Week 11 Article
JOURNAL OF APPLIED BEHAVIOR ANALYSIS
BEHAVIORAL MOMENTUMJN THE TREATMENT OF NONCOMPLIANCE
F. CHARLES MACE RUTGERS UNIVERSITY
MICHAEL L. HOCK UNIVERSITY OF VERMONT
JOSEPH S. LALLI, BARBARA J. WEST, P-Huup BELFIORE, AND ELIzABETH PINTER LEHIGH UNIVERSrrY
D. KIRBY BROWN LANCASTER-LEBANON INTERMEDIATE UNIT
Behavioral momentum refers to the tendency for behavior to persist following a change in envi- ronmental conditions. The greater the rate of reinforcement, the greater the behavioral momentum. The intervention for noncompliance consisted of issuing a sequence of commands with which the subject was very likely to comply (i.e., high-probability commands) immediately prior to issuing a low-probability command. In each of five experiments, the high-probability command sequence resulted in a "momentum" of compliant responding that persisted when a low-probability request was issued. Results showed the antecedent high-probability command sequence increased compliance and decreased compliance latency and task duration. "Momentum-like" effects were shown to be distinct from experimenter attention and to depend on the contiguity between the high-probability command sequence and the low-probability command. DESCRIPTORS: behavioral momentum, compliance latency, excessive task duration, noncom-
pliance, high-probability command sequence
Noncompliance is one of the most commonly reported behavior problems in developmentally dis- abled populations (Schoen, 1983). In addition to its prevalence, treatment of noncompliance is im- portant because of its covariation with other aber- rant and adaptive behaviors. For example, several studies have demonstrated that increased compli- ance often results in collateral reductions in aggres- sion, disruption, self-injury, and tantrums (e.g., Cataldo, Ward, Russo, Riordan, & Bennett, 1986; Parrish, Cataldo, Kolko, Neef, & Egel, 1986; Rus-
This research was funded in part by a grant from the Pennsylvania Office of Mental Health/Mental Retardation. The authors gratefully acknowledge the support of Richard J. Smith, Timothy Boyer, and Al Deibler of Lehigh County MH/MR that made this work possible.
Experiment 1 and the concepts of applied behavioral mo- mentum and the high-probability command sequence were presented at the 12th annual ABA convention (Hock & Mace, 1986). Experiments 1-5 were presented in a symposium at the 13th annual ABA convention (Mace, 1987).
Reprint requests may be addressed to F. Charles Mace, Graduate School of Applied and Professional Psychology, Rutgers University, Box 819, Piscataway, NewJersey 08854.
so, Cataldo, & Cushing, 1981). Conversely, re- duced noncompliance has been associated with in- creased appropriate behavior (Baer, Rowbury, & Baer, 1973). Thus, intervention to increase com- pliance appears to be an efficient means of im- proving a range of socially important behaviors. A variation of noncompliance is slowness to re-
spond to instructions or complete assigned tasks. Individuals who are excessively slow at completing tasks may receive less reinforcement (e.g., income from vocational tasks) and may incur punitive social responses from peers or staff.
Considerable research has evaluated procedures for increasing compliance and, to some extent, for reducing excessive compliance latency and task du- ration. However, much of this research has been conducted with children (Breiner & Beck, 1984; Fjellstedt & Sulzer-Azaroff, 1973; Forehand & McMahon, 1981). Procedures commonly used to increase compliance indude time-out (e.g., Parrish et al., 1986) and guided compliance (e.g., Neef, Shafer, Egel, Cataldo, & Parrish, 1983). However,
123
1988,211,123-141 NUMBER2 (summER 1988)
F. CHARLES MACE et al.
a potential liability of these procedures is that they often require physical contact with a client to achieve treatment integrity, which for large, uncooperative, or aggressive clients may be ill-advised. Alterna- tively, the effectiveness of differential reinforcement of compliant behavior depends on reinforcement for compliant responses being rich relative to re- inforcement produced by noncompliant or daw- dling behavior (cf. Ayllon, Garber, & Pisor, 1976; Cuvo, 1976; Holt, 1971). Unless a more powerful reinforcer or richer schedule can be applied to com- pliant behavior compared to the reinforcer and schedule maintaining noncompliance, differential reinforcement may not have the desired effect and punishment-based alternatives may need to be con- sidered (Myerson & Hale, 1984).
Alternative approaches to increasing compliance with developmentally disabled adults may be de- rived from consideration of advances in basic op- erant research (e.g., Deitz, 1978; Hayes, Rincover, & Solnick, 1980; Michael, 1980; Pierce & Epling, 1980). For example, Nevin has discussed the re- lationship between response strength and rate of reinforcement (see Nevin, 1974, 1979, for re- views). Behavior maintained at steady states by interval or ratio schedules of reinforcement has been shown to persist over time following a change in reinforcement conditions (de Villiers, 1977; Nevin, 1979; Zeiler, 1977). This resistance to change in the face of altered contingencies has been referred to as "response strength" (Herrnstein, 1970; Nev- in, 1979). Response strength may be relatively low when response patterns change readily or relatively high when response rates are slow to change under modified conditions. In general, behavior controlled by a multiple schedule will be more resistant to change during the schedule component that has a comparatively higher rate of reinforcement. That is, a relatively higher rate of reinforcement will result in relatively greater resistance to change or greater response strength.
Nevin, Mandell, and Atak (1983) suggested a parallel between a behavior's resistance to change and the momentum of objects in motion as de- scribed by Newton's first law of motion. They argued that it may be worthwhile to consider be-
havior at possessing the property of momentum. Accordingly, behavioral momentum can be ana- lyzed in terms analogous to the product of mass and velocity in classical physics (Nevin et al., 1983, p. 49). Behavioral mass was considered formally analogous to response strength and behavioral ve- locity as corresponding to response rate. Nevin et al. demonstrated that behavior controlled by a two- component multiple schedule procedure was more resistant to change in the component with a rela- tively higher rate of reinforcement when reinforce- ment was provided noncontingently, or when all reinforcement was discontinued. Thus, factors that influence rate of reinforcement may be expected to affect a behavior's resistance to change.
Consideration of Nevin et al.'s (1983) work on behavioral momentum prompted us to develop a novel intervention for noncompliance and excessive compliance latency and task duration. This pro- cedure, referred to as the high-probability com- mand sequence, indirectly manipulates rate of re- inforcement to establish what appears to be a "momentum" ofcompliant behavior that may per- sist when subjects are asked to perform a task with a low probability of compliance. Our objectives in the following series of experiments were (a) to eval- uate the effectiveness of the high-probability com- mand sequence in increasing compliance to "do" and "don't" commands (Neef et al., 1983) (Ex- periment 1), (b) to conduct preliminary investi- gations regarding the appropriateness of the be- havioral momentum analogy (Experiments 2 and 3), and (c) to evaluate the generality of the pro- cedure to reduce excessive compliance latency and task duration (Experiments 4 and 5).
EXPERIMENT 1
METHOD Subject and Setting
Bart, a 36-year-old man with severe mental re- tardation (IQ = 42), served as the subject in this experiment. Bart had resided in large, state-oper- ated institutions for most of his life and had a long history of noncompliance and aggression. Bart's
124
BEHAVIORAL MOMENTUM
large physical stature (height 6'1", weight 200 lb) contributed to the severity of his noncompliance and aggression. In his first community placement, these behaviors eventually resulted in his recom-
mitment to a private institution.
At the time of the present experiment, Bart had lived in a university-affiliated group home for ap-
proximately 18 months. The program was behav- ior-analytic in nature and was operated by univer- sity graduate students and faculty. Typical staffing patterns consisted oftwo graduate students working with six adults with moderate to severe mental retardation. After 6 months in this program, Bart became increasingly noncompliant and aggressive. A structured self-management program consisting of positive reinforcement for completion of house jobs and personal hygiene, without aggressive in- cidents, was effective only for periods of 2 to 3 months.
Sessions were conducted in the living room (5 m by 4 m), family room (3.5 m by 3 m), and kitchen (5 m by 4 m) of the home. An experi- menter, one or two data collectors, and zero to two
other clients were present during these sessions. Interactions between staff and other dients were
minimal; client-clent interactions were unrestrict-
ed. Because of the applied nature of the research, the subject was allowed free movement in these rooms to assess experimental effects under natural conditions.
Response Definitions, Measurement, and Interobserver Agreement
The principal dependent measure was the per-
centage of compliance to low-probability (low-p) "do" and "don't" commands. In Experiment 1, low-p commands were instructions or requests is- sued by the experimenter to the subject with which, in the experimenter's experience, the subject was
unlikely to comply. (In the remaining four exper-
iments, the probability value of both low-p and high-p commands was empirically determined.) Examples of low-p "do" and "don't" commands are "Bart, please put your lunch box away" and "Bart, please don't leave your lunch box on the table." Commands called for performance of sim-
ple tasks that could be completed within 30 to 60 s (i.e., "do" commands) or discontinuation of an undesirable behavior or condition (i.e., "don't" commands). Command compliance was defined as the subject initiating the response called for by the command within 10 s of the stated command and eventually completing the requested response(s).
The independent variable in this experiment was a sequence of high-probability (high-p) commands that was issued prior to a low-p command. High-p commands were instructions or requests with which the subject had a history of complying. These com- mands were always stated as a "do" request and are exemplified by the following: "Give me five, Bart," "Come here and give me a hug," and "Show me your pipe (or wallet, notebook, etc.), Bart." The mean percentage compliance to high-p com- mands during the entire experiment was 98%. Two trained observers recorded (a) experimenter
commands or requests directed to the subject for low-p and high-p behaviors, (b) compliance to "do" and "don't" low-p commands, and (c) compliance to high-p commands. A count of all responses was made during continuous 10-s intervals. A per- centage compliance measure was derived for each session by dividing the number of compliant re- sponses (of a given dass) by the number of exper- imenter requests for responses (of the same dass) and multiplying by 100. Observers stood within 2 to 5 m of the experimenter and subject but did not speak or make eye contact with the subject.
The second observer independently collected in- terobserver agreement data from a position no closer than 2.5 m from the primary observer during an average of 66% of the sessions across all phases and conditions of the experiment. For the first three experiments, total, occurrence, and nonoccurrence agreement were calculated on a point-by-point basis within all intervals per session (Page & Iwata, 1986). Table 1 presents the mean and range of interob- server agreement values for the dependent and in- dependent variables for all experiments.
Procedures Baseline. During each baseline session, the ex-
perimenter stood or sat within 1 to 2 m of the
F. CHARLES MACE et al.
Table 1 Interobserver Agreement: Mean and Range Percentages for Total Agreement (TA), Occurrence Agreement (OA),
Nonoccurrence Agreement (NOA), and Agreement (A) within ±1 s across the Dependent and Independent Variables of Experiments 1 through 5
Experiment 1 Experiment 2
TA OA NOA TA
Dependent variables Compliance with "do" commands 99 99 94 99.5
(93-100) (93-100) (63-100) (95-100) Compliance with "don't" commands 97 79 96
(92-100) (53-100) (91-100) Compliance with "do" commands 99.1
(during attention control) (97-100) Latency to initiate task Minutes to complete task
Independent variables Compliance with high-p commands 93 85 89 94.7
(83-98) (71-94) (77-97) (75-100) Occurrence of high-p 93 85 89 96.4
(83-98) (71-94) (77-97) (87-100) Occurrence of attention - 98.4
(97-100) Occurrence of 5-s IPT
Occurrence of 20-s IPT
Occurrence of prompts Occurrence of contingency statement Delivery of reinforcement
subject. The primary data collector prompted the experimenter to issue a command to the subject on a fixed-time (FI) 1-min schedule. The experi- menter made eye contact with the subject and issued a low-p command or request to Bart in a pleasant tone of voice. Low-p commands were selected at random from a pool of 20 low-p commands or, in the case of many low-p "don't" commands, were chosen on the basis of the subject's behavior (e.g., "Bart, don't put your feet on the coffee table"). If the subject satisfied the definition of command compliance, the experimenter provided immediate descriptive praise (e.g., "That's good Bart, thanks for putting your lunch box away"). Descriptive praise was used as a consequence for compliance for subjects in all five experiments because, in the experimenters' experience, praise appeared to be an effective reinforcer for these individuals. "Do" and "don't" command sessions differed only in the dass
of commands issued to the subject (i.e., either all "do" or all "don't" low-p commands).
Psychotropic intervention-Haldol. On Day 7 of the experiment, Bart's psychiatrist prescribed 10 mg of Haldol b.i.d. to control his aggressive be- havior. This represented a return-to-Haldol inter- vention, which Bart had experienced during the past 7 years, after a 6-week period of medication withdrawal. Baseline procedures remained in effect. Psychotropic intervention continued during all sub- sequent phases of the experiment.
High-probability command sequence. This condition was identical to the baseline procedures except that each low-p command was preceded by a sequence of high-probability (high-p) commands. The high-p command sequence consisted of the experimenter issuing a series of three or four high-p commands or requests to the subject immediately preceding presentation of the low-p command.
126
BEHAVIORAL MOMENTUM
Table 1 (Continued)
127
Experiment 2 Experiment 3 Experiment 4 Experiment 5 OA NOA TA OA NOA TA A ± I s TA A ± I s
96.7 99.6 99 98.7 97 - (69-100) (96-100) (92-100) (92-100) (80-100) - - 99 97 97 -
(93-100) (88-100) (82-100) 90.9 99.4 - - 100
(71-100) (97-100) 100
-_--- - - - - 100
95 97.1 95 95 93 100 - 100 (77-100) (86-100) (90-100) (88-100) (79-100)
95.4 97.1 96 96 94 100 100 - (82-100) (92-100) (91-100) (92-100) (87-100)
94 98.6 100 - (88-100) (95-100)
- - 99 96 99 -
(98-100) (88-100) (96-100) 99 95 99 - -
(98-100) (82-100) (97-100) - - - -- - - 100 -
-_--- -- - 100 100
High-p commands were issued at 10-s intervals (i.e., the interval between completion of a high-p task and the next high-p command).
Experimental Design The experimental conditions described above were
presented to the subject during two 1 5-min sessions daily that were separated by a 15- to 30-min free time period. Because "do" and "don't" commands have been shown to be members of different stim- ulus dasses (Neef et al., 1983), sessions with either all "do" commands or all "don't" commands were alternated in a multielement design (Sidman, 1960). The order in which "do" and "don't" command sessions were conducted was determined randomly each day. In addition, the independent variable was alternately applied and withdrawn during "do" and "don't" command sessions in the context of a re- versal design (Sidman, 1960).
REsuLsT Figure one represents Bart's percentage of com-
pliance to low-p commands during "do" and "don't" command sessions across all phases of the experiment. During baseline, Bart's compliance to low-p requests during "do" sessions averaged 47% and during "don't" sessions 53.5%. With the ad- dition of psychotropic medication, mean compli- ance to "do" commands was 68% versus 53.5% for "don't" commands.
During Phase 3, application of the high-p com- mand sequence prior to each "don't" command resulted in an increase in mean compliance to 87.5%. Compliance to "do" commands, which remained under baseline conditions, averaged only 61%. In Phase 4, the pattern of compliance reversed with "do" command sessions increasing to a mean of 90.5% following application of the high-p com-
128 F. CHARLES MACE et al.
High Probability Command Sequence Preceding
A Low Probability Command
A
00) 100
z < 80 -lu a.
>- 60 c -
u K5 w m 40
w 0 so 31 0'"Jo CO _j
Baseline
"Don't"
"Do"
5
Psychotropic Applied to Applied to Applied to Applied to
Intervention-Holdol "Don't" Commonds 'Do" Commonds 'Don't"Commands 'Do" a"Don't" Commands
L~~~~ WI
>I KY 10 15 20 25 30 35
SESSIONS
Figure 1. Bart's percentage compliance to low-probability "do" and "don't" commands under baseline and psychotropic intervention conditions, and alternate application and withdrawal of the high-probability command sequence.
mand sequence. "Don't" command compliance re-
turned to low levels during this period (M = 44%). In the fifth phase, compliance to "don't" com-
mands returned to high levels when low-p com-
mands were preceded by a series of compliant re-
sponses (M = 91%). Compliance to "do" commands, which were not preceded by the high-p command sequence, averaged only 56%. In the final experimental phase, use of the high-p com-
mand sequence resulted in high levels ofcompliance to both "do" and "don't" commands. Mean com-
pliance ranged from 87% to 97% for "do" com-
mand sessions (M = 93%) and 85% to 97% for "don't" command sessions (M = 90%).
DISCUSSION
This experiment demonstrated the effectiveness of preceding a low-probability command with a
sequence of high-probability commands in the treatment of noncompliance. Establishing a pattern
ofcompliant responding by the subject immediately prior to the issuance of a low-p request resulted in increases in the subject's compliance. Our objective in the second experiment was to assess the subject generality of the high-p procedure and to examine possible effects of positive attention alone on com-
pliance.
EXPERIMENT 2
MEHOD
Subject and Setting The subject of the second experiment was Ned,
a 44-year-old severely retarded (IQ = 21) male with Down Syndrome. Ned had lived in institu- tions for most of his life. When asked to perform a task, Ned typically shook his head "no" and looked away. Occasionally, he would throw items, curse, spit, hit others, or lie on the floor when such commands were issued. The setting for the study was the same house as
in Experiment 1. During the first four phases of the experiment, baseline and treatment high-p com- mand sessions were conducted in the kitchen. At- tention control sessions were held in the subject's second floor bedroom (4 m by 3.5 m). Persons present and their interactions during the sessions were similar to those in the first study.
Response Definitions, Measurement, and Interobserver Agreement
Ned primarily did not comply with "do" re- quests. The procedure for identifying commands to which the subject had a low probability of com- plying consisted of the experimenter approaching
I II
BEHAVIORAL MOMENTUM
Ned and asking him to perform each of 25 tasks on separate occasions. Ten separate trials were con-
ducted for each of the 25 tasks; those commands that were complied with four or fewer times in 10
trials were designated as low-p commands. This procedure resulted in a pool of 15 low-p "do" commands that were used in the experiment. A similar procedure used with high-probability com- mands (i.e., at least 80% compliance) resulted in the following high-p command sequence: (a) "Ned, give me five," (b) "Give me a bump" (i.e., the experimenter and subject bumped hips in a dancing motion), and (c) "Ned, show me your radio."
The definition of command compliance and the data collection procedures for the primary and sec-
ondary observers were identical to those in Exper- iment 1 (see Table 1 for interobserver agreement values).
Procedures Baseline. The actions of the experimenter and
data collectors during this condition were virtually identical to those described for the baseline con-
dition for Experiment 1. High-probability command sequence. The pre-
sentation of the high-p command sequence and the experimenter's response to compliance were iden- tical to the procedures used during this condition in the first experiment.
Attention control. This condition was designed to provide experimenter attention prior to issuance
of a low-p command without providing specific discriminative stimuli for behaviors presumably maintained by high rates of reinforcement. On an
FT 1-min schedule, the experimenter sat or stood within 1 to 2 m of the subject and directed a
sequence of three or four neutral or positive com-
ments to the subject. Comments were randomly selected from a pool of 2 5 statements. The interval between comments ranged from 10 to 15 s. Ex- amples of these comments included "Ned, that's a nice shirt you're wearing," "We're going bowling this afternoon," and "I'm going to visit my parents
this weekend." Within 10 s of the last comment, the experimenter issued a randomly selected low-p command to the subject. Compliance to low-p com-
mands resulted in descriptive praise on a continuous reinforcement (CRF) schedule. Experimental Design
The experimental procedures were presented dai- ly during two 1 5-min sessions separated by a 15- to 30-min free time period. An ongoing attention control condition was alternated with either the baseline or the high-p command sequence condition in a multielement design. The order in which con- ditions were conducted was determined randomly each day. The effects of the high-p command se- quence were evaluated with an A-B-A-B reversal design. In the final phase of the experiment, the settings in which the high-p command sequence condition and the attention control condition were conducted were reversed to control for possible ef- fects of setting-specific commands.
REsuLTs Figure 2 depicts Ned's compliance to low-p
commands during all baseline, high-p command sequence, and attention control conditions. In the initial baseline phase, issuing low-p commands without a preceding high-p command sequence re- sulted in a mean compliance of 26%. When ex- perimenter attention preceded each low-p com- mand, compliant behavior was similar to baseline (M = 35%). During Phase 2, application of the high-p command sequence effected an increase in mean compliant responses to 73%. Compliance during the attention control sessions remained es- sentially unchanged from the previous phase (M = 38%). A return to baseline condition in the third phase
produced an immediate decrease in the subject's percentage compliance (M = 39%). Comparable levels of compliance (M = 43%) continued during the subsequent attention control condition. In the fourth phase, high levels of compliance occurred when the high-p command sequence was reinstated (M = 84%). Average percentage compliance in- creased slightly during the ongoing attention control condition (M = 51%). Finally, the setting reversal had little effect on the subject's pattern of compli- ance during the high-p command sequence (M =
79%) and attention control conditions (M = 47%).
F. CHARLES MACE et al.
No High- P Command Sequence Preceding A
o CX Low - P Command I .equence Qequec I ze IiI
10 I Reversal
z G 800 I
IY3 0 ~ Attention Preceding A 0- 0-ij
C-40W/ o Low-P Command
Cr 0 w/o Attention li W 0 '
5 10 15 20 25 30
SESSIONS Figure 2. Ned's percentage compliance to low-probability commands during the attention control condition and alternate
application and withdrawal of the high-probability command sequence. In the final experimental phase, the settings in which the attention control and high-p sequence were conducted were reversed.
DISCUSSION Results of the second experiment support the
subject generality of the effects produced by the high-p command sequence inasmuch as the effects for Ned and Bart were similar. A second important finding was that experimenter attention was not itself sufficient to occasion compliance to low-p requests. That is, experimenter comments presented in the same manner as high-p commands failed to influence the probability of subject compliance. This finding suggests that presentation of discriminative stimuli for high-probability behaviors is critical to the momentum-like effects observed.
In the third experiment, we investigated another parameter of the high-p command procedure that may determine its effectiveness as an applied pro- cedure and further examines the value of the be- havioral momentum analogy. Nevin et al. (1983) found that resistance to change or behavioral mo- mentum was directly related to rate of reinforce- ment. The higher the relative rate of reinforcement, the greater the resistance to change. Therefore, it may be logical to predict that momentum-like ef- fects will decrease with an increase in the interval between the last high-p command in the sequence (or between any high-p commands in the sequence)
and the statement of the low-p command. Increas- ing this interval presumably has the effect of de- creasing rate ofreinforcement which, in turn, should decrease behavioral momentum.
EXPERIMENT 3
METHOD
Subject and Setting The subject and setting in which experimental
sessions were conducted were identical to those de- scribed in Experiment 1. Bart continued to take 10 mg of Haldol b.i.d. for the duration of the study. This experiment was conducted 1 month after completion of the first study.
Response Definitions, Measurement, and Interobserver Agreement
As in Experiment 1, the principal dependent measure in the third experiment was the percentage compliance to low-p "do" and "don't" commands. The procedure described in Experiment 2 to iden- tify low- and high-probability commands was used to define a pool of 15 low-p "do" commands, 10 low-p "don't" commands, and seven high-p "do"
130
BEHAVIORAL MOMENTUM
High Probability Command Sequence Preceding
A Low Probability Command A
20-s IPT a"Do" 20-s IPT"Don't 5-s IPT "Donti 5-s IPT "Do"
to"Don't"l
'~~ ~~I
5
20-s IPT "Do" 120-sIPT"Don't"20-s IPT "Do" | 5-s IPT 5-s IPT "Don't" 5-s IPT "Do" 5-s IPT" Don't Do a "Don't
f~~~~~~fS~Dot D Dn
I, .al, . .l I . , , , , . , , 10 15 20 25 30 35
SESSIONS Figure 3. Bart's percentage compliance to low-probability "do" and "don't" commands during 5-s and 20-s IPT
applications of the high-probability command sequence.
commands. Additional low-p "don't" commands were extemporaneously selected during "don't" command sessions corresponding to the subject's aberrant behavior.
Definitions and procedures used to measure com-
mand compliance for "do" and "don't" low-p commands and high-p commands were identical to
Experiment 1. The independent variable manipu- lated in this study was interprompt time (IPT). IPT was defined as the time interval beginning with the cessation of the last high-p command in the high-p command sequence and ending with the onset of the low-p command. Independent observ- ers measured this interval using a stopwatch. Time measurements within ±2 s were considered in agreement. Interobserver agreement measures were
taken for the dependent and independent variables on an average of 53% of the sessions during all phases and conditions of the study (see Table 1).
Procedures High-probability command sequence-20-s
IPT. All procedures in this condition were identical to those described for the high-p command se-
quence in Experiment 1 with one exception. After the last high-p command in the high-p command sequence was issued, the experimenter paused 20 s without speaking to the subject, and then stated
a randomly selected low-p "do" command or a
low-p "don't" command that corresponded to the subject's inappropriate behavior. The primary data collector timed the IPT interval and nonvocally cued the experimenter to deliver the low-p com-
mand. High-probability command sequence-5-s
IPT. This condition consisted of the same proce-
dures described for the high-p command sequence
in the first study except that the 5-s IPT interval was timed by the primary data collector.
Experimental Design The high-p command sequence preceded each
low-p command in all sessions of the experiment. "Do" command sessions and "don't" command sessions were alternated in a random order daily according to a multielement design. The effects of 5-s and 20-s IPTs were compared by alternately applying each IPT condition to "do" and "don't" command sessions across successive phases of the study in the context of a reversal design.
RESULTS Bart's compliance to low-p "do" and "don't"
commands under 5-s and 20-s IPT conditions is presented in Figure 3. During Phase 1, application of the high-p command sequence with a 5-s IPT
/
O en I a0 z
w C)z 2 _ 0
L >-0 >- 0 - o _j
W OZ a: <L X
C-
CL J
100
80
60
40
20
0
0
F. CHARLES MACE et al.
to "don't" commands resulted in consistently higher compliance (M = 83%) than the high-p sequence with a 20-s IPT applied to "do" commands (M = 53%). When IPT conditions were reversed in the second phase, mean compliance to "do" commands using a 5-s IPT increased to 89%, whereas com- pliance to "don't" requests dropped with the use of a 20-s IPT to an average of 27%. The reversal pattern continued during Phases 3 through 5. Mean percentage compliance with the 5-s IPT was 80%, 86%, and 78% for Phases 3, 4, and 5, respectively. By contrast, the high-p command sequence with a 20-s IPT resulted in low levels of compliance. Per- centage compliance averaged 22%, 29%, and 37% during the third through fifth phases of the ex- periment, respectively. In the final phase, compli- ance averaged 91% for "do" commands and 78% for "don't" commands when the high-p procedure was used with the 5-s IPT for both stimulus classes.
DISCUSSION Two important findings may be gleaned from
the third experiment. First, the momentum-like effects produced by the high-p command sequence appear to depend on the temporal contiguity be- tween the high-p command sequence and the low-p command. The relatively longer IPT interval failed to elevate compliance levels above those achieved by this subject during the baseline phase of Ex- periment 1. Thus, on the basis of this subject's data, it appears that practitioners must ensure that low-p commands are issued immediately after the high-p sequence. Extensions of the IPT interval appear to negate the controlling effect that high-p commands have on compliant behavior. Second, these findings may have been due to differences in the rate of reinforcement between the 5-s and 20-s IPT condition. Issuing three high-p commands at 10-s intervals followed by a low-p command at an IPT of 20 s results in a reinforcement rate that is approximately half the rate using a 5-s IPT. Thus, the results of Experiment 3 are predicted by the behavioral momentum analogy, if rate of reinforce- ment is analogous to behavioral mass as Nevin et al. (1983) have argued.
The fourth and fifth experiments examined the
application of the high-p command procedure to a problem related to noncompliance, excessive re- sponse latencies. These studies investigated whether the high-p command procedure could reduce sub- jects' latency to respond to experimenter commands or requests to perform tasks.
EXPERIMENT 4
METHOD
Subjects and Setting Two adult men with moderate mental retarda-
tion served as subjects. Tim was a 34-year-old male with Down Syndrome (IQ = 53) who lived with his parents until age 33. He performed most self- care skills and household tasks independently. However, the speed with which he responded to staffrequests was extremely slow. During the period following staff instructions, Tim would typically engage in various forms of stereotypy or stare into space and move very slowly toward initiation of the task.
The second subject, Mitch, was 45 years old (IQ = 47) and had lived most of his life in state- operated institutions. He had grand mal seizures that were controlled by 500 mg of Tegretol and 250 mg of Mysoline per day. His psychiatrist also prescribed 100 mg of Mellaril per day to control his "psychotic" behavior, which consisted of talking to himself or talking out of context. Mitch was skilled at most self-care and household tasks; how- ever, he was sometimes very slow to respond to staff requests or spent excessive periods of time performing tasks such as showering, making his bed, or preparing his lunch. When off-task, Mitch would typically stare into space or talk to himself.
Both subjects lived in the community group home described in Experiment 1. All sessions were con- ducted in the kitchen.
Response Definitions, Measurement, and Interobserver Agreement
The dependent measure for both subjects was compliance latency defined as the interval beginning with the completion of an experimenter's instruc-
132
BEHAVIORAL MOMENTUM
tion and ending with initiation of the specified task. Task initiation for Tim was defined as lifting his plate or glass from the dining table. For Mitch, task initiation entailed performing one of the fol- lowing depending on the task selected: (a) lifting the kitchen trash container, (b) lifting a broom, or (c) touching the mirror-cleaning materials. Com- pliance latency was measured in seconds by the experimenter using a stopwatch. A trained inde- pendent observer collected interobserver agreement measures for the dependent and independent vari- ables on an average of 52% and 40% of the sessions across conditions of the study for Tim and Mitch, respectively. All interobserver latency measures agreed to within ± 1 s (see Table 1).
Measures of the integrity of the independent variables were obtained for all sessions. Event rec- ords were collected for the following variables dur- ing their respective experimental sessions: (a) oc- currence of high-p commands, (b) compliance with high-p commands, and (c) occurrence of attention statements. The integrity measures indicated that the experimenter issued high-p or attention state- ments according to the procedures on 100% of the compliance trials. Compliance to high-p commands was 100% for both subjects. Interobserver agree- ment calculated on a trial-by-trial basis was 100% for all independent variables (see Table 1) (Page & Iwata, 1986).
Procedures Baseline: No high-probability command se-
quence. For Tim this condition was conducted im- mediately after he finished eating his breakfast, lunch, or dinner. Tim was seated along one side of an oblong dining table and the experimenter was seated across from him. Within 5 s of the subject placing his napkin on his plate indicating the end of the meal, the experimenter made eye contact with Tim and issued the following instruction: "Tim, please dear your place at the table." The experimenter remained seated and directed no other comments to Tim until he complied with the task request (i.e., rinsed his plate and glass and placed them in the dishwasher). Descriptive praise was provided immediately after Tim performed the task.
During baseline for Mitch, the experimenter took Mitch into the kitchen where all task materials were located and issued one of the following five ran- domly selected task commands: (a) "Mitch, please empty the trash," (b) ". . . sweep the downstairs (or upstairs) bathroom floor," or (c) "... clean the downstairs (or upstairs) bathroom mirror." Pro- cedures for descriptive praise were identical to those described for Tim.
High-probability command sequence. Proce- dures in this condition were identical to baseline, except preceding the statement ofeach task request, the experimenter delivered the following sequence of high-p commands in a manner identical to that described in Experiments 1 through 3: "Tim (or Mitch), shake my hand," "Tim (or Mitch), give me five,' and "Tim (or Mitch), give me a hug.' Within 10 s of the subject's compliance to the last high-p command in the sequence, the experimenter issued the task request described for each subject during baseline.
The probability value of each high-p command was determined empirically prior to the experiment. Ten separate trials for each of the high-p requests were conducted for both subjects. Trials were sep- arated by at least 15 min. Both subjects complied with all high-p requests 100% of the time during this preliminary assessment.
Attention control. Procedures in this condition were the same as those described in Experiment 2.
Experimental Design Experimental conditions were presented in the
context of a multielement design. During the first 9 days of Tim's study, the baseline, high-p com- mand sequence, and attention control conditions were administered one per day in a random order across days without balancing the number of times each condition was conducted. On Days 10 through 2 7 these conditions were administered in a random and balanced order.
Six sessions were conducted for Mitch each day, with two of each of the three types of tasks rep- resented (i.e., empty trash, sweep floor, clean mir- ror). Each day of the experiment a baseline and high-p command sequence condition were con-
133
F. CHARLES MACE et al.
TI
High-P Command Sequence
5 10 15 20 25
Sessions Figure 4. Latency in seconds to initiate task following staff instruction during baseline (no high-p) and high-p command
sequence condition (Tim and Mitch), and attention control condition (Tim). Different data symbols represent different tasks for Mitch.
ducted for each of the three task pairs. The order in which these six sessions were conducted was
determined randomly on a daily basis.
RESULTS
Figure 4 represents the subjects' latency to ini- tiate each task following a staff instruction during the different experimental conditions. During the
baseline or no high-p command sequence condition, Tim's compliance latency varied greatly from 12 s
to 848 s (M = 156 s). Experimenter comments
prior to the task command during the attention control condition produced results similar to base- line. Mean compliance latency was 117 s with a
range of 16 s to 416 s. By contrast, Tim consistently responded quickly to experimenter instructions that
134
200
150
100
50
0
to c 0 4-
0 4-
(a
c
0S
CP
I-
cs
0
0
L._
4- a 0
4-
._
0-
C0
c
4n
c
0 Cl) C
0
40- a -j
;
-
BEHAVIORAL MOMENTUM
were preceded by the high-p command sequence. His average latency to compliance was 17 s, with a narrow range of 11 to 25 s.
Similar results were obtained for Mitch. Without the preceding high-p command sequence, compli- ance latency was quite variable and often lengthy. During baseline, the average latency to compliance across all three task types was 151 s (range, 5 s to 377 s). Use ofthe high-p command sequence sharply reduced the subject's latency to comply. Mean com- pliance latency with the high-p procedure was 10 s across all task types. Momentum-like effects were also usually consistent across the three types of tasks used in the experiment. With the high-p command sequence, average latencies to initiate emptying trash, sweeping floors, and cleaning mirrors were 5 s, 17 s, and 8 s, respectively. Without the high-p pro- cedure, compliance latencies averaged 98 s, 160 s, and 194 s for the three tasks, respectively.
The fifth experiment extended the application of the high-p command sequence to reduce the time a subject spent performing an entire task. When applied to reduce task duration, the high-p command procedure was presented when off-task behavior or dawdling occurred in the course of performing the task. Because the high-p command sequence requires continual supervision of the task and is more complex to administer than simple prompts to resume task-related behavior, the ap- plied value of the high-p procedure for reducing excessive task duration depends on it being highly effective. For this reason, the fifth study compared the effectiveness of the high-p command sequence with simple prompts and a contingency manage- ment procedure.
EXPERIMENT 5
METHOD
Subject and Setting Mitch (of Experiment 4) served as the subject
in this study. The target behavior of interest was the excessive amount of time Mitch spent taking a shower.
The study was conducted in the group home's second floor bathroom (2.5 m by 3.5 m) and Mitch's bedroom (5 m by 5 m). The bathroom was equipped with a tub, shower head, and plastic shower curtain, and was located 6 m down the hallway from Mitch's bedroom. In general, only the experimenter and secondary observer were present during experimen- tal sessions.
Response Definitions, Measurement, and Interobserver Agreement
Showering sessions were divided into three task segments which Mitch, on average, spent com- parable amounts of time performing. The depen- dent measure was the time elapsed to complete each of the three task segments. Task Segment 1 was shower preparation and was defined as the period beginning with the experimenter's instruc- tion "Mitch, it's time to take your shower" and ending with the subject entering the bathroom wearing his bathrobe and slippers and carrying a towel and washcloth. During this segment the sub- ject undressed in his room, put his clothes away, dressed in his bathrobe and slippers, and obtained a towel and washdoth from his drawer. The second task segment was showering, which began with the end ofTask Segment 1 and ended when the subject turned off the shower water. During this period, Mitch undressed, washed most body parts, and shampooed his hair. Task Segment 3 began with the end of Task Segment 2 and ended when the subject was dressed in his pajamas and slippers. This task had become very routine for Mitch, and no steps were omitted during any session of the experiment.
Task segment durations were measured by the experimenter using a stopwatch. During baseline, the experimenter assumed a position in the hallway that would permit observation of Mitch's bedroom and the bathroom. During intervention phases, timing took place in the room in which the subject was located. Interobserver agreement measures for the dependent and independent variables were col- lected in no less than 29% of the sessions across all conditions and phases of the study. All inter-
135
F. CHARLES MACE et al.
observer duration measures were within ± 1 s (see Table 1).
Event recording was used to measure the integ- rity of the independent variables during all sessions. The following variables were measured during the conditions in which they occurred: (a) occurrence of vocal prompts, (b) occurrence of contingency statement, (c) occurrence of high-p commands, (d) compliance with high-p commands, and (e) deliv- ery of contingent reinforcement. The integrity mea- sures indicated that (a), (b), (c), and (e) were ad- ministered the number of times described in the procedure section on 100% of the sessions. Per- centage compliance to high-p commands (d) was 96% for Mitch. Interobserver agreement computed on a trial-by-trial basis was 100% for all indepen- dent variables (see Table 1).
Procedures Baseline. Sessions were begun at approximately
8:00 p.m. each evening. The experimenter ap- proached Mitch, made eye contact, and provided the instruction to take a shower. No other instruc- tions or contingencies were announced. The exper- imenter followed the subject upstairs and continued timing task duration from the hallway position. When Task Segment 3 was completed and the subject exited his bedroom, the experimenter said "Mitch, I'm glad to see you finished your shower."
Contingency management. Procedures in this condition were identical to baseline with the fol- lowing exceptions. The experimenter (and second- ary observer) stood approximately 1 to 3 m from the subject. Contingent on the first occurrence of off-task behavior the experimenter showed Mitch two cupcakes, two quarters, and one of his favorite books and said "Mitch, if you finish (last step in the task segment) by the time the buzzer sounds you can have your choice when you're done with your shower." "Off-task" was defined as 15 con- tinuous seconds of any behavior that was unrelated to completion of the task. Examples included (a) repetitive motor movements such as removing or replacing his watch, wallet, comb, etc., (b) rear- ranging items on his dresser, (c) talking to himself or out of context without working on the task, and
(d) staring into space. After stating the contingency, the experimenter set a kitchen timer for 16 min, positioned the timer within Mitch's view, and left the room. The 16-min criterion was 2 min lower than the subject's lowest baseline data point. When the timer sounded, the experimenter entered the room, told the subject whether or not the reinforcer had been earned, and praised successful task com- pletion. On 85% of the sessions, the experimenter stated the contingency within 60 s of the onset of the task segment and between 60 s and 120 s during the remaining sessions.
Prompts. These procedures paralleled baseline except that the experimenter stood within 1 to 3 m of the subject and provided a combination vocal and gestural prompt to resume the task contingent on each occurrence of off-task behavior. The prompt was repeated every 15 s until the subject resumed on-task behavior. Descriptive praise was delivered for compliance with the prompts. An average of 4.7 prompts per session were required to sustain Mitch's involvement in the task (range, 1 to 13).
High-p command sequence. The procedures in this condition were the same as the prompt con- dition except that the high-p command sequence was applied instead of a prompt, contingent on each instance of off-task behavior. The high-p com- mands, timing of high-p commands, and descrip- tive praise were identical to Experiment 4. Dura- tions of each high-p command sequence were induded in the measures of task duration. The mean number of high-p command sequences ad- ministered per session was 1.8 and 1.5 during Phas- es 2 and 4, respectively.
Experimental Design Experimental conditions were administered in
the context of a four-phase multielement design. Baseline conditions were in effect during all task segments for the first and third phases of the ex- periment. Phase 2 randomly assigned the contin- gency management, vocal prompts, and the high-p command sequence conditions to Task Segments 1 through 3 for each day of the experiment. In the fourth phase, the high-p command sequence was applied during all three task segments per session.
136
BEHAVIORAL MOMENTUM
Comparison of High- P Command Sequence, Prompts, ed Cestingemey Msongement
It!
o~ . .1. ......... 1|
U)5401 0 2 W3 04
30 -
a Yolk SegomS
I
5 tO iS 20 25 30 38 40 45
Non - Consecutive Days
Figure 5. Minutes to complete each of three showering task segments during baseline, alternating treatments, and application of the most effective treatment during the entire task. Different data symbols correspond to different task segments.
REsuLTs Durations for each of the three task segments
during all experimental conditions are presented in Figure 5. In the first baseline phase, durations were similar although quite variable across the three task segments. The average time spent performing Task Segment 1 was 35 min. Mean durations for the second and third task segments were 31.8 min and 33.9 min, respectively.
All three interventions resulted in faster perfor- mance of task segments compared to baseline. The most effective procedure was the high-p command sequence, which reduced task durations to a mean duration of 10.3 min per task segment (range, 4.3 min to 15.2 min). Prompts were the next most effective, reducing task duration to approximately one half of baseline. With prompts, task segment durations averaged 16.7 min. Least effective of the three interventions was the contingency manage- ment procedure. Reinforcement of short task du- rations resulted in an average of 18.4 min per task segment. Mitch met the criteria for reinforcement on 57% of the sessions during this condition.
The return to baseline condition in the third phase of the study again resulted in longer task durations. However, unlike the first baseline phase, Mitch spent considerably more time performing Task Segment 1 than Segments 2 and 3. Average
duration for the first task segment was 40.7 min compared to 21.6 min and 28.7 min to complete Task Segments 2 and 3, respectively. Although performance in the second baseline was differen- tiated on the basis of task segment, the overall time required to complete the three task segments was similar for both baseline phases. Mean overall task duration was 100.7 min for Baseline 1 and 91 min for Baseline 2.
In the final phase of the study, the most effective intervention was applied during all task segments. Administration of the high-p command sequence during the entire task resulted in uniformly short task segment durations. The mean durations for Task Segments 1 through 3 were 9.7 min, 12.2 min, and 12.1 min, respectively. This resulted in an overall task duration mean of 33.9 min, which again was approximately one third of the baseline level.
GENERAL DISCUSSION
Concepts and findings from the basic behavior analysis literature stimulated the development of an innovative intervention for adult noncompliance. A nonhuman model of behavioral momentum (Nevin et al., 1983) was useful to predict how persons with severe developmental disabilities would respond to low-probability commands under dif-
137
F. CHARLES MACE et al.
ferent antecedent conditions. Presentation of a se- quence of high-probability commands immediately prior to issuance of a task request increased the probability of compliance for some subjects and reduced compliance latency and task duration for other subjects. The precision of our analogy with Nevin's behavioral momentum, as well as the fit between behavioral and physical momentum, may at some point prove to be less than perfect. How- ever, there may be applied and theoretical value in viewing behavioral momentum as a distinct phe- nomenon.
The applied value of the analogy lies in its in- spiration of innovative intervention procedures. The high-probability command sequence used in the present research seemed to establish a series of re- sponses with high behavioral mass. Commands that have a high probability of occasioning compliant responses are, we assume, discriminative stimuli for behavior that has produced reinforcement in the past. Although the exact reinforcers and their sched- ules were not analyzed in this research, the subjects quickly and reliably responded to the high-p re- quests and, anecdotally, seemed to enjoy doing so. Thus, it appears that by manipulating the type of command issued it is possible to reliably evoke behavior that effects reinforcement and, according- ly, establish a pattern of responding that has a relatively high behavioral mass. Interpreted from a behavioral momentum framework, increased com- pliance to low-p commands following the high-p sequence may illustrate resistance to change in the face of altered environmental conditions (i.e., when a low-p command is presented).
The results of Experiments 2, 3, and 4 offer some preliminary support for the appropriateness of the behavioral momentum analogy. First, Ex- periment 3 illustrated that when reinforcement rate was reduced by increasing the interval between the high-p sequence and low-p command, compliance to low-p commands decreased. This effect is pre- dicted by the behavioral momentum analogy be- cause decreases in reinforcement rate should pro- duce corresponding decreases in resistance to change or behavioral momentum. Second, in the attention control conditions ofExperiments 2 and 4, pleasant,
neutral statements delivered to subjects on the same schedule as the high-p commands failed to alter compliance to low-p requests. This suggests the important role of the high-p command, which pre- sumably serves as a discriminative stimulus for be- havior maintained by high rates of reinforcement. We should emphasize, however, that these analyses are preliminary. Further research should directly manipulate reinforcement rates and intervals be- tween high-p commands and compare reinforce- ment associated with neutral statements versus high-p commands.
Several dimensions of the present experiments differed from the basic work ofNevin et al. (1983). First, Nevin et al. directly manipulated subjects' access and rate of reinforcement. By contrast, we manipulated discriminative stimuli assumed to be correlated with reinforcement (i.e., high-p com- mands). Thus, without direct manipulation of re- inforcement rates we must be cautious in our con- dusion that the high-p procedure produced a relatively high behavioral mass. Second, if we can assume that reinforcement rates were manipulated indirectly with the high-p commands, the rein- forcement schedule for compliance to high-p com- mands approximated a CRF schedule. This differed from Nevin et al.'s work in which resistance to change was examined under different, and highly intermittent, variable-interval (VI) schedules of re- inforcement. Finally, Nevin et al. used a two-com- ponent, multiple-schedule procedure in which each component (i.e., reinforcement schedule) was cor- related with a different discriminative stimulus (i.e., a red or green response key). Subjects' rate of re- sponding was controlled by the discriminative stim- uli only via their associated reinforcement schedule. In the present experiments, the rate of compliant responding was controlled directly by the number of high-p and low-p commands issued.
Given the differences between Nevin et al.'s (1983) basic research and the present attempt to apply these concepts in the high-p command pro- cedure, alternative explanations for the results of the present research merit discussion. One plausible account may be stimulus generalization, which re- fers to the spread of the effects of reinforcement to
138
BEHAVIORAL MOMENTUM
stimulus conditions that have not been associated with reinforcement (Catania, 1984). Thus, stim- ulus generalization indicates a lack of stimulus con-
trol. When the subjects in the present experiments complied with low-p commands following the high-p command sequence, it could be said that compliance to high-p commands generalized to
low-p commands and that the stimulus control of high-p commands was weak. However, as Nevin (1974) noted, stimulus generalization appears to
be an instance of resistance to change rather than an alternative to it. During extinction, resistance to
change is greatest at the training stimulus and de- creases as the test stimulus departs from the training stimulus (Nevin, 1974, p. 406). Thus, the ante-
cedent presentation of high-p commands may
weaken the distinction between high- and low- probability commands, thereby increasing resis- tance to change and inducing stimulus generali- zation. Our results also bear some resemblance to the
effects reported in the generalized imitation litera- ture. Several studies have shown that, following imitation training, subjects made imitative re-
sponses to unreinforced models (e.g., Baer & Sher- man, 1964; Brigham & Sherman, 1968). Further, the probability of imitation to unreinforced models increased when unreinforced models were inter- spersed among models that were reinforced (Pe- terson, 1968) and decreased when discrimination between reinforced and unreinforced models was
facilitated (Burgess, Burgess, & Esveldt, 1970). It may be possible to view the present findings in this context. The dose temporal contiguity (5 s) be- tween the high-p commands and the low-p com-
mand (i.e., interspersal) may have facilitated com- pliance to low-p commands whose historical association was presumably with relatively weak reinforcement. In Experiment 3, extending the IPT interval to 20 s may have induced discrimination between high-p and low-p commands, resulting in lower percentages of compliance to low-p com-
mands. These speculations could be tested by ran-
domly interjecting a low-p command in the high-p sequence and introducing stimuli antecedent to the low-p command that may enhance its discrimina-
tion (e.g., verbal statements or different experi- menters correlated with different command types).
Future investigations of the high-p command sequence and/or applications of behavioral mo- mentum could improve on some aspects of the methodology used in these experiments. First, all sessions were conducted by an experimenter who was aware of the experimental hypotheses. Where possible, staff who are uninformed of the experi- mental hypotheses should conduct sessions to avoid possible expectation effects and to assess the prac- tical value of the procedures for applied settings. Second, as a novel intervention, the acceptability of the high-p command procedure should be as- sessed by those who use it and observe its use. The topography of high-p requests may need to be altered to be consistent with the subject's age and functioning level to gain widespread acceptance of the procedure. Finally, general conclusions regard- ing the comparative efficacy of the high-p procedure and the contingency management intervention (Ex- periment 5) should be made with caution. The degree of effectiveness of the contingency manage- ment procedure may have been influenced by the level at which the criterion was set. Perhaps a lower criterion would have resulted in shorter task du- rations and represented an optimally effective rep- resentation of contingency management (Van Hou- ten, 1987).
Finally, we hope that the present findings will stimulate additional research on the use of high-p command sequences as well as investigations of behavioral momentum in applied settings. Con- ceivably, modifications could be made to the high-p command procedure that would make it applicable to a range of target behaviors and populations. In addition to studies with an applied focus, more research is needed to establish the appropriateness of the behavioral momentum analogy. Specifically, more experiments are needed that directly manip- ulate variables affecting behavioral mass and ex- amine their relationship to the degree of persistent responding in applied settings. Enthusiasm for the applied value of the behavioral momentum analogy must await the outcome of these studies. However, at the very least, we must credit the heuristic value
139
140 F. CHARLES MACE et al.
of Nevin et al.'s (1983) basic research in stimu- lating the development of an innovative treatment for noncompliance.
REFERENCES
Ayllon, T., Garber, S., & Pisor, K. (1976). Reducing time limits: A means to increase behavior of retardates. Jour- nal of Applied Behavior Analysis, 9, 247-252.
Baer, A. M., Rowbury, T., & Baer, D. M. (1973). The development of instructional control of deviant children. Journal of Applied Behavior Analysis, 6, 289-298.
Baer, D. M., & Sherman, J. A. (1964). Reinforcement control of generalized imitation in young children. Jour- nal of Experimental Child Psychology, 1, 37-49.
Breiner, J., & Beck, S. (1984). Parents as change agents in the maintenance of grocery shopping skills by severely mentally retarded adolescents. Applied Research in Men- tal Retardation, 5, 259-278.
Brigham,T.A.,&Sherman,J.A. (1968). Anexperimental analysis of verbal imitation in preschool children.Journal of Applied Behavior Analysis, 1, 151-160.
Burgess, R. L., Burgess, J. M., & Esveldt, K. C. (1970). An analysis of generalized imitation. Journal ofApplied Behavior Analysis, 3, 39-46.
Cataldo, M. F., Ward, E. M., Russo, D. C., Riordan, M., & Bennett, D. (1986). Compliance and correlated problem behavior in children: Effects of contingent and noncontingent reinforcement. Analysis and Intervention in Developmental Disabilities, 6, 265-282.
Catania, A. C. (1984). Learning (2nd ed.). Englewood Cliffs, NJ: Prentice Hall.
Cuvo, A. J. (1976). Decreasing repetitive behavior in an institutionalized mentally retarded resident. Mental Re- tardation, 14, 22-35.
Deitz, S. M. (1978). Current status of applied behavior analysis: Science versus technology. American Psychol- ogist, 33, 805-814.
de Villiers, P. (1977). Choice in concurrent schedules and a quantitative formulation of the law of effect. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook of operant behavior (pp. 233-287). Englewood Cliffs, NJ: Prentice Hall.
Fjellstedt, N., & Sulzer-Azaroff, B. (1973). Reducing the latency of a child's responding to instructions by means of a token system. Journal of Applied Behavior Anal- ysis, 6, 125-130.
Forehand, R. L., & McMahon, R. J. (1981). Helping the noncompliant child: A clinician's guide toparent train- ing. New York: Guilford Press.
Hayes, S. C., Rincover, A., & Solnick, J. V. (1980). The technical drift of applied behavior analysis. Journal of Applied Behavior Analysis, 13, 275-286.
Herrmstein, R. J. (1970). On the law of effect. Journal of the Experimental Analysis of Behavior, 13, 243-266.
Hock, M. L., & Mace, F. C. (1986). Increasing command compliance through the use of a high-probability com- mand sequence. Poster presented at the annual convention ofthe Association for Behavior Analysis, Milwaukee, WI.
Holt, G. (1971). Systematic probability reversal and con- trol of behavior through reinforcement menus. The Psy- chological Record, 21, 465-469.
Mace, F. C. (1987). Applications ofbehavioral momentum concepts in the treatment of aberrant behavior. Chair, symposium presented at the annual convention of the Association for Behavior Analysis, Nashville, TN.
Michael, J. L. (1980). Flight from behavior analysis. The Behavior Analyst, 3, 1-21.
Myerson, J., & Hale, S. (1984). Practical implications of the matching law. Journal of Applied Behavior Anal- ysis, 17, 367-380.
Neef, N. A., Shafer, M. S., Egel, A. L., Cataldo, M. F., & Parrish,J. M. (1983). The dass specific effects of com- pliance training with "do" and "don't" requests: Ana- logue analysis and classroom application. Journal ofAp- plied Behavior Analysis, 16, 81-99.
Nevin, J. A. (1974). Response strength in multiple sched- ules.Journal ofthe Experimental Analysis ofBehavior, 21, 389-408.
Nevin,J.A. (1979). Reinforcementschedulesandresponse strength. In M. Zeiler& P. Harzem (Eds.), Reinforcement and organization ofbehavior (pp. 117-158). New York: John Wiley & Sons.
Nevin, J. A., Mandell, C., & Atak, J. R. (1983). The analysis of behavioral momentum. Journal of the Ex- perimental Analysis of Behavior, 39, 49-59.
Page, T. J., & Iwata, B. A. (1986). Interobserver agree- ment: History, theory, and current methods. In A. Poling & R. W. Fuqua (Eds.), Research methods in applied behavior analysis (pp. 99-126). New York: Plenum Press.
Parrish, J. M., Cataldo, M. F., Kolko, D., Neef, N. A., & Egel, A. (1986). Experimental analysis of response co- variation among compliant and inappropriate behavior. Journal of Applied Behavior Analysis, 19, 241-254.
Peterson, R. F. (1968). Some experiments on the orga- nization of a class of imitative behaviors. Journal of Applied Behavior Analysis, 1, 225-235.
Pierce, W. D., & Epling, W. F. (1980). What happened to analysis in applied behavior analysis? The Behavior Analyst, 3, 1-9.
Russo, D. C., Cataldo, M. F., & Cushing, P. J. (1981). Compliance training and behavioral covariation in the treatment of multiple behavior problems.Journal ofAp- plied Behavior Analysis, 14, 209-222.
Schoen, S. F. (1983). The status of compliance technology: Implications for programming. The Journal of Special Education, 17, 483-496.
Sidman, M. (1960). Tactics of scientific research. New York: Basic Books.
Van Houten, R. (1987). Comparing treatment techniques:
BEHAVIORAL MOMENTUM 141
A cautionary note. Journal of Applied Behavior Anal- ysis, 20, 109-1 10.
Zeiler, M. (1977). Schedules of reinforcement. In W. K. Honig & J. E. R. Staddon (Eds.), Handbook of operant behavior (pp. 201-232). Englewood Cliffs, NJ: Prentice Hall.
ReceivedJune 10, 1987 Initial editorial decision August 25, 1987 Revisions received November 24, 1987; January 8, 1988 Final acceptance March 8, 1988 Action Editor, Brandon F. Greene