Week 3 Article Review

137

JOURNAL OF APPLIED BEHAVIOR ANALYSIS 2002, 35, 137–154 NUMBER 2 (SUMMER 2002)

USE OF A STRUCTURED DESCRIPTIVE ASSESSMENT METHODOLOGY TO IDENTIFY VARIABLES

AFFECTING PROBLEM BEHAVIOR

CYNTHIA M. ANDERSON AND ETHAN S. LONG

WEST VIRGINIA UNIVERSITY

This study evaluated a variation of functional assessment methodology, the structured descriptive assessment (SDA). The SDA is conducted in an individual’s natural environ- ment and involves systematically manipulating antecedent variables while leaving conse- quences free to vary. Results were evaluated by comparing the results of an SDA with results obtained from an analogue functional analysis with 4 children who exhibited problem behavior. For 3 of 4 participants, the results of the two assessments suggested similar hypotheses about variables maintaining problem behavior. Interventions based on the results of the SDA were implemented for 3 children and resulted in significant reductions in rates of problem behavior.

DESCRIPTORS: functional assessment, functional analysis, intervention, problem behavior

Research has demonstrated the utility of the analogue functional analysis methodol- ogy developed by Iwata, Dorsey, Slifer, Bau- man, and Richman (1982/1994) for identi- fying sources of reinforcement that maintain aberrant behavior. This methodology sys- tematically assesses situations hypothesized to be analogueous to those in the natural environment by directly manipulating pu- tative antecedents and consequences for problem behavior. The major advantage of this methodology compared to other meth- ods of functional assessment is that it allows greater control over the environment, result- ing in a more direct inference of functional relations.

In contrast to the analogue functional analysis, descriptive assessments involve di- rect observation of behavior and events in the individual’s natural environment and in-

Ethan Long is now at the Kennedy Krieger Insti- tute, Baltimore, Maryland.

We thank Carie English, Shannon Haag, Bridget Hayes, Ellen McCartney, and Mary Mich for their assistance with data collection and analysis.

Address correspondence to Cynthia M. Anderson, Department of Psychology, Box 6040, West Virginia University, Morgantown, West Virginia 26506-6040 (e-mail: [email protected]).

volve less control over environmental vari- ables. Descriptive assessments may yield in- formation about naturally occurring sched- ules of reinforcement and idiosyncratic var- iables associated with problem behavior (e.g., Fisher, Adelinis, Thompson, Worsdell, & Zarcone, 1998; Mueller, Sterling-Turner, & Scattone, 2001). As a result, descriptive assessments may enhance understanding of how reinforcement operates in the natural environment.

Recent research suggests that descriptive assessment may be beneficial in augmenting analogue functional analyses. For example, hypotheses about environment–behavior re- lations might be developed via descriptive assessment when results of an analogue func- tional analysis are inconclusive (Mace & Lal- li, 1991). Descriptive assessments also may be useful when it is difficult or impossible to conduct analogue functional analyses.

Despite these potential benefits, a number of limitations are associated with commonly used methods of descriptive assessment. First, data obtained via descriptive assess- ment provide only correlational information about environment–behavior relations, and therefore do not allow causal statements to

138 CYNTHIA M. ANDERSON and ETHAN S. LONG

be made (Mace, Lalli, & Shea, 1992). Sec- ond, if behavior is reinforced only occasion- ally, it may be difficult to identify reinforc- ing consequences (Lerman & Iwata, 1993; Sulzer-Azaroff & Mayer, 1977). Third, be- cause antecedent and consequent stimuli are not controlled, functionally relevant stimuli may not occur during the observation peri- od. This may occur, for example, if care pro- viders have modified the environment such that discriminative stimuli or establishing operations that occasion problem behavior are removed in an attempt to prevent the occurrence of such problems.

One strategy that might increase the util- ity of the descriptive assessment is to pro- gram the occurrence of specific antecedent conditions (i.e., task presentation, removal of tangible items, attention deprivation) that have been demonstrated, by research em- ploying the analogue functional analysis, to be functionally related to challenging behav- ior. To illustrate, Freeman, Anderson, and Scotti (2000) evaluated 2 individuals’ prob- lem behavior using a structured descriptive assessment (SDA). The SDA was conducted by typical care providers in the individuals’ natural environment and involved imple- menting specific antecedent conditions sim- ilar to those found in the analogue function- al analysis while observing naturally occur- ring consequences. Results of the SDA were compared to those of analogue functional analyses and typical descriptive assessments. Results indicated that the SDA resulted in greater occurrence of targeted environmental events (e.g., antecedents mentioned above and specific consequences such as attention delivery, task removal, delivery of tangible items) when compared to the descriptive as- sessment. Also, the patterns of responding in the SDA were similar to those observed in the analogue functional analysis, resulting in similar hypotheses about the function of problem behavior for the 2 participants.

Although the results obtained by Freeman

et al. (2000) tentatively support the utility of the SDA, the results were limited for sev- eral reasons. First, the authors did not de- termine if specific environmental events oc- curred more often in conjunction with prob- lem behavior (i.e., conditional probabilities were not calculated), so it is difficult to draw conclusions about potential functional rela- tions. Second, because the study was con- ducted with only 2 participants, the gener- ality of these findings were limited. Finally, the utility of this method remains unclear because interventions were not evaluated.

Given that descriptive assessments are a potential alternative or adjunct to analogue functional analyses, it is important to con- tinue to evaluate the utility of the SDA. The purpose of this study was to conduct further evaluations of the SDA by comparing results obtained with the SDA to results obtained with the analogue functional analysis with 4 participants. A second purpose of this study was to evaluate the extent to which the SDA led to development of an effective treatment. If so, a case may be made for use of the SDA as an adjunct to or replacement for an ana- logue functional analysis under certain cir- cumstances. Finally, research on the SDA may provide additional insights as to how behavior is reinforced in the natural environ- ment.

STUDY 1: FUNCTIONAL ASSESSMENT

METHOD

Participants and Setting

Four children who had been referred for the assessment and treatment of severe be- havior problems participated. Drew was an 8-year-old boy who had been diagnosed with autism and moderate to severe mental retardation. He had been referred for treat- ment of self-injurious behavior (SIB), ag- gression, and disruption. Jane, a 13-year-old

139STRUCTURED DESCRIPTIVE ASSESSMENT

girl who had been diagnosed with Down syndrome and profound mental retardation, had been referred for treatment of SIB. Lyle was a 13-year-old boy who had been diag- nosed with moderate to severe mental retar- dation. He had been referred for treatment of aggression and disruption. Mitch was a 6- year-old boy who had been diagnosed with autism and had been referred for treatment of aggression.

Analogue functional analyses for all par- ticipants were conducted in a therapy room equipped with a one-way mirror. SDA ses- sions with Drew were conducted in his home. SDA sessions for Jane, Lyle, and Mitch were conducted in their classrooms.

Response Definitions, Data Collection, and Interobserver Agreement

Analogue functional analysis. Definitions for problem behavior were developed from interviews with parents (Drew), teachers (Jane, Lyle, and Mitch), and direct obser- vations conducted prior to beginning the study. Target behaviors included SIB (Drew and Jane), defined as head banging, head hitting, ear scratching, slamming the body against hard surfaces, and banging an arm or hand against a hard object; aggression (Drew, Lyle, and Mitch), defined as hitting, kicking, pinching, and pulling hair of others; and disruption (Drew, Lyle, and Mitch), de- fined as throwing objects and knocking over furniture. Due to the severity of her SIB, Jane wore a boxing helmet to protect her ears. All target behaviors were scored using continuous frequency recording.

Frequency data were collected on the fol- lowing therapist responses for all partici- pants: (a) prompt, defined as providing a verbal, gestural, or physical request; (b) at- tention delivery, defined as 3-s to 5-s verbal statements that were not prompts; (c) deliv- ery of tangible items, defined as placing a preferred object within the reach of the par- ticipant; and (d) removal of tangible items,

defined as removing a preferred object from the participant. Partial-interval data were collected on the occurrence of the following variables across consecutive 5-s intervals: (a) attention deprivation, defined as the absence of attention for a complete interval; (b) es- cape, scored only if a prompt occurred in a preceding interval, the participant was not engaged in a previously requested task, and prompts to engage in the task were not pre- sented for the entire interval; and (c) tangi- ble deprivation, defined as the removal of a preferred tangible item for at least one com- plete interval. Therapist behaviors were cod- ed as soon as they occurred, and the order of occurrence was recorded if they occurred within the same interval as a child problem behavior. When a therapist response contin- ued to occur across adjacent intervals, it was rescored in each new interval until the event ended. For example, after attention had been absent for one complete interval, attention deprivation was scored in each subsequent interval until attention was reinstated. If both the therapist and participant continued to respond across intervals, the original order of responding was maintained at the begin- ning of each interval.

Sessions were coded either during the ses- sion or from videotapes using a computer- ized data-collection system. Interobserver agreement was calculated for frequency mea- sures by breaking sessions into consecutive 10-s intervals and dividing the smaller num- ber of responses recorded by an observer by the larger number of responses recorded by the other observer. The resulting fractions were averaged across intervals and multiplied by 100%. Mean interobserver agreement across participants was 91% for problem be- havior (range, 83% to 100%), 98% for prompts (range, 95% to 100%), 97% for attention delivery (range, 93% to 100%), 99% for delivery of tangible items (range, 98% to 99%), and 95% for removal of tan- gible items (range, 93% to 100%). For par-

140 CYNTHIA M. ANDERSON and ETHAN S. LONG

tial-interval measures, percentages of occur- rence and nonoccurrence agreement were calculated by dividing the number of inter- vals in which both observers agreed on the occurrence or nonoccurrence of the response by the total number of intervals and multi- plying by 100%. Mean occurrence and non- occurrence agreement scores for therapist be- haviors across subjects were as follows: atten- tion deprivation, occurrence 98% (range, 92% to 99%), nonoccurrence 99% (range, 97% to 100%); tangible deprivation, occur- rence 100%, nonoccurrence 100%; escape, occurrence 100%, nonoccurrence 100%.

Structured descriptive assessment. Response definitions for problem behavior were iden- tical to those described above. In addition to the boxing helmet, Jane wore arm splints due to concerns raised by her teacher. How- ever, the splints were loosened so that Jane could contact her head with her hand. Be- haviors exhibited by participants and thera- pists were videotaped and scored using a computerized scoring system. Data on ther- apist behavior were collected using 5-s par- tial-interval recording. Target therapist (teacher or parent) behaviors included prompts, attention delivery, tangible deliv- ery, attention deprivation, escape, and tan- gible removal. Prompts were defined as an instruction to complete an action, including physical prompts, and an ongoing instruc- tional context. Requests to verbalize a state- ment (e.g., ‘‘say hi’’) were coded only for Mitch, who was verbal. Attention delivery was scored when the therapist interacted with the participant in a noninstructional manner. This included reprimands, verbal statements toward the child (e.g., ‘‘you look nice today’’), and physical interaction (e.g., hug, pat on the shoulder). Tangible delivery was defined as allowing the participant ac- cess to the predefined preferred stimulus. Tangible delivery was coded if the therapist handed the item to the child, told the child that he or she could have it, or simply al-

lowed the child to independently obtain the item. Response definitions for attention dep- rivation, escape, and tangible removal were identical to those used in the analogue func- tional analysis.

Interobserver agreement, calculated as de- scribed above, was obtained for at least 30% of all sessions for all participants. Mean agreement for problem behavior was 84% (range, 74% to 100%), 94% (range, 86% to 100%), 80% (range, 50% to 99%), and 92% (range, 82% to 100%) for Drew, Jane, Lyle, and Mitch, respectively. Mean occur- rence and nonoccurrence agreement scores for therapist behaviors across participants were as follows: prompt, occurrence 94% (range, 91% to 99%), nonoccurrence 98% (range, 98% to 99%); attention delivery, oc- currence 88% (range, 80% to 94%), non- occurrence 91% (range, 85% to 100%); tan- gible delivery, occurrence 70% (range, 0% to 100%), nonoccurrence 99% (range, 98% to 100%); instruction delivery, occurrence 94% (range 91% to 99%), nonoccurrence 99% (range, 98% to 99%); attention dep- rivation, occurrence 94% (range, 83% to 99%); nonoccurrence 97% (range, 94% to 99%); tangible removal, occurrence 100%, nonoccurrence 100%; escape, occurrence 93% (range, 82% to 100%), nonoccurrence 90% (range, 75% to 100%).

Procedure

The analogue functional analysis was con- ducted prior to the SDA for all participants except Jane, whose assessments were con- ducted simultaneously. All sessions were 10 min in length. In both analyses, sessions were conducted until response differentia- tion was observed or (for Jane) until rates of problem behavior were relatively stable across conditions. Definitions for preferred stimuli were obtained through parental in- terviews and direct observation conducted prior to beginning the study. Preferred stim- uli for each participant were as follows:

141STRUCTURED DESCRIPTIVE ASSESSMENT

Drew, cookies and a mechanical singing plant; Jane, paper; Lyle, automobile maga- zines and toy cars; and Mitch, a train video and toy trains.

Analogue functional analysis. Participants were exposed repeatedly to four or five ex- perimental conditions similar to those de- scribed by Iwata et al. (1982/1994). Con- ditions were randomly presented in a mul- tielement design. All participants were ex- posed to attention, task, and play conditions. A tangible condition was imple- mented with Jane, Lyle, and Mitch, and an alone condition was implemented with Jane and Drew, both of whom exhibited SIB. The tangible condition was not conducted with Drew, because his mother indicated that removal of tangible items never evoked problem behavior and requested that the tangible condition not be conducted. Trained graduate and undergraduate stu- dents served as therapists.

Structured descriptive assessment. Partici- pants were exposed repeatedly to four exper- imental conditions: attention, task, tangible, and play. The classroom teacher served as therapist for Jane, Lyle, and Mitch. Drew’s mother served as his therapist. For all par- ticipants except Drew, sessions were con- ducted during times of the day when certain activities pertaining to each SDA condition normally occurred (e.g., attention sessions occurred when the therapist typically was unable to interact directly with the child). Sessions with Drew were conducted in ran- dom order, because his mother reported that no specific activity schedule was followed. Prior to each session, therapists were given specific instructions about the antecedent condition (detailed below) and were asked to respond to problem behavior as they typ- ically would. Because antecedent conditions were controlled as part of the assessment, therapists were asked to reestablish anteced- ent events if the designated event had not

occurred for 2 min in the absence of prob- lem behavior.

The purpose of the attention condition was to establish the antecedent of attention deprivation. This condition was conducted during times when the therapist typically did not directly interact with the child (e.g., was working with another child). Two minutes prior to beginning the session, the therapist was told, ‘‘Please play with the child as you typically do. Do not give requests or ask the child to engage in any demands at this time.’’ Upon initiation of the session, the therapist was told, ‘‘In this role-play we would like you to pretend this is a time that you cannot directly interact with the child. You may interact with other children or en- gage in another activity, such as working at your desk.’’ The therapist also was asked to keep preferred tangible items out of sight and reach of the child. The task condition, designed to establish the antecedent of pre- sentation of requests to work, was conducted when the participant was expected to com- plete tasks. Jane, Lyle, and Mitch worked on preacademic tasks, specified in their individ- ualized education plans. Drew worked on daily living skills (e.g., copying letters) and receptive language tasks. Similar tasks were used in the analogue functional analysis and the SDA. Removal of preferred tangible items or activities did not occur within 2 min of initiating this condition. At the start of the task condition, the therapist was told, ‘‘In this role-play we want to see how the child responds to requests. Please work with the child on activities that you typically en- gage the child in and use prompting strate- gies you normally use.’’ The tangible con- dition, conducted when access to preferred items was ending (access had occurred for at least 2 min), was designed to establish the antecedent of removal of preferred tangible items. At the beginning of the condition, the therapist was told, ‘‘In this role-play we want to see how the child reacts when preferred

142 CYNTHIA M. ANDERSON and ETHAN S. LONG

activities end. When we tell you to begin, please remove [preferred item]. You may in- teract with the child as you desire, but please refrain from attempting to engage the child in work activities.’’ The play condition was designed to simulate an enriched environ- ment, similar to the play condition of the analogue functional analysis. Preferred items were available, and the therapist was told, ‘‘In this role-play we would like to see how the child responds when you are not making requests and preferred items are available. Please play with the child as you normally do.’’

Data Analysis

Mean rate (number of responses per min- ute) of problem behavior was compared across conditions for the analogue functional analysis and the SDA. However, one impor- tant way that the SDA differed from the an- alogue functional analysis was that the oc- currence of antecedent and consequent events was not controlled. Although we did exert some control over antecedent condi- tions, the frequency and timing of these events were somewhat uncontrolled. Thus, conditional probabilities also were calculated for the SDA to determine the relation be- tween environmental events and problem behavior (Blakeman & Gottman, 1997; Freeman et al., 2000; Lerman & Iwata, 1993). All probabilities were calculated based on the first occurrence of child behav- ior in each interval (i.e., as though child be- havior was coded as a partial-interval mea- sure).

Conditional probabilities were calculated for consequences as described by Lerman and Iwata (1993), except that proportions were calculated for environmental events that occurred within 5 s of the problem be- havior. Also, proportions were calculated only for problem behavior that occurred in the presence of a given antecedent variable, thus taking into account the presence of pu-

tative establishing operations. For example, if a child already had access to a preferred tangible item, emitted problem behavior, and continued to have access to the item, these data were not included in the calcu- lations of tangible delivery as a consequence for problem behavior. This was necessary to control for the fact that therapists may not have implemented the relevant establishing operation throughout some portion of the sessions.

Conditional probabilities were calculated in two ways. First, behavior-based probabil- ities were calculated to determine the pro- portion of problem behavior that occurred within 5 s prior to environmental events by dividing the number of intervals containing problem behavior that preceded the environ- mental event by 5 or fewer seconds by the total number of intervals scored with prob- lem behavior. Second, event-based probabil- ities were calculated to reveal the proportion of intervals containing environmental events that followed problem behavior and were conducted to control for the possibility that the occurrence of environmental events might vary across conditions. These proba- bilities were calculated by dividing the num- ber of intervals containing problem behavior that preceded the environmental event by 5 or fewer seconds by the total number of in- tervals scored with the event.

RESULTS

Figures 1 through 4 depict mean rates of problem behavior across conditions of the analogue functional analysis and the SDA and results of the conditional probability calculations from the SDA for each partici- pant. The mean percentage of session time that antecedent variables were scored in each condition of the analogue functional analysis and the SDA are shown in Table 1. Per- centages were obtained by dividing the num- ber of intervals in which an antecedent stim-

143STRUCTURED DESCRIPTIVE ASSESSMENT

Table 1 Mean Percentage of Intervals Containing Antecedent Events Across Conditions of the Structured Descriptive

Assessment and Analogue Functional Analysis

Condition Antecedent

Drew

Structured Analogue

Jane

Structured Analogue

Lyle

Structured Analogue

Mitch

Structured Analogue

Attention Attention deprivation

97 93 81 83 81 11 91 99

Prompt 0 0 0 0 0 0 0 0 Tangible

deprivation 0 0

Demand Attention deprivation

38 28 18 68 38 23 28 13

Prompt 75 68 77 23 54 78 68 70 Tangible

deprivation 0 0 0 0

Tangible Attention deprivation

67 100 58 100 83 100

Prompt 0 0 0 0 0 0 Tangible

deprivation 94 57 100 93 100 23

Control Attention deprivation

12 75 18 77 21 73 0 73

Prompt 9 0 0 0 0 0 0 0 Tangible

deprivation 0 0 0 0 0 0 0 0

ulus was scored by the total number of in- tervals.

Drew Drew exhibited the highest rates of prob-

lem behavior in the task condition of the analogue functional analysis, suggesting that his problem behavior was maintained by es- cape from tasks (Figure 1). Drew also emit- ted the highest rates of problem behavior in the task condition of the SDA. In interpret- ing the conditional probability data, one would have the most confidence in a hy- pothesized function (e.g., escape) if that event consistently followed problem behav- ior, but only in the relevant condition (e.g., task), and other events (e.g., attention deliv- ery, tangible delivery) rarely followed prob- lem behavior in any of the conditions. Fig- ure 1 shows that occurrences of problem be- havior were often followed by escape in both

the task condition (M 5 53%) and the play condition (M 5 67%), whereas escape never occurred in the attention and tangible con- ditions. In addition, attention and tangible delivery rarely occurred in any condition. These results are consistent with the hypoth- esis that Drew’s problem behavior was main- tained by escape, with one exception: The conditional probabilities of problem behav- ior and escape were high in the play condi- tion (not solely in the task condition). How- ever, it should be noted that some prompts were presented in the play condition, which likely evoked escape-maintained problem be- havior.

Jane Jane emitted high rates of SIB in all con-

ditions of the analogue functional analysis, but rates were highest in the task condition (Figure 2). Further analysis of within-session

144 CYNTHIA M. ANDERSON and ETHAN S. LONG

Figure 1. Mean response rate (number of responses per minute) in the analogue functional analysis for Drew (top panel); mean rate of problem behavior across conditions of the SDA (middle panel); proportion of event intervals following problem behavior (bottom left panel); and proportion of problem behavior intervals preceding events (bottom right panel) during the SDA.

145STRUCTURED DESCRIPTIVE ASSESSMENT

Figure 2. Mean response rate in the analogue functional analysis for Jane (top panel); mean rate of problem behavior across conditions of the SDA (middle panel); proportion of event intervals following problem behavior (bottom left panel); and proportion of problem behavior intervals preceding events (bottom right panel) during the SDA.

146 CYNTHIA M. ANDERSON and ETHAN S. LONG

patterns of responding revealed that SIB oc- curred almost continuously throughout the task condition (i.e., during escape intervals as well as when prompting occurred; data are available from the first author upon request). Thus, results suggested that Jane’s SIB was either multiply maintained or maintained by nonsocial reinforcement. As in the analogue functional analysis, Jane emitted high rates of SIB across all conditions of the SDA, and rates were slightly higher in the task condi- tion. Event-based conditional probabilities for the SDA revealed that less than half the occurrences of targeted social variables that occurred in the presence of relevant anteced- ent conditions followed SIB, and escape was only slightly more likely to follow SIB com- pared to other consequences. Further, very small proportions of SIB were followed by any social consequence when the relevant es- tablishing condition was present. These re- sults suggest that responding was maintained by nonsocial reinforcement because SIB was not differentially related to the presence or absence of social variables.

Lyle

Lyle exhibited problem behavior almost exclusively in the attention condition of the analogue functional analysis, suggesting that problem behavior was maintained by atten- tion (Figure 3). Although he also emitted high rates of responding in the attention condition of the SDA, similar rates were ob- served in the tangible condition. This find- ing suggested that Lyle’s problem behavior was maintained by both attention and access to tangible items. Conditional probabilities for both the attention and tangible condi- tions showed that, given the antecedent of attention deprivation, the vast majority of intervals containing attention delivery fol- lowed problem behavior, and problem be- havior in both conditions was frequently fol- lowed by attention delivery (65% in atten- tion, 85% in tangible). Conversely, tangible

delivery never followed problem behavior in any condition. This finding suggested that problem behavior was maintained by atten- tion only. Inspection of overall levels of the antecedent events during the SDA (see Table 1) indicates that, in the tangible condition, tangible deprivation was in effect for 100% of the intervals; tangible items were never provided to Lyle contingent on problem be- havior in the tangible (or any other) condi- tion. Further, attention deprivation occurred in more than half (58%) of the intervals in the tangible condition. Thus, this condition resembled the attention condition. The pres- ence of attention deprivation and contingent access to attention in the tangible condition likely accounts for the high rates of problem behavior.

Mitch

Mitch emitted the highest rates of aggres- sion in the tangible condition of the ana- logue functional analysis, suggesting that problem behavior was maintained by access to preferred tangible items (Figure 4). In the SDA, Mitch emitted aggression primarily during the task condition, suggesting that aggression was maintained by escape from or avoidance of tasks. The conditional proba- bilities showed that the only consequence that occurred contiguous to problem behav- ior was escape from tasks. As shown in Table 1, tangible deprivation was in effect during a large proportion of the tangible sessions of the SDA, indicating that the relevant estab- lishing operation was in effect (see further discussion below). Nevertheless, Mitch rarely emitted problem behavior in this condition, and when he did, it was not followed by tangible delivery. Further, prompts were pre- sented in almost 70% of the intervals during the task condition, indicating that the rele- vant establishing operation also was in effect in the task condition.

147STRUCTURED DESCRIPTIVE ASSESSMENT

Figure 3. Mean response rate in the analogue functional analysis for Lyle (top panel); mean rate of problem behavior across conditions of the SDA (middle panel); proportion of event intervals following problem behavior (bottom left panel); and proportion of problem behavior intervals preceding events (bottom right panel) during the SDA.

148 CYNTHIA M. ANDERSON and ETHAN S. LONG

Figure 4. Mean response rate in the analogue functional analysis for Mitch (top panel); mean rate of problem behavior across conditions of the SDA (middle panel); proportion of event intervals following problem behavior (bottom left panel); and proportion of problem behavior intervals preceding events (bottom right panel) during the SDA.

149STRUCTURED DESCRIPTIVE ASSESSMENT

Control over Antecedent Variables

The data presented in Table 1 suggest that therapists implemented antecedent condi- tions in the SDA with a high degree of in- tegrity; that is, relevant establishing opera- tions were in effect. In fact, levels closely re- sembled those observed in the analogue functional analysis, during which presenta- tion of antecedents was tightly controlled. In the attention condition, therapists never is- sued instructional prompts or manipulated preferred tangible items, and attention dep- rivation was in effect for a large portion of the sessions. In the task condition, prompts occurred in the majority of intervals for all participants, and preferred tangible items were never manipulated. In the tangible con- dition, tangible items were never re-present- ed after the session began for all participants except Jane. Thus, tangible deprivation was in effect for all intervals scored. During the tangible condition, none of the therapists is- sued instructional prompts, and attention deprivation was scored in the majority of in- tervals for all participants. This finding sug- gested that the tangible condition resembled the attention condition. In the play condi- tion, therapists never issued instructional prompts (with the exception of Drew’s mother, who issued prompts in 9% of in- tervals), and tangible deprivation did not oc- cur. Attention deprivation occurred in less then 25% of intervals for all participants.

STUDY 2: TREATMENT

METHOD

Participants and Setting

Drew, Lyle, and Mitch participated. Jane did not participate because the frequency of her SIB decreased markedly shortly after completing the assessments and, as a result, her parents declined treatment. Intervention for Drew was conducted in his home during

speech therapy sessions. Drew’s speech ther- apist served as the therapist. Drew’s mother implemented the intervention outside ses- sion times, but she declined to participate in the study because the presence of observers reportedly made her nervous. Lyle’s treat- ment was conducted in the classroom, and his primary teacher served as therapist. Mitch’s assessment was conducted at the end of the academic year, and he began the next school year in a new class. Thus, the treat- ment evaluation took place in a new class- room for typically developing children. A teacher’s aide implemented the intervention during sessions.

Response Measurement and Interobserver Agreement

Response definitions for problem behav- iors were identical to those in Study 1. Ses- sions were 10 min long and were conducted one to three times per day, two to four times per week. Observers used frequency record- ing to collect data during sessions. Interob- server agreement was assessed on child be- haviors during at least 30% of all sessions. Agreement coefficients were calculated as in Study 1. Mean percentage agreement across baseline and treatment was 98% (range, 91% to 100%), 93% (range, 88% to 100%), and 93% (range, 82% to 100%) for Drew, Lyle, and Mitch, respectively.

Procedure

For each participant, intervention was matched to the functions of problem behav- ior as identified via the analogue functional analysis and SDA. For Drew and Lyle, both assessments suggested similar hypotheses, so only one intervention was implemented. The assessments suggested contrasting hy- potheses for Mitch; however, an intervention based on access to tangible items was not evaluated in the classroom because his teach- er reported that problem behavior occurred only when Mitch was working on tasks at

150 CYNTHIA M. ANDERSON and ETHAN S. LONG

his desk, not when tangible items were re- moved. Thus, it was not possible to fully evaluate the intervention suggested by the analogue functional analysis (intervention based on responding maintained by access to preferred tangible items). However, when follow-up observations were conducted at summer camp approximately 21 months af- ter the intervention was completed, Mitch’s camp director indicated that Mitch had dif- ficulty sharing his trains with others. As a result, follow-up data were collected during two situations: a task situation, during which Mitch worked on word recognition tasks, and a social skills situation, which fo- cused on sharing toys with peers.

All data for each participant were collect- ed at times problem behavior was reported to occur (during therapy sessions for Drew, throughout the day for Lyle, and during ta- ble work times for Mitch). An ABAB design was used to evaluate treatment effects with both Drew and Lyle. An ABC design was used with Mitch because, although his teacher had agreed to a brief return to base- line prior to implementing intervention, she was unwilling to do so once the intervention had actually been implemented.

Drew. Data were collected when Drew was working on match-to-sample and recep- tive language tasks, which were identified by the speech therapist as tasks that evoked problem behavior most often. During base- line, the therapist was instructed to respond as she normally did when problem behaviors occurred. If the therapist ceased to deliver instructional prompts for 2 consecutive min- utes in the absence of problem behavior, she was asked to reinitiate prompting. Interven- tion consisted of escape extinction combined with praise and brief (10 s to 30 s) breaks contingent on compliance. Drew’s speech therapist also was taught to use a sequential prompting procedure, consisting of verbal, gestural, and physical prompts. If Drew emitted a problem behavior, the therapist

immediately physically guided him to com- plete the task.

Lyle. Sessions were conducted throughout the academic day, but occurred most often when Lyle was not working on academic tasks (e.g., play time). During baseline, Lyle’s teacher was asked to respond to problem be- havior as she normally did; however, if she interacted with Lyle in a noninstructional manner (i.e., did not prompt him to com- plete tasks) for 2 consecutive minutes in the absence of problem behavior, she was asked to stop. Intervention consisted of differential reinforcement of other behavior. The teacher was taught to periodically praise Lyle when he was not emitting problem behavior and to ignore instances of problem behavior. Al- though an attempt was made to have the teacher set a timer to indicate when to de- liver praise (based on mean rate of respond- ing in baseline), she indicated that this would be too difficult. Thus, she simply de- livered praise periodically throughout the day during times that Lyle was not engaging in problem behavior. On average, the teach- er delivered verbal praise once every 2 min.

Mitch. Sessions were conducted when Mitch was expected to be working at his desk. During baseline, Mitch’s aide was asked to respond to problem behavior as she typically did. Intervention consisted of es- cape extinction and differential reinforce- ment of appropriate behavior. The aide stood behind Mitch, positioning herself at least 1 m from him. If Mitch began to turn his head away from the teacher or the ma- terials in front of him, the aide moved for- ward and prompted him back to work using a sequential three-step verbal prompting procedure. When Mitch began to work in- dependently, the aide moved away. In addi- tion, Mitch was provided periodic 20-s breaks for working appropriately, during which he was allowed to walk away from his desk with his aide. The aide typically hugged Mitch and engaged him in a brief activity

151STRUCTURED DESCRIPTIVE ASSESSMENT

(e.g., playing ‘‘rock, scissors, paper’’) during breaks. Initially, breaks occurred once every 2 min, but breaks were faded over time until they were occurring once every 5 min.

Follow-up data were obtained with Mitch at a summer program approximately 21 months after the intervention was complet- ed. Data were collected during two situa- tions: a task situation, during which Mitch worked on word recognition tasks, and a so- cial skills situation, which focused on shar- ing toys with peers. During the latter situa- tion, a toy train was used and Mitch was expected to allow a peer to play with the train when the peer requested to do so. Mitch could have the train back if he stated, ‘‘train please.’’ Mitch’s camp director indi- cated that initially this situation evoked high rates of problem behavior (providing anec- dotal support for the hypothesis obtained in the analogue functional analysis), but that a combination of escape extinction for prob- lem behavior combined with differential re- inforcement for manding had resulted in substantial decreases in problem behavior. Thus, during follow-up, treatments based on both assessments were used.

RESULTS

Results obtained for each participant are shown in Figure 5. In baseline, Drew emit- ted an average of 3.6 problem behaviors per minute. Intervention resulted in significant decreases in problem behavior, and a reversal to baseline conditions was implemented to assess functional control. During the return to baseline, Drew’s therapist was instructed to no longer implement physical guidance upon occurrence of problem behavior, but instead to respond as in baseline. Following reimplementation of escape extinction, rates of problem behavior once again decreased, and a mean reduction of over 90% below baseline rates was achieved.

Lyle emitted an average of 7.8 problem behaviors per minute in baseline. Interven-

tion also resulted in significant reductions in problem behavior, and a reversal was con- ducted to demonstrate functional control. During the final phase of intervention, Lyle averaged 0.8 responses per minute, slightly greater than a 90% reduction over baseline.

Intervention for Mitch also resulted in significant reductions in problem behavior. In fact, Mitch did not emit any instances of problem behavior during the last eight ses- sions conducted prior to follow-up. During follow-up, he continued to emit low rates of problem behavior during work periods. In addition, he used a mand to request trains, averaging 0.7 mands per minute during fol- low-up sessions.

GENERAL DISCUSSION

For 3 of 4 participants, results of the an- alogue functional analysis and SDA led to similar hypotheses about the function of problem behavior. Findings from the SDA also were helpful for developing effective in- terventions for 2 of the 3 children who par- ticipated in Study 2. Results for Mitch, how- ever, suggested that both assessments were necessary to identify all variables that were functionally related to problem behavior and to develop an intervention that would be ef- fective in all relevant contexts. Thus, the findings of this study suggest that the SDA may be useful as a substitute for or an ad- junct to the analogue functional analysis.

For Drew and Jane, evaluation of overall levels of problem behavior across conditions of the SDA and the analogue functional analysis resulted in similar hypotheses about the function of problem behavior. Exami- nation of overall rates in the SDA conducted with Lyle would have led to identification of an irrelevant function (access to tangible items). It was necessary to analyze condi- tional probabilities to identify the function of his problem behavior. A comparison of results obtained with Mitch from the two

152 CYNTHIA M. ANDERSON and ETHAN S. LONG

Figure 5. Mean rate of problem behavior during the intervention for Drew (top panel), Lyle (middle panel), and Mitch (bottom panel). On Mitch’s graph, closed triangles represent aggression per minute in work settings, open squares represent aggression per minute in the train-sharing condition, and open circles depict mands emitted per minute.

153STRUCTURED DESCRIPTIVE ASSESSMENT

assessments suggested that the SDA might have identified a functional relation that was not identified in the analogue functional analysis. Specifically, the SDA suggested that tasks often evoked problem behavior and that, in the presence of tasks, problem be- havior was maintained by escape from or avoidance of tasks. Conversely, results also indicated that the analogue functional anal- ysis identified a functional relation that was not identified in the SDA—that problem behavior might have been maintained by ac- cess to tangible items.

Anecdotal information provides some possible reasons for the discrepancies be- tween the findings of the analogue function- al analysis and the SDA for Mitch. First, the prompting procedure used in the classroom (repeated verbal prompts issued quite rap- idly) was markedly different than that used in the analogue functional analysis. It is pos- sible that an escape function would have been identified if a prompting strategy more similar to that used in the classroom had been used in the analogue functional analy- sis. With regard to the tangible function identified in the SDA, upon removal of the tangible item during the tangible condition of the SDA, Mitch’s teacher immediately en- gaged him in another activity, so as to dis- tract him from the loss of the item. It is possible that tangible removal did, at one time, evoke problem behavior (supporting the results of the analogue functional anal- ysis), but that she had developed an effective antecedent intervention with which to de- crease the likelihood of problem behavior.

Results of Study 2 suggested that the SDA may be useful when the goal of assessment is to design effective treatments. One poten- tial advantage of the SDA is that it is not necessary to remove the individual from his or her natural environment or disrupt his or her routine. Thus, the utility of the SDA may prove beneficial for practitioners who are not able to conduct analogue functional

analysis. The SDA might also be useful in providing information about reinforcement contingencies in the natural environment (e.g., schedules of reinforcement, stimuli that commonly follow problem behavior). For example, with the exception of Jane, tan- gible items were never delivered contingent on problem behavior. Thompson and Iwata (2001) evaluated naturally occurring re- sponse–consequence relations and also found infrequent presentation of tangible items following problem behavior. Future re- search should examine the extent to which procedures in the tangible condition of the analogue functional analysis replicate natu- rally occurring events.

Further refinements and evaluation of the SDA are needed to examine the extent to which the SDA can augment or be used as an adjunct to analogue functional analysis. First, research should evaluate the external validity of the SDA and analogue functional analysis. One strategy for doing so would be to conduct the SDA and the analogue func- tional analysis and then implement interven- tion evaluations using baselines matched to hypothesized functions of problem behavior. Research also is needed to evaluate the ex- tent to which the SDA might be used to develop relevant hypotheses without deter- mining conditional probabilities. A reason- able next step would be to conduct the SDA with more participants and to evaluate the results obtained from examining overall re- sponse patterns compared to conditional probabilities, and then conducting treatment evaluations to evaluate the utility of discrep- ant hypotheses.

REFERENCES

Blakeman, R., & Gottman, J. M. (1997). Observing interaction: An introduction to sequential analyses (2nd ed.). New York: Cambridge University Press.

Fisher, W. W., Adelinis, J. D., Thompson, R. H., Worsdell, A. S., & Zarcone, J. R. (1998). Func- tional analysis and treatment of destructive behav-

154 CYNTHIA M. ANDERSON and ETHAN S. LONG

ior maintained by termination of ‘‘don’t’’ (and symmetrical ‘‘do’’) requests. Journal of Applied Be- havior Analysis, 31, 339–356.

Freeman, K. A., Anderson, C. M., & Scotti, J. R. (2000). A structured descriptive methodology: In- creasing agreement between descriptive and ex- perimental analyses. Education and Training in Mental Retardation and Developmental Disabilities, 35, 55–66.

Iwata, B. A., Dorsey, M. F., Slifer, K. J., Bauman, K. E., & Richman, G. S. (1994). Toward a func- tional analysis of self-injury. Journal of Applied Be- havior Analysis, 27, 197–209. (Reprinted from Analysis and Intervention in Developmental Dis- abilities, 2, 3–20, 1982)

Lerman, D. C., & Iwata, B. A. (1993). Descriptive and experimental analysis of variables maintaining self-injurious behavior. Journal of Applied Behavior Analysis, 26, 293–319.

Mace, F. C., & Lalli, J. S. (1991). Linking descriptive and experimental analyses in the treatment of bi-

zarre speech. Journal of Applied Behavior Analysis, 24, 553–562.

Mace, F. C., Lalli, J. S., & Shea, M. C. (1992). Func- tional analysis and treatment in self-injury. In J. Luiselli, J. Matson, & N. Singh (Eds.), Assessment, analysis, and treatment of self-injury (pp. 122– 152). New York: Springer-Verlag.

Mueller, M. M., Sterling-Turner, H. E., & Scattone, D. (2001). Functional assessment of hand flap- ping in a general education classroom. Journal of Applied Behavior Analysis, 34, 233–236.

Sulzer-Azaroff, B., & Mayer, G. R. (1977). Applying behavior-analysis procedures with children and youth. New York: Holt, Rinehart, & Winston.

Thompson, R. H., & Iwata, B. A. (2001). A descrip- tive analysis of social consequences following problem behavior. Journal of Applied Behavior Analysis, 2, 169–178.

Received March 26, 2001 Final acceptance February 6, 2002 Action Editor, Dorothea Lerman

STUDY QUESTIONS

1. What limitations of descriptive analysis methodology were noted by the authors, and how might the structured descriptive analysis (SDA) address these limitations?

2. Describe some of the similarities and differences between the functional analysis and the SDA.

3. Briefly describe the SDA conditions and the purpose of each condition.

4. Why did the authors calculate conditional probabilities? What was the difference between the two types of calculations?

5. Summarize the results obtained from the functional analysis and the SDA.

6. How did the authors explain discrepancies between results of the functional analysis and the SDA for Lyle and Mitch?

7. Briefly describe the treatments evaluated for each participant.

8. What appears to be the major advantage and disadvantage of the SDA relative to the func- tional analysis?

Questions prepared by Claudia Dozier and Jessica Thomason, The University of Florida