institu-

tionalized retarded, Journal of Applied Behavior Analysis, 4, 89–99. Dietz, D. E. D., & Repp, A. C. (1983). Reducing behavior through reinforcement.

Exceptional Education Quarterly, 3, 34–46. Ferster, C. B., & Skinner, B. F. (1957). Schedules of reinforcement. Englewood

Cliffs, NJ: Prentice-Hall. Foxx, R. M., & Azrin, N. H. (1973). The elimination of autistic self-stimulatory

behavior by overcorrection. Journal of Applied Behavior Analysis, 6, 1–14. Hall, R. V., & Hall, M. C. (1980). How to select reinforcers. Lawrence, KS: H & H

Enterprises. Kazdin, A. E. (1975). Behavior modification in applied settings. Homewood, IL:

The Dorsey Press. Premack, D. (1959). Toward empirical behavior laws: I. Positive reinforcement.

Psychological Bulletin, 66, 219–233. Skinner, B. F. (1953). Science and human behavior. New York: Macmillan. Wolery, M., Bailey, D. B., Jr., & Sugai, G. M. (1988). Effective teaching principles

and procedures of applied behavior analysis with exceptional students. Needham, MA: Allyn & Bacon.

52 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

CHAPTER

3 Methods for Recording Behaviors

IMPORTANT CONCEPTS TO KNOW THE IMPORTANCE OF OBSERVABLE, MEASURABLE BEHAVIOR

Interobserver Agreement Taylor’s Vignette

Choosing a Recording Procedure

EVENT-BASED METHODS FOR RECORDING AND REPORTING BEHAVIOR Frequency Recording Rate Interval Recording

Taylor Revisited #1

Evaluating Permanent Products

TIME-BASED METHODS FOR RECORDING AND REPORTING BEHAVIOR Duration Recording Latency Recording

RECORDING PROCEDURES FOR SPECIFIC PURPOSES Trials to Criterion Recording Cumulative Recording

Taylor Revisited #2

A FEW WORDS ABOUT THESE METHODS Choosing the Appropriate Recording Procedure Summary

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

The Importance of Observable, Measurable Behavior As with independent variables, there is virtually an infinite variety of depen- dent variables that may be used in single subject research. When we refer to recording dependent variables, we are discussing the methods used to quan- tify the target behavior and thus determine the effectiveness of the indepen- dent (i.e., treatment or intervention) variable(s). As noted in Chapter 1, it is important that the dependent variables (target behaviors) be precisely defined, observable, and measurable.

At times, the general behavior of concern may be difficult or even impossible to actually observe. The behavior may be related to some inter- nal cognitive process, which of course cannot be directly observed. Then the dependent variable or target behavior must be some outcome that reflects the occurrence of the more general behavior. For example, a social worker may be interested in helping a client cope with a stressful marital situation by having him “think” before he “acts.” The overall goal may be for the client to reflect on possible actions to be taken before actually taking action in a stressful situation (i.e., the behavior of concern is reflecting on one’s possible actions). However, since reflection cannot be directly observed, it must be operationally defined in such a way that it can be observed and measured. One way might be to identify the time that elapses between the onset of the stressful situation and the beginning of the action as the depen- dent variable (in this case “waiting time” is the target behavior that can be recorded). Also, the researcher might measure the accuracy of the indivi- dual’s ability to verbally recite a series of internal steps to follow that are indicative of reflection. The researcher might even measure the outcomes of reflection, such as responding in a nonthreatening manner, after the stressful event occurs. In each of these cases, the researcher identifies a target behav- ior that is observable and measurable and is indicative of the overall goal being achieved. These direct quantitative measurement techniques are those employed in single subject research. Before discussing the general definitions and use of the different recording procedures, it is important to address the issue of interobserver agreement. This procedure is used to ensure that the behavior being observed is precisely defined and measured.

Interobserver Agreement Interobserver agreement, sometimes referred to as interrater agreement, is determined when the occurrence of the target behavior is viewed by at least two individuals and measured independently by those individuals. This type of interobserver agreement thus refers to the level of agreement that is reached concerning the measurement of the dependent variable (e.g., how long, or how frequently the target behavior occurred).

54 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Before determining interobserver agreement, observers should be trained regarding the mechanics and issues (e.g., observer drift) related to observing and recording behavior. We include our discussion of how observers are trained in Chapter 4, “Issues in Single Subject Research.” The following serves as a general example to demonstrate the mechanics of calculating inter- observer agreement. Specific methods for determining interobserver agree- ment are addressed in our description of each of the recording procedures.

To calculate interobserver agreement, one must have two observers recording the same behavior using whichever recording method is chosen. The recordings of the two observers are then compared to determine the percent of agreement. Suppose that a student, Zack, was being observed to determine the amount of “off task” behavior he was exhibiting. After care- fully defining off task behavior, the two observers recorded whether or not Zack was off task at any time during a series of 10-second intervals (called partial interval recording, described later). A plus (!) is used to indicate that the off task behavior occurred and a minus (") to indicate that it didn’t. The following data were collected:

Interval 1 2 3 4 5 6 7 8 9 10 Observer 1: ! ! " ! ! ! " ! ! ! Observer 2: ! ! ! ! ! ! ! ! " "

Interobserver agreement could be determined by using the following formula:

Interobserver # Number of Agreements

Number of Agreements ! Number of Disagreements $ 100%

Agreement

Note that it is important to observe, record, and compare each interval rather than simply looking at the total number of agreements/disagreements. In the example, if we merely compared how many intervals Zack was off task versus on task each observer recorded, we would find that each recorded 8 off- task and 2 on- task intervals. Theoretically, one might arrive at a 100% interobserver agreement figure. However, if we examine the observers’ recordings more carefully, we find this is not the case. Really, the observers disagreed about intervals 3, 7, 9, and 10, creating four dis- agreements. They agreed about Zack’s off task/on task behavior in intervals 1, 2, 4, 5, 6, and 8. Using the formula, this would indicate that the interob- server agreement was 60%.

6 6 ! 4

$ 100% # 6 10

$ 100% # 60%:

The “standard” used to determine acceptable agreement in the single subject research is a minimum of 80%, and that interobserver agreement checks be conducted on at least 20% of the total observation sessions in a study (Kennedy, 2005). Some have suggested a more stringent criteria of

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 55

90% or higher as desirable, 70-89% as adequate, and less than 70% as questionable as to whether the procedures were consistent, fair, and rigor- ous (Zirpoli & Melloy, 1993).

4 C H E C K I T O U T # 1 Calculate the interobserver agreement for the following interval data from two recorders (A and B).

Interval 1 2 3 4 5 6 7 8 9 10 Recorder A ! ! ! ! " ! ! ! " ! Recorder B ! ! " ! " ! ! " " ! Would you characterize the level of agreement as desirable, adequate, or questionable?

Taylor’s Vignette Taylor is a 6-year old girl with severe intellectual disabilities and paraple- gia as the result of cerebral palsy. Although she was eligible for early intervention services and for preschool services, the delivery of those ser- vices was very sporadic due to family issues, loss of communication, and frequent moves, and only recently began receiving consistent program- ming. Taylor has developed a pronounced behavior of “hand-flapping.” Her educational team consisting of a school psychologist, her mother, teachers, and other related services personnel have all agreed that this behavior interferes severely with Taylor’s learning. A functional behavior analysis suggests the hand-flapping is exhibited when Taylor is excited or is asked to perform a behavior she does not appear to want to do. It also likely is being used as self-stimulation. It occurs at a high rate. Taylor can be observed flapping her hands during virtually every activity and time period of the day. The team has agreed on a definition of hand-flapping as a target behavior, but now must decide how best to record the behav- ior. One team member suggested the frequency of the behavior be counted. Another suggested that the duration be counted. Still another suggested that the behavior be measured using either a partial or whole interval recording procedure. The team must decide which procedure is most appropriate. They have also brainstormed interventions and have decided that response interruption (gently, physically stopping Taylor from hand-flapping when she begins) and using differential reinforcement of incompatible behavior (praising Taylor enthusiastically for engaging in more developmentally appropriate behaviors such as play, sorting, string- ing beads, helping prepare snacks, etc.) would be appropriate.

56 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Choosing a Recording Procedure In this chapter, we will discuss how quantitative measures may be used to assist in determining the social and ecological validities and outcomes of sin- gle subject research. Several issues, suggestions, and specific procedures will be addressed. First, and foremost, we will discuss target behaviors and specific methods that are used to record changes in different types of target behaviors. For each type of recording procedure, six questions will be answered:

1. When should it be used?

2. What do you record?

3. How is it calculated?

4. How is interobserver agreement determined?

5. What is an example?

6. What are some special considerations?

We will also provide sample data recording sheets for many of the pro- cedures that may be used (these are included in the Appendix at the end of the chapter). The recorded data should also be visually displayed. Although there are options available to the researcher (e.g., bar graphs), we assume quantitative data typically will be reported using an x-y or line graph that was discussed more thoroughly in Chapter 1. We must stress that the possi- ble available recording procedures are not limited to those discussed here, but these do represent those more commonly used in single subject research. Quantitative dependent variables may be generally categorized into two groups: (a) those primarily designed to measure the frequency, accuracy, or intensity of a target behavior (event-based); and (b) those that are more time-based and used to measure how long a behavior occurs or how long it takes a behavior to begin. It is important to remember that these depen- dent variables should be observable, measurable, and reliable, and some effort and time is often devoted to specifying the target behavior and mak- ing sure the most appropriate procedure is used. Anecdotal recording often is useful in this process.

Event-Based Methods for Recording and Reporting Behavior As noted previously, there are two general groups of measurement procedures—event-based methods and time-based methods for recording behavior. We begin our discussion with those that are generally considered event-based methods (typically concerned with the frequency, accuracy, or intensity of the target behavior). Two procedures, frequency recording and inter- val recording, generally address the issue of how often a behavior occurs with frequency recording perhaps being the most widely used event-based method.

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 57

Frequency Recording When Should it Be Used? Frequency recording, a basic and simple record-

ing procedure, is often used in single subject research. It is used when the number of behaviors that occur is important as the dependent variable. The term frequency is often used synonymously with the term number (Ayres & Gast, 2010) and also is often referred to as rate (although tech- nically there is a difference between frequency and rate). When using fre- quency recording, one simply counts the number of occurrences of the target behavior within a specific time period. However, the length of the observation sessions must remain constant. For example, if a student interrupted his teacher six times during a 30-minute math lesson and three times during a 30-minute science lesson, frequency recording could be used and the number of interruptions could be directly compared. If, however, the student interrupted his teacher six times during a 30-minute math lesson and three times during a 20-minute science lesson, frequency recording could not be used and the numbers could not be directly com- pared. This demonstrates the basic difference between frequency and rate. Although both can be used when you are concerned with the number of behaviors, rate is used when recording the number of behaviors that occur during different amounts of time. Thus, the number of interrup- tions in the second example above could be directly compared by using rate (discussed next).

What Do You Record? Because the length of time is held constant, only the number of behaviors needs to be counted and recorded using frequency recording.

How Is it Calculated? Again, because the length of the observation peri- ods remains the same, only the number of behaviors are counted so no cal- culation is needed.

How Is Interobserver Agreement Determined? Interobserver agreement for frequency recording is determined by having two observers each record the number of occurrences of the target behavior and comparing these figures. It would be worthwhile for observers to compare recordings and discuss when and what they recorded. If the time of occurrence for each response is recorded, then an occurrence-by-occurrence comparison is possible. Other- wise, if significantly different frequencies are recorded, the observers may then attempt to determine why each recorded what she or he did, which was the correct recording, and how to avoid such disagreements in the future. This procedure should be followed when establishing interobserver

58 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

agreement using any recording procedure. Interobserver agreement for fre- quency is determined as follows:

Observer 1: 9 responses recorded Observer 2: 10 responses recorded

Smaller!Number!of!Responses !Larger!Number!of!Responses

$ 100% #

9!Responses 10!Responses

$ 100% # 90%

If disagreements about when responses occurred are known, the researcher may use the more conservative approach by comparing the observers’ recordings occurrence by occurrence. In this case, the researcher may use the following formula:

Number!of!Agreements Number!of!Disagreements ! Number!of!Agreements

! $ !100% #

What Is an Example? Franklin is a first-grade student who was adminis- tered the Dynamic Indicators of Basic Early Literacy Skills Next (Good & Kaminski, 2010) as a universal screener in his school’s Response to Inter- vention model. He was identified as being high risk for reading failure and received a supplemental phonics program. His teacher, Mr. Russell, moni- tored Franklin’s oral reading fluency every day by having him orally read a different grade-level passage each day. Mr. Russell simply counted the num- ber of words Franklin read correctly during the one minute. Based on these data, Mr. Russell could evaluate the effectiveness of the phonics program.

What Are Some Special Considerations? The above example used a discrete variable that was easily and appropriately recorded within the time period. It would be far more difficult to count, for instance, the number of words spoken by a client in a therapy session if the goal was to increase verbal output. As general guidelines, frequency should be used when the target behavior is relatively short in duration (i.e., is not exhibited for several min- utes per response), can be easily observed and counted (does not occur so often that many responses would be missed or counting them would prove very difficult), yet occurs often enough that significant changes are detect- able (occurs more than once a day or once a week for example) (Alberto & Troutman, 2009). This last guideline may be waived when behavior is so severe that any occurrence is inappropriate (e.g., serious physical injury to self or others). In these cases, the goal of the program should be to achieve a zero or extremely low rate of responding. Also, opportunities to

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 59

emit the target behavior usually should not be restricted or controlled (i.e., the individual must be able to emit the behavior at any point during the observation rather than having only a specific number of opportunities to produce the target behavior). If the opportunities are controlled, the time ele- ment should remain the same. For example, if a student’s number of correct math calculations on 30 problems was recorded, the time given to complete the problems (5 minutes) should remain constant. However, note that using such an approach might lead to a ceiling effect. In other words, if the student correctly answered all 30 problems, there would be no room to increase the fre- quency. In our example, the reading passages used with Franklin would have to be of sufficient length so that he could not read them all in one minute.

Rate When Should it Be Used? Rate is a number-related measure that is directly

linked to the length of observation periods. When observation sessions vary in length, rate must be used rather than frequency. Rate is sometimes used for higher frequency behaviors that would be difficult to accurately count for frequency over an extended period of time.

What Do You Record? Two measures need to be recorded: the number of behaviors and the length of the observation (amount of time observed).

How Is it Calculated? Rate is determined by dividing the number of responses observed by the length of the observation (length is generally expressed in terms of behaviors or responses per minute or second but could include hours or even longer periods).

Observer 1: 10 responses recorded Length of Observation: 20 minutes.

Rate # Number!of!Responses Length!of!Observation

! # ! 10!Responses

20!min ! # !:50!responses=min

How Is Interobserver Agreement Determined? Interobserver agreement for rate is determined just as it is for frequency. Both observers would record the responses during the same observation period. The researcher should determine if substantially abbreviated observation sessions yield better inter- observer agreement than lengthier ones. In such a case, it would indicate some change may be needed in the recording procedures or in the opera- tional definition of the target behavior.

What Is an Example? Bill is a 35-year-old business man who has a notice- able facial tic that he finds embarrassing and is starting to limit some of his interactions with his associates. He is asked by his therapist to count the

60 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

number of facial tics during his 9-5 workday. Bill, however, is not consistent about the time he records his tics. Sometimes he forgets, and other times he works more or fewer than 8 hours a day. Bill keeps track of these times and gives his therapist the number of tics and the time involved. The frequency of his tic behavior cannot be determined since the time was not held constant, although the rate of tic behavior can. These data might then be used as base- line to determine the effectiveness of a therapeutic plan that was initiated.

What Are Some Special Considerations? Two issues about Bill’s example should be mentioned. First, rate becomes an estimate of frequency rather than a measure of actual frequency in relation to time observed. This has the advantage of being less time consuming, but may also less accurately depict the actual frequency of the target behavior (e.g., 9 tics might have occurred during the 3-hour observation on Monday morning, followed by 25 tics in the next 3 hours when Bill’s behavior was not recorded). Thus, if the goal of rate recording is to estimate frequency, this limitation must be acknowledged. Another point is that in Bill’s situation, it would be important to determine when each observation occurred to help determine if the number of tics might at least be partially a function of when the tics were observed. Similarly, the therapist might investigate what Bill was doing during each recording period. Such information might also prove helpful in determining the most therapeutic approach. These suggestions might assist in data interpretation.

Interval Recording Occasionally the dependent variable or target behavior may occur with such a high degree of frequency that measurement is virtually impossible across an extended length of time (e.g., self-stimulatory behavior that occurs very rapidly), even using a rate measurement. In this situation, the researcher may elect to use interval recording. Similar to rate, interval recording involves estimating the frequency of the response and, accordingly, must be used in a fashion that will ensure the most accurate estimate. In interval recording, the occurrence or nonoccurrence of the behavior is recorded within very short periods of time (or intervals) across a longer observation session. The researcher then reports the number of intervals where occur- rences of the target behavior were recorded. Either partial or whole interval recording or momentary time sampling may be used.

Whole Interval Recording

When Should it Be Used? Whole interval recording is selected when the target behavior has a degree of duration (or occurs for a long enough period of time) that such a measure would be sensitive. If a behavior occurs, but only for very brief periods of time, then whole interval recording would

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 61

probably not be the best choice. Therefore, the target behavior should be one for which the researcher is primarily concerned with events (or epi- sodes) of at least the length of the interval.

What Do You Record? In whole interval recording, the observer divides the observation period into an equal number of intervals (e.g., a 10-minute observation session is divided into 60 ten-second intervals). The observer then records if the target behavior occurs during the entire interval (usually given a ! sign). If the behavior is not occurring at the beginning of the inter- val or discontinues during any point during the interval, a nonoccurrence of the behavior is recorded (usually a " sign).

How Is it Calculated? The number of intervals in which ! signs are recorded can be added and divided by the total number of intervals. This can be expressed as a percentage of intervals.

How Is Interobserver Agreement Determined? Interobserver agreement in whole interval recording is based on an interval-by-interval appraisal of agreement. That is, each observer’s recording of occurrence or nonoccur- rence should be compared for each interval.

Interval 1 2 3 4 5 6 7 8 9 10 Observer 1: ! ! ! " " " ! ! " ! Observer 2: ! ! ! " " " ! ! ! "

Number!of!Agreements Number!of!Agreements ! Number!of!Disagreements

$ !100% #

8 8 ! 2

$ 100% # 80%

Note that the observers each scored a total of six occurrences of the target behavior. A false impression would be given if the researcher were to report that each observer scored six occurrences and four nonoccurrences, although technically this would be true. This might give the impression that 100% interobserver agreement was obtained. In fact, agreement was achieved only in intervals 1, 2, 3, 4, 5, 6, 7, and 8. Disagreements occurred in intervals 9 and 10. Therefore, a lower interobserver agreement of 80% was actually obtained. For this reason, the interval-by-interval method of comparison should be made for whole interval, partial interval, and momen- tary time sampling procedures for recording.

What Is an Example? Alicia is a 4-year-old girl with severe intellectual disabilities. She exhibits the characteristic stereotypic behavior of rocking back and forth when sitting and even occasionally while standing. This is occurring more and more frequently and is interfering with many of her educational and developmental goals. Alicia’s rocking behavior generally

62 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

occurs for a number of seconds and sometimes for entire minutes if not interrupted by someone. Also, there are occasions when Alicia rocks only briefly and does not appear to interfere with other activities. The researcher may select whole interval recording as the preferable method; it wouldn’t record the unimportant events of extremely short duration but would record the rocking that does interfere with important activities.

What Are Some Special Considerations? The researcher should ensure that each observation period is of the same length of time. Also, if the period of time for which there is concern about the target behavior considerably exceeds the typical observation period (e.g., observation period is 10 min- utes in length but the period of concern when the target behavior is being emitted is 1 hour), then the researcher should vary the time of day in which the observation period occurs. This will provide a more accurate esti- mate of the frequency of the target behavior. Also, the possibility clearly exists that the target behavior may occur but may not be recorded using this system, hence, the estimate of frequency.

Whole interval recording may provide an estimate of duration as well (see “Duration Recording”). The number of consecutive intervals in which occurrences are recorded estimates how long the behavior was occurring (e.g., occurrences scored in five consecutive 10-second intervals would indi- cate the target behavior was emitted continuously for at least 50 seconds, but still does not exactly measure the actual duration). In whole interval recording, the researcher must ensure that the recording method does not artificially underestimate the actual frequency as a result of the length of the observation intervals. That is, if the interval lengths are longer than most emissions of the target behavior, many nonoccurrences may be scored when the target behavior is actually being emitted. For this reason, whole interval recording is more likely to underestimate actual frequency compared to the partial interval recording discussed next (Alberto & Troutman, 2009).

Partial Interval Recording

When Should it Be Used? The partial interval recording procedure is pref- erable for obtaining the higher estimate of frequency. In other words, partial interval recording should be selected when any instance of the behavior dur- ing an interval should be scored by the observer. In partial interval record- ing, unlike whole interval recording, you would be interested in those behaviors that might occur for shorter periods of time. Whole interval recording should be used when the researcher wishes to score primarily those responses that occur for the length of the interval or longer. Also, whole interval recording is typically easier in that observers are not likely to miss an occurrence that must occur for the entire interval length.

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 63

In partial interval recording, a distraction could cause an observer to not see a response of very short duration.

What Do You Record? Partial interval recording is very similar to whole interval with one notable exception (the sample data sheet for whole inter- val recording could be used for partial interval recording as well). Occur- rences are scored if the target behavior is exhibited at any point during the interval. Again, the occurrence (!) and nonoccurrence (") of the target behavior are recorded.

How Is it Calculated? The number of intervals in which ! signs are recorded can be added and divided by the total number of intervals. This can be expressed as a percentage of intervals.

How Is Interobserver Agreement Determined? Interobserver agreement for partial interval recording is calculated in the same manner as whole interval recording.

What Is an Example? Hank is a seventh-grade student who is earning Bs and Cs in most of his classes. The notable exception is history, in which the majority of information is delivered through a lecture format. His teacher, Mr. Brodsky, has noted that Hank often spends time staring out the win- dow and seemingly not paying attention. Using a partial interval recording procedure, Mr. Brodsky divides his 40-minute history into 15-second inter- vals. Mr. Brodsky’s teacher aide observed Hank to determine if he looked out the window at any time during each interval. Suppose Hank began a 15-second interval not looking out the window. After 8 seconds, he looks out for 4 seconds, and concludes the final 3 seconds by not engaging in the target behavior. An occurrence would be scored because the target behavior did occur at some point during the interval (whole interval recording would have resulted in a nonoccurrence because the target behavior did not occur for the duration of the interval). If Hank was observed for five minutes at the beginning, middle, and end of the period, then the percentage of inter- vals that Hank looked out the window could be determined for the total number of intervals as well as for the beginning, middle, and ending inter- vals. Mr. Brodsky could use this latter information to determine if there was a temporal pattern to Hank’s off task behavior.

What Are Some Special Considerations? Because occurrences are more likely to be scored using partial interval versus whole interval recording, research- ers may prefer this method as a better estimate of frequency. However, par- tial interval recording may also suggest that the duration of the behavior is greater than it actually is, particularly if the target behavior occurs only briefly during the interval. One must be aware of this distinction to avoid misinterpreting the data. Partial interval recording does not estimate duration.

64 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

In fact, an individual may be emitting the target behavior during a tiny per- centage of the actual observation period but still be scored in 100% of the interval observations. For example, the extremely brief off task events alluded to earlier would be scored in a partial interval system. The result could be many consecutive intervals scored when the target behavior occurred (e.g., 1 or 2 seconds of every 15-second interval), even though Hank could have actu- ally spent only a relatively small fraction of time looking out the window. In Hank’s case, Mr. Brodsky would use partial interval recording if he wished to get the best estimate of how often his off task behavior occurred. However, if Mr. Brodsky wanted to estimate the off task behavior that occurred for longer periods (e.g., 15 seconds or longer), he would select whole interval recording.

Momentary Time Sampling

When Should it Be Used? Momentary time sampling is a third method of using an interval-based system for estimating the frequency of a target behavior. Momentary time sampling is probably the easiest of the interval- based recording systems to implement but may yield the roughest estimate if the intervals are long and relatively infrequent. As with the other two meth- ods, the observation period is divided into intervals but should be short enough to ensure that responses of the target behavior are not frequently missed. Target behaviors that occur for some duration rather than quickly occurring ones are more appropriate for recording using momentary time sampling. In other words, the behavior should not occur infrequently because many times the behavior would not be observed.

What Do You Record? In this system, the observer looks at (or listens to) the student only at one time during each interval (usually the end) and records whether the behavior is occurring (scored !) or is not occurring (scored ") at that moment.

How Is it Calculated? The number of “moments” in which ! signs are recorded can be added and divided by the total number of moments observed. This can be expressed as a percentage.

How Is Interobserver Agreement Determined? Interobserver agreement for momentary time sampling is calculated in the same manner as whole inter- val and partial interval recording.

What Is an Example? Using the previous example provided for partial interval recording, Mr. Brodsky would again divide the 40-minute period into 15-second intervals. Instead of indicating whether the off-task behavior (looking out the window) occurred at any time during the interval, an occurrence would be recorded if Hank was off task at the moment of the observation within each 15-second interval. For example, Hank might be recorded at 15s., 30s., 45s., 60s., etc. If the off-task behavior occurred

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 65

before or after the momentary observation but not at that moment, a non- occurrence would be scored. Conversely, if Hank was not staring out the window for most of the elapsed interval but was doing so at the moment of the observation, an occurrence would be scored.

What Are Some Special Considerations? If the number of observations are sufficient and the target behavior is one that typically occurs for at least the length of the intervals and longer, this system can produce a reliable esti- mate of frequency of behavior (but not an estimate of duration). The observer should be eminently aware if the intervals are too long or the nature of the target behavior is such that the system is significantly underestimating the frequency. An advantage of momentary time sampling is that data may be collected on more than one individual at the same time. For example, one individual could be observed after 30s., 90s., 150s., and so on, while a second individual could be observed at 60s., 120s., 180s., and so on.

We would add a final note to our discussion of interval recording systems. They should not be confused with interval reinforcement systems discussed in Chapter 2. The systems could be used simultaneously, but this is not a requirement. In other words, observation interval systems bear no direct relationship to reinforcement systems that rely on intervals or duration of time.

Taylor Revisited #1 Taylor’s team must decide on a recording procedure. Frequency was one suggestion. However, the team decided that the hand-flapping occurred at a high rate and would be very difficult to count. Another suggestion was whole or partial interval recording. The team considered these as well. However, because the intervention was going to include response inter- ruption, whole interval would not be appropriate because an adult would likely stop her target behavior before an interval elapsed. At this point, the team is focusing on partial interval recording.

Evaluating Permanent Products The previously described recording procedures are used primarily when a teacher, clinician, or researcher is actively observing the target subject(s). There are many times, however, when permanent products are available that can be analyzed. For example, permanent product data may be gath- ered through the use of videotapes and audiotapes. There are other forms of permanent products that are used to record target behaviors. A typical example is the frequent use of paper-pencil (or typed or word processed) products in school environments to determine whether an instructional

66 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

intervention has yielded the desired outcomes. Permanent products offer the advantage of allowing later analysis and usually repeated review of that analysis (Alberto & Troutman, 2009). For example, videotapes and audio- tapes can be paused and replayed to ensure accurate recording and to allow for careful analysis of the relationship between the independent variable and the dependent variable.

A disadvantage of using permanent products may be that the individual is not observed actually performing the target behavior. For example, an individual may be working on solving word problems. The researcher may decide to record the percent of correct responses. If the individual’s perfor- mance is recorded based on paper-pencil products that are analyzed ex post facto, the researcher may not actually see what the individual is doing (e.g., what mistakes were made, what distractions occurred). Of course, a researcher or practitioner may also directly observe the individual complet- ing the permanent product. Using videotapes will avoid this limitation.

As just mentioned, permanent products can be analyzed using the same procedures as those for recording and analyzing active, real-time observations (depending on the nature of the permanent product). Additional recording procedures that are easier to use with permanent products are also available. One example is calculating percent correct and percent incorrect.

Percent Correct and Incorrect

When Should it Be Used? Percentages, which are frequently used with paper-pencil permanent products, may be one of the more widely used mea- sures in school settings. A common measure is percent correct, the metric most often reported to students as their scores on classroom spelling tests, math tests, etc. When combined with percent incorrect, the accuracy is determined. As a guideline, we suggest that the percent method of calculat- ing and reporting accuracy (particularly changes in accuracy) be used when there are a minimum of 10 and preferably 20 opportunities in a given obser- vation session (Wolery, Bailey, & Sugai, 1988). In fact, percentages most accurately reflect changes in behavior when there are 100 or more opportu- nities (Cooper, Heron, & Heward, 2007). Fewer than 10 opportunities tend to create too large a fluctuation in the percent calculated if incorrect responses have occurred. For example, a student who makes 4 out of 5 cor- rect responses gets 80% accuracy. He can only achieve 100% or reduce his accuracy by 20 percentage points per incorrect response when only 5 oppor- tunities are available. It might sound impressive to say that a student’s math performance was increased by 20% but misleading when the increase was actually only an improvement of one problem. We should also note that percent correct does not provide insight into the absolute number of events, the total number of opportunities, or the total length of time within which events were observed. Additionally, one may improve in terms of speed or

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 67

fluency, but 100% accuracy remains as good as the performance can be reported (Wolery et al., 1988). In other words, there is an artificial ceiling (100%) as well as an artificial floor (0%) when percentages are used.

What Do You Record? After deciding what constitutes a correct and incor- rect response, the observer then determines into which category each of the responses fall. This could simply be reporting how many responses were correct and/or incorrect. This method may be used when the number of opportunities to perform the target behavior is unchanging (e.g., the individ- ual always has 20 word problems to solve). In this situation, percent correct could be determined but is not necessary. If the number of opportunities for correct/incorrect varies from session to session, the percent has to be calcu- lated to involve a common metric for monitoring purposes.

How Is it Calculated? The overall percent correct of the individual’s responses could be calculated by

Number!of!Correct!Responses Number!of!Correct!Responses ! Number!of!Incorrect!Responses

$ 100%

How Is Interobserver Agreement Determined? Assume that after the criteria for correct and incorrect were established, two observers reported the fol- lowing (! # correct; " # incorrect)

Opportunity 1 2 3 4 5 6 7 8 9 10 Observer 1: " ! " ! ! ! ! ! ! ! Observer 2: ! ! ! ! ! ! ! ! " !

Interobserver agreement could be determined by using the following formula:

Interobserver # Number of Agreements

Number of Agreements ! Number of Disagreements $ 100%

Agreement

The observers agreed on 80% of the opportunities. Note: Opportunities could be defined as number of items, number of problems, etc.

What Is an Example? Ms. James is interested in using Key to Fractions (Rasmussen, 2011) with two of her students, Monica and Bart, who are having particular difficulty in the computation of fractions. Before she began the program, Ms. James gave both students a worksheet of 25 differ- ent fraction problems to complete each day for three days. She scored each worksheet to determine the number of correct and incorrect answers.

What Are Some Special Considerations? Technically, we are discussing a method of reporting behavior rather than a specific recording procedure. As noted previously, a researcher may also simply report how many responses were correct and/or incorrect if the number of opportunities to

68 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

perform the target behavior is unchanging. In the above example, since the number of problems was held constant, Ms. James only needed to count the number of correct answers. If Monica had 15 correct responses on Monday, 17 on Tuesday, and 18 on Wednesday, her performance could be directly compared across the 3 days. If the number of opportunities varied, percent correct must be used. Suppose, for example, that Monica and Bart were using the practice tests found in their textbooks, each of which had differ- ent numbers of problems (20, 30, and 25). If percent correct is not used, simply comparing the number correct across these days might be mislead- ing (e.g., 19 correct out of 20 might actually appear as a poorer perfor- mance than 25 out of 30 correct because 25 is more than 19 in total number of correct responses). However, 19 out of 20 is a better perfor- mance than 25 out of 30 (95% versus 83.3%). Also note that Ms. James could also include rate as well as accuracy. If she had the two students complete as many problems as possible in a 2-minute period, she could compute both. If Bart answered 12 correctly and 4 incorrectly in the 2 minutes, his rate would be 6 per minute and his accuracy would be 75%. The reason both are important is to again avoid misinterpretation. If Bart answered 16 correctly in the 2 minutes but also missed 8, his rate would increase to 8 per minute but his accuracy would decrease to 67%.

4 C H E C K I T O U T # 2 Assume a student is going to be reading passages of various lengths (100–250 words), and an teacher will be recording how many words are read correctly and his overall fluency (number correct and number incorrect) for a 1-minute reading. What might be the measures the teacher would use and report to best represent the student’s performance on the various passages?

Time-Based Methods for Recording and Reporting Behavior Although many experts may include interval recording systems as time- based, we chose to include them under event-based recording. We elected this option because we believe the overall purpose of interval systems is to estimate frequency (with the previously noted rough estimate of duration in whole interval recording). In the following recording systems, the concern is with the length of time the behavior occurs or with how long it takes for a target behavior to begin. Therefore, the measurement of time itself is of con- cern. In interval recording systems, time is used to determine when and how observations are scored, but the time itself is not recorded as a variable for measuring the target behavior.

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 69

Duration Recording When Should it Be Used? Generally, duration recording is used when the

length of time a behavior occurs (either longer or shorter) is the primary concern rather than how often a behavior occurs.

What Do You Record? Duration recording involves measuring the length of time from when a behavior begins to its termination. The observer may start with a timer as soon as the behavior begins and stop the timer when the behav- ior terminates. When the behavior begins again, the timer is once again started (but not reset to zero) and the process is repeated. At the conclusion of the observation period, the total time on the timer represents the total duration.

How Is it Calculated? Time!Engaged!in!Target!Behavior

Total!Time!Observed

How Is Interobserver Agreement Determined? Interobserver agreement for duration recording is calculated as follows (observers’ recordings equal duration in minutes):

Occurrences: 1 2 3 4 5 6 7 8 9 10 Observer 1: 7 6 3 4 6 4 8 2 7 3 Observer 2: 7 6 3 4 6 4 8 3 7 4

Shorter!Time!Recorded Longer!Time!Recorded

! $ !100%

In the above example, interobserver agreement could actually be determined in two ways. In the more conservative (and more acceptable method), the inter- observer agreement for each occurrence would be determined. This would be 100% for all occurrences except numbers 8 and 10. The eighth occurrence would yield a interobserver agreement of 66.66% (2 divided by 3 $ 100%) and the tenth a interobserver agreement of 75%. In this case, the researcher should report the agreement findings for each observation. She or he could also include the agreement for total duration as well. The first observer would have obtained a total duration of 50 minutes. The second observer would have recorded 52 minutes of duration. Interobserver agreement for total duration would have been 96%. This would suggest very good interobserver agreement, but would also be somewhat misleading because the actual agreement had been substantially less than this figure for 20% of the occurrences.

What Is an Example? Len is a 21-year-old individual who is living with his parents and who has some mental health issues. Len is often on the Internet and gaming whenever he can seize the opportunity. His team agrees that Len is entitled to use of the Internet and gaming, but his current aver-

70 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

age duration of 8 hours per day is excessive and interferes with his perform- ing daily routines such as bathing and cleaning his room, as well as looking for employment. His team agrees that his Internet and gaming behavior should be limited to 2 hours per day. Len himself, and his parents, are responsible for recording how long he is using the computer each day.

What Are Some Special Considerations? Frequency can also be determined within duration recording. For example, if the length of each period of Len’s computer use is recorded, then one can determine how long each day he is using the computer and how often. This method allows the determination of the number of episodes, the duration per episode (the amount of time each computing episode lasts), and/or the average duration per occurrence (total time spent engaged in the target behavior divided by the number of occur- rences). This method of determining duration per episode yields more data than the total duration (total time engaged in computer behavior) but is proba- bly more difficult to implement. Consider this example of the different interpre- tations of the two methods. You have your two roommates record total time spent studying during each day of the week. If total duration data alone are collected, you could determine how much time was spent studying, but not how much time was spent studying per occurrence of studying. Both students could each spend 4 hours studying although one could accomplish this dura- tion in one session whereas the other might take eight sessions. The choice of which method to choose depends on the reason for the data collection.

Latency Recording When Should it Be Used? Latency recording is used when the researcher is

interested in the amount of time that elapses between the signaling of a stim- ulus (e.g., the school bell ringing) and the emission of a target behavior (e.g., sitting in the desk). Latency recording is often used in compliance studies.

What Do You Record? Latency recording involves measuring the length of time from the delivery of an antecedent stimulus that should elicit the target behavior, to the actual beginning of the target behavior.

How Is it Calculated? There is no mathematical calculation; it is simply the amount of time that elapses between a stimulus and a response (target behavior).

How Is Interobserver Agreement Determined? Interobserver agreement is calculated using the same method as with duration recording.

Shorter!Time!Recorded Longer!Time!Recorded

! $ !100%

What Is an Example? Mr. Donovan is concerned because one of his stu- dents, Ben, is missing valuable instructional time because he wastes time

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 71

“preparing” to take his lecture notes. Ben will spend several minutes putting his books away, finding a sharp pencil, getting the right subject notebook, etc. Mr. Donahue wants to decrease the time it takes for Ben to begin his note taking (target behavior) after he says, “Let’s begin today’s science lesson” (the antecedent stimulus). Mr. Donovan records how long it takes for Ben to actually begin taking notes. The latency is the period of time from the ante- cedent delivery to the beginning of work. In this case, a reduction in the latency would be the desired outcome.

What Are Some Special Considerations? In the above example, the goal of any intervention would be to decrease the latency time. In other cases, an increase in latency might be desired. For example, if the overall objective is to reduce impulsive behavior, the target behavior might be to wait before acting in a particular situation. In this case, the desired outcome is to increase the length of time between the delivery of the antecedent and the onset of the target behavior (e.g., time between the delivery of an insult to action taken by the insulted individual). See Figure 3-12 in the Appendix for an example of a latency recording sheet. As with duration, latency may be recorded by occurrence or by total latency.

4 C H E C K I T O U T # 3 Assume a researcher is concerned about the length of time a man spends washing his hands, despite the medication he takes to control obsessive-compulsive beha- viors. The researcher wishes for the man to use a stopwatch and record how much time he spends washing his hands each episode. The man argues he is in a service industry and is around many people each day. Washing his hands is a necessary and healthful practice and that trying to reduce the actual number of hand- washings in not appropriate. The researcher concedes that for his health and the health of customers, the man does need to frequently wash his hands. How might these two resolve the recording issue?

Recording Procedures for Specific Purposes There are situations when the goal of an intervention program might require the recording of behavior in a specific or unique way. Many of these are tied to specific types of behavior analytic procedures or behavior intervention tech- niques. This involves using the previously described event-based and/or time-based recording procedures for specific purposes. Two examples are trials to criterion recording and cumulative recording.

72 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Trials to Criterion Recording Trials to criterion refers to measuring the number of responses needed by the individual to achieve some preset level of acceptable performance (crite- rion). Trials to criterion recording is often used in task analysis, in which a goal or skill area is broken down into sequential, clearly defined steps. Each of these steps are considered discrete so the number of trials to criterion could be counted for each step as well as for the goal itself. For example, a job coach may be teaching an adult with an intellectual disability to fold towels as part of his supported employment. The target behavior might be broken down into 10 sequential steps starting with “Lay the towel down with the long side closest to you” and proceeding through each step until successful completion of independent towel folding. The job coach may record how many trials it takes before the man successfully completes each step at the predetermined criterion level.

Generally, trials to criterion is used to obtain an idea of how quickly the individual acquires a skill. When fewer trials are needed, the assumption is that learning is occurring more rapidly or knowledge is being recalled more quickly (Cooper, Heron, & Heward, 2007). Cooper et al. also noted that this method may be useful in determining which type of instruction meets the cri- terion performance in the fewest number of trials. The researcher reports how many opportunities were given and what criterion was achieved by the indi- vidual. The criterion could be an ultimate criterion (e.g., a four-sentence par- agraph free of errors) or a criterion set for each session (e.g., utter a preset number of words with an initial s sound). Interobserver agreement for trials to criterion is determined in the same manner that frequency is determined except that the researcher divides the smaller number of trials recorded by the larger number of trials recorded rather than by the number of responses. If there is disagreement as to whether or not criterion is ever achieved (e.g., one observer believes it was achieved and the other records it was not achieved at all), then the researcher must compare the smaller recorded crite- rion (e.g., 80% correct or 8 of 10 steps completed correctly) by the larger recorded criterion (e.g., 100% correct or 10 of 10 steps completed correctly). When there is disagreement as to the criterion level the individual actually obtained, the researcher should immediately address this problem by ensuring all observers understand the standards for determining the criterion. Other- wise, subsequent data become suspect in their use for judging performance.

Cumulative Recording In cumulative recording, you are recording and reporting the total number of responses that occur across each observation session toward some goal or final level of achievement. This is usually thought of as a graphing option.

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 73

The number of occurrences observed in a given observation is graphed after being added to the total number of occurrences for all observations to that point (Alberto & Troutman, 2009). Suppose that a teacher, Ms. Freeman, was interested in determining the effectiveness of a group contingent token sys- tem. She provided tokens several times a day each day for the students’ compli- ance to several mutually determined class rules. Each day the number of tokens was added to the previous total to provide a cumulative record. The class knew that if a certain number of tokens were accumulated by Friday, there would be no homework over the weekend. There are other instances in which a researcher might be interested in recording the cumulative number of behaviors or responses. One might well use a bar graph or an x-y or line graph that shows the cumulative total as it changes with each observation to report findings.

Taylor Revisited #2 Taylor’s team, for the same reason they discarded the idea of using whole interval recording (because response interruption of her hand-flapping would decrease the duration immediately), also declined the use of dura- tion recording. It appeared that partial interval recording was the only pos- sibility. However, several team members noted that partial interval recording often requires a recorder to do nothing but observe and record. They questioned if this was feasible and if they limited the recording to only a few minutes each day or at various points throughout the day, they would not “capture” the extent of her behavior or the impact of the intervention. Finally, the team arrived at a unique solution. They decided that they would use frequency recording after all. Although counting the actual number of individual flaps of her hand would still be difficult, they decided they could record each episode of her hand flapping. Because it would be interrupted, each episode should be limited in length. This type of frequency recording could be used by each adult working with her through using a simple counter that could be passed along from one adult to another throughout the day. The total frequency of her hand flapping could then be recorded each day.

A Few Words about These Methods Up to this point, we have focused on the quantitative measures used in applied behavior analysis and single subject research. The researcher may also wish to consider other measures that may further enhance her or his ability to explain and understand the complex outcomes of a single subject research design. For example, anecdotal recording might be used to help provide important contextual information. To that end, we have included

74 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

a discussion of qualitative measures as well as other methods of analyzing single subject data in Chapter 13.

Methods for recording behavior are not limited to those we have dis- cussed, although these measures and variations are found frequently in the literature. The researcher should investigate studies in which similar measures to those being contemplated have been used. In the Application Chapters in this text, we provide examples of the use of the various single subject designs from the professional literature. Included in these examples is the identifica- tion of the recording procedures chosen. As with the designs themselves, “pure” examples in the literature are not always easily identified, as research- ers must adapt procedures to the individual, setting, and goals of the study.

Choosing the Appropriate Recording Procedure Researchers must be concerned with how the recording procedures dovetail with the intervention used and the nature of the target behavior. As an example, we will describe a series of scenarios using a therapist, Mr. Monday, and his client, Cindy, who has the habit of biting her nails. A particular treatment for Cindy’s nail biting has been identified and agreed to by all concerned. Mr. Monday must now decide how the treatment and target behavior will impact the recording procedures. The following options are among those possible.

1. If Mr. Monday was observing Cindy throughout a day or specific period of time, he would likely use frequency recording. However, if the observa- tion period varied (and the number of opportunities was uncontrolled), then he could elect to use rate. Mr. Monday would design an intervention intended to reduce the number or rate of nail biting episodes.

2. If Cindy’s behavior occurred at a relatively high frequency or was diffi- cult to measure in terms of a specific number of responses, then Mr. Monday might elect to use interval recording. Whether he used whole or partial interval, or momentary time sampling, his overall goal would be to reduce the number of intervals in which nail biting occurred.

3. If Mr. Monday is interested in obtaining a record of and analyzing Cindy’s behavior, he may then use a videotape or audiotape as a perma- nent product. Percent correct would probably not be used in this case.

4. Mr. Monday may also use time-based procedures for recording target behaviors. Duration per occurrence recording could be used to measure how often nail biting occurred, what the total time spent being engaged in the behavior was, and what the average duration of each episode was. In this case, the overall aim of intervention would be to reduce the length of time spent engaged in nail biting (as measured per episode or by total duration).

5. Let’s suppose Mr. Monday was interested in increasing the time spent from when Cindy entered a stressful situation to when she began her

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 75

nail biting. In this case, he could use latency recording with the intention of increasing the latency from the onset of the stressful situation until she exhibited the behavior.

6. If Mr. Monday was aiming toward assisting Cindy in achieving a zero rate of nail biting for 5 consecutive stressful situations, he might elect to record trials to criterion. The researcher would measure how many opportunities were required before the performance criterion was achieved.

These examples are intended to show the reader how different procedures might be used with the same target behavior and different intervention goals. Knowledge of these methods for recording behavior is important, but the researcher must be able to adapt and adjust these methods, be able to use them in combination with other possible measures (e.g., qualitative measures), and ensure they accurately reflect what is actually happening to the individual. This last point is perhaps the most important. The ultimate aim in data collec- tion is gathering of information that can later be analyzed to tell the story of what happened to the individual(s) involved. Quantitative methods should lend objectivity to that storytelling. That is the strength of using these quantita- tive measures. Objectivity is, however, not useful if the researcher fails to mea- sure what is really important about the changes the individual is undergoing. Interobserver agreement is essential, but the researcher must not select methods primarily because they are easier to conduct or interobserver agreement is more certain, if other methods may be more meaningful for measuring actual out- comes. For example, a momentary time sampling method may be very reliable, but may not actually reveal the true nature of changes in the target behavior.

Summary In our next chapter, we address many of the issues that the researcher must consider in designing a study and ensuring ethical behavior and treatment. The methods for recording the target behavior are an integral part of that process, but cannot be separated from decisions concerning the treatments involved as well as many other variables. In Chapter 4, we discuss issues related to single subject research.

Key Concepts/Terms Event-based Methods for Recording and Reporting Behavior Interobserver agreement—Two or more observers record the target behavior

independently and simultaneously; later, results are compared to determine if the target behavior is being measured reliably; interobserver agreement is calculated differently depending on the method used for recording behavior.

76 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Frequency—Used when behavior can be easily counted and the length of observation is constant; opportunities to perform the behavior should not be controlled.

Rate—Used as frequency recording except that length of observations vary; the number of responses per unit of time is calculated and reported.

Interval recording—Method of estimating the frequency of the target behavior using a period that has been divided into equal intervals of time for individual observations; not to be confused with interval sche- dules of reinforcement.

Whole interval recording—Provides a smaller estimate of frequency; behav- ior is recorded only if it occurs throughout the interval.

Partial interval recording—Provides the greater estimate of frequency; behavior is recorded if emitted at any point during the interval.

Momentary time sampling—Provides the roughest estimate of frequency; behavior is recorded if occurring at a specific moment at the conclusion of an interval; may be used to record behavior of more than one individ- ual simultaneously.

Permanent products—These are permanent data forms that may be reviewed repeatedly (e.g., paper-pencil products, videotapes, or audiotapes).

Percent correct/incorrect—Used when the number of opportunities to per- form the target behavior may vary from observation to observation.

Time-based Methods of Recording and Reporting Behavior Duration recording—Used when how long the behavior is emitted is the

primary concern; the researcher may record and report duration per occurrence, average duration, or total duration.

Latency recording—How long it takes for the behavior to occur following the antecedent stimulus; the researcher may record and report latency per occurrence, average latency, or total latency.

Recording Procedures for Specific Purposes Trials to criterion—Used to measure events (occurrence of target behavior)

when the researcher wishes to determine how many occurrences are required to achieve a criterion level of performance.

Cumulative recording—Used when the aggregate number of emissions is of primary importance.

4 Possible Answers to Check It Out

¶ The interobserver agreement is 80% which is adequate. Note therecorders disagreed on intervals 3 and 8. Remember not to equate the nonoccurrence of the target behavior (a minus in this case) as a

CHAPTER 3 METHODS FOR RECORDING BEHAVIORS 77

disagreement when both recorders indicate nonoccurrence of the target behavior.

· The teacher has several options. Because the readings are always oneminute, she could report the number of words correctly and the num- ber of words incorrectly. In themselves, these would be minimally adequate measures. However, she could also include the percent of words read correctly and incorrectly within the passage as the student might read 50 words in one passage and 60 in another. Finally, because the student very well might read more words in one passage versus another, the rate of words read correctly and incorrectly per minute would also be helpful. In fact, these are common measures used in short cycle assessments in the early grades of school.

¸ The researcher and the man might agree to simply use total durationrecording. Because washing hands is a necessary component of his job, the man must have some discretion as to when it is appropriate to do so. However, the length of time spent washing his hands is also an important factor. The two agree that total duration will likely yield sufficient infor- mation as to whether he is washing his hands so much that it is actually interfering with his daily functioning. The man agrees to carry a stopwatch, stop it and start it each time he washes his hands, and record the total duration at the end of his workday.

References Alberto, P. A., & Troutman, A. C. (2009). Applied behavior analysis for teachers

(8th ed.). Upper Saddle River, NJ: Pearson Education. Ayres, K., & Gast, D. (2010). Dependent measures and measurement procedures

(pp. 129–165). In D. Gast (ed.) Single subject research methodology in behav- ioral sciences, NY: Routledge.

Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis (2nd ed.). Upper Saddle River, NJ: Pearson Education.

Good, R., & Kaminski, R. (2010). Dynamic Indicators of Early Literacy Skills Next. Eugene, OR: Measurement Group.

Kennedy, C. (2005). Single case designs for educational research. Boston: Allyn & Bacon.

Rasmussen, S. (2011). Key to Fractions. Emoryville, CA: Key Curriculum, Inc. Wolery, M., Bailey, D. B., Jr., & Sugai, G. M. (1988). Effective teaching principles

and procedures of applied behavior analysis with exceptional students. Needham, MA: Allyn & Bacon.

Zirpoli, T. J., & Melloy, K. J. (1993). Behavior management applications for teachers and parents. New York: Merrill.

78 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Appendix

Sample Recording Sheets

FIGURE A3-1 Anecdotal recording sheet

Name of Person(s) being observed: Date:

Name of Observer: Location:

Time Antecedents Behaviors Consequences

80 APPENDIX SAMPLE RECORDING SHEETS

FIGURE A3-2 Recording sheet for

accuracy

Name of person being observed: Location:

Name of Observer:

Time/Date Number of Opportunities Number Correct % Correct

APPENDIX SAMPLE RECORDING SHEETS 81

FIGURE A3-3 Recording sheet for correct

responses

Name of person being observed: Location:

Name of observer:

Time/Date Number of Correct Responses (opportunities are constant)

82 APPENDIX SAMPLE RECORDING SHEETS

FIGURE A3-4 Recording sheet for

frequency

Name of person being observed: Location:

Name of observer:

Date/Time Number of Events/Episodes (length of observation remains constant)

APPENDIX SAMPLE RECORDING SHEETS 83

FIGURE A3-5 Recording sheet for rate

Name of person being observed: Location:

Name of observer:

Time/Date Length of Observation

Number of Episodes/ Events

Rate of Responding (Events/Time)

84 APPENDIX SAMPLE RECORDING SHEETS

FIGURE A3-6 Recording sheet for trials

to criterion

Name of person being observed: Location:

Name of observer:

Time/Date Criterion Level Number of Trials Required of Meet Criterion

APPENDIX SAMPLE RECORDING SHEETS 85

FIGURE A3-7 Recording sheet for cumulative responses

Name of person being observed: Location:

Name of observer:

Time/Date Number of Responses Cumulative Responses

(Total responses + this observation)

86 APPENDIX SAMPLE RECORDING SHEETS

FIGURE A3-9 Recording sheet for

momentary time sampling

Name of person being observed: Location:

Name of observer: Time/Date:

10-minute Intervals

FIGURE A3-8 Recording sheet for

interval data

Name of person being observed: Location:

Name of observer: Time/Date:

10-Second Intervals

+ = Occurrence of the target behavior at end of interval 0 = Nonoccurrence of the target behavior at end of interval

H O U R S

M IN

U T E S

+ = Occurrence of the target behavior 0 = Nonoccurrence of the target behavior

APPENDIX SAMPLE RECORDING SHEETS 87

FIGURE A3-10 Recording sheet for level

of assistance

Name of person being observed: Location:

Name of observer: Levels of Assistance: I = independent V = verbal prompt G = gesture M = modeling PP = partial physical prompt FP = full physical prompt

Steps in Task Analysis 1 2 3 4 5 6 7 Comments

Time/Date Level of Assistance I M FP

88 APPENDIX SAMPLE RECORDING SHEETS

FIGURE A3-11 Recording sheet

for duration

Name of person being observed: Location:

Name of observer:

Date Time Behavior Started Time Behavior Ended Duration

APPENDIX SAMPLE RECORDING SHEETS 89

FIGURE A3-12 Recording sheet for latency

Name of person being observed: Location:

Name of observer:

Date Time of Delivery of

Antecedent Time Behavior Ended Latency

90 APPENDIX SAMPLE RECORDING SHEETS

CHAPTER

4 Issues in Single Subject Research

IMPORTANT CONCEPTS TO KNOW THE CONCEPTS OF PREDICTION, VERIFICATION, AND REPLICATION

Ramona’s Vignette

Prediction Verification Replication

Ramona Revisited #1

RELIABILITY AND VALIDITY Reliability Validity

Ramona Revisited #2

ETHICS

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

The Concepts of Prediction, Verification, and Replication Single subject research is an applied approach that is concerned with demon- strating functional relationships between the independent and dependent variables. In fact, as you will see when you study each of the designs and their variations, this approach may demonstrate such relationships better than other approaches (e.g., group study approaches). That is, the effective- ness of single subject designs relies on demonstrating that changes in the dependent variable are directly attributable to the presence or absence of changes in the independent variable. Continuous measurement of the depen- dent variable and subsequent changes in accordance with the researcher’s manipulation of the independent variable allows this direct demonstration within the chosen design. Prediction, verification, and replication of effects are present when this functional relationship exists. Also, the researcher must demonstrate reliability and validity as well as intervention fidelity. Ethics and humane treatment are also important aims in single subject research. When all these effects are achieved, the researcher is able to make a strong case for the functional relationship between independent and dependent variables. We will discuss each of these issues as it relates to single subject research.

The concepts of prediction, verification, and replication relate to the issues of reliability and validity as specifically applied to single subject research. When these concepts can be demonstrated within a single subject research design, the functional relationship between the independent and dependent variables is evident. The functional relationship is one of the quality indica- tors identified for demonstrating evidence-based practices that are crucial in single subject research (Hudson, Lewis, Stichter, & Johnson, 2010). Hence, the reliability and validity of the study are verified assuming the extraneous variables and systematic bias discussed earlier in Chapter 1 cannot be reason- ably perceived to have accounted for the changes that have occurred. Each concept plays a unique part in this verification process, yet the concepts are interdependent. That is, the demonstration of each concept is critical but the presence of all is equally important. Each concept is related to changes in the dependent variable that are directly attributable to the independent variable. In other words, the changes that are present in the data path can be explained by systematic manipulations included in the study and eliminate common threats to internal validity such as a rival hypothesis as to why the dependent variable might change (Horner, Carr, Halle, Mcgee, Odom, & Wolery, 2005). The following discussions include a summary of Tawney and Gast (1984) with some of our own modifications and additions.

Ramona’s Vignette Ramona is an 18-year-old university student. She has been visiting her counseling center because she has experienced high test anxiety and she

92 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

believes this is adversely affecting her grade point average. She has decided to enroll in a course through the counseling center that addresses study skills, test-taking skills, lifestyle management strategies, strategies for reducing stress and anxiety prior to a test, strategies for handling the results of tests, and strategies for talking with professors throughout a course about testing and any other issues. Ramona and three other students enrolled in the course dur- ing an intersession in hopes of improving their test performances in the new semester. The course also includes biweekly group sessions in the new semes- ter for discussions among the students and the counselor results of testing in classes, how the course strategies helped or did not, and how to maintain or improve testing results. The counselor has approval through the university review board overseeing studies involving human subjects and has obtained informed consent from Ramona and her three peers.

The counselor, who is also conducting a research study to establish the worthiness of the counseling course, teaches Ramona and the three other students to chart their test results from the previous semester. As the new semester begins, the counselor also asks them to chart their testing results for the new semester. At the end of the semester, Ramona and her three peers each compare individually with the counselor, their prior semester’s test results to the current semester. Ramona finds that in comparing the two semesters (using letter grades for each test), she has improved on aver- age from a grade of B- to a grade of A- for each course in which she is enrolled. She is thrilled. Two of the other peers experienced similar improve- ments of a full letter grade on tests on average. The third peer experienced no change in testing performance and dropped out of the university.

Prediction Prediction refers to the idea that if there is no effect attributable to the inde- pendent variable, the dependent variable’s data path will remain unchanged. For example, a researcher has collected baseline data on an individual’s target behavior. After stability has been achieved during the baseline phase, the intervention phase would be introduced. If the intervention had no effect on the dependent variable, one could logically assume that the data path from baseline to intervention phases would depict no appreciable change (see Figure 4-1). Therefore, one could predict the data path will remain unchanged despite a phase change (Horner et al., 2005). Should the data path change and that change be maintained, one could then reject the prior hypothesis that phase change had no effect on the dependent variable (also assuming appropriate control of extraneous variables and systematic bias). When a data path changes predictably in conjunction with a phase change, it is possible that there is verification the intervention has an effect on the dependent variable or target behavior.

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 93

Verification Verification is the confirmation that the dependent variable is changing in a predictable direction as the independent variable is systematically applied (see Figure 4-1). For example, a baseline phase is completed and an inter- vention is introduced. As we have discussed, if the independent variable has no effect, we could predict there will be no change in the data path with that phase change from baseline to intervention. We may predict conversely that the data are likely to reveal a change in the desired direction in the data path as a result of the implementation of the independent variable. When this happens, we have verified our hypothesis that the dependent variable will change predictably with the introduction of the independent variable. Replica- tion is needed also to complete our demonstration of a functional relationship.

Replication Replication refers to the repeating of the observed predictions and verifica- tions within the same study. This concept is essentially what separates single subject research designs from the typical teaching situation. For example, a practitioner gathers baseline data on math accuracy and implements a rein- forcement program, and the desired changes are observed. Please note that one could state that both a prediction (baseline will remain unchanged if the reinforcement program is not started and the data will show improved

Data path if prediction, veri!cation, and replication are not achieved. Data path if prediction, veri!cation, and replication are achieved.

Predicted path if no replication is achieved

D ep

en d

en t V

ar ia

b le

Observations

A1 B1 Prediction B2 Veri!cation B2 Replication

Predicted path if intervention is e"ective (original prediction of e"ectiveness is veri!ed)

Predicted path if intervention does not account for changes

Replication of e"ects achieved in A1-B1

Veri!ed path if intervention does account for changes (replication of A1)

Predicted path if intervention is ine"ective

FIGURE 4-1 Graph Illustrating

Prediction, Verification, and Replication

19 99

94 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

accuracy when positive reinforcement is started) and a verification (the accu- racy does in fact improve) are present. Replication of the effect of the inde- pendent variable on the dependent variable, however, is not present. For replication to occur, the researcher might return to baseline conditions (and one could predict the data will remain unchanged if the positive rein- forcement is not actually accounting for the improved accuracy, and we could predict that math accuracy will decrease toward baseline levels if, in fact, positive reinforcement is accounting for the change). Should the indivi- dual’s accuracy decrease with this phase change, we have verified the predic- tion that math accuracy is increasing as result of the introduction of a positive reinforcement program (and decreasing when the positive reinforce- ment program is not present). We may also allow the data path to return to the same or a similar level as during our original baseline. We then may reintroduce the intervention. This allows the researcher to replicate the orig- inal prediction and verification. If replication does occur (accuracy once again increases as a result of the reintroduction of the independent variable), the researcher has created a stronger case for the demonstration of a func- tional relationship between the independent and dependent variables. This example would be an A-B-A-B, or withdrawal, design. In each of the chap- ters related to how to set up a particular design, we will discuss how predic- tion, verification, and replication are used to strengthen the case for influence of the independent variable on the dependent variable.

4 C H E C K I T O U T # 1 Assume a sixth-grade student is learning to write essays of five paragraphs with correct punctuation including the use of periods, exclamation points, question marks, quotation marks, and commas. Assume the researcher working with the stu- dent is implementing a reinforcement program as the student achieves at least 80% correct or better to earn reinforcers of free time and a homework pass. The researcher has established a baseline for three assessments that includes 50%, 40%, and 40% correct punctuation. Can you describe how prediction, verification, and replication might occur in such a study?

Replication has at least one other important connotation in single subject research. Because group experimental designs use a larger number of indivi- duals as subjects who frequently are randomly assigned to treatment groups, there is an assumption that the influences of intra-individual variables are evenly distributed across groups (error variance in the parlance of paramet- ric statistics). In other words, whatever peculiar or unique influences occur for each individual participant in the study, these do not influence the overall outcome of the study because those influences are balanced by

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 95

the size of groups and the random assignment of individual participants to those groups. We have trouble making these assumptions when there is only an individual or a few subjects involved in a study. Perhaps positive rein- forcement was effective only for this individual’s math accuracy, or because it was in a particular setting, or because of the time of year, or the matura- tion of the individual, or a myriad of other conceivable influences (confounding variables). The more replications included within a study, the less likely changes in the dependent variable are attributable to extraneous or confounding variables. For example, Horner et al. (2005) suggest a design include at least 3 replications of this effect on the dependent variable as a quality indicator of the study. Also, the case for the power (or robust- ness) of an independent variable to influence a dependent variable is strengthened when other researchers seek to replicate effects with other indi- viduals, with similar behaviors, in different settings, and so on. The more different studies replicate one another or achieve a similar and predictable effect, the greater confidence we may place in the intervention being used. The behavior change strategies discussed in Chapter 2, for example, have effects that have been replicated many times over and their power verified.

Prediction, verification, and replication are key elements to demonstrat- ing a functional relationship between the independent and dependent variables. These concepts are linked to the concepts of reliability and valid- ity in single subject research.

Ramona Revisited #1 In Ramona’s case, the independent variable is the counseling course and biweekly meetings. The dependent variable is her test performance by letter grade. One could predict that her test performance would not improve as a result of the independent variable if, in fact, it has no or little value as an inter- vention. When, in fact, her performance does improve, there is verification that the counseling course may have impacted the dependent variable. Repli- cation in this vignette is achieved by including other subjects in the study. Interestingly, because one peer did not perform better on the dependent vari- able (verification was not achieved), there is left some concern as to whether the replication across the participating students is sufficient to suggest the counseling course will result in improved test performance among university students experiencing test anxiety adversely affecting their performance.

Reliability and Validity Reliability and validity may have some meaning to you already. In single subject research, we have three major areas of concern related to these concepts. First, interobserver reliability (sometimes referred to as interrater

96 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

reliability or interobserver agreement) is important to establish confidence in the measurement of the dependent variable (Horner et al., 2005). Second, internal validity is important to establish the believability of the functional relationship and, therefore, establish confidence in the results and conclusions drawn. Intervention fidelity should be established in controlling internal validity (Gersten, Fuchs, Compton, Coyne, Greenwood, & Innocenti, 2005). We will also discuss external validity, which is related to the confidence others may have that the same independent variable will yield similar results in similar studies (e.g., with different individuals, with different dependent variables, in different settings). External validity is also related to the concept of replication.

Reliability In Chapter 3, we discussed the specifics for determining the interobserver reliability for the more standard methods of recording target behaviors used in applied behavior analysis (e.g., frequency, duration, latency). Inter- observer reliability is not our only measurement concern, but the researcher should provide evidence that her or his measures of the dependent variable are accurate. As we outlined in Chapter 3, the researcher must identify and define the target behavior in such a way that there is confidence that at least two people can observe the individual participant and agree whether or not the behavior has occurred, or to what extent, or for how long, and so on. This process is essential to establishing reliability and also internal validity. Practitioners may frequently use observation to determine whether progress toward a goal has occurred, but they may also rely on their own personal judgment regarding the individual’s performance. This may result in subjec- tive measures or observer drift. Establishing interobserver reliability helps to ensure that the process has been fair, ethical, and rigorous.

As the term would imply, interobserver reliability relies on the use of more than one observer. Typically, the researcher identifies at least two indi- viduals (which may include herself or himself) who will be involved in measuring the dependent variable, or who are at least available to serve as observers for reliability purposes. More than two is advisable so that the loss of one observer will not disrupt the study. Once the observers are iden- tified, a training program is established in which the observers gain practice in observing the individual participant and scoring the dependent variable. In our experience, this training process involves a number of steps. The researcher or practitioner must identify and define the target behavior. An example might be sitting in one’s chair. After observing the individual par- ticipant, the researcher may wish to operationally define sitting in one’s chair so that minimal judgment is required to decide if in fact the individual is sitting in her chair (e.g., sitting might be defined as buttocks in contact

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 97

with seat, the back in contact with the chair back, at least one foot on the floor, and the trunk and head erect and not in contact with any part of the chair or other furniture). Operational definitions were more fully described in Chapter 3, but the reader should understand here that the target behavior should be so well defined that it is easy to determine if it has occurred and to what extent (e.g., when it began and ended, how frequently it occurred, or how long it took for the target behavior to begin). Defining the target behavior may require having other observers also participate so that poten- tial ambiguities and unusual occurrences or circumstances may be observed or anticipated. Sometimes, videotapes of the individual participant may assist in defining the target behavior, practicing observations and measuring the dependent variable without intruding in the experimental environment. Generally, it is important to also define what are non-occurrences (or non- examples) of the target behavior. That is, the observers should practice rec- ognizing when the target behavior did not occur (e.g., the individual meets all aspects of the definition of sitting in seat but both feet are propped up on a chair in front of her and hence she is not sitting based on the definition). Observers should practice until they are very comfortable and efficient with recognizing the target behavior. This may take some time. Next, the researcher must select a system for measuring the target behavior (depen- dent variable). For example, if an interval system is used (see Chapter 3), more practice may be required because of the more complicated procedures. Following practice sessions, the observers should compare results for each and every occurrence or measurement of the target behavior (dependent variable). They should reach an understanding of both why and why not the target behavior had occurred. When disagreements occur, it is particularly important to understand and reach agreement as to what the correct mea- surement was and why. This comparison process can take some time, but it also serves to more precisely define the target behavior, to refine the mea- surement procedure, and to provide confidence that a reliable system is in place. As the practice observations continue, interobserver reliability should be calculated among and between observers. The interobserver agreement coefficients should be calculated in the more rigorous manner as discussed in Chapter 3 to ensure adequate reliability. In our experience, it is also criti- cal that the researcher maintains an open mind and avoids being judgmental with observers. Although the researcher ultimately must make decisions as to what is being measured, how, and by whom, the input from the observers is invaluable. The researcher who adopts an attitude that the observer is always wrong whenever a disagreement in scoring occurs, is a researcher likely in for trouble. The researcher must remember that she or he needs good observers who are committed to performing their task with integrity and confidence. Those attributes are acquired through reinforcement, practice, and honest and open discussions when disagreements do occur.

98 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

We cannot overemphasize this point, especially to the student-researcher who may be relying on peers to help with a study.

It should be noted that some experts (e.g., Barlow & Hersen, 1984) argue that observers should be naive to the intervention and purpose of the study, should not be involved in calculating interobserver agreement, and should be unaware of when reliability checks are scheduled. In our experi- ence, the researcher herself or himself is frequently involved in data collec- tion (which does have its own limitations in terms of introducing a possible bias), so such precautions often are not possible. However, if the researcher is not actually involved in the data collection, then such precautions may be advisable to strengthen internal validity. Again, interobserver reliability of 80% or better is considered a quality indicator of a study (Horner et al., 2005) with 90% or better being preferred.

Once interobserver agreement is ensured, the researcher and/or observers may begin collecting data. However, they must still be aware of reactivity and observer drift, which may serve to confound their observational procedures.

Reactivity refers to the individual being observed altering his or her behavior (i.e., target behavior) as a response to being observed. Using our previous example, the individual may immediately improve (or worsen) her performance on sitting in her seat if she is aware of or suspects that she is being observed on that target behavior. Reactivity generally diminishes as the number of observations increases. Actual practice observations in the experimental setting can, therefore, reduce the possibility of reactivity when the experimental data collection begins, although the presence of observers regularly before the introduction of the independent variable may have an unknown and confounding result as well. Reactivity may also be overcome by extending baseline measurements until there is stability but also a reasonable assumption that the performance on the dependent vari- able truly represents previous levels unaffected by observer presence. Estab- lishing a pattern of responding under baseline conditions is a quality indicator of a single subject study (Horner et al., 2005). The use of video- taping, audiotaping, or two-way mirrors may also help to avoid reactivity, although the presence of cameras, recorders, or large mirrors can also influ- ence the individual. Once reactivity may be ruled out as a major contributor to performance on the dependent variable, the researcher must still continue to guard against observer drift.

Observer drift refers to a change in interpretation of when the target behavior occurs or not (or at what level, or how intensely, or for how long, etc.) from the original operational definitions. Observer drift occurs generally when a number of observations have been made, particularly if the communication and comparison of scores among raters has not been reg- ularly reviewed and discussed. In our example of sitting in seat, one observer

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 99

may begin to score occurrences even if the individual has begun to slump over (i.e., not erect in trunk and head) and props her head with her elbow on the chair’s armrest. After all, the observer may think, the individual is paying attention to what is going on and maybe she is a little tired, and her buttocks and back are in proper contact with the seat and both feet are on the floor. So, the observer scores the individual as sitting in her seat. Drift has occurred. If another observer was simultaneously scoring the individual’s performance, he may very well disagree on what the actual performance was and therefore threaten interobserver reliability. In commonsense terms, observer drift occurs when one or more observers begin(s) to exert a personal definition or at least personal modification of the definition of the target behavior. As we have already suggested, the best guard against drift is regular com- munication and perhaps even repeating training exercises throughout the study to ensure all observers remain in agreement about the definition and occurrence of the target behavior. If, during a study, interobserver reli- ability should fall below the 80% level in any observation session, the researcher should be alert to the possibility that drift may be occurring, and take appropriate measures to avoid a serious lapse in reliability. If, as mentioned earlier, the observers are naive to the reliability checks and coefficients, then such measures might include additional training, ensur- ing observers have memorized operational definitions and observation pro- cedures, and finally interviewing observers to ascertain any biases or other potential confounding variables that may influence their scoring (Barlow & Hersen, 1984). Precisely defining the dependent variable and precisely defining how the dependent variable is to be measured are quality indica- tors of a single subject study (Horner et al., 2005).

Validity, like reliability, is of critical importance to single subject researchers. As with reliability, single subject research has some unique con- siderations in terms of validity. We will discuss these considerations as they regard internal and external validity.

Validity As you may already be aware, there are several types of validity associated with measurement of any particular behavior (e.g., face validity, content validity, and predictive validity). These are frequently used when providing evidence that a test (e.g., an intelligence test or test of vocabulary knowl- edge) measures what it purports to measure. The reader should become familiar with the various types of validity and the concepts determining the validity of measurement procedures. These procedures are used less fre- quently in single subject research than in group studies, but understanding them is important. Internal and external validity are concepts that apply to both group and single subject research and are of primary concern to us.

100 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Internal Validity Internal validity refers to the degree to which the researcher has adequately controlled the independent, dependent, and extraneous (or confounding) variables so that changes in the dependent variable are directly attributable to the presence or absence or intensity of the independent variable. In sim- pler terms, the researcher convincingly demonstrates a functional relation- ship in that the treatment (independent variable) led to the individual changing his or her behavior (dependent variable). For example, a researcher is attempting to help an individual improve her eating behavior (consume no more than 1,800 calories per day for 30 consecutive days). The researcher implements a treatment (weighing portions, eating from selected food groups, self-calculating calories consumed) that is intended to accomplish this objective. After several weeks of intervention within an appropriate design, the researcher’s measurement of the individual’s eating behavior does in fact verify that the target behavior is substantially improved. Examining the individual’s life across those several weeks, there is no reason to suspect that any other factor or variable accounted for the improved eating behavior. As an example of diminished internal validity, let us examine the same example. In this case, the same substantial change in the target behavior is noted by the researcher. However, the individual reports she began chatting online with others who had to restrict their calo- ric intake and was truly inspired by their examples and by ideas they pro- vided as to how to achieve her caloric-intake goal. She also reports that her mother also decided to follow the same eating regimen and they were encouraging each other. The researcher must now acknowledge that other variables may have had an effect on the dependent variable.

If these online and maternal support activities had been going on for weeks, months, or even years before the outset of the study, the influence of the activities may be explained as potentially less or nonexistent. If, however, the inspirational activities commenced during the study, the researcher must be honest and acknowledge that the treatment alone may not account for changes in the target behavior (eating behavior). The control of extraneous or confounding variables is paramount to establishing a functional relation- ship between the independent and dependent variables (Horner et al., 2005).

Extraneous Variables Extraneous variables refer to virtually anything that may affect the demonstration of the functional relationship between the independent and dependent variables. In other words, the presence of these variables may elicit questions as to whether it was the influence of the independent variable alone that led to changes in the dependent variable (i.e., confound the interpretation and validity of your results and conclusions). Cooper, Heron, and Heward (2007) noted that the strength of an experimental design is evident to the

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 101

extent that it reduces or eliminates the influence of confounding variables while still allowing the researcher to investigate the research questions. Some of these extraneous variables are commonly recognized among researchers. One example noted by Cooper et al. (2007) is maturation. Interference from the use of multiple interventions and the extent to which the intervention is applied consistently and as planned (treatment integrity or procedural fidelity) (Cooper et al., 2007) are also potential confounding variables. Measurement effects that were addressed in Chapter 3 may also occur. It is worth noting that the presence of extraneous variables may or may not influence the desired treatment outcomes. In other words, the goal of the study may be achieved (e.g., better accuracy on solving math problems or more fluent speech), but demonstrating that it was the influence of the systematic application of the independent variable (i.e., a functional relationship) may not be achieved. Several typical extraneous or confounding variables are discussed in the following passages.

History is used to refer essentially to the passage of time and both fore- seen and unforeseen events that arise. For example, a researcher may begin an experiment in the fall of the year and conclude it in the winter. The indi- vidual participant is engaged in a new therapy with great promise for assist- ing those with symptoms of depression. The change in seasons may have some influence on the individual as she is known to be affected by seasonal changes including the diminished levels of sunlight associated with winter. Or, the individual’s parents may have decided to get divorced. Or, the indi- vidual may have started a new job that had an influence on her beliefs about herself so that her actions in the study setting change. Clearly, some events are beyond the control of the researcher. What the researcher must achieve is (a) obtaining as much information as possible that may shed light on changes in the individual’s behavior (see Chapter 13 for qualitative mea- sures discussion); (b) seeking to control any foreseen events; and (c) if possi- ble, altering the presentation of the independent variable in such a way that changes in the dependent variable (e.g., multiple replications) are more clearly linked to the independent variable than the other events.

Maturation is similar to history except that maturation refers to the nat- ural development of an individual that occurs over time. In our example concerning accuracy in solving math problems, an individual might mature sufficiently during the length of an experiment so that her accuracy is improving because she is psychologically and cognitively better equipped to learn and excel, rather than as a result of the use of positive reinforcement. Maturation is generally controlled by attempting to limit the length of a study so that such influences are minimized. Still, the researcher does not necessarily have control over the length of a study and must be aware if maturation forces are at work. The researcher can account for these influ- ences by obtaining as much information as possible about the individual participant’s behavior not only in the experimental setting but also in other

102 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

settings. For example, if the individual began to excel in all areas of school, in her communication skill development, in her behavior at home, etc., there may be some reason to believe that positive reinforcement for improved math accuracy would hardly account for such universal changes. However, if the area of math accuracy improves while other areas of endeavor remain relatively unchanged, the researcher may be on firmer ground for ruling out the influence of maturation. Clearly, maturation is a greater threat as parti- cipants in a study are younger in age when even relatively short time periods may result in substantial development.

Attrition, as the term implies, is the loss of subjects during the course of a study. In experimental designs with large numbers of subjects, such events are not uncommon and, unless dramatic, tend to have little influence on the overall outcomes. In single subject research, the loss of an individual partic- ipant can be devastating. Later, we discuss the ethics of single subject research and, as a part of that discussion, we stress the need for informed consent. As part of the consent process, the researcher should stress the need for the individual (and his or her significant others) to commit to the behavior change. This is one reason why the social and ecological validities of desired changes are so critical. People must see clearly the worthiness of the outcomes of the study to be committed to seeing the study concluded. At times, attrition is unavoidable. We have had the unfortunate and sad instance of a participant dying during a study, another who required surgery and had to be hospitalized, and another whose family moved out of the state. Attrition can best be controlled by knowing with whom you are work- ing and whether or not any conditions exist that may suggest the likelihood of such an event. We stress, however, that the presence of any of these extraneous variables does not constitute a valid reason to deny treatment to a person in need. Rather, they are controlled as best as possible and sometimes merely explained as well as possible following the intervention.

Multiple Treatment Interference Multiple treatment interference is a very real threat in many single subject studies. Interference might occur when more than one independent variable (e.g., positive reinforcement followed by negative reinforcement) is used. For example, the negative reinforcement might be more effective when it is pre- ceded by positive reinforcement than if it is the only treatment used. Also, it could be that positive reinforcement did have an effect (albeit not a powerful one) that is compounded when negative reinforcement is implemented. In other words, negative reinforcement appears to have an effect that could be partially attributable to positive reinforcement. Other times, a package treatment (or independent variable) is used (e.g., positive reinforcement plus verbal prompting plus response interruption). Because more than one intervention is being used, it may be difficult if not impossible for the

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 103

researcher to accurately determine which of the components of the indepen- dent variable or variables actually accounted for changes in the dependent variable. Cooper et al. (2007) noted that researchers must ensure that the independent variable is carried out consistently by all involved and exactly as planned. Otherwise, the integrity of the treatment program may be jeop- ardized or treatment drift may occur. Treatment drift refers to individuals involved in administering the independent variable producing personal mod- ifications (consciously or unconsciously) that may influence the impact of the independent variable on the dependent variable (Cooper et al., 2007). The same precautions used to control observer drift should be used to control any potential treatment drift. Researchers also refer to treatment drift as intervention fidelity (or fidelity of implementation) (Gersten et al., 2005). Intervention fidelity is concerned with a trained observer rating the imple- mentation of the intervention (independent variable) to ensure it is being car- ried out correctly (with fidelity to the intervention procedures prescribed by the researcher). Intervention fidelity should include assessing the implementa- tion of key components of the intervention, adequate time allocation for the implementation of the intervention (e.g., by day or week), and coverage of a specified amount of material, curriculum or similar guide when appropriate (e.g., a minimum of one chapter in history) (Gersten et al., 2005). Similar to interobserver reliability, two observers should be used to rate, compare rat- ings, and ultimately verify intervention fidelity. Observations should occur throughout the study as well. Finally, some interventions, such as a therapy session, may not be easily observed and may not have a “strict” protocol. Nevertheless, videotaping and rating whether key features of the therapeutic approach were implemented may be needed (Gersten et al., 2005). In general, the more individuals involved in implementing an intervention (independent variable), the greater the possibility that there would be a lack of intervention fidelity. Both interobserver agreement on measuring the dependent variable and intervention fidelity are key components in ensuring internal validity.

Other extraneous variables are myriad and virtually infinite in number. They may be intrinsic to the individual or present in the setting. The researcher must be extremely careful that these inevitable influences have minimal impact on the study. More importantly, the researcher must be rig- orous in her or his self-examination to ensure that she or he is not in some way influencing the outcome of the study in an undue way. A researcher may find herself wanting so badly for an outcome to be achieved that she may be tempted at some level to arrange matters in such a way that those outcomes are more likely to occur. This could be done consciously (which would be very unethical) but is more likely to occur because the researcher did not carefully guard against such a possibility. For example, in our math accuracy example, we might select for our experimental setting one that we know is very well liked by the student and with a teacher who is so effective

104 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

that the outcome is virtually guaranteed. Such arrangements are sometimes referred to as systematic bias. That is, the researcher has introduced some element to the experimental conditions that is likely to influence changes in the dependent variable regardless of the manipulation of the independent variable. Perhaps the best manner in which to guard against such bias is to apply the old adage “Two heads are better than one.” The more people involved in or at least reviewing the procedures used in the study, the less likely that such influences will go unnoticed before the implementation of the study. Also, intervention fidelity observations should be of assistance. Although the impetus to improve individuals’ lives through desired changes in behavior is admirable, the researcher must be careful that this desire does not unduly influence her or his actions.

External Validity External validity refers to the degree to which the researcher may have con- fidence that she or he or other researchers will obtain the same or similar results if they use the same or very similar experimental procedures with other individuals, with other target behaviors, or in other settings. In simpler terms, the researcher must convince the reader that the treatment used will likely be effective if used by the reader under similar circumstances to those described by the researcher. When researchers conduct studies using the same experimental procedures as those in another study, we refer to this as replica- tion. The more an experimental effect is replicated by the same or other researchers, the greater the external validity. However, as Johnston and Pennypacker (1980) noted, the term replication may be interpreted to mean the replication of the procedures rather than specifically replicating effects. That is, one cannot precisely replicate results, but one might obtain similar results from precisely replicating experimental procedures.

Cooper et al. (2007) and Sidman (1960) described two types of replica- tion in establishing external validity, direct and systematic replication. In direct replication the researcher attempts to duplicate the procedures as pre- cisely as possible. Intrasubject replication occurs when the same subject is used in the subsequent study. Intersubject replication studies involve main- taining every aspect of an earlier study but with different although similar subjects. Intersubject replication is used more commonly to demonstrate that research findings are generalizable across particpants (Cooper et al, 2007). For example, five different experiments may be conducted using the same procedures (and same independent and dependent variables) in the same setting with five different subjects. Some designs, such as multiple baseline across subjects, include intersubject replication as a typical aspect of a study.

Systematic replication involves varying the conditions from an earlier study but still obtaining similar results. For example, a researcher may use the same experimental procedures as those from an earlier study, but apply

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 105

them in a different setting, or with subjects whose characteristics vary in some significant manner from the earlier study (e.g., applying the proce- dures with an individual subject with a different cultural and socioeconomic background from the subject in the earlier study, yet both exhibit behavior disorders). Target behaviors, administration of the experimental procedures, or virtually any aspect of a study may be varied (Cooper et al., 2007). If such variations are used, and similar results are obtained, then the generality of the procedures (external validity) is enhanced. If, however, different results are obtained, one may be unable to discern which variation may have caused different results. Of some importance then, is for researchers to carefully describe and define the independent and dependent variables, the characteristics of the participants and setting, and the various phases of the study (e.g., baseline and intervention phases) (Horner et al., 2005). Cooper et al. noted that systematic replications may occur as researchers conduct a series of studies, or may be conducted by different researchers. While the field of applied behavior analysis has moved increasingly toward studies involving socially valid behavior changes (Horner et al., 2005), it is still equally important to analyze procedures and results to make the best possible determination concerning which variables affect those changes. Although similar results may be obtained, knowing why those results were obtained is critical to expanding the knowledge base (Cooper et al., 2007).

4 C H E C K I T O U T # 2 Do you think it is possible that a dependent variable could be measured reliably and the intervention be implemented reliably, and yet the study lack validity? Can you describe an example?

Over time, researchers have focused on other aspects of validity that may be discussed in texts concerned with group research procedures. Frequently, the educational significance of obtained results is considered. Educational significance refers to the concern that, although statistically significant results may be achieved, the results should merit conclusions that the interventions used also translated into real-world significance. The question is asked, “Did the interventions result in outcomes that are meaningful to practitioners and to the lives of the participants in the study?” This concern may also be consid- ered in establishing social validity. In single subject research, this question is equally important. It is also related to the ethical treatment of individual parti- cipants. Ethical treatment is critical to obtain informed consent of participants and/or those responsible for their well-being and to ensure research is con- ducted in a humane manner that results in educational significance.

106 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Ramona Revisited #2 In the vignette of Ramona, there are a number of issues related to reliability and validity. Although the counselor did teach Ramona and her peers how to chart their own test performance, the counselor did not ensure any reli- ability checks. The counselor had to rely on honest charting by Ramona and the others. The counselor could have a second counselor observe and monitor both the course implementation and the biweekly meetings to estab- lish reliable implementation of the independent variable. Observer drift would not be likely in assessing the dependent variable (assuming Ramona and her peers did record their test grades honestly), but reactivity on the part of Ramona and her peers would be a possibility. Ramona might improve on her test performance at least partly as a result of knowing she had to record and chart her test performance. Maturation could also be a confounding variable. While Ramona is 18 years old, the experience gleaned from her previous semester and the second semester in which her improved test per- formance was established could be a factor in her improved outcomes. She may have improved as a result of more experience as a university student. Similarly, history could also have an impact. Clearly, Ramona was taking different courses in the new semester with different professors. It is possible these courses were more interesting, that Ramona was more motivated, or that the professors were easier in their grading. There is also the possibility of multiple treatment interference. While the intervention might be imple- mented with fidelity, the counselor might have considerable difficulty identi- fying which strategies included in the course and biweekly meetings have impacted or had little impact on Ramona’s performance.

Replication through Ramona’s peers helps with validity. If Ramona’s improvement continues to be replicated in subsequent semesters, that helps establish validity. Asking Ramona to also record the strategies she uses for each test could help in reducing multiple treatment factors and more clearly establishing those strategies that impact Ramona’s perfor- mance. Perhaps more importantly, if the counseling course yields consistent improvement in test performance among different students, enrolled in dif- ferent majors, and in different semesters, then replication is achieved such that the viability and value of the course may be more clearly established.

Ethics Single subject studies may incorporate elements of qualitative and quantita- tive procedures. As in much qualitative research, single subject research emphasizes change within the individual or individuals participating and the value placed on those changes by participants themselves and significant

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 107

others. Therefore, the need to obtain data that are not numerical is common in single subject research. Interviews and observations with family members, teachers, and the subjects themselves may be used. The compilation and presentation of data from multiple sources to verify changes and the social validity of those changes are found in single subject studies. As in much quantitative research, there is a focus on a target behavior or dependent variable by which changes in individuals can be objectively verified. Like quantitative group designs, single subject designs may combine the perfor- mance of individuals on some measure to determine the effectiveness of interventions across a sample of individuals.

In short, single subject designs are versatile and flexible, allow for the use of a variety of data collection and presentation techniques, may involve an individual or a number of persons, focus on socially valid changes, incor- porate a wide variety of interventions and outcome measures, are applicable in educational and clinical settings, have been used with people across all ages and with many types of strengths and challenges, and provide for both internal and external validity. Single subject designs also are relatively easy to understand.

Because single subject designs may incorporate robust and powerful methods for influencing the behavior of individuals, ethical concerns are extremely important. Walker and Shea (1991) posed the following questions regarding ethical and humane treatment:

• Who shall decide who will manage behavior?

• Who shall decide whose behavior is to be changed?

• Who will guarantee that the behavior manager behaves ethically?

• What type of interventions will be used?

• Who will determine if these interventions are ethical?

• What are the outcomes sought?

Walker and Shea also noted that these considerations are especially critical when applied to children. They modified the above questions with some additions to consider specifically working with children. These include

• What is a child?

• Is a child free to make choices?

• Should a child be allowed to make choices concerning interventions and treatments?

• Does a child behave in a manner that is observable and predictable in accordance with principles of applied behavior analysis?

• Can a child’s behavior be changed by external forces?

• Can an educator modify a child’s behavior?

108 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

• Can another child or parent modify a child’s behavior?

• Who shall determine which and whose behavior is to be modified?

• Which interventions are to be applied in school or in other settings?

• Who will approve the use of these interventions and monitor their ethical use?

• What are the outcomes sought?

These questions pose both philosophical and ethical issues to be addressed by each single subject researcher. Walker and Shea suggested the following guidelines that they found in the research literature (cited in Walker & Shea, 1991):

• Explore alternative interventions before selecting aversives (Schloss & Smith, 1987).

• Consider potential side effects and injury that may occur as a conse- quence of any intervention (Sabatino, 1983).

• Determine whether the individual understands the treatment program (Hewett, 1978).

• Anyone involved in applying an intervention should be trained and comfortable with the procedures (Morris & Brown, 1983; Rose, 1989).

• Empirical evidence should be available that indicates the intervention is effective (Morris & Brown, 1983).

• Any formal plans (e.g., Individualized Education Program, Individual- ized Family Services Plan, Individual Written Rehabilitation Plan) should be consistent with the planned treatment and should be agreed to by the principals involved in those plans (Morris & Brown, 1983; Singer & Irvin, 1987).

• The planned intervention should be carefully monitored, its results should be documented, and it should be regularly evaluated (Morris & Brown, 1983).

• Informed consent should be obtained and include information about the nature of the program, benefits, risks, expected outcomes, and possible alternatives to the planned treatment (Axelrod, 1983; Kazdin, 1980; Morris & Brown, 1983; Rose, 1989; Singer & Irvin, 1987).

• The principle of normalization should be applied (Allen, 1969; Nirje, 1967). This principle demands that individuals with disabilities be given opportunities and treatments that are as close to normal as possible.

• The procedure used should be fair and appropriate to the degree of concern regarding the behavior targeted for change (Allen, 1969; Sabatino, 1983). In other words, are due process safeguards in place

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 109

and will the outcomes result in an improved life for the individual whose behavior is changing (Walker & Shea, 1991)?

• The dignity and fundamental human worth of the individual should be protected (Allen, 1969). Do the proposed procedures embody respect for the individual as a human being (Walker & Shea, 1991)?

• Committee review of all procedures should occur (Rose, 1989; Singer & Irvin, 1987). Human rights committees and institutional review boards for studies involving human subjects are commonly found in school, res- idential, community agency, and higher education settings and may well be required by law. These committees may monitor the necessity, qual- ity, and social validity of the procedures used. Peer review of procedures is also recommended (Axelrod, 1983).

• Finally, the principle of the least restrictive environment should be applied (Singer & Irvin, 1987). This principle may be viewed as apply- ing the procedures in the least restrictive manner and environment, and such that outcomes of the research project increase the likelihood that the individual will remain in the least restrictive environment or will move to a less restrictive one.

When each of these points is addressed, the researcher or practitioner is establishing what some refer to as the empirical and social validity of the proposed study (Evans & Meyer, 1985). Social validity refers to the degree to which other people think that the targeted changes in behavior are impor- tant and that the methods used to encourage behavior change are accept- able. Social validity also implies that: (a) the magnitude of change in the dependent variable be socially important, (b) the implementation of the independent variable is cost effective, practical, and humane, (c) those implementing the intervention are typical change agents (e.g., psychologist, teacher, social worker who has direct involvement with the individual), (d) that changes in the dependent variable are maintained over time, and that (e) environmental and social contexts of the study are also typical (e.g., school classroom, home, community) (Horner et al., 2005). Similarly, empirical validity refers to the measurements that actually demonstrate that the proposed behavioral changes will indeed positively affect the individual’s life (Evans & Meyer, 1985). Evans and Meyer noted that the charting of individual behavioral responses may leave unanswered questions. Such mea- sures do not answer whether clinically or educationally significant changes have occurred; if good or bad side effects have occurred; or whether the interventions used were appropriate, humane, and carried out in accordance with philosophical and legal assumptions regarding the rights of individuals (Evans & Meyer, 1985).

110 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

4 C H E C K I T O U T # 3 How does the inclusion of a child in a single subject study influence how the study must be conducted?

There are many issues in the use of applied behavior analysis that are dis- cussed at greater length and detail in the research literature or in texts devoted to interventions using applied behavior analysis techniques. Although we dis- cussed methods for changing target behaviors in Chapter 2, the reader is strongly advised to obtain any and all information relevant to the use of any particular procedure. Texts and articles related to group designs may include information (e.g., obtaining and selecting subjects when whole groups are to be treated as a single case) that may prove useful. Also, many organizations (e.g., the American Association on Intellectual and Developmental Disabil- ities) publish materials related to informed consent and the ethical treatment of individuals with disabilities. The reader is encouraged to avail herself or himself of these materials. This chapter should provide an overview of those issues relevant to single subject research.

Key Concepts/Terms Prediction—The idea that if the independent variable has no effect on the

dependent variable, the data path will remain unchanged across phases. Verification—The confirmation that the dependent variable is changing in a

predictable fashion as the independent variable is systematically applied. Replication—The repeating of the predictions and verifications within the

same study. Reliability—In single subject research, we are concerned primarily with

interobserver reliability; the researcher must ensure observational proce- dures and results are reliable.

Reactivity—An individual altering his or her behavior as a result of being observed.

Observer drift—A change in interpretation of the agreed upon operational definition of the target behavior; this is a threat to reliability.

Internal validity—The degree to which the researcher has adequately con- trolled the independent, dependent, and extraneous variables so that there is confirmation of a functional relationship.

History—The passage of time and both foreseen and unforeseen events; this is a threat to internal validity.

Maturation—The natural development of the individual over time; this is a threat to internal validity.

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 111

Attrition—The loss of a subject during the course of a study; this is a threat to internal and external validity.

Multiple treatment Interference—Effects from previously used interven- tions (e.g., in an A-B-C design where interference from B might influence the outcomes in the C phase) or when package interventions are used (e.g., BC phase when it is difficult to determine whether B or C has the greater influence or if only the combination has the effect).

Treatment drift—A situation when individuals involved in administering the independent variable are producing personal modifications (consciously or unconsciously) that may influence the impact of the independent variable on the dependent variable.

Intervention fidelity—The degree to which the researcher, through system- atic observations by two or more raters, can verify that the independent variable was consistently carried out according to the prescribed procedures.

External validity—The degree to which the researcher (or consumer of the research) may have confidence that similar results will be obtained if the experimental procedures are used with other individuals, in other settings, with other behaviors, and so on.

Direct replication—Occurs when a researcher duplicates as precisely as possible the procedures used in a previous study and similar results are obtained.

Systematic replication—Occurs when experimental conditions are varied but similar results are obtained.

Educational significance—Refers to the concern that, although statistically significant results may be achieved, the results should merit conclusions that the interventions used also translated into real-world significance.

Social validity—Refers to the degree to which other people think that the targeted changes in behavior are important and that the methods used to encourage behavior change are acceptable.

Empirical validity—Refers to the measurements that actually demonstrate that the proposed behavioral changes will indeed positively affect the individual’s life.

4 Possible Answers to Check It Out

¶ The student’s performance on correct punctuation usage (currently at50%, 40%, and 40% for baseline) could be predicted to remain at levels similar to the baseline if the independent variable (the reinforcement program) has no effect. As a result of the introduction of the independent variable, the researcher verifies there has been a change in the independent variable in a positive direction (e.g., 70%, 80%, and then 90% correct for

112 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

five consecutive assessments). Finally, if the researcher would then withdraw the reinforcement program, she could predict that the student’s percentage correct will remain at the higher levels if the reinforcement program is having no effect. She then verifies that the percentage correct deteriorates toward baseline levels (e.g., 60%, 50%, 40% for three consecutive assess- ments) when the independent variable is withdrawn. When she reintro- duces the independent variable, the student’s performance once again improves to above 90% correct for five consecutive assessments. The researcher has now replicated the original prediction (no change in the data trend if the independent variable has no effect) and the original veri- fication (the dependent variable does change with the implementation of the independent variable).

· It is possible to have a reliably measured dependent variable and areliably implemented intervention and still not achieve validity. Extra- neous variables may account for changes in the dependent variable even in the presence of the reliable measurement of the dependent variable and intervention fidelity. For example, a young child could mature over a period of a nine-month study such that assuming an intervention accounted for changes in the dependent variable (e.g., communication skill development) might prove difficult. Another example might be the same young child was reunited with a beloved grandparent who moved back into the family home. Historical events could influence the child’s behavior.

¸ The inclusion of a child as a subject in any study has considerableinfluence on the procedures used in planning and implementing the study, although some of the same procedures would be required for adult participants as well. First, informed consent must be obtained from the caregivers of the child. The social validity of the behavior change should be clearly evident to the caregivers and that the first and foremost concern is that the child will be treated in a humane manner. Because caregivers are not always as well versed in research methods, interventions, etc. as they might be, there should be procedures in place (e.g., review board or team) to review the study and ensure that the researcher is acting in the child’s best interests prior to, perhaps during the study, and after the study’s comple- tion. The right to withdraw from the study at any time should also be clearly communicated. The researcher should regularly review measurement of the dependent variable, fidelity of the intervention implementation, and other study procedures to ensure that the child is treated in a humane and ethical manner at all times, especially when the intervention might involve some adverse procedure (e.g., withholding privileges commonly available to all children in the environment). The confidentiality of the child should be protected in disseminating any results beyond the child and his/her caregivers. There are other safeguards that might be required in any given

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 113

circumstance. Institutions that serve children and adults typically have review boards and procedures in place to review study procedures and ensure ethical behavior and treatment.

References Allen, R. C. (1969). Legal rights of the disabled and disadvantaged (GPO 1969-

0-360-797). Washington, DC: U.S. Government Printing Office. Axelrod, S. (1983). Behavior modification for the classroom teacher. New York:

McGraw-Hill. Barlow, D. H. & Hersen, M. (1984). Single case experimental designs: Strategies

for studying behavior change (2nd ed.). New York: Pergamon Press. Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis

(2nd ed.). Upper Saddle River, NJ: Pearson. Evans, I. M. & Meyer, L. H. (1985). An educative approach to behavior problems:

A practical decision model for interventions with severely handicapped learners. Baltimore: Brookes.

Gersten, R., Fuchs, L. S., Compton, D., Coyne, M., Greenwood, C., & Innocenti, M. S. (2005). Quality indicators for group experimental and quasi- experimental research in special education. Exceptional Children, 71, 149–164.

Hewett, F. M. (1978). Punishment and educational programs for behaviorally dis- ordered and emotionally disturbed children and youth: A personal perspective. In F. Wood & K. Lakin (Eds.), Punishment and aversive stimulation in special education (pp. 101–117). Minneapolis: University of Minnesota.

Horner, R. H., Carr, E. G., Halle, J., Mcgee, G., Odom, S., & Wolery, M. (2005). The use of single-subject research to identify evidence-based practice in special education. Exceptional Children, 71, 165–179.

Hudson, S. S., Lewis, T., Stichter, J. P., & Johnson, N. W. (2010). Putting quality indicators to the test: An examination of 30 years of research. Journal of Emo- tional and Behavioral Disorders, 19, 143–155.

Johnston, J. M. & Pennypacker, H. S. (1980). Strategies and tactics for human behavioral research. Hillsdale, NJ: Erlbaum.

Kazdin, A. E. (1980). Behavior modification in applied settings (2nd ed.). Home- wood, IL: Dorsey Press.

Morris, R. J. & Brown, D. K. (1983). Legal and ethical issues in behavior modifica- tion with retarded persons. In J. Matson & F. Andrasik (Eds.), Treatment issues and innovations in mental retardation. New York: Plenum.

Nirje, B. (1967). The normalization principle and its human management implica- tions. In R. Kugel & W. 1. Wolfensberger (Eds.), Changing patterns in residen- tial services for the mentally retarded. Washington, DC: President’s Committee on Mental Retardation.

Rose, T. L. (1989). Corporal punishment with mildly handicapped students: Five years later. Remedial and Special Education, 10, 43–52.

Sabatino, A. C. (1983). Discipline: A national issue. In A. C. Sabatino & L. Mann (Eds.), Discipline and behavior management (pp. 1–27). Rockville, MD: Aspen.

Schloss, P. J. & Smith, M. A. (1987). Guidelines for ethical use of manual restraint in public school settings for behaviorally disordered students. Behavioral Disor- ders, 12, 207–213.

114 PART 1 CONDUCTING SINGLE SUBJECT RESEARCH: ISSUES AND PROCEDURES

Sidman, M. (1960). Tactics of scientific research. New York: Basic Books. Singer, G. S. & Irvin, L. K. (1987). Human rights review of intrusive behavioral

treatments for students with severe handicaps. Exceptional Children, 57, 298–313.

Tawney, J. W. & Gast, D. L. (1984). Single subject research in special education. Columbus, OH: Merrill.

Walker, J. E. & Shea, T. M. (1991). Behavior management: A practical approach for educators. Englewood Cliffs, NJ: Prentice-Hall.

CHAPTER 4 ISSUES IN SINGLE SUBJECT RESEARCH 115

P A R T 2 Overview and Application of Single Subject Designs

CHAPTER 5 Overview of Withdrawal Designs 119

CHAPTER 6 Application of Withdrawal Designs 139

CHAPTER 7 Overview of Changing Conditions and Changing Criterion Designs 165

CHAPTER 8 Application of Changing Conditions and Changing Criterion Design 185

CHAPTER 9 Overview of Multiple Baseline Designs 205

CHAPTER 10 Application of Multiple Baseline Designs 235

CHAPTER 11 Overview of Alternating Treatments Designs 263

CHAPTER 12 Application of Alternating Treatments Designs 283

117

CHAPTER

5 Overview of Withdrawal Designs

IMPORTANT CONCEPTS TO KNOW A AND B DESIGNS

THE A-B DESIGN The A-B Design and Action Research

MECHANICS OF THE WITHDRAWAL DESIGNS The A-B-A Design The A-B-A-B Design

PREDICTION, VERIFICATION, AND REPLICATION

ADVANTAGES OF THE WITHDRAWAL DESIGN

DISADVANTAGES OF THE WITHDRAWAL DESIGN

ADAPTATIONS OF THE TYPICAL WITHDRAWAL DESIGN The B-A-B Design (No Initial Baseline) The A-B-A-B-A-B Design (Repeated Withdrawals)

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

119

The withdrawal design is widely recognized across disciplines as a basicexperimental procedure for demonstrating treatment effects. The goal isto show that specific treatment(s) or intervention(s) have a direct influence on changing behavior. The typical withdrawal design is usually designated by the letters A-B-A-B, where A represents baseline and B represents a treatment or intervention. There are variations of this design that will be described later.

Withdrawal refers to the withdrawal of treatment during one or more phases of a study to demonstrate the effects that it has on the target behavior. This design also has been referred to as the equivalent time samples design (Campbell & Stanley, 1963), the interrupted time series with multiple replications design (Cook & Campbell, 1979), and the within series elements design (Barlow, Hayes, & Nelson, 1984). It is often referred to as the reversal design, which was initially described by Baer, Wolf, and Risley (1968). The term withdrawal is preferred, however, because it describes the mechanics of the design (withdrawal of intervention) rather than the intended outcomes of the design (reversing the direction of the target behavior). In other words, the intent is to examine the effect of introducing and subsequently removing or withdrawing the intervention, not to actively reverse the effect of the intervention. There is actually a distinction between withdrawal and reversal designs. In the latter, there is an active attempt to use procedures (usually some differential reinforcement procedure) to reverse the effects of a treatment. Examples of true reversal designs, sometimes designated as A-B-A’-B, actually are quite rare in the professional literature (Gast & Hammond, 2010). The A’ signifies a reversal (of treatment) to baseline, whereas A signifies a withdrawal (of treatment) to baseline. Subsequently, although the term reversal design is often used, the term withdrawal design will be used throughout this book.

The withdrawal design is a powerful design because it allows the investigator to directly and easily demonstrate cause-effect relationships between behavior and intervention. Because of the nature of the withdrawal design, the effects of variables such as history and maturation are ruled out by demonstrating that the behavior change occurs only with the introduction or withdrawal of treatment. Further, the more often and longer the intervention is introduced and withdrawn (the more replications of the effect of the independent variable on the dependent variable), the stronger the case for a cause-effect relationship.

Before discussing the mechanics of the withdrawal design, we will first present information on the components of the design, the A and B conditions. In addition, a brief discussion of the most basic single subject design, the A-B design, will be presented. In this design, there is a baseline and a systematic introduction of an intervention, but not a subsequent withdrawal of that intervention.

120 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

A and B Designs As mentioned earlier, when using single subject design terminology, “A” refers to a baseline condition and “B” refers to a treatment or intervention condition. It is possible that information may be collected that only include an A condition or might only include a B condition, even though these so called A designs and B designs should not be considered research designs per se. The one characteristic of these approaches is the systematic observa- tion and measurement of behavior and might better be thought of as quantitative-descriptive paradigms (Gast & Hammond, 2010). In an A design, behavior is simply observed and recorded, which might provide important information about the status of some behavior of interest. For example, a teacher who is concerned about the disruptive behavior of one of her students might record the number of distractions requiring her redi- rection during the school day for a few weeks to document his behavior prior to a referral for possible special education placement. Within this paradigm, no attempt is made to change or provide an intervention for the behavior. Rather, the intent is simply to describe the current status of the behavior.

In a B design, behavior is observed and recorded during or after some change or intervention has taken place but not before. Rubin and Babbie (2011) provided an example of the utility of the B design in social work. They described a situation in which a client might have been provided pro- longed exposure therapy for Post-Traumatic Stress Syndrome but no base- line data were collected prior to therapy. Suppose data collection of the frequency of traumatic symptoms during therapy was subsequently started and a decreasing trend was noted. It would suggest that “you would have no reason to suppose that prolonged exposure therapy was the wrong choice for your client” (p. 314). Conversely, if the data trend did not decrease, particularly for a long period of time, it would suggest that a dif- ferent approach should be used. One important point should be made that is demonstrated in this design. As defined in Chapter 1, baseline data are those that are collected prior to the implementation of an intervention (treatment). However, baseline data also refer to those collected under the current con- ditions (before a new treatment is introduced). In the above example, one could argue that the data obtained were actually baseline data collected under the current conditions. Subsequently, in the first case, in which no new trend was noted, there was no treatment warranted, whereas in the sec- ond case, the trend indicated that a new treatment was warranted.

Regardless of the semantics used, the importance of A and B designs lies in the systematic observation and recording of behavior, providing informa- tion on which educational and clinical decisions can be made. They are not

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 121

intended to investigate any relationship between behavior and an interven- tion, or serve as an experimental research design.

The A-B Design As noted, the A-B design is considered the most basic of the single subject designs, involving the collection of baseline data (A) over a brief period of time and then the implementation of a treatment (B) to determine its effect on the target behavior, usually over a brief period of time. In fact, all other single subject designs can be viewed essentially as variations of the A-B design. Unfortunately, from an experimental point of view, it is also the weakest of the single subject designs because the functional relationship between the dependent and independent variables is not firmly established. Even though baseline and treatment conditions are both present, and changes in the target behavior might be attributed to the treatment, there are many other factors (e.g., maturation, history, practice effects) that could also play a role. Although there is no cause-effect relationship clearly established, the A-B design has advantages over a casual or subjective eval- uation of teaching or therapeutic effectiveness. First, the A-B design requires that a specific behavior be targeted, and that data be collected over time. This at least provides an objective record of an individual’s behavior/perfor- mance and, in itself, could be valuable. For example, if baseline data were inconsistent or extremely variable, then reasons for the inconsistency might be pursued, even before an intervention was introduced. Second, because behavior is observed and recorded over time, the effects of the intervention might be more apparent if the behavior changed immediately after its intro- duction. If the behavior changed gradually over time, then other factors might be involved. It is less likely that any immediate behavioral change is due to other factors. However, if the change only occurred immediately, then there may be a novelty effect that must be considered.

4 C H E C K I T O U T # 1 A parent comes to you very excited because she recently started using a program she heard about on TV that “guarantees to increase your preschool child’s vocabulary.” She stated that her child spoke only two or three words before the program but after 2 months, now has a vocabulary of almost 20 words. What could you tell the mother about the effectiveness of the program?

The A-B design may be useful for teachers or other change agents whose primary goal is to change performance or behaviors of their students or cli- ents. A common attitude by many might be “I don’t care why the behavior

122 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

changed, I’m only interested that it did.” In this case, the A-B design might provide an alternative between a more haphazard, uncontrolled approach and a highly controlled situation. There are also additional steps that can be taken to enhance the A-B design. Assume that a teacher finds that a particular inter- vention is useful in changing a specific behavior of a student. One step would be to use that intervention with other behaviors in the same student or to use it with the same behavior with other students. If positive results are replicated, then the intervention would have more educational or clinical utility. In other words, the external validity would be strengthened. Another step would be to check on the behavior at some future time to see if it has maintained or continued to change in the desired direction. The use of such probes are useful in many single subject designs that will be discussed later in this book.

The A-B Design and Action Research Another consideration of the A-B design might be related to its role in action research, an approach frequently used in education and other applied settings. In this approach, the teacher or other change agent is an active par- ticipant who initially identifies some problem or topic that is a concern in the immediate, natural environment (e.g., the classroom). There are many different models of, and approaches to, action research although most begin with an inquiry approach of identifying a problem. Note: Those inter- ested in finding out more about action research should consult Action Research: A Guide for the Teacher Researcher (Mills, 2007).

After the central topic is identified, three other steps are usually involved in action research (Mertler, 2009). These are 1) fact finding, or observing and monitoring current practice, 2) synthesizing the information and data, and 3) taking some sort of action. In other words, the goal is to identify an action that will generate some improvement that the researcher believes is important (Mills, 2007). The first two steps are essentially those that occur within the “A” design. This involves the careful identification of a problem (behavior), collecting information, including data, analyzing those baseline data, and interpreting the data and making decisions (the action plan). After an action plan is determined, it can then be initiated and evaluated (step 3 above and the “B” segment of the design), with the goal of communicating and sharing the results and reflecting on the process (Mertler, 2009). This essentially becomes a type of A-B design. Consider the following example:

Ms. Tracy has been very concerned that her fifth-grade class is very negligent in turning in their math homework. She has to spend a good deal of class time informally evaluating the information covered in the homework to make sure the students understand it so she can move to the next lesson. When she mentioned this in the teacher’s lounge, there was overwhelming agreement from other teachers that homework, in

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 123

general, was becoming a problem, particularly for fifth graders. Ms. Tracy interviewed a few of her students—some who typically complete their homework and some who do not. She concluded that the students’ social life was becoming very important, and they typically called or texted their classmates at home. Homework was not viewed as a high priority. The students also said they had problems “getting started;” when and if they did get started, they could usually complete the homework in about 30 minutes. Ms. Tracy discussed this feedback with the other fifth- grade teachers and synthesized all the information. She decided that the action she would take would be to allow the students to work in small groups, or teams, for 10 minutes each day to ask any questions they might have and to actually begin their math homework. She talked over the new plan with her students who liked the idea. She then calcu- lated from her grade book how many students in her class on average had been turning in completed math homework and found that only about 33% had been doing so on any given school day. Subsequently, Ms. Tracy started the program and saw an immediate increase in math homework completion. Over the next 3 weeks, the class average increased to 90% of students completing math homework on any given school day.

Technically, this should be considered an A-B design because baseline data were obtained and an intervention was initiated with data collected on the same target behavior and used for comparison against the baseline data. Ms. Tracy was satisfied that her new approach to encouraging home- work completion was responsible for the change in her students’ behavior.

In this sense, action research could be seen as a reflective method of iden- tifying a problem, gathering and synthesizing the information and data, and coming up with a solution. The A-B design can be seen as a method to objec- tify the process. In some instances, data may only be collected during the intervention leading essentially to a “B” design which is less desirable from an action research perspective, but still may result in a “therapeutic” effect that remedies the identified problem and is sufficient to satisfy the teacher or change agent. Another characteristic of action research is that the action plan frequently generates a new cycle (Hinchey, 2008). This implies that the action plan would be initiated and evaluated and then modified or changed if neces- sary. This describes an A-B-C design. This design and other changing condi- tion designs are discussed in Chapter 7.

4 C H E C K I T O U T # 2 The action plan Ms. Tracy implemented appeared to be successful. What additional steps could she take to determine the effectiveness of her program?

124 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Mechanics of the Withdrawal Designs The limitations of the A-B design do have to be acknowledged, particularly if experimental control is desired. To better establish the functional relationship between the target behavior and the treatment, a researcher might withdraw the intervention to see if the behavior changes toward or returns to the base- line level (A-B-A). In such a design, the goal is to show that there is a func- tional relationship between the target behavior and the intervention. In other words, the behavior will change as a function of the presence or absence of the intervention (verification as discussed in Chapter 4). Although it shows more experimental control than the A-B design, the A-B-A design is still not recommended in educational or clinical settings because the study concludes with the subject in a nontreatment phase. This is particularly true if the inter- vention is shown to be successful because its withdrawal will leave the subject at the unwanted, pretreatment level. As a way of addressing these shortcom- ings, the researcher might also reintroduce the treatment to determine if the pattern of behavior will again change (A-B-A-B). In this chapter, we will refer to the A-B-A-B design as the typical withdrawal design.

Specifically, the A-B-A-B design includes the following steps:

1. Baseline data are collected on a target behavior before an intervention is introduced (Al);

2. The intervention is introduced for a specific period of time and data are collected on the same target behavior (Bl);

3. The intervention is withdrawn for a short period of time to determine if the target behavior returns back to the baseline level (A2); and

4. The intervention is reintroduced to see once again the effects on the target behavior (B2).

Figure 5-1 shows data that demonstrate the functional relationship of the target behavior and the intervention. Note that the behavior increased during the presentation of the intervention and decreased during its with- drawal. Figure 5-2, on the other hand, shows data that do not demonstrate a functional relationship. In this case, the withdrawal of the intervention (A2) did not increase the target behavior toward baseline levels. There are several reasons why the trend in Figure 5-2 might occur and we will later discuss ways to minimize this and provide situations in which a withdrawal design would not be the appropriate choice.

The A-B-A Design In the first step in an A-B-A design, the investigator must precisely define both the target behavior to be altered and the treatment to be implemented. The next step is to collect baseline data (A) for a predetermined number of

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 125

sessions or, preferably, until a stable baseline trend is established (see Chapter 1 for determining a stable baseline). Additionally, if the baseline data are con- sistently moving in a counterproductive or undesirable direction, then the researcher should move to the next step. The third step would then be to

B eh

av io

Baseline A1

Intervention B1

Intervention B2

Baseline A2

Sessions

FIGURE 5-1 Data that indicate a

functional relationship

B eh

av io

Baseline A1

Intervention B1

Intervention B2

Baseline A2

Sessions

FIGURE 5-2 Data that do not indicate a

functional relationship

19 99

126 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

introduce the treatment or intervention (B) and collect continuous data either until a specific criterion has been met (e.g., reduction of out-of-seat behavior to 25% or less) or until a stable mode of responding in the desired direction is recorded. The last step is to withdraw the treatment or return to the baseline phase (A). The logic of this design is that if the target behavior improves in the desired direction during the intervention phase and changes toward baseline levels once the treatment is withdrawn, then the investigator can conclude that the treatment was indeed responsible for the improvement of the target behavior.

The following is a hypothetical example of an A-B-A design.

Johnny is a 7 year-old second-grade student with a learning disability who is in an inclusion class with 20 students. Johnny has great difficulty staying on task and completing his assignments during independent seatwork. His teacher, Ms. Little, wanted to improve his on-task behavior and decided to use praise as a reinforcer for that purpose. The first step was to carefully define the target behavior related to working in his math workbook so that the instances of on-task behavior could be recorded. She identified the inter- vention as providing verbal praise (“good working, Johnny”, etc.) when he was on-task. Every 30 seconds of a 10-minute session, Ms. Little would observe Johnny to see if he was “on-task” based on her definition and crite- ria. During the baseline phase (A) she simply observed and noted that Johnny spent considerable time fidgeting, tapping his pencil, and staring out the window. During this time (five consecutive days) he was on task only about an average of 4 times out of the possible 20 observations. Ms. Little then introduced the treatment (B) by praising him each time he was on task dur- ing her 30-second observations. During this intervention phase, she col- lected data for 10 days. During this time, Johnny’s on-task behavior increased to 90% of the observations, averaging around 15/20. During the 4th week, for 5 days, she again only observed and recorded Johnny’s on-task behavior, thus there was a return to the baseline condition (A). During this period, Johnny again became off-task and returned to similar baseline levels. Those data are graphically presented in Figure 5-3.

This design allows the teacher to conclude that the intervention was indeed responsible for the change in Johnny’s on-task behavior. However, this design ends in a baseline phase where he was primarily off task during the instruction time. To avoid this negative situation, an A-B-A-B design, a more clinically appropriate design that reintroduces the treatment, should be used.

4 C H E C K I T O U T # 3 What type of recording procedure did Ms. Little use with Johnny?

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 127

The A-B-A-B Design As explained previously, the investigator must precisely define a target behavior and the intervention to be used before collecting data. The com- mon four-phase A-B-A-B withdrawal design involves a no-intervention base- line (A1) and an intervention phase (B1), with each phase being repeated (A2 and B2). Introducing the intervention twice to compare the target behavior with two baseline phases is done to strengthen or validate the func- tional relationship between the target behavior and treatment.

Consider the following example.

Carla, a 6 year-old student with autism, has a moderate vocabulary but uses oral language very infrequently. She often uses body language or points to objects rather than communicating verbally when interacting with adults. The speech and language clinician, Mr. Mathis, wanted to increase Carla’s verbalizations for communication purposes. He worked with Carla’s special education teacher, Ms. Diaz, to develop a program to encourage this goal. Because Carla often brought fruits and raw vegetables for snacks, Ms. Diaz decided to use a raisin as a positive rein- forcer for Carla’s verbalizations. During the first baseline (A1), Ms. Diaz collected data on the total number of verbalizations in a 30-minute free- time period. During the first intervention phase (B1), she gave Carla a raisin after each verbalization and collected data. During the second baseline (A2) phase, the reinforcer was withdrawn. As can be seen in

Baseline A Intervention B Baseline A 20

0 1 2 3 4 5 6 7 8 10

Days

N um

b er

o f O

n -t

as k

se ss

io n

9 11 12 13 14 15 16 17 18 19 20

FIGURE 5-3 Johnny’s data from an

A-B-A Design

20 14

128 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Figure 5-4, the pattern of the second baseline (A2) data closely resembles the pattern of Carla’s first baseline (A1) data. Ms. Diaz was able to assume that there was a functional relationship between the use of the reinforcer and Carla’s increased verbal responses. However, to further establish the functional relationship and leave Carla in a more clinically appropriate situation, Ms. Diaz reintroduced her treatment (B2) and col- lected data. Figure 5-4 also shows that Carla’s verbal responses increased to a level consistent with the first intervention phase (B1).

Prediction, Verification, and Replication The issues of prediction, verification, and replication can be easily described for the typical withdrawal, or A-B-A-B, design. After stable baseline (A1) data are collected, one would predict that the pattern would remain the same if the intervention had no effect. However, one would also predict that if the treatment (B1) has an effect, a different pattern would emerge. Verification occurs first when the change from baseline to intervention phase results in a change in responding. Then, again, verification occurs when return to the baseline condition (A2) results in a pattern similar to the original baseline condition (A1). Finally, replication occurs when the sec- ond treatment phase (B2) results in a similar response pattern as the previ- ous treatment phase (B1). These concepts can be better understood by referring back to Figure 5-1. As will be discussed later in this chapter,

Baseline A1

Baseline A2

Intervention B1

Intervention B2

Free Time Periods

N um

b er

o f V

er b

al iz

at io

n s

0 1 2 3 4 5 6 7 8 109 11 12 13 21 2214 15 16 17 18 19 20

FIGURE 5-4 Carla’s data from

an A-B-A-B Design

20 14

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 129

some withdrawal designs might include the reintroduction and withdrawal of the intervention more than once. This further strengthens the evidence of prediction, verification, and replication.

Advantages of the Withdrawal Design The withdrawal design is one of the easiest single subject designs to imple- ment. Before a discussion of the advantages of the withdrawal design, however, a brief note is necessary about the assumptions made when using it. The withdrawal design is a powerful design by demonstrating that the withdrawal of a particular treatment will return the target behav- ior to baseline, or pre-intervention, levels, and then back again with the reintroduction of the treatment. This means that the target behavior itself must be reversible. Unfortunately, this is often neither possible nor desir- able. For example, if a target behavior is learned (e.g., reading), it is not likely or desirable that it would be unlearned when the intervention is withdrawn. There are also times when the target behavior might be main- tained even after the treatment is withdrawn because of factors in the natural environment. For example, an individual is shown to be verbally abusive to his peers (A1) and treatment is introduced that reduces his verbal abuse (B1). After the treatment is withdrawn (A2), however, the behavior remains at a reduced level. It is possible that the reduction of ver- bal abuse was maintained because it resulted in the individual receiving positive reinforcement from, and engaging in positive interactions with, his peers. In addition to the target behavior being reversible, the nature of the treatment must also be such that the effects are reversible. For exam- ple, if a student was taught a mnemonic strategy to increase memory skills, then it is unlikely that removal of the treatment will decrease the memory skills. In this situation, the treatment has been learned or internalized, thus affecting the target behavior.

If the criteria for reversibility of the behavior and treatment can be met, the withdrawal design is a powerful design that documents the functional relationship between the independent and dependent variables. As Cooper, Heron, and Heward (2007) noted, “An investigator who reliably turns the target behavior on and off by presenting and withdrawing a specific variable makes a clear and convincing demonstration of experimental control” (p. 185). This is particularly true when the treatment is presented and with- drawn a number of times in the same study. This is referred to as the repeated withdrawal design and is discussed later in this chapter. The alter- ation of baseline and treatment conditions also provides direct evidence of prediction, verification, and replication of treatment effects. In summary, the

130 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

withdrawal design is a powerful design when it is used in the following situations:

• When a clear functional relationship between the independent and dependent variables needs to be demonstrated;

• When the nature of the target behavior is such that it can be reversed when the treatment is withdrawn;

• When the nature of the treatment is such that its effects are not present on the target behavior after it is withdrawn;

• When withdrawal of treatment does not compromise ethics.

Disadvantages of the Withdrawal Design Even though the withdrawal design is both powerful and easy to implement, it has several disadvantages. As discussed, there are practical and ethical issues related to the required reversibility of the target behavior in order to demonstrate its functional relationship with the independent variable. It is certainly desirable once a behavior has changed in a more positive direction that it continues in that direction or stays at that level. In fact, maintenance of treatment effects is an important goal in and of itself. One compromise suggested by Mayer, Sulzer-Azaroff, and Wallace (2012) is that treatment can be reintroduced after the behavior demonstrates “an unambiguous shift in the direction of baseline” (p. 165) rather than a complete return to baseline level. In this way, the subject does not have to spend as much time in a nontreatment phase. However, if this procedure is used, then the evi- dence of the functional relationship between the behavior and treatment is weakened. Decisions need to be made about whether more clinical effective- ness or experimental control is desirable.

Another disadvantage, reported by Yaden (1995), has to do with what is called resentful demoralization (Cook & Campbell, 1979). The term actu- ally refers to the attitudes of individuals who are subjects in control groups in group studies, who may resent not receiving a treatment in an experiment and thus react negatively. This concept could be extended to those involved in withdrawal designs as well. Essentially, it means that an individual’s behavior during subsequent baseline conditions might be negatively affected by resentment over having the treatment withdrawn. In addition, many tea- chers and clinicians are hesitant to withdraw effective intervention for the sake of experimental control. It might seem antithetical and illogical, from an educational or clinical point of view, to stop providing an effective treat- ment or technique purely to establish experimental control. Another, and perhaps the greatest, concern is of an ethical nature when using the

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 131

withdrawal design with dangerous behaviors, such as physical violence or self-abuse. Particularly for behaviors such as these, removal of a successful treatment would be highly questionable.

In summary, the major disadvantages of the withdrawal design have to do with the following situations:

• When the target behavior is not reversible; • When the treatment effects will continue after the treatment is withdrawn; • When it is not educationally or clinically desirable for the behavior to return to

baseline levels;

• When the target behavior is such that withdrawal of effective treatment would be unethical (e.g., dangerous behavior).

Adaptations of the Typical Withdrawal Design Because of the basic, almost simplistic, nature of the withdrawal design involving the A and B conditions, which clearly shows the effects of a single independent behavior on a single dependent variable, there are several adap- tations that are frequently used and reported in the professional literature. These could involve the repeated introduction and withdrawal of various treatments A-B-A-B-A-B, or the elimination of the initial baseline data phase by introducing the treatment phase first (B-A-B).

The B-A-B Design (No Initial Baseline) There are times when the collection of initial baseline data is either impossi- ble or inappropriate. As noted previously, if an investigator is working with an individual whose behavior results in physical harm or danger to self or others, then an A-B-A or A-B-A-B design should not be used due to ethical reasons. In this situation, the B-A-B design can be used to determine the effectiveness of an intervention on the target behavior. In this design, the experiment begins with the application of the independent variable (B phase). Once this intervention phase produces stable results at an accept- able (usually predetermined) level, the treatment is withdrawn. If the behav- ior reverts or moves toward the pre-intervention level in the absence of the treatment (A phase), the intervention is reintroduced to demonstrate the functional relationship between the behavior and the intervention. Consider the following example.

Mark is a 15-year-old boy with a severe intellectual disability. Although Mark has been described as being a quiet, gentle young man, he recently began engaging in a potentially serious behavior of trying to hit others, particularly when working in a one-to-one situation. This concern was

132 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

reinforced when he struck and hurt a much smaller student when he wouldn’t sit next to Mark during a movie shown in class. His special edu- cation teacher, Mrs. McDermott, was very concerned about safety and wanted to eliminate Mark’s hitting behavior. After a series of meetings with concerned individuals, including Mark’s parents, she sought assis- tance from the school’s behavior analyst, Ms. Meeks, to determine a method to control his dangerous behavior. After observing Mark, Ms. Meeks hypothesized that his hitting behaviors were his way of seeking physical attention. She developed a program using physical attention paired with praise as a reinforcer. Using differential reinforcement of other behavior (DRO) (see Chapter 2) Mrs. McDermott started the pro- gram. This involved touching his hand and saying ”Good hands, Mark” for every 1-minute interval in which Mark did not attempt to hit her during a 20-minute period of one-on-one instruction. This intervention phase (B) lasted for 15 sessions. During this period, Mark rapidly reduced his hit- ting attempts from 19 in the first session to an average of slightly more than one for the remaining 14 sessions, totally eliminating them for the last 5 sessions. During the withdrawal phase, for 3 sessions, Mrs. McDermott withdrew the attention during the sessions. During this time, Mark’s hitting behavior increased to 12, 14, and 14 attempts, moving more toward the pretreatment level that Mrs. McDermott recalled. She then reintro- duced the treatment (B) and noticed that Mark’s hitting behavior decreased to a similar level observed during the first treatment phase (see Figure 5-5).

4 C H E C K I T O U T # 4 Why do you think that the baseline condition (A) in the example of Mark only involved three sessions?

The B-A-B design was chosen over an A-B-A design for Mark because the nature of the target behavior (hitting) indicates that it is not one that should be “left alone” while collecting baseline data. Also, as indicated ear- lier, it is more appropriate to end the study with the intervention condition in effect, particularly given Mark’s behavior.

The disadvantage of the B-A-B design, from an experimental point of view, is the absence of pre-intervention baseline data so it is not possible to determine the effects of the intervention on Mark’s “typical” hitting behav- ior. In other words, the first intervention could have had an effect on the post-intervention baseline. That is one reason why the A-B-A-B design demonstrates the most experimental control of the three withdrawal designs discussed thus far—A-B-A, B-A-B, and A-B-A-B.

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 133

The A-B-A-B-A-B Design (Repeated Withdrawals) In this design, the treatment variable is introduced and withdrawn repeat- edly. For example, the A-B-A-B design previously used and described with Carla could be extended to include additional withdrawals and presenta- tions of the same treatment. These repeated withdrawals increase our con- fidence in the functional relationship between the treatment and the target behavior. Theoretically, the more times the treatment is applied and with- drawn (with the predicted pattern demonstrated), the more there is evi- dence of experimental control (repeated replication). For example, if there was a trend noted in an A-B-A-B design, but one that was somewhat ambiguous, more A and B conditions might be necessary to achieve a clearer demonstration of the treatment effects. Again, the issue of ethics and treatment efficacy must be considered. This means that a researcher must decide when a clear functional relationship has been established so that it is no longer necessary to continue the withdrawal phases. In other words, if experimental control is sufficiently demonstrated, is it desirable, ethical, and thus necessary to continue the additional withdrawal and treatment conditions?

Reversal A

Intervention B

15 14

17 18 19 20

13 12 11 10

9 8 7 6 5 4 3 2 1 0

1 2 3 4 5 6 7 8 10 Sessions

N um

b er

o f H

it ti

n g

A tt

em p

9 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

FIGURE 5-5 Mark’s data from a B-A-B Design

20 14

134 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

4 C H E C K I T O U T # 5 Looking at Carla’s data in Figure 5-4, do you think additional A and B conditions are necessary? Why or why not?

Key Concepts/Terms Basic goal—Demonstration of a functional relationship between the target

behavior and intervention by withdrawing (and reintroducing) the treatment.

Reversal design—Different from a withdrawal design;. In the reversal design there is an active attempt to reverse the effects of the treatment.

A and B designs—Nonexperimental designs in which only baseline data are recorded and described (A design) or when behavior is recorded and described during or after, but not before, an intervention is in place.

A-B design—Sometimes called the teaching design; baseline (A) data are col- lected and a treatment (B) is introduced; very little experimental control.

Action research—An inquiry-based approach of identifying a problem, gathering information, and developing an action plan.

A-B-A design—There is a return to baseline condition after the treatment phase; better demonstrates functional relationship between behavior and treatment; subject is left in baseline condition.

A-B-A-B design—Typical withdrawal design adds a second treatment phase; more evidence of a functional relationship is provided; subject is left in the treatment phase.

Prediction—After baseline data are stable, the prediction would be that the same or a similar pattern would emerge in other baseline conditions; conversely, a different pattern should emerge during treatment phases.

Verification—This occurs when treatment results in a change in the target behavior and the return to baseline condition results in a similar pattern as the previous baseline condition.

Replication—This occurs when the return to the treatment phase results in a similar pattern as the previous treatment phase.

Advantages of the withdrawal design—Easy to implement; a powerful design to demonstrate the functional relationship between target behav- ior and the treatment; has many possible variations.

Disadvantages of the withdrawal design—Often times the target behavior is not reversible; returning to baseline condition is not clinically or edu- cationally desirable; often times the treatment effects remain after it has been withdrawn.

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 135

Adaptations B-A-B—Used when collection of initial baseline data is either impossible or

inappropriate; subject is left in treatment condition. Repeated withdrawals—Involves the repeated application and withdrawal

of the treatment (e.g., A-B-A-B-A-B); increases evidence of functional relationship; additional baseline phases can raise ethical concerns.

4 Possible Answers to Check It Out

¶ You could be supportive and encouraging and acknowledge that herchild has made nice gains. However, you might want to mention that the program may or may not be the reason for the gains, i.e., her vocabulary could have increased because of developmental factors. There would be no need to discontinue or change the program, however, if the child was mak- ing adequate progress.

· Two possible suggestions would be to have Ms. Tracy try the programwith any other subject from whom she requires homework. She might also talk to the other concerned fifth-grade teachers and have them imple- ment it. However, this would still not demonstrate that it was the specific program that had an effect from an experimental point of view.

¸ Ms. Little apparently used a momentary time-sampling procedure.

! Baseline probably only involved three sessions because a) the targetbehavior was dangerous and no-treatment sessions should be kept at a minimum, and b) the data for the three sessions were somewhat stable with two of the three yielding the exact results.

º There probably would not be a need to include additional phasesbecause the number of verbalizations were very consistent across the two baseline and the two treatment conditions.

References Alberto, P. & Troutman, A. (2009). Applied behavior analysis for teachers (8th

ed.). Columbus, OH: Pearson: Merrill. Baer, D. M., Wolf, M. W., & Risley, T. R. (1968). Some current dimensions of

applied behavior analysis, Journal of Applied Behavior Analysis, 1, 91–97. Barlow, D. H., Hayes, S. C, & Nelson, R. O. (1984). The scientist practitioner:

Research and accountability in clinical and educational settings. New York: Pergamon Press.

Campbell, D. T. & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Chicago: Rand-McNally.

136 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Cook, T. D. & Campbell, D. T. (1979). Quasi-experimentation: Design and analy- sis issues for field settings. Chicago: Rand-McNally.

Cooper, J., Heron, T., & Heward, W. (2007). Applied behavior analysis (2nd ed.). Columbus, OH: Merrill.

Gast, D., & Hammond, D. (2010). Withdrawal and reversal designs (pp. 234–275). In D. Gast (Ed.) Single subject research methodology in behavioral sciences. NY: Routledge.

Hinchey, P. (2008). Action research primer. NY: Peter Lang. Mayer, G. R., Sulzer-Azaroff, B., & Wallace, W. (2112). Behavior analysis for last-

ing change (2nd ed.), Cornwall-on-Hudson, NY: Sloan Publishing. Mertler, C. (2009). Teachers as researchers in the classroom (2nd ed.). Thousand

Oaks, CA: Sage Publications. Mills, G. (2007). Action research: A guide for the teacher researcher (3rd ed.).

Upper Saddle River, NJ: Merrill/Prentice-Hall. Rubin, A. & Babbie, E. (2011). Research methods for social work. Belmont, CA:

Brooks/Cole. Yaden, D. B. (1995). Single subject experimental research: Applications for literacy.

In S. B. Neuman & S. McCormick (Eds.), Single subject experimental research: Applications for literacy (pp. 32–46). Newark, DE: International Reading Association.

CHAPTER 5 OVERVIEW OF WITHDRAWAL DESIGNS 137

CHAPTER

6 Application of Withdrawal Designs

139

F or each chapter focusing on different single subject designs, anapplication chapter is included. These chapters summarize the resultsof studies published in the professional literature that use each designated design. Users of this textbook are encouraged to read the original studies that are referenced to more fully understand the intricacies of the research designs. At the end of this chapter, a vignette is presented that allows the reader to apply the knowledge obtained from the overview of the withdrawal designs chapter and this application chapter. The vignette presents a sample study along with questions to be answered by the reader. The questions are the same as those used in summarizing studies from the professional literature presented in this chapter. Suggested answers are subsequently provided. This feature, called Application Practice, appears at the end of each application chapter.

In the previous chapter, we described the different types of withdrawal designs and their variations. Withdrawal designs are important single subject designs because they allow the investigator to demonstrate a strong functional relationship between treatment and changes in behavior (i.e., between the independent and dependent variables). In this chapter, we will provide specific examples from the professional literature for each of the basic types of withdrawal designs and an A-B-A-B design vignette for two students with autism.

The A-B Design Paterson, J., Hamilton, M. M., & Grant, H. (2000). The effectiveness of the

Hierarchic Dementia Scale in tailoring interventions to reduce problem beha- viors in people with Alzheimer’s disease. Australian Occupational Therapy Journal, 47, 134–140.

Purpose of the Study The authors designed this study to investigate the effectiveness of an individually-tailored intervention program to reduce targeted problem behavior in a person with Alzheimer’s disease (AD).

Subject A 79-year-old woman who was diagnosed with AD served as the subject. She excessively walked around her residential senior care facility (Alzheimer’s Unit) for lengthy periods with no apparent purpose. This excessive walking prevented her from fully participating in most unit activities and contributed to her weight loss and fatigue. Also, the unit staff observed her body lan- guage and facial expressions while walking and hypothesized that she was

140 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

experiencing distress. This further prompted concern among the care provi- ders to reduce her excessive walking. Other than having AD, she was in good health throughout the study and took only a mood stabilizer medication.

Setting The study was conducted in a secured Alzheimer’s Unit that provided care for 24 residents. This unit was composed of a large day room with a nursing station located in the middle. Bedrooms were situated off the day room on an adjacent corridor.

Dependent Variable The dependent variable identified in this study was excessive walking (defined as walking repeatedly with no apparent aim around the rest home unit in which she resided). Duration recording was used to record the number of minutes during 1-hour observation periods that the subject walked around the unit.

Independent Variable A tailored intervention program (independent variable) was prepared based on the participant’s past history, interests, and current situation and on results from the Hierarchic Dementia Scale (HDS). The HDS is a Piagetian-based instrument used to measure cognitive functioning in persons with dementia. There were several components of the intervention program including: (a) seating her at a table with another resident for morning and afternoon tea, (b) directing her to take a 20-minute walk outside the unit prior to lunch on Mondays and Fridays, (c) playing soothing music, using aroma therapy techniques, providing massage, encouraging her to use a rocking chair, and (d) accompanying her to the table at meal times and other activities.

The Design To investigate the effectiveness of the tailored intervention program, the authors used an A-B design. During 21 days of baseline (phase-A), the staff at the AD unit was instructed to carry out their normal routines with the subject. During 22 days of the intervention (B) period, the intervention program was implemented by all unit staff to reduce her excessive walking.

Intervention A tailored intervention program was designed by two occupational thera- pists that had extensive experience working with AD patients based on HDS results. A series of environmental adaptations and behavioral intervention

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 141

techniques were developed for implementation by unit staff following the baseline period. All staff at the unit participating in the implementation of the intervention program were trained. One of the authors regularly moni- tored the fidelity of program implementation by checking that the procedures outlined in the intervention program were consistently followed.

Obtaining the Data and Plotting the Results A research assistant was trained to collect the data. She observed the subject at the AD unit 3 days a week in the mornings and measured her duration of walking behavior with a stopwatch. Data were recorded on structured observation sheets. All unit staff participated in the intervention by modify- ing the unit environment and altering their interactions with the subject as prescribed in the intervention protocol. Data were plotted on a line graph to show the amount of time the participant spent on walking around the AD unit during the A and B conditions (See Figure 6-1).

Results During baseline (A), the mean length of time the participant spent walking around the unit was about 37 minutes. The time was reduced to a mean of about 14 minutes during the treatment (B) phase. This represented a decrease of 40% of the walking behavior after the intervention was introduced. Although fluctuations in time spent walking continued during phase B, the average time spent walking was lower than in phase A, tapering off over time. Also, the number of days the participant did not aimlessly walk at all increased from 2 in phase A to 5 in phase B.

Ti m

e (m

in )

0 1 4 7 10

Phase A Phase B

13 16 19 Day 22 25 28 31 34 37 40 43

FIGURE 6-1 Data from the subject’s amount of time spent

wandering in theA-Bstudy. Note. From The

effectiveness of the Hierarchic Dementia Scale in tailoring interventions to

reduce problem behaviors in people with Alzheimer’s

disease, Paterson, J., Hamilton, M. M., & Grant, H.

Therapy Journal. Reproduced with permission of John Wiley & Sons Ltd.

142 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Why Use an A-B Design? The goal of this study was to determine the clinical effectiveness of a multi- component intervention program developed from results of the HDS, the patient’s history, interests, and current situation. The researchers were primarily interested in changing the subject’s behavior (excessive walking) that prevented her from fully participating in unit activities and contributed to weight loss and fatigue. The researchers used an A-B design to demon- strate the effectiveness of the intervention and, more importantly, to produce a lasting reduction of her excessive walking behavior. Based on the intervention protocol, all staff participated in the intervention and modified the unit environment and altered their interactions with the partic- ipant. Repeated measures of the target behavior, using duration recording, continued throughout phases A and B. The researchers were unable to with- draw the intervention because all the unit staff were required to participate in the intervention phase. Because the staff had continued influence on the participant’s target behavior, it became impossible to reliably withdraw the intervention during a withdrawal phase. For this reason, the authors justified using the A-B design and not introducing a second baseline or intervention phase.

Limitations of the Study The A-B design used in this study is considered to be weaker than other withdrawal designs because there is no return to baseline to demonstrate a functional relationship between the independent and dependent variables. The researchers were unable to withdraw the intervention as required in an A-B-A or other withdrawal design because all the unit staff were involved in the intervention and they had undue influence on the participant’s target behavior. As a result, it was not feasible to stop the intervention to allow a withdrawal phase. Because of this limitation, one could argue that other factors may have influenced or contributed to the reduction in the partici- pant’s behavior other than the intervention strategies. However, readers should consider the multi-component intervention was successful in reduc- ing the participant’s excessive walking and the participant’s behavior did not reliably decrease until the introduction of the intervention. Because the intervention was specifically tailored for the subject, the possibility of gener- alizing the results to other individuals is limited (although the title of the article implies generalization). Also, it is impossible to determine the effectiveness of the individual components of the intervention since it was implemented as a package. It should also be noted that although stability of baseline data is recommended in most single subject designs, baseline data were not fully stabilized. Finally, because a measure of success was

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 143

achieved (an overall average of 40% decrease in the dependent variable), the reader should note why such a clinical study is important for identifying potential treatments and contributing to the knowledge base even when the functional relationship is not as clearly established as might be in a with- drawal design.

Summary A summary of the relevant dimensions of this A-B study can be found in Table 6-1.

The A-B-A Design Reichow, B., Barton, E. E., Good, L., & Wolery, M. (2009). Brief Report: Effects of

pressure vest usage on engagement and problem behaviors of a young child with developmental delays. Journal of Autism and Developmental Disorders, 39, 1218–1221.

Purpose of the Study The authors sought to examine the effects of wearing a pressure vest on engagement behaviors and problem behaviors of a child with develop- mental delays.

Subject The subject was a 57-month-old boy who was diagnosed with developmen- tal delays with impairments in cognition, language, and fine motor skills. He received speech therapy, occupational therapy, and special education from his preschool.

TABLE 6-1 Summary of “The

Effectiveness of the Hierarchic Dementia Scale in Tailoring Interventions

to Reduce Problem Behaviors in People with

Alzheimer’s Disease.”

FEATURE DESCRIPTION Type of design A-B design

Purpose of the study Determine the effectiveness of a tailored intervention program on the excessive walking of a person with Alzheimer’s disease

Subject A 79-year-old female who was diagnosed with Alzheimer’s disease

Setting A secured Alzheimer’s unit in a rest home

Dependent variable Excessive walking around the unit

Independent variable An individually tailored intervention program

Results and outcome The tailored intervention program reduced excessive walking © Cengage Learning 2014

144 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Setting The study took place in an inclusive preschool classroom at a university- affiliated childcare center. The classroom consisted of one teacher, an aide, and 12 students with and without disabilities. While the study was in progress, beginning with session 6, the subject’s classroom and another classroom with similar aged students merged to become one classroom. To accommodate the merger, the new classroom had two teachers, one teaching assistant, two teacher aides, and 17 students with and without disabilities. Since the merger occurred during the summer months, the attendance was erratic, and there were only 12-14 students in the classroom each day with 4-6 students per group.

Dependent Variables The dependent variables were the subject’s task engagement or non- engagement and problem behaviors (e.g., banging on tables, running away from adults, and having loud outbursts).

Independent Variable The use of the pressure vest was the independent variable. It was made of neo- prene and was designed to provide even amounts of pressure to the torso. Velcro straps that controlled the tightness of the vest were located over each shoulder from the back and wrapped around sides, and passed under the crotch area.

The Design To examine the relation between wearing the pressure vest and the subject’s target behaviors, an A-B-A withdrawal design was used.

The Intervention The intervention sessions occurred during a teacher-led morning art activity for two groups of six children each at small child-sized tables. During the intervention phase, a graduate student placed the pressure vest on the parti- cipant’s shoulders immediately before a table-time activity. The activities ranged from 10-15 minutes.

Obtaining the Data and Plotting the Results During the baseline condition, the participant did not wear the pressure vest for the table-time activity. This condition provided the authors an estimate of the child’s typical behavioral pattern during the activity. At the beginning of the intervention conditions, the pressure vest was placed on the child every time before the table-time activity began. Four categories of behavior were coded: (a) engagement, (b) non-engagement, (c) problem behavior, and

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 145

(d) unable to see child. Only one category was recorded per interval. To code problem behavior whenever it occurred regardless of engagement or non- engagement during the art activity, the researchers created a mutually exhaus- tive and exclusive coding system to ensure problem behavior was recorded even in the presence of engagement during an observation. Each session was videotaped and downloaded to an external hard drive. Behaviors were recorded using a 10-second momentary time sample during both baseline and intervention phases. A line graph was plotted to show the effect of the pressure vest on the child’s behavior during the table-time activity (see Figure 6-2).

Results During the initial baseline condition, the data for task engagement while at the art activity had a range of 44-87% of intervals scored with a slight decreasing trend in engagement. The data for problem behavior were more stable with a range of 0-21%, and there were no problem behaviors during three sessions. During intervention condition with the pressure vest on, task engagement increased variably with a range of 8-88%. Moreover, three of the four highest levels of engagement and six lowest levels of engagement were also observed during this period. Similarly, data for problem behavior for 11 of 14 sessions showed an increased level from the baseline condition including the three highest levels during the intervention phase. The second baseline condition lasted only for two sessions due to the end of the school year, and the data for engagement and problem behaviors were within the ranges of the treatment phase. This high variability of data made it difficult to conclude that there was a systematic change in task engagement or reduc- tion in problem behavior from the baseline condition to the pressure vest condition; thus, no apparent systematic difference was shown.

100

Baseline

Engagement

Problem Behavior

Classroom Merge

Pressure Vest Baseline

70 Pe

rc en

ta ge

o f I

nt er

va ls

0 0 1 2 3 4 5 6 7 8 9 10

Sessions 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

FIGURE 6-2 Use of pressure vest and engagement behavior data

in the A-B-A study. Note. From Springer and Journal of Developmental

Disorders, 39, 2009, pp. 1218–1221, “Brief

Report: Effects of Pressure Vest Usage on Engagement and Problem Behaviors of a

With kind permission from Springer Science and

Business Media.

146 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Why Use an A-B-A Design? The goal of an A-B-A design is to demonstrate a functional relationship between independent and dependent variables. This provides more experi- mental control than the A-B design. In this study, the investigators were interested in examining the relationship between wearing the pressure vest and the child’s behavior. The A-B-A design allows for the establishment of an extended baseline performance to better establish a functional relation- ship. The A-B-A design was also appropriate because data during the inter- vention phase did not suggest a therapeutic outcome on engagement and behavior, thus making a withdrawal phase desirable to determine if, in fact, engagement and problem behavior outcomes would improve with the withdrawal of the vest.

Limitations of the Study This study had several limitations. First, during the initial baseline, data for task engagement lacked stability. Second, the second baseline phase had only two sessions due to the end of the school year. Because of the shortness of this phase, the results must be interpreted with caution. Third, although the use of percent of intervals scored was an appropriate measure, the length and nature of the table-activity sessions could have affected the results. For example, it might be possible that the subject became less engaged and/or displayed more problem behaviors during long sessions. Another limitation was the inclusion of only one subject who had a history of pressure vest use before the study. Therefore, it is unclear what effect this history had and how these results would generalize to other individuals with developmental delays. Additionally, the authors did not address the possibil- ity of peer stigmatization or physical side effects due to the use of pressure vest. Although not technically a limitation, it should be pointed out that no functional relationship was noted. During the intervention phase, both the highest and lowest levels of task engagement were recorded and the data did not show a convincing trend either in increasing task engagement or improvement in the problem behavior. In fact, the problem behavior increased during this phase. This high variability does not support the use of the pressure vest to increase task engagement or decrease problem behaviors of this child. This study does demonstrate that systematic investi- gation is necessary to determine if an intervention is not effective or has equivocal effects.

Summary A summary of the relevant dimensions of this A-B-A study can be found in Table 6-2.

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 147

The A-B-A-B Design De Prey, R. L., & Sugai, G. (2002). The effect of active supervision and pre-

correction on minor behavioral incidents in a sixth grade general education classroom. Journal of Behavioral Education, 11(4), 255–267.

Purpose of the Study This study was designed to examine the effects of using active supervision, pre-correction, and daily data review on minor occurrences of behavioral incidents in a sixth-grade general education classroom.

Subjects Twenty-six general education students from a sixth-grade class participated in this study.

Setting This study was conducted in a rural elementary school in a Pacific North- west state that participated in school-wide implementation of effective behavioral support. One sixth-grade teacher with over 20 years experience volunteered to participate in this study due to high rates of minor problem behaviors during her social studies period.

Dependent Variables Minor behavioral incidents of the students and procedural integrity data related to the teacher’s use of an intervention package were the dependent

TABLE 6-2 Summary of “Effects of

Pressure Vest Usage on Engagement and Problem Behaviors of a Young Child with

Developmental Delays.”

FEATURE DESCRIPTION Type of design A-B-A design

Purpose of the study Examine the effects of wearing a pressure vest on task engagement and problem behaviors of a child with developmental delays

Subject One 57-month-old boy who had impairments in cognitive, language, and fine motor domains

Setting An inclusive university-affiliated preschool classroom

Dependent variables Task engagement, non-engagement, and problem behaviors

Independent variable Use of a pressure vest

Results and outcomes The pressure vest had no systematic effect on task engage- ment or non-engagement and problem behaviors. In fact, the problem behaviors increased during the pressure vest condition sessions compared to the initial baseline condition

148 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

variables. The minor behavioral incidents targeted were: (a) not being academically engaged, (b) eating in the classroom, (c) not following the teacher’s directions, (d) note passing, (e) getting out of seat, and (f) copying another person’s work. Additionally, data concerning the procedural integ- rity of the teacher’s use of the intervention package were recorded during the intervention phases.

Independent Variable The independent variable consisted of an intervention package with three major components: (a) active supervision, (b) pre-corrections, and (c) daily data review. Active supervision consisted of the teacher circulating around the classroom, scanning the classroom, interacting with students, and reinforcing demonstration of appropriate social and academic behaviors. Pre-correction involved the presentation of an instructional prompt prior to a situation in which problem behaviors were likely to occur. Together, these were referred to as “planned responding.”

The Design An A-B-A-B design consisting of repeated baseline and planned responding phases was used.

The Intervention The researchers trained the teacher to use active supervision and pre- correction instructional strategies (planned responding) for one 30-minute session in the teacher’s classroom prior to the first intervention phase. How- ever, no training was provided during the first baseline phase. For daily data review, each morning before students arrived in class, the researcher and teacher met in her class to talk about the data that were collected during the previous day on student and teacher behaviors. The researcher presented a graph that showed rates of minor problem behaviors in the classroom and taught her to analyze the data for trend, level, and variability. The teacher was informed that the data review was part of the study, and she would have an opportunity to review the data and adjust her teaching as necessary every morning during intervention phases. For the second baseline phase, the teacher and researcher did not meet to talk about the data. However, for the second intervention phase, the researcher and teacher met and fol- lowed the intervention protocol of reviewing the data daily.

Obtaining the Data and Plotting the Results During the study period, observers sat in the back of the classroom and recorded data for the students’ behavior and the teacher’s procedural

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 149

integrity in implementing the intervention. While the teacher was engaged in her typical teaching routine during the baseline phases, the observers systematically scanned the classroom using a partial interval recording pro- cedure (5 seconds for observation, 5 seconds for recording) for 45 minutes, recording students’ behavior only. Data were collected for the minor incidents of students’ behavior and the teacher’s use of planned responding during the intervention phases. During the baseline phases, only the students’ behavior was reported. The percentage of intervals in which prob- lem behavior occurred was calculated. The data were plotted on a line graph (see Figure 6-3) to display the effects of planned responding on the minor behavioral incidents.

Results During the first baseline phase, data patterns indicated high levels of minor behavioral incidents with a mean percentage of intervals scored at approxi- mately 96%. Data patterns for the initial intervention phase (B1) indicated an immediate change in level and a decreasing trend for minor behavioral incidents to about 62%. Although behavioral incidents decreased, the data in this phase were highly variable. During the second baseline phase, the data indicated a mean percentage of intervals scored at 72%, although behavioral incidents did not increase to the first baseline level. The reintro- duction of the intervention phase (B2) showed an immediate decrease of minor behavioral incidents with a mean percentage of intervals scored at 34.5%. The results suggest a functional relationship between the use of

Baseline Planned Responding

Planned Responding

Baseline100

0 1 2 3 4 5 6 7 8 10

Sessions

Pe rc

en t o

f I nt

er va

ls -M

in or

B eh

av io

9 11 12 13 14 15 16 17 18

FIGURE 6-3 Effect of active supervision and pre-correction data in

the A-B-A-B study. Note. From Springer and

Journal of Behavioral Education, 11, 2002,

pp. 255–267, “The Effect of Active Supervision and Pre-Correction on Minor Behavioral Incidents in a

Sixth Grade General Education Classroom,”

Springer Science and Business Media.

150 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

the planned responding and simultaneous decreases in minor behavioral incidents. Data on teacher behavior about active supervision and pre- corrections were collected across 25% of the sessions and the procedural integrity averaged 92% but is not shown on the graph.

Why Use an A-B-A-B Design? This design is superior to the A-B or A-B-A designs because it controls many threats to internal validity and allows the reader to make confident statements about the functional relationship. As explained in Chapter 5, the A-B-A-B design is perhaps the most powerful design to use if the target behavior is reversible. For classroom teachers or applied researchers, the most important ethical consideration is whether the removal of the treat- ment would harm the student or others. In this study, the reversibility of the minor behavioral incidents did not harm anyone, and they were subse- quently reduced with the teacher’s planned responding during the second intervention phase. Since the teacher used the same intervention twice, the issue of prediction, verification, and replication is strengthened and thus validated the functional relationship. Another positive reason to use this design is that it ends with the treatment phase in effect. Therefore, it is a desirable design for classroom or applied research because any educational or therapeutic effects should continue and be maintained over time.

Limitations of the Study There are several limitations to this study that were noted by the authors. First, this study was conducted at the end of the school year for a short period of time. As a result, one can question whether the reduction in students’ behavior incidents were directly related to only the planned responding or possibly related to the changing environment associated with the end of the school year activities. Second, although the intervention data trends were observed in the desired directions, the firmness and endurance of the effects on student and teacher behavior over several sessions were not established because the study ended suddenly. Third, even though minor behavioral incidents decreased during the first teacher planned responding phase, the data showed considerable variability. Finally, because the inter- vention was used as a package, the relative impact of one or more of the elements on student behavior cannot be determined. However, the A-B-A-B design was appropriate for the purpose of the study and despite the limita- tions, suggested that a functional relationship might exist, especially if the duration of the study could have been extended. Finally, the reader should note this study is an example where a group of participants ( 26 sixth-grade students) could be treated as a single case because the overall responding rate of the class served as the dependent variable.

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 151

Summary A summary of the relevant dimensions of this A-B-A-B study can be found in Table 6-3.

The B-A-B Design Robinson, P. W., Newby, T. J., & Granzell, S. L. (1981). A token system for a class

of underachieving hyperactive children. Journal of Applied Behavior Analysis. 14(3), 307–315.

Purpose of the Study In this study, the authors developed a token system to increase academic per- formance of 18 children with hyperactivity using reinforcement contingencies.

Subjects The subjects were 18 third-grade boys diagnosed with hyperactivity. Five students were on medication for hyperactivity. As a group, the students dem- onstrated little or no cooperative play. The reading abilities of these students ranged from low first-grade to third-grade level indicating varied abilities.

Setting This study was conducted in a newly formed special third-grade reading class (combined from three existing third-grade classes) in which 18 under- achieving children with hyperactivity were taught.

Dependent Variables The dependent variables were reading and vocabulary performance of students. Specifically these consisted of: (a) learning seven words of a unit

TABLE 6-3 Summary of “The Effect of

Active Supervision and Pre-correction of Minor

Behavioral Incidents in a 6th-Grade General Education Classroom.”

FEATURE DESCRIPTION Type of design A-B-A-B design

Purpose of the study Examine the use of active supervision, pre-corrections, and daily data review on minor behavioral incidents of students

Subjects Twenty-six general education students

Setting Sixth-grade general education classroom

Dependent variables Minor behavioral incidents and the teacher’s use of inter- vention package

Independent variables Active supervision, pre-corrections, and daily data review

Results and outcomes Planned responding resulted in decreased incidents of minor behaviors of the sixth-grade students

152 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

(learn to spell and define words), (b) helping teach a second student those same words, (c) learning to use the same words in sentences, and (d) teach- ing a second student to use those words in sentences.

Independent Variable The independent variable was the use of a token system. Four different color metal edged cardboard paper discs were used as tokens. There were four ways a student could get different colored tokens for complet- ing vocabulary assignments (a-d under the Dependent Variable heading). When all four colored tokens were earned, the student was allowed to play an electronic version of tennis on a pinball machine (Pong) for 15 minutes.

Design In this study, a B-A-B design was used to determine whether the token system was effective in increasing the academic behaviors of all students in the classroom. For the first (B) condition, the token system was instituted for 14 school days. It was removed for five days for the (A) condition. Then it was reinstated for 13 school days for the final (B) condition.

The Intervention The school district selected 220 vocabulary words that every student in the district was required to read and understand by the end of third grade. The classroom teacher divided these words into five levels or groups, with 40 to 45 words per level. The five levels were identified as preprimer, primer, level one, level two, and level three. The teacher divided each level down into 10 units (consisting of seven words per unit). To complete the reading and vocabulary assignments, each student would take out his file folder from the cabinet and go to the unit assignment box to take out the unit he was to learn. He would study the words with the help of a classmate who had already learned that unit, and then the student would go to the teacher to get quizzed. The teacher would give a green token if he could repeat and write the words. A feedback chart was posted on the class wall (histogram). The student would then check the wall chart to find out who had not yet learned that unit. Students’ names were listed on the abscissa. The ordinate dimension indicated how many reading assignments had been completed. The teacher would record in the student’s files and on the wall chart when each student completed his particular part of a unit. The wall chart allowed a student to find a classmate to proctor. Then the student would teach those same words to the selected classmate who had yet to learn them. If this classmate learned those words and passed the teacher’s quiz, the classmate

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 153

received a green token while the student who helped him learn the words received a yellow token. After learning the vocabulary words, the student picked out a packet containing cut-up sentences to be arranged and then again found a classmate to help. The classmate would then receive a red token if the classmate he helped could correctly arrange the sentences for the teacher. To receive a white token, the target student sat down with another classmate and helped him to arrange the newly learned sentences correctly. Following instruction by the teacher, the teacher would say “that’s right” or “that’s wrong” when the student arranged the words. When the target student earned all four colored tokens, he got to play Pong for 15 minutes.

Obtaining the Data and Plotting the Results The data collected from the class during the investigation were grouped together under two specific measures: (a) the number of assignments the class completed and (b) the number of vocabulary level tests mastered by the whole class. The vocabulary tests were conducted on a weekly basis for each student by a testing supervisor from the school district. However, these results were not plotted on the graph. When the baseline condition started, the students were told that the machines were not available to play. Once the B condition was reinstated, the machines were made available for the remainder of the study period. Data for all 18 students

Treatment with Tokens

(B)

Tokens withheld

(A)

Treatment with Tokens

(B)

0 0 5 10 15 20 25 30

Experimental Sessions

N um

b er

o f F

ul !

lle d

A ss

ig n

m en

FIGURE 6-4 Total number of completed

assignments data in the B-A-B study.

Note. From Robinson, P. W., Newby, T. J., &

Granzell, S. L. (1981). A token system for a class

of underachieving hyperactive children. Journal of Applied

Behavior Analysis, 14(3), 307–315. © Society for the Experimental

Analysis of Behavior, Inc. used by permission.

154 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

collected were plotted on a line graph (see Figure 6-4) to show the com- pleted assignments.

Results The token system produced substantial changes in class achievement in both measures. First, the class completed nine times as many assignments while working under the token system than when the token system was removed. As can be seen in Figure 6-4, the 18-member class completed an average of approximately 35 assignments per day during the 14 days in the first exper- imental stage (B). When tokens and back-up reinforcers (playing Pong) were withdrawn during the baseline phase (A), the average number of assignments completed decreased to about 4 per day for the entire class. Once the token system was reinstated (B) the average rate of assignments completed raised to 39.5 assignments per day.

Why Use B-A-B Design? The authors’ choice of a B-A-B design was largely due to restraints provided by the local school district and practical factors that existed in the class. The school district said the program’s success must be demonstrated in four weeks or permission for the program would be withdrawn. In addition to the district’s restrictions, the classroom teacher had great difficulty con- trolling her students’ misbehavior and frequently called in sick to escape from teaching the class. The principal had difficulty obtaining a substitute teacher to return a second time to teach the class. The teacher, in fact, informed the principal that she was going to resign because of the level of misbehavior in her class. The principal requested her to stay and promised her to seek professional help from the local university. Due to these over- whelming demands, the authors chose a B-A-B design so that immediate intervention could be introduced. With this design, the authors were able to increase the academic behaviors of all students and subsequently reduce the time available for fights or out-of-seat behaviors within the restricted time period.

Limitations of the Study The teacher’s overwhelming difficulty controlling the class coupled with the principal’s difficulty finding a substitute teacher led to the request for the local university’s professional help. However, the school district placed the following restrictions upon the request on any project: (a) whatever pro- gram was developed must not hinder the students’ being intermittently

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 155

removed from class by resource teachers, (b) weekly vocabulary testing required by the local school board could not be disrupted, and (c) no pro- gram could be implemented unless it could eventually be run by a single teacher. As a result, the investigators had no control over when the students would be removed, for how long or how many would be gone at any given time. In fact, at any given time, several students were removed from the classroom for vocabulary testing or to attend another class with a resource teacher. Because of these limitations, the teacher had only a small number of students in the class who worked for tokens during these times. A more typ- ical limitation of a B-A-B design included a lack of baseline data that did not establish an initial level of performance by which changes in the dependent variable could be compared. However, the baseline introduction (A) for five days did suggest that the dependent variable would decrease without the presence of the independent variable. Therefore, while the functional rela- tionship was not as clearly established as it might be, it is a stronger design than an A-B design because it allowed the researchers to demonstrate one phase of prediction, replication, and verification during the baseline and second intervention phases. Also, because of the restrictions of the school district, the results may be less generalizable. Nevertheless, the strength of the improvements in the dependent variable (down to 4 assignments per day on average during the baseline phase to 39.5 assignments per day on average in the second intervention phase) might suggest a robust effect from the independent variable that might be applicable with similar students or situations.

Summary A summary of the relevant dimensions of this B-A-B study can be found in Table 6-4.

TABLE 6-4 Summary of “A Token

System for a Class of Underachieving

Hyperactive Children.”

FEATURE DESCRIPTION Type of design B-A-B design

Purpose of the study To determine the effectiveness of a token system to increase academic performance of underachieving children with hyperactivity

Subjects Eighteen boys

Setting Third-grade reading class

Dependent variables Reading and vocabulary performance

Independent variable Token system with back-up reinforcers

Results and outcomes Token system increased the academic performance of all students and the teacher had a reinforcement system to control her students

156 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Application Practice: Brian and Brittany Brian is a 6-year-old boy who was diagnosed with autism shortly after his third birthday. He learned and used a few words for a few months but then stopped speaking altogether. At about the same time, he started to fre- quently rock back and forth often shaking his fingers in front of his face. He tended to cry a lot and, at times, it was very difficult to calm him down. Brian, who is now 6 years old, is currently in a first grade elementary autism cluster class with five other children with autism. His teacher, Ms. Ravel, has a highly structured classroom with a predictable schedule posted throughout the day. Brian is nonverbal and is prone to having temper tantrums and sometimes screams and hits his desk. Although some of the children in Ms. Ravel’s class use simple verbal communication, Brian and another student, Brittany, do not. Brittany is also 6 years old and was diagnosed with autism at age 28 months. Brittany frequently cries, has tantrums, hits her chest, and appears not to listen to Ms. Ravel’s directives and communication efforts. Currently, Brittany has no intelligi- ble speech.

A functional behavior assessment (FBA) was conducted to determine the function of Brian’s and Brittany’s tantruming behaviors. Based on the FBA, it was determined that they engaged in tantruming behaviors as a means of accessing desired items (e.g., certain toys, food). To improve functional communication for Brian and Brittany, the speech language pathologist (SLP), Mr. Sloane, decided to use the Picture Exchange Communication System (PECS). Both Brian and Brittany were able to mas- ter each of the six phases of PECS including constructing simple sentences and responding to questions using the pictures provided. During the PECS training, Ms. Ravel noticed that both students seemed to be calmer and quieter.

Unfortunately, even though the PECS materials were available to Brian and Brittany, they did not use them spontaneously and consistently, and they often reverted back to their tantruming behaviors. To help them improve their functional communication skills and decrease their tantrum- ing behaviors, Ms. Ravel decided to use MotivAider (a small timer that vibrates at set times) as a means of reminding them that PECS was available to indicate their requests. This 6-cm by 6-cm device (made by Behavioral Dynamics, Inc.) can be placed in the pocket or clipped onto a belt or waist band.

Ms. Ravel decided to develop a study to determine if the use of the MotivAider would result in more use of the PECS binder and result in a subsequent decrease in Brian’s and Brittany’s tantruming. During baseline (the first A phase), the PECS binder was available to Brian and Brittany during a 30-minute free-play period. They could use the binders to pick

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 157

a picture, make a spontaneous request, or make a sentence strip to help communicate their needs. A psychology graduate student from a local university who was doing her internship at the school collected data on the total number of functional communication requests during the 30- minute free-play period. She also recorded the amount of time they were engaged in tantruming behaviors (defined as screaming, shouting, crying, hitting self or objects such as the desk). This free-play period lasted for 3 consecutive days.

During intervention (the first B phase), Brian had the MotivAider clipped onto his belt and Brittany on her waist band during the 30-minute free-play period. The SLP and the teacher interacted with the whole class. The PECS binder was again made available for both of them during this period. Both were instructed at the beginning of the period to listen for the vibration from the MotivAider as a reminder to use the PECS materials to ask for any items they wanted/needed to play with, eat, or drink. The MotivAider was programmed to vibrate for 4 seconds every 5 minutes. If Brian or Brittany communicated using the binder, then they were given their requested item. Data were again collected on both the number of communication requests and the duration of tantruming. The data on both subjects revealed that the use of PECS increased dramati- cally and their tantruming behavior decreased sharply. This phase lasted 5 days.

During the second baseline (the second A phase), MotivAider was not available to Brian and Brittany. However, the PECS binder was available for them to communicate if they wanted. During this 3-day phase, the data reversed to near initial baseline levels.

To validate the functional relationship between the use of the MotivAider and the increase of communication requests and decrease in tantruming behaviors for Brian and Brittany, Ms. Ravel reintroduced the MotivAider and collected data (the second B phase). This also left both Brian and Brittany in a clinically appropriate situation in that the PECS binder was available, and its use was also encouraged. Both their communication increased and tantruming behaviors decreased to levels similar to the first intervention B phase level. Ms. Ravel gradually faded the use of MotivAider with the ultimate goal of increased spontaneous communication in the class- room. The results of this study are presented in Figures 6-5 and 6-6.

Brian and Brittany: The Questions After reading the description of Brian and Brittany and the study that Ms. Ravel implemented, answer the following questions. Then compare your answers to those provided in the answers section.

158 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

0 1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

N um

b er

o f P

EC S

R eq

ue st

Days

A MotivAider A MotivAider

Brian Brittany

FIGURE 6-5 Brian’s and Brittany’s use of PECS requests

0 1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

D ur

at io

n o

f T an

tr um

s (m

in ut

es )

Days

A MotivAider A MotivAider

Brian Brittany

FIGURE 6-6 Brian’s and Brittany’s

duration of tantruming behavior

20 14

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 159

What is the purpose of the study?

Who are the subjects?

What is the setting?

What are the dependent variables?

What are the independent variables?

What kind of intervention is provided to the subjects?

How were data collected and presented on graphs?

What were the results?

Why use an A-B-A-B design for this study?

What are the limitations of this study?

Brian and Brittany: The Answers

Purpose of the Study The classroom teacher designed this study to examine the use of the MotivAider to remind children with autism to use the PECS binder to communicate their wants and subsequently reduce tantruming behaviors.

Subjects Two first-grade children with autism participated in this study. Both were nonverbal and prone to have temper tantrums in the classroom.

Setting The study was conducted in a highly structured first-grade autism cluster classroom. This classroom had three other children with autism who had simple verbal communication skills.

Dependent Variables The dependent variables identified in this study were the use of the PECS binder to communicate and tantruming behaviors.

Independent Variable The MotivAider served as the independent variable. This device vibrated 4 seconds every 5 minutes to remind the participants that the PECS binder was available to indicate their requests. This intervention could also be described as using differential reinforcement of alternative or incompatible

160 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

behaviors as described in Chapter 2. The use of the PECS was an alternate or incompatible behavior with tantrums. The researcher may have chosen to use this behavioral terminology depending on whom she targeted to report her results.

The Design An A-B-A-B design was used to establish the functional relationship between the use of the MotivAider on increased communication and decreased tan- truming behaviors.

The Intervention To help Brian and Brittany increase functional communication skills and decrease tantruming behaviors, Ms. Ravel used the MotivAider to remind them that the PECS binder was available to indicate their requests for desired items. The MotivAider was programmed to vibrate 4 seconds every 5 minutes, and both participants were initially told that the vibration was a reminder to use PECS if they wanted to ask for any desired items.

Obtaining the Data and Plotting the Results During the baseline condition, the PECS binder was available for Brian and Brittany to communicate. The number of communicative attempts and the duration of tantruming behavior were both recorded. During intervention phases, the MotivAider was provided to remind them to use PECS to indicate their requests. Again, data were collected on the total number of functional communication requests during the 30-minute free-play period. Also, the observer used a stop watch to record the amount of time each was engaged in tantruming (duration recording). Data were plotted on a line graph to show how the MotivAider increased their communica- tion attempts and decreased the duration of tantruming behaviors (see Figures 6-5 and 6-6).

Results During the initial baseline phase, Brian and Brittany used the PECS binder only one time each to communicate. At the same time, Brian engaged in tan- truming behaviors that lasted an average of 7 minutes and Brittany approxi- mately 9 minutes during the three, 30-minute periods. When the MotivAider was introduced (the first intervention phase (B)), Brian’s use of the PECS binder to communicate increased to an average of 5.6 times per observation, and his tantruming behaviors decreased to slightly over 2 minutes. Brittany’s

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 161

use of the PECS binder increased to an average of 5 times and her tantruming behaviors decreased to 2.6 minutes during the 5-day intervention period. During the second baseline phase (A), the use of the PECS binder decreased to about 2 times each, and tantruming behaviors increased to 7 minutes for both Brian and Brittany. The reintroduction of the second intervention phase (B) showed an immediate increase in using PECS to 6.8 and 6 times, as well as a decrease in tantruming behaviors to .6 minutes and 1 minute for Brian and Brittany, respectively. The results suggest a strong functional rela- tionship between the use of the MotivAider on increased spontaneous com- munication requests and decreased tantruming behaviors. If one looks closely at Figures 6-5 and 6-6, it is clear that for both Brian and Brittany, there was an inverse relationship between PECS use and tantruming. In other words, as PECS use increased on a given day, the tantruming decreased, and vice versa. This provides further evidence of the communica- tive intent of the tantruming as found in the FBA. During photographing the Figures 6-5 and 6-6, apparently the two data lines merged into the single line. Readers should note that the data paths merge at points.

Why Use an A-B-A-B Design? The goal of an A-B-A-B design is to provide a strong demonstration that the introduction of the intervention causes changes in the behaviors of the par- ticipants. Using this design, Ms. Ravel was able to establish a functional relationship between the use of the MotivAider on Brian and Brittany’s increased communication requests and decreased tantruming behaviors. Withdrawal of the MotivAider in the second baseline phase and reintroduc- tion of the intervention provided more evidence of prediction, replication, and verification to further establish a causal relationship. Also, by ending the study in an intervention phase, Ms. Ravel left Brian and Brittany in an educationally appropriate situation.

Limitations of the Study This study has several limitations. First, even though the MotivAider helped Brian and Brittany communicate better and reduced their tan- truming behaviors, the teacher did not use this device in other settings to assess generalization. Second, the teacher will have to fade the use of MotivAider because it is not the goal to make Brian and Brittany depen- dent on the MotivAider to initiate communication, and it is unnatural for Brian and Brittany to wear this device on a daily basis. Third, the second baseline phase (A) lasted only 3 days, yet the tantruming beha- viors returned to initial baseline phase levels quickly. This suggests that they were highly prompt dependent and it would be difficult to fade the use of MotivAider.

162 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Summary A summary of the relevant dimensions of this vignette study can be found in Table 6-5.

Withdrawal designs may be used in a variety of situations. As we have noted, however, the ethics involved in withdrawing an intervention and the possibility of irreversibility of the intervention sometimes make their appli- cation problematic. For these and other reasons, additional designs have been devised. One example is the multiple baseline design, which is proba- bly the most commonly used design found in the literature. As the reader shall see in chapter 9, multiple baseline designs help overcome the problems associated with withdrawal designs.

TABLE 6-5 Summary of “The

MotivAider’s Use in Increased Communication Attempts and Decreased Tantruming Behaviors of Children with Autism.”

FEATURE DESCRIPTION Type of design A-B-A-B design

Purpose of the study To determine if the use of the MotivAider would result in more use of the PECS binder to communicate wants and subsequently decrease tantruming behavior

Subjects Two first-grade children diagnosed with autism

Setting First-grade autism cluster classroom

Dependent variables Number of PECS requests and duration of tantruming behavior

Independent variable Use of MotivAider

Results and outcome The MotivAider increased the PECS binder use and decreased tantruming behavior

CHAPTER 6 APPLICATION OF WITHDRAWAL DESIGNS 163

CHAPTER

7 Overview of Changing Conditions and Changing Criterion Designs

IMPORTANT CONCEPTS TO KNOW THE CHANGING CONDITIONS DESIGNS

The A-B-C Design The A-B-A-C (Multiple Treatment) Design Prediction, Verification, and Replication Advantages and Disadvantages of the Changing

Conditions Design

CHANGING CRITERION DESIGNS Issues Related to Changing Criterion Designs Prediction, Verification, and Replication Advantages of the Changing Criterion Design Disadvantages of the Changing Criterion Design

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

165

C hapters 5 and 6 focused on withdrawal designs or those thatinvolve the various manipulations of baseline (A) and a treatmentcondition (B). However, there will be times in both educational and clinical settings in which the independent variable (treatment) will not have the desired effect on the dependent variable (behavior). What happens then? Usually, the teacher or other change agent will either alter or change the treatment in some way or will introduce a new treatment. When either of these occurs a changing conditions design can be used to evaluate the effectiveness of more than one treatment, or even a combination of treatments. At other times, there will be situations in which a behavior that is in the individual’s repertoire needs to be gradually increased or decreased. The design used to evaluate the effectiveness of such a treatment is the changing criterion design. Both of these designs will be discussed in this chapter and application examples will be presented in Chapter 8.

The Changing Conditions Designs As an extension of the basic A-B design involving a baseline and a treat- ment, various changing conditions designs can be used to evaluate the effects of more than one treatment or a combination of treatments. Thus, the simplest changing conditions design, the A-B-C design, is similar to the A-B design with the added advantage of evaluating the effectiveness of more than one treatment. In this design, “A” represents baseline, “B” indicates Treatment 1, and “C” indicates Treatment 2. Any number of treatment con- ditions could be introduced (A-B-C-D or A-B-C-D-E) with the goal to end the design in an effective treatment condition. It should be noted that the hyphen between letters (B-C) indicates that the treatments are presented independently. When the letters are presented without a hyphen (BC), it indicates that the two treatments are presented in combination. In addition, withdrawal of treatments can be introduced to better evaluate the relation- ship of the independent and dependent variables. Changing condition designs that include the reintroduction of baseline conditions are frequently referred to as multiple treatment designs.

The A-B-C Design The A-B-C design is frequently used in education settings, when a teaching technique might be tried and not found to be effective, so that a different or a modified technique is subsequently introduced. For example, as an attempt to reduce Troy’s frequent cursing behavior (A), Mr. Barbaro sends “notes home” to inform his parents (B). After determining that this approach had no effect on the cursing, Mr. Barbaro introduces a different approach (C), in which he used tokens for periods of noncursing that

166 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

could be exchanged for extra free time. He found that this tactic was highly successful.

The A-B-C design also might be used to evaluate the additive effect of an “instructional package” (Alberto and Troutman, 2009). Consider the fol- lowing example:

Andre is a student with a moderate intellectual disability who is involved in a prevocational program. After collecting baseline data (the number of successful steps completed of folding a towel), his teacher, Ms. Landers, uses a videotape as a model to demonstrate the steps. After viewing the videotaped model for five sessions, no increase in towel folding was noted, so the teacher introduced verbal prompts in addition to the videotape. After only three sessions, the towel-folding performance increased.

Alternatively, using the previous example, the A-B-C design also could be used to evaluate the reductive effect of a teaching practice. For exam- ple, Andre might have been provided with both the videotape and the prompts first (B). Then the videotape could be faded so that only the prompts were in effect (C). The C condition, verbal prompts only, if found to be as effective or more effective than the B condition, should be used because it would more closely approximate teaching that typically occurs in the classroom.

As noted in Chapter 5, the A-B-C design could also be used in action research. Using this design, the first step would be to determine and implement an action plan. If this action plan was unsuccessful, then a modification of the plan could be introduced and subsequently evaluated. Thus, A ! baseline, B ! action plan, and C ! modified action plan. Further modification (A-B-C-D, etc.) could be included until the desired result is obtained.

4 C H E C K I T O U T # 1 In the instructional package described above, why couldn’t Ms. Landers assume that the verbal prompts alone were responsible for Andre’s increase in towel folding?

The A-B-C Design and Response to Intervention The Response to Intervention (RtI) model is widely used in general educa- tion and special education to help in the identification of students with learning and behavior problems. The RtI is a tiered approach in which students identified as having initial problems progress through stages of more intense intervention with the goal of eventually remediating those pro- blems and thus avoiding the need for a referral for special education as a student with a disability. Using the A-B-C design within the RtI model,

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 167

each tier (each with a new or additional intervention) can be thought of as a different letter. Consider the following example:

Stephen is a third-grade student who was screened as having difficulties in reading (A). His teacher, Ms. Carrington, provided a more intensive, evidence-based reading program for Stephen (B). After six weeks, adequate progress was not being obtained. At that point, the special education teacher, Mr. Dunn, consulted with Ms. Carrington and also worked indi- vidually with Stephen for 30 minutes a week (C). After two weeks, Stephen began to make satisfactory progress and the decision was made to keep him in the general-education classroom and not refer him for special education.

Limitations of the A-B-C Design It might seem logical that the A-B-C design could be used to see which treat- ment (B or C) is more effective. However, this is not possible for at least two reasons. First, the changes between B and C might be due to other factors (similar to an A-B design). For example, a teacher might conclude that tech- nique C is superior to technique B because a student performed much better in the C condition. However, the improvement in C might be due to any of a number of factors such as maturation or new medication. Second, there could be a sequencing or a cumulative effect. In other words, since the B technique has already been used prior to the C technique, the effect of the C technique might be influenced by B. In fact, the major problem with this design, again from an experimental point of view, is that there is very little evidence of the functional relationship between baseline (A) and any of the treatments (B, C, etc.). The way to address this concern is to reintroduce the baseline condi- tion(s), resulting in an A-B-A-C design or other multiple treatment designs.

The A-B-A-C (Multiple Treatment) Design One of the major advantages of changing conditions designs, in general, is their flexibility. At times, due to several reasons, the investigator may choose to, or have to, change the nature of the treatment, even after the experiment has started. As mentioned previously, in this design, the C phase may be an alteration of the original treatment introduced in the B phase or it may rep- resent a new treatment altogether. The difference between the A-B-C and the A-B-A-C design is that, in the latter case, another baseline condition is introduced between the treatments. Thus, the investigator collects baseline data (A), introduces the intervention variable (B), and then withdraws it as in an A-B-A design. Next the investigator introduces a second intervention (C) (or significant alteration of the first). The obvious advantage of this design over the A-B-C design is that the functional relationship can be better established because of the return to the baseline before the C condition.

168 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Extensions of the A-B-A-C Design There are an endless number of adaptations of this design. For example, the experimenter could evaluate the effects of three interventions (A-B-A-C-A-D) or even more. Note that although the potential for verification is present, the potential for replication is not because no specific treatment is evaluated more than once. It is also possible that the experimenter might combine treatments. For example, in an A-B-A-C-A-BC-A-BC design, A phases would be baseline, the B phase might be the use of prompts, the C phase the use of praise, and the BC phase a combination of prompts and praise. In these types of designs, it is important to be aware of which treatments or treatment combinations are being evaluated in isolation as well as the order in which they are pre- sented to avoid making false interpretations. Consider the following multiple treatment designs using the A, B, C, and BC conditions described above for Ben, a young adult who is involved in a psychiatric sheltered workshop with the goal of increasing his job production involving the insertion of a small microphone into cell phones. What conditions does each design actually eval- uate? This example only serves to demonstrate what functional relationships can and cannot be evaluated given different designs.

A-B-A-C: Can compare relationship of prompts to baseline and praise to baseline

A-B-A-B-C: Can compare relationship of prompts to baseline but not praise to baseline. The data in Figure 7-1 indicates that the effectiveness of the prompts was established. Although the praise also increased Ben’s job production, it is unclear if praise alone was responsible (possibility of an A-B-C sequencing effect).

A-B-A-BC-A-BC: Can compare relationship of prompts to baseline and prompts and praise to baseline but not praise to baseline

As mentioned, the various possible phases that can be evaluated in a mul- tiple treatment design are virtually limitless. However, the complexity and length of a study will be dependent on the number of treatments to evaluate and the degree of experimental control that is desired. It may be necessary to use several baselines and to evaluate each treatment or treatment combination as an isolated phase (Cooper, Heron, & Heward, 2007). These possibilities need to be considered when planning a research study.

4 C H E C K I T O U T # 2 Use the designations for A, B, and C noted above for Ben. Provide the design nota- tion for a study that would allow you to determine the relationship of prompts to baseline, praise to baseline, and the combination of prompts and praise to baseline.

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 169

Prediction, Verification, and Replication The prediction, verification, and replication of changing conditions designs, in general, depend on the specific combinations of treatments and whether and where baseline conditions are reintroduced. For example, in an A-B-C design, the prediction criterion can be met but the verifica- tion criterion can only partially be met. In other words, you can predict that the A condition will change as a result of the introduction of B and C, and verification is partially met when the pattern changes. However, the other prediction—that the pattern would remain the same given a return to baseline—cannot be verified. A design that better establishes this relationship is the A-B-A-C design; it adds the other element of potential verification (at least for the B condition) by returning to the baseline condition. The verification of other combinations depends on the return to baseline following a specific treatment condition. So, for an A-B-A-C-A-D design, verification for both the B and C conditions (but not D) could be established; replication occurs when the treatment condi- tion(s) are reintroduced. For instance, an A-B-A-C-A-C design would allow replication of treatment C. The extent of verification and replication that is desirable depends on the degree to which the researcher/clinician wants to demonstrate experimental control. Similar to withdrawal designs,

N um

b er

o f M

ic ro

p h

on es

In se

rt ed

Sessions

A B A B C

A -Baseline B -Prompts C -Praise

FIGURE 7-1 Example of data from a

changing conditions design

20 14

170 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

the decision also must be mediated by the practical and ethical issues associated with withdrawing an effective intervention to demonstrate a functional relationship.

Advantages and Disadvantages of the Changing Conditions Design The primary advantage of changing conditions designs is the ability to eval- uate the effectiveness of more than one treatment, a modification of a treat- ment, or a combination of treatments. A primary disadvantage of the A-B-C design, in particular, is the possibility of sequencing effects. That is, the effects of C could be influenced by B. The advantages and disadvantages of multiple treatment designs that reintroduce baseline conditions mirror those of the withdrawal design. Those are the positives of demonstrating a strong relationship between the behavior and the treatment(s) and the relative ease of implementation. The negatives are the issue of the irreversibility of either the behavior or the treatment. The first occurs when the behavior, once changed, is likely to remain even after the treatment is withdrawn. The sec- ond occurs when the effects of a treatment are maintained even after it is withdrawn. The sequencing effects noted in the A-B-C design is minimized by the reintroduction of the baseline between treatments but does not elimi- nate the possibility.

The following are the advantages of changing conditions designs:

• When the effects of more than one treatment or treatment combinations needs to be determined;

• For multiple treatment designs:

• When a clear functional relationship between the independent and depen- dent variables needs to be demonstrated;

• When the nature of the target behavior is such that it can be reversed when the treatment is withdrawn;

• When the nature of the treatment is such that its effects are not present on the target behavior after it is withdrawn;

• When withdrawal of treatment does not compromise ethics.

Conversely, the disadvantages of changing conditions designs are:

• Doesn’t take into account sequencing effects of more than one treatment (espe- cially in the A-B-C design);

• For multiple treatment designs:

• Inappropriate when the target behavior is not reversible;

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 171

• Inappropriate when the treatment effects will continue after the treatment is withdrawn;

• Inappropriate when it is not educationally or clinically desirable for the behavior to return to baseline levels;

• Inappropriate when the target behavior is such that withdrawal of effective treatment would be unethical (e.g., dangerous behavior).

Changing Criterion Designs The changing criterion design was first named by Hall (1971) and later described in greater detail by Hartmann and Hall (1976). A similar, unnamed design was described a decade earlier by Sidman (1960). The changing criterion design involves the evaluation of the effects of a treat- ment on the gradual, systematic increase or decrease of a single target behavior. This is accomplished by carefully changing, in a step-wise fashion, the criterion levels necessary to meet contingencies to increase behavior (e.g., positive or negative reinforcement) or to decrease behavior (e.g., differential reinforcement procedures or punishment). In other words, the effect of the intervention is demonstrated when the target behavior changes to the prede- termined criterion levels specified by the experimenter.

The changing criterion design also can be considered a variation of the A-B design discussed in Chapter 5. In the changing criterion design, after baseline data are collected (A), the treatment phase (B) is divided into subphases, with each subphase requiring changes in the target behavior that more closely approximate the terminal behavior or goal (Poling, Methot, & LeSage, 1995). Note that the terms phase and subphase are sometimes used interchangeably in the single subject literature when describing changing criterion designs. Because of the mechanics of the design, in which the different criterion levels are actu- ally subphases of the B phase, we will use the term subphase in this chapter.

The following steps outline the procedures to be taken when using a changing criterion design.

Step 1—Carefully define the target behavior. The behavior should be one that can be changed gradually and in a step-wise fashion.

Step 2—Collect baseline data. These data should be gathered until they are stable or moving in a counterproductive fashion (Hartmann & Hall, 1976).

Step 3—Determine level of performance (criterion levels) required for the contingent presentation of the reinforcement or punishment.

1. Determine terminal behavior or goal. For example, a teacher might want to increase a student’s typing rate to 50 words per minute or decrease talking-out behavior to zero.

172 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

2. Determine criterion level for the first subphase. To do this, Alberto and Troutman (2009) suggested the following options: a. Set the criterion level at the mean of the stable portion of the base-

line data, particularly if the initial rate of responding is low. For example, if a target behavior occurred 2, 0, 4, 0, 4, 2, and 2 times during baseline, the average would be 2. Therefore, 2 would be used as the initial criterion level.

b. Take 50% of the average of the baseline data and add to the aver- age. In the above example, the initial criterion level would be 3 (50% of 2 ! 1; 1 " 2 ! 3).

c. Choose the highest or lowest data point (depending on the goal) and use that as the criterion level. This is particularly useful with social behaviors.

d. Use professional judgment. For example, if the baseline rate is zero, there are no objective data to use as guidelines. In this situation, the experimenter must make the best estimate based on the available information about the subject.

3. Establish the criterion levels for the subsequent subphases. This usually involves a gradual increase or decrease in the criterion levels in the direction of the goal. There need to be at least two subphases, although three or more are generally used. The issues of the number of sub- phases to be used and the magnitude of the criterion changes are dis- cussed later in this chapter.

Step 4—Begin the intervention. The criterion required to obtain the treat- ment (i.e., contingencies to increase or decrease behavior) is applied.

Step 5—Introduce the next subphase level after the initial criterion level is met. One important point relates to the length of each subphase. In other words, how long must the subject respond at the criterion level in one subphase before mov- ing to the next subphase? Alberto and Troutman suggested that, at a minimum, the behavior should occur for two consecutive sessions or two out of three con- secutive sessions. It is important, however, to continue until a stable rate has been established because each subphase actually acts as a baseline for the subse- quent subphase (Hartmann & Hall, 1976). This strengthens the functional rela- tionship between the dependent and independent variables when a behavior consistently occurs at the criterion level (not above it or below it). The issue of the length of the subphases also will be discussed later in more depth.

Step 6—Continue through each subphase in a step-wise fashion until the terminal goal is reached.

The following is an example of the use of a changing criterion design.

Mr. Barkley was concerned that Danny, one of his third-grade students with Attention Deficit/Hyperactivity Disorder, was not completing his math worksheets during class. Typically, the students had about 15 minutes

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 173

to complete a mixed review worksheet at the end of the math lesson. Danny could do the problems correctly but wasn’t particularly interested in the task. He would frequently lose interest or attention, completed very few, and subsequently had to finish the worksheets for homework, which was quickly becoming a burden for both Danny and his parents. Mr. Barkley first developed several worksheets that were similar to those used in class. They included 20 addition, subtraction, multiplication, and division problems (5 each in a random order). He gave Danny a worksheet each day and allowed him 15 minutes to com- plete it. The same contingencies were in effect as before. In other words, if he didn’t finish the worksheet, he had to take it home to complete. The first week, he completed 6, 7, 5, 8, and 6 problems. On the basis of these baseline data, Mr. Barkley established an initial criterion level of 8 correct problems. Danny was told that if he completed 8 problems correctly, he would be given 10 minutes of free time at the end of the day to play video games. The contingency immediately had an effect. After stable responding was noted (defined as at least four sessions meet- ing criteria in a week), the next subphase, requiring 11 problems com- pleted, was introduced. This procedure was continued with the next two subphases with criterion levels set at 14 and 17 problems completed. The final subphase (the target goal of all 20 problems completed) was then put in effect. Figure 7-2 shows the data from this example.

1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

N um

b er

o f M

at h

P ro

b le

m s

C om

p le

te d

Sessions

FIGURE 7-2 Example of data from

a basic changing criterion design

19 99

174 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

4 C H E C K I T O U T # 3 What technique did Mr. Barkley use to determine the initial criterion level?

Issues Related to Changing Criterion Designs Hartmann and Hall (1976) noted that there were three very important issues that must be considered when using changing criterion designs: the length of each subphase, the magnitude of the criterion changes, and the number of subphases or criterion changes. They pointed out that these three issues are highly interdependent. Another issue that should be consid- ered is the placement of the subphases.

Length of Each Subphase As noted previously, in the changing criterion design, the level of responding of each subphase actually serves as a baseline for the subsequent subphase. It is therefore important that each subphase should continue until stable responding has occurred. The nature of the design, however, should generally allow this to occur quickly. In other words, the presentation of a new criterion level should result in an almost immediate change in the target behavior to that new level. This is particularly true for behaviors that can change rapidly. In the first example shown in Figure 7-2, Danny already had the target behavior in his repertoire. The issue was one of compliance, not his ability to do math. Thus, when the new criterion levels were set, the behavior could change rapidly. For the most part, if the behavior can change rapidly, the subphases can be shorter than if the behavior is slow to change. In that situation, the subphase may need to be longer to demonstrate experimental control (Cooper et.al., 2007).

The evidence for the functional relationship between the independent and dependent variables is strengthened when the behavior changes to exactly the criterion level and stays at that rate until the criterion level changes in the next subphase. For this reason, the actual lengths of the sub- phase should vary to demonstrate that control. More evidence for the effec- tiveness of the video game reinforcer for Danny could have been made by varying the number of sessions within each subphase. Figure 7-3 indicates that when the length of the subphases varied from three to six sessions, Danny’s math performance changed to the criterion level and remained rel- atively stable within each subphase, regardless of its length. This stability demonstrates greater internal validity.

Magnitude of Criterion Changes This issue focuses on the question “How much change in the target behavior is required before the subject receives the contingency (intervention)?” This is a very important question—if the required change is small, the subject

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 175

might progress, but it would be difficult to determine if the change was not due to other factors such as maturation or practice effects. If the required change is too large, however, there are at least two possible problems. First, because the target goal will be reached in fewer subphases, there might not be enough subphases to demonstrate experimental control in the study. Second, requiring drastic changes might contradict good instructional practices (Cooper et al., 2007). In our example of Danny, if the criterion levels had been 10, 15, and 20 (and the behavior changed to those levels), the effects would look impressive because the behavior changed so dramati- cally. However, the fact that the behavior was controlled in only three sub- phases compromises the validity of the study. One logical method of determining the criterion changes is to use the baseline data as an indicator. In general, smaller criterion changes should be used for more stable beha- viors whereas larger changes would be necessary to demonstrate control for behaviors that are more variable (Hartmann & Hall, 1976).

4 C H E C K I T O U T # 4 The magnitude of the criterion change for Danny was three (8, 11, 14, 17, and 20 problems). Suppose Mr. Barkley had chosen a magnitude of four (9, 13, 17, and 20). What do you think would be an advantage and disadvantage of that change?

1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

N um

b er

o f M

at h

P ro

b le

m s

C om

p le

te d

Sessions

FIGURE 7-3 Example of data from a changing criterion design with different

subphase lengths

19 99

176 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Number of Criterion Changes The number of criterion changes actually refers to the number of subphases that should be included in the study. The determination will depend on both the length of the subphases and the magnitude of the criterion changes. This demon- strates the interrelationship of these issues. For example, if there is a limited amount of time available for a study (e.g., only a month left before intense prep- aration for the State Assessment Test), then the fewer the number of subphases there can be. Also, as previously discussed, the greater the magnitude of the cri- terion changes, the fewer the number of subphases before the target goal is met.

In general, the more times the target behavior changes to meet a new criterion, the more control is demonstrated. The researcher should be aware of artificial floors and ceilings, however. This could give the faulty impression that more experimental control is present than there might be. Cooper et al. (2007) described this concern.

An obvious mistake of this sort would be to give a student only five math problems to complete when the criterion for reinforcement is five. Although the student could complete fewer than five problems, the pos- sibility of exceeding the criterion has been eliminated, resulting perhaps in an impressive-looking graph, but one that is badly affected by poor experimental procedure. (p. 222)

In other words, if the student only completed five problems (criterion level) when more than five were possible, then more experimental control would be demonstrated.

Placement of the Subphases As we have discussed, the changing criterion design is used to systematically increase or decrease a target behavior toward a terminal goal. Experimental control is dependent, in part, on the variation of the number and length of the subphases and the magnitude of the criterion levels. However, if the direction of the criterion changes is always in the same direction, it is more difficult to demonstrate that the changes “are not naturally occurring due to either his- torical, maturational or measurement factors” (Hartmann & Hall, 1976; p. 530). A very powerful addition to the basic changing criterion design is to include a subphase or subphases in which there is either a) a reversal to a previous criterion level or b) a return to baseline level. These techniques, although not a requirement of the basic design, is sometimes used to help demonstrate that the treatment was responsible for changes in behavior.

Again, consider the previous example of Danny. The criterion levels for both Figures 7-2 and 7-3 were 8, 11, 14, 17, and 20. Suppose, however, that the criterion levels had been 8, 11, 14, 17, 14 ,17, and 20. By showing that the level of behavior changed only to the specified level (i.e., increased to 17 in Subphase 4 and then decreased back to 14 in Subphase 5), an even

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 177

stronger relationship can be demonstrated than when a linear trend is in effect (see Figure 7-4). This approach also delays the time it takes for Danny to meet the goal of completing 20 math problems, however. Note that if additional reversal subphases were added (e.g., 8, 11, 14, 11, 14, 17, 14, 17, and 20), additional experimental control would be demonstrated but reaching the goal would be delayed further.

Although used rarely, a changing criterion design with a return to base- line conditions is possible. This is a more radical reversal of conditions than just returning to a previous criterion level. In a sense, it is similar to the A-B-A or A-B-A-B designs discussed in Chapter 5. In this case, the B sub- phase is the step-wise criterion changes required for the presentation of the contingencies. This design does show a powerful functional relationship between the independent and dependent variables similar to a withdrawal design. However, it also has the ethical liabilities of the withdrawal design because the subject is placed, at least temporarily, in a nontreatment sub- phase. A study should never end in a baseline condition. The following is an example of a changing criterion design with a return to baseline.

Barbi, a fourth-grade student with Down syndrome, frequently uses inappropriate verbalizations during instructional time. This consists of calling out short phrases that are typically not related to the information presented. Even when her verbalizations are appropriate, they typically disrupt the class. Her special education teacher, Mr. Warde, decided to

0 1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

N um

b er

o f M

at h

P ro

b le

m s

C om

p le

te d

Sessions

FIGURE 7-4 Example of data from a changing criterion design

with a reversal to a previous criterion level

19 99

178 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

develop a program in which Barbi’s inappropriate verbalizations would be ignored and she would be provided with immediate reinforcement if she decreased them to specified levels. Barbi loves Disney characters, so the reinforcement chosen was three Disney stickers that she could place into her sticker book. The teacher collected baseline data during a 20-minute period at the beginning of each day when announcements were made and the events of the day were planned. The baseline data were relatively consistent; she averaged 15 verbalizations per 20-minute period (17, 13, 15, 14). Barbi was told that she must initially reduce her verbalizations to 12 and she would receive the three stickers at the end of the 20-minute period. The teacher also gave her the following verbal prompt when she would lose the reinforcer if she made another verbaliza- tion: “Barbi, if you speak out of turn one more time, you will not get your stickers today.” After the second day, she was meeting the criterion level. The next levels that were specified were 8, 4, and 0. Figure 7-5 shows the data indicating that Barbi was able to successfully decrease her inappro- priate verbalizations. To determine the effectiveness of the reinforcement even more, Barbi was then told that there would be no stickers available. This return to baseline conditions resulted in a sharp increase in her ver- balizations, although not quite to baseline level. For obvious reasons, the criterion level conditions and the terminal goal were then reinstated. This reduced her verbalizations almost immediately.

A2A1

0 1 2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

N um

b er

o f I

n ap

p ro

p ri

at e

V er

b al

iz at

io n

Sessions

FIGURE 7-5 Example of data from a

changing criterion design with a return to

baseline phase

19 99

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 179

4 C H E C K I T O U T # 5 What would be the advantage for Mr. Warde of the return to baseline criterion subphase for Barbi?

Prediction, Verification, and Replication Although prediction and replication are easily addressed within a changing criterion design, verification is somewhat more difficult to demonstrate (Cooper et al., 2007). In general, prediction of the levels of future behaviors is made when stable responding is attained within each subphase. Verifica- tion is possible when either of two of the previously discussed suggestions to increase internal validity is made. By varying the lengths of the subphases, verification of the treatment effects is made. Similarly, and perhaps more convincingly, verification is demonstrated when the direction of the criterion levels is reversed and the behavior returns to a previously set criterion level. Replication occurs every time that the behavior changes in the predicted direction based on the predetermined criterion levels.

Advantages of the Changing Criterion Design Used appropriately, the changing criterion design can be an effective tool for the practitioner or researcher. It is particularly helpful when working with target behaviors that can increase or decrease in a step-wise fashion. The behavior should already be in the subject’s repertoire and be measurable using a number of recording procedures, including frequency, accuracy, duration, or latency. Examples of behaviors that are ideally suited for the changing criterion design were offered by Hartmann and Hall (1976). These include increases in writing or reading rate, improvements in peer relationships, as well as decreases in smoking, overeating, and latency of compliance behaviors. The changing criterion design is particularly helpful when determining the effects of contingency programs specifically designed to increase or decrease behaviors.

One somewhat controversial use of the design is its evaluation of the effectiveness of shaping procedures. Its use for this purpose has been recom- mended by some (e.g., Alberto & Troutman, 2009; Hartmann & Hall, 1976) but questioned by others (e.g., Cooper et al., 2007). The argument against its use has to do with terminology related to the nature of shaping procedures. Shaping involves the development of a new behavior by reinfor- cing successive approximations toward that behavior. In other words, a dif- ferent form of behavior is required at each subphase instead of different levels of the same behavior. This violates the assumption that a behavior is already in an individual’s repertoire to use the changing criterion design.

180 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Alberto and Troutman, however, discussed the issue of shaping fluency, rate, or speed of a behavior. For example, steadily improving a student’s performance on a timed test administered over and over would be an example of shaping the fluency of a behavior. In this situation, the behavior is already in the student’s repertoire so that a changing criterion design would be appro- priate. Regardless of the terminology used, the changing criterion design should focus on the systematic increase or decrease of a single behavior.

Two other advantages are worthy of note. First, it can be helpful to use when the terminal goal that is set takes a relatively long time to reach (Alberto & Troutman, 2009). Thus, the gradual movement toward the goal results in educationally or clinically desirable outcomes while at the same time demonstrating experimental control. A second, related advantage is that the treatment does not have to be withdrawn to show its requisite functional relationship with the target behavior. In fact, after baseline, treat- ment conditions can stay in effect so that the subject is always moving toward the goal. However, if the researcher wants to demonstrate greater experimental control, the direction of the criterion changes can be reversed or baseline conditions could be reintroduced within the study.

The changing criterion design would be appropriate to use in the follow- ing instances:

• When the target behavior can change gradually in a step-wise fashion; • When the behavior is already in the subject’s repertoire and needs to be

increased or decreased;

• When the effects of contingent reinforcement or punishment procedures need to be evaluated;

• When withdrawal of treatment is not appropriate.

Disadvantages of the Changing Criterion Design The same characteristics of the changing criterion design that make it partic- ularly useful for some behaviors and treatments also limit its use with other behaviors and treatments. For instance, it is not appropriate unless the behavior can be changed in a gradual, step-wise fashion. Also, as noted ear- lier, the target behavior should already be in the subject’s repertoire, so the goal should be to increase or decrease a behavior, not develop a new one.

A related disadvantage is that the behavior should change only to the specified criterion level in order to demonstrate the greatest experimental control. However, this might not always be educationally or clinically desir- able. If the terminal goal can be met faster, then is it appropriate to “hold back” the subject for research purposes? A final disadvantage relates to the planning and implementation of the changing criterion design. As discussed,

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 181

considerable thought must be given to determine the number and length of subphases and the magnitude of the criterion changes. For example, if the criterion changes are set too high, the subject might be adversely affected because the contingency is applied infrequently or not at all. Conversely, if the levels are set too low, the subject’s optimal responding may be slowed. Similarly, a problem might exist when subphases are short and similar in length. This might result in a linear trend that could also be explained by factors such as maturation or practice effects. This is why varying the length, number, and direction of the subphases as well as the magnitude of the criterion levels is important to demonstrate experimental control.

It is not appropriate to use the changing criterion design in the following situations:

• When the target behavior cannot be changed in a gradual, step-wise fashion; • When the target behavior is not in the subject’s repetoire; • When treatments other than the presentation of contingencies are being

evaluated;

• When time and effort cannot be given to determine and manipulate the important parameters of the design.

Key Concepts/Terms Changing conditions design—Allows the evaluation of more than one

treatment (e.g., A-B-C); evidence for functional relationship is weak. A-B-C design—Has many practical applications; can be used in action

research, the RtI model, and evaluating additive and reductive instruc- tional packages.

A-B-A-C (Multiple treatment) design—Allows the evaluation of more than one treatment with reintroduction of baseline condition(s); more experi- mental control than changing conditions; many variations possible.

Prediction—Occurs when you predict that the pattern of behavior will change with the introduction of different conditions.

Verification—Is only partially met by an A-B-C design; in an A-B-A-C design, verification of B is determined by return to baseline levels.

Replication—Occurs when the treatment condition(s) are reintroduced and the behavior returns to previous treatment levels.

Advantages and Disadvantages—Advantage of changing conditions design is the evaluation of more than one treatment; disadvantage is the possi- bility of sequence effects. Advantages and disadvantages of multiple treatments design similar to those of withdrawal designs.

182 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Changing criterion design—Evaluate effects of a treatment on the gradual increase or decrease of a single target behavior; frequently used when reinforcement or punishment contingencies are in effect. Experimental control is demonstrated when the target behavior changes to each new criterion level to meet the contingency in effect.

Length of each subphase—Subphase should be long enough to allow sta- ble responding; length of subphase should be varied if possible.

Magnitude of the criterion changes—Should not be too large or too small; use baseline data to help make the determination.

Number of criterion changes—Related to the length of the subphases and the magnitude of the criterion changes (e.g., the greater the magnitude of the criterion changes, the fewer the number of subphases before the tar- get goal is met).

Placement of subphases—Instead of having subphases that increase or decrease in a linear fashion, a reversal to a previous criterion level or to baseline level is included.

Prediction—Occurs when stable responding within each subphase predicts the behavioral levels of subsequent subphases.

Verification—Accomplished by varying the lengths of the subphases and/or by changing the direction of the criterion levels and showing changes in the direction of the target behavior.

Replication—Occurs when the behavior changes to the predetermined crite- rion levels.

Advantages—Withdrawal of treatment not necessary; good design to eval- uate contingency programs to increase or decrease behavior; gradual change in target behavior results in educationally or clinically appropri- ate outcomes.

Disadvantages—Target behavior must be able to change in a gradual, step- wise fashion; requires time and effort to determine important parameters of the design (e.g., the number and lengths of subphases); not appropri- ate for treatment approaches that do not use contingent procedures to increase or decrease behaviors.

4 Possible Answers to Check It Out

¶ In this A-B-C design, the verbal prompts were the C condition andwere presented after the B condition (use of the videotape). Therefore, a sequencing effect is possible.

· The correct notation would be A-B-A-C-A-BC.

¸ Mr. Barkley determined the initial criterion level for Danny by choos-ing his highest data point during baseline.

CHAPTER 7 OVERVIEW OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGNS 183

! An advantage of increasing the magnitude of the criterion level is thatDanny could reach his goal in fewer subphases, A disadvantage is that since Danny did not have the opportunity to meet the criterion level in more subphases, experimental control is compromised.

º The advantage of returning to baseline would be that Mr. Wardewould be more confident that the contingency was effective and might be used for other of Barbi’s behaviors. Also note that baseline was only three sessions, and her behavior returned to the previous criterion level very rapidly.

References Alberto, P., & Troutman, A. (2009). Applied behavior analysis for teachers

(6th ed.). Columbus, OH: Merrill. Cooper, J., Heron, T., & Heward, T. (2007). Applied behavior analysis (2nd ed.).

Columbus, OH: Merrill. Hall, R. V. (1971). Managing behavior: Behavior modification and the

measurement of behavior. Lawrence, KS: H & H Enterprises. Hartmann, D., & Hall, R. V. (1976). The changing criterion design, Journal

of Applied Behavior Analysis, 9, 527–532. Poling, A., Methot, L., & LeSage, M. (1995). Fundamentals of behavior analytic

research. New York: Plenum Press. Sidman, M. (1960). Tactics of scientific research. New York: Basic Books.

184 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

CHAPTER

8 Application of Changing Conditions and Changing Criterion Design

185

In chapter 7 we described the changing conditions and changing criteriondesigns and several of their adaptations. The changing conditions designis generally used when the independent variable fails to produce the desired effect on the dependent variable. The changing criterion design is used when evaluating the effects of an intervention on the gradual systematic increase or decrease of a target behavior. In this chapter we provide examples for both designs and provide an application practice and suggested answers for each design.

The A-B-A-C Design Handen, B. L., Parrish, J. M., McClung, T. J., Kerwin, M. E., & Evans, L. D. (1992).

Using guided compliance versus time out to promote child compliance: A prelim- inary comparative analysis in an analogue context. Research in Developmental Disabilities, 13, 157–170.

Purpose of the Study The purpose of this study was to determine whether guided compliance was a better procedure for promoting child adherence to adult requests than a traditional time out procedure.

Subjects The subjects were five children (four males, one female) ranging in age from 3-years 8-months to 6-years 4-months. All the subjects were classified as having mild developmental disabilities. Three of the subjects were functioning in the mild range of mental retardation; other disabilities reported were language delays and early signs of nonspecific learning disabilities.

Setting The study was conducted in a small individual treatment room furnished with a one-way mirror, a worktable, and two chairs. For each session, the subjects were brought into this room, given requests, and allowed to play with books, blocks, puzzles, cars, and trucks without restrictions.

Dependent Variable The dependent variable under investigation was child compliance behavior to adult requests as measured through the percentage of correct responses. The authors defined compliance as “satisfactory completion of a requested task

186 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

within 10 s of the request” (p. 160)*. Examples of the compliance requests are “Give me the ________,” and “Put the ________ on the ________.”

Independent Variables The independent variables were guided compliance (B), a technique that incorporates physical guidance, and time out (C). Subjects were reinforced with praise for compliance during both intervention conditions.

The Design The impact of guided compliance and time out on all subjects’ compliance responses was evaluated using an A-B-A-C design, with the order of the two treatments counterbalanced across subjects.

The Intervention During all experimental sessions, the experimenter established eye contact, called the subject by name, and then issued a request. During the guided compliance phase, if the subject responded to a request within 10-seconds, verbal praise was provided and the child was allowed to play until the next request at the beginning of the subsequent minute. If noncompliance resulted, praise was withheld and the subject was physically guided to com- plete the task. During the time out phase, the subject was also praised and allowed to play until the next request if he or she responded to a request within 10-seconds. If noncompliance resulted, the subject was placed in a chair facing a corner in the room for 30-seconds (if the subject refused to stay in the seat, then the subject was held gently in the chair). Subsequent to this time out, the next request was initiated. Also, at the conclusion of training phases, five generalization requests were given in 1-minute intervals.

Obtaining the Data and Plotting the Results The observer recorded each compliance trial as correct (compliant) or incorrect (noncompliant). Data were collected on one subject at a time. Ten requests were presented at a rate of about one per minute; the subjects were instructed to respond while the observer collected data on each response. Data for each subject were placed on x-y graphs to determine the effectiveness of the intervention procedures in increasing compliance. For the purpose of visual inspection, as a representative sample, data for subject 1 are shown in Figure 8-1.

*Handen, B. L., Parrish, J. M., McClung, T. J., Kerwin, M. E., & Evans, L. D. (1992). Using guided compliance versus time out to promote child compliance: A preliminary comparative analysis in an analogue context. Research in Developmental Disabilities, 13, 157–170

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 187

Results During baseline the overall mean percentage (M) of compliance (i.e., within a phase) was below 41% for all subjects. During time out the mean percentage of compliance was 85% (range 71–92%), and during guided compliance the mean percentage of compliance was 59%. Time out was more effective than guided practice for all subjects. Figure 8-1 shows the mean percentage of compliance recorded by the observer across conditions for Subject 1. As can be seen, Subject 1 exhibited no change in compliance rates over baseline (M ! 47%) with guided compliance (M ! 48%). However, following a return to baseline, presentation of time out let to a considerable increase in compli- ance rate (M ! 71%). This graph depicts why overall means and ranges can be useful in evaluating results when data are highly variable.

Why Use an A-B-A-C Design? Clearly, the researchers were interested in determining the effects of two behavior management procedures to increase compliance behavior of chil- dren with developmental disabilities. This design allowed the investigators to test more than one independent variable after each baseline phase. Also,

Subject 1

Baseline Guided

Compliance Baseline Time Out Baseline

100

0 5 10 15 20 25 30 35

% C

om p

lia n

Sessions

GeneralizationTraining

FIGURE 8-1 Data from Subject 1 in the

A-B-A-C study. Note. Reprinted from Research in Developmental Disabilities, 13, B. L. Handen, J. M.

Parrish, T. J. McClung, M. E. Kerwin, and L. D. Evans, “Using Guided Compliance Versus Timeout to Promote

Child Compliance: A Preliminary Comparative Analysis in an Analogue

from Elsevier.

188 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

by using more than one subject, the researchers were able to counterbalance the presentation of the two procedures to help avoid sequencing effects.

Limitations of the Study As stated earlier, Subject 1 made no change in compliance rates during baseline and guided compliance phases. As the authors described, “guided compliance may have served on occasion to maintain noncompliance among these children” (pp. 167–168)*. Also, the baseline data for each subject was highly unstable, indicating the variability of compliant behavior of each subject in general.

Summary A summary of the relevant dimensions of this A-B-A-C study can be found in Table 8-1.

Application Practice: Kumar Ms. Rhoten was recently hired to teach literacy skills at an elementary school. When she arrived at the school, Ms. Rhoten was approached by

TABLE 8-1 Summary of “Using Guided Compliance Versus Time Out to Promote Child

Compliance: A Preliminary Comparative Analysis in an Analogue Context.”

FEATURE DESCRIPTION

Type of Design A-B-A-C design

Purpose of the study Determine the effectiveness of guided compliance and a time out procedure to promote child adherence to adult requests

Subjects Four boys and one girl ranging in age from 3-years 8-months to 6-years 4-months who had developmental disabilities

Setting A treatment room (2.6 m X 2.9 m) in an outpatient clinic; it had a one-way mirror, a worktable, and two chairs

Dependent variable Compliance behavior (satisfactory completion of a request within 10-seconds) as measured through percentage of correct responses

Independent variables Guided compliance and time out procedures

Results and outcomes The time out procedure produced more compliance by all subjects than the guided compliance procedure

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 189

the third-grade special education and inclusion teacher, Ms. Cheng. She was concerned about Kumar, a student with specific learning disabilities in the area of spelling who was in Ms. Rhoten’s literacy class. Kumar has difficulty spelling words correctly. Kumar also gets frustrated during spelling assessments as he takes considerable time writing his responses, although he has demonstrated he can write all letters accurately.

Ms. Rhoten decided to use a systematic approach to help Kumar improve his spelling accuracy in an action research study. She reviewed Kumar’s last five weekly spelling assessments in which Kumar and his classmates were expected to spell 20 third-grade vocabulary words. In those previous five weekly assessments, the number of words Kumar spelled cor- rectly from each 20 word list were 10, 9, 8, 7, and 8 respectively. Ms. Rhoten treated these assessments as her baseline data.

For an intervention, Ms. Rhoten used an oral spelling strategy through which she taught Kumar to break each word into syllables and sound them out. She believed that Kumar’s spelling would improve if he could better pair phonemes and graphemes. However, after five more weekly assess- ments, Kumar had only improved his correctly spelled words to 10, 11, 10, 14, and 13. Although there was a very slight upward trend in the data, Ms. Rhoten feared that the intervention was not sufficient for helping Kumar improve his spelling to a predetermined mastery level of 18 or more words correct per assessment and he would continue to fall further behind his peers. She believed there had been sufficient time for more substantial improvement, but she was not yet willing to give up on the oral spelling strategy intervention altogether.

Because Kumar did not make adequate progress, Ms. Rhoten decided to meet with the special education teacher, Ms. Cheng, to identify an additional strategy to assist Kumar. Ms. Cheng looked at the baseline and oral spelling strategy data and suggested having Kumar use a key- board to type the spelling assessments. She believed that if Kumar had letters in front of him, he would be able to more easily identify the graph- eme(s) with phoneme(s) and that typing the spelling words would also relieve the stress of handwriting. Ms. Rhoten worked with Kumar on using the keyboard to spell (he was already familiar with computer and keyboard use) in addition to continuing with the oral spelling strategy. In this third phase of the project, the combination of the interventions appeared to result in more rapid and satisfactory progress. Over the next five weekly spelling assessments, Kumar scored 17, 18, 19, 18, and 20 words correctly. Clearly, it appeared the combination of interventions was effective. His weekly spelling assessments continued to show 18 or more words spelled correctly. The results of this study are presented in Figure 8-2.

190 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Kumar: The Questions After reading the description of Kumar and the intervention strategies Ms. Rhoten implemented, answer the following questions. Then compare your answers to those provided in the answers section.

What is the purpose of the study?

Who are the subjects?

What is the setting?

What was the dependent variable?

What are the independent variables?

What kind of intervention is provided to the subject?

How were data collected and presented on a graph?

What were the results?

Why use an A-B-C design for this study?

What are the limitations of this study?

Kumar: The Answers Purpose of the Study The purpose of this study was to evaluate the effectiveness of an oral spell- ing strategy intervention to increase the words spelled correctly on Kumar’s weekly spelling assessments. A combination of interventions were then introduced and evaluated when the first intervention alone did not produce the desired results.

Subject Kumar, a third-grade boy who was diagnosed with a specific learning disability with specific problems in spelling, served as the subject.

Setting The study was conducted in a literacy teacher’s classroom.

Dependent Variable The dependent variable identified in this study was the number of vocabulary words spelled correctly out of 20 presented in weekly assessments.

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 191

Independent Variables An oral spelling strategy was the original intervention and independent variable in phase two of the study. A combination of this same strategy paired with typing the weekly spelling assessments was the intervention in the third phase. The two interventions in combination was the independent variable at the study’s end.

The Design An A-B-C design was used because after baseline, an intervention was introduced (B) as typically occurs in an action research study. A second intervention phase (C) began in hopes of producing more satisfactory results than from the B phase intervention alone.

The Intervention To help Kumar learn to spell third-grade vocabulary words correctly, Ms. Rhoten used an oral spelling strategy through which she taught him how to break vocabulary words into syllables and sound them out. Kumar made some, but inadequate, progress during this intervention. Based on a conversation with Ms. Cheng, Ms. Rhoten changed to a combination of the oral spelling strategy and typing his weekly spelling assessments. In five weeks Kumar improved his spelling skills considerably and was able to cor- rectly spell 18 or more of 20 vocabulary words on tests for four consecutive weekly assessments.

Obtaining the Data and Plotting the Results For baseline, Ms. Rhoten used Kumar’s existing five weekly spelling assess- ments. While this is not typical, it was practical in that the dependent variable would remain the same during intervention and it was ethical in that Ms. Rhoten did not continue a prolonged baseline during which she believed Kumar would continue to fail. During the first intervention condi- tion, an oral spelling strategy was used and Kumar had limited improve- ment on the dependent variable. Since he did not improve adequately, Ms. Rhoten introduced a combination intervention as the independent variable. Data were plotted on a line graph to show Kumar’s performance during baseline and the two intervention conditions (see Figure 8-2).

Results During the baseline, Kumar correctly spelled a very limited number of the 20 weekly vocabulary words. During the oral spelling strategy intervention, Kumar improved during the next five weekly assessments, but not to a

192 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

predetermined mastery level. Since he did not make substantial improve- ment, his literacy teacher consulted with the special educator and introduced a new combination intervention. In five weeks, Kumar was spelling at the mastery level of at least 18 of 20 words correct.

Why Use an A-B-C Design? An A-B-C design is more likely to be used in response to results during the B phase than as a pre-planned design. Although Ms. Rhoten tried the oral spelling strategy with Kumar for five weekly assessments, he did not make adequate progress. As a result, she introduced a different strategy using a combination of the oral spelling strategy and typing his weekly assessments resulting in improvement to the desired level of performance. For an action research study such as this, the A-B-C design is appropriate because the first intervention did not result in adequate progress with Kumar and he was still in need of improvement. Through this design, Ms. Rhoten was able to introduce a new intervention to assist Kumar in improving his spelling. It is worth noting a withdrawal design (e.g. A-B-A-C) design would present ethical issues for reintroducing a baseline condition. Also, it is unlikely Kumar would completely forget the oral spelling strategy he had learned in the B phase and that his responding during a second baseline phase,

Baseline Oral Strategy Oral Strategy and Keyboard

1 0

2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

15 161 32 4 5 6 7 8 9 10 11 12 13 14

# of

W or

d s

Sp el

le d

C or

re ct

Kumar

Weekly Assessments

FIGURE 8-2 Kumar’s Spelling.

20 14

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 193

if one was reintroduced, may not even present a downward trend. Finally, some researchers may have used the notation B " C for the third phase of the study to indicate the original intervention was still in use but now in combination with an additional intervention, rather than just a C notation.

Limitations of the Study This study has two obvious limitations. First, there was not a baseline phase in the typical manner. This should not, however, have affected the results of the study and was in response to ethical considerations. Second, there could have been a cumulative effect from the oral spelling strategy intervention that extended into the C phase. Kumar had been improving during the B phase. Ms. Rhoten could not determine if the C phase improvements were due solely to adding the typing intervention, to the combination of the oral spelling strategy and typing, or even possibly primarily to a delayed or long-term effect of the oral spelling strategy alone. As stated in chapter 7, the primary problem with the A-B-C design is that there is very little evi- dence of a functional relationship between the independent variable(s) and the dependent variable.

Summary A summary of the relevant dimensions of this application practice study can be found in Table 8-2.

TABLE 8-2 Summary of “The Impact of an Oral Spelling Strategy and Typing Interventions

on Improving the Performance on Weekly

Spelling Assessments of a Student with a Specific

Learning Disability.”

FEATURE DESCRIPTION

Type of design A-B-C design

Purpose of the study Evaluate the effectiveness of oral spelling strategy and typing interventions on Kumar’s spelling accuracy on weekly spelling assessments

Subject Kumar, a third-grade boy diagnosed with a learning disability

Setting Literacy teacher’s classroom

Dependent variable The number of words spelled correctly from a list of 20 presented each week

Independent variables Oral spelling strategy and typing of weekly assessments

Results and outcome A combination of interventions appeared to be most effective in improving Kumar’s spelling performance and on weekly assessments

194 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Changing Criterion Design with a Return to Baseline Phase De Luca, R. V., & Holborn, S. W. (1992). Effects of a variable-ratio reinforcement

schedule with changing criteria on exercise in obese and nonobese boys. Journal of Applied Behavior Analysis, 25, 671–679.

Purpose of the Study The authors designed this study to determine if variable ratio (VR) schedules would increase and maintain high rates of exercise. This study was an exten- sion of these authors’ previous study in which they used a fixed ratio (FR) schedule to increase students’ exercise skills. They wanted to compare the results of the two studies to determine if one type of schedule was preferable.

Subjects The subjects were six boys, each 11-years old. Three of these subjects were obese and three of them were of normal weight.

Setting The study was conducted in an elementary school nurse’s room. Each sub- ject exercised on a stationary bicycle in the nurse’s room.

Dependent Variable The target behavior was pedaling on a stationary exercise bicycle. Pedaling was measured by (a) the subjects’ overall responding rate per session (total number of wheel revolutions per session divided by the number of minutes spent exercising) and (b) the total time spent exercising per session.

Independent Variable The independent variable was the use of different schedules and criteria for reinforcement. A token system in which subjects earned points by pedaling a bicycle that could be exchanged for reinforcers from a reinforcement menu (a handheld battery-operated game, kite, bicycle bell, flashlight, model car, model plane, puzzle, adventure and comic books) was used for meeting the different criteria for reinforcement.

The Design A changing criterion design with a return-to-baseline phase was used in this study. A VR schedule of reinforcement was implemented with a different criterion for performance for each subject based on their performance in the preceding subphase.

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 195

The Intervention Subjects were tested individually once a day, from Monday through Friday, for about 12 weeks. At the beginning of each session, an identical instruc- tion, “exercise as long as you like,” was given to all subjects. Sessions were terminated when the subject dismounted from the bicycle or the assigned time of 30-minutes had elapsed. After stable baseline was established, the VR schedule of reinforcement (i.e., a rate of 70–85 revolutions per minute) was implemented for eight sessions. For each of the three VR subphases, a criterion of 15% increase over mean responding during the previous subphase was established. “In other words, a different criterion for perfor- mance was specified for each subject, based on his performance in the previous subphase” (p. 673)*. To further determine the effectiveness of the VR schedule, a return-to-baseline phase was implemented after the third intervention subphase. Finally, the last VR subphase was reinstated.

Obtaining the Data and Plotting the Results The experimenter sat behind the equipment panels out of the subject’s sight and started recording data when each subject began pedaling the bicycle. For accurate data collection, the stationary bicycle was programmed to signal when the variable number of responses had occurred. The mean number of revolutions per minute and the total number of minutes spent exercising were determined for each subject (obese and nonobese) and plotted on an x-y graph (see Figure 8-3).

Results The data for the obese and nonobese subjects were collapsed and discussed as group data. During baseline, the nonobese boys responded at a mean of 71.9 revolutions per minute and the obese boys responded at a mean of 59.2 revolutions per minute (see Figure 8-3). During the first subphase, the mean rates of responding for the nonobese and obese boys were 98.89 and 85.51 revolutions per minute, respectively. During the second subphase, the rates of responding per minute increased to 114.2 and 101.2. The mean response rates during the third subphase for the nonobese and obese boys were 130.0 and 117.0, respectively. Following this phase, a brief return to baseline produced a reduction in response rate to 95.3 for nonobese boys and 83.6 for obese boys. However, when the reinforcement was reintro- duced, it produced the highest response rates in all boys (nonobese boys, 138.7, and obese boys, 123.6). The authors compared this VR schedule

*De Luca, R. V., & Holborn, S. W. (1992). Effects of a variable-ratio reinforcement schedule with changing criteria on exercise in obese and nonobese boys. Journal of Applied Behavior Analysis, 25, 671–679.

196 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Sessions

M ea

n R

ev ol

ut io

n s

Pe r M

in ut

0 40 60 80

100 120 140 160

Baseline VR 80 VR 115 VR 130 BL VR 130

SCOTT Non-obese

0 40 50 70 90

110 130 150

Baseline VR 85 VR 115 VR 125 BL VR 125

SHAWN Non-obese

0 40 50 70 90

110 130 150

Baseline VR 85 VR 115 VR 125 BL VR 125

STEVE Non-obese

0 40 60 80

100 120 140

Baseline VR 70 VR 95 VR 100 BL VR 100

PETER Obese

0 40 60 80

100 120 140

Baseline VR 80 VR 105 VR 120 BL VR 120

PAUL Obese

0 1 5 10 15 20 25 30 35 40

40 60 80

100 120 140

Baseline VR 70 VR 90 VR 110 BL VR 110

PERRY Obese

FIGURE 8-3 Data for the subjects in the study using a changing

criterion design with return to base line. Note. From

“Effects of a Variable-Ratio Reinforcement Schedule

with Changing Criteria on Exercise in Obese and

Nonobese Boys,” by R. V. De Luca, and S. W.

Society for the Experimental Analysis of Behavior. Reprinted with

permission.

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 197

study with their previous FR schedule of reinforcement study (De Luca & Holborn, 1990) and indicated that the VR subphases of the changing crite- rion design produced greater increases in the rate of exercise.

Why Use Changing Criterion Design with a Return to Baseline? The investigators were interested in determining the effects of a variable ratio reinforcement procedure to increase the bicycle pedaling of nonobese and obese boys. The changing criterion design provided the opportunity to set initial criterion levels based on the subjects’ own baseline performance, and allowed each subject to increase his own rate through small successive increments. Through this procedure, rates of exercise were increased gradu- ally and systematically. By reinstating the baseline conditions and noting decreases in responding in all subjects, more convincing evidence of the functional relationship between the independent and dependent variable was provided.

Limitations of the Study One limitation of the study was gender bias. It was noted in the study that there are more overweight girls than boys, yet the authors included only boys in the study. Therefore, the findings may not be generalizable to the general population. Second, the return-to-baseline phase included only three sessions before the final VR subphase was reintroduced due to end of school year time constraints. This brief baseline phase demonstrated a reduction in both mean revolutions and time spent exercising, but was not long enough to allow stabilization of the subjects’ target behaviors. Finally, the authors do not indicate if the reinforcer (token system with reinforcement menu) was the same in their previous study using FR schedules.

Summary A summary of the relevant dimensions of this changing criterion study can be found in Table 8-3.

Application Practice: Jerry Jerry, a fifth-grade student diagnosed with behavior disorders, is in Mr. Hunter’s general education classroom. Although he likes being in the inclusion classroom with his peers and does fairly well in other subjects, Jerry gets frustrated and gives up on his math work if he can’t solve the pro- blems quickly. It seems that he can actually do the math work but becomes

198 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

frustrated easily. Also, he rarely turns in his math homework and never gives a valid reason for not completing it. His IEP team decided to place him in Mrs. Smith’s resource room for a 40-minute math session of review and instruction five days a week. Mrs. Smith contacted his parents. Jerry’s mother told her that he dislikes math and avoids doing math homework in every possible way even after she sits with him to help. Mrs. Smith is con- cerned about Jerry’s lack of compliance and task completion.

Since the current lesson unit involves solving multiplication problems, Mrs. Smith wants to focus on increasing the number of multiplication pro- blems Jerry solves when given his assigned math worksheets. She contacted the school’s Behavior Intervention Specialist, Mr. Jackson, and told him about her concerns. After reviewing Jerry’s recently completed math work- sheets, Mr. Jackson met with Mr. Hunter and Mrs. Smith and they agreed to develop an intervention plan.

A week later, Mr. Jackson observed Jerry in his resource room and confirmed Mrs. Smith’s worries about Jerry’s frustrations when he is work- ing on math. At the end of the class, Mr. Jackson sat next to Jerry and explained to him about an intervention plan which would allow him to play games on his iPad if he correctly completed the number of assigned multiplication problems each day. At this point, Mr. Jackson and Mrs. Smith agreed that the goal was to increase the number of math problems Jerry solved in a given time.

Mrs. Smith first gave Jerry one worksheet with 15 multiplication pro- blems per day for a week. He completed 6, 3, 4, 5, and 2 problems correctly

TABLE 8-3 Summary of “Effects of a Variable-Ratio

Reinforcement Schedule with Changing Criteria on

Exercise in Obese and Nonobese Boys.”

FEATURE DESCRIPTION

Type of design Changing criterion design with a return-to-baseline phase

Purpose of the study Determine the effectiveness of VR schedules of reinforcement to increase exercise of obese vs. nonobese children

Subjects Three obese and three nonobese 11-year-old boys

Setting Nurse’s room in an elementary school

Dependent variable Pedaling on a stationary exercise bicycle as measured by (a) the subjects’ overall responding rate per session (total number of wheel revolutions per session divided by the number of minutes spent exercising) and (b) the total time spent exercising per session

Independent variable Use of a token system to exchange for reinforcers from a reinforcer menu along with the changing VR schedule of reinforcement

Results and outcomes VR schedule of reinforcement using the token system produced increased exercise in both obese and nonobese boys.

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 199

over this five day period. These data were used as baseline. During this base- line phase, Jerry continued to demonstrate his frustration and refusing to complete his work. He even told Mrs. Smith that he was “dumb” on several occasions. At this point, Mr. Jackson and Mrs. Smith began the interven- tion. Mrs. Smith provided her math review/instruction for 5-10-minutes and then allowed Jerry to work independently for 15-minutes to complete his assigned work for the remainder of the lesson. The assigned work included only multiplication problems that Jerry had already mastered.

During the first subphase, the criterion for the number of multiplication problems to be solved correctly by Jerry was set at six. In order to move to the next subphase, Jerry had to meet the criterion (or higher) for three consecutive days. The reinforcement for correctly solving six math problems on the worksheet during a 15-minute period would allow him access to an iPad and the opportunity to play a game of his choice for 10-minutes. This same contingency was in effect for each subphase. During this subphase, he solved 6, 5, 6, 6, and 6 problems correctly. The criterion for the second supphase was set at nine. Jerry completed 8, 9, 7, 10, 9, and 9 problems correctly, and the criterion level was then set at 12. Jerry’s performance was 11, 13, 12 and 12. Finally, Jerry was required to complete all 15 pro- blems correctly to receive his reinforcement. During this final subphase, Jerry answered 15, 13, 15, 14, 15, 15, and 15. At this point, Jerry had com- pleted all of the problems correctly over a three day period. During the inter- vention Mrs. Smith noted that Jerry’s frustrations disappeared, he did not give up on his math work, and he showed improved self-esteem by com- menting on how well he did his work. Also, he solved the math problems rapidly and never needed the 15-minutes to solve the preset number of cri- terion problems after the second intervention phase. Mrs. Smith decided to keep the reinforcement program in effect. The results of this study are pre- sented in Figure 8-4.

Jerry: The Questions After reading the description of Jerry and the study Mrs. Smith executed, answer the following questions. Then compare your answers to those pro- vided in the answers section.

What is the purpose of the study?

Who are the subjects?

What is the setting?

What was the dependent variable?

What are the independent variables?

What kind of intervention is provided to the subject?

200 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

How were data collected and presented on a graph?

What were the results?

Why use changing criterion design for this study?

What are the limitations of this study?

Jerry: The Answers Purpose of the Study The present study examined whether the use of an iPad to play games as a reinforcer would increase a student’s correct completion of multiplication problems.

Subject A fifth-grade student diagnosed with behavior disorders participated in this study.

Setting The study was conducted in a resource room.

Baseline iPad Reinforcer

Observations

Jerry

# of

P ro

b le

m s

So lv

ed C

or re

ct ly

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

FIGURE 8-4 Jerry’s Math Problems

Solved.

20 14

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 201

Dependent Variable The dependent variable identified in this study was the number of multipli- cation problems correctly completed from a 15-problem worksheet.

Independent Variable The use of an iPad with preloaded games served as the independent variable.

The Design A changing criterion design was used to establish the functional relationship between the contingent use of an iPad on increasing multiplication problems correctly completed.

The Intervention To help improve Jerry’s correct completion of multiplication problems, an iPad was used as a reinforcer to correctly solve multiplication problems at preset criterion levels. Once he met the criterion, he was provided the oppor- tunity to play a game of his choice on the iPad for 10-minutes each day.

Obtaining the Data and Plotting the Results Mrs. Smith gave a worksheet with 15 multiplication problems per day during the baseline and intervention subphases. She used a frequency data recording method and recorded the number of multiplication problems Jerry completed. Data were plotted on a graph to show the number of math problems he correctly completed during baseline and intervention subphases (see Figure 8-4).

Results During baseline, Jerry correctly solved four multiplication problems on aver- age from the 15-math problem worksheet. Once the first intervention was introduced with the contingency in place, he met his criterion of six problems for three consecutive sessions in four days. In subsequent subphases, the criterion was increased to 9, 12, and 15 multiplication problems answered correctly. He met these criterion levels in 6, 4, and 7 days respectively.

Why Use the Changing Criterion Design? In a changing criterion design, each subphase provides a baseline for the fol- lowing subphase. When the participant’s multiplication problem solving skills increased to the preset criterion level, it demonstrated experimental

202 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

control and established a functional relationship between the independent and dependent variables. The changing criterion design is appropriate because it allows the student to build on the learned behavior (already mas- tered multiplication problems) to successive approximations toward the tar- get goal, in this case, completing 15 math problems. The goal was to increase the number of correct multiplication problems within a 15-minute time limit. By establishing criterion levels for the number of correct pro- blems necessary to receive the reinforcer, Jerry’s performance could be mon- itored to see if it increased to those criterion levels. Thus, his performance would increase in a step-wise fashion.

Limitations of the Study Although this study was planned and implemented well, it has some limita- tions. First, the use of an iPad as a reinforcer helped Jerry solve the set crite- rion number of math problems in the resource room. However, the teacher did not use this intervention procedure in his general education classroom to assess generalization. Second, this study was carried out with only one par- ticipant, making it difficult to know if similar results would be achieved with other participants. Third, the teacher reported that the participant’s frustrations associated with mathematics disappeared, but there was neither data collection nor visual representation of this decrease in behavior. Finally, to demonstrate the best experimental control, the performance should increase to the criterion level quickly and neither go below or above that level. Of the 22 intervention sessions, Jerry was below the crite- rion level 5 times (approximately 23%) and above 2 times (9%). However, he achieved at or above the criterion the vast majority of sessions. He also met the requirement of three days in a row of performing at least at the cri- terion level relatively rapidly. The overall performance demonstrates the effectiveness of the intervention rather than a limitation of the study. A fur- ther step that could be used to strengthen the study would be to either return to a previous criterion level subphase or a baseline subphase to dem- onstrate Jerry’s behavior was functionally related to the criterion for rein- forcement. However, given that this study was more an applied study than a research study, quickly reaching the goal was desirable and was in Jerry’s best interests.

Summary A summary of the relevant dimensions of the changing criterion design is presented in Table 8-4.

The changing conditions design is generally used when the first interven- tion phase does not produce the desired effect on the dependent variable.

CHAPTER 8 APPLICATION OF CHANGING CONDITIONS AND CHANGING CRITERION DESIGN 203

The changing criterion design has been used in a variety of circumstances and is useful for producing stepwise changes in a target behavior. Regard- less of which design is used, the researcher is collecting data that must be analyzed to determine if the functional relationship exists between the independent and dependent variables.

TABLE 8-4 Summary of “The Use of an

iPad as a Reinforcer to Increase the Number of Correct Multiplication

Problems of a Student with Behavior Disorders.”

FEATURE DESCRIPTION

Type of design Changing criterion design

Purpose of the study Examine whether the use of an iPad to play computer games increases the number of correct multiplication problems completed

Subject Jerry, a fifth-grade student diagnosed with behavior disorders

Setting Special education resource classroom

Dependent variable Number of correctly completed multiplication problems

Independent variable Use of an iPad with preloaded games as a reinforcer

Results and outcome The iPad used to play games increased correctly completed multiplication problems to the predetermined criterion levels

204 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

CHAPTER

9 Overview of Multiple Baseline Designs

IMPORTANT CONCEPTS TO KNOW THE BASIC MULTIPLE BASELINE DESIGN

MECHANICS OF THE MULTIPLE BASELINE DESIGN

PREDICTION, VERIFICATION, AND REPLICATION Covariance Among Dependent Variables

ADVANTAGES OF THE MULTIPLE BASELINE DESIGN

DISADVANTAGES OF THE MULTIPLE BASELINE DESIGN

THE DIFFERENT MULTIPLE BASELINE DESIGNS Multiple Baseline Across Behaviors Multiple Baseline Across Settings Multiple Baseline Across Subjects

ADAPTATIONS OF THE MULTIPLE BASELINE DESIGN Multiple Probe Design Delayed Multiple Baseline Design Summary

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

205

M ultiple baseline designs are A-B designs that are replicated within thesame study. The three major types of multiple baseline designs aremultiple baseline across behaviors, settings, and subjects. In the basic multiple baseline design, the researcher takes repeated measures of baseline performance concurrently on two or more baselines (same individual with two or more behaviors in need of intervention; same individual with the same behavior in need of intervention in two or more settings; or two or more individuals with the same or very similar behavior in need of intervention in the same setting—i.e., across behaviors, across settings, or across subjects). When a stable, predictable baseline is obtained, the researcher implements the intervention (independent variable) and records the results over a period of time to determine the effects of the intervention. Typically, a criterion is established a priori to establish a dependent variable level at which one may judge the intervention or treatment has been successful in altering the dependent variable, or the researcher may simply apply the independent variable until a stable performance is obtained on the dependent variable following the introduction of the intervention. A withdrawal may be included following an intervention phase within a multiple baseline design. The process is repeated for each baseline.

The Basic Multiple Baseline Design Multiple baseline designs may be the most appropriate single subject designs to use for a variety of reasons (Baer, Wolf, & Risley, 1968). These include the following: (a) when withdrawal or reversal designs may not be feasible due to ethical concerns about withdrawing treatment that is working (Harvey, May, & Kennedy, 2004); or (b) there may be practical considera- tions, such as more than one person or setting needing interventions; or (c) in those cases where the independent variable (treatment) should not be withdrawn or the achieved target behavior cannot be reversed (e.g., allow- ing verbal threats to once again increase after a decrease has been obtained, learning to match consonant sounds with the appropriate alphabet letter), it may be more appropriate to implement a multiple baseline design. Multiple baseline designs are versatile, are relatively easy to understand, and are gen- erally practical in real-world settings (Cooper, Heron, & Heward, 2007; Hammond & Gast, 2010).

Because multiple measures are used to obtain data over two or more baselines (and usually three or more), the end result appears visually as a series of A-B designs stacked on top of one another. Actually, in a multiple

206 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

baseline across behaviors design, two or more target behaviors of the same individual receive the same treatment in the same setting. In the multiple baseline across settings design, the researcher is applying the same interven- tion to the target behavior of the same individual in two or more settings. In a multiple baseline across subjects design, the researcher is applying the same intervention to the target behavior of two or more individuals in the same setting. Many experts (e.g., Alberto & Troutman, 2013) prefer to conceptualize each baseline (be it different behaviors in the same individ- ual, different settings where the same individual’s target behavior is occurring, or different individuals exhibiting the target behavior in the same setting) as a different target behavior or dependent variable. This concept is used because each baseline followed by intervention is treated as a separate applied behavior analysis (Alberto & Troutman, 2013). Therefore, we will discuss the designs in terms of recording and interven- ing with the first, second, and third dependent variables (i.e., across beha- viors, settings, or individual subjects). Keep in mind that this is convenient for the sake of discussion. In reality, the same method for recording the target behavior is used with each baseline and intervention (Alberto & Troutman, 2013). See Figure 9-1 for a depiction of the vari- ous designs and what variables actually remain the same and which change based on the design used.

Mechanics of the Multiple Baseline Design Assume three dependent variables have been chosen. After baseline data have been obtained for all three dependent variables, the researcher imple- ments the intervention for the first dependent variable while maintaining baseline conditions for the other two. When the criterion is obtained on the first behavior, in the first setting, or with the first individual subject fol- lowing intervention, the intervention may be implemented and analyzed as to its effect on the second dependent variable. Meanwhile, baseline condi- tions are maintained with the third dependent variable. This is then fol- lowed by introducing the intervention to the third dependent variable after it is clear that the intervention is working with the second dependent vari- able (see Figure 9-2). Subsequent dependent variables, if any, would con- tinue with the same format and design in this time-staggered method (Tankersley, Harjusola-Webb, & Landrum, 2008). Frequently, a follow-up stage or a generalization phase is included to ensure that the effects of the independent variable are maintained over time, settings, behaviors, and/or individuals.

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 207

Sa m

e Su

b je

ct /S

am e

Se tt

in g/

D i!

er en

t B eh

av io

rs Target Behavior

One Target Behavior

Two Target Behavior

Three

Target Subject One

Target Subject Two

Target Subject Three

Target Setting One

Target Setting Two

Target Setting Three

D i!

er en

t S ub

je ct

s/ Sa

m e

Se tt

in g/

Sa m

e Be

ha vi

or Sa

m e

Su b

je ct

/D i!

er en

t S et

tin gs

/ Sa

m e

Be ha

vi or

O b

se rv

at io

n s

O b

se rv

at io

n s

O b

se rv

at io

n s

M ul

tip le

B as

el in

e A

cr os

s Be

ha vi

or s

D es

ig n

M ul

tip le

B as

el in

e A

cr os

s Su

b je

ct s

D es

ig n

M ul

tip le

B as

el in

e A

cr os

s Se

tt in

gs D

es ig

FI GU

RE 9- 1

Va ria bl es in m ul tip le

ba se lin e de sig n

©CengageLearning2014

208

Criterion Level

D ep

en d

en t V

ar ia

b le

T w

D ep

en d

en t V

ar ia

b le

O n

e D

ep en

d en

t V ar

ia b

le T

h re

Observations

Criterion Level

FIGURE 9-2 Typical multiple baseline design

20 14

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 209

4 C H E C K I T O U T # 1 A colleague approaches you about an individual in counseling who has continuing issues with severe cursing at home, in therapy sessions, and at work. This target behavior has become so pronounced it is possible the individual may lose his job. His parents are also embarrassed and frustrated by his cursing at home. In his group therapy sessions, his peers also are complaining that the individual is violat- ing the rules of the group. Which multiple baseline design might be appropriate in this case?

Prediction, Verification, and Replication Beginning the collection of baseline data simultaneously across all depen- dent variables adds an important feature to multiple baseline designs. It allows the researcher to infer a verification of the prediction that the base- line behaviors would have remained stable and unchanged if the interven- tion had not been implemented (Kucera & Axelrod, 1995). As discussed in Chapter 4 (and here we have modified the discussion of Tawney & Gast, 1984), verification is evident if the data path changes in a predictable man- ner through a phase change, as from baseline to intervention. In other words, if the data path line remains constant across the baseline and intervention phases, then the independent variable yields no effect on the dependent variable (i.e., no change in the target behavior when the intervention is implemented) and there is not verification. If there is a change in the data path, then the possibility exists that the independent variable is responsible for the change and there is verification that the data path has changed with the introduction of the treatment. In a multi- ple baseline design, replication of this prediction and verification may occur when the data paths of subsequent dependent variables follow pat- terns similar to the first (and subsequent) variables. According to Cooper et al. (2007), inferences can be made concerning the complementary roles of prediction and verification in multiple baseline designs. First, if poten- tial confounding variables are held constant across all variables and a target behavior remains unchanged from Dependent Variable 1 to Vari- ables 2 and 3 (and so on), then the prediction is valid. Second, if changes occur with Dependent Variable 1, the observed changes in that target behavior are brought about by the independent variable because only that dependent variable was exposed to the independent variable. In other words, baseline levels of each dependent variable remain stable and then change when and only when the independent variable is intro- duced (Alberto & Troutman, 2013).

210 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Consider the following example:

Joseph is an 18 year old male who is performing poorly in three differ- ent content area classes in high school. He fails to engage in appropriate comments in each of the three classes. His team implements a multiple baseline across settings design. Baseline measures of his appropriate comments are gathered in each class. An intervention is implemented in the first class following establishment of a stable baseline in all three classes. The intervention is only in effect in the first class. The dependent variable (frequency of appropriate comments) should remain stable at baseline levels in the second and third classes. As there is notable improvement in the first class, the intervention is implemented in the sec- ond class. These first two dependent variables should continue to main- tain (in the first and second settings) or improve. The third class should still have stable baseline level responding as the intervention has not yet been applied in that setting. Replication is achieved when similar results ultimately are obtained with each dependent variable following the introduction of the same independent variable (see Figure 9-3).

This form of replication provides a convincing argument for the pres- ence of a functional relationship between the dependent and independent variables. It is important to note the importance of maintaining control of the extraneous variables that have the potential to influence results across variables. The only changes that should be evident are found in (a) the dependent variable (and again those may be different behaviors of the same individual, the same behavior in the same individual in different settings, or the same behavior in the same setting with different individuals) and (b) the treatment, or independent variable. Other variables should remain constant to ensure that the results can indeed be attributed to the independent variable (Cooper et al., 2007; Barlow, Nock, & Hersen, 2009).

Covariance Among Dependent Variables It is also important to note the relationship among the dependent variables. Realizing that variables in the treatment may covary, it is important to select dependent variables that exhibit some degree of independence (Tawney & Gast, 1984). In our example, if a student was likely to increase appropriate comments in all three classes when the independent variable is introduced only in the first class, then the dependent variables would covary (or change at the same time and in the same direction even in the absence of the inter- vention in classes two and three) (Datilo, Gast, & Malley, 2000). These variables would not be considered sufficiently independent of one another for research purposes because one could not achieve the desired replication of effects across the dependent variables (see Figure 9-4 for an example of

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 211

Fr eq

ue n

cy o

f A p

p ro

p ri

at e

C om

m en

Observations

1st Class

2nd Class

3rd Class

A B

Criterion

FIGURE 9-3 Data from Joseph example

displaying a functional relationship within a

multiple baseline design

20 14

212 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Fr eq

ue n

cy o

f A p

p ro

p ri

at e

C om

m en

Observations

1st Class

2nd Class

3rd Class

Criterion

FIGURE 9-4 Example of covariance

among dependent variables

20 14

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 213

covariance among dependent variables). This independence is important in maintaining experimental control, yet selecting independent variables that are completely unrelated would also be undesirable.

For example, the dependent variables in a multiple baseline design, such as hitting children on the playground, eating too quickly at lunch, and start- ing work on time, are all behaviors that through some type of intervention one might aim to improve, but one might not expect them to change in response to the same independent variable. For example, intervention designed to teach this student to obtain attention in a more adaptive manner might decrease hitting, but may not result in similar changes in eating too quickly or starting work on time. Also, in our example, measurement of the dependent variables would probably not be the same. Whereas hitting might be measured through frequency, one would likely measure eating too quickly through a duration procedure and starting work on time through a latency procedure.

It is necessary that each of the dependent variables be measured using the same method of recording behavior (Alberto & Troutman, 2013) and the researcher have a reasonable expectation that each dependent variable will respond similarly to the independent variable.

Therefore, selecting dependent variables that are completely unrelated can present as many problems as selecting variables that are significantly interrelated. A balance between these two areas of concern will produce the best choice. The researcher should select dependent variables (indivi- duals with the same behavior, different behaviors in the same individual, or different settings in which the same individual exhibits the same behav- ior) that are functionally similar so that they would likely change similarly in response to the same treatment, while at the same time not be likely to change until that treatment is specifically introduced to the particular depen- dent variable (e.g., ed endings and spelling changes may very well respond similarly to instruction that improves recognition of s endings but would be unlikely to change until the instruction is delivered specifically for each of those behaviors). Tawney and Gast (1984) referred to these dependent vari- ables (or baselines) as being at once functionally similar and functionally independent of one another. Again, these dependent variables must be mea- surable using the same method for recording behavior as well so that results are easily compared from one dependent variable to the next. Also, the dependent variables should be measured concurrently, and the possible influ- ences of other variables should be reasonably equal or their influence controlled for each dependent variable (Alberto & Troutman, 2013; Barlow et al., 2009).

In addition to these concerns, the researcher must also be aware that the various design options available will influence the decisions made regarding experimental procedures. As noted earlier, multiple baseline designs

214 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

typically involve multiple baselines across behaviors, individuals, or settings. Although the basic design is unchanged, each of these particular variations has its unique issues and concerns that should be addressed.

Advantages of the Multiple Baseline Design There are several advantages to the multiple baseline design. First, the with- drawal of an effective treatment is not required to demonstrate the func- tional relationship between the independent and dependent variables (Baer et al., 1968). Second, the sequential implementation of the independent var- iable parallels the practice of many teachers (Alberto & Troutman, 2013). Third, generalization of behavior change is monitored through the design. Fourth, the design is easily conceptualized and used (Cooper et al., 2007).

A multiple baseline design should be used in the following situations:

• When withdrawal designs are not feasible due to ethical concerns; • When there is more than one target behavior, setting, or individual in need of

treatment;

• When the effects of the independent variable cannot be withdrawn or reversed.

Disadvantages of the Multiple Baseline Design Multiple baseline designs have their disadvantages. These include the possi- bility of covariance and the aforementioned result that a functional relation- ship is not clearly demonstrated (Datilo et al., 2000). Verification is reliant on dependent variable levels not changing until the independent variable is introduced and then changing in a similar manner to any previously treated behaviors. Second, the multiple baseline design does yield data related to general effectiveness of the independent variable in treating various beha- viors, in different environments, or with various individuals, but allows us to analyze the dependent variable less so than other designs might allow because the treatment is applied in only one intervention phase as a rule. The withdrawal design can use multiple intervention and baseline phases to clearly demonstrate the functional relationship (Cooper et al., 2007). Finally, implementing a multiple baseline design can be time consuming and may require substantial resources because two or more dependent vari- ables are being measured simultaneously. Despite these limitations, the mul- tiple baseline design remains a commonly used design and one that lends itself well to clinical practice where intervention is required for multiple behaviors, multiple individuals, or the same behavior in multiple settings.

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 215

The multiple baseline design should not be used in the following situations:

• When selected target behaviors are not functionally similar nor independent of one another;

• If there is only one individual, in one setting, and one target behavior selected for treatment;

• When more than one intervention phase is desirable to demonstrate the func- tional relationship;

• When constraints on resources make implementation impossible.

4 C H E C K I T O U T # 2 A student has become so disruptive that her teacher is threatening to quit, the class as a whole is not functioning academically, and the parents of the other stu- dents are voicing concern about their children’s learning and achievement. If dis- ruptive behavior could be broken down into disruptive comments, disruptive noises, and disruptive questions, would a multiple baseline across behaviors design be appropriate? Why or why not?

The Different Multiple Baseline Designs Multiple baseline designs have variations within the basic design itself. First, we will discuss the variations of the basic design that are commonly found in the literature (across behaviors, settings, or individual subjects). Second, we will discuss adaptations (i.e., multiple probe and delayed multiple base- line) of the basic design that may be applied to any of these first three varia- tions. Each design has its own requirements for implementation. Taken together, multiple baseline designs and their variations are applicable for research use in many settings and with many individuals and behaviors.

Multiple Baseline Across Behaviors Generally, with the multiple baseline across behaviors design, three or more behaviors are identified that are exhibited by the same individual in the same setting and then systematically subjected to the same intervention or independent variable. The behaviors selected as targets need to be both functionally similar and functionally independent (Tawney & Gast, 1984). That is, the behaviors should be similar enough that the same treatment or intervention is likely to influence the occurrence of each in a similar fashion (e.g., increase, decrease, or maintain the level of each dependent variable).

216 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

At the same time, the behaviors should be unlikely to change (i.e., covary) until the intervention is actually introduced to influence that specific depen- dent variable. For example, an individual might exhibit visual threats, verbal threats, and vandalism. One might reasonably expect that these behaviors would be functionally similar (e.g., each involves abuse of another person or property). However, one might be less certain that if visual threatening is treated, verbal threats and vandalism would not covary when verbal threatening alone is treated. If they do not covary, then the behaviors are functionally independent. This type of covariance is of special concern with a multiple baseline across behaviors design because achieving functional similarity and independence among target behaviors is not necessarily easy. In fact, doing so a priori may be difficult (Cooper et al., 2007). From a treatment or outcomes perspective, such covariance is not necessarily unde- sirable; however, from a research perspective, the establishment of a func- tional relationship between the independent variable and dependent variables becomes problematic. The researcher will not have a clear demon- stration that the dependent variables or target behaviors change when and only when the independent variable is systematically applied to each. One possible advantage to covariance is that the researcher may assess and sub- sequently analyze concomitant changes that occur when the treatment is introduced to a specific dependent variable. For example, the researcher may indeed discover that verbal threats diminish when the independent var- iable is applied to visual threats. Yet, the third baseline (vandalism) remains unchanged during the treatment of physical aggression. This does suggest some generality of treatment but, again, may not allow one to carefully ana- lyze the direct influence of the treatment on verbal threats, and it may weaken the demonstration of the functional relationship.

The critical issues in implementing a multiple baseline across behaviors design include

1. Selection of an individual participant who displays multiple behaviors (at least two but preferably three or more for a convincing argument for a functional relationship) in a single setting;

2. Functional similarity and functional independence of those behaviors as one might be able to determine a priori;

3. A reasonable expectation that the same variables (extraneous or sys- tematic) will exert equal influence on each of the dependent variables;

4. Selection of a treatment or independent variable that can be expected to produce a similar and independent effect on each of the dependent variables;

5. A consistent recording procedure for each of the target behaviors and a criterion level for decision making; and

6. Confidence that the resources and time needed to record multiple baselines and subsequent intervention will be maintained throughout the study.

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 217

The following scenario serves as an example of the use of a multiple baseline across behaviors design.

Mrs. Davis has been working with Steve for several weeks now in her classroom. Steve has exhibited a number of behaviors that interfere with his learning and the learning of those around him, and, if untreated, may ultimately result in referral for evaluation for a behavior disorder. He has pinched other students, has repeatedly told others he would “get them” after school, and has made obscene gestures to appar- ently get attention from peers and Mrs. Davis. Mrs. Davis has requested that Dr. Lester assist her in intervening. After using anecdotal recording, the two agree on operational definitions for Steve’s pinching, threaten- ing, and gesturing behaviors. They also want to ensure that any inter- vention attempted will be successful before investing all their time and effort, so they wish to use it with the more severe target behavior first (pinching) before investing in using it with each of the other behaviors. They implement baseline recording on each of the three target behaviors in the classroom. Once stable responding is achieved (keeping in mind no other students are expressing significant distress or harm from Steve’s behavior), they implement their intervention (differential reinforcement of other behaviors with response interruption) with the pinching target behavior. The frequency of pinching abruptly decreases to zero levels after six sessions. Meanwhile, baselines have been maintained on threat- ening and gesturing and those behaviors have remained stable. Next, the two implement the intervention on threatening and that target behavior changes to zero levels after only four sessions. Pinching has remained at zero levels and gesturing continues to exhibit a steady baseline. Finally, the intervention is implemented with gesturing, which decreases to zero levels after only three sessions. Both pinching and threatening have remained at zero levels. See Figure 9-5 for a depiction of this study.

A major advantage of the multiple baseline across behaviors design is that generality of intervention effects for similar behaviors within the same individual can be demonstrated. The major disadvantage is the aforemen- tioned possibility of covariance among behaviors, which weakens the dem- onstration of a functional relationship.

Multiple Baseline Across Settings The multiple baseline across settings design is similar to the across behaviors design in that only one subject is identified. The researcher identifies two or more settings (again, generally at least three) in which the individual emits the same behavior. The same subject is treated for the same behavior in different settings. For example, a student might be treated for compulsive

218 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

15 DRO with Response Interruption

Pi n

ch in

g G

es tu

ri n

g Fr

eq ue

n cy

T h

re at

en in

DRO with Response Interruption

Sessions

FIGURE 9-5 Data from Steve example

displaying a multiple baseline across behaviors design. DRO = differential

reinforcement of other behavior

20 14

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 219

talking in the classroom, in the hallways, and in the cafeteria. Settings need not be literally interpreted, however, to mean different physical environ- ments. Settings may include functionally similar situations that are still inde- pendent of one another. For example, a student’s speech fluency might be addressed in whole group, cooperative group, and presentation situations. The physical environment may not actually change, although the situations may be different enough that the researcher may not expect the behavior to change in any of those situations until the intervention is applied. The inter- vention is applied after baseline data have been obtained in all settings, but initially applied in only one setting. When criterion or an acceptable level of responding is achieved in the first setting, the researcher implements the intervention in the second setting. Following acceptable responding in the second setting, the intervention is applied in the third setting. Baselines are maintained in settings prior to the introduction of the intervention in that setting. Intervention is maintained in prior settings as it is applied in subse- quent settings (or some follow-up or maintenance phase is in effect in the prior setting if optimal responding has already been achieved).

Critical issues in the implementation of the multiple baseline across set- tings design include

1. Selection of an individual subject who displays the same target behavior in multiple settings;

2. Selection of settings that are functionally similar but also independent of one another as one may best determine a priori;

3. A reasonable expectation that the same variables will be exerting the same influence in each of the settings;

4. Selection of a treatment or independent variable that can be expected to produce similar effects in each setting;

5. A consistent recording procedure for each setting and a criterion level for decision making; and

6. Confidence that the resources and time needed to record data in multiple settings will be maintained throughout the study.

The following scenario serves as an example of the use of a multiple baseline across settings design.

Mr. Stephens, Mrs. Roberts, and Mr. Michaels have all been working with Sara since she arrived at the high school 2 months ago. Each of the teachers works on a team that serves tenth-grade students, including Sara, who has mild cognitive disabilities and is included in their classes. In one meeting, they all express concern that Sara is consistently slow in beginning work. Each notes that she usually falls behind almost immedi- ately because she dawdles at getting her class materials out and therefore

220 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

is not focusing on what the teacher and other students are saying. She sometimes wanders slowly to her desk as the bell rings, further com- pounding her problem. Dr. Lester is asked to assist the team in imple- menting a program. Each teacher wishes to help Sara change her behavior, but they agree as a team to try the intervention (differential reinforcement for lowering the latency spent from when the bell rings to beginning work) in one class at a time to ensure it is effective. After beginning work has been clearly defined, baseline measures are taken in each class. The baselines are all steady. Mr. Stephens first implements the intervention, and Sara achieves the criterion level of responding (1-minute latency) in four sessions. Meanwhile, her latency for begin- ning work has continued at higher and steady levels in Mrs. Roberts’ and Mr. Michaels’ classes. Next, Mrs. Roberts implements the interven- tion and Sara’s latency falls to the acceptable level of 1 minute within five sessions. Her beginning work has continued to remain within the 1-minute latency in Mr. Stephens’ class and her baseline continues to be higher and steady in Mr. Michaels’ class. Finally, Mr. Michaels implements the intervention and obtains a result similar to that achieved in the first two settings. Figure 9-6 depicts the data from this example.

A major advantage to the multiple baseline across settings design is that generality of intervention effectiveness with the same individual in different settings may be demonstrated. A major disadvantage is that extraneous vari- ables that may influence responding in different settings may be difficult to control or predict. The presence of different people, times of the day, instructional or clinical activities, and so on, all may have some unforeseen influence on the individual’s responding. The more control the researcher may exert over such possible influences, the greater the likelihood that a functional relationship may be demonstrated.

Multiple Baseline Across Subjects The multiple baseline across subjects design differs from those for multiple behaviors or settings in that more than one subject participates. In this design, two or more (again, three or more is desirable) individuals are iden- tified who emit the same target behavior in the same setting.

The baseline measures reflect the responding of the multiple subjects. For example, three subjects may exhibit an articulation problem (or verbal threats or inaccurate math problem solving, etc.) in the same setting. The subjects should be enough alike that one might reasonably expect each to respond similarly to the same intervention, yet independent enough of one another that one is not likely to change his or her behavior as a result of perceiving changes in one of the other subject’s behavior (covariance).

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 221

15 Stephens’ Class

Roberts’ Class

Michaels’ Class

La te

n cy

in M

in ut

Observations

FIGURE 9-6 Data from Sara

example displaying a multiple baseline across

settings design

20 14

222 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

All other variables are held as constant as possible. One should be aware that same target behavior need not be literally interpreted to mean exactly the same behavior. Such an example might include one student who is dis- ruptive by making noises, one who is disruptive by talking out during instruction, and a third student who disrupts by talking to classmates during instruction. Although these target behaviors of different individuals are not exactly the same, they may be functionally similar yet still independent of one another. The researcher should logically be able to operationally define the target behavior of disrupting class in such a manner that each of the individual responses given would be examples of the target behavior.

The critical issues in implementing a multiple baseline across subjects design include

1. Selection of individual participants who display the same target behavior in the same setting;

2. Selection of individuals who are similar enough to one another to expect each would change his or her behavior in response to the same interven- tion and yet not likely to change his or her behavior until the interven- tion is specifically implemented to treat his or her behavior;

3. A reasonable expectation that the same variables will exert the same influence on each of the subjects;

4. Selection of an independent variable that is likely to have a similar effect on each subject;

5. A consistent recording procedure for all subjects’ behavior and a criterion level for decision making; and

6. Confidence that the resources will be available to maintain data collection and intervention across the life span of the study.

An example of the use of a multiple baseline across subjects design follows.

Mrs. Ziegler, a speech and language pathologist, works with Ed, Charles, and Lydia each day in their regular sixth-grade class. Each has difficulty with writing personal stories that can be followed well by others. After operationally defining storytelling, Mrs. Ziegler and the teacher decide to try teaching the students a strategy by which they can construct personal stories. They also decide they want to try the inter- vention with only one student at a time to ensure it is effective. They take baseline measures on the permanent products obtained from each student’s written stories. Each consistently exhibits a number of errors in storytelling which becomes the dependent variable in the study. The intervention is implemented with Ed first. Within five sessions where he has received the strategy instruction, Ed reduces his error level to an

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 223

acceptable level of two per story. Meanwhile, both Charles and Lydia continue to make many errors for each story they write. Next, Charles is taught the strategy. He improves to a two errors or fewer level of responding in only three sessions. Ed continues his improved perfor- mance, and Lydia continues to exhibit a steady and high level of errors. Finally, Lydia is taught the strategy and she too improves rather quickly. All students continue to maintain their performance. Figure 9-7 depicts the results of this study.

A major advantage to the multiple baseline across subjects design is that it allows the researcher to demonstrate the effectiveness of an intervention with more than one individual who displays a similar need for behavior change. A major disadvantage is that covariance among subjects may emerge if individuals learn vicariously through the experiences of other sub- jects. Identifying multiple subjects in the same setting who are functionally similar yet independent of one another can prove difficult.

Clearly, the multiple baseline design is versatile and has many potential applications. Still, a primary difficulty in implementing these designs is the availability of time and resources to maintain multiple baselines and data collection across behaviors, settings, or individuals. This can be prohibitive, so researchers have developed adaptations to the multiple baseline design that help to overcome this difficulty.

Adaptations of the Multiple Baseline Design There are two major adaptations of the multiple baseline design that are rel- evant to this discussion. These are the multiple probe and delayed multiple baseline designs. Each is designed to contend with issues concerning main- taining baseline data collection with three or more dependent variables simultaneously. However, these adaptations may also include problems that potentially weaken the possibility of demonstrating the functional rela- tionship between the independent and dependent variables.

Multiple Probe Design The multiple probe design (Horner & Baer, 1978) may be used as an adap- tation to designs addressing multiple behaviors, settings, or individuals. The primary variation in the multiple probe design is to decrease the collection of data across multiple baselines (and possibly during follow-up or mainte- nance phases). In this adaptation, the researcher collects data across the multiple baselines at the study’s outset, but does not maintain continuous recording of all baseline measures before the introduction of the interven- tion. Rather, the researcher makes periodic recordings of baseline levels to

224 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

15 Ed

Charles

Lydia 0

Fr eq

ue n

cy o

f Er

ro rs

in S

to ry

te lli

n g

Strategy Instruction

Sessions

Baseline

FIGURE 9-7 Data from

example displaying a multiple baseline across

subjects design

20 14

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 225

ensure that no significant changes have occurred before the introduction of the intervention (see Figure 9-8, which uses data from a previous example in this chapter modified to demonstrate a multiple probe design). The periodic measures are referred to as probes, hence the design’s name. The periodic probes may be used because they reduce the need for resources that may be unavailable to maintain continuous recording of behavior during baseline phases, or because baseline measures are causing severe reactivity, or because there is a strong a priori assumption of stability (e.g., the target behavior is not likely to be emitted until the intervention is introduced, as it does not yet exist in the individual’s behavioral repertoire; Horner & Baer, 1978). Also, once optimal or criterion responding is achieved during an intervention phase, the researcher may resort to data probes to ensure that changes are being maintained. This last procedure is used commonly within many designs and is not unique to the multiple probe design. This is often referred to as a follow-up or maintenance phase and is intended to demonstrate the robustness of the behavior change over a more extended period of time and/or in the absence of the independent variable.

Prediction, verification, and replication are achieved through the same processes discussed with the basic multiple baseline design, with one note- worthy element. Because there is not continuous recording of baseline data, the researcher should be aware that the demonstration of a functional relationship between the independent and dependent variables is at greater risk. For example, if one or more of the probes of a baseline measure appeared inconsistent with other probes, one may have difficulty explaining this phenomenon or providing an adequate argument that covariance was not occurring, or that a stable baseline level of responding had been achieved. Also, this lack of continuous measurement of data may make less obvious a subsequent change in responding following the intervention.

The critical issues in implementing a multiple probe design are the same as those for multiple behaviors, settings, and individuals. The major advantage to the multiple probe design is that fewer resources are required as there is not continuous measurement of multiple baselines. The major disadvantage is that the functional relationship may be more difficult to demonstrate. The researcher may wish to take several precau- tions. First, the researcher must ensure that an adequate number of probes are conducted so that one may easily infer that those probes do represent a true depiction of baseline responding. Second, should a probe result in a measurement that significantly deviates from other measures, the researcher may need to implement continuous recording for that baseline (or at least more frequent recording) to obtain insight into why this may be occurring and to establish a true baseline level of responding (Horner & Baer, 1978). Third, the researcher may wish to conduct a short but continuous baseline measure for each behavior, setting, or individual just

226 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

15 Stephens' Class

Roberts' Class

Michaels' Class

La te

n cy

in M

in ut

Observations

FIGURE 9-8 Data from Sara earlier example displaying a multiple probe design

20 14

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 227

before the introduction of the independent variable to assist in establish- ing a better depiction of baseline level responding. Although in the latter two instances the advantage of reduced need for resources may be diminished or forfeited, the greater disadvantage of failure to demonstrate a functional rela- tionship between independent and dependent variables may be avoided. Con- sider the following example:

Joyce, Tracy, and Josh are all working for a nonprofit organization that employs individuals with mental health issues and prepares them for independent work in businesses in the community. Each has been asked to fill out a job application. None of the three is able to fill out the application. Their social worker decides to implement a strategy by which they can use pre-printed information to complete most applica- tion questions. The social worker recognizes she will be using a multiple probe across subjects design for her study. It is evident that none can complete an application and are unlikely to learn to do so without inter- vention. Joyce is taught the strategy first and is successful within a few sessions of training. The social worker checks Tracy and Josh once each during the intervention with Joyce to confirm they are still unable to complete an application. She then implements the strategy training with Tracy. Tracy also rapidly learns to complete an application. Dur- ing the intervention with Tracy, the social worker does a follow-up probe with Joyce and confirms she can still fill out an application suc- cessfully using the strategy. The social worker also does one more probe with Josh to once again confirm he is not able to fill out the application. The social worker implements the strategy training with Josh who also quickly acquires the ability to complete an application. A couple of weeks after all three have demonstrated their ability to fill out applications, the social worker has them once again do so as a final probe and finds all three have maintained their ability using the strat- egy they were taught.

Delayed Multiple Baseline Design The delayed multiple baseline design also may be used across behaviors, set- tings, or individuals. This design may be employed when inadequate resources are available for continuous recording of baselines, but is usually used when either of two following contingencies arise (Cooper et al., 2007). First, the researcher begins the study with the intention of using a with- drawal design. Some unforeseen occurrence makes the use of such a design impossible (e.g., other persons concerned with the outcomes of the study are concerned about introducing a withdrawal phase). In this situation, the researcher may attempt to salvage the ability to demonstrate a functional

228 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

relationship by introducing the intervention following a baseline on another behavior, in another setting, or with another individual. Second, new behaviors, settings, or individuals could emerge during the study that would appear to possibly be responsive to similar treatment. For example, a student may begin to display a new behavior that is functionally similar to, but independent of, the target behavior. An individual may begin emit- ting the target behavior in settings where it had not previously occurred. Another individual who had not previously done so may begin to emit the same target behavior. Also, other settings or individuals in need of treatment may become available if principals concerned with the original study are convinced that the treatment is effective and should be applied with other behaviors, settings, or individuals. In these contingencies, the delayed multi- ple baseline design is probably not planned a priori, which presents a sepa- rate difficulty with prediction, verification, and replication. Following is an example of a delayed multiple baseline design.

Ms. Johnson is a first-year teacher who works with fourth-grade stu- dents. She has had considerable difficulty in managing Robert’s behav- ior. Mr. Hilary has been asked to assist her with her problem. After observing and recording contingencies in the class, the two agree as do Robert’s parents, that Robert is a student with learning difficulties whose behavior is much in need of changing. Robert is very disruptive by making loud and inappropriate comments during class. Mr. Hilary and Ms. Johnson have agreed to use differential reinforcement to diminish disruptive talking out. After four sessions, the disruptive behavior has dropped to an acceptable level of no more than one response per class period. Unfortunately, Robert has begun disrupting in Ms. Davis’s music class now, where heretofore he had been quite pleasant. Mr. Hilary con- fers with Ms. Davis and the two measure baseline levels of disruptive behavior, implement the intervention, and meet with success in only five sessions. Meanwhile, the low level of responding has been maintained in Ms. Johnson’s class. Finally, Robert begins to exhibit the same disruptive behavior in his class with Mr. Michaels. Again, Mr. Hilary obtains base- line data and the intervention is again successful and continues to be so in all settings. In this scenario, new settings emerged that were not evident at the outset of the study. Figure 9-9 depicts the results of this delayed mul- tiple baseline study.

The major advantages to the delayed multiple baseline design is that it may allow the use of fewer resources and it may allow the researcher to extend the study to new behaviors, settings, and individuals that had not been targeted a priori. Cooper et al. (2007) noted three limitations to delayed designs. First, delaying treatment for other behaviors, in other set- tings, or with other individuals may be problematic, although this difficulty

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 229

15 Johnson's Class

Davis' Class

Michael's Class

Fr eq

ue n

cy o

f D is

ru p

ti ve

T al

ki n

g O

Observations

Baseline

FIGURE 9-9 Data from Robert example

displaying a delayed multiple baseline design

20 14

230 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

is inherent in multiple baseline designs in general. Second, fewer data points may have been gathered and the length of the various baselines may differ. Third, the use of delayed baselines when new behaviors, settings, or individuals emerge may mask the effects of the independent variable on the dependent variable. That is, the researcher is less likely to demonstrate that each baseline remains unchanged even when the treatment is introduced to other baselines. The fact that new possibilities emerge may suggest some influence occurring in the study that the researcher may have difficulty explaining. As Cooper et al. (2007) stressed, the best use of the delayed mul- tiple baseline design may be to add tiers to an already existing multiple base- line design, which enhances further the demonstration of a functional relationship established during the course of the study as originally planned.

4 C H E C K I T O U T # 3 Ariel is a young woman who is consuming too many soda pops. Her mother reports Ariel’s cavities have increased and her dentist has warned her if she con- tinues to drink so many sugary beverages, she will lose her teeth at an early age. The social worker and psychologist work with Ariel and she does drastically reduce the number of sodas she consumes each day. However, Ariel’s mother reports Ariel has now started to consume considerable amounts of candy. The social worker and psychologist measure her candy consumption for only 3 days and then work with Ariel and she reduces her candy consumption to an accept- able level. Then, Ariel begins eating ice cream several times a day. Again, the social worker and psychologist take 3 days of baseline data and then intervene using the same treatment strategy as with sodas and candy. Ariel then reduces her ice cream consumption. Over the following few weeks, the social worker and psychologist pick one day randomly each week to record Ariel’s soda, candy, and ice cream consumption. They find each has maintained at acceptable levels. Which adaption of the multiple baseline design was used?

Summary Multiple baseline designs are versatile and relatively easy to understand. They are found frequently in the literature and are perhaps the most com- mon design in use today. As the reader will note in Chapter 10, pure designs are not always found when one reviews a study in which a multiple baseline design was used. Changing criteria, changing conditions, and even withdra- wals are sometimes found, in addition to the alternatives discussed here in Chapter 9. Do not let these variations throw you as you review the research literature. Rather, take note of how well these designs may be manipulated and altered to meet the needs of individuals in many settings.

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 231

Key Concepts/Terms Basic goal of multiple baseline design—Demonstration of a functional

relationship between the target behavior and intervention by replicating the intervention effects with two or more behaviors, in two or more settings, or with two or more individuals.

Basic design—The basic multiple baseline design includes two or more A-B designs where baseline data are simultaneously measured and the inter- vention is introduced to one behavior, in one setting, or with one individual at a time.

Prediction—After baseline data are stable, the prediction would be that there would be no change in the data path for the dependent variables if there was no intervention effect.

Verification—When the intervention is implemented, the data path changes predictably for the dependent variable.

Replication—The prediction and verification are repeated for each depen- dent variable.

Covariance—This occurs when baseline measures change in the same direc- tion as during an intervention phase although the intervention has not yet been implemented with that dependent variable; this is a threat to internal validity.

Advantages—Withdrawal of treatment is not required; sequential imple- mentation of the independent variable parallels the practice of teachers; generalization of behavior change is monitored within the design; the design is easily conceptualized and used.

Disadvantages—Possibility of covariance of dependent variables; the dem- onstration of the functional relationship is not as direct as in a with- drawal design; there is typically only one intervention phase with each dependent variable, reducing the opportunities to study the functional relationship; requires substantial resources to maintain multiple base- lines and is time consuming.

Multiple baseline across behaviors design—The same intervention is applied to similar behaviors in the same individual in the same setting.

Multiple baseline across settings design—The same intervention is applied to the same behavior in the same individual in different settings.

Multiple baseline across subjects design—The same intervention is applied to the same or similar behaviors, in the same setting, to different individuals.

Adaptations Multiple probe design—May be adapted to any of the basic designs; data

probes are taken during baselines rather than continuous measurement; reduces need for resources; potentially leads to problems if probes are too infrequent or do not suggest steady baseline responding.

232 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Delayed multiple baseline design—Used when a withdrawal design is no longer possible or when other behaviors, settings, or individuals emerge that are in need of intervention; baselines are not measured simulta- neously; potential problems in demonstrating a functional relationship.

4 Possible Answers to Check It Out

¶ Because you will be working with the same individual with the sametarget behavior in three different environments (job, home, therapy session), then a multiple baseline across settings would be appropriate. You might also consider which environment should first have the intervention introduced into it, which second, and which third based on the severity and need in each environment.

· There are two issues that would suggest the use of a multiple baselineacross behaviors would not be appropriate. First, and perhaps more importantly, the behavior is so disruptive that intervention is required immediately making extended baseline phases untenable. Second, while disruptions might be further defined as comments, noises, and questions, these are so similar that one might well anticipate they would covary and change in accordance with the introduction of the intervention to any of the three behaviors. Actually, a B-A-B design might be more appropriate with a very limited baseline phase in the withdrawal of the intervention.

¸ The social worker and psychologist used a delayed multiple baselinedesign. Because new but functionally similar behaviors emerged (candy and ice cream consumption), there was a need to address these new target behaviors. Interestingly, the final follow-up phase of collecting data ran- domly one day a week also provides an example of using multiple probes.

References Alberto, P. A., & Troutman, A. C. (2013). Applied behavior analysis for teachers

(9th ed.). Boston: Pearson. Baer, D. M., Wolf, M. W., & Risley, T. R. (1968). Some current dimensions of

applied behavior analysis. Journal of Applied Behavior Analysis, 1, 91–97. Barlow, D. H., Nock, M. K., & Hersen, M. (2009). Single-case experimental

designs: Strategies for studying behavior change (3rd ed.). Boston: Pearson. Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis

(2nd ed.). Upper Saddle River, NJ: Pearson- Merrill-Prentice-Hall. Datilo, J., Gast, D. L., Loy, D. P., & Malley, S. (2000). Use of single-subject

research designs in therapeutic recreation. Therapeutic Recreation Journal, Third Quarter 2000.

Hammond, D., & Gast, D. L. (2010). Descriptive analysis of single subject research designs: 1983–2007. Education and Training in Autism and Developmental Disabilities, 45, 187–202.

CHAPTER 9 OVERVIEW OF MULTIPLE BASELINE DESIGNS 233

Harvey, M. T., May, M. E., & Kennedy, C. H. (2004). Nonconcurrent multiple baseline designs and the evaluation of educational systems. Journal of Behav- ioral Education, 13, 267–276.

Hersen, M., & Barlow, D. H. (1975). Single-case experimental designs: Strategies for studying behavior change. New York: Pergamon Press.

Horner, R. D., & Baer, D. M. (1978). Multiple-probe technique: A variation of the multiple baseline, Journal of Applied Behavior Analysis, 11, 189–196.

Kucera, J., & Axelrod, S. (1995). Multiple-baseline designs. In S. B. Neuman & S. McCormick (Eds.), Single-subject experimental research: Applications for literacy (pp. 47–63). Newark, DE: International Reading Association.

Tankersley, M., Harjusola-Webb, S., & Landrum, T. J. (2008). Using single-subject research to establish the evidence base of special education. Intervention in School and Clinic, 44, 83–90.

Tawney, J. W., & Gast, D. L. (1984). Single subject research in special education. Columbus, OH: Merrill.

234 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

CHAPTER

10 Application of Multiple Baseline Designs

235

In this chapter we will discuss studies related to multiple baseline designs.The studies will demonstrate multiple baseline designs across behaviors,across settings, across subjects and a multiple probe design. In the interest of space, we have not included other adaptations (e.g., delayed multiple baseline design). These examples from the literature are more representative of commonly used and “classic” multiple baseline designs.

We will present studies that lead to both increases and decreases in the dependent variable (target behavior) that are related to educational and other needs of individuals. The background information for and various aspects of each study will be presented in a chart that describes the subjects, purpose of the study, setting, independent and dependent variables, results, limitations, and any interesting aspects of the study related to the use of the design.

Multiple Baseline Across Behaviors Mazzotti, V. L., Test, D. W., Wood, C. L., & Richter, S. (2010). Effects of

computer-assisted instruction on students’ knowledge of postschool options. Career Development for Exceptional Individuals, 33(1), 25–40.

Purpose of the Study The purpose of this study was to evaluate the effects of computer-assisted instruction (CAI) on high school students’ with mild to moderate intellectual disabilities knowledge of postschool options in employment, education, and independent living.

Subjects Four Caucasian high school students with prior experience in computer use participated in the study. Their ages ranged from 16 to 19 years, and their reading levels varied from second to fifth grade. All of them knew how to use computers and answer simple questions. Rick, a 17-year-old male, had autism and an intellectual disability. Dylan, a 16-year-old male, had an intellectual disability and motor delays. As a result, he had difficulty in writing. Kim, a 19-year-old female, had a mild intellectual disability. Finally, Mary Lou, a 17-year-old female, had moderate intellectual disability.

Setting The study took place in a nonprofit private school located in a church. The study began during summer school and the intervention was provided in a life skills classroom. The study extended over six weeks into the regular school year.

236 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Dependent Variables The subjects’ knowledge of postschool options in employment, education, and independent living measured by a probe checklist of 13 items and scored on a 30 point scale was the primary dependent variable. Students’ responses to items (e.g., “What does living in a supported apartment mean?” p. 28)* could be scored as incorrect, partially correct, or correct. Some items were scored as either incorrect or correct only. A measure of setting and situation generalization (data were not plotted on the graph) was the secondary depen- dent variable. Checklist probes served as a pretests and posttests used prior to baseline and again at the end of the study and included six items to measure increased ability to make informed choices (e.g., “What do you want to do to further your education after high school?” p. 29)**. Both interrater reliability and social validity measures were taken. Social validity measures were obtained from the participants and special education teachers.

Independent Variable A laptop computer with Microsoft PowerPoint was used to provide the CAI program to increase students’ knowledge of postschool options. This soft- ware was used to create the visual and audio effects of the intervention. Treatment fidelity measures were used to ensure each student could navigate the CAI program independently.

The Design A multiple baseline across behaviors design that was replicated across parti- cipants was used in this study. The participants’ knowledge of employment, education, and independent living were assessed during baseline and inter- vention sessions. Predetermined criteria were established for implementing the CAI with each new target behavior.

Intervention Prior to baseline, a pretest was conducted to assess the participants’ knowledge of postschool options in order to make informed decisions. Then, a minimum of three baseline probes was administered to determine their knowledge of postschool options before the intervention. These data were used to identify which targeted area (i.e., employment, education, or independent living) was lowest in existing knowledge with a stable baseline. At first the CAI consisted of all three outcome areas as a single intervention called CAI 1 which was not yielding the desired outcomes. As a result, the researchers made a decision to

*,**Mazzotti, V. L., Test, D. W., Wood, C. L., & Richter, S. (2010). Effects of computer-assisted instruction on students’ knowledge of postschool options. Career Development for Exceptional Indivi- duals, 33(1), 25–40.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 237

provide instruction on each outcome area (called CAI 2) independently for each target behavior and provide intervention to all four participants at the same time. CAI 2 included visual and audio components presented in a model-lead-test format. This instructional format was used for each dependent variable (i.e., employment, education, independent living). Once the partici- pant met the predetermined criteria of 8 out of 10 correct responses for two consecutive sessions on that target behavior, the intervention was stopped and the participant was moved to the maintenance phase. Booster sessions were conducted for two participants as they did not maintain mastery of target behaviors for two consecutive sessions during maintenance.

Obtaining the Data and Plotting the Results The researcher assigned 10 points for each dependent variable and mea- sured the participants’ responses item-by-item based on each oral response to the questions. Based on a 30-point probe checklist, the researcher scored the participants’ correct responses to the probe questions as 0 points for incorrect and 1 point for correct responses. Also, the researcher trained a doctoral student to collect interrater reliability data on 30% of the probes for baseline, intervention, and maintenance phases on an item-by-item scoring sheet. The interrater reliability for the intervention yielded an overall mean of 91%. The numbers of correct responses to the 30-point probes were graphed using 10 points for employment, 10 points for education and 10 points for independent living. Each graph showed the number of correct responses for the primary dependent variable for each participant (see Figures 10-1, 10-2, 10-3, 10-4). Treatment fidelity measures yielded a mean of 95.9%.

Results Rick’s results indicate he achieved the predetermined criteria level in each target behavior area although his maintenance results were variable (see Figure 10-1). Dylan also achieved the predetermined criteria level and maintained a high and stable performance during maintenance phases (see Figure 10-2). Similarly, Kim achieved the predetermined criteria and also maintained her gains (see Figure 10-3). Mary Lou also met the criteria in each area but had a more variable performance during maintenance (see Figure 10-4). The authors indicated a functional relationship between the independent and dependent variables was achieved with all 4 subjects.

Why Use a Multiple Baseline Across Behaviors Design? Interestingly, the researchers wanted to evaluate the use of the CAI 1 pro- gram on knowledge of postschool options for employment, education, and independent living across four participants within a multiple probe design.

238 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

When it became apparent that employment, education, and indepen- dent living knowledge measures would need to be treated separately (the CAI 2 intervention), the study was altered to a multiple baseline across behaviors design. This design allowed the researchers to administer the

BL CAI1 CAI2 Maintenance Booster session

Employment 0 1 2 3 4 5 6 7 8 9

Sessions

N um

b er

C or

re ct

- R

ic k

Booster session

0 1 2 3 4 5 6 7 8 9

5 10 15 20 1-week 2-week 3-week

1 2 3 4 5 6 7 8 9

Education

Independent Living

FIGURE 10-1 Data from the number correct on computer-

assisted instruction for Rick in the multiple baseline across behaviors design study. Note. From V. L.

Mazzotti, D. W. Test, C. L. Wood, and S. Richter, Career Development for

Pro-Ed Inc. Reprinted by Permission of SAGE

Publications.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 239

independent variable to each dependent variable when, and only when, the predetermined criteria had been met on one of the dependent vari- ables. The researchers did implement the intervention with all four participants simultaneously. Also, through the use of this design, the researchers were able to demonstrate the functional relationship between

Baseline CAI2 Maintenance

Employment 0 1 2 3 4 5 6 7 8 9

Sessions

N um

b er

C or

re ct

- D

yl an

0 1 2 3 4 5 6 7 8 9

5 10 15 20 25 1-week 3-week

1 2 3 4 5 6 7 8 9

Education

Independent Living

FIGURE 10-2 Data from the number correct on computer- assisted instruction for Dylan in the multiple

baseline across behaviors design study. Note. From V. L. Mazzotti, D. W. Test, C. L. Wood, and S. Richter,

Career Development for Exceptional, 33(1), p. 32,

Permission of SAGE Publications.

240 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

the intervention and improved knowledge of postschool options across behaviors (employment, education, and independent living) and also across the four subjects providing additional replication and verification of results.

0 1 2 3 4 5 6 7 8 9

Employment

0 1 2 3 4 5 6 7 8 9

Education

0 1 2 3 4 5 6 7 8 9

Independent Living

Sessions 5 10 15 20 25 1-week 3-week

N um

b er

C or

re ct

- K

Baseline CAI2 MaintenanceFIGURE 10-3 Data from the number correct on computer-

assisted instruction for Kim in the multiple baseline across behaviors design study. Note. From V. L.

Mazzotti, D. W. Test, C. L. Wood, and S. Richter, Career Development for

Pro-Ed Inc. Reprinted by Permission of SAGE

Publications.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 241

Limitations of the study The study had limitations identified by the authors. First, the CAI 2 pro- gram was difficult for two participants who required more sessions to reach the predetermined criteria levels and subsequently also produced vari- able maintenance data. Second, because of the small number of subjects and

Baseline CAI2 Maintenance

Employment

Sessions 5 10 15 20 25 1-week 3-week

0 1 2 3 4 5 6 7 8 9

N um

b er

C or

re ct

- M

ar y

Lo u

0 1 2 3 4 5 6 7 8 9

Booster session

Education

Independent Living

FIGURE 10-4 Data from the number correct on computer- assisted instruction for

Mary Lou in the multiple baseline across behaviors design study. Note. From V. L. Mazzotti, D. W. Test, C. L. Wood, and S. Richter,

Career Development for Exceptional, 33(1), p. 32,

Permission of SAGE Publications.

242 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

the setting, generalizability of results may be limited. Third, the researchers did not obtain setting/situation generalization data they would have pre- ferred (actual choice-making during the Individualized Education Program process). Finally, there were two additional concerns with specific materials used and perceptions on the social validity measure administered to the special education teachers that are discussed in the article.

Summary A summary of the relevant components of this study can be found in Table 10-1.

Multiple Baseline Across Settings Cushing, L. S., & Kennedy, C. H. (1997). Academic effects of providing

peer support in general education classrooms on students without dis- abilities, Journal of Applied Behavior Analysis, 30, 139–151.

Purpose of the Study The purpose of this study was to examine whether peer support provided by a student without disabilities to a student with disabilities would have posi- tive or negative effects on academic engagement and associated measures for the student without disabilities.

TABLE 10-1 Summary of “Effects of

Computer-Assisted Instruction on Students’ Knowledge of Postschool

Options.”

FEATURE DESCRIPTION

Type of design Multiple baseline across behaviors design

Purpose of the study Evaluate the effects of CAI on high school students’ knowledge of postschool options in employment, education, and independent living

Subjects Four high school students with mild to moderate intellectual disabilities

Setting A life skills classroom in a private school located in a church

Dependent variable Participants’ knowledge of postschool options in employ- ment, education, and independent living as measured through specific checklist items (8 of 10 correct as predetermined criteria level)

Independent variable The CAI program that was independently used by each participant on a laptop computer following training

Results and outcome CAI program increased the knowledge of postschool options for all four participants in each of the three option areas. It was an effective strategy in general

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 243

Subjects The subjects in the part of the study being presented here were two 11- year-old students, Louie and Leila. It should be noted that the study included other student dyads, although, for our purposes, we are focusing on only on Louie and Leila because of the design used. Leila was identified as a student with moderate intellectual disabilities, and Louie was identified as a peer to work with Leila. Louie was selected because he was in the same class as Leila, he had previously expressed a desire to work with a student with disabilities, and his engagement in class activities was below average. Leila was identified as sociable but she had a limited vocabulary and arti- culation difficulties. Louie was identified as being occasionally disruptive, was not turning in assignments, was receiving very low grades, and had difficulty paying attention.

Setting The study was carried out at a suburban intermediate school with 1,100 students from diverse backgrounds. Students in the school who were eligible for special education were provided with services while participating full- time in general education settings. The English, science, and social studies classes served as the three experimental settings.

Dependent Variable The dependent variable in this study was percentage of time academically engaged for Louie, which was defined as attending to ongoing classroom activities, engaging in work-related assignments, or both. A 1-minute momen- tary time-sampling recording procedure was used (the subject was observed for 1-second at the end of each minute throughout the 55-minute class period). The observers were special education personnel who had been trained in the use of the observational system. A Likert-type item scale was also used to assess adult’s perceptions of Louie’s performance.

Independent Variable The independent variable was the peer support system of academic engage- ment that included Louie’s participation with Leila, training and supervision by a special education teacher, and supervision by the general education teacher. More specifically, Louie worked with Leila on assignment comple- tion, classroom participation, and adaptation of assignments. Special educa- tion personnel commented on Louie’s performance approximately once every 10 minutes. Assistance was provided by personnel if Louie had diffi- culty with course content or adapting content, and brief daily feedback was

244 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

provided regarding his performance. In addition, general education teachers were asked to praise Louie at least once in each class when he was serving as a peer support.

Design A multiple baseline across settings design was used in this study for this particular student dyad (other designs were used with other dyads). The intervention was sequentially applied in each setting (English, science, and social studies classes) as Louie’s academic engagement increased from base- line levels.

Intervention The intervention applied in this study was the peer support provided to Leila by Louie. This support centered around special education personnel teaching Louie how to interact with Leila, including behavioral management strategies, adapting assignments to meet Individualized Education Program goals, and communication strategies. For example, Louie was taught to adapt assignments through verbal descriptions, modeling, and praise for correct performance. He also took notes for Leila and revised those notes to accommodate her. Training in adapting assignments was provided by special education personnel and occurred over several days.

Obtaining the Data and Plotting the Results The percentage of time academically engaged for Louie when he worked alone was recorded across different settings, which included his English, science, and social studies classes. These data served as baseline measures, and observations were conducted under routine class conditions. The baseline data were com- pared to the observations conducted when Louie served as a peer supporter for Leila in each of the three settings. The number of observations conducted during baseline and peer support varied for each setting. A graphic depiction of the results of this study is presented in Figure 10-5. The broken line in the data path included in the social studies graph indicates an absence for Louie.

Results The results indicated that the percentage of time in which Louie was academically engaged increased during the periods he was serving as a peer support for Leila. Also, a Likert-type scale was used to measure adults’ perceptions of Louie’s classroom performance. These data also indicated consistent overall increases for Louie across a variety of behaviors, including listening to and following directions, participating in activities, completing

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 245

assignments, bringing supplies, and following rules. The data from this scale were intended to lend increased validity to the quantitative changes observed.

Why Use a Multiple Baseline Across Settings Design? For this study, it would have been a less convincing demonstration of a functional relationship if observations were conducted in only one setting. Because the settings were similar but functionally independent of one

Louie Working Alone Louie Supporting Leila 100

Days

Pe rc

en ta

ge o

f T im

e A

ca d

em ic

al ly

E n

ga ge

5 10 15 20 25

English

100

Science

Social Studies

FIGURE 10-5 Data from the multiple baseline across settings

study. Note. From “Academic Effects of

Providing Peer Support in General Education

Classrooms on Students without Disabilities,” by L. S. Cushing and C. H.

Society for the Experimental Analysis of Behavior. Reprinted with

permission.

246 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

another, the multiple baseline design offered the advantages of providing the intervention in all settings of need without having to withdraw treatment. It is worth noting that covariance among the dependent variables was not evident, particularly in the social studies class, which was the last setting in which the intervention was implemented. There was also a notable increase in percentage of time engaged immediately following implementation of the intervention, particularly in the last two settings. This strengthened the demonstration of the functional relationship.

Limitations of the Study One limitation of the study is the selection of students used for the study. The authors contend that because the sample of students was so selective, the robustness of the findings requires systematic replication across a variety of students without disabilities. Another area of concern is whether positive effects would occur for students who are already performing at high levels in general education classes. In this case, Louie was not performing to his perceived potential, and, as a result, there was an expectation that this inter- vention would result in improvement.

TABLE 10-2 Summary of “Academic Effects of Providing Peer

Support in General Education Classrooms on

Students without Disabilities.”

FEATURE DESCRIPTION

Type of design Multiple baseline across settings design

Purpose of the study

To examine whether peer support provided by a student without disabilities to a student with disabilities would have positive or negative effects on academic engagement and associated measures for the student without disabilities

Subjects Two 11-year-old students, Louie (student without disabilities) and Leila (student with disabilities)

Setting Suburban intermediate school with 1,100 students from diverse backgrounds; English, science, and social studies classes

Dependent variable

Percentage of time academically engaged for Louie, defined as involvement in on-going classroom activities and/or work-related assignments; a Likert-type scale was used to evaluate adult’s perceptions of Louie’s performance

Independent variable

The peer support system of academic engagement for Louie that included participation with a student with disabilities, Leila, and training and/or supervision from special education and general education teachers

Results and outcomes

The percent of time in which Louie was academically engaged increased during the periods he was serving as a peer support for Leila; adults’ perceptions were that Louie’s performance in those classes improved on a variety of behaviors

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 247

Summary This study is summarized in Table 10-2. Please note that withdrawal designs were also included in this study, although we chose to focus on Louie and Leila because this represented a multiple baseline across settings design.

Multiple Baseline Across Subjects Bennett, K. D., Ramasamy R., & Honsberger, T. (2012). The effects

of covert audio coaching on teaching clerical skills to adolescents with autism spectrum disorder. Journal of Autism and Developmental Disorders. DOI 10.1007/s10803-012-1595-6.

Purpose of the Study The purpose of the study was to determine the effects of covert audio coach- ing (CAC) on the development and maintenance of employment skills of high school students with autism spectrum disorders (ASD).

Subjects Three high school boys diagnosed with ASD participated as subjects in this study. All three were in a special diploma program in a special school for students with ASD. The students spent their days primarily learning daily living and employment skills. All of them were able to communicate, follow directions and answer simple questions. Subject one was 13-years-old, sub- ject two was 22-years-old, and subject three was 16-years-old; all subjects were male. Subjects were selected through teacher nomination and specific criteria for participation in the study.

Setting The study took place in the school’s faculty lounge. This large lounge was also used as a workroom for students learning employment skills. The room had tables, chairs, computers, mailboxes, a copy machine, and a laminating machine. The room was selected because the employment skills being taught during the study were related to copying.

Dependent Variable The dependent variable was the percentage of task analysis steps correctly completed in making photocopies. A 10-step task analysis was repeated for each copy job that varied from 3 to 6 jobs per session.

248 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Independent variable Privately delivered, real-time performance feedback given through two-way radios and an ear bud speaker served as the independent variable. The authors referred to this as covert audio coaching (CAC). The performance feedback included support statements, antecedent prompts, and correction statements.

The Design A multiple baseline across subjects design was used in this study. The authors collected repeated measures of photocopying performance on the 10-step task analysis.

Intervention During baseline, each subject was brought into the workroom one at a time and verbally instructed to make some photocopies. No coaching statements were provided and the participants were allowed to make mistakes. When- ever a participant asked for help, he was prompted to do his best and allowed to complete the task steps even if errors occurred. When the partic- ipant made continuous errors for 10-seconds on a critical step (e.g., placing the papers to be copied below the feeder instead of placing them in the feeder), one of the researchers stopped the copy job and instructed the par- ticipant to start the next copy job. For one or two baseline sessions, each participant wore the two-way radio and earbud speaker to probe for any reactivity to the equipment. Baseline sessions were conducted for stability. Stability “was defined as 80% of the data occurring within 20% of the median for at least five sessions” (DOI 10.1007/s10803-012-1597-6). The intervention was implemented to the participant whose baseline data stabilized first. During intervention sessions, support statements were deliv- ered at the end of each copy job. Antecedent prompts were immediately given once a participant made two consecutive errors before the third chance to perform that step. Correction statements were delivered as a consequence whenever an error occurred. All these coaching statements were delivered approximately 23 feet away from the participants so that these statements could be heard only through the ear bud speaker. The intervention sessions continued until a participant achieved 90% of the 10 steps completed correctly for at least five successive sessions. Follow-up probes were continued after the termination of the intervention.

Obtaining the Data and Plotting the Results For interobserver agreement, the study authors independently recorded a mean agreement of 98.4% for the dependent variable. For procedural fidel- ity, two authors independently observed the coach’s delivery of performance

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 249

feedback. The agreement was 100% that the coach did not deliver coaching statements during baseline and maintenance sessions (when those statements should not occur). There was a mean 97.4% agreement on all coaching statements delivered during intervention conditions. During other baseline, intervention, and probe sessions, one observer recorded the percentage of correct steps completed for each 10 step task analysis job. Data were plotted on a graph to show the percent of task analysis steps completed correctly by each participant across all three phases of the study (See Figure 10-6).

Results As can be seen in Figure 10-6, during baseline sessions each participant’s data showed ascending trends but were stabilized before CAC was intro- duced. The reason for the ascending trend was attributed to the participants’ previous exposure to operating copy machines through their teachers. When CAC was delivered to subject one, his accuracy level increased to 98–100% within six sessions and he maintained above 90% during a three-week follow-up period. Similarly, once CAC was introduced, subject two’s accu- racy steadily increased to 100% within four sessions and he maintained his skills at the same level for three weekly follow-up probes. When the CAC was introduced, subject three’s performance level increased to 90% within two sessions and maintained at 98-100% for the remaining four sessions. However, when CAC was discontinued, his performance decreased, and he was unable to maintain his skills. As a result, CAC was reintroduced and his accuracy increased immediately to 100% and he maintained it for five con- secutive sessions. For the three weekly follow-ups, he maintained his accu- racy between 98–100%.

Why Use a Multiple Baseline Across Subjects Design? This design allowed the researchers to examine the effectiveness of CAC with three participants who needed to improve their employment skills. Moreover, it was likely that the dependent variable, the percentage of cor- rectly completed task-analysis steps for making photocopies, would not reverse to baseline levels even if CAC was terminated making a withdrawal design untenable. Given that multiple individuals were in need of the employment training, the multiple baseline design across subjects was appropriate.

Limitations of the Study The authors pointed out several limitations to the study. First, generaliz- ability to other subjects, settings, or dependent variables may be limited.

250 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Second, the number of follow-up sessions was limited. Third, CAC in photocopying skills was not provided in any other setting. As a result, it is difficult to conclude whether these participants would use the learned skills to make photocopies in any other work setting.

Summary A summary of the relevant dimension of this multiple baseline across sub- jects can be found in Table 10-3.

Baseline

BIE Probe

CAC Follow-up (weekly) 100

90 80 70 60 50 40 30 20 10

0 5 10 15 20 25 30

Jason

100 90 80 70 60 50 40 30 20 10

0 5 10 15 20 25 30

Shaun

CAC (2) Follow-up (2) 100

90 80 70 60 50 40 30 20 10

0 5 10 15 20 25 30

David

BIE Probe

BIE Probes

Sessions

% T

as k

St ep

s Co

rr ec

FIGURE 10-6 Data from the multiple baseline across subjects

design. Note. From Springer and the Journal of Autism and Developmental Disorders, 2012, “Effects of Covert Audio Coaching on Teaching Clerical Skills to Adolescents with Autism Spectrum Disorder,” K. D. Bennet, R. Ramasamy, and T. Honsberger, © 2012,

With kind permission from Springer Science and

Business Media.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 251

Multiple Probe Design Hines, S. J. (2009). The effectiveness of a color-coded, onset-rime decoding

intervention with first-grade students at serious risk for reading disabil- ities. Learning Disabilities Research & Practice, 24(1), 21–32.

Purpose of the Study The purpose of this study was to examine the effectiveness of an onset-rime instruction program for mastery and transfer of word reading skills.

Subjects Four first-grade students that were referred by their teachers as the most at-risk for reading failure served as the subjects in this study. The four sub- jects were all 6 years old; two were male and two female. They were ethni- cally diverse. None of these students were identified as having disabilities.

Setting The training sessions were provided in an empty room in the same school near the first-grade classrooms. This setting allowed privacy to the students, avoided distractions to the students, and avoided generalization of the train- ing to other participants.

Dependent Variables There were three dependent variables identified in this study; (1) the ability to read twenty randomly presented consonant/vowel/consonant (CVC) instructional words and consonant/vowel/consonant/consonant (CVCC)

TABLE 10-3 Summary of “The Effects of Covert Audio Coaching on Teaching Clerical Skills to Adolescents with Autism

Spectrum Disorder.”

FEATURE DESCRIPTION

Type of design Multiple baseline across subjects design

Purpose of the study Determine the effects of CAC on employment skill development and maintenance

Subjects Three high school males diagnosed with ASD

Setting School’s faculty lounge/workroom

Dependent variable Percentage of correctly completed steps in a 10-step task analysis

Independent variable Coaching statements provided through CAC

Results and outcomes CAC was effective in developing and maintaining (photocopying) skills

252 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

instructional words from the eight short a and short e rime word patterns; (2) the ability to read eight randomly selected uninstructed short a and e CVC/CVCC words from instructed rime patterns for near-transfer; and (3) the ability to read six randomly selected short a and e CVC/CVCC words from uninstructed rime patterns for far-transfer. Data were graphed as the percentage of words read correctly for each dependent variable.

Independent Variable The independent variable was a color-coded, onset-rime reading program that targeted the reading skills measured by the dependent variables. The intervention materials were all books that were organized by color-coded rime patterns from a published reading series.

The Design This study used a multiple probe across subjects design. This design is a varia- tion to the multiple baseline design and allows the researcher to gather baseline data through intermittent probes rather than extended continuous recording.

Intervention Each child received individual instruction from the author 4–5 times a week. This avoided the generalization of the training to other participants. During each session, the subject read one of the books from the reading series, while the author followed a detailed intervention script. Following a correction pro- tocol, subjects reread books with which they had difficulty. After the instruc- tional procedure, each subject was presented with words from the dependent variable lists. Subjects were prompted to sort the color-coded words into word families, read the words, and then read the words again with random presentation. The researcher corrected subjects’ errors. The instructional pro- cedure was then repeated without color-coding.

Obtaining the Data and Plotting the Results The researcher recorded the percentage of words read correctly for each of the dependent variables. Baseline and post training measures were audio recorded and an independent rater listened to and scored one-third of data collection sessions for the instructional dependent variable, and 100% of the data collection sessions for the near- and far-transfer dependent variables. Interrater reliability was calculated and averaged for each of the three mea- sures. Reliability averages ranged from 86% to 91% for the dependent vari- ables. Treatment fidelity was measured by an independent rater listening to audio recordings of the sessions. The first two sessions with subject one were scored and then 33% randomly selected sessions across subjects were

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 253

scored. The overall average treatment fidelity reliability was 96%. Data were plotted on graphs to show the percentage of words read correctly for each of the four subjects on three measures administered during baseline, after intervention, at 1-week and 1-month maintenance (See Figure 10-7). Data collected for the two transfer dependent variables were included in the study in bar graphs but are not presented here.

Results The results indicated that each subject made substantial gains on the first dependent variable (the average correctly read instructional words). The overall average increase in words read correctly across subjects from base- line was 73%. Gains were maintained during the maintenance probes. Gains were also evident in the near-transfer dependent variable but not for three subjects on the far-transfer dependent variable. Again, the results from the near and far transfer variables are not graphically presented here.

Why Use a Multiple Probe Design? As discussed in chapter 9, the major advantage of the multiple probe design is to allow the researcher to collect baseline (and often follow-up or maintenance) data through intermittent probes rather than continuously. In this study, the target behaviors were not likely to improve significantly from baseline without intervention. Therefore, this was an appropriate design for this study. It is important to note the researcher collected data continuously on all subjects prior to intervention. This is a recommended practice to enhance the researcher’s ability to compare performance across the baseline and intervention phases.

Limitations of the Study This study had limitations. First, the researcher collected all data rather than having an independent rater. However, interrater and treatment fidelity measures indicate this was not a significant threat to internal validity. Sec- ond, because the intervention used several components, it was not possible to discern precisely if the color coding of words had an effect. Third, the researcher could not draw any conclusions as to the relative effectiveness of this intervention versus another program. Finally, there were a limited number of probes during maintenance phases.

Summary A summary of the relevant dimensions of this multiple probe design study is found in Table 10-4.

254 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

% C

or re

Sessions

Tammy

Arthur

Maria

1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031 0

10 20 30 40 50 60 70 80 90

100

1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031 0

10 20 30 40 50 60 70 80 90

100

1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031 0

10 20 30 40 50 60 70 80 90

100

Instructional Words

1 2 3 4 5 6 7 8 9 10111213141516171819202122232425262728293031 0

10 20 30 40 50 60 70 80 90

100 Baseline Post-Int. Maint.

John

FIGURE 10-7 Data from the multiple

probe design. Note. From “Effectiveness of Color-

Coded, Onset-Rime Decoding Intervention with

permission of John Wiley & Sons Ltd.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 255

Application Practice: Donald Donald is a third-grade student who has been enrolled in the same school since kindergarten. Almost from the outset of his schooling, Donald was a struggling reader who was lagging developmentally behind his peers in language arts and math. Also, Donald’s parents and teachers noticed that Donald did not appear to interact with his peers beyond what was necessary to complete assignments or tasks in school. He ate lunch by himself in the school cafeteria. His parents reported that Donald was not engaged in any extracurricular activities despite their encouragement. At home, Donald preferred to play video games and watch television. His parents, in an effort to further encourage him to play and interact more with other children, restricted his time allowed for gaming and television. Despite their efforts, Donald still did not engage in play or interaction with other children by choice.

Donald was interviewed by his school counselor and the school psychol- ogist to determine if Donald himself was aware of his limited interactions and what was the reason for this. Donald reported to them he didn’t know what to say to other children. Donald’s educational team decided that Donald could benefit from strategy instruction in pragmatic language use. The team had noted that while Donald could answer questions and engage in a conversation focused on specifics, such as how to complete a task or

TABLE 10-4 Summary of

“The Effectiveness of a Color-Coded, Onset-Rime

Decoding Intervention with First-Grade Students at Serious Risk for Reading

Disabilities.”

FEATURE DESCRIPTION

Type of design Multiple probe design

Purpose of the study Examine the effectiveness of an instructional program for mastery and transfer of word reading skills

Subjects Four first-grade students who were at-risk for reading failure

Setting An empty classroom near the first-grade classrooms in an elementary school

Dependent variables 1. The correct percentage of twenty instructional words from eight short a and e rime patterns; 2. The correct percentage of eight uninstructed short a and short e words from instructed rime patterns for near-transfer; and 3. The correct percentage of six short a and e words from uninstructed rime patterns for far-transfer

Independent variable A color-coded onset-rime reading instruction program

Results and outcomes All four subjects made gains on the first dependent variable which is graphically presented here.

256 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

play a video game, he was very limited in his ability to engage in social conversations that relied on the give and take of questions, answers, topics, and initiating and ending a conversation.

The team decided to try out the pragmatic strategy instruction in his homeroom class as this was the more social but still controlled environ- ment in his school day. The team also decided that if the pragmatic strat- egy instruction was successful in his homeroom class period, the team would extend it to his language arts class period, his math period, and finally to his lunch period. The school psychologist and counselor each observed Donald in these various school periods to record how often Donald engaged in a social conversation. After a period of 5 school days and recording no more than 1 social conversation during any of the 4 periods (homeroom, language arts, math, and lunch), the team decided it was appropriate to implement the pragmatic strategy instruction in his homeroom period.

With the implementation of the strategy instruction, Donald’s social conversations increased in frequency to 2, 3, 4, 4, and 4. The team believed this was very quick success indeed. Earlier observations had revealed that 4 social conversations during homeroom were typical of Donald’s peers. The school psychologist and counselor continued to make observations during language arts, math, and lunch. They reported no changes in the fre- quency of Donald’s social conversations; he continued to engage in no more than 1 conversation per period. Next, the team implemented the strategy instruction in language arts and again, his conversations increased quickly to 2, 4, 3, and 3. The team again determined that Donald was achieving his goal to engage in an appropriate number of social conversations as this was comparable to his peers. Meanwhile, the observations by the school psy- chologist and counselor revealed Donald continued to engage in social con- versations during homeroom but not during the math or lunch periods. The team then implemented the strategy intervention during math and Donald’s social conversations increased to 3, 3, 4, and 3. Again, this was comparable to his peers’ frequency of social conversations during math period. His fre- quency continued to be maintained in homeroom and language arts periods. Interestingly, the counselor, who was observing Donald during his lunch period recorded that Donald’s social conversations there had increased to 4, 4, 3, and 5 during the same time as the implementation of the strategy instruction in his math period. The team reasoned that Donald had generalized the strategy instruction to the lunch period and that quite possibly, the effects of having more social conversations during other periods had “spilled over” into lunch as Donald’s peers were more inter- ested in conversing with him and vice versa. The results of this study are presented in Figure 10-8.

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 257

Donald: The Questions After reading the description of Donald and the pragmatic strategy instruc- tion implemented, answer the following questions. Then compare your answers to those provided at the end of this chapter.

What is the purpose of the study?

Who are the subjects?

What is the setting?

What are the dependent variables?

What are the independent variables?

What kind of intervention is provided to the subjects?

How were data collected and presented on graphs?

What were the results?

Why use a multiple baseline across settings design for this study?

What are the limitations of this study?

Donald: The Answers

Purpose of the Study The team implemented this study to determine if pragmatic strategy instruc- tion would increase a third grade student’s social conversations.

Subject A third-grade student, Donald, who struggled in language arts and math, and who engaged in very few social conversations was the subject.

Setting The study was conducted in Donald’s elementary school during homeroom, language arts, math, and lunch periods.

Dependent Variables The overall dependent variable recorded in this study was the frequency of Donald’s social conversations. One should note that each period of the day represents a separate and distinct dependent variable as well and actually accounts for four dependent variables.

258 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Independent Variable The pragmatic strategy instruction served as the independent variable.

The Design A multiple baseline across settings design was used.

The Intervention To help Donald with increasing the frequency of his social conversations, the team implemented pragmatic strategy instruction. This was appropriate because Donald reported he did not know what to say with other children and data indicated infrequent social conversations.

Obtaining the Data and Plotting the Results During the baseline condition, observations were made to record the fre- quency of Donald’s social conversations in homeroom, language arts, math, and lunch periods. As the intervention was implemented in homeroom and subsequent settings, observations continued to determine if the frequency of social conversations increased, decreased, or remained the same in relation to whether the baseline or intervention phase was in effect (see Figure 10-8).

Results During the initial baseline phase, Donald had a low frequency of conversa- tions in all four settings. When the intervention was implemented during the homeroom period, Donald successfully increased his conversations while a low frequency continued in the other three settings. Following implementa- tion in subsequent settings, a similar pattern emerged with higher and appropriate frequencies continuing in intervention phases and low frequency in baseline phases with one exception. In the final setting, lunch period, Donald’s frequency of social conversations increased simultaneously with the implementation of the intervention during math period (and its contin- ued implementation in homeroom and language arts periods). Reviewing Figure 10-8, it is clear that Donald’s frequency of social conversations increased with the introduction of the intervention except in the fourth set- ting, lunch period, when covariance occurred.

Why Use a Multiple Baseline Across Settings Design? The goal of a multiple baseline design is to provide a strong demonstration that the introduction of the intervention causes changes in the dependent variables when, and only when, the intervention is introduced. Using this

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 259

# of

S oc

ia l C

on ve

rs at

io n

Donald Baseline

Baseline

Observations

Lunch

Math

Strategy Instruction Homeroom

Language Arts

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 2 3 4 5 6 7 8 9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18

1 2 3 4 5 6 7 8 9

FIGURE 10-8 Donald’s Frequency

of Social Conversations.

20 14

260 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

design, the team was able to establish a functional relationship between the pragmatic strategy instruction and the frequency of social conversations during various periods of the school day although the covariance occurring during the lunch period is somewhat confounding.

Limitations of the Study This study has limitations. First, the covariance that occurred during the lunch period confounds the prediction that Donald’s behavior would change only when the intervention was introduced. The team did provide a reasonable explanation for the covariance. Second, there was limited follow-up in the study to demonstrate the increase in social conversations over time was maintained. Third, although the home environment was an important setting concerning Donald’s social conversations and interac- tions with other children, the team did not address this setting in the study. The team might reasonably expect that if Donald’s behavior gen- eralized to the less formal setting of the lunch period, it might also generalize to the home and other environments. However, at least includ- ing probes in other environments might have confirmed or refuted this expectation.

Summary A summary of the relevant dimensions of this vignette study can be found in Table 10-5.

TABLE 10-5 Summary of “The Pragmatic Strategy

Instruction Use to Increase the Frequency of Social Conversations of a Third

Grader in Multiple Settings.”

FEATURE DESCRIPTION

Type of design Multiple baseline across settings design

Purpose of the study To determine if the use of pragmatic strategy instruction would result in an increase in the frequency of a third grader’s social conversations

Subject A third grader who had academic struggles and very low frequencies of social conversations

Settings Four school periods including homeroom, language arts, math, and lunch

Dependent variables Frequencies of social conversation in the four settings

Independent variable Pragmatic strategy instruction

Results and outcome The pragmatic strategy instruction resulted in appropriate increases in the frequency of social conversations

CHAPTER 10 APPLICATION OF MULTIPLE BASELINE DESIGNS 261

Multiple baseline designs are versatile and relatively easy to understand. Yet, the time and resources required to conduct multiple baseline studies are sometimes problematic. In addition, there may be no other behaviors, settings, or individuals available who are in need of the intervention. At times, the need to identify an effective intervention as quickly as possible is very important. For these reasons, other designs may be appropriate, including the alternating treatments design discussed in the following chapter.

262 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

CHAPTER

11 Overview of Alternating Treatments Designs

IMPORTANT CONCEPTS TO KNOW ALTERNATING TREATMENTS DESIGN WITH NO BASELINE

Interpreting Data from Alternating Treatments Designs

ALTERNATING TREATMENTS DESIGN WITH A BASELINE

ALTERNATING TREATMENTS DESIGN WITH A BASELINE AND A FINAL TREATMENT PHASE

PREDICTION, VERIFICATION, AND REPLICATION

ADVANTAGES OF THE ALTERNATING TREATMENTS DESIGN

DISADVANTAGES OF THE ALTERNATING TREATMENTS DESIGN

ADAPTATIONS OF THE ALTERNATING TREATMENTS DESIGN Multielement Design Simultaneous Treatments Design Adapted Alternating Treatments Design

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

263

T he alternating treatments design allows for the comparison of theeffects of two or more independent variables (treatments) on the samebehavior. This is a very important design for educators and clinicians who frequently are concerned with which of several intervention procedures is the most effective. The alternating treatments design has also been referred to as the multielement design (Ulman & Sulzer-Azaroff, 1975), although there are slight differences, primarily based on the nature of the independent variables being investigated (Wolery, Gast, & Hammond, 2010). The multielement design is described briefly later in this chapter. It has also been erroneously called the simultaneous treatments design (Kazdin & Hartmann, 1978). This is actually a modification of the alternating treatments design and is also discussed later in this chapter.

The basic use of the alternating treatments design requires the “rapid alternation of two or more distinct treatments (i.e., independent variables) while their effects on a single target behavior (i.e., dependent variable) are measured” (Cooper, Heron, & Heward, 2007; p. 188). The treatments can be alternated within sessions, across different times of the same day, or across different days. There are three important points that should be made about the alternating treatments design. First, the presentation of the treatments should be counterbalanced. If, for example, there were three treatments (A, B, and C), they could be presented randomly (e.g., ABBCABCAC) or in blocks. There are six possible blocks of the three treatments: ABC, BCA, CAB, ACB, BAC, CBA. The researcher should make sure that each block of treatments is presented at least once (Alberto & Troutman, 2009). It is important to note that all aspects of the treatment should be counterbalanced. For example, if the treatments are given at different times of the day or by more than one person, then those aspects of the study should be counterbalanced as well.

The second important point is that the subjects should be able to discriminate between or among the treatment conditions. This is made easier if the treatments are sufficiently different from one another. In some cases, the nature of the treatment will help the subject discriminate (e.g., use of a worksheet vs. manipulatives to increase math computation skills). It might also be necessary to use verbal cues (e.g., “Today you will receive bonus points for completing your science workbook assignment”). Other means such as cue cards or signs can also be used. The important point is that a distinct stimulus is associated with each treatment (Cooper et al., 2007). Finally, the dependent variable(s) should be reversible. In other words, the behavior should be able to increase or decrease when the conditions are presented (Wolery, Gast, & Hammond, 2010).

One interesting characteristic of the alternating treatments design is that, unlike typical single subject designs, it does not require the collection of baseline data, although it is recommended. Three different types of the basic alternating treatments designs are discussed in this chapter, two of which

264 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

incorporate baseline data and one of which does not. The three types are the alternating treatments design without baseline, alternating treatments design with baseline, and the alternating treatments design with a baseline and a final treatment phase. It should also be noted that the alternating treatments design is frequently used in conjunction with other single subject designs. For example, the first phase of a study might use an alternating treatments design to determine which of several interventions is the most effective. Next, a withdrawal design (discussed in Chapter 5) might be used to further establish the functional relationship between the most effective treatment and the target behavior.

Alternating Treatments Design with No Baseline Because it is not necessary to collect baseline data when using the alternat- ing treatments design, the treatment phases can be implemented immedi- ately. Baseline data in this sense are those data obtained prior to any treatment conditions (preintervention baseline). There are times when the collection of preintervention baseline data is not educationally or clinically appropriate. The first is when the nature of the target behavior (e.g., self- abusive behavior) is so severe that, ethically, baseline data should not be collected and treatment should begin immediately. This is similar to the sit- uation in which a B-A-B design is used (discussed in Chapter 5). It should be emphasized, however, that many individuals who use the alternating treatments design actually include a type of baseline data by having a no-treatment phase as one of the alternating treatment conditions. Thus, a distinction is often made between an alternating treatment designs without a no treatment condition and an alternating treatments design with a no- treatment condition design (Cooper et al., 2007). In the first type, no data are recorded other than during treatment conditions. In the second type, the no-treatment condition acts as a baseline of sorts, although that condition should not be considered the same as a preintervention baseline condition. When the no-treatment phase is alternated with various treatment phases, there may be multiple treatment interference (Barlow & Hayes, 1979), in which the effects of one condition carry over or in some way affect the other condition(s). This could result in the data from the no-treatment phase being different from baseline data recorded before any treatment is introduced. In fact, the use of the no-treatment condition is often used to evaluate if multiple treatment interference is occurring (Wolery, Gast, & Hammond, 2010). The issue of multiple treatment interference was discussed in Chapter 4 and is described in more depth later in this chapter. An example of both a no baseline alternating treatment design with and without a no-treatment condi- tion are presented in the next chapter on applications.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 265

The following is an example of an alternating treatments design with no baseline. Note that in this example, a no-treatment condition is included.

Jimmy and Sue are two 5-year-old students with mild autism who have minimal but evident and emerging verbal skills and speak in single words and short sentences. They frequently use more negative statements (e.g., “Not like you” and “Shut up”) than positive sentences (e.g., “Like to play?” and “What’s your name?”) when interacting with their peers. As a result, their peers have started avoiding them, which is counterproduc- tive to their social and communication goals. Their teacher, Mr. Ferrell, wanted to increase the number of positive statements made by the two students. He decided to determine the effectiveness of the use of verbal praise versus a token system as reinforcers for positive sentences. He had used both approaches in the past and noted that both seemed to work but was unsure of their relative effectiveness. After carefully defin- ing what was to be considered positive statements, Mr. Ferrell randomly assigned three treatment conditions: no consequence for a positive state- ment (A), verbal praise for a positive statement (B), and presentation of a token for a positive statement (C). The various contingencies and data collection were initiated during the 30-minute free-play time that was scheduled each day. He carefully explained the token system that had been used in the class before, which used a reinforcement menu individu- alized for each student. At the beginning of the day, Mr. Ferrell told the students which condition would be in effect. For example, “Today, I am going to tell you what a great job you are doing every time you say something nice to your classmates” or “Today I am going to give you a token so you can earn something on your wish list every time you say something nice to your classmates.”

As noted, it is important that the treatments be randomly assigned or counterbalanced in some way to avoid order effects. Data would then be collected to determine which of the three treatment conditions was the most effective. These data are graphically presented in Figure 11-1. This same design could also be used by omitting the no-treatment phase and just comparing the two treatment conditions, although it would not provide the valuable information regarding treatment versus no-treatment gains.

Interpreting Data from Alternating Treatments Designs Interpretation of data from an alternating treatments design can sometimes be ambiguous, particularly if the data paths overlap. The question becomes “How superior does one treatment need to be to assume that there are sig- nificant differences?” As will be discussed in Chapter 13, there are ways of determining statistical significance in single subject designs, although for

266 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

most practitioners the question is more related to practical significance. Alberto and Troutman (2009) noted that in order to assume that one treat- ment is more effective than the other(s) when interpreting information from an alternating treatments design, their data paths must be separate except at the beginning of the study. This is frequently determined by calculating the

A B C B C B A C A B B A C A C B C A A B C B A C B C Sessions (Days)

Subject 1 (Jimmy)

# of

P os

it iv

e St

at em

en ts

A B C B C B A C A B B A C A C B C A A B C B A C B C Sessions (Days)

Subject 2 (Sue)

# Po

si ti

ve S

ta te

m en

A-No Treatment B-Verbal Praise C-Token

FIGURE 11-1 Example of data from an

alternating treatments with no baseline design

19 99

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 267

percentage of non overlapping data (PND). To do this, the first data point from one condition is compared to the first data point in another condition. The second data point from one is then compared to the second data point of the other, etc. Figure 11-2 shows an example of both ambiguous and unambiguous results. In the graph of the ambiguous data, condition C was

A B C B C B A C A B B A C A C B C A C A B Sessions

Ambiguous Results

D ep

en d

en t V

ar ia

b le

A B C B C B A C A B B A C A C B C A C A B Sessions

Unambiguous Results

D ep

en d

en t V

ar ia

b le

A-No Treatment B-Treatment 1 C-Treatment 2

FIGURE 11-2 Example of both ambiguous and

unambiguous data from an alternating treatments

design

19 99

268 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

superior to condition B approximately 57% of the time (4/7 higher data point pairs). In the unambiguous data, condition C was superior approxi- mately 83% of the time (5/6 higher data point pairs). Further, the only data point pair in the unambiguous data set where B was superior was the first pair. Ideally, 100% PND would show clear superiority of one condition over another. Note that in the data set representing ambiguous results, it is still possible to determine that using an intervention (either B or C) was bet- ter than using none (A). The PND of both conditions B and C compared to baseline (A) was 100%.

4 C H E C K I T O U T # 1 Using PND, what conclusions can Mr. Ferrell make based on the data presented in Figure 11-1?

Alternating Treatments Design with a Baseline It is generally agreed that, whenever possible and appropriate, initial base- line data should be collected before introducing the alternating treatments. Although not a requirement, ideally, as in other single subject designs, the baseline data should demonstrate a stable rate of responding. There are two situations, however, when this might not be possible. The first occasion when baseline data do not have to be stable is when the trend is moving in a counter therapeutic direction. Suppose, for example, that the frequency of a target behavior, verbal insults, did not stabilize under baseline conditions. In fact, it increased at a rather steady rate. Then, the collection of baseline data could be discontinued and the intervention phases introduced. The second sit- uation is when the target behavior changes simply as a result of being recorded. This could also be viewed as practice effects. Suppose, for example, that a teacher was interested in determining the effects of a phonics versus a whole language approach on the target behavior of word identification. If the student first reads lists of words to establish a baseline, it is possible that the student’s word identification skills might actually improve during the baseline condition due to practice effects (particularly if this was a skill that the student hadn’t spent much time practicing). This could result in a baseline that never stabilizes. Because a stable baseline is not a prerequisite for the alternating treatments design, it might be a better choice than other designs when using dependent variables that are more susceptible to practice effects.

An example of data collected using an alternating treatments design with a baseline using the previously described study is presented in Figure 11-3. The major difference from the data depicted in Figure 11-1 is that baseline data are

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 269

gathered before the introduction of any treatment phase and the no-treatment condition is eliminated. One option in using this design with a preintervention baseline is to also include a no-treatment condition as one of the alternating treatments. This becomes an even more powerful design and can be used to evaluate the presence of multiple treatment interference (discussed later).

A A A A A A A B C B C B C C B B C C B B C C B C B

A A A A A A A B C B C B C B B C C B C C B C B B C Sessions (Days)

Subject 1 (Jimmy)

# of

P os

it iv

e St

at em

en ts

Sessions (Days)

Subject 2 (Sue)

# of

P os

it iv

e St

at em

en ts

A-Baseline B-Verbal Praise C-Token

FIGURE 11-3 Example of data from a

baseline followed by an alternating

treatments design

19 99

270 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Alternating Treatments Design with a Baseline and a Final Treatment Phase

For obvious clinical, educational, and ethical reasons, it is important to con- tinue the most effective treatment after that determination has been made. Thus, using this design, the researcher would first collect initial baseline data, then introduce the alternating treatments to determine which is the most effective, and, finally, continue the study using only the most effective treatment. Data using the baseline followed by alternating treatments and a final treatment phase are presented from the previously described study (see Figure 11-4). Note that once the determination was made that verbal praise was, in fact, the most effective treatment, it was used exclusively in the final treatment phase. In this way, the time and effort of planning and implement- ing the less effective token system could be eliminated.

There is also a methodological reason to continue implementing the most effective treatment as a final phase. It is possible that the most effective treatment determined from the alternating treatment conditions might lose its effectiveness once it is presented in isolation. In other words, multiple treatment interference could have been a factor. On the other hand, if the effect is maintained, then there is further evidence that multiple treatment interaction is not a major concern (Wolery, Gast, & Hammond, 2010).

4 C H E C K I T O U T # 2 Based on the data presented in Figure 11-4, do you think multiple treatment inter- ference was a factor? Why or why not?

Prediction, Verification, and Replication There are some pros and cons of the alternating treatments design when it comes to prediction, verification, and replication. On the one hand, Alberto and Troutman (2009) stated that the determination of a functional relation- ship between the dependent and independent variables is relatively weak because of the lack of replication. They did point out however, that including the most effective treatment in a final phase addresses this concern.

On the other hand, Cooper et al. (2007) argued that the very nature of alternating treatment designs addresses each of the three issues in the fol- lowing way.

1. Prediction—Each data point serves as a predictor of future behavior under the same treatment.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 271

2. Verification—Each successive data point serves to verify previous predictions of performance under the same treatment.

3. Replication—Each successive data point provides the opportunity to replicate the differential effects produced by the treatments.

A A A A A A A B C B C B C B B C C B C B B B B B B

A A A A A A A B C B C B C B B C C B C B B B B B B Sessions (Days)

Subject 1 (Jimmy)

# of

P os

it iv

e St

at em

en ts

Sessions (Days)

Subject 2 (Sue)

# of

P os

it iv

e St

at em

en ts

A-Baseline B-Verbal Praise C-Token

FIGURE 11-4 Example of data from a baseline followed by an alternating tratments design with a final

treatment phase design

19 99

272 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Consider, for example, the data for Jimmy presented in Figure 11-1. Draw a vertical line after the 9th day or data point (after each condition has been presented three times). The last data point for each condition serves as a predictor for the next data point for each condition. If you then draw another line after the next set of data points (4 days later), it can verify the previous prediction. As you look at the rest of the data points over time, you notice the same trend continue as the three sets of data points separate. This provides replication of the relative effects of the different treatments.

Issues related to internal and external validity have also been addressed. Neuman (1995) pointed out that the alternating treatments design demon- strates good internal validity; if one treatment is consistently associated with an improved level of responding, then the design demonstrates good experimental control. As with other single subject designs, external validity is an area that should be specifically addressed. It is important to replicate the results of alternating treatments designs with different subjects, differ- ent experimenters, and/or different conditions. It is recommended that at least five participants be included in a study using an alternating treat- ments design to better elucidate the treatment effects (Wolery, Gast, & Hammond, 2010). Interestingly, the possibility of multiple treatment inter- ference has both positive and negative implications. It might produce carryover effects that obscure the relationship of the dependent variable to the various treatments. A flip side to the possible negative effect of multiple treatment interference, however, is the possible positive effect of eliminating sequence effects through the treatment counterbalancing that would strengthen the relationship.

Advantages of the Alternating Treatments Design There are many advantages of using an alternating treatments design. As noted previously, it is ideal for the teacher or clinician who is interested in determining which of several interventions is the most effective. In fact, because the treatments are alternated rapidly, the determination of their rel- ative efficacy can usually be made faster than when using other designs. This process is especially fast if preintervention baseline data are not collected. It should be kept in mind, however, that researchers who want to make a stronger case for the effectiveness of the treatments should collect those pretreatment data. Another advantage of the alternating treatments design previously noted is that if baseline data are collected, then they do not need to be stable before the intervention can be initiated. As noted in Chapter 4, stability of baseline is an important prerequisite for prediction, verification, and replication for most single subject designs.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 273

Some single subject designs (e.g., the A-B-A-B or A-B-A-C) require with- drawal of a treatment to demonstrate that a functional relationship exists between the independent and dependent variable(s). This is an ethical and practical concern that is avoided when using the alternating treatments design because no withdrawal of treatment is necessary. In addition, because the determination of the most effective treatment can often be made quickly, less time may be spent administering an ineffective treatment. This is partic- ularly true when using an alternating treatments design that incorporates a final treatment phase. Again, the alternating treatments design seems to be a good design to use in an educational or therapeutic setting.

Another point, previously mentioned, is that the counterbalancing used in the alternating treatments design helps to eliminate sequencing effects. Suppose that an A-B-A-C changing design discussed in Chapter 7 was used in our previously described study to determine the effects of verbal praise (B) versus tokens (C) to increase positive statements (with the A condition representing baseline). It is possible that the presentation of the verbal praise phase (B) might have an effect on the student’s behavior in the token phase (C). Similarly, if the treatments had been presented in the other order, with the token phase preceding the verbal praise phase, then the first treatment condition might have an effect on the second. By using an alternating treatments design this problem could be avoided, because the B and C phases (or the A, B, and C phases) would be counterbalanced.

In summary, the following guidelines can help to determine when an alternating treatments design should be used:

• When you want to determine the relative effectiveness of more than one treat- ment on a given behavior;

• When baseline data are either unavailable or might be unstable; • When the treatments are sufficiently different from each other; • When the subjects can discriminate the treatment conditions; • When the effects of sequencing the interventions might obscure the results.

Disadvantages of the Alternating Treatments Design Although the advantages of the alternating treatments design are numerous, there are disadvantages as well. As noted previously, one major concern when using this design is the possibility of multiple treatment interference. In fact, the very nature of the alternating treatments design that requires the rapid alteration of interventions can lead directly to this situation. Mul- tiple treatment interference results in the masking of the effects of a specific

274 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

treatment because other treatments might influence, confound, or carry over. Also, as noted earlier, multiple treatment interference can also have an effect on the no-treatment condition, thus making that condition different from a true, preintervention baseline condition. However, it is possible to minimize multiple treatment interference in an alternating treatments design. If the various treatments are considerably different from one another then the likelihood of multiple treatment interference is minimized. Take, for example, a situation in which a researcher was interested in reducing the number of temper tantrums in a student with Down syndrome. He chose two types of time-out as the treatments. The first, contingent observation, allowed the stu- dent to see and hear what was going on in the classroom but not to receive any reinforcement. The second type, exclusionary time-out, allowed the stu- dent to hear what was happening in the classroom but not to see what was going on or to receive any reinforcement. The chance of multiple treatment interference would be greater in this situation than if contingent observation and a more dissimilar treatment such as differential reinforcement of other behavior (DRO) were used.

It is also possible to evaluate the presence of multiple treatment interfer- ence. One suggestion is to include the presentation of the most effective treatment at the end of the design as a final treatment phase. In this way, the effects of the final treatment in isolation will help to eliminate the multi- ple treatment interference (Cooper et al., 2007). If both preintervention baseline data and no-treatment condition data are collected, then they could be compared. If the level of no-treatment condition data were stable and similar to the preintervention baseline data, then it would suggest that multiple treatment interference did not occur. However, if the data level were different, then it would suggest that multiple treatment interference was present (Wolery, Gast, & Hammond, 2010).

A related issue has to do with reversibility. As was noted earlier, the withdrawal of treatment to demonstrate that the target behavior reverses toward baseline levels is not a requirement in the alternating treatments design. However, if dependent variables are chosen in which reversibility is not expected or desired, then the alternating treatments design would not be appropriate. Suppose, for example, that a teacher was interested in teaching phonics using two different methods. Once the student correctly learned a sound-symbol relationship, it is both likely and desirable that it would be retained, so the differential treatment effects could not be determined adequately.

In addition, alternating treatments designs are not effective for evaluat- ing independent variables that produce change slowly or that need to be administered over a continuous period of time (Neuman, 1995). In fact, the treatments should be able to produce change on a session-by-session basis. Consider the following two examples to demonstrate this point.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 275

The first example would be appropriate for an alternating treatments design, whereas the second would not be.

Example 1: Ms. Smith is a third-grade teacher for students with learning dis- abilities. She has two students in her class who perform poorly in the area of mathematics computation because they frequently perform the wrong operation (e.g., adding instead of multiplying, subtracting instead of adding). Ms. Smith wanted to see which of two interventions would be most effective. On worksheets of 25 problems, the two students often would make between 7 and 10 mistakes because they performed the wrong operation. The first intervention involved the use of color-coded operation signs (e.g., blue !, red ", green #) to serve as a visual cue to remind them which operation needed to be performed. The second intervention involved the use of a visual reminder “THINK SIGN” written at the top of the worksheet.

Example 2: Mr. Norris is a seventh-grade math teacher. He has three stu- dents who are having great difficulty converting fractions to percentages. He has been working on this skill with the students for over 3 weeks with little success. He decided to try two different approaches to teaching this skill. The first involved having the students memorize a conversion chart that showed the various fractions presented as percentages. The other was the use of a mnemonic device to teach the process of “Divide Top By Bottom, Multiply Times Hundred.” To do this, he made up the acrostic “Detroit Tigers Bat Boy Makes The Hit.” The students used the first letter of each word in the saying to help remember the correct process to use.

In Example 1, an alternating treatments design could be used. Ms. Smith could have collected baseline on a series of worksheets and then randomly assigned the students to the two treatment conditions. The students should be able to discriminate the two interventions and had in their response rep- ertoire the ability to perform differently in each session. Even here, however, the possibility of multitreatment interference is evident. It is possible that the effects of one intervention might carry over to the other.

In example 2, an alternating treatments design would probably not be the best choice. First, the students did not have the dependent variable—conver- sions of fractions to percentages—in their repertoire. Therefore it is unlikely that the behavior could change from session to session, at least initially. It would be necessary for the students to first learn the skill over repeated trials.

4 C H E C K I T O U T # 3 What is another possible reason why an alternating treatments design would be inappropriate for example 2?

276 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Other disadvantages of the alternating treatments design relate to its implementation and interpretation. Regarding implementation, Cooper et al. (2007) suggested that because this design uses rapidly alternating treat- ments, it is somewhat artificial and not typical of the method in which treat- ments are typically presented in the natural environment. They also pointed out that typically a maximum of three treatments should be evaluated. Otherwise, the number of sessions necessary to conduct this study becomes excessive. Also, it is often difficult to do the necessary counterbalancing in most natural settings and if too many treatment conditions are involved, then the participants might have more difficulty discriminating the various conditions.

Yet another disadvantage of the alternating treatments design is that it would be inappropriate to use with individuals who may not have the abil- ity to discriminate between or among the treatment conditions. As noted earlier, this is a prerequisite for the use of the design. For example, its use with an individual with severe cognitive limitations who would have to dis- criminate among relatively subtle treatment conditions might not be appro- priate. One factor that has been shown to affect a subject’s ability to discriminate the treatment conditions is the intercomponent interval length. That is, the longer each specific treatment is in effect, the easier it is for the subject to discriminate the conditions, thus minimizing multitreatment inter- ference (McGonigle, Rojahn, Dixon, & Strain, 1987).

In summary, the following are examples of instances when an alternat- ing treatments design is not the most appropriate to use:

• When the treatments might interact, thus obscuring the results; • When the subjects cannot discriminate the treatment conditions; • When the treatments typically produce slow behavior changes; • When the treatments need to be administered over a continuous period of time

to be effective;

• When it becomes difficult to counterbalance the various aspects of the study.

Adaptations of the Alternating Treatments Design

Multielement Design As mentioned earlier in this chapter, the term alternating treatments design is sometimes used interchangeably with the term multielement design. There are indeed many similarities although there are also some subtle differences in the two designs. One such distinction is the nature of the independent variable investigated. Whereas the alternate treatments design is typically

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 277

used to evaluate the effectiveness of two or more interventions with a partic- ipant, the multielement design is used to determine the factors that influence participant’s behavior to develop an effective intervention (Wolery, Gast, & Hammond, 2010). Thus the multielement design is particularly relevant when used as part of a functional behavior assessment in which the purpose of a specific behavior is identified so that a behavioral intervention program can be developed. Specifically, the multielement design can be used to “eval- uate whether a response occurs differentially across conditions, in order to identify a functional relation” (Wacker, Berg, & Harding, 2008; p. 71). Once the purpose of the behavior is determined, then an appropriate inter- vention can be developed that might incorporate a replacement behavior that would serve the same purpose. The multielement design is sometimes used as a first phase of a larger study. As an example, Peyton, Lindauer, and Richman (2005) used the design with a 10-year-old girl with autism as phase 1 of a 3-phase study. They found that the target behavior, noncompli- ant vocal behavior, occurred much more frequently under the condition of escaping nonpreferred tasks than during conditions of free play, receiving attention, or being left alone. Thus, the researchers concluded that the pur- pose of the subject’s noncompliant vocal behavior was to escape or avoid nonpreferred tasks.

Simultaneous Treatments Design The major adaptation of the alternating treatments design is the simulta- neous treatments design (Kazdin & Hartmann, 1978), also called the con- current schedules design (Hersen & Barlow, 1976). As noted previously, these terms have been used erroneously to describe the alternating treat- ments design. The use of the simultaneous treatments design, in fact, is rela- tively rare in the professional literature. As the name implies, in this design the treatment conditions are presented at the same time instead of being alternated. The simultaneous treatments design actually has several advan- tages over other single subject designs (Tawney & Gast, 1984). First, it best approximates the conditions of the natural environment. Second, it takes less time to determine treatment effectiveness than other designs, including the alternating treatments design. Conversely, the use of this design requires more skill, planning, and organization than the use of other designs. It is a difficult task to systematically analyze the effects of several interventions presented at the same time.

The following example shows both the value of the simultaneous treat- ments design and the challenge that is present in its planning.

Ms. Ziegler, a general education teacher, and Ms. Massey, a special education teacher, were teamed together in a fifth-grade inclusion

278 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

classroom. There were 20 students in the class, including 2 with learning disabilities, 1 with behavior disorders, 1 with mild intellectual disability, and 3 who were receiving speech and language services. Joey, the stu- dent with behavior disorders, would frequently curse at the teachers, dis- rupting the classroom. Ms. Ziegler and Ms. Massey requested a meeting with the school psychologist to discuss some type of intervention plan. They mutually agreed to determine which of two intervention proce- dures—verbal reprimand or contingent exercise—would be most effec- tive in reducing Joey’s cursing behavior. The school psychologist suggested that both teachers collect baseline data for a week, noting each instance of cursing behavior directed at each teacher during the day. During the 2nd week, Ms. Ziegler verbally reprimanded Joey’s cursing behavior. This consisted of her saying loudly and abruptly “No Joey, do not curse.” Also during the 2nd week (at the same time Ms. Ziegler was using verbal reprimands), Ms. Massey initiated the contin- gent exercise condition. During this condition, every time Joey cursed he was instructed to sit down and stand up 10 times. This procedure has been shown to be effective with cursing behavior (Luce, Delquadri, & Hall, 1980). The 3rd week, Ms. Ziegler provided the contingent exercise treatment while Ms. Massey used verbal reprimands. This continued for 2 more weeks, with each teacher providing each of the two interventions for a week.

In the above example, the treatment conditions were presented to the student at the same time rather than being alternated. In other words, both the verbal reprimand and contingent exercise conditions were presented simultaneously. The goal was to determine, over time, which treatment becomes superior, regardless of the teacher who is implementing it.

Adapted Alternating Treatments Design Another version of the basic alternating treatments design that has been used in the professional literature is called the adapted alternating treatments design (Sindelar, Rosenberg, & Wilson, 1985). In this design, each inter- vention is applied to different behaviors that are considered to be of equal response difficulty but functionally independent. Unfortunately, the process of equating the behaviors can be time consuming and problematic, although it is a very important prerequisite to the use of this design. Holcombe, Wolery, and Gast (1994) noted that grade levels or other nor- mative data might be used to help equate the dependent behaviors. For example, if a student had a very similar standard score on an achievement test measuring both written spelling and math computation, those two behaviors might be targeted for an adapted alternating treatments design.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 279

If two interventions were being evaluated (e.g., use of worksheets vs. flash cards), then each behavior would receive each treatment in a random order for the same number of sessions. If one treatment is shown to be more effective with both behaviors, it would provide additional evidence of a functional relationship.

Key Concepts/Terms Basic goal—Comparison of the effects of two or more treatments on the

same behavior. Alternating treatments design with no baseline—Treatments are pre-

sented randomly to the subjects across different days, across times of day, or within sessions; preintervention baseline data are not col- lected, although a no-treatment phase (A) can also be alternated (e.g., BACBABCAC).

Multiple treatment interference—A situation in which the effects of one condition carry over or in some way affect the other condition(s).

Alternating treatments design with a baseline—Baseline data are collected before the presentation of the treatments; provides additional informa- tion about changes from pretreatment to treatment.

Alternating treatments design with a baseline and a final treatment phase—Ends in a final treatment phase that has been determined to be the most effective during the study.

Prediction—Each data point in each condition serves as a predictor for future behavior under the same condition.

Verification—Succeeding data points in each condition verify the prediction made from the previous data points.

Replication—The differential effects of the treatment are demonstrated if the trend continues.

Advantages of the alternating treatments design—Good to use when determining which of two or more treatments is effective; collection of baseline data not necessary; avoids sequencing effects found in other single subject designs.

Disadvantages of the alternating treatments design—The possibility of multiple treatment interference when the treatments interact, thus obscuring the results; not good for behaviors that change slowly; some- times difficult to implement (e.g., counterbalancing the treatments).

Intercomponent interval length—The length of time a specific intervention is in effect. The longer the interval, the easier it is for the subject to discriminate between or among conditions.

280 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Adaptations Multielement design—Used in functional behavior analysis to determine

the purpose of a behavior that can lead to the choice of an intervention. Simultaneous treatments design—Also called the concurrent schedules

design; treatment conditions are presented simultaneously rather then being alternated; somewhat difficult to plan and implement.

Adapted alternating treatments design—Intervention is applied to multi- ple behaviors that are similar in nature but functionally independent.

4 Possible Answers to Check It Out

¶ For Jimmy, the results were clearcut. Except for the first data pair(which were the same), the verbal praise condition was superior to the token system for all the data pairs. For Sue, verbal praise was more effective in approximately 89% (8/9) of the comparisons. For Jimmy, both the verbal praise and the token system was superior to the no-treatment condition. For Sue, however, the token system condition could not be determined to be more effective than no treatment because of the overlap in the data lines.

· Analysis of the data would indicate that multiple treatment interferenceis not a factor. For both Jimmy and Sue, the verbal praise condition continued its positive treatment effect when given in isolation.

¸ The goal of the mnemonic strategy intervention is for the students toremember it and use it whenever a fraction-to-percentage conversion is required. Thus, the possibility of multiple treatment interference is not only possible but expected.

References Alberto, P., & Troutman, A. (2009). Applied behavior analysis for teachers

(8th ed.). Columbus, OH: Pearson: Merrill. Barlow, D., & Hayes, S. (1979). Alternating treatment design: One strategy for

comparing the effects of two treatments in a single behavior. Journal of Applied Behavior Analysis, 12, 199–210.

Cooper, J., Heron, T., & Heward, W. (2007). Applied behavior analysis (2nd ed.). Columbus, OH: Merrill.

Hersen, M., & Barlow, D. (1976). Single case experimental designs: Strategies for studying behavior change. New York: Pergamon Press.

Holcombe, A., Wolery, M., & Gast, D. (1994). Comparative single subject research: Description of designs and discussion of problems. Topics in Early Childhood Special Education, 14, 119–145.

Kazdin, A., & Hartmann, D. (1978). The simultaneous treatment design. Behavior Therapy, 9, 912–922.

CHAPTER 11 OVERVIEW OF ALTERNATING TREATMENTS DESIGNS 281

Luce, C, Delquadri, J., & Hall, R. (1980). Contingent exercise: A mild but powerful procedure for suppressing inappropriate verbal and aggressive behavior. Journal of Applied Behavior Analysis, 13, 583–594.

McGonigle, J., Rojahn, J., Dixon, J., & Strain, P. (1987). Multitreatment interfer- ence in the alternating treatments design as a function of the intercomponent interval length. Journal of Applied Behavior Analysis, 20, 171–178.

Neuman, S. (1995). Alternating treatments designs. In S. Neuman & S. McCormick (Eds.), Single subject experimental research: Applications for literacy (pp. 64–83). Newark, DE: International Reading Association.

Peyton, R., Lindauer, S., & Richman, D. (2005). The effects of directive and nondi- rective prompts on noncompliant vocal behavior exhibited by a child with autism. Journal of Applied Behavior Analysis, 38, 251–255.

Sindelar, P., Rosenberg, M., & Wilson, R. (1985). An adapted alternating treatments design. Education and Treatment of Children, 8, 67–86.

Tawney, J., & Gast, D. (1984). Single subject research in special education. Columbus, OH: Merrill.

Ulman, J., & Sulzer-Azaroff, B. (1975). Multielement baseline design in educational research. In E. Ramp & G. Semb (Eds.), Behavior analysis: Areas of research and application (pp. 371–391). Englewood Cliffs, NJ: Prentice-Hall.

Wacker, D., Berg, W., & Harding, J. (2008). Single-case research methodology to inform evidence-based practice (pp. 61–82). In J. Luiselli, D. Russo, W. Christian, and S. Wilczynski (Eds.), Effective practices for children with autism: Educational and behavioral support interventions that work. NY: Oxford Press.

Wolery, M., Gast, D., & Hammond, D. (2010). Comparative intervention designs. In D. Gast (Ed.), Single subject research methodology in behavioral sciences (pp. 329–381). NY: Routledge.

282 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

CHAPTER

12 Application of Alternating Treatments Designs

283

In the last chapter, we described the different types of alternatingtreatments designs. The alternating treatments design is an importantsingle subject design that is used to determine the effects of more than one treatment (independent variable) on the same behavior (dependent variable). In this chapter, we will provide specific examples from the professional literature for four basic types of alternating treatments designs. Similar to previous application chapters, a vignette is provided at the end of the chapter to give the reader the opportunity to apply information related to this design.

Alternating Treatments Design with No Baseline (without a no-treatment condition)

Ingersoll, B. (2011). The differential effect of three naturalistic language interven- tions on language use in children with autism. Journal of Positive Behavior Interventions, 13, 109–118.

Purpose of the Study The purpose of this study was to compare the effect of three natural lan- guage teaching approaches—responsive interaction (RI), milieu teaching (MT), and a combination of the two on expressive language of two children with autism. The investigator hypothesized that MT would result in greater total language, prompted language, and requests, whereas RI would result in greater spontaneous language and comments.

Subjects Two young children with autism served as subjects. Leon was 42 months old, whose language targets consisted of single words. Griffon was 40 months old, whose target behavior was noun-verb and noun-adjective combinations.

Setting The study was conducted in a small treatment room in a communication disorders center. A number of motivating toys were provided in the room.

Dependent Variables The dependent variable for Leon was production of single words; for Griffon, it was production of noun-verb and noun-adjective combinations. In addition, the produced language was categorized by type (spontaneous vs. prompted) and function (requesting vs. commenting). Percentage of the occurrence of language production in 10-second intervals was reported.

284 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Independent Variables Response interaction, milieu teaching, and the combination of the two were the independent variables. RI consisted of modeling and expanding the child’s language. MT involved the direct elicitation of language with reinforcement.

Design An alternating treatments design without baseline or a no-treatment condi- tion was used. The investigator used a random number generator to deter- mine the order of the treatment conditions.

Intervention The subjects visited the communication disorders center twice a week for approximately a month. Each day, three to four 10-minute sessions were conducted using the randomized order of the interventions that was deter- mined. For the RI condition, the therapist sat next to the child and narrated, labeled, and described the child’s play. For example, if the child was pushing a toy car back and forth, then the therapist might say “car” for Leon or “roll car” or “red car” for Griffon. Using the MT approach, the therapist would elicit the language by modeling, requesting, or asking questions. For example, if Leon was rolling the car back and forth, the therapist might interrupt, say the word “car,” tell Leon to say “car,” or ask a question such as, “What do you want?” A similar procedure would be used with Griffon using two word combinations. The combined approach used both RI and MT.

Obtaining the Data and Plotting the Results Data were collected for the percentage of 10-second intervals within the 10- minute session in which the target language production occurred (one word for Leon, noun-verb and noun-adjective combinations for Griffon). These data were further coded by type (prompted or spontaneous) and by function (requests or comments.) Data were presented on a series of x-y graphs.

Results For total language, MT and combined were significantly more effective than RI for Leon. For Griffon, MT and combined started out much more effec- tive, but RI increased throughout the study until there were essentially no differences by the end of the study (see Figure 12-1). MT and combined were also significantly more effective for prompted language and for spon- taneous language with Leon. For Griffon, MT and combined were more effective for prompted language, although the results for spontaneous

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 285

language were not as clear. Similar to total language, MT and combined were more effective initially, but RI increased over the course of the study. Regarding the function of the produced language, MT and combined were significantly higher than RI for requests for both subjects. RI was slightly higher for comments for both children. (see Figure 12-2).

Why Use an Alternating Treatments Design with No Baseline (without a no-treatment design)? The investigator was interested in determining the relative effectiveness of three interventions. The amount of time for the study (in reality only about eight days total) precluded the use of other designs. The alternating treat- ments design does address the investigator’s questions. However, the lack of baseline or a no-treatment condition limits some of the conclusions that can be made about the effectiveness of the interventions. Apparently, the limited time affected that decision.

100

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Session

Pe rc

en t o

f I n

te rv

al s

Pe rc

en t o

f I n

te rv

al s

Leon

Gri!on

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Session

Responsive Interaction Milieu Teaching Combined

FIGURE 12-1 Total language data for the subjects in the alternating treatment with no baseline (without a no-treatment

condition) study. Note. From “The

Differential Effect of Three Naturalistic Language

Interventions on Language Use in Children with

Autism”, by B. Ingersoll, 2011, Journal of Positive

Sage Publications. Reprinted with permission.

286 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

60 10 0 80 60

60 50 40 30 20 10 0

40 20 050 40 30 20 10 0 1

3 5

7 9

11 13

15 17

19 21

1 3

5 7

9 11

13 15

17 19

21 23

1 3

5 7

9 11

13 15

17 19

21 23

1 3

5 7

9 11

13 15

17 19

60 50 40 30 20 10 0

Se ss

io n

Re q

ue st

Percent of Intervals

Percent of Intervals Percent of Intervals

Percent of Intervals C

om m

en ts

Se ss

io n

Le on

G ri!

on G

ri! on

Le on

Se ss

io n

S es

si o

Re sp

on si

ve In

te ra

ct io

n M

ili eu

T ea

ch in

g C

om b

in ed

FI GU

RE 12 -2

Co m m un ica tiv e fu nc tio n of

la ng ua ge

da ta fo r th e

su bj ec ts in th e al te rn at in g

tre at m en tw

ith no ba se lin e

(w ith ou t a no -tr ea tm en t

co nd iti on ) st ud y.

No te .F ro m “T he

Di ffe re nt ia lE ffe ct of Th re e

Na tu ra lis tic

La ng ua ge

In te rv en tio ns on

La ng ua ge

Us e in Ch ild re n w ith

Au tis m ”, by

B. In ge rso ll,

20 11 ,J ou rn al of Po sit ive

Be ha vio r In te rv en tio ns ,1 3,

p. 11 6. Co py rig ht 20 11

by Sa ge

Pu bl ica tio ns .

Re pr in te d w ith

pe rm iss io n.

287

Limitations Although alternating treatments design, by definition, does not require a baseline, such data would provide important information about the relative effectiveness of the three interventions. In fact, the language production did not improve that much over time even though the differential effects of intervention could be determined. Baseline data would indicate how much, if any, gains were made. Another possible limitation is multiple treatment interference, particularly since the conditions were presented in close prox- imity (3–4 per day). Griffon’s data showing a gradual increase in the RI con- dition over time could be due to multiple treatment interference.

Summary A summary of the relevant dimensions of this study is found in Table 12-1.

Alternating Treatments Design with No Baseline (with a no-treatments phase)

Caldwell, M. L., Taylor, R. L., & Bloom, S. R. (1986). An investigation of the use of high- and low-preference food as a reinforcer for increased activity of indivi- duals with Prader-Willi syndrome. Journal of Mental Deficiency Research, 30, 347–354.

TABLE 12-1 Summary of “The

Differential Effect of Three Natural Language

Interventions on Language Use in Children with

Autism.”

FEATURE DESCRIPTION

Type of design Alternating treatments with no baseline (without a no- treatment condition)

Purpose of the study Determine effectiveness of using three natural language interventions on language production of children with autism

Subjects Two young children with autism

Setting Treatment room at a communications disorder center

Dependent variable Single word production for one child and noun-verb and noun-adjective combinations for the other

Independent variables Response interaction, milieu teaching, and a combination of the two

Results and outcomes Milieu teaching and the combination were superior for total language, prompted language, spontaneous language and requesting for one child. Results were not as clear with the other. Response interaction was slightly more effective for commenting for both children.

288 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Purpose of the study In this study, the investigators determined the effectiveness of using high- and low-preference food for increased exercise in individuals with Prader- Willi syndrome (PWS). PWS is a disorder characterized by varying levels of cognitive deficits, extreme obesity, and inappropriate food-related beha- viors. One area in which there have been conflicting reports is the lack of food preferences for individuals with PWS.

Subjects The subjects were 11 adolescents and young adults with PWS (6 males, 5 females) ranging in age from 14 to 32. The IQs of the subjects ranged from 54 to 83. Although only one subject is necessary to use an alternating treatments design, the use of multiple subjects increases the external validity of the study.

Setting The 11 subjects were participating in a 5-week residential training program for individuals with PWS that focused on weight management and the development of appropriate social and leisure skills. The subjects resided in a university dormitory during the program, and all activities related to the study were conducted on the university campus.

Dependent Variables The study was composed of two phases. In the first, the investigators deter- mined what, if any, food preferences each subject had. Previous research had produced equivocal results in this area. The investigators did find that each subject showed a definite preference for snack foods (pretzels, candy, potato chips) over healthy foods (carrots, oranges, apples). The caloric amount of each type of food was the same. The second phase used the alter- nating treatments design to determine if the food reinforcers could be used to increase activity levels of the subjects. Specifically, the dependent variable was the number of “activity units” that were generated by each subject. An activity unit was defined as engagement in a prescribed activity (e.g., walk- ing, bicycling, swimming) for a 20-minute period. Up to six activity units could be earned each day. The calories expended in the activity was greater than the caloric intake from the reinforcer.

Independent Variables The subjects were assigned to two treatment conditions and a no-treatment condition. These were high-preference food reinforcement, low-preference food reinforcement, and no food reinforcement.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 289

The Design The three treatment conditions were randomly assigned to 20 sessions (days). The following order was determined using A as the high-preference food condition, B as the low-preference food condition, and C as the no- food condition: ABCAABBCCCABACBBACBA. (Note: In most alternating treatments designs with a no-treatment condition, that condition would be specified as the A condition.)

The Intervention Each day, the subjects were told which treatment condition was in effect. Pos- ters were also used throughout the day as a reminder. They were given several opportunities throughout the day to earn up to six activity units that could be traded for the reinforcer under effect in the treatment condition for that day. The activity unit consisted of a 20-minute supervised exercise program. Each activity unit expended approximately 80–150 calories, whereas the food rein- forcers were only approximately 30 calories. Consequently there was a net loss of calories for each exchange. The study was conducted over a 20-day period. Four staff members and a program director supervised all the activities.

Obtaining the Data and Plotting the Results The collection of data consisted of simply determining how many, if any, activity units were earned each day. Data for each subject were placed on x-y graphs to determine if any treatment condition consistently produced more activity units.

Results The investigators found that the data from the 11 subjects clustered into four patterns. For 3 subjects, there was little or no effect under any of the three treat- ment conditions. For another subject, the results were unclear, with increases noted for both the high- and low-preference foods. For 2 other subjects, the high-preference food condition was superior to low-preference, although the low-preference food condition was higher than the no-treatment condition. Finally, for 5 of the subjects, the high-preference condition demonstrated clear superiority. Figure 12-3 shows the data for these 5 subjects.

Why Use an Alternating Treatments with No Baseline (with a no-treatment phase) Design? The investigators were interested in determining (a) if this sample of indi- viduals with PWS had a preference for one type of food over another and,

290 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

if so, (b) whether high-preference food could be used to increase their activity levels. Because of the nature of the program, there were a limited number of days to conduct this study. The clinical importance of having the subjects increase their activity level was evident. Because collection of baseline data before intervention is not a prerequisite for an alternating

A ct

iv it

y U

n it

6 4 2 0

Days

6 4 2 0

High-preference Condition Baseline Low-preference Condition

FIGURE 12-3 Data for five of the subjects

in the alternating treatments with no baseline (with a no-

treatment condition) study. Note. From “An

Investigation of the Use of High and Low-Preference Food as a Reinforcer for

Increased Activity of Individuals With Prader-

Willi Syndrome,” by M. L. Caldwell, R. L. Taylor, and S. R. Bloom, 1986, Journal

Reprinted with permission.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 291

treatments design, the investigators chose to begin their alternating treat- ments immediately. However, a no-treatment condition was included as one of the alternating treatments to help determine the relative efficacy of the two food reinforcement conditions. This was important because of the conflicting reports about the presence of food preferences in this population.

Limitations of the Study As noted earlier, three of the subjects did not respond to any of the three treatment conditions. This could have been due to their inability to discrim- inate among the treatment conditions. The authors did note that these three subjects had lower IQs and mention other areas in which individuals with PWS with lower IQs differ from those with relatively higher IQs. In this study, using multiple subjects confounded the overall results because there was not a consistent pattern across all subjects. However, the demonstration of a functional relationship was strengthened for those subjects who did respond consistently.

Summary A summary of the relevant dimensions of this study demonstrating an alter- nating treatments design with no baseline can be found in Table 12-2.

TABLE 12-2 Summary of “An

Investigation of the Use of High- and Low-Preference Food as a Reinforcer for

Increased Activity of Individuals with Prader-

Willi Syndrome.”

FEATURE DESCRIPTION

Type of design Alternating treatments with no baseline (with a no- treatment condition)

Purpose of the study Determine effectiveness of using food reinforcers to increase activity level

Subjects Eleven adolescents and young adults with Prader-Willi syndrome

Setting Residential summer program on a university campus

Dependent variable Activity units (20 minutes of exercise ! one activity unit) Independent variables High- and low-preference foods

Results and outcomes High-preference foods were effective in increasing activity in 5 of the 11 subjects; no or unclear effects were found in 4 subjects; 2 subjects showed increases under both high- and low-preference food conditions.

292 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Alternating Treatments Design with a Baseline Weismer, S. E., Murray-Branch, J., & Miller, J. (1993). Comparison of two meth-

ods for promoting productive vocabulary in late talkers. Journal of Speech and Hearing Research, 36, 1037–1050.

Purpose of the Study The investigators determined the effectiveness of two methods of promoting productive vocabulary in young children identified as being late talkers. The instructional methods were modeling and modeling plus evoked production.

Subjects The subjects were two boys and one girl, ranging in age from 27–28 months at the beginning of the study. Each had been identified as having restricted productive vocabularies based on a variety of test scores. The actual vocab- ularies of the three subjects consisted of 51, 52, and 87 words.

Setting The actual setting of the study was not described. The subjects received both individual and group training by three trained graduate students.

Dependent Variables The investigators developed a different set of two control words, three target words for individual instruction, and four words for group instruction for each subject. These consisted of object and action words from the Early Language Inventory. In addition, the sets of words for each treatment con- dition were also different. Thus for each subject, there was one set of two control words, two sets of three words for individual instruction, and two sets of three words for group instruction. There were three dependent vari- ables used to evaluate the effectiveness of the treatment programs:

1. The frequency of use of the targeted words per probe session;

2. The number of different words (lexical diversity) produced per probe session;

3. The number of targeted words acquired. This occurred when a subject used a target word in two or more probe sessions.

Independent Variables As noted previously, there were two treatments being evaluated. The first, modeling, occurred when the target words were stated by the investigator

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 293

but with no required response from the child. The second, modeling plus evoked production, was similar to the first in that the target words were stated by the investigator. Under this condition, however, the child was given the opportunity to repeat the word and receive feedback. A third con- dition, called approximation, was also used. This phase might be considered a no-condition phase in which the set of control words was used but with- out any modeling.

The Design The three conditions (modeling, modeling plus evoked production, and approximation) were presented in semi-random order, making sure that no more than three sessions of any one type occurred consecutively. The initial order of treatments was also counterbalanced across subjects. Baseline data were collected over four sessions before the introduction of the treatment conditions for the individual instruction only; no baseline data were col- lected for the group instruction.

The Intervention Three trained graduate students conducted the instructional sessions. Each subject was assigned a graduate student to implement all of the individual sessions. Group instruction was rotated among the three trainers. In both the modeling and modeling plus evoked production conditions, the trainer engaged in an activity in which the target words were introduced a specific number of times. In the modeling plus evoked production condition, the subject was given the opportunity to produce the target word and receive feedback. For example, for the word pen, the subject was shown a witch’s kettle and was told that she or he was making a brew by adding several things while stirring. Eventually a pen (the target word) was put in the kettle and the trainer said that the magic word is pen. The subject then was told it was her or his turn to say the magic word and to put the object in the kettle. If the subject said pen, she or he was told, “Right, that’s a pen.” If no response was given, she or he was told, “It’s the pen, isn’t it?”

Obtaining the Data and Plotting the Results After each session, production probes were conducted (e.g., “What is this? What am I doing?”), as objects or actions were shown or demonstrated. Each target word was sampled twice. The same probes were used for the control words. There were 20 sessions for individual instruction (plus 4 baseline sessions) and 24 group instruction sessions. For each subject, two graphs were generated for each of two dependent variables—frequency of target word per probe session and lexical diversity per probe session for

294 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

both individual and group instruction. As an example, Figures 12-4 and 12-5 present the frequency and lexical diversity data for Subject 3. For the third dependent variable, acquisition of the target words, the authors pro- vided a table that summarized the data.

Results The results indicated that each subject responded differently to the treatments. Subject 1 was more responsive to modeling, Subject 2 did not appear to respond consistently to either treatment, and Subject 3 responded to the model- ing plus evoked production condition. Figures 12-4 and 12-5 visually demon- strate the relative superiority of the modeling plus evoked production condition for Subject 3, particularly during individual instruction. Taken as a whole, these results would have to be interpreted as inconclusive regarding treatment efficacy. The control (approximation) condition, however, did not result in noticeable gains, so the improvement when using either of the two treatments over a no-treatment condition is evident, primarily for Subjects 1 and 3.

6 7

5 4 3 2

0 2 4 6 8 10 12 14 16 18 20

1 0#

C or

re ct

P ro

d u

ct io

n s

LT1: Indv. Instruction

Session

Baseline Treatment

AA CA

6 7

5 4 3 2

0 2 4 6 8 10 12 14 16 18 20 22 24

1 0#

C or

re ct

P ro

d u

ct io

n s

LT1: Group Instruction

Session

Treatment

A A

M MEP APPR.

FIGURE 12-4 Frequency data for Subject

3 in the baseline followed by alternating

treatments study. Note. From “Comparison

of Two Methods for Promoting Productive Vocabulary in Late

Talkers,” by S. E. Weismer, J. Murray-Branch, and

J. Miller, 1993, Journal of Speech, Language, and Hearing Research, 36,

Language, and Hearing Association. Reprinted

with permission.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 295

Why Use an Alternating Treatments Design with a Baseline? Clearly, the authors’ goal was to determine which of the two treatment con- ditions would be most effective in increasing vocabulary. The condition that included prompting the subject to respond was sufficiently different from the modeling alone condition to allow for their discrimination by the sub- jects. The authors also state that, although baseline data are not required in an alternating treatments design, they collected them for the individual sessions to “further document the lack of target vocabulary in the child’s repertoire before teaching” (p. 1040). It is unclear why baseline data were not collected for the group sessions as well.

Limitations of the Study The results of this study were somewhat inconsistent with those previously reported in the literature. The authors predicted that the modeling plus evoked production treatment would be superior. Although they do not make specific recommendations, the authors do provide cogent arguments for the presence of specific subject characteristics, such as learning style and personality factors, that might differentially affect response to

0 2 4 6 8 10 12 14 16 18 20

0 #

C or

re ct

P ro

d u

ct io

n s

LT1: Individual Instruction

Session

Baseline Treatment

A A CA

0 2 4 6 8 10 12 14 16 18 20 22 24

# C

or re

ct P

ro d

u ct

io n

s LT1: Group Instruction

Session

Treatment

A A

M MEP APPR.

FIGURE 12-5 Lexical diversity data for Subject 3 in the baseline followed by alternating

treatments study. Note. From “Comparison

of Two Methods for Promoting Productive Vocabulary in Late

Talkers,” by S. E. Weismer, J. Murray-Branch, and

J. Miller, 1993, Journal of Speech, Language, and Hearing Research, 36,

Language, and Hearing Association. Reprinted

with permission.

296 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

treatments. It is also unclear why baseline data were not collected for the group instruction.

Summary A summary of the relevant dimensions of this baseline with alternating treatments design can be found in Table 12-3.

Alternating Treatments Design with a Baseline and a Final Treatment Phase

Cihak, D. F., Smith, C. C., Cornett, A., & Coleman, M. B. (2012). The use of video modeling with the Picture Exchange Communication System to increase indepen- dent communicative initiations in preschoolers with autism and developmental delays. Focus on Autism and Other Developmental Disabilities, 27, 3–11.

Purpose of the Study The purpose of this study was to determine if the use of the Picture Exchange Communication System (PECS) accompanied with visual model- ing was effective in increasing imitative communication initiation in pre- schoolers with autism and developmental delays.

Subjects Four participants were involved in this study. Amy and Carl were diagnosed with autism; Ben and Doug were receiving services due to developmental delay. None of the four used words for verbal communication.

TABLE 12-3 Summary of “Comparison

of Two Methods for Promoting Productive

Vocabulary in Late Talkers.”

FEATURE DESCRIPTION

Type of design Alternating treatments design with a baseline

Purpose of the study Determine effectiveness of modeling procedures to increase vocabulary in young children

Subjects Two boys and one girl ranging in age from 27 to 28 months who had very limited expressive vocabularies

Setting Unspecified

Dependent variables (1) Frequency of targeted words; (2) number of different words produced; (3) number of targeted words acquired

Independent variables Modeling alone and modeling plus evoked production1

Results and outcomes Equivocal results, with each subject responding differently

1A no-treatment condition was also used.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 297

Setting The study was conducted within three inclusive preschool classrooms during center time. Amy and Ben were in the same classroom. Carl and Doug were in different classrooms. The study was conducted in each of their own classrooms.

Dependent Variable The dependent variable was the number of independent communicative initiations made by each student. This was defined as the student taking a picture of a preferred food or toy (determined through a preference assess- ment) independently (no teacher assistance or prompts) and giving it to the teacher in exchange for the item.

Independent Variables The use of the PECS alone and PECS plus video modeling were the indepen- dent variables. With PECS, pictures are used to communicate, make requests, and eventually express ideas. The video modeling component involved using a video clip depicting a model demonstrating the picture exchange.

The Design The design used was an alternating treatments design with a baseline and final treatment phase. The interventions—PECS and PECS plus visual modeling—were randomly presented. The foods and toys were counterba- lanced across students to decrease the likelihood of a student receiving the same picture more often in a specific condition.

The Intervention Baseline data were collected for three days. Students were given 10 opportu- nities to exchange a picture for a desired item (food or toy). The picture was placed in front of the student and the corresponding item was placed out of reach about 3 feet away. For the PECS only condition, the same exchange procedure was used as in the baseline condition. However, if the student failed to respond after 30 seconds, he or she was physically prompted to pick up the picture and give it to the teacher. The item was then provided to the student as a reinforcer. The PECS plus visual modeling condition used a video clip of a 4-year-old girl using a picture card to request the desired item. This video was shown immediately prior to the same PECS procedure. After criteria were met (100% independent initiation for three consecutive sessions) for the more effective intervention, it was used in a final phase to replicate the results using the materials of the less effective intervention.

298 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Obtaining the Data and Plotting the Results The percentage of independent initiations was calculated by dividing the number of independent initiations by the total number of opportunities (10). These percentages were placed on a line graph that delineated the base- line phase, the alternating treatment phase, and the final treatment phase (see figure 12-6).

Results Baseline data indicated that none of the four students made any independent initiation of picture exchanges. During the PECS only condition, indepen- dent initiations improved to a mean of 53.9%; the mean was 77.6% for the PECS plus video modeling condition. Thus, both interventions increased the initiations considerably. Individual data indicated that the PECS plus video modeling intervention was clearly superior for Amy and Carl, and was slightly more effective than PECS alone for Ben and Doug. The PECS plus video modeling intervention was the only one that met the criteria of 100% for three consecutive sessions. The PECS only intervention did reach 100% for Ben and Doug but only for one session. During the final replica- tion phase, the time required for the criteria to be met ranged from four ses- sions (Carl) to seven sessions (Ben).

1 2 3 4 5 6 7 8 9 10 11 12

Baseline Alternating Treatments VM + PECS

VM + PECS

PECS Only

Carl

13 14 15 16 17 18 19 20 21 22 23 24 25 Sessions

100

Pe rc

en ta

ge o

f I n

d ep

en d

en t E

xc h

an ge

FIGURE 12-6 Carl’s percentage of

independent exchanges in the alternating treatments

design with a baseline and a final treatment

phase study. Note. From “The Use of

Video Modeling With the Picture Exchange

Communication to Increase Independent

Communicative Initiations in Preschoolers With

Autism and Developmental Delays” by D. Cihak,

C. Smith. A. Cornett, and M. B. Coleman, 2012, Focus

on Autism and Other Developmental Disorders,

Reprinted with permission.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 299

Why Use an Alternating Treatments Design with a Baseline and a Final Treatment Phase? The authors state that an alternating treatments design was used to assess the rela- tionship between PECS only and PECS plus video modeling. The nature of the interventions allowed for rapid alternation, and the inclusion of the video model- ing component with the video clip made that condition clearly different from the PECS only condition. Baseline data were collected so the authors could clearly see the effects of the two interventions. The final phase was used to replicate the findings with different materials and to determine when, and if, the criteria would be met. Also, as noted in Chapter 11, effective treatment phases also help demon- strate that the results were not affected by multiple treatment interference.

Limitations of the Study The study was well designed and well described. Interobserver agreement and treatment fidelity were also determined and reported (93% and 95%, respectively). The authors do mention the typical limitation of single subject designs of a small sample size and questionable generalization to popula- tions other than those in the study (preschoolers with autism and develop- mental delay). Perhaps most importantly, they noted that known reinforcers were used for the exchanges. Use of the PECS plus video modeling for a variety of exchanges should be addressed.

Summary The salient components of this study that demonstrated the addition of a final treatment phase can be found in Table 12-4.

TABLE 12-4 Summary of “The Use of Video Modeling with the

Picture Exchange Communication System to Increase Independent Communicative Initiations

in Preschoolers with Autism and

Developmental Delays.”

FEATURE DESCRIPTION

Type of design Alternating treatments design with a baseline and a final treatment phase

Purpose of the study Determine if the use of PECS alone or PECS plus video modeling would be more effective in increasing independent communicative initiations

Subjects Four preschool students—two with autism and two with developmental delays

Setting Three preschool inclusive classrooms

Dependent variable Number (percentage) of independent initiations of a picture exchange

Independent variables The PECS program and PECS plus visual modeling

Results and outcomes Both interventions were effective, although the PECS plus video modeling was superior and met the criteria for success. The PECS plus video modelling was also effective when used as the only intervention.

300 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

Application Practice: Latisha Mr. Wieland is a fourth-grade teacher at Lincoln Elementary School. Lincoln Elementary is an inner-city school that has primarily minority, low-income students. Many of Mr. Wieland’s students are struggling academically; he is constantly looking for innovative ways to teach his students and to make them more motivated. Recently, Mr. Wieland attended a conference and heard about a computer software program for spelling called Personal Best Spelling. He found out that this Australian program is used worldwide. It sounded like it would be very appropriate for his students. It is designed to both teach new words and correct persistent spelling errors. To teach new words, it uses a modification of the well-known Look Say Cover Write Check method that emphasizes drill and practice. To correct spelling errors, it uses the old way/new way approach that includes error imi- tation followed by the modeling of correct spelling. Mr. Wieland was also interested in investigating the use of peer tutoring; he thought that both the tutor and the tutee would benefit from such an approach by providing social in addition to academic support.

Mr. Wieland mentioned his ideas to the school psychologist, Dr. Solomon, who suggested that he might want to conduct an experimental study to deter- mine if the two approaches were effective and, if so, which one was more effective. Dr. Solomon recommended an alternating treatments design; by using this design, the study could be conducted relatively quickly and should answer Mr. Wieland’s questions.

To set up the study, Mr. Wieland selected 40 words from the fourth- grade word lists provided in the computer program. He randomly assigned two sets of 20 words to each of the two intervention conditions—computer instruction and peer tutoring. He chose a student, Latisha, to participate in the study. Latisha has been struggling in spelling this year; school records indicated that spelling had been a problem for Latisha for several years. Mr. Wieland also picked another student, Donna, to act as the peer tutor. Donna is very strong in all academic areas including language arts, and is very outgoing. Using a coin flip, Mr. Wieland randomly assigned the order of the two intervention conditions (B ! computer instruction, C ! peer tutoring).

First, Mr. Wieland collected baseline data on Latisha’s spelling of 20 words randomly selected from the initial list of 40 words (A). These data were collected for a week. Latisha averaged between six and seven words correct. Mr. Wieland then followed the order of intervention conditions that was randomly determined. On the B days, Latisha received the com- puter instruction for 45 minutes, followed by a test over the 20 words. On the C days, Latisha worked with Donna, going over the different list of 20 words for 45 minutes, again followed by a test. Donna coached Latisha on the strategies she uses to help her spell. Data were collected for three weeks.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 301

Although both interventions were successful, the computer instruction proved to be superior, resulting in an average of 11.13 words correct with a high of 14 words correct. This represented a 68.6% increase over baseline (average 6.6 words correct). The peer tutoring resulted in an average of 8.63 words correct with a high of 10 words correct. This represented a 30.8% increase over baseline (see Figure 12-7). Subsequently, the computer instruc- tion alone was continued for a week and the results were maintained. As a result of the study, Mr. Wieland decided to adopt the software program for his entire classroom’s use.

Latisha: The Questions After reading the description of Latisha and the study that Mr. Wieland implemented, answer the following questions. Then compare your answers to those provided immediately afterward.

What is the purpose of the study?

Who are the subjects?

What is the setting?

What is the dependent variable?

2 3 4 5 6 7 8 9

10 11 12 13 14 15 16 17 18 19 20

A A A A A B C B C B B C C B C B C B B C B B B B B Days

N um

b er

o f C

or re

ct P

ro b

le m

Treatment Phase Final Phase

Baseline

Baseline Computer Instruction Peer Tutoring

FIGURE 12-7 Latsiha’s spelling data using an alternating

treatments design with a baseline and a final

treatment phase

302 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

What are the independent variables?

What kind of intervention is provided to the subject?

How were data collected and presented on graphs?

What were the results?

Why use an alternating treatments design with a baseline and final treat- ment phase for this study?

What are the limitations of this study?

Latisha: The Answers

Purpose of the Study This study was designed to determine if a computer software program— Personal Best Spelling—and peer tutoring are effective in increasing spelling accuracy, and, if so, which is more effective.

Subject The subject was a fourth-grade student who was struggling with spelling and had a history to confirm her problems in this area. Another student was chosen to act as the peer tutor based on her academic strengths, although no data were collected from her.

Setting The study took place in a fourth-grade classroom in an inner-city elemen- tary school primarily serving minority students.

Dependent Variable The dependent variable used for this study was the number of correctly spelled words from the lists of 20 words.

Independent Variables The two interventions, use of the Personal Best Spelling computer program and peer tutoring, were the independent variables for this study.

The Design An alternating treatments design with a baseline and final treatment phase was used in this study. This design allowed the rapid alternation of the two interventions to determine their relative effectiveness. By collecting baseline data first, their effectiveness could be better established. The final treatment

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 303

phase, in a sense, was used to validate the effectiveness of the superior inter- vention by demonstrating that results were maintained when the more effec- tive intervention was used exclusively.

The Intervention The interventions were the use of two approaches to increasing spelling accuracy. One intervention was the use of the Personal Best Spelling soft- ware. The other was peer tutoring. Latisha received 45 minutes of the inter- vention assigned for each day, followed by a test over the 20 words addressed during the intervention.

Obtaining the Data and Plotting the Results During the baseline phase, the subject took daily tests over 20 words ran- domly selected from a 40-word list. During the alternating treatment phase, a test was administered on the 20-word list specific to each interven- tion. In the final treatment phase, the 20-word test was administered after the more effective intervention was used exclusively. Data related to each phase was presented on an x-y graph (see Figure 12-7).

Results The intervention using the computer program proved to be superior to the peer tutoring, although both resulted in gains over baseline performance. Use of the computer program resulted in over 68% improvement, whereas peer tutoring resulted in almost 31% improvement. Use of the computer program alone also maintained and slightly increased Latisha’s high performance.

Why Use an Alternating Treatments Design with a Baseline and Final Treatment Phase? This design allowed the determination of the effectiveness of both treatments (by comparing to baseline data). It also allowed the comparative effectiveness of the two treatments during the alternating treatments condition. Finally, it allowed the determination if the superior intervention would maintain its effectiveness when used alone. If not, multiple treatment interference might be present, affecting the results of the alternating treatments phase.

Limitations of the Study This study only used one subject. Using more subjects and receiving similar results would strengthen the findings. In addition, the peer tutoring condi- tion was based solely on the tutor’s skills. Use of other tutors might result

304 PART 2 OVERVIEW AND APPLICATION OF SINGLE SUBJECT DESIGNS

in different findings. An outline or script that describes the different components of the peer tutoring in detail might result in a higher probability of obtaining similar results with different tutors.

Summary A summary of the relevant dimensions of the vignette study is presented in Table 12-5.

The alternating treatments design is sometimes rather complicated and can be difficult to manage and implement. In addition, there may be situa- tions in which a researcher wants to change behavior in a more systematic, stepwise fashion. When this is the case, the changing criterion design dis- cussed in chapter 7 might be selected.

TABLE 12-5 Use of a Computer

Software Program and Peer Tutoring to Increase

Correct Spelling in a 4th Grade Student

FEATURE DESCRIPTION Type of design Alternating treatments design with a baseline and a final

treatment phase

Purpose of the study Determine the effectiveness of a computer program Personal Best Spelling and peer tutoring on spelling performance of a struggling speller

Subjects A fourth-grade student who has a history of spelling problems

Setting Inner-city elementary school fourth-grade classroom

Dependent variable Number of correct spelling words from a list of 20 randomly selected words

Independent variables The Personal Best Spelling program and peer tutoring

Results and outcomes Both interventions were effective, although the computer program was superior. The Personal Best Spelling program was also effective when used as the only intervention.

CHAPTER 12 APPLICATION OF ALTERNATING TREATMENTS DESIGNS 305

P A R T 3 Analyzing Results from Single Subject Studies

CHAPTER 13 Methods for Analyzing Data 309

307

CHAPTER

13 Methods for Analyzing Data

IMPORTANT CONCEPTS TO KNOW VISUAL ANALYSIS

When to Use Visual Analysis Applying Visual Analysis within Phases Applying Visual Analysis Across Phases Advantages of Visual Analysis Limitations to Visual Analysis

STATISTICAL ANALYSIS When to Use Statistical Analysis How to Use Statistical Analysis Statistical Procedures

QUALITATIVE ANALYSIS When to Use Qualitative Analysis How to Use Qualitative Analysis Limitations of Qualitative Methods

KEY CONCEPTS/TERMS

POSSIBLE ANSWERS TO CHECK IT OUT

309

In this chapter, we will discuss the analysis and interpretation of datathrough three possible avenues. First, we will discuss visual analysis ofdata, which is perhaps the most commonly used method for immediately examining results. Second, we will discuss statistical analysis of data. Because many experts have criticized the sole use of visual analysis, these methods are present in the literature more frequently but also have limitations. Third, we will examine the use of qualitative data analysis. Qualitative research methods have increased in popularity, and their application to single subject studies may enhance the researcher’s ability to analyze aspects of behavior change that are not always captured in numerical variables. This type of research may be referred to as mixed methods research because it uses both quantitative and qualitative data.

The researcher should be prepared to use each and all of these methods in any given study. Each has its own particular advantages and limitations. Using these methods in combination, the researcher will likely enhance her or his ability to demonstrate a functional relationship, by increasing the amount and types of data collected. The significance of behavior change may be examined quantitatively and qualitatively. As stated many times in this text, the overall goal of the researcher is to demonstrate an experimental criterion has been met (i.e., a functional relationship exists between independent and dependent variable) as well as clinical criterion (i.e., the changes in the dependent variable are of practical value and importance to the individual participant) and the social validity of the behavior change (i.e., those involved with the participant evaluate the outcome of the study as having relevance and importance; Alberto & Troutman, 2013). The reader should not feel compelled to select one method of analysis over another, but use all reasonably applicable methods in reporting and interpreting results.

Visual Analysis Richards, Taylor, and Ramasamy (1997) noted that professionals involved in applied research and practice frequently visually analyze data in making inferences about behavioral changes. However, these same authors stressed that visual analysis has its limitations because of the subjectivity that may be involved. There may appear to be common agreement among researchers about characteristics of graphed data that should be inspected and evalu- ated, yet there are no operational criteria that appear consistently in the lit- erature (Alberto & Troutman, 2013). Therefore, visual analysis can be somewhat subjective. However, there are general principles that a researcher may use that will help to offset this potential subjectivity. We will first address when to use visual analysis, how visual analysis is applied, and what it offers the researcher. We will then address the limitations.

310 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

When to Use Visual Analysis Visual analysis of data is used generally when continuous numerical data are gathered (i.e., using the methods described in Chapter 3), data are graphically depicted, and the researcher wishes to make formative as well as summative analyses of study outcomes. The researcher uses visual analy- sis to examine two overall aspects of the data. These are the level (i.e., per- formance on the dependent variable) and trend (i.e., changes or consistent patterns in the data path; Tawney & Gast, 1984). More specifically, the researcher should examine the number of data points and their arithmetic mean in phases and across phases, the variability of performance, the level of performance, the direction and degree of trends that occur, and the over- lap of data across study phases (Alberto & Troutman, 2013; Cooper, Heron, & Heward, 2007). Tawney and Gast (1984) also stressed that these aspects should be examined both within and between phases within the study (e.g., within and between baseline and intervention phases).

Cooper et al. (2007) offered several questions that the viewer of graphed data should answer before visual analysis. Although these questions are par- ticularly relevant for a reader of research, it is worth noting that the researcher herself or himself should also consider these questions when determining how a graph should be constructed. We have modified the questions somewhat so that they appear in a manner more applicable to the researcher. These questions are

1. Are the legend, axes, and all phases labeled clearly?

2. Have you tracked the data visually to ensure that all data points are connected appropriately and whether the data points include aggregated data across observations, a single observation, a daily observation, etc.?

3. Is the scaling of the y-axis (i.e., dependent variable) appropriate? That is, do changes in performance appear to be commensurate with their clini- cal significance (e.g., a minuscule change does not appear as a huge change in behavior unless such a change would be significant). A rela- tively small change in the occurrence of a health- or life-threatening behavior might be depicted as a rather important one on the scale. Con- versely, a change of a few percentage points in the accuracy of solving math problems might not represent such a significant change.

4. Do the data depict the performance of an individual or a group treated as a single subject? In the latter case, is the range or variability of the individual performances within the group also included?

Cooper et al. (2007) stress that visual analysis should proceed only after the researcher is satisfied that the graph is correctly constructed and accu- rately depicts the results of the study and the events occurring in the study.

CHAPTER 13 METHODS FOR ANALYZING DATA 311

Applying Visual Analysis within Phases As noted previously, visual analysis may be applied primarily in two ways: (a) to inspect changes within a phase or condition and (b) to inspect changes across phases or conditions. In the former, one should examine the number of data points, the variability of performance, the level of performance, and the direction and degree of any trends (Cooper et al., 2007).

Number of Data Points within Phases The number of data points within a phase should be sufficient to make a reasonable determination of whether the data path accurately represents performance on the dependent variable. Clearly, the greater the number of data points, the more likely one may have confidence in such a determination (Cooper et al., 2007). We have heard students remark that a baseline phase should contain at least 3 data points. One might just as well say 5, 10, or 15 data points. The number of data points may be limited if it is clear that the dependent variable is not likely to change unless there is also a change in phases. For example, if the target behavior was one that the individual could not exhibit (e.g., phoneme-grapheme matches) without instruction, then the number of data points in the baseline may be very limited. There may be a need to demonstrate only that the target behavior has not been acquired. When the target behavior is one that the individual exhibits, the need to collect more data to obtain an accurate portrayal of performance cannot be set at any fixed number. For example, if the target behavior is disrupting the class, the frequency may be quite variable and achieving a steady baseline cannot be guaranteed for a preset number of observations. Of importance is the variability present within the performance.

Variability in Performance If an individual participant exhibits little variability in performance (i.e., a flat data path) or a steady data path with a clear trend (i.e., always increas- ing or decreasing), then the researcher may rely on fewer data points. When an individual’s performance fluctuates, then a greater number of data points are needed. Cooper et al. (2007) noted that the number of times a phase has been repeated may also influence the number of data points needed. In a withdrawal design (see Chapter 5), one might encounter such a circum- stance. For example, an individual is attempting to decrease her episodes of nail-biting. The intervention phase clearly depicts such a reduction to a pre- determined criterion level. A second baseline phase is implemented (the withdrawal of intervention phase). Immediately the data path depicts an upward trend in episodes of nail-biting. Although one might argue a return to baseline levels of performance is desirable from a research perspective, one might argue more persuasively that such action could be detrimental to the individual and deemed highly undesirable by other concerned

312 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

persons. Therefore, a sufficient number of data points to indicate a change in level (i.e., number of episodes of nail-biting immediately increases) and trend (across several data points the trend continues upward) is generally acceptable when there are such ethical considerations. One must also consider ethical treatment and social validity of procedures in other circum- stances as well. Take, for example, an individual who is engaged in variable rates of self-abusive (but not health-threatening) behavior. Because, during baseline phase, possibly no new effort may be made to alter the target behavior (i.e., the status quo is maintained), the researcher may limit the number of data points as well (Cooper et al., 2007).

Level of Behavior Level of behavior refers to the performance of the target behavior and pos- sible changes that would be viewed vertically. That is, a significant and immediate change in behavior should be depicted as a jump in the data path either upward or downward (see Figure 13-1). However, such changes in level can sometimes be ambiguous when there is variability in perfor- mance across phases (Morgan & Morgan, 2009).

When there is variability within a phase, the level may be determined by calculating the mean performance and drawing a horizontal line across the phase. Other options are to determine the median point in the data to draw the horizontal line or to examine the range of performance (Cooper et al., 2007). See Figure 13-2 for depiction of a mean, median, and range of level performance lines. These lines may help the researcher better determine overall performance when variability in the data path is present. Also, they become important when data are examined across phases. Generally, the greater the variability the more preferable the median line or the range may be to a mean level line of performance. Tawney and Gast (1984) sug- gested that if at least 80% of data points fall within a 15% value range of the mean level line, then the data may be considered stable and the mean level line acceptable. This is a rule of thumb and not a standard, however. Familiarity with one’s own dependent variable, the phase in effect, the nature of the desired outcomes, and the amount of data actually depicted with each point (e.g., an entire day or a single session of many in a day) all should be used in determining stability and the meaning of level as well as trend.

Trend Trend is determined by examining the direction of the data path. The researcher is concerned with whether the trend is flat, increasing, or decreas- ing and whether it is variable or stable. Trend may be “eyeballed” to an extent. When the data path depicts a clear and steady direction, the overall trend may be relatively obvious. Also, trend direction changes (e.g., from increasing to decreasing) may be reasonably clear. However, in most

CHAPTER 13 METHODS FOR ANALYZING DATA 313

instances, data points are variable and the overall trend may not be appar- ent. Therefore, the researcher may need to construct a trend line. The split- middle line method is one method for constructing a trend line. Figure 13-3 includes procedures for producing the split-middle line (Figure 13-3 is adapted from Tawney & Gast, 1984). The split-middle line provides a bet- ter portrayal of the overall trend in the data. Although rates of acceleration

D ep

en d

en t V

ar ia

b le

Observations

D ep

en d

en t V

ar ia

b le

Important change in level when objective is to

increase the target behavior (assuming appropriate

calibration of y-axis)

A B

Important change in level when objective is to decrease the

target behavior

FIGURE 13-1 Example demonstrating

change in level of behavior

314 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

y range = 1–7

y 5 5 4 3 = median 1 1 1

Sum of y = 20 Mean = 4

Sum of y = 12

Mean = 2

y 7 6 5 5 = median 5 4 3

y range = 4–12

15 B

D ep

en d

en t V

ar ia

b le

Observations

FIGURE 13-2 Example demonstrating

mean, median, and range of level of performance

CHAPTER 13 METHODS FOR ANALYZING DATA 315

and deceleration cannot be calculated accurately with equal-interval graphs as commonly used, one might use logarithmic charts to produce such rates if desirable (Tawney & Gast, 1984).

Assessing level and trend within phases is important to understand spe- cifically what is occurring during baseline phase(s) and within intervention phase(s). Of equal importance is analyzing level and trend across adjacent phases. More specifically, evaluation of the data levels and trends as the study progresses from baseline to intervention phases or vice versa is at the heart of visual analysis. Generally, when there is an obvious change in level (e.g., an individual who exhibits episodes of cursing at very high rates dra- matically decreases those to nearly zero in a few observations following the introduction of the intervention), then visual analysis of level of perfor- mance across phases can be revealing. However, if the level of performance on the dependent variable includes considerable overlap (the episodes of cursing may be decreasing but is still quite high during intervention), then analysis of trend may be useful. The more the trends are similar between phases (e.g., a trend ascending in baseline phase continues with an ascend- ing trend in the following intervention phase), the less likely trend analysis will reveal a functional relationship. The more obvious there are changes in trend (e.g., an ascending baseline trend changes to a descending interven- tion phase trend), the more helpful trend analysis can be in visual analysis in general.

20 y = 3, 7,10, 8, 10, 8, 10, 13, 12, 15

Observations

D ep

en d

en t V

ar ia

b le

Step1: Divide data into halves. Draw a vertical line either through the middle data point, or between the two middle points (if there is an even number of data points).

Step 2: Find the intersection of the mid-data point and the mid-observation for each half and draw an appropriate horizontal and vertical line at the intersection.

Step 3: Draw a line that passes through both intersections.

Step 4: Count to see if the same number of data points fall below as well as above the line. If so, stop. If not, draw a new line parallel to the orginal that meets this criterion.

FIGURE 13-3 Example of the split- middle method for

determining trend lines

316 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

Applying Visual Analysis Across Phases The principles just discussed are equally applicable when one visually ana- lyzes data as the study progresses from one phase to another. Of importance in this visual analysis is to determine if the level and trends in the data path changed in predictable fashion. If so, the researcher is more likely to have obtained verification. If those changes in level and trend are similar with other individuals, settings, and behaviors, and/or occur in response to changes in the intervention, then replication is more likely to have been achieved. We emphasize more likely because visual analysis is not always sufficient to determine these outcomes that demonstrate a functional rela- tionship between independent and dependent variable. Nevertheless, it can be useful for understanding changes in the data path.

The researcher should examine the data for (a) an immediate change in level of performance on the dependent variable when there is a phase change, (b) the overall level of performance within a phase and how that compares to overall performance in other phases, and (c) changes in trend subsequent to phase changes.

An Immediate Change in Level An immediate change in level when a phase change occurs is generally a visual indicator that the intervention (or perhaps the withdrawal of the intervention) is having some effect (or perhaps no effect or even a detrimen- tal effect). Suppose, for example, that the target behavior is consuming junk food (keep in mind we would need to define more precisely what we meant by “junk food”). A baseline reveals a high level of consumption that is stable (or possibly increasing). An intervention is introduced and the amount of junk food consumed decreases dramatically. The change in the data path would indicate that the intervention possibly had an effect on the target behavior (replication of this effect would be necessary to firmly establish the functional relationship). Conversely, assume that the target behavior is iden- tified as playing appropriately. Baseline measures reveal a steady but very low rate of appropriate play. An intervention is introduced and appropriate play changes little. This would be an indicator that there was no or a small inter- vention effect. Tawney and Gast (1984) suggested that the level change between conditions may be analyzed by using the following steps.

1. Identify the last data point in the first phase and the first data point in the subsequent phase (i.e., the two data points separated by the phase change line on the graph).

2. Subtract the smaller value from the larger value.

3. Note whether the change indicates an improvement, no change, or deterioration in performance.

CHAPTER 13 METHODS FOR ANALYZING DATA 317

Of course, the number of data points collected over time would also be an important consideration. An immediately significant change might not occur, but a steady change may be evident over a number of data measurements, which would indicate verification of a prediction of behavior change in response to the intervention. Such an instance may exemplify an overall change in level of performance, although the change may or may not be dramatic.

Comparing Performance Across Phases In general, when the range of performance in one phase does not overlap with the range of an adjacent phase, the researcher may assume a visual indi- cator that a change in behavior has occurred (see Figure 13-4). When such an effect has not been achieved, a mean or median level line constructed for each phase may allow a better comparison. Tawney and Gast (1984) sug- gested calculating the percentage of overlap of data points between phases. This is determined by (a) determining the range of data point values in the first phase, (b) counting the number of data points in the subsequent phase which fall within that range, (c) counting the number of data points in the first phase, and (d) dividing the number from (b) by the number from (c) and multiplying by 100%. Following is an example of this procedure.

David is a student with a physical disability who is placed in Mr. Edwards’ eighth-grade class. Mr. Edwards is working with David on reducing the target behavior of spending time in the bathroom. Dura- tion recording (minutes per occurrence) is used. During the baseline phase, David spent 10, 12, 10, 11, 13, 11, 9, 13, 16, and 14 minutes per occurrence. Mr. Edwards implemented differential reinforcement for low rates of behavior within a withdrawal design (overall acceptable

Variable data paths but no overlap of data

across phrases

Observations

D ep

en d

en t V

ar ia

b le

A BFIGURE 13-4 Example of nonoverlapping

data across phases

318 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

criterion level was set for 2 minutes per occurrence in three out of four consecutive observations). In the first intervention phase, David spent 9, 10, 8, 5, 3, 3, 2, 2, and 2 minutes per occurrence. For (a) above, the range of data points in the baseline phase would be 9–16 minutes. For (b), only the 9 and 10-minute observations would fall within that range. For (c), the number of data points in the baseline phase equals 10. For (d), we divide 2 by 10 and multiply by 100% to obtain a result of 20% overlap.

Generally, the smaller the percentage, the better the indication that the intervention has had an impact (assuming the phases compared are a baseline and intervention phase). Another method, is calculating the percentage of non-overlapping points (Scruggs, Mastropieri, & Castro, 1987). In this method, the researcher determines the number of data points in the interven- tion phase that are greater than the highest level in the baseline phase (or using the number of lower level of data points in intervention versus the lowest in baseline phase if the intervention is designed to diminish performance on the dependent variable) divided by the total number of data points in the treatment phase and multiplying by 100 percent. In the above example, the lowest performance during baseline for David was 9 minutes. During inter- vention, 7 of 9 data points were below 9 minutes. Therefore, the percentage of non-overlapping data points would be approximately 78%. This would suggest that when the intervention was introduced, the overall level of perfor- mance improved as a substantial number of data points during intervention were beyond the best performance during the baseline phase. Scruggs and Mastropieri (1998) suggested that when the percentage of non-overlapping data points is at least 70%, then one might reasonably judge the treatment to be effective or very effective (when the percentage is 90% or above). It is possible that outliers in data points (e.g., if David’s performance on one day during baseline phase had been 3 minutes), overall trends, and other variables can affect this type of analysis (Shadish & Rindskopf, 2007). However, if one is using visual analysis properly, a researcher should be able to discern the impact of outliers, unusual data trends, and other variables that might affect the internal validity of the study and the analysis of the outcomes.

4 C H E C K I T O U T # 1 Assume you are reviewing the results of a study you just completed. The depen- dent variable was the frequency of telling lies during one day. The frequency of telling lies during baseline was 12, 11, 10, 12, 9, 10, and 13. The frequency of telling lies following the implementation of the intervention was 10, 8, 7, 5, 4, 2, 1, and 0. Calculate the percentage of non-overlapping data and determine if you believe the intervention was effective based on these results.

CHAPTER 13 METHODS FOR ANALYZING DATA 319

Trend Changes Trend changes are evaluated in much the same way as they might be in within-phase procedures, although the researcher is typically predicting a shift in the trend (e.g., from a flat or accelerating baseline trend to a deceler- ating trend during the intervention phase). Analyzing trend changes can be very useful when level changes are less obvious. In the example of the girl who is nail-biting, there may be a change in trend, but the overall level of performance may fall within a similar range. In other words, using a per- centage of overlapping or non-overlapping points may not prove helpful. A girl could have exhibited episodes of biting her nails per day at 10, 11, 13, 12, 15, 16, and 19 during a baseline phase. There is a small but definite upward trend in her performance. When the intervention is introduced, her episodes per day are 19, 17, 14, 13, 10, 7, 6, and 4. As you can see, by using the overlapping method of analyzing level of performance, the per- centage will be substantially high for overlapping data (most data points would fall within the range of performance found during baseline). Simi- larly, the non-overlapping method would indicate only 3 data points (7, 6, and 4 episodes) would not overlap yielding a low percentage result. Yet, examining the data, we can see that there does appear to be a trend toward a reduction in episodes of the target behavior.

An immediate shift in trend is generally an indicator that the interven- tion is having the desired effect (or the opposite of the effect expected). For example, an individual’s target behavior is solving math problems. Both rate and accuracy are of importance. During baseline, the rate and accuracy are low although steady. When the intervention is introduced (a changing crite- rion for reinforcement), rate increases but accuracy decreases. The individ- ual is attempting to solve more problems but is making more careless errors. Such an immediate shift in the accuracy would certainly clue the researcher that some other intervention must be implemented. Conversely, if the indivi- dual’s rate and accuracy should both begin to accelerate, the researcher might reasonably assume that the intervention is having the desired effect, although replication of that effect would be desirable to demonstrate a func- tional relationship.

The split-middle line procedures discussed earlier are equally applicable when visual analysis is applied across phases. The overall trends may be less apparent than one might perceive at a casual glance at the data. These lines may assist the researcher in determining what the trends are and whether there are changes in trend that indicate desirable or undesirable changes in the dependent variable. The steepness of the trend line may also be exam- ined for some indicator of the strength (or lack thereof) of the intervention. For example, a data path that indicates a number of consecutive accelerating level changes (when the overall goal is to increase the target behavior) may also serve to indicate intervention effects. One should, however, also keep in

320 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

mind our earlier warning that graph construction on equal-interval graphs can be misleading if the intervals are either exceedingly large (steepness in the trend line appears more dramatic) or are exceedingly small (less steep trend line). Again, the use of logarithmic graphs may prevent this, although equal-interval graphs are still more commonly found in the literature.

Although visual analysis has been used for many years and is still com- monly employed, there are researchers that question the use of visual analysis alone as the sole means for assessing study outcomes. However, many researchers also acknowledge that visual analysis can reveal robust interven- tion effects and that subtle effects that may be ferreted out through statistical analyses, may, in fact, be less educationally or clinically significant. It is impor- tant to understand both the advantages of visual analysis and its limitations.

Advantages of Visual Analysis Perhaps the primary advantage to visual analysis is that it is relatively simple to use (assuming correctly scaled and accurate graphic displays of data). Sec- ond, educators, social workers, psychologists, and other direct service provi- ders are concerned with changes in target behaviors that are socially valid and meaningful in the real world. Visual analysis does appear sufficient for revealing strong and robust intervention effects that can literally be seen in many instances. Also, its simplicity tends to decrease the likelihood of error in analysis of data (Brossart, Parker, Olson, & Mahadevan, 2006). Visual analysis can include descriptive statistical analyses (trend lines, percentage of non-overlapping data points) that can assist visual inspection alone. Finally, visual analysis requires that the researcher look beyond the numerical data and consider outliers, unusual events, changes in level or trend, or other vari- ables that might have affected the outcomes of the study. In other words, the overall results of the study may not be clearly limited to the “numbers” alone.

Limitations to Visual Analysis There are limitations to visual analysis, however. DeProspero and Cohen (1979) noted that visual analysis may not be as reliable as statistical meth- ods of analysis. These authors were concerned that two analysts may not reliably evaluate the same data because the guidelines are neither as strin- gent nor as precise as those employed in statistical formulas. Ottenbacher (1990) found that 30 individuals could not visually rate significant changes from baseline to intervention phases on 24 graphs when compared to an objective method of analysis. Similarly, Richards et al. (1997) reported that neither undergraduate nor graduate students studying applied behavior analysis could accurately rate behavior changes in those same 24 graphs compared to the objective method used. In fact, the actual performance of

CHAPTER 13 METHODS FOR ANALYZING DATA 321

the visual raters in comparison to the objective method was less than chance. Richards et al. found that the raters were more likely to suggest a significant change had occurred when in fact one had not. Of course, in an applied research study, there are many variables to consider in determining significant behavior change, and the Ottenbacher and Richards et al. studies have their own limitations. Still, these results provide useful caution. If visual analysis is used, then the raters must be well trained and versed in the proce- dures discussed. Also, it would be helpful if two or more analysts viewed the data independently and drew conclusions that could be compared for reliabil- ity. If two or more independent raters agree on the nature and degree of changes in level and trend, and the consequent significance of the behavior change, then readers may place greater confidence in the conclusions drawn from visual analysis. Richards et al. concluded, however, that visual analysis should perhaps not be the only method used to evaluate results.

Because there are other options available to the researcher, these should be considered in addition to the use of visual analysis. The researcher may compare the statistical results to those yielded through visual analysis. Such a combination, when possible, may allow for an objective confirmation (or refutation) of a significant behavior change through statistical procedures, while providing the flexibility to examine changes in level and trend that may help explain more subtle aspects of the study.

Statistical Analysis As with visual analysis, statistical (or quantitative) analysis has its advan- tages and limitations. Morgan and Morgan (2009) point out that such anal- ysis may reduce the somewhat subjective and perhaps even idiosyncratic nature of visual analysis. Such analyses include both nonparametric and parametric procedures. We will discuss when statistical analysis may be of use, some of the more commonly used statistical analyses, and what are the advantages and limitations to the use of statistical analysis.

When to Use Statistical Analysis Kazdin (2009) suggested several situations in which statistical tests may be suitably applied in single subject research studies. First, when the researcher has failed to establish a stable baseline and a trend is evident (or perhaps not evident through visual inspection), the use of statistical analyses may assist in determining any subsequent change in the dependent variable is significant when the intervention is introduced. For example, an individual who emits self-stimulatory behavior may actually perform at a somewhat decreasing rate during baseline. No stability is present and the behavior may be changing due to variables already present in the individual or environment. Yet, the rate of change may be too slow so that intervention is still desirable. Statistical tests

322 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

may allow the researcher to scrutinize continuous shifts across phases when a change in trend is neither visually evident nor rapidly changing in level (Kazdin, 2009). Second, the researcher may find that visual inspection does not allow a clear analysis of whether or not interventions had weak effects. This may be of particular importance when the interventions used have not been thoroughly researched previously and therefore their potential impacts are less predictable. Equivocal results may make visual inspection difficult if not impossible. Statistical analysis may lead the researcher to discover reliable, although weaker, intervention effects (Kazdin, 2009). Kazdin acknowledged the argument that large effects are generally sought in single subject research, but researchers must not rule out smaller effects that are reliable as also being potentially useful for treatment. For example, interventions that are cheap, relatively easy to administer, and may assist large numbers of people may be useful as well. Kazdin shares the example of physicians encouraging patients to quit smoking as one such type of intervention that may have weak but important results in the population. Third, Kazdin argued that single subject researchers increasingly conduct their studies in less well-controlled settings with greater numbers of confounding variables. Consequently, visual inspec- tion may be more difficult to apply than statistical analysis in determining if reliable changes have been obtained.

How to Use Statistical Analysis There have been new developments in the types of statistical analyses avail- able while older methods are still discussed in the literature. In general, researchers debate almost as vigorously about the appropriateness of certain statistical analyses (particularly inferential statistical analyses) as they do about the accuracy and objectivity of visual analysis. Possible statistical pro- cedures we discuss include determining effect size of a treatment/intervention and inferential methods.

Determining Effect Size Effect size can be roughly thought as the degree to which the intervention has a quantifiable effect on the dependent variable. There are a variety of procedures for determining effect size (e.g., Parker, Vannest, & Davis, 2011, discuss nine techniques). Despite arguments about their limitations and, in the case of some authors, calls for the abandonment of their use, these remain widely in use. The percentage of non-overlapping data points first proposed by Scruggs, Mastropieri, and Castro (1987) was discussed earlier in the section on visual analysis concerning trends in data. Effect size analyses are numerical and are typically used to support visual analysis rather than as a substitute for visual analysis (Parker et al., 2011). Two more commonly used methods are the percentage of non-overlapping data points (PND) and a regression analysis approach suggested by Allison and

CHAPTER 13 METHODS FOR ANALYZING DATA 323

Gorman (1993). These methods are intended to overcome issues in analysis that might be confounded by trend existing in repeated measures of the dependent variable. Researchers disagree as to which method is more appro- priate (Campbell, 2004). The reader should refer to various sources given here and many others that are available in the literature to determine which, if any, of the possible effect size procedures might be appropriate for any given study. In general, the PND method is relatively simple, although critics (who have been answered in the literature) have argued this technique is affected by trend changes and the number of observations in the various phases being compared (Scruggs & Mastropieri, 2001). Parker et al. (2011) concluded that non-overlap techniques (including PND) do not require interval levels of data nor large data sets making these techniques relatively good fits for single subject research designs.

Statistical Procedures There are two general types of statistical procedures typically used by research- ers. Descriptive measures (e.g., mean, median, mode, frequency) are used to describe aspects of the data without inference to their statistical significance. Inferential statistics are used to test for statistical significance. Additionally, inferential statistics are used when the researcher wishes to generalize her or his findings to other individuals (or samples or populations). Descriptive statis- tics are typically not used with a specific aim toward generalization of results. Inferential statistics may be used to test the significance of relationships between or among variables or to test the significance of differences between or among groups of participants (e.g., analysis of variance [ANOVA], t test). Assumptions to be met to use descriptive statistics are minimal. That is, the researcher may statistically describe her or his subjects (e.g., mean perfor- mance, median performance on the dependent variable) without having to meet prescribed conditions. Therefore, we will not discuss the use of descriptive statistics. Their use is comparatively simple and easily understood.

The use of inferential statistics does demand that certain assumptions be met. These assumptions differ depending on which statistical procedures are used. Inferential statistics may be further subdivided into parametric and non- parametric, with each type having its own general assumptions. The reader should refer to a source that thoroughly examines the assumptions that must be met before employing any particular statistical test (e.g., see Kratchowill & Levin, 1992 and Kazdin, 1984 for very thorough discussions of assumptions and applications of many different statistical procedures in single subject designs). To do so here would require considerable devotion of words to procedures that would take us far afield of the issues most relevant to the novice single subject researcher. However, serial dependency or autocorrela- tion with single subject data must be addressed, as it serves as a violation of assumptions for the use of some inferential statistical tests (Kamil, 1995).

324 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

Particularly with parametric tests, there is an assumption that observa- tions are independent of one another. This may not be the case with single subject research, as the researcher is working with the same individual for each observation of data, rather than making observations of different indi- viduals as in group studies. Also, the fact that the same observer often is collecting the data for each observation also introduces the likelihood of autocorrelation (or serial dependency) in the data (Busk & Marascuilo, 1992). In fact, we have stated repeatedly that prediction, verification of pre- dictions, and replication of those verifications are mandatory to demonstrat- ing a functional relationship. Successive observations in a single subject design tend to be correlated, which means they are serially dependent (Kazdin, 1984). The correlation among data points tends to allow a predic- tion of future performance based on present performance (and hence the observations are not truly independent). The extent of this dependency among successive observations is assessed by examining autocorrelation (Kazdin, 1984). Kazdin states that autocorrelation refers to a correlation between data points separated by time intervals or lags. When these viola- tions occur, findings of significance may be more likely (Kamil, 1995). This presents the problem of committing a Type I error. That is, because of auto- correlation, use of parametric statistics may be inappropriate as their use assumes independence of each observation (Barlow, Nock, & Hersen, 2009).

A Type I error, in everyday language, refers to the researcher concluding that a significant effect has been achieved from the independent variable when this is not actually the case (a false rejection of the null hypothesis). Autocor- relation increases the likelihood that a statistically significant result may be obtained, leading to a Type I error. Busk and Marascuilo (1992) reviewed the findings of several experts and concluded that the risk of Type I error may be doubled or higher when single subject data are autocorrelated. A lag 1 analysis is typically used to determine if this violation has occurred.

Autocorrelation of lag 1 refers to the pairing of data points from the same distribution. The first of the data points is paired with the second, the second with the third, the third with the fourth, and so on. A correlation coefficient is then calculated. Kamil (1995) recommends the use of Bartlett’s Test (r). If the correlation coefficient is not zero or close to it, it is possible autocorrelation has occurred. Single subject designs tend to include time-series data where data points may be more related to those observed closer in time than to their overall mean, and that researchers should use procedures that address autocorrelation in analyzing their data (Barlow et al., 2009).

t tests and ANOVAs Experts have discussed the use of t tests and ANOVAs that may be used to compare data between phases (e.g., baseline and intervention phases or between more than two treatments) with the same dependent variable. In essence, the measures within each phase are aggregated and compared

CHAPTER 13 METHODS FOR ANALYZING DATA 325

(Busk & Marascuilo, 1992). The unit of analysis (what is actually com- pared) may be the means or medians for each phase and, to an extent, depends on whether changes in level of responding or changes in trend are of concern (Busk & Marascuilo, 1992). The t statistic or F test (in the case of ANOVA) is then computed; should a significant difference be obtained, the researcher may then determine which phase represented the higher mean (overall average across all data points within a phase). Subsequently, the researcher would consider whether the direction of difference was support- ive of the prediction made. For example, a significantly higher mean might be obtained for the baseline phase. If the overall goal was to significantly reduce the target behavior, then one might surmise the goal had been met. Similarly, if the mean were significantly higher for the treatment phase when the overall goal was to increase the target behavior, then the goal would be met. However, Barlow et al. (2009) caution that the assumption underlying the use of t-tests and F-tests are likely to be violated in single subject research even though the tests are robust. Perhaps the most significant issues are that these tests may not account for trends in within-phase data and autocorrelation among data.

Busk and Marascuilo (1992) pointed out that these statistical procedures are to be used when autocorrelation is insignificant, which may be seldom with single subject data. Other procedures may be used to help overcome this potential problem. One of these is time series analysis.

Time Series Analyses Kazdin (2009) noted that with the use of time series analysis, a significant change in the mean across phases may be detected. See Figure 13-5 for exam- ples of data paths across phases. In example A in Figure 13-5, the continuing trend from baseline to intervention makes visual inspection difficult and con- founds the use of the t test or ANOVA. In example B, a change in trend occurs, but no significant change in level. In example B, both visual analysis and t tests or ANOVAs are confounded by the fact that overall performance across phases is not substantially different (i.e., the data points obtained in each phase are very similar to those obtained in the other phase). In example A, the lack of change in trend confounds the analysis. In example B, the lack of change in level confounds the analysis. Time series analysis is useful in asses- sing both changes in level and slope across phases (Kazdin, 2009). Further, time series may be especially helpful when visual analysis is complicated or difficult because of data outcomes, when variability in the data is large, or when intervention effects are neither rapid nor marked (Kazdin, 2009). The final graph in Figure 13-5 includes an example of data that could lend itself to statistical analysis as the mean performance across the phases could be com- pared but the variability in the data paths would have to be considered as well regarding calculating the means.

326 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

Observations

Example B

Example A

D ep

en d

en t V

ar ia

b le

D ep

en d

en t V

ar ia

b le

Variable data paths but no overlap of data

across phrases

Observations

D ep

en d

en t V

ar ia

b le

A B

Change in level occurs but trend confounds analysis

Change in trend occurs but overlap in data across phases confounds analysis

FIGURE 13-5 Example of confounding data paths across phases

CHAPTER 13 METHODS FOR ANALYZING DATA 327

Time series analysis also has its limitations. Kazdin (2009) noted that time series analysis requires a relatively large number of data points in each phase ranging from 20 to 100, but variations have been used with smaller numbers of data points. However, short phases may confound the reliability of results. That is, the researcher may not be able to identify the processes within the observation series itself to select a model that fits the data (Kazdin, 1984). When neither variations of t tests, ANOVAs, nor time series analyses are suitable, the researcher may use randomization tests.

Randomization Tests Because randomization tests are nonparametric, the assumptions for their use are less rigorous than those for the use of parametric statistics. Haardor- fer and Gagne (2010) point out randomization tests may be useful when data are from a sample and the population parameters are unknown. Kamil (1995) stated that in a randomization test, a conventional statistic is calculated for repeated orderings of the data. The proportion of statistically significant results is then used as the test for how rarely the value obtained for the actual ordered pairs would be obtained due to a random effect (i.e., not due to a treatment effect). That is, the results indicate whether one can say with confidence whether the obtained result would be due to a chance effect. If the probability of this is small (e.g., p < .05), then one may conjec- ture the obtained result is due to the intervention effect. Computer analyses are necessary because the number of permutations of the data becomes quite large with even a small number of observations. Randomization tests offer an alternative to t tests, ANOVAs, or time series analyses. Edgington (1992) also provided a discussion of the possible nonparametric procedures that may be used in randomization tests. Haardorfer and Gagne (2010) suggest that randomization tests are a valid option and not a panacea for statistical analysis of single subject research.

4 C H E C K I T O U T # 2 You have recently completed a study where the number of minutes spent in med- itation per day was the dependent variable. You have done a lag 1 analysis and determined the autocorrelation for the data in baseline was .20 and during inter- vention was .40. Based on these autocorrelations, would you recommend the use of a parametric test of significance to determine if the intervention was effective?

Limitations to Statistical Analysis As previously discussed, there are assumptions that must be met to use many statistical procedures. These assumptions may be violated in single sub- ject research, making use of those procedures tenuous. Statistically significant

328 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

results may be quite useful, but they could conceivably mislead the researcher into assuming a significant intervention effect had been achieved (i.e., the indi- vidual’s personal functioning had been significantly improved) when in fact it had not. Statistical analyses generally require knowledge of computer statisti- cal packages. In general, time series analyses are probably the more appropri- ate analyses for single subject research although the research question, data collection methods, and the experimental design should be the determining factors of which is the most appropriate technique (Barlow et al., 2009). Statistical analyses may best be used to supplement visual analysis.

In addition to visual and statistical analyses, qualitative analyses may prove useful to the researcher. These procedures, in simple terms, are more concerned with the telling of the story of the study and the people involved than analyzing quantitative data for a statistically significant result. For this reason, their use may become increasingly popular as yet another supplement to aid in analyzing and reporting the outcomes of single subject research.

Qualitative Analysis Over time, the use of qualitative research methods has increased in popular- ity and acceptance. Qualitative research methods may also be combined with quantitative methods in what is typically referred to as mixed methods research. Single subject data are typically quantitative but as we have noted, there are limited guidelines in what is best practice in how best to analyze those data. Therefore, novice researchers would be wise to consider using qualitative methods to supplement their efforts (i.e., use mixed methods) to determine the real world significance of the outcomes of their studies.

Qualitative research can be defined as research that has as its primary purpose the determination of relationships, effects, and causes that focus on individual variables (McWilliam, 1991). Denzin and Lincoln (1994) stated that qualitative researchers study events in their natural settings, and attempt to make sense of, or interpret, phenomena in terms of the meanings that people involved place on them. Qualitative analyses may involve case studies, personal experiences, introspection, life stories, interviews, and observational, historical, interactional, and/or visual texts (Denzin & Lincoln, 1994). In qualitative research, the experiences and reactions of the individual researcher as well as other participants (potentially including the individual subject) are very important. This inevitably opens the door to judgments that are more subjective than the types we have previously dis- cussed. However, the judgments can also be viewed as outcomes of the research. For example, a mother believing her child has improved substan- tially in performance on a dependent variable may, in itself, be a desirable outcome above and beyond what visual or statistical analyses may indicate. Such an outcome might be the result of the mother’s personal observations

CHAPTER 13 METHODS FOR ANALYZING DATA 329

beyond the research setting or perhaps her judgment as to what is an impor- tant change in behavior. This qualitative data may be as important in under- standing study outcomes for that individual participant as would more quantitative data outcomes.

We will discuss when to use qualitative methods, what types of methods might be used, and what are the limitations of these methods. Although the use of qualitative procedures has become increasingly popular and is worthy of inclusion within single subject research, we would not suggest abandon- ing the designs and methods to which this text is primarily devoted. The use of qualitative methods may enhance the interpretation of the data, the social validity of the outcomes, and the telling of the story of the individuals involved. Single subject research is by its very nature personal and individu- alized, which in turn lends itself to qualitative research procedures.

When to Use Qualitative Analysis Qualitative methods may be used in virtually any single subject study. In fact, the use of qualitative methods is built into single subject methodology. For example, descriptions of the problems, the settings, the individuals involved, people’s perceptions of the ethics of the methods used, and the validity of the outcomes may all be found in the literature. The story of the study is told in many journal articles as supplementary to the discussion of the visually or statistically obtained results. We stress the complementary nature of qualitative research and single subject research. Readers wishing to employ exclusively or primarily qualitative methods should consult texts that are devoted to those methods.

McWilliam (1991) noted that the methods of both qualitative and quan- titative research may be combined as follows. First, qualitative principles would be violated if the researcher spent too little time observing. A study comprising only 2–3 weeks may not allow the gathering of information that would allow the interpretation of qualitative data. Second, the researcher should maintain field notes. Personal interpretation of the events is impor- tant in qualitative research; numerical data alone collected via the methods discussed in Chapter 3 would not be sufficient for this type of analysis. Third, the researcher must be involved in the study and familiar with the individuals who are subjects, those who may be involved in collecting data and applying the treatment variable, and those whose lives may be affected by the outcomes (e.g., family members). If the researcher is distant, then she or he will be unable to access the information necessary to relate the inter- pretation of events by others. Fourth, the researcher must be prepared to collect data on variables that become important as the study evolves. Fifth, the researcher should interpret what is seen and heard. Finally, biases related to the study should be acknowledged (McWilliam, 1991). For the

330 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

combination of qualitative and quantitative methods to be successful, quan- titative principles should be adhered to simultaneously. First, interobserver reliability data must be available for formal quantitative observations. Sec- ond, operational definitions of variables must not be subjective. Third, a sufficient number of individuals (or in some designs, a sufficient number of phases) must be involved. Fourth, phases must not be too long to accommo- date collection of qualitative data. Fifth, confounding variables must be identified and controlled. Finally, the discussion of the formal observations must not stray too far afield of the objective data (McWilliam, 1991).

Creswell (2008) suggested that mixed methods may be used in at least two different types of studies. First, in an embedded mixed methods design, the researcher collects both quantitative and qualitative data simultaneously, but the overall research outcomes are associated with the quantitative data outcomes. The qualitative data has secondary status and is used in a support- ive form. In this supportive form, the qualitative data is used to answer a dif- ferent research question than that answered by the quantitative data. For example, a researcher might use quantitative data to determine if an interven- tion had a significant effect on the dependent variable. Secondarily, the researcher might use qualitative data to determine if the subject him- or her- self believed the change in behavior was significant. A second design identified by Creswell (2008) that might lend itself well to single subject research is the explanatory mixed methods design. In this design, the quantitative data also has the major emphasis in answering the primary research question. The qualitative data is then collected later to better explain key results, typical cases, outliers, or other unusual or important aspects of the study.

How to Use Qualitative Analysis Denzin and Lincoln (1994) summarized a variety of techniques that may be useful in qualitative research. We have edited their summary to those we believe to be more relevant in complementing single subject research. Denzin and Lincoln delineated five phases in the research process, and we will include their suggestions (with some modification) as they outlined them by those phases.

Phase 1—The researcher examines his or her own multicultural nature and places himself or herself in historical context. Conceptions of self and others as well as the ethics and politics of the research may be included.

Phase 2—Theoretical paradigms and perspectives are discovered and applied as appropriate. These may include different theories and approaches to counseling for example.

Phase 3—Research strategies are identified. These include the study design, case study methods, ethnography, participant observation, phenomenology,

CHAPTER 13 METHODS FOR ANALYZING DATA 331

biographical method, historical method, applied methods, and clinical meth- ods. Specific to single subject research, we have already discussed the research strategies that are typically used. Mixed methods are to be consid- ered an asset rather than a liability. That is, there use assists in explaining the results and outcomes rather than investing complete and total confi- dence in quantitative analyses alone. Phase 4—Methods of collection and analysis are identified. These may include interviewing; observing; examination of artifacts, documents, and records; visual methods; personal experience methods; data management; computer-assisted analysis; and textual analysis. Analysis may include statistical analysis, visual analysis, and various qualitative methods. Phase 5—The researcher practices the art of interpretation and presenta- tion. The researcher must identify criteria for judging the adequacy of the research, study the art and politics of interpretation, write interpretively, analyze policy as appropriate, evaluate traditions, and apply the results.

Denzin and Lincoln (1994) elaborated on Phase 5. They noted that the researcher creates a field text where observations, notes, and documents are gathered. Next, the researcher moves from this field text to a research text. Notes and interpretations are made based on the field text. Next, this text is used as a working interpretive text that includes initial attempts to make sense out of what has been learned. Finally, the researcher produces the public text to be read by others. This final tale may be confessional, realistic, impression- istic, critical, formal, literary, or analytic (Denzin & Lincoln, 1994).

It is also essential to understand that while the interpretations of the research by the researcher are important, the interpretations of the research outcomes by other participants including the “subject(s)” are often equally important. Therefore, familiarizing oneself with the possible methods for gathering qualitative data and analyzing those data (e.g., from interviews, questionnaires, observations) is desirable for the novice researcher.

The following serves as an example of how quantitative and qualitative approaches might be blended.

David was a student in fourth grade with attention-deficit/hyperactivity disorder. Three teachers worked with David, and his mother was very involved in his education. The teachers had noted that David, despite receiving medication for 3 months, still seemed to have attentional diffi- culties and subsequent problems in disrupting classes with inappropriate comments and movements. Dr. Katzen, the school psychologist, was consulted to assist with a program to improve the situation. Dr. Katzen interviewed (on several occasions) all three teachers, David’s physician, his mother, and David himself. She discovered that each had differing opinions as to why David behaved the way he did, who actually had a problem, and what should be done (these opinions differed based on the

332 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

relationship to David, how often the interviewee interacted with David, years of experience and training, and among the teachers their respective areas of certification). After observing David and consulting with each principal, an operational definition was devised for disrupting class. A differential reinforcement program, which was to be applied in a mul- tiple baseline across settings design, was agreed to by all. After a steady baseline was obtained in each setting, Dr. Katzen spearheaded the imple- mentation of the reinforcement intervention. Dr. Katzen conducted observations and took notes; she continued to interview the principals and David himself. She collected information about his academic gains as well. David’s responding reached criterion levels in the first, second, and third classrooms. In the first and third classrooms, the change took some time. In the second classroom, the change occurred rapidly. Over- all, the intervention was effective and a functional relationship (based on visual analysis and PND) was achieved. Dr. Katzen concluded the study with additional interviews and observations of David, his teachers, his physician, and his mother. After long and careful analysis of her obser- vations, field notes, and interviews, Dr. Katzen reached a number of conclusions about the study and wanted to include the perceptions and conclusions of those involved. When she reported the results of this study, Dr. Katzen found it important that David’s mother felt that the primary cause of David’s behavior change was that the teachers had learned some skills that were needed to provide structure and effectively communicate their expectations to David. The teacher in the second classroom concurred to some extent, as she perceived that her colleagues were initially ill-equipped to teach David and had actually exacerbated his problems through inappropriate reinforcement and punishment before the study’s outset. David’s physician was pleased with the results, but noted that individuals have different reactions and adjustment peri- ods to medications and the effect of David’s drug therapy may have been partly if not mostly responsible for the changes. She had seen many such cases. David himself was pleased that he no longer was get- ting in trouble, but he perceived that the real difference was the teachers were being nice to him now rather than mean. Finally, Dr. Katzen confessed that she was biased toward the idea that David had been inap- propriately reinforced and punished for some time (her training was in applied behavior analysis) and that the program was as responsible for changing the behavior of his teachers as it was David himself. She noted that David’s mother, the teacher in the second classroom, and, to an extent, David appeared to agree with this conclusion through their own perceptions and comments. She pondered to what degree the medication might have actually affected David. Ultimately, she theorized that mutual shaping had indeed occurred among the individuals. The

CHAPTER 13 METHODS FOR ANALYZING DATA 333

teachers provided reinforcement to David, David was in turn more pleasant to the teachers, this in turn made David’s mother happier, and her interactions with his teachers were more satisfactory, creating still more positive feelings toward David. All others involved concluded that Dr. Katzen’s help had been invaluable, although she felt her role had largely been one of facilitation and documentation rather than true change agent. Finally, Dr. Katzen noted that despite improved behav- ioral changes, David’s academic performance was not significantly improved at the study’s conclusion. This affected her recommendations for generalization of the use of the intervention and her conclusions about whose behavior had changed. She also decided she would need to continue following David’s case for an extended period of time to determine if academic improvement occurred and whether the behavior and/or perceptions of those involved changed over time.

In our example, we have suggested that objectively determined signifi- cance may not always tell the story of what actually happened. In fact, those involved may not agree as to why something happened or even exactly what did happen, even though they may agree the overall desired outcomes were achieved. By use of qualitative methods, the researcher is able to more fully develop and explain the data and the events (or realities), although our example by no means illustrates the breadth of qualitative research.

Limitations of Qualitative Methods Hepburn, Gerke, and Stile (1993) suggested that the major limitation of qualitative research is the time demand. Specifically, baseline and interven- tion phases may require extension in order to obtain qualitative data to understand such phenomena as mutual simultaneous shaping. The need for such may compromise the requirements for single subject methods; to not do so might compromise qualitative methods (McWilliam, 1991). Reid and Bunson (1993) noted that the resources and time needed to carry out labor- intensive approaches are scarce, that special educators have limited training in research methods, that the field tends toward the use of quantitative methods, and that quantitative methods fare better with funding agencies.

4 C H E C K I T O U T # 3 You are conducting a study examining whether a new type of cognitive therapy reduces self-reported measures of anxiety in a youth who has been incarcerated. During the intervention phase of the study, the youth is moved from one facility to another. The youth was moved as the authorities received reliable information that another youth intended to physically assault the one in therapy. You note in the

334 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

data first obtained during intervention that the self-reported level of anxiety was quite high although somewhat lower than baseline levels. After the move, the anxiety level at first increased to nearly baseline levels, and then continued to decrease to very low levels over time. Qualitatively, what might you need to include in your report of the results of the study? What qualitative data collection procedures might be used?

In this chapter, we have discussed the use of visual, quantitative, and qualitative analysis in single subject design studies. Visual analysis continues to be the most popular method for analyzing the quantitative data. Statistical analyses may be used to supplement visual analysis. Statistical analyses must be used with caution, however, because the assumptions for their use may not be met. If they are met, the decision to use statistical analyses should be made a priori. Finally, the reader is encouraged to learn more about qualitative research from sources devoted to those methods. Over time, the use of quali- tative and mixed methods has increased and the contributions made through the inclusion of mixed methods are generally recognized as important.

Key Concepts/Terms Visual analysis of data—The researcher examines aspects of graphed data

to determine if a significant change in the target behavior has occurred; visual analysis involves some subjectivity.

When to use visual analysis—When continuous numerical data are gath- ered, the data are graphically depicted, and the researcher wishes to make formative and summative analyses.

Applying visual analysis—The researcher typically inspects changes in the data within a phase or condition and changes in the data across phases and conditions.

Number of data points within a phase—There must be a sufficient number of data points to determine if the data path accurately represents the individual’s performance.

Variability in performance—The degree to which the data path indicates var- iability affects whether the researcher’s analysis may be accurate; the more variable the data within each phase, the more difficult the visual analysis.

Level of behavior—The performance of the target behavior and where along the y-axis the data points fall; when the level changes within phases, variability is created and mean, median, or range lines may be needed to assist in the visual analysis.

Trend—The direction of the data path (generally upward, downward, flat, or variable or stable); variability may necessitate the use of a split- middle line to determine trend.

CHAPTER 13 METHODS FOR ANALYZING DATA 335

Applying visual analysis across phases—The researcher is particularly interested in examining the changes in level and trend as the study moves from baseline to intervention phases.

Immediate changes in level—Assuming the change is in the desired direc- tion, immediate changes may be indicative of (but not prove) a functional relationship.

Comparing performance across phases—Examining the range of perfor- mance across phases; the fewer data points in the intervention phase that overlap with the range of performance in baseline phase (and in the desired direction), the more likely the intervention is indicative of a functional relationship.

Percentage of non-overlapping data points—In this method, the researcher determines the number of data points in the intervention phase that are greater than the highest level in the baseline phase (or using the number of lower level of data points in intervention versus the lowest in baseline phase if the intervention is designed to diminish performance on the dependent variable) divided by the total number of data points in the treatment phase and multiplying by 100 percent.

Trend changes—Changes in the direction of the data path across baseline and intervention phases; when the direction changes (e.g., from flat dur- ing baseline to increasing during intervention), the greater the indication of a functional relationship.

Limitations to visual analysis—May be less reliable than statistical analysis; two analysts may not arrive at same conclusions from visual analysis (subjectivity).

Statistical analysis—May be used primarily as a supplement to visual analysis.

When to use statistical analysis—May be used when there is variability in baseline data but a trend is evident; when changes in trend across phases are difficult to visually analyze; when visual analysis does not establish a clear intervention effect but one may be present; because applied research typically involves many extraneous variables that may affect variability of performance.

How to use statistical analysis—Descriptive and inferential statistics may be used; inferential statistics may be divided into parametric (e.g., t test, ANOVA) and nonparametric (e.g., randomization test) procedures; parametric procedures have more rigorous assumptions to be met for their use.

Autocorrelation—Occurs because the individual data collection observations in single subject research may not yield independent observations (serial dependency is present); because the same rater may be making each observation, this may also affect independence; autocorrelation may lead to violations of assumptions for the use of statistical procedures.

336 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

Type I error—The researcher infers there is a significant intervention effect when in fact there is not one.

t tests and ANOVAs—Parametric procedures used to test for mean perfor- mance differences (typically across phases).

Time series analysis—A nonparametric procedure that may be more appro- priate than parametric procedures but requires a large number of obser- vations and is somewhat complicated for the novice.

Randomization test—A nonparametric procedure, with less rigorous assumptions than parametric procedures, that is recommended by sev- eral experts.

Limitations to statistical analysis—Assumptions for use may be violated; results may not indicate educational or clinical significance; generally require knowledge and use of statistical computer packages.

Qualitative methods/analysis—Interpretivist approach in which the researcher typically attempts to tell the story of the study through the collection of data and the examination of variables that often are not quantitative. Mixed methods are increasing in popularity.

When to use qualitative analysis—May be used in virtually any study so long as the researcher has identified some variables/methods a priori and is flexible enough to identify others as the study progresses; used as a supplement to either or both visual and statistical analyses.

How to use qualitative analysis—Typically, field notes and observations are made, compiled, analyzed, and presented by the researcher; mixing both qualitative and quantitative methods.

Limitations of qualitative analysis—Time demands may be greater; may require extensions of phases that are inappropriate; resources needed may be scarce; funding agencies may be less enthusiastic toward qualita- tive approaches than quantitative approaches; reader should study these procedures further.

4 Possible Answers to Check It Out

¶ Because the study was aimed at lowering the dependent variable, youwould look for the lowest level of performance during baseline, which is 9. Then examining the 8 data points during intervention, you should find that only 1 data point is at 9 or above. Therefore, 7 of 8 intervention data points are non-overlapping which yields a percentage of approximately 87.5%. Based on the criterion that at least 70% of the intervention data should be non-overlapping, you should conclude it is an effective interven- tion. Because the percentage does not quite reach the 90% level, you should not conclude it is a very effective treatment. However, continued interven- tion data might result in achieving that 90% or better of non-overlapping data and more clearly establish performance over time.

CHAPTER 13 METHODS FOR ANALYZING DATA 337

· Autocorrelation is a serious issue that may undermine the use ofparametric statistical tests in single subject research. The autocorrela- tions should be close to zero for an assumption of independence of obser- vations (data points). The .20 and .40 levels are actually moderate enough correlations to indicate parametric tests should not be used. Instead, a ran- domization test might be useful if visual analysis and/or percentage of non- overlapping data were not reasonable options.

¸ Clearly, you would need to report the move occurred and that wasbased on the information you received from the authorities. Analyzing the numerical data prior to and after the move and in comparison to the baseline would be helpful. Still, using an interview with the youth might also help shed light on those numerical results. You would want to know if he was aware of the threat and if so, how did that affect his anxiety level. You would also want to know how the move affected him, why his anxiety first increased and then decreased over time.

References Alberto, P. A., & Troutman, A. C. (2013). Applied behavior analysis for teachers

(9th ed.). Boston: Pearson. Allison, D. B., & Gorman, B. S. (1993). Calculating effect sizes for meta-analysis:

The case of the single case. Behaviour Research & Therapy, 31, 621–631. Barlow, D. H., Nock, M. K., & Hersen, M. (2009). Single case designs: Strategies

for studying behavior change (3rd ed.). Boston: Pearson/Allyn and Bacon. Brossart, D. F., Parker, R. I., Olson, E. A., & Mahadevan, L. (2006). The relation-

ship between visual analysis and five statistical analyses in a simple AB single- case research design. Behavior Modification, 30, 531–563.

Busk, P. L., & Marascuilo, L. A. (1992). Statistical analysis in single-case research: Issues, procedures, and recommendations, with applications to multiple beha- viors. In T. R. Kratchowill & J. R. Levin (Eds.), Single-case research design and analysis: New directions for psychology and education (pp. 159–185). Hillsdale, NJ: Erlbaum.

Campbell, J. N. (2004). Statistical comparison of four effect sizes for single-subject designs. Behavior Modification, 28, 234–246.

Cooper, J. O., Heron, T. E., & Heward, W. L. (2007). Applied behavior analysis (2nd ed.). Upper Saddle River, NJ: Pearson Education Group, Inc.

Creswell, J. W. (2008). Educational research: Planning, conducting, and evaluating quantitative and qualitative research (3rd ed.). Upper Saddle River, NJ: Pearson/ Merrill/Prentice Hall.

Denzin, N. K., & Lincoln, Y. S. (1994). Introduction: Entering the field of qualita- tive research. In Handbook of qualitative research (pp. 1–17). Thousand Oaks, CA: Sage.

DeProspero, A., & Cohen, S. (1979). Inconsistent visual analysis of intrasubject data. Journal of Applied Behavior Analysis, 12, 273–279.

Edgington, E. S. (1992). Nonparametric tests for single-case experiments. In T. R. Kratchowill & J. R. Levin (Eds.), Single-case research design and

338 PART 3 ANALYZING RESULTS FROM SINGLE SUBJECT STUDIES

analysis: New directions for psychology and education (pp. 133–157). Hillsdale, NJ: Erlbaum.

Haardorfer, R., & Gagne, P. (2010). The use of randomization tests in single- subject research. Focus on Autism and Other Developmental Disabilities, 25, 47–54.

Hepburn, E., Gerke, R., & Stile, S. W. (1993). Interpretive single-subject design: A research tool for practitioner-guided applied inquiry in rural settings. ERIC Document 358981.

Kamil, M. L. (1995). Statistical analysis procedures for single-subject designs. In S. B. Neuman & S. McCormick (Eds.), Single-subject experimental research: Applications for literacy (pp. 84–103). Newark, DE: International Reading Association.

Kazdin, A. E. (1984). Statistical analyses for single-case experimental designs. In D. H. Barlow & M. Hersen (Eds.), Single case experimental designs: Strategies for studying behavior change (2nd ed., pp. 285–324). New York: Pergamon Press.

Kazdin, A.E. (2009). Single-case research designs: methods for clinical and applied settings (2nd ed.). New York: Oxford University Press.

Kratchowill, T. R., & Levin, J. R. (Eds.). (1992). Single-case research design and analysis: New directions for psychology and education. Hillsdale, NJ: Erlbaum.

McWilliam, R. A. (1991). Mixed method research in special education. ERIC Doc- ument 357554.

Morgan, D. L., & Morgan, R. K. (2009). Single-case research methods for the behavioral and health sciences. Thousand Oaks, CA: Sage.

Ottenbacher, K. J. (1990). When is a picture worth a thousand p values? A compar- ison of visual and quantitative methods to analyze single subject data. Journal of Special Education, 23, 436–449.

Parker, R. I., Vannest, K. J., & Davis, J. L. (2011). Effect-size in single-case research: A review of nine non-overlap techniques. Behavior Modification, 35(4), 303–322.

Reid, D. K., & Bunson, T. D. (1993). Pluralizing research options in special educa- tion: A roundtable discussion. ERIC Document 364008.

Richards, S. B., Taylor, R. L., & Ramasamy, R. (1997). Effects of subject and rater characteristics on the accuracy of visual analysis of single subject data. Psychol- ogy in the Schools, 34, 355–362.

Scruggs, T. E., & Mastropieri, M. A., & Castro, G. (1987). The quantitative syn- thesis of single subject research: Methodology and validation. Remedial and Special Education, 8, 24–33.

Scruggs, T. E., & Mastropieri, M. A. (1998). Summarizing single-subject research: Issues and applications. Behavior Modification, 22, 221–242.

Scruggs, T. E., & Mastropieri, M. A. (2001). How to summarize single-participant research: Ideas and applications. Exceptionality, 9, 227–244.

Shadish, W. R., & Rindskopf, D. M. (2007). Methods for evidence-based practice: Quantitative synthesis of single-subject designs. New Directions for Evaluation, n. 113, 95–109.

Tawney, J. W., & Gast, D. L. (1984). Single subject research in special education. Columbus, OH: Merrill.

CHAPTER 13 METHODS FOR ANALYZING DATA 339

Index

A and B withdrawal design, 120, 121–122, 135 A-B withdrawal design, 120, 122–123, 135,

140–144, 206 and action research, 123–124 applications of, 140–144 defined, 122, 135

A-B-A withdrawal design, 125, 125–128, 135, 144–148 applications of withdrawal designs and, 144–148

A-B-A-B four-phase withdrawal design, 95, 128–129, 135, 148–152

applications of withdrawal designs and, 148–152 A-B-A-B-A-B repeated withdrawal designs,

134–135 A-B-A-C multiple treatment design, 14, 168–170 applications of changing condition design, 186–189 defined, 168 extensions of, 169–170

A-B-C design (changing conditions design), 166–168 and Response to Intervention (RtI), 167–168 limitations of, 168

Action research and A-B withdrawal design, 123–124 defined, 123

Action Research: A Guide for the Teacher Researcher (Mill), 123

Adapted alternating treatments design, 277–280, 281 advantages of

alternating treatments design, 273–274, 280 changing conditions design, 171–172, 182 changing criterions design, 180–181 generalized reinforcers, 35–36 multiple baseline designs, 215 qualitative analysis, 329–330 statistical analysis, 322–323 visual analysis, 321 withdrawal design, 130–131

Advantages and limitations A-B design, 143–144 A-B-A design, 147 A-B-A-B design, 151–152 A-B-A-C design, 189 A-B-C design, 168 alternating treatment with baseline, 296–297 alternating treatment with final treatment phase, 300 alternating treatment with no baseline, 292,

296–297 alternating treatments design, 273–274, 280 B-A-B design, 155–156 changing conditions design, 171–172,182 changing criterion design, 180–181, 183, 198 multiple baseline across behaviors, 242–243 multiple baseline across settings, 247 multiple baseline across subjects, 250–251 multiple baseline designs, 215 multiple probe design, 254 qualitative analysis of data, 334–335, 337 statistical analysis of data, 328–329, 337 visual analysis of data, 321–322, 336 withdrawal design, 130–131

Alternating treatments designs, 264–281 adaptations of, 277–280 adapted alternating treatments, 279–280 multielement design, 277–278, 281 simultaneous treatments, 278–279, 281

advantages of, 273–274, 280 defined, 264 disadvantages of, 274–277, 280 other names for concurrent schedule design, 278 multielement design, 264 simultaneous treatment design, 264, 278–279,

281 prediction in, 271–273 replication in, 271–273

341

Alternating treatments designs (continued) types of

alternating treatments with baseline, 269–270 alternating treatments with no baseline, 265–269, 284–288

alternating treatments and final treatment phase, 271

verification in, 271–273 Alternating treatment design, applications of, 284–305 with baseline, 293–297 with baseline and final treatment phase, 297–300 with no baseline, 284–292

without a no-treatment condition, 284–288 with a no-treatment phase, 288–292

Analysis of data, descriptive frequency, 324 mean, 313, 324 median, 313, 324 mode, 324

Analysis of data, qualitative, 329 how to use, 331–334 limitations of, 334–335 phases of research process, 331–332 when to use, 330–331

Analysis of data, quantitative. See Analysis of data, statistical

Analysis of data, statistical, 322–329 how to use, 323–324

analysis of variance (ANOVA), 324, 325–326, 328 autocorrelation, 325 lag 1 analysis, 325 Bartlett’s test, 325 determining effective size, 323–324 f-ratios, 326 randomization tests, 328 t-tests, 324, 325–326, 328 time series analyses, 326–327 type I error, 325

limitations of, 328–329 types, defined, 323, 324

descriptive, 324 inferential, 324 nonparametric, 324 parametric, 324 when to use, 322–323

Analysis of data, visual, 310–322 advantages of using, 321

applications calculating percentage of overlap of data points, 318

comparing performance across phases, 318–319 immediate change in level, 317–318 level of behavior, 313 logarithmic graphs, 313–316 number of data points within a phase, 312

trend changes, 320–321 acceleration, 314–316 deceleration, 314–316 split middle line procedures, 316, 320–321 variability in performance, 312–313

disadvantages and limitations of, 321–322 questions prior to use, 321 when to use, 311

Analysis of variance, 92,324, 325–326, 328 Anecdotal recording, 74 Antecedent, 27

defined, 27 in operant conditioning, 27

Application practice of alternating treatment designs, 297–305 of changing condition designs, 189–194 of changing criterion designs, 198–204 of multiple baseline designs, 256–262 of withdrawal design, 157–163

Applications alternating treatment designs, 284–305 changing criterion designs, 195–204 changing condition designs, 186–189 multiple baseline designs, 236–262 withdrawal designs, 140–163

Applied behavioral analysis, dimensions of, 4 analytic, 4–5, 21 applied, 4, 21 behavioral, 4–5, 21

Attrition, 103, 112 defined, 103, 112 as extraneous variables, 103

Autocorrelation, 325

B-A-B (adaptation of typical withdrawal design), 132–134,136, 152–156

applications of withdrawal designs and, 152–156

342 INDEX

Bartlett’s test, 325 Baseline data, 11, 12 Baseline measure, purposes of, 11–12 Baseline phase, 11–12, 14, 21 Behavior defined, 9 as dependent variable, 11,16 importance of measurable, 54–57 importance of observable, 54–57 in operant conditioning, 27

Behavior, methods of decreasing, 41–48 aversive, 26, 41 differential reinforcement schedules, 44–47

DRO, 45, 45–46, 50 DRI/DRA, 45, 46, 50 DRL, 45, 46–47, 51

extinction, 43–44, 50 loss of privileges, 42 overcorrection, 44, 47–48, 48, 51

positive practice overcorrection, 48, 51 restitutional overcorrection, 48, 51 simple restitution, 48, 51

punishment, negative, 42–43, 50 punishment, positive, 28, 29–30, 42, 50 response cost, 43, 50 response interruption, 44, 45, 47, 51

Behavior, increasing, methods of, 26–33 negative reinforcement, 28, 30–31, 42–43 positive reinforcement, 28, 29–30, 42–43 Premack principle, 32, 34, 49 shaping, 33, 49

Behavior, maintaining. See Methods of increasing behavior

Behavior, operant, 4, 27 Behavior, recording and reporting, quantitative

methods of event-based, 4, 57–69

cumulative response recording, 73–74, 77 frequency, 57, 58–60, 77 interval recording, 61–66, 77 momentary time sampling, 65–66, 77 permanent products, 66–69 rate, 60–61, 77 trials to criterion, 73, 77

importance of measurable, 54–57 importance of observable, 54–57

time-based, 69–72 duration, 70–71, 77 latency, 71–72, 77

Behavior of Organisms, 4 Behavioral, aspect of applied behavioral analysis,

defined, 21 Behaviorists, goal of, 4

Case studies Everett vignette, 29, 32, 36–37, 47 Mark vignette, 6–7, 10–11, 13–14 Ramona vignette, 92–93, 96, 107 Taylor vignette, 56, 66, 74

Changing condition designs, 166–172, 182 A-B-A-C multiple treatment design, 14, 168–170,

186–189 defined, 168, 182 extensions of, 169–170

A-B-C design, 166–168 defined, 166, 182 limitations of, 168 and response to intervention, 167–168

advantages of, 171–172, 182 applications of, 186–189 application practice of, 189–194 defined, 166, 182 disadvantages of, 171–172, 182 goal of, 166 prediction in, 170–171, 182 procedures in using, 15 replication in, 170–171, 182 verification in, 170–171, 182

Changing criterion designs, 172–182, 182 advantages of, 180–181, 183 applications of, 185–204 return to baseline conditions, 195–198

application practice of, 198–204 defined, 172 disadvantages of, 181–182, 183 issues in, 175–180 length of phase, 175, 183 magnitude of criterion changes, 175–176, 183 number of criterion changes, 177, 183 placement of subphases, 177–180, 183

INDEX 343

Changing criterion designs (continued) prediction in, 180, 183 procedures in using, 15 replication in, 180, 183 verification in, 180, 183

Committee review, in single subject research, 110 Concurrent schedule design, 278 Confounding variable. See Extraneous variable Consequences, 27 Consequent stimulus, and satiation, 34 Content validity, 100 Continuous schedule (CR, CRF), of reinforcement,

37–38, 50 Covariance, among dependent variables,

211–215, 232 Criterion, changing, 206 in multiple baseline design, 206 in withdrawal designs, 203

Criterion, performance, 33 defined, 33 in multiple baseline designs, 206

Criterion responding, in multiple baseline designs, 226 Cumulative effect, 168 Cumulative recording, 73–74, 77

DRA. See Differential reinforcement DRI. See Differential reinforcement DRL. See Differential reinforcement DRO. See Differential reinforcement Data analysis, methods of qualitative, 329–335 quantitative, 329 statistical, 322–329 visual, 310–322

Data path, 16, 19–20, 22 Delayed multiple baseline design, 228–231, 233 Dependent variable, 9, 16 A-B withdrawal design applications and, 141 as behavior or response, 9–10 defined, 9 graphing of, in x-y graph, 16, 22 as method of measuring effects of independent

variables, 54 as target behavior, 9

Differential reinforcement, 44–47, 50 of alternative behavior (DRA), 45, 46, 50 of incompatible behavior (DRI), 45, 45, 46, 50 of low rates of behavior (DRL), 45, 46–47, 51 of other behavior (DRO), 45, 45–46, 50 when to use, 44

Direct replication, 105, 112 Disadvantages and limitations of

alternating treatments design, 274–277, 280 changing conditions design, 171–172,182 changing criterions design, 181–182, 183 multiple baseline designs, 215–216 qualitative analysis of data, 334–335 statistical analysis of data, 328–329 visual analysis of data, 321 withdrawal design, 131–132

Descriptive data analysis, 324 Due process, and single subject research, 109–110 Duration methods, of recording behavior, 70–71, 77

calculation of interobserver agreement, 70 example of, 70–71 special considerations, 71 what to record, 70 when to use, 70

Duration schedules of reinforcement, 37, 39–40

Ecological validity, 103 Educational significance, 106 Empirical validity, 110 Error variance, 95 Ethics, in single subject research, 92, 107–111 Extinction, 43–44, 50 Extraneous variables, 9–10

attrition, 103 defined, 10, 21, 103 confounding variables, 10 history, 102 interference from multiple interventions, 103–105 maturation, 102–103 multiple treatment interference, 103–105

F tests, 326 Face validity, 100

344 INDEX

Fixed duration, schedule of reinforcement, 39, 50 Fixed interval (FI), schedule of reinforcement, 39 50 Fixed ratio (FR), schedule of reinforcement, 38, 50 Follow-up phase, 13 defined, 13, 22 goal of, 13

Formulas interobserver agreement for duration recording, 70 interobserver agreement for frequency, 58–59 interobserver agreement for latency recording, 71 interobserver agreement for momentary time

sampling, 65 interobserver agreement for partial interval

recording, 64 interobserver agreement for percent correct and

incorrect responses, 68 interobserver agreement for rate, 60 interobserver agreement for whole interval

recording, 62 Four-phase A-B-A-B withdrawal design, 95,

128–129, 135 Frequency, as descriptive measure of data analysis,

324 Frequency methods of recording behavior, 57,

58–60, 77 calculation of interobserver agreement, 58–59 example of, 59 special considerations, 59–60 what to record, 58 when to use, 58

Functional relationship, 9–10 in alternating treatment designs, 274 defined, 9 as goal of single-subject research, 27 in changing conditions designs, 181 in multiple baseline designs, 224 in withdrawal designs, 122, 130

Generalization of target behavior, 40 “Grandma’s rule.” See Premack principle Graphing axis, in relation to variables, 16 baseline data, 11–12 data paths, 16, 19–20, 22

legend, 16, 22 phase change lines, 16, 19 plotting data, 16 tick marks, 16 x-y graph (line graph), 6, 16–21, 22

Graphs, types of bar, 16 contingency tables, 16 histograms, 16 line, 16 X-Y (line), 6, 16–21

History, of subject and internal validity, 102, 111

Independent variable, 9–10 A-B withdrawal design applications and, 141 defined, 9, 22, 26 in multiple baseline designs, 207 and phase change lines, 16, 22 as treatment or intervention, 9 in withdrawal designs, 130 on x-y graph, 16

Individual educational plan, and single subject research, 109

Individual written rehabilitation plan, and single subject research, 109

Individualized family services plan, and single subject research, 109

Inferential data analysis, 324 Informed consent, and single subject research, 109 Intercomponent interval length, 277, 280 Interobserver agreement, 54–56, 76 Interval recording of behavior types of, 61–66, 77

calculation of interobserver agreement, 67–68 differentiated from interval or duration rein-

forcement systems, 64–65 momentary time sampling, 65–66, 77 partial interval, 63–65, 77 when to use, 61, 63–64 whole interval, 61–63, 77

Interval schedules of reinforcement, 38–39

INDEX 345

Intervention, 28, 54 A-B withdrawal design applications and, 141–142 fidelity, 104 as independent variable, 9, 26 as treatment, 9 types of

methods for decreasing behavior, 26 methods for increasing or maintaining behavior, 26–33

Intervention phase, 12–13, 22, 269

Journa1 of Applied Behavior Analysis (JABA), 4

Latency, methods of recording behavior, 71–72, 77 calculation of interobserver agreement, 71 example of, 71–72 special considerations, 72 what to record, 71 when to use, 71

Least restrictive environment, 110 Legend, 7, 16, 22 Limited hold (LH) contingency of reinforcement, 39 Line graph, 16, 22. See also X-Y graph

Maturation, 102–103, 111 defined, 102, 111 as extraneous variable, 102–103

Mean, 313, 324 Measurable behavior, importance of, 54–57 interobserver agreement, 54–56 recording procedure, choosing, 57, 75–76

Median, 313, 324 Medication, 29 Mental illness, 29 Menus, reinforcer, 33–34 Methods of increasing behavior, 33–40 generalization of target behaviors, 40 primary, secondary and generalized reinforcers,

35–36 quality of reinforcers, 36

reinforcement schedules, 37–40, 49 reinforcer menus, 33–34 satiation, 34–35

Mode, 324 Momentary time sampling, method of recording

behavior, 65–66, 77 calculation of interobserver agreement, 65 example of, 65–66 special considerations, 66 what to record, 65 when to use, 65

Multiple baseline designs, 206–234 adaptations, 224–231 delayed multiple baseline, 224, 228–231, 233 multiple probe, 224–228, 232

advantages of, 215 multiple baseline across behaviors, 216–218 multiple baseline across settings, 218–221 multiple baseline across subjects, 221–224 multiple baseline, basic designs, 206–207 multiple baseline, delayed design, 232

applications of, 235–262 multiple baseline across behaviors, 236–243 multiple baseline across settings, 243–248 multiple baseline across subjects, 248–252 multiple probe designs, 252–256

critical issues in implementation across behaviors, 217 across settings, 220 across subjects, 223 in multiple probe designs, 226

defined, 206 designs, types of, across behaviors, 216–218, 232 across settings, 218–221, 232 across subjects, 221–224, 232 basic, 232

disadvantages of, 215–216 multiple baseline across behaviors, 218 multiple baseline across settings, 221 multiple baseline across subjects, 224 multiple baseline, basic design, 206–207, 215–216

multiple baseline delayed design, 232 functional relationship in, 224 goal of, 232

346 INDEX

mechanics of, 207–210 covariance among dependent variables, 211–215 prediction, 210–211 replication, 210–211 verification, 210–211

Multielement design, 264, 277–278, 281 Multiple probe design, 224–228, 232, 252–256 Multiple treatment interference, 103–105 defined, 103, 112 as extraneous variable, 103–105

Natural reinforcement schedules, 40 Negative punishment, 42–43, 50 defined, 50 extinction, 43–44 removal of privileges, 42 response cost, 43

Negative reinforcement, 28, 30–31, 49 Nonparametric data analysis, 324 Notations, in single subject designs, 7, 14–16, 22 Null hypothesis, 325

Observable behavior, importance of, 54–57 interobserver agreement, 54–56 recording procedure, choosing, 57

Observer drift, 97, 99–100, 111 Operant behavior (voluntary behavior), 4, 27 Operant conditioning, principles of, 4, 27 Overcorrection, 44, 47–48, 51 defined, 47, 51 positive practice, 48, 51 restitutional, 48, 51 simple restitution, 48, 51

Package interventions, 14 Parametric data analysis, 324 Partial interval recordings of behavior, 63–65 Participants, defined, 9 Percent correct/incorrect, method of recording

behavior, 67–69, 77

Percentage of non-overlapping data points (PND), 268, 269, 323, 324, 333

Permanent products, method of recording behavior, 66–69, 77

advantages of, 66–67 calculation of interobserver agreement, 68 disadvantages of, 67 example of, 68 special considerations, 68–69 what to record, 68 when to use, 67–68

Phase change lines 16, 19 and independent variable, 19

PND. See Percentage of non-overlapping data points Positive practice overcorrection, 48, 51 Positive punishment, 30, 42, 50 Positive reinforcement, 28, 29–30, 49 Prediction

in alternating treatment designs, 271–273 in changing conditions designs, 180 in multiple baseline designs, 210–211, 232 in multiple treatment designs, 170–171 in single subject research, 92, 93–94, 96, 111 in withdrawal designs, 129–130, 135

Predictive validity, 100 Premack principle, 32, 49

defined, 32, 49 “Grandma’s rule,” 32 as operant conditioning strategy, 32

Principle of normalization, 109 Probes, 226 Punishment, 26

and aversive stimuli, 41 defined, 41, 42 individual history, 42 as an intervention, 41 issues in use of, 41 negative, types of, 42–43 positive, 42

Quality of reinforcers, 36, 49 Qualitative data analysis, 329–335

advantages of, 329–330 how to use, 331–334, 337

INDEX 347

Qualitative data analysis (continued) phases of research process, 331–332 principles of, 330–331 when to use, 330–331

Qualitative research defined, 329 limitations of, 334–335, 337 methods of, 331–332 positivist assumptions of, 334

Randomization tests, 328, 337 Rate, as method of recording behavior, 58, 60–61, 77 calculation of interobserver agreement, 60 example of, 60–61 special considerations, 61 what to record, 60 when to use, 60

Ratio schedules of reinforcement, 37–38 Ratio strain, 38, 50 Reactivity, 99, 111 Recording procedure, choosing, 57, 75–76 Reflexive behavior. See Respondent behavior Regression analysis approach, 323 Reinforcement concepts and terms associated with, 28 critical issues in, 28

Reinforcement, negative, 28, 30–31 Reinforcement, positive, 28, 29–30, 49 Reinforcement schedules, for increasing or

maintaining behavior, 32, 37–40, 49 duration, 37, 39–40

fixed, 39 variable, 39

interval, 38–39 fixed (FI), 39, 50 variable (VI), 39, 50

natural, 40 ratio, 37

continuous (CR or CRF), 37–38, 50 fixed (FR), 38, 50 variable (VR), 38, 50

response duration, 39–40 Reinforcers, strategies for choosing deprivation, 35

menus, 33–34, 49 primary, secondary and generalized, 35–36 quality of, 36, 49 satiation, 34–35

Reinforcing stimuli, types of generalized, 35–36, 49 primary, 35–36, 49 secondary, 35–36, 49

Reliability, 92 areas of concern in, 96–97

Replication in alternating treatment designs, 271–273 in changing criterion designs, 180 in multiple baseline designs, 210–211 in multiple treatment designs, 169, 170–171 in single subject research, 5, 92, 96, 111 stability of baseline and, 94–96 in withdrawal designs, 129–130, 135

“Resentful demoralization,” 131–132 Respondent behavior, 4 Response, 26

as behavior, 26 as dependent variable, 26

Response cost, 43, 50 Response duration schedules, 39–40 Response interruption, 44, 45, 47, 51 Restitution, simple, 48, 51 Restitutional overcorrection, 48, 51 Reversibility, 275

Satiation, 34–35, 49 Schedules of reinforcement. See Reinforcement

schedules Science and Human Behavior, 4 Sequencing effect, 168 Serial dependency, 325 Shaping, 33, 49 Simultaneous treatments design, 264, 277–278, 281 Single-subject research, concepts of designs, types of

alternating treatment design applications of, 284–302 overview of, 264–282

analyzing data, methods for, 310–339 basic concepts and definitions of, 7–14

348 INDEX

changing condition designs applications of, 186–194 overview of, 166–172

changing criterion designs applications of, 195–204 overview of, 172–182

goal of, 27 historical aspects of, 4–7 issues in, 92–115 multiple baseline design

applications of, 236–262 overview of, 206–234

notations, 6, 14–16 phases

baseline, 11, 14, 21 follow-up, 13 intervention (treatment), 12 multiple intervention, 12 package intervention, 14

prediction in alternating treatments design, 271–273 in changing criterion design, 180, 183 defined, 93–94 in multiple baseline designs, 210–211 in single subject research, 92, 93–94, 96, 111 in withdrawal designs, 129–130

qualitative data analysis, 329–335 recording behavior, methods of, 54–78 reliability, 92, 97–100, 111

defined, 111 interobserver agreement, 97 observer drift, 97, 99–100, 111 reactivity, 99, 111

replication defined, 94, 111 in alternating treatments designs, 271–273 in changing criterion designs, 180 in multiple baseline designs, 210–211, 232 in single subject research, 94–96 in withdrawal designs, 120

statistical analysis of data, 322–329 target behaviors, methods for changing, 26–52 validity, in single subject research, 92, 100–107 validity, external, methods of establishing, 97,

105–107, 112 validity, extraneous, methods of establishing,

101–103

visual analysis of data, 310–322 withdrawal designs applications of, 140–163 overview of, 118–137

X-Y or line graph, 16–21 Skinner, B. F., 4 Social validity, 103, 110 Spontaneous recovery, 44, 50 Stability, 12 Statistical analysis of data, 322–329, 336

analyses of variance (ANOVAs), 324, 325–326, 328, 337

Bartlett’s test, 325 descriptive, 324 how to use, 323–324, 336 inferential, 324 autocorrelation, 325, 326, 336 assumptions of, 324 nonparametric inferential statistics, 324 parametric inferential statistics, 324 and randomization tests, 328, 337 serial dependency in inferential statistics,

324 tests of significance between or among

groups, 324 time series analyses, 326–328, 337 type 1 error in, 325, 337

limitations of, 328–329, 337 null hypothesis in, 325 serial dependency in, 325 when to use, 322–323

Statistical procedures, 324–329 Systematic bias, 92, 105 Systematic replication, 105–106

t test, 325–326 Target behavior, 9–10, 54

defined, 10 as dependent variable, 10 generalization of, 40

Tick marks, 16 defined, 16, 18 on x axis, 18 on y axis, 18

INDEX 349

Time series analyses, 326–328, 337 Treatment, 9. See also Intervention, defined Treatment drift, 104 Trials to criterion method of recording behavior,

73, 77 Type 1 error, 325

Validity, external, methods of establishing, 97, 105–107, 112

direct replication, 105 intersubject replication, 105 intrasubject replication, 105 systematic replication, 105–106

Validity, extraneous, methods of establishing, 101–103

attrition, 103 history, 102 multiple treatment interference, 103–105 validity, internal, 97, 101, 112

Validity, in single subject research, 92, 100–107 concerns, 99–100 defined, 100 educational significance, 106 ethics, 107–111 social validity, 103

Variables, in single subject design, 9, 14 dependent, 9,13 extraneous (confounding) variables, 10 independent variables, 9–10, 22

Variable duration (VD), schedule of reinforcement, 39

Variable interval, schedule of reinforcement, 39, 50

Variable ratio (VR), schedule of reinforcement, 38, 50

Verification, in single subject design in alternating treatments designs, 271–273 in changing criterion designs, 180 defined, 94, 111 in multiple baseline designs, 210–211, 232 in multiple treatment designs, 170–171 in single subject research, 92, 94 in withdrawal designs, 129–130, 135

Visual analysis of data, 310–322, 335 advantages of, 321 applications of, 317–321, 335 inspecting change across phases, 317–321 comparing performance across phases, 318–319, 336

equal interval graphs, 321 immediate changes in level, 317–318, 336 logarithmic graphs, 321 mean and median line construction, 313, 318 percentage of overlap of data points, 318 split middle lines, 316, 320 trend changes, 320–321, 336

inspecting change within phases, 312–316 level of behavior, 313, 335 number of data points within a phase, 312, 335

trend and trend lines, 313–316 rates of acceleration and deceleration, 314–316

split-middle line, 316 questions prior to using, 311 variability in performance, 312–313

limitations to, 321–322, 336 when to use, 311, 335

Voluntary behavior. See Operant behavior

Walden Two, 4 Withdrawal design, 120–137

advantages of, 130–131, 135 applications of, 139–163 application practice, 157–163 defined, 120 disadvantages of, 131–132, 135 practical and ethical issues, 131 “resentful demoralization,” 131–132

functional relationship in, 122, 130, 134 goal of, 125, 135 mechanics of, 125–129, 310 other names for, 120 prediction, 129–130 replication, 129–130 situations appropriate for use, 130

350 INDEX

types and adaptations of, 132–135, 136 A and B withdrawal design, 120, 121–122, 135 A-B (basic withdrawal, teaching design), 120, 122–123, 135

A-B-A, 125–128, 135 A-B-A-B (four-phase withdrawal design), 122–123, 128–129, 135

A-B-A-B-A-B (repeated withdrawals design), 134–135, 136

A-B-A-C (multiple treatments design), 168–170

A-B-C (changing conditions), 166–168

B-A-B (no initial baseline), 132–134, 136 B-C-D-C-B-E, 14

uses of, 130 verification, 129–130

X-Y graph, 6, 16–21, 22 scaling of, 18 x axis in line graphs, 18 y axis, in line graphs, 18, 22

INDEX 351