JOIiRX.4L
OF
EXPERIMENTAL
Some
Variables
CHILD
15, 1t!-21
PSYCHOLOGY
Affecting with
Compliance
Instructions
BRADLEY University
Children’s
(1973)
of
BUCHER Western
Ontario
Children were given instructions to perform different tasks. Compliance was studied as a function of several variables: the prevailing pattern of reinforcement for different instructions, the availability of a competing reinforced activity, and the similarity among tasks. Rate of compliance tended to decrease when the probability of reinforcement for compliance decreased, or when a competing reinforced activity was available. Differential reinforcement over the tasks did not always produce discriminative responding even under favorable conditions. Novel instructions obtained compliance at the same rate as that for frequently repeated instructions, independent of the novel instruction’s past or present reinforcement history. Greater discrimination in responding was found when tasks were less similar. Results imply that sets of instructions may form a response class, with characteristics similar to those found in studies of generalized imitation.
The receptive use of speech is of great importance for the development of normal behavior patterns in children. Adults use speech to shape skills and stimulus controls, and to point out response consequences. One of the variables controlling the child’s use of speech cues is availabliity of reinforcement for compliance with instructions. Schutte and Hopkins (1970) used teacher attention to reinforce inst.ruction-following in kindergarten children. Compliance decreased when attention was not provided, and rose after its reinstatement. Other studies show similar results (Hopkins, 1968; Wahler, 1969). 1 The author thanks the Wilton Place Children’s Center at the Neuropsychiatric Institute at UCLA, and Mark James Hawkins, who assisted in this project. The work NIH Grants MH-19045 and MH-11440. Preparation of tated by the Children’s Psychiatric Research Institute of may be obtained from Bradley Bucher, Department of Western Ontario, London 72, Ontario, Canada. Copyright All rights
@ 1973 by Academic Press, of reproduction in any form
10 Inc. reserved.
of Los Angeles, the nurses Bailow, Tom Lyman, and was supported in part by the manuscript was faciliLondon, Ontario. Reprints Psychology, University of
COMPLIANCE
WITH
INSTRUCTIONS
11
Innumerable instructions are potentially available, exceeding those to which any organism has actually been exposed. Responses evoked by these cues may often be unusual or novel. Thus compliance with classes of instructions is of interest. These may be studied using techniques devised to study classes of imitative behaviors (Baer & Sherman, 1964; Metz, 1965; Baer, Peterson & Sherman, 1967; Lovaas, Berberich, Perloff, & Schaeffer, 1966; Brigham & Sherman, 1968; Peterson, 1968). These studies show that nonimitative children will learn to imitate a set of experimenter-acts when reinforced for imitating only a subset. Imitations that are never reinforced persist through many sessions when mixed with reinforced imitative cues. Peterson (1968) and Martin (1971) found that compliance with nonimitative instructions persisted when reinforcement was given for imitative behaviors, and conversely; implying that modeling and compliance behaviors may belong to t.he same response class. In the generalized imitation paradigm the experimenter provides a cue and the child must match either the action or a visual stimulus configuration that action produces. In compliance, as defined here, an instruction cues which of a number of available tasks is to be carried out. The experimenter does not perform the behavior, and the stimulus context is the same for all trials. Compliance is not imitative either of the experimenter’s behavior or of a stimulus event his behavior has produced. Several generalized imitation studies have used children like those used here, who modeled and followed instructions well. These studies have found that extinction of imitative behavior is not readily obtained. Steinman (1970a,b) and Peterson and Whitehurst (1971) got no noticeable reduction in imitations even with extinction and DRO procedures. Steinman also failed to get differential imitation of sets of reinforced and unreinforced experimenter-acts, even after many sessions. Such differentiation does not. appear to have been produced in generalized imitation studies, excepts with special procedures, e.g., instructions not to respond for unreinforced experiment-acts (Steinman, 1970a,b). In this paper some typical procedures of generalized imitation research were followed, to investigate variables controlling compliance and differential compliance with reinforced and unreinforced instructions, and the generalizat,ion of compliance or noncompliance from familiar to novel instructions. Specifically, aims were (1) to reduce the high initial rate of compliance by withdrawing reinforcement for compliance and making available an alternative reinforced activity; (2) to study variables affecting differential compliance with reinforced (SD) and unreinforced (S*) instructions and (3) to examine the effect of introducing unreinforced novel instructions when compliance with the continuing instructions was either high, or low.
12
BRADLEY
BUCHER
METHOD
Subjects The S, was a 7-year-old normal boy, who scored in the upper 170 in tested int,elligence. The SS--~ were emotionally disturbed nonretarded children, aged 9-12, hospitalized for short-term treatment. Results for S, were collected during one period, and those for S,S, were obtained somewhat later. General Procedure Children were observed individually, for 20-30 min sessions.A brief standardized statement, about. the nat,ure of the instructions, tasks, and possible reinforcement was given at the beginning of each session.No details of specific reinforcement contingencies were given. Care was taken to exclude any implication that noncompliance with instructions would lead to aversive consequences. An instruction to perform one of a set of tasks was given at, regular intervals in an unsystematic order. Reinforcement was given upon task completion when appropriate. Instructions were given at fixed intervals, that is, independently of compliance or noncompliance. Interaction during the sessionwas avoided, to eliminate inadvertent social reinforcement. Experimental conditions for each subject were chosen to obtain differentiated responding. Three factors were introduced for the purpose: the number of SD and SA tasks was kept small-three each for S, and one each for S,-S,; extinction conditions were used to increase noncompliance so that subjects could experience its nonaversive consequences; and an alternative t,ask was used to provide a competing reinforced response.The sequence followed for each subject was chosen to obtain a moderate rate of compliance, to facilitate investigation of conditions affecting this rate. These conditions included differential reinforcement for every compliance (CJW ; reinforcement for a subset of instructions (DRF) ; and no reinforcement for compliance (Ext). Reinforcement for the alternat’ive reinforced task was sometimesavailable (Alt) Insfmctions
and
Tasks-Generd
Tasks and instructions were chosen that could be easily and reliably scored and could be performed with little training. Pretraining was given to determine how much time to provide for each. Initial compliance with instructions was high in all subjects. The alternative task was introduced in an early sessionafter the child had mastered it in pretraining. At the beginning of the sessionthe subject was informed that he could work at, the t,ask when he had spare time and that performance would sometimes
COMPLIANCE
WITH
INSTRUCTIONS
13
be rewarded. “Spare time” was defined on request as time not spent in carrying out instructions. This information was not repeated during the session. Reinforcement for performance in the alternative task was independent of compliance. These rewards were delivered routinely, without. comment or social interaction. The time permitted for compliance was fixed after the first one or two sessions to permit task completion if the subject worked continuously. Variability in the time required was not, more than 10-15 sec. At the end of the allotted interval, if the subject was working on the instructed t.ask, he was told his time was up but he could finish if he wished, with no reinforcement. This instruction was very rarely given. Specific Instructions
and Tasks
S,. This child cooperated throughout the program. Six tasks were presented with 24 trials in each session. The subject put together woodblocks in one of a set of numbered configurations (“Task X, go”). The configurations were displayed on a large card. The instruction was the only cue to the configuration to be matched. Reward was brief social approval and tokens exchangeable for candy and small toys. Thirty seconds were provided for each task. By the end of the first session performance time had stabilized at about 10 sec. The alternative task was a jigsaw puzzle, introduced with the instruction : “If you have spare time after doing the tasks, you can work on this puzzle.” The reinforcement for connecting two pieces was also stated, with the information that this reward would not always be provided. Reinforcement, was a token, without social approval. Conditions CRF, DRF and Ext were presented, with and without the alternative task (Alt). In the DRF conditions three of the six tasks were reinforced. Ext and Ext + Alt were introduced to facilitate noncompliance. Ss,-,. Subjects were instructed to do one of two number puzzles. Eight cards, each with a four-digit number, were available. Four cards were to be combined into a 16-digit number. A large display panel showed eight possibilities, numbered 1-8. Instructions to do additional puzzles were presented in later conditions as probes. The instruction alone cued the task to be performed. Reinforcers were tokens exchangeable for pennies, small edibles and toys. No social reinforcement was given. For S, a similar set of tasks using letters was introduced without pretraining in session 36, and time per trial was made the same as for the number tasks. Sessions consisted of 30 trials. Time permitted for each trial was fixed after the first one or two sessions at 3645 set, different for each subject depending on speed of performance. The alternative task used a commercially available puzzle box. Single
14
BRADLEY
BUCHER
letters were printed on small flat plastic squares, in a shallow rectangular casing. The squares could be moved along rows or columns on the inside surface of the case. The case contained 4 X 6 = 24 spaces, and had 23 squares, so that one space was always available into which an adjoining square could be moved. Letters were initially in scrambled order. The desired final placement was presented on a large panel. Each four-letter column was to be a four-letter word, except, the last. column. A token was given for each word, for two children, or for each letter, for the third, according to the child’s pretested skill. When all the words in one puzzlebox were completed another was provided. Subjects were routinely promised a reward for remaining seated during the session. One subject did not earn a reward on one day. The relation between the reward and the contingency (remaining seated) was clearly stated each day, to avoid t,he implication that the reward depended on compliance with instructions. Noncompliance that did not require interruption of the procedure was ignored. Conditions DRF and Ext were presented, with and without Alt. Two instructions were routinely presented and one was never reinforced. In sessions where novel tasks (probes) were introduced the original tasks were continued with equal frequency. For S, a new set, of t’asks was presented after session 36, using comhinations of letters, similar to the number task. Eight possible combinations were indexed by the letters A-H on a large display panel. The numerical tasks were still present,. The two sets of tasks were separated spatially on t,he table, and the cards differed in size and color. In sessions 36-38 eight unreinforced word tasks were presented randomly, with number task 3. Task 3 was presented on half the trials. In sessions 39-60 tasks 3 and D were presented (three reinforced). Tasks 4 and E, (four reinforced) were used in sessions 61-79, and, with reversed contingencies, in sessions 8&92. Beginning in session 93, four pairs (7-8, 5-A, B-C, F-6) were presented, for four sessions each (16 sessions), to investigate effects of task similarity. Then the sequence was repeated with reinforcement contingencies reversed. Then this entire sequence of 32 sessions was repeated. RESULTS
Resulm for S, and S, are summarized below. Figures 1 and 2 show graphs for S, and for sessions l-60 for X,. The S, was observed for 21 sessions. Average compliance ranged from 92 to 100% for four sessions of reinforcement for all six tasks (CRF). In five sessions of DRF, compliance for the three SD instructions stayed at 100%. Compliance for the S* instruct,ions ranged from 92 to 100%. In three Ext sessions average compliance ranged from 50 to 34% with no
COMPLIANCE
WITH
15
INSTRUCTIONS ixt+Alt
DRF*Alt
DRF :, :
6 x
4
8
12 SESSIONS
16
FIG. 1. Frequency of compliance for subject 2, for various experimental conditions. Experimental conditions: Differential reinforcement for one of two instructions (DRF) ; and no reinforcement (Ext). The presence of reinforcement for the alternative task is noted as (Alt). Probe instructions (unreinforced) are shown in the session in which they appeared.
FIG. 2. Frequency of compliance for Subject 4, for various experimental conditions. Experimental conditions: Differential reinforcement for one of two instructions (DRF) ; and no reinforcement (Ext). The presence of reinforcement for the alternative task is noted as (Alt). Probe instructions (unreinforced) are shown in the session in which they appeared. In sessions 3&38 a new set of tasks was introduced with task 3: tasks A-H. Sessions 60-156 are described in the text.
16
BRADLEY
BUCHER
downward trend. The alternate task was then added for one session. Compliance fell to 8%. Then DRF + Alt was presented for eight sessions. Discrimination developed quickly. Compliance for S” instructions rose steadily from 50 to 100%. Compliance for SA instructions fell steadily from 40 to 0%. Figure 1 shows results for &. No discrimination occurred although compliance rate was readily manipulable. Compliance with the unreinforced probes and the routine instructions are essentially the same. The S, was observed for 32 sessions.Compliance for both instructions was 100% for three sessions.Then in DRF + Alt (eight sessions)average compliance fell gradually from 100 to 65%. No discrimination appeared. In Ext + Alt (sessions 12-17) average compliance dropped from 48% in the first sessionto 0% in the last five. Then in DRF (sessions18-23) compliance rose from 0 to 100% (for the last four sessions)with no discrimination. A return to Ext + Alt (sessions23-28) gave 100% compliance in the first two sessionsand 0% in the last three. DRF (sessions 2930) again produced no discrimination, with average compliance in each session over 90%. Ext + Alt (session 31) gave 23% compliance, then DRF gave 90% (session32) for both tasks. Thus rate of compliance was affected by the reinforcement condit.ions, but discriminat’ion did not occur. One unreinforced probe was introduced in each session beginning with session15. Five probes were used, for one to six sessionseach. Compliance rate for probes paralleled compliance for Sn and SA. Probes introduced when the rate was low were ignored. If the rate was high, they were obeyed. Compliance for a given probe shifted with the rate for SD and SA: in one case from 100% to 0% in adjacent sessions. Figure 2 shows results for S,, t,hrough session60. In session61 t,asks 3 and D were changed to 4 and E, with task 4 reinforced. Compliance for SD (task 4) fell to 67% then rose to 100% on the next, session,and remained high (averaging over 90%). Compliance for SA st.ayed in the range 30400/o for five sessionsthen fell to 0% and remained low (averaging below 10%). In session80 reinforcement was reversed: task 4 became SA. Compliance with both tasks rose immediately. Compliance with t.he new SD remained high throughout (averaging over 85%). Compliance with the new SA fell, to 0% by session88, and averaged below 20% to session92. In sessions93-156 the effects of task similarity were examined. The results are discussedin a later section. The data are relevant. for exploration of variables affecting generalization, discrimination and rate of compliance. Compliance
under
Various
Experimental
Conditions
Compliance with all instructions was initially high. Decreases were most marked when reinforcement rate was reduced while reinforcement
COMPLIANCE
WITH
INSTRUCTIONS
17
for the alternative task was introduced (e.g., DRF to Ext + Alt). Increases occurred when reinforcement increased while reinforcement for the alternative task was removed. Four such changes occur for S,, five for S,, and one for 8,. Compliance changed immediately in seven of these cases. In three a marked change occurred within one to two sessions. Change in reinforcement alone also shows an effect. The 8, complied less when DRF was changed to Ext. Later, compliance increased when Ext + Alt was changed to DRF + Alt. The X, stopped complying within one session when DRF + Alt was changed to Ext + Alt. The S, gradually stopped complying when DRF + Alt was changed to Ext + Alt. Removing reinforcement for the alternative task shows a less clear effect. The S, decreased in compliance when Ext was changed to Ext + Alt.. The X, decreased in compliance when DRF was changed to DRF + Alt and increased when DRF + Alt. was changed to DRF. The S, decreased when DRF was changed to DRF + Alt. The S, did not shift compliance when DRF changed to DRF + Alt, in either of two cases. Introduction
of linreinforced
Instructions
(Probes)
Unreinforced probes were introduced for S,, S,, and 8,. Probe trials used the same stimulus materials presented as the routine trials. Probes were continued for one or several sessions, and some were withdrawn and then reintroduced. In almost every case performance rate for probes was about the same as for the S* instruction. When compliance for SA shifted over sessions probe compliance shifted also. In two cases, (S, and S,), the same probe was introduced when S* compliance was low, and later when compliance was high. Compliance for the probes showed a similar shift. Thus the ongoing rate of compliance controlled responding to probes, rather than the probe’s previous rate of compliance. There is no evidence of extinction of compliance for probes. lliscrivainntion
Between
SD and XA
Compliance rates for SD and SA instructions were generally about the same. The X, and S, showed no evidence of discrimination, although changes in rate of compliance were readily obtained by manipulating experimental conditions. Two subjects did show discrimination. The S, began to discriminate after session 14. The variables controlling this effect could not be investigated. For S,, compliance for SD was higher than for S* in most early sessions, but t,here was no increase in this effect over sessions. Discrimination appeared after session 39, after a set. of distinctively different tasks was introduced. Discrimination developed gradually to session 60. Then the two tasks were each replaced by a different task from the same set (not shown in Figure 2). As before, the number task was reinforced. Compliance for the two tasks continued as before
18
BRADLEY
BUCHER
with little change. In session 80 the reinforcement contingency was reversed, and compliance rates reversed after session 86. Thus, differential responding to SD and SA persisted over a substitution of tasks and was reacquired quickly after reinforcement reversal. Discrimination
as a Function
of Task &&la&y
For S,, beginning as session 93, task pairs 7-8, 5-A, B-C and F-6 were introduced for four sessions each. Two pairs were similar tasks, and two were dissimilar. The first task in each pair was reinforced. The alternative task was reinforced throughout. This sequence was then repeated with reversed contingencies; and then this 32 session sequence was repeated. Discrimination developed quickly for each pair. Average compliance was 79.4 and 7.9% for SD and SA. Compliance with SD was greater in 63 of 64 sessions. There was no significant trend in the discrimination rate for successive pairs. To examine the effects of task similarity, compliance with SD and SA was examined separately. Average compliances for SD in the 16 sets of task pairings were calculated and ordered numerically. Compliance in sets for similar tasks was compared with compliance in sets of dissimilar pairings, using Wilcoxon’s rank-sum test (Bradley, 1968). The reappearance of the same tasks in different sets was ignored for this analysis. The result. showed no effect of degree of similarity. A similar analysis was carried out for SA. Compliance rate was higher for SA (P < .Ol) when SD and SA were similar. Thus, the effect of task similarity appears to have affected primarily compliance to SA. DISCUSSION
The results show that, children’s responsesto instructions are affected by the contingencies available for compliance. Initially compliance was high when all or a subset of instructions was reinforced. Compliance decreased during extinction, and increased when reinforcement was reinstated. Removal and reinst.atement of the alternative reinforced task also showed an effect, though less clearly. Reinforcement variables also controlled compliance for unreinforced probes. Compliance rate followed the rate for routine instructions: probes when compliance was low were ignored; and when compliance was high, probes also evoked compliance. The same probes, introduced into different sessions, showed sudden shifts in compliance following the intervening changes in compliance for the continuing inst,ructions. The present st.udy included several manipulations to produce favorable condit,ions for differential responding. Extinction reduced the originally high rates of compliance and provided experience with the consequences
COMPLIANCE
WITH
INSTRUCTIOKS
19
of noncompliance; small sets of tasks were used; and a reinforced alternative task provided reinforcement for noncompliance. Discriminated responding could not be readily obtained. Two subjects eventually showed discrimination. For one child a discrimination developed between dissimilar tasks which persisted when new tasks from the same sets were substituted; and a reversal of reinforcement contingencies was quickly mastered. Similar tasks were also discriminated in later sessions. Greater similarity between SD and SA increased responding to SA, but did not affect SD responding. These findings resemble results of studies of generalized imitation, for originally imitative subjects. Reinforcement for one or a few cues generalized to a larger set. Subject,s discriminated slowly or not at, all between SD and SA experimenter cues even with extensive t.raining. These results justify the extension of the concept. of response class to compliance with instructions. Compliance with a large set of instructions depends on contingencies associated with a subset of the instructions, in that compliance for the other elements in the set rises and falls with changes in the reinforcement contingencies for the subset,. In discussing generalized imitation Baer and Sherman (1964) conjectured that similarity between the experimenter’s and child’s behavior became discriminative for reinforcement, and acquired seconda.rp reinforcement value. This explanation is not adequate for the present study, nor to Martin (1971) on compliance, or Brigham, Saunders and Sherman (1968) on generalized matching-to-sample. The behavior of the subjects does not match an experimenter’s cue. A second explanation emphasizes the difficulty in discriminating among the numerous, similar and intermixed SD and SA cues (Bandura, 1968). This hypothesis does not seem applicable to the data from L‘%-~. Extreme measures short of specific instructions were taken to facilitate the child’s of the differential reward contingencies. The same two in“awareness” structions were presented in all sessions, and compliance w&h one was never reinforced. This discrimination seems trivial for a normal school age child, but little or no discrimination appeared for two of these three subjects, and S, discriminated only after about 45 sessions. Steinman (1970a,b) found that children may show discrimination between SD and SA (i.e., their performance rates differ) under certain experimental conditions, but yet not do so under others. Subjects who imitated all cues when presented singly, imitated the reinforced cue when both were presented. Subjects stopped responding when so instructed. Steinman suggested that instructions such as “DO this” and other situational features set the occasion for compliance based on transfer from the child’s prior learning history. Aversive contingencies for noncompli-
20
BRADLEY
BUCHER
ante or conditioned reward value for compliance might mediate this transfer. Of course such setting effects operate in any human study. An important question here is whether such effects in laboratory studies produce generalization phenomena that would not occur widely under less artificial conditions. A high degree of social control does not seem necessary. Brigham, Saunders and Sherman (1969) found generalized matching or mismatching in preschoolers on a mechanical apparatus, with no experimenter present. The results of the present st.udy imply that the use of reinforcement for a subset of instructions may be highly generalized, beyond the child’s apparent awareness of the contingencies, Generalization in performance beyond that shaped by specific contingencies has been observed in research in many other areas: language learning (Schumaker & Sherman, 1970; Guess, 1969) ; rule learning (Bourne, 1970; Gag&, 1969) ; and in performance at many cognitive t,asks. Children appear to acquire wide repertoires of behavioral skills without, direct reinforcement for each specific interact,ion. Some understanding of the manner in which these skills are developed and maintained should be valuable. Research into generalized imitative behavior, generalized compliance with instructions, and such an generalized matching behavior, may assist. in developing understanding. REFERENCES J. .4. The development of imitation by rrinforcing behavioral similarity to a model. Jo~rnnl o,f the Experimental Annlysis o.f Behavior, 1967. 10, 405-416. UAER, D. M., & SHERMAN. J. A. Reinforcement control of generalized imitation in young children. Jortrnnl of Ex:perimental Child Psychology, 1964. 1, 37-49. BANDVRA, A. Social-learning theory of identificatory processes. In D. A. Goslin (Ed.), Hnndbook of socializatiorl theory n~tcl research. Chicago: Rand McNally. 1968. ROT-RYE. I,. E.. JR. Knowing and using concepts. Psychologicnl Review, 1970, 77, 546 556. BRADLEY. J. V. Dbtribufion-free stnfisficnl tests. Englewood Cliffs. N. J.: PrenticeHall, 1968. BRIGHAM, T. A., & SHERMAS. J. 8. An experimental analysis of verbal behavior in preschool children. JolLrnal of Applied Behavior Analysis, 1968, 1, 151-158. GAGN~~. R. M. Learning and human development. Psychological Review, 1968, 65, 177-191. HOI~KINS. B. L. Effects of candy and social reinforcement, instructions, and reinforrement, schedule learning on the modification and maintenance of smiling. Jo~(rnnl of App&d Behavior Amlysis. 1968, 1, 121-130. LOVAAS. 0. I.. BERBERICH. J. P.. PERLOFF, B. F.. & SCHAEFFER. B. Acquisition of imitative speech by schizophrenic children. Science, 1966, 151, 705-707. MARTIX, J. il. The control of imitative and nonimitative behaviors in severely retarded children through “generalized-instruction following.” Journal of Experimentrtl Child Psychology. 1971. 11, 390-400. B.IER.
11. M.,
PETKRSOS,
11. F..
I!Z SHERMAS,
COMPLIAh’CE
METZ,
WITH
INSTRUCTIONS
21
J. R. Conditioning generalized imitation in autistic children. Jouraal of Experimental Child Psychology, 1965, 2, 389-399. PETERSON, R. F. Some experiments on the organization of a class of imitative behaviors. Jowxal of Applied Behnvior Analysis, 1968, 1, 22.5236. SCRUMAKER, J., & SHERMAN, J. A. Training generative verb usage by imitation and reinforcement procedures. Jou,nal of Applied Behavior Analysis, 1970. 3, 273288. SCHTTTE, R. C., & HOPKINS, B. L. The effects of teacher attention on following instructions in a kindergarten class. Journal of Applied Behavior Analysis. 1970, 3, 117-122. SHERMAN, J. A., SAUNDERS, R. R., & BRIGHAM, T. A. Transfer of matching and mismatching behavior in preschool children. Jownal of Eqwimental Child Psychology, 1970, 9, 489498. STEINMAN, W. M. Generalized imitation and the discrimination hypothesis. Journal of Experimental Child Psychology, 1970, 10, 79-99. (a) STEINMAN, W. M. The social control of generalized imitation. Journal of Applied Behavior Analysis. 1970, 3, 159-168. (b) WAHLER, R. G. Oppositional children: A quest for parental reinforcement control. Journal of Applied Behavior Analysis, 1969, 2, 159-170.