Journal of Abnormal Psychology 1977, Vol. 86, No. 5, S28-53S

Reinforcement Analysis of Interaction in Problem and Nonproblem Families James J. Snyder Pius XII School, Holy Cross Campus The patterns of consequents that problem and nonproblem family members provided for each other's behavior and the responsivity of the members to those consequents were examined. The interactions of 10 problem and 10 nonproblem families were observed in a laboratory setting and coded using a behavioral rating system. Problem families provided more aversive and fewer positive consequents for prosocial behavior and more positive and fewer consequents for deviant behavior than nonproblem families. Problem family members were less responsive to the consequents than nonproblem family members. The results are discussed in terms of the development and maintenance of dysfunctional family systems. From the social learning point of view, family interaction can be conceptualized as a series of verbal and nonverbal behaviors emitted by the family members. Each member's behavior is controlled to a degree by the other family members. Numerous behavior modification studies have demonstrated that deviant child behaviors can be modified by reprogramming the social environment (Patterson, Cobb, & Shaw, 1973; Patterson, Reid, Jones, & Conger, 197S). Other studies, focusing on marital problems, have also shown that spouses after being taught behavior modification procedures are able to increase marital satisfaction (Patterson & Hops, 1973; Weiss, Hops, & Patterson, 1973). The implication is that in families with deviant children or in which there is marital conflict, deviant behavior is reinforced at a high rate and prosocial behaviors are often ignored or rein forcers are used in a noncontingent manner. Parents of children who demonstrate infrequent deviant behavior have been found to This research was submitted in partial fulfillment of the PhD degree at Southern Illinois University. The author wishes to express gratitude to Donald Shoemaker, Scott Monroe, Bill Saunders, Dru Copeland, and Ginny Nunez for their help in this study. Requests for reprints should be sent to James J. Snyder, who is now at the Department of Psycholog}', Wichita State University, Wichita, Kansas 67208.

reinforce prosocial behaviors more appropriately and to ignore or punish deviant behaviors (Johnson, Wahl, Martin, & Johansson, 1973; Wahl, Johnson, Johansson, & Martin, 1974). Nondistressed couples provide more positive and fewer aversive consequents than distressed couples (Birchler, 1972; Vincent, 1972). These "nonproblem" families make reinforcement contingent on the type of behavior displayed. Theoretically, the different rates of deviant behavior displayed by the members of problem and nonproblem families are due, in part, to the differential use of consequents that the members provide for each other's behavior. Direct comparisons of problem and nonproblem families support this hypothesis. The parents of problem children provide more positive consequents for deviant behavior than the parents of nonproblem children (Sallows, 1972; Shaw, 1971). In addition, the problem children are more frequently punished for prosocial behaviors than are nonproblem children (Shaw, 1971; Taplin, 1974). However, the schedules of consequents provided for deviant and prosocial behavior may not be the only variable that differentiates problem and nonproblem families. Taplin (1974) found a nonsignificant relation between the amount of positive consequents that parents provided for their problem children and the rate of deviant behavior displayed by those children. Patterson and Cobb

528

REINFORCEMENT ANALYSIS OF FAMILY INTERACTION

529

(1971) and Patterson (1976) report that the degree of marital satisfaction (a score of 120 or class of consequents that suppress prosocial more on the MAS). A total of 20 families, 10 in each group, were obbehaviors in problem families accelerate detained. The two groups did not differ on the demoviant behaviors in those families. These re- graphic variables measured. The parents' mean age sults are not found in nonproblem families. was 36 years; they had been married for an average Thus, as well as receiving different schedules of 13 years and had an average of three children. of consequents, problem and nonproblem fam- The fathers averaged 17 years and the mothers averaged 15 years of schooling, and their socioecoilies are differentiated by their responsivity to nomic index (Hollingshead & Redlich, 1958) was 20. the consequents. The average age of the observed child was 8 years. The present study is an attempt to replicate The problem children were best described by the conduct problem factor of the BPC. and extend the data on the differentiation of problem and nonproblem families on the basis of the schedules of consequents provided for Observation Procedure deviant and prosocial behaviors and on the After a 15-minute habituation period, a sample of basis of the families' responsivity to those 45 minutes of interaction of each of the problem and consequents. The results of past studies are nonproblem families (including only the father, confounded by a difference in the treatment mother, and the targeted 5-to-10-year-old male child) status of the samples (Jacob, 197S); the pres- was videotaped through a one-way mirror. The inent study uses families that have never been teraction took place in a research room in the psychology department. The room was designed as a involved in treatment. Past studies have used combination living room and playroom, containing a home observation methodology and have chairs, end tables, newspapers, magazines, toys, and focused mainly on the behavior of the child games. The family members were instructed to do as and the behavioral determinants. In this they pleased but not to leave the room. The families were given a general rationale for the study and were study, interaction was observed in a labora- invited to see the videotape apparatus if they wished. tory setting, and the analysis was extended to include the behavior and the determinants of Rating Procedure the behavior of all the family members. The videotaped interaction was rated using the time sampling observation procedures developed by Method Patterson, Ray, and Shaw (1969). Each videotape was rated three times, once for each family member Subjects as the target. The order in which the family memFamilies with a male child between 5 and 10 bers were observed was randomly predetermined. years of age were recruited by means of a letter and This procedure required the observer to code the a follow-up phone call. These families were identi- target member's behavior and the response of the fied using birth announcements in past local news- other family members every 6 sec using a signal papers. Those families that agreed to participate were superimposed on the videotape. This procedure prosent two copies of both the Locke-Wallace Marital vided a running account of family interaction as it Adjustment Scale (MAS; Locke & Wallace, 1959) revolved around each family member. and the Peterson-Quay Behavior Problem Checklist The videotapes were rated by five graduate stu(BPC; Peterson & Quay, Note 1). The parents were dents from the psychology department who were asked to complete both questionnaires independently, extensively trained prior to viewing the videotapes applying the BPC to their 5-to-10-year-old male and who were blind to the status of the families child. being rated. Each rater was assigned two problem Only those families meeting preestablished criteria and two nonproblem families. As well as coding the for problem and nonproblem families and who had interaction of their assigned families, each rater coded never been involved in psychotherapy were included 45 minutes (IS minutes for each family member as in the observation portion of the study. Problem the target) from one of the families of each of the families were denned as those in which (a) both other raters. All raters were aware that their records parents reported a high frequency of behavior prob- would be checked for reliability but did not know lems for their child (a score of 21 or more on the which videotaped segments would be coded by the BPC) and (b) one or both parents reported a high other raters. All ratings were completed within 6 degree of marital conflict (a score of 95 or less on weeks of training; reliability checks were randomly the MAS). Nonproblem families were defined as distributed within this period. those in which (a) both parents reported few beAgreement between pairs of raters was calculated havior problems for their child (a score of 11 or less by dividing total agreements by total agreements on the BPC) and (b) both parents reported a high plus disagreements for each 30-sec segment of inter-

JAMES J. SNYDER

530

Table 1 Comparisons of Mean Percentages and Rates (in minutes) of Displeasing Behavior in Problem and Nonproblem Families Nonproblem families

Problem families

Family member

Rate/minute

%

Rate/minute

%

Z

Sons Fathers Mothers Total

.327 .187 .182 .698

3.3 1.9 1.9 2.3

.560 .593 .273 1.427

5.6 5.9 2.7 4.7

5.35*** 9.95*** 2.87** 1.97*

*p < .05. **p < .01. *** p < .001. action. The event-by-event analysis required agreement on behavior, family member, and sequence for each 6-sec rating period. The percentage of agreement ranged from 61% to 86% with an average of 70%. Chance agreement, calculated by summing the squared value of the mean base rates observed for each coding category, was 19%.

Abstractions oj the Coding Categories Simple, nontechnical examples of each of the behavioral codes were presented to each of the observed family members. Each member was asked to rate each behavioral code twice (one each for the other two members) on a three-point scale ranging from dislike (1) to like (3). The experimenter assisted younger children when necessary. The pleasing (prosocial) or displeasing (deviant) function of the target member's behavior was defined by the mean rating of each coding category for the target by the other two family members. Any category with a mean rating from 1.00 to 2.00 was denned as displeasing; a mean rating from 2.01 to 3.00 was denned as pleasing. For example, the children's potential behaviors were denned by the mean ratings of the categories for children by the fathers and mothers. The ratings were also used to define the nontargets' responses as positive, neutral, or aversive in consequent function. The consequent function of the nontargets' responses was defined by the mean rating of each category for the nontarget members by the target. Any coding category receiving a mean rating from 1.00 to 1.66 was defined as aversive, from 1.67 to 2.33 as neutral, and from 2.34 to 3.00 as positive. For example, the consequent function of the fathers' and mothers' responses to the children was denned by the mean ratings of the categories for the fathers and mothers by the children. These definitions were obtained separately for problem and nonproblern families. The experimenter then converted the basic categories into pleasing or displeasing behavior and positive, neutral, or aversive consequents on the basis of the family members' ratings.

The abstractions obtained from the rating procedures and data on their validity are reported elsewhere (Snyder, Note 2). There was nearly complete agreement on the definition of behavior as pleasing or displeasing both across family members and across types of families. The definition of the consequent function of the categories was less consistent. There was disagreement due both to the member providing the consequent and the type of family from which they came. However, the number of categories that the two samples defined as positive, neutral, or aversive was nearly the same. Those categories that were defined differently by the two samples were observed at such a low rate that they had minimal effect on the differences reported to distinguish problem and nonproblern families.

Results Problem and nonproblem families were differentiated by the rate of displeasing behavior displayed, by the percentage of the types of consequents provided for pleasing and displeasing behavior, and by their responsiveness to the consequents. Most of the behaviors displayed by both problem and nonproblem families were pleasing. Comparisons between the amounts of displeasing behavior observed in problem and nonproblem families were made using Z tests for proportions. Table 1 presents a summary of the rates and percentages of displeasing behavior displayed by the families. The rate of displeasing behavior in problem families was twice that observed in nonproblem families, Z(18) = 1.97, p < .05. In addition, this difference was observed in the comparisons of specific family members from the two samples. Problem sons, 2(18) = S.3S, p < .001, problem fathers, Z(18) = 9.9S, p
C A )

Problem families No. events J°(Pk/Pli-^Cf,) .955 .948 .934 .945

Z -1.07 -2.14* -1.61 -2.64*

Sons Fathers Mothers Total

1,124 930 1,176 3,230

.946 .927 .919 .931

1,754 1,424 2,143 5,321

Sons Fathers Mothers Total

6,965 7,379 7,277 21,621

P(PZ 2 /PZi-»C P ) .963 .982 .976 .974

6,178 6,030 5,996 18,204

.954 .968 .961 .961

2.57* 5.83** 5.00** 2.41*

Sons Fathers Mothers Total

8,089 8,309 8,453 24,851

Baseline .961 .976 .968 .968

7,932 7,454 8,139 23,525

Baseline .954 .964 .954 .957

4.19** 4.44** 4.67** 6.11**

P(Ph/Pll^CF)

*p < .05. **p < .01.

more aversive consequents for displeasing behavior than problem sons, Z(18) — 3.SS, p < .001. Problem and nonproblem mothers did not differ on the amount of positive or aversive consequents received for displeasing behavior. The lack of clear differences in the consequents received for displeasing behavior was due, in part, to the small sample of displeasing behavior observed. The differences between problem and nonproblem families described above were not due to differing rates of interaction. All t tests comparing the observed frequency of target behavior and consequents were insignificant. However, potential differences in the rate of interaction were minimized by the priority coding and time sampling demands of the rating system (Jones, 1973). The responsivity of problem and nonproblem families to the consequents provided for their behavior was assessed by calculating the conditional probability (P) for the immediate recurrence of pleasing (PI) behavior [P(Plz/ PI)] or displeasing (D) behavior [P(D%/ DI)] in the next rating time frame, given either an aversive (C A ) or a positive (CP)

consequent. The conditional probability of the recurrence of pleasing behavior as a function of aversive and positive consequents is presented in Table 4. Nonproblem family members were more likely to persist in pleasing behavior, given a positive consequent, than were problem family members. Although these differences were consistent and statistically significant, the absolute differences in probability of recurrence were small, ranging from .01 to .02. The conditional probability of the recurrence of displeasing behavior as a function of aversive and positive consequents is presented in Table 5. Nonproblem sons, Z(18) = 3.60, p < .001, and nonproblem fathers, 2(18) = 4.04, p < .001, were less likely to persist in displeasing behavior than their problem counterparts, given an aversive consequent. In addition, nonproblem families as a unit demonstrated a greater tendency to change from displeasing to pleasing behavior than did problem families, Z(18) = 1.88, p < .10, given an aversive consequent. The absolute differences in the persistence of displeasing behavior as a function of aversive consequents

REINFORCEMENT ANALYSIS OF FAMILY INTERACTION

ranged from .07 to .11. More important, aversive consequents for displeasing behavior actually increased the probability of its recurrence (compared to baseline) in problem families. In nonproblem families, aversive consequents had a decelerative effect on displeasing behavior. The samples differed less consistently in their probability of persisting in displeasing behavior as a function of positive consequents. Nonproblem fathers, Z(18) = 2.43, p < .05, and nonproblem family units, Z(18) = 2.43, p < .05, were less likely to continue displeasing behavior, given a positive consequent, than their problem counterparts. The observed differences were small in absolute value, ranging from a probability of .02 to .06. Discussion The rate of displeasing behavior was higher in problem than in nonproblem families whether analyzed by family unit or specific

533

family member. Problem and nonproblem families also differed in their contingent use of consequents. Members from nonproblem families reacted to pleasing behavior by providing more positive reinforcement and less punishment than the members of problem families. Nonproblem family members were also less likely to reward and more likely to punish displeasing behavior than were problem family members. In the problem families, there were no contingencies. The probability of receiving a positive, neutral, or aversive consequent was independent of the behavior displayed. Members from nonproblem families were more discriminating. Their responses fostered the development of prosocial behavior and discouraged the display of deviant behavior. For a member from a nonproblem family, it paid to act in a way that was pleasing to the other family members. In problem families, the payoff was nearly the same for both pleasing and displeasing behavior.

Table 5 Conditional Probabilities (P) that Displeasing (D) Behaviors Persist as a Function of Aversive (£A) and Positive (Cp) Consequents Nonproblem families Target

No. events P(ZV-Di -> CA)

Problem families No. events

P CDs/A -*£A)

Z

Sons Fathers Mothers Total

69 43 29 141

.276 .223 .211 .241

110 84 57 251

.357 .333 .278 .331

-3.60*** -4.04*** -.21 -1.88*

Sons Fathers Mothers Total

171 103 102 376

P(Dt/Dl -> Cp) .318 .364 .288 .319

340 391 150 881

P(Di/Dl -» CP) .268 .240 .243 .251

1.16 2.43** .78 2.43**

Sons Fathers Mothers Total

240 146 131 517

Baseline .304 .322 .267 .300

*p < .10. **p < .05. * * * # < . 01.

Baseline 450 475 207 1,132

.289 .257 .251

.269

1.50 4.81*** 1.00 1.29

534

JAMES J. SNYDER

The rates of displeasing behavior observed the consequents provided suggest that probin this study are comparable to those found lem and nonproblem families can also be difin studies using home observation (Johnson ferentiated on the basis of their responsivity et al., 1973; Patterson, 1976). This suggests to consequents. This difference is most apthat there are negligible differences in the type parent in the effect of aversive consequents of behavior observed in the home and that on displeasing behavior. In nonproblem famobserved in the laboratory (given a stimulus ilies, punishment was associated with a supsetting comparable to the one used in the pression of displeasing behavior. In problem present study). This is at variance with past families, punishment was associated with an studies that compared home and laboratory acceleration rather than a suppression of disobservation (Eyberg & Johnson, 1974; Mar- pleasing behavior. These data are consistent tin, Johnson, Johansson, & Wahl, 1976), but with those of Patterson and Cobb (1971) these studies used a laboratory task that was and Patterson et al. (197S) for children and dissimilar from free interaction in the home. demonstrate a similar effect for their parents. The data on the consequents provided for The responsivity of problem and nonproblem deviant and prosocial behavior obtained in families to positive consequents for displeasthis study are also consistent with those ing behavior and to positive and aversive found in previous research (Patterson et al., consequents for pleasing behavior was small 197S; Sallows, 1972; Shaw, 1971; Taplin, and of less practical significance. 1974). Both found that problem and nonThe lack of contingencies in problem famproblem families differ in their unprogrammed ilies appears to adversely affect their memcontingency management, and both support bers' responsiveness to social reinforcers. the hypothesis that the observed differences Cairns and Paris (1971) and Cairns (1972) in the rate of deviant behavior is a result of have shown that children with a history of this contingency management. Although the noncontingent reinforcement are less respondifferential use of positive and aversive con- sive when reinforcers are made contingent. sequents by problem and nonproblem families It is also possible that the social reinforcers is consistent with past research, the actual used by problem families are less effective percentages of the consequents reported in due to insufficient pairing with primary reinthis study differ from those reported in stud- forcers (Hill, 1968). ies using home observation (Johnson et al., The differences found between problem and 1973; Martin et al., 1976; Sallows, 1972; nonproblem families in the rates of deviant Taplin, 1974). The parents observed in this behavior, in the contingent use of reinforcestudy tended to provide more positive and ment and punishment, and in their responsivfewer aversive consequents for the behavior ity to consequents were observed in all famof their children than the parents observed ily members. Past research has focused mainly in previous studies. This tendency was not the result of the classification of positive and on the behavior and determinants of the beaversive stimuli; the classification used was havior of children. The present study sugsimilar to that used in the studies cited. The gests that the family system is disrupted and tendency to respond more positively was that all family members contribute to the probably due to the differences in observation development and maintenance of deviant belength and procedures. Reactivity decreases havior. with the length of observation (Johnson & Bolstad, 1973), and the tendency found in Reference Notes this study is similar to but not as extreme as that found by Martin et al. (1976) using 1. Peterson, D. R., & Quay, H. C. Manual for the Behavior Problem Checklist. Unpublished manua laboratory analogue measure of parent-child script, University of Illinois, 1967. interaction. 2. Snyder, J. J. The relationship between self reThe data on the conditional probability of port and operant definitions of reinforcers in family interaction. Unpublished manuscript, 1976. the recurrence of behavior as a function of

REINFORCEMENT ANALYSIS OF FAMILY INTERACTION

References Birchler, G. R. Differential patterns of instrumental affiliative behavior as a function of marital distress and level of intimacy. Unpublished doctoral dissertation, University of Oregon, 1972. Cairns, R. B. Fighting and punishment from a developmental perspective. In W. J. Arnold & D. Levine (Eds.), Nebraska symposium on motivation, Lincoln: University of Nebraska Press, 1972. Cairns, R. B., & Paris, S. G. Informational determinants of social reinforcement effectiveness among retarded children. American Journal of Mental Retardation, 1971, 76, 361-369. Eybcrg, S. M., & Johnson, S. M. Multiple assessment of behavior modification with families. Journal of Consulting and Clinical Psychology, 1974, 42, 594-606. Hill, W. F. Sources of evaluative reinforcement. Psychological Bulletin, 1968, 69, 132-146. Hollingshead, A. D., & Redlich, F. C. Social class and mental illness. New York: Wiley, 1958. Jacob, R. Family interaction in disturbed and normal families: A methodological and substantive review. Psychological Bulletin, 1975, 82, 33-65. Johnson, S. M., & Bolstad, 0. D. Methodological issues in naturalistic observation: Some problems and solutions for field research. In L. A. Hamerlynck, L. C. Handy, & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice. Champaign, III.: Research Press, 1973. Johnson, S. M., Wahl, G., Martin, S., & Johansson, S. How deviant is the normal child? A behavioral analysis of the preschool child and his family. In R. D. Rubin, J. P. Brady, & J. D. Henderson (Eds.), Advances in behavior therapy (Vol. 4 ) . New York: Academic Press, 1973. Jones, R. P. Behavioral observation and frequency data: Problems in scoring, analysis, and interpretation. In L. A. Hamerlynck, L. C. Handy, & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice. Champaign, 111.: Research Press, 1973. Locke, H., & Wallace, K. Short marital adjustment and prediction tests: Their reliability and validity. Marriage and Family Living, 1959, 21, 251-255. Martin, S., Johnson, S. M., Johansson, S., & Wahl, G. The comparability of behavioral data in laboratory and natural settings. In E. J. Mash, L. A. Hamerlynck, & L. C. Handy (Eds.), Behavior modification and families. New York: Brunner/ Mazel, 1976.

535

Patterson, G. R. The aggressive child: Victim and architect of a coercive system. In E. J. Mash, L. A. Hamerlynck, & L. C. Handy (Eds.), Behavior modification and families. New York: Brunner/Mazel, 1976. Patterson, G. R., & Cobb, J. A. A dyadic analysis of aggressive behaviors. In J. P. Hill (Ed.), Minnesota symposia on child psychology (Vol. 5). Minneapolis: University of Minnesota Press, 1971. Patterson, G. R., Cobb, J. A., & Shaw, D. A. A social engineering technology for retraining families of aggressive boys. In H. E. Adams & I. P. Unikel (Eds.), Issues and trends in behavior therapy. Springfield, 111.: Charles C Thomas, 1973. Patterson, G. R., & Hops, H. Coercion: A game for two, intervention techniques for marital conflict. In H. E. Ulrich & P. Montjoy (Eds.), The experimental analysis of behavior. New York: Appleton-Century-Crofts, 1973. Patterson, G. R., Ray, R. S., & Shaw, D. A. Manual for the coding of family interactions. (Available from ASIS National Auxiliary Publications Service, 1969, under Document No. 01234.) Patterson, G. R., Reid, J. B., Jones, R. R., & Conger, R. E. A social learning approach to family intervention (Vol. 1). Eugene, Oreg.: Castilia, 1975. Sallows, G. Comparative responsiveness of normal and deviant children to naturally occurring consequences. Unpublished doctoral dissertation, University of Oregon, 1972. Shaw, D. Family maintenance schedules for deviant behaviors. Unpublished doctoral dissertation, University of Oregon, 1971. Taplin, P. Changes in parental consequation as a function of intervention. Unpublished doctoral dissertation, University of Oregon, 1974. Vincent, J. P. Problem solving in distressed and nondistressed married and stranger dyads. Unpublished doctoral dissertation, University of Oregon, 1972. Wahl, G., Johnson, S. M., Johansson, S., & Martin, S. An operant analysis of child-family interaction. Behavior Therapy, 1974, 5, 64-78. Weiss, R. L., Hops, H., & Patterson, G. R. A framework for conceptualizing marital conflict, a technology for altering it, and some data for evaluating it. In L. A. Hamerlynck, L. C. Handy, & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice. Champaign, 111.: Research Press, 1973.

Received November 8, 1976 Revision received May 16, 1977 •

Reinforcement analysis of interaction in problem and nonproblem families.

Journal of Abnormal Psychology 1977, Vol. 86, No. 5, S28-53S Reinforcement Analysis of Interaction in Problem and Nonproblem Families James J. Snyder...
592KB Sizes 0 Downloads 0 Views