Symposia on Child Psychology, Volume 7

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY, VOLUME 7

This page intentionally left blank

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Volume 7

ANNE D. PICK, EDITOR

THE UNIVERSITY OF MINNESOTA PRESS

•

MINNEAPOLIS

©Copyright 1973 by the University of Minnesota. All rights reserved. Printed in the United States of America at Lund Press, Minneapolis.

Library of Congress Catalog Card Number: 67-30520 ISBN 0-8166-0699-4

PUBLISHED IN THE UNITED KINGDOM AND INDIA BY THE OXFORD UNIVERSITY PRESS, LONDON AND DELHI, AND IN CANADA BY THE COPP CLARK PUBLISHING CO. LIMITED, TORONTO

Preface

THE SEVENTH of the annual Minnesota Symposia on Child Psychology was held at the University of Minnesota in October 1972, and the papers in this volume are the contributions of the participants in that symposium. For that occasion, a group of eminent researchers was invited by the members of the Institute of Child Development to present reports of their own research programs and to consider together matters of mutual concern. A variety of topics is represented among the papers of this volume, as among those of earlier volumes. This diversity of contents is intentional, and it reflects the range of interests of those currently engaged in the study of development. Not accidentally, some similarities of emphasis also are revealed among these reports of investigations of quite different problems. Parents and educators share with researchers a desire to identify the effects of socialization practices on children's development. Diana Baumrind's interest is focused on relations between parents' methods of exercising authority and the development of independence and competence in their children. She urges the use of converging strategies; the validity of conceptual and operational definitions of types of parental control can be established by empirical observations of methods of control used by parents and by observations of child behaviors. She has identified patterns of control by parents which are associated with social responsibility and independence in young children, and she suggests specifically that punishment is not necessarily harmful but that it may be an effective means of control under appropriate conditions. C. G. Beer cautions against generalization by analogy from the findings of ethology to questions about human development. Nonetheless, the v

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY strategy used in his own research on vocal communication between parents and young Laughing Gulls can be generalized without modification to the study of development in any species. Having established that young Laughing Gulls recognize by ear their own parents, Beer then studied the development of this individual recognition. He observed its existence in young birds less than 24 hours post-hatching and subsequently documented changes over time in the behavior of the chicks in response to the vocalizations of the parents. He conducted experiments in the laboratory and observations in the settings in which the behaviors under study occur naturally. The use of this procedure allowed him first to validate the existence and course of development of individual recognition and then to interpret it intelligently in the context of social development. A second paper in this volume in which is reflected a concern for identifying relations between socialization practices and child development is that of Norma Feshbach. She hypothesizes that socioethnic differences in socialization practices are functionally related to socioethnic differences in cognitive development. Reinforcement style, the giving of praise and of criticism, is the aspect of socialization studied. It was selected as a general variable, one which can be observed in parents, teachers, and children. First she observed a relation between socioethnic identity and reinforcement style among children and their mothers in one culture. Then she documented the generality of the relation by observing it in two other cultures. Finally, she supported the interpretation of a functional relation between reinforcement style and cognitive performance by finding that mothers of children who had difficulty learning to read used different styles of reinforcement than did mothers of children who read easily. John Hagen and Gordon Hale have collaborated in an investigation of the development of selective attention in children. Two stages of information processing are described. At the first stage, relevant information is identified, and at the second stage, attention to the relevant information is maintained. Hagen and Hale suggest that developmental changes are to be found primarily at the second stage. From the results of a series of studies in which children performed a variety of tasks, it was suggested that, with development, children become more efficient in the use of selective attention. The generality of this account of the development of selective attention and the relation of attention to environmental variables was explored by extending the investigation to other populations. David Klahr uses information-processing models to formulate theories vi

PREFACE

and to test hypotheses about human cognitive development. He constructs models which are specific to an experimental task and which, unlike metaphor models of task performance, are detailed and disprovable by the children whose behavior is being simulated. One goal of this strategy is to propose and to test principles of developmental change based on models containing explicit assumptions about task performance at different times. The tasks for which Klahr has constructed programs for performance include several which derive from the developmental theory of Jean Piaget. Harriet Rheingold proposes a new interpretation of a fact about human infants which is often observed but little noticed. Infants who can locomote leave their mothers to seek out nearby new objects and places. She notes that such separations from the mother are not stressful, at least not for the infant, and she suggests that they be viewed not simply in the context of the dependence of the child on the mother, but as reflections of the emergence of independence in the child. After observing behavior rightfully termed independent during the first year of life, she studied systematically one class of independent behavior during the second year. She recorded children's explorations in a free play situation, and she describes active, eager children who, upon finding new nearby rooms and new toys, spent more time with the toys than with their mothers. The children also shared their discoveries with their mothers, suggesting that "Show and Tell" as well as other components of independent behavior develop much earlier than commonly has been assumed. Financial support for this seventh symposium was provided again by a Public Health Service grant from the National Institute of Child Health and Human Development (HD-01765), by a grant from the Graduate School of the University of Minnesota, and by the Institute of Child Development. In addition, the work of the editor was carried out at the Center for Advanced Study in the Behavioral Sciences, and the support from that institution and from its staff is acknowledged with appreciation. A number of individuals bore responsibility for conducting this sympo sium, a task made more complicated for them by the editor's absence from the scene before the symposium. Helen Dickison was director-inresidence, and she was assisted by John Drozdal, Virginia Eaton, Elizabeth Haugen, and Douglas Sawin among others. The capable staff of the University of Minnesota Press have guided the publication process from manuscripts to book with maximum efficiency. vii

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Finally, gratitude is extended to the investigators who conducted the research and prepared the papers which are collected here. The talent, conscientiousness, and enthusiasm of these individuals made the editorial task a pleasure. ANNE D. PICK Stanford, California May 1973

viii

Table of Contents

The Development of Instrumental Competence through Socialization BY DIANA BAUMRIND

3

A View of Birds BY c. G. BEER

47

Cross-Cultural Studies of Teaching Styles in Four-Year-Olds and Their Mothers BY NORMA D. FESHBACH

87

The Development of Attention in Children BY JOHN W. HAGEN AND GORDON H. HALE

117

An Information-Processing Approach to the Study of Cognitive Development BY DAVID KLAHR

141

Independent Behavior of the Human Infant BY HARRIET L. RHEINGOLD

178

List of Contributors

205

Index

209

This page intentionally left blank

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY, VOLUME 7

This page intentionally left blank

DIANA BAUMRIND

The Development of Instrumental Competence through Socialization

THE major objective of the program of research to be described here is to identify the effects of alternative patterns of parental authority on the development of instrumental competence in the child. Emphasis will be placed on the upbringing of girls to encourage independence. The program of research has been in progress for over a decade. (Detailed presentations of the research data are found in Baumrind [1967, 197la] and Baumrind & Black [1967].) The young child's development is the result neither of spontaneous maturing of inborn capacities nor of automatic adaptation to programed stimuli. It is, rather, the result of increasingly complex interactions with socializing adults, primarily parents, who, during the early years, have the power to control these interactions. Children are not the originators of their own actions in the sense that adults should be. An adult can contribute to his own development by altering the stimuli which impinge upon him and by defining objectives for himself which, once formulated, then structure his actions. A child, on the other hand, will be presented with stimuli and asked to accomplish goals formulated for him by his upbringers. Innate and maturational predispositions, present at birth, and mediated by neurophysiological processes, interact throughout the individual's life with environmental factors to determine the course of development. Although maturation of the child's nervous system provides opportunities NOTE: The program of research discussed in this paper was supported by the National Institute of Child Health and Development under Research Grant HD02228. 3

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY for development, these opportunities can be realized only in a facilitating social environment designed by knowledgeable adults. Socialization is the process by which the young person, through education, training, and imitation, acquires his culture as well as the habits and values congruent with adaptation to that culture; thus, he attains skill in instrumental and expressive functions. In the studies reported here, instrumental competence is defined by social responsibility, independence, achievement orientation, and vitality. By social responsibility is meant behavior which is friendly rather than hostile to peers, facilitative rather than disruptive of others' work, and cooperative rather than resistive of adult-led activity. By late childhood, the qualities of objectivity and self-control are important correlates of social responsibility. By independence is meant behavior which is ascendant rather than submissive, purposive rather than aimless, and self-determining rather than conforming. By achievement orientation is meant behavior in which the child seeks rather than avoids intellectual challenge and problem solves persistently and efficiently rather than inefficiently and impulsively. Vitality refers to the child's level of biological energy and vigorous appearance. Authoritarian, Permissive, and Authoritative Models of Child Training In 1961,I began to study child-rearing practices associated with instrumental competence in children. In this section, the two studies based upon a typological approach to child rearing will be discussed. The first study, with a small sample, was a pilot study. The second, longitudinal study is still in progress. FIRST STUDY

Methods. Subjects were 32 three- and four-year-olds selected from among all the children enrolled at the Child Study Center, Institute of Human Development, University of California, Berkeley, during the fall semester of 1961. The selection procedures began with the assessment of all the children on five dimensions intended to measure instrumental competence at ages three and four: self-control, approach-avoidance tendency, self-reliance, vitality (from buoyant to dysphoric), and peer affiliation. Each dimension was illustrated and given concrete meaning for the nursery school teachers by reference to relevant time sample categories and by 4

DIANA BAUMRIND

instances of actual observed behavior. After being observed for fourteen weeks, the children were ranked on each dimension both by their nursery school teacher and by the observing psychologist. The 52 children who received one of the five highest or the five lowest rankings on at least two of the five dimensions were observed further, individually, in a laboratory setting where they were given some standardized tasks. For example, one task included three puzzles graded in difficulty so that each child experienced easy success, probable success, and certain failure. Their responses to success and failure were observed and rated on the five dimensions by the testing psychologist and by the observing psychologist. In order for a child to remain in the study, the observing and the testing psychologists' ratings of the child in the two settings had to concur. Using these multiple assessment procedures, three contrasting groups of children were obtained, each with a clear-cut, stable pattern of interpersonal attributes. The three groups of children were selected in order to test a set of hypotheses concerning the interacting effects on child behavior patterns of parental control, parental maturity demands, parent-child communication, and parental nurturance. Pattern I contained all children who were ranked high on Vitality (from Buoyant-Dysphoric), Self-Reliance, Approach-Avoidance Tendency, and Self-Control (6 girls, 7 boys). Pattern II contained all children who ranked low on Peer Affiliation and Vitality and did not rank high on Approach-Avoidance Tendency (7 girls, 4 boys). Pattern III contained all children who ranked low on Self-Reliance, Self-Control, and Approach-Avoidance Tendency (3 girls, 5 boys) (Baumrind, 1967). Each mother-child pair was observed in a two-hour structured teaching experience followed by a play experience in the laboratory setting (see Baumrind, 1967, pp. 68-70 for description of the structured observation). Also, two home visits were made to each family by a psychologist who had not previously rated the child's behavior. The home visits were structured identically for each family, and they occurred during a period from shortly before the dinner hour until just after the child's bedtime. This two- to three-hour period is commonly known to produce instances of parent-child divergence and was selected for observation in order to elicit a wide range of critical control interactions under maximum stress. Each mother and father was interviewed separately and the interviews were tape recorded. These interviews were rated using a fifty-six-item scale based on the scales used by Sears, Maccoby, and Levin (1957). 5

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Families were rated on a twenty-four-item Parent Behavior Rating Scale (PBRS) based in part on the Fels Parent Behavior Rating Scales (Baldwin, Kalhorn, & Breese, 1949). These observations and interviews were used to rate parent's early activities on four dimensions of child-rearing practices selected for their theoretical importance as possible predictors of competence in preschool children. At the time this study began, considerable concern about the negative effects of strict discipline on children was being expressed, particularly in the clinical and educational literature (Becker, Peterson, Luria, Shoemaker, & Hellmer, 1962; Goodman, 1964; Kagan & Moss, 1962; Maslow, 1954; Neill, 1964; Rogers, 1960). In the study reported here, parental control was defined so that punitiveness and arbitrariness were not components of it. Thus, the effects on the child of strict discipline which was punitive could be distinguished from the effects of equally strict discipline which was neither punitive nor arbitrary. The operational definition of the four dimensions of parent-child interaction derive from observations made in both laboratory and home settings and consist of Summary Ratings for the Structured Observation (SRSO) and Home Visit Sequence Analysis (HVSA). The dimensions are defined as follows: Parental control refers to those parental acts intended by the parent to shape the child's goal-oriented activity; to modify his expression of dependent, aggressive, and playful behavior; and to promote internalization of parental standards. Parental control as defined here is not a measure of restrictiveness, punitive attitudes, or intrusiveness, but is a measure of strict discipline. The behavior of the mother in the teaching and play situations was rated on the following variables from the SRSO: (a) mother consistently enforced house rules (see Baumrind, 1967, p. 69, for description of house rules), (b) mother gave structure to child's activities, (c) mother (apparently) felt in control of the child's behavior, and (d) child acted in accord with mother's stated wishes. Both parents were rated on five HVSA variables. Two examples from the five HVSA variables are as follows (see Baumrind [1967] for a complete description of the variables): Positive Outcome is the percentage of parent-initiated control sequences where the child complies. The purpose of this measure is to assess the parent's ability to enforce directives. Positive Outcome by Persistence is the percentage of total parent-initiated control sequences where the parent achieves compliance after repeating the directive or increasing the use of 6

DIANA B A U M R I N D power, minus those parent-initiated control sequences where the parent does not persist and the child does not comply. The purpose of this measure is to assess the parent's ability to enforce directives when the child initially does not obey. Maturity demands refers primarily to the pressures put upon the child by the parents to perform up to his ability intellectually, socially, and emotionally (independence-training) and secondarily to the leeway given the child to make his own decisions (independence-granting). SRSO variables rated were (a) mother's expectation of child's intellectual attainment, (b) mother's demand for self-reliant child behavior, and (c) mother's demand for self-control on part of child. One example from the four relevant HVSA variables is: Independence Training, Control, the percentage of parentinitiated control sequences where the message concerns cognitive insight into cause-and-effect relations or factual knowledge about the world. The purpose of this measure is to assess the extent to which the parent's control efforts are directed at teaching the child more about natural and social reality. Clarity of parent-child communication means the extent to which the parent uses reason to obtain compliance, solicits the child's opinions and feelings, and uses overt rather than manipulative techniques of control. The SRSO variable items rated were (a) ease and spontaneity of verbal communication and (b) clarity of directives that the mother used. One of the three HVSA variables is Uses Reason to Obtain Compliance, the percentage of parent-initiated control sequences where the parent uses reason with directive. This measures the extent to which the parent offers a reason for a directive before the child objects to the directive. Nurturance refers to caretaking; to those parental acts and attitudes that express love and are directed at guaranteeing the child's physical and emotional well-being. Nurturance is expressed by warmth and involvement. By warmth is meant the parent's love and compassion for the child, expressed by sensory stimulation, verbal approval, and tenderness of expression and touch. By involvement is meant pride and pleasure in the child's accomplishments, manifested by words of praise and interest, and by conscientious protection of the child's welfare. The SRSO variables rated were (a) attentiveness to child; (b) warmth in the form of support, reassurance, nurturance; (c) absence of hostile behavior (such as belittling and sarcasm); and (d) solicitousness in spending time with the child and being involved with his performance and pleasure. One of the three HVSA 7

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY variables rated is Uses Positive Incentive and Reinforcement, the percentage of parent-initiated sequences involving either positive or negative incentives in which the positive reinforcements are used. This score measures the relative use of positive versus negative reinforcement. Results. The results derived from all these measures were as follows: Parents of Pattern I children (children designated mature), by comparison with other parents studied, were controlling and warm. They were rated on all measures as communicating more clearly with their children than did parents of children in either Pattern II or Pattern III. Despite parent behavior ratings indicating high readiness of reinforcement, especially by comparison with parents of Pattern III children, the homes of Pattern I children, according to the ratings, lacked discord and disciplinary friction. According to the interview, these parents used corporal punishment more often (persisting until compliance was obtained), than parents of Pattern III children, but ridicule, frightening the child, and withdrawal of love less frequently (n.s.). However, the parents of Pattern I children used positive reinforcement more than negative reinforcement to obtain compliance. Pattern I children subjected to consistent pressure for mature and obedient behavior were both socially responsible and assertive. This combination of high control and positive encouragement of the child's independent strivings was called Authoritative parental control. Parents of Pattern II children (children designated dysphoric and disaffiliated) were rated as lower on use of rational rather than coercive methods of control and as less nurturant and sympathetic with their children, but not as less controlling, than were parents of Pattern I children. According to interview data (significant only for mothers), the parents were more inclined to give respect for parental authority and a religious belief as reasons for their demands than were parents of Pattern I children. Parents of Pattern II children were rated as more controlling than were parents of Pattern III children. They did not attempt to convince the child through use of reason to obey a directive, nor did they encourage the child to express himself when he disagreed. Their expressed attitudes were less sympathetic and approving, and they more often admitted to frightening the child. These parents who, relative to other parents studied, were detached and controlling and somewhat less warm were called Authoritarian. By comparison with parents of Pattern I children, the parents of Pattern III children (designated immature) behaved in a markedly less controlling manner and were not so well organized or so effective in running their 8

DIANA B A U M R I N D households. According to the interviews, they did not feel in control of their child's behavior or appraise their influence on the child as high. Neither parent demanded much of the child, and fathers were lax reinforcing agents. Parents engaged in less independence training, although they granted the child's demands for independence and, according to ratings, babied their children more. By comparison with mothers of Pattern I (mature) children, mothers of Pattern III (immature) children used withdrawal of love and ridicule rather than overt power, physical punishment, or reason to obtain compliance. The most significant difference between parents of children who were dysphoric and disaffiliated (Pattern II) and parents of children who were immature (Pattern III) is that the controlnurturance ratios were in opposite directions: parents of the dysphoric and disaffiliated children (Pattern II) were more controlling and less warm, whereas parents of the immature children (Pattern III) were less controlling and warmer, although not so warm as parents of Pattern I children. These parents of Pattern III children were called Permissive. SECOND STUDY

In order to replicate and to extend the findings of this first exploratory study, a second study using similar methods and variables was initiated (Baumrind, 197la). The design of Study II differed from the design of Study I in that: (a) Parent-child relationships in Study II were examined for boys and girls separately, (b) Study II, still in progress, is longitudinal. The children are now eight and nine years of age and are being seen a second time. It is hoped to observe them again during adolescence. (c) The subjects in Study II were selected from thirteen nursery schools, privatecooperative, public school—cooperative, as well as university-operated facilities, and thus the sample is more varied than that of Study I. (d) Patterns in Study II were defined by parent scores rather than by child scores. (e) An attempt was made in Study II to identify patterns of parental authority in addition to the three identified in Study I. The distinction between two patterns of parental authority high in control but differing in encouragement of independence and individuality, Authoritarian and Authoritative, was maintained in Study II (Baumrind, 1966; 1968). In addition, a search was made for patterns low in control which would differ in whether lack of control reflected laxness and neglect or an ideologically motivated belief in giving children maximum freedom. Subjects in the second study were enrolled in the fall of 1967 and in the 9

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY spring of 1968 in one of the thirteen nursery schools. Of the 246 families who consented to observation of their children and completed a lengthy Parent Attitude Inquiry, there were 150 who were willing to participate in the home visit phase of the study, and whose nursery school age child obtained a Stanford-Binet IQ score of at least 95 and had no diagnosed developmental anomaly. The final sample consisted of 60 white girls, 74 white boys, and their families and 16 black children and their families.* By comparison with the population from which it was drawn, the sample contained children who were more cooperative and fathers who were less authoritarian. Child Behavior Clusters. During a period of three to five months, an observer recorded the interpersonal and social behavior of the children as they engaged in activities in nursery school. In addition, each child was rated while taking the Stanford-Binet. The entire protocol describing the child's behavior over the school semester and while taking the StanfordBinet was used to make ratings. In a previous study (Baumrind & Black, 1967), a two-dimensional, eight-cluster model of preschool behavior was developed. Analyses of the behaviors described in the protocols of this study were directed at determining the dimensionality of the behavior space covered by the items and its similarity to the previous model. As in the previous study, the first two clusters from a BC TRY (Tryon & Bailey, 1966) cluster analysis were uncorrelated. They accounted for 89 per cent of the mean of the squared original correlations for both sexes and for over 65 per cent of the initial estimate of communality. This more than met the criteria for a two-dimensional model. A principal-components solution was used to provide the most stable two-factor solution. Then all Q-sort items were plotted in this two-factor space with their factor coefficients used as coordinates. The items were formed into clusters on the basis of position on the circular plot, pattern of intercorrelation of contiguous items, and similarity of pattern for both sexes. The actual ordering of the six empirical clusters and a seventh overlapping cluster of theoretical interest is illustrated in Figure 1. Parent Behavior Clusters. Information about family interaction was obtained from observations and interviews in the home as in Study I (BaumThe data for the black children and their families were analyzed separately because the parent-child relationships were, as expected, not the same as for whites.

10

DIANA B A U M R I N D

rind, 1967). Fifteen constructs, rather than four as in Study I, described the relevant parental behaviors. Fifty Parent Behavior Rating (PER) scales were devised to assess the behavior of mother and father separately; twenty-five additional scales were devised to measure the joint influence of the parents. Each of the fifteen constructs was itself defined as a dimension so that the observers could summarize their impressions of the family by rating the family on the dimension. The items defining the constructs were subjected to the BC TRY cluster analysis procedures. The major empirical clusters which emerged are described as follows in terms of the initially conceived constructs. a. Firm Enforcement. This cluster included items which had been constructed to measure Firm versus Lax Enforcement Policy, and Obedience

Figure 1. Child behavior model, sexes combined, nursery school setting. Dashed lines inside circle represent principal axes from previous study (Baumrind & Black, 1967). 11

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY as a Salient Positive Value versus Obedience as a Nonsalient Negative Value. The mother and father solutions were closely comparable. b. Encourages Independence and Individuality. This second cluster was designated Encourages Independence and Individuality for both parents, although more than half of the defining items differ for the mother and the father. For the mother, all but two defining items had been constructed to measure three related constructs — Encourages versus Discourages Independence, Encourages versus Discourages Verbal Exchange and Use of Reason, and Promotes Individuality versus Social Acceptability. For the father, items defining Encourages versus Discourages Verbal Exchange and Use of Reason were important. In addition, the Flexibility and Clarity versus Inflexibility and Lack of Clarity of Parent's Views, and Confidence versus Lack of Confidence in Self as a Parent were prominent components of Father Cluster 2. c. Passive-Acceptant. For both parents, the majority of the items defining the Passive-Acceptant cluster were drawn from items devised to measure Willing versus Reluctant to Express Anger or Displeasure to Child. These items were intended to measure parental inhibition of aggression and to reflect passivity and mildness in the parent. d. Rejecting. Clusters designated Rejecting measured for both parents Expresses Punitive versus Nurturant Behavior. e. Father Promotes Nonconformity; f. Father A uthoritarianism. For the father, two additional clusters emerged which did not have their counterparts in clusters for mothers. The cluster designated Promotes Nonconformity was composed almost entirely of items designed to measure Promotes Individuality versus Social Acceptability. The cluster designated Authoritarianism was defined primarily by items designed to measure Encourages versus Discourages Independence, and Promotes Respect for Established Authority versus Seeks to Develop a Cooperative Working Relationship with the Child. g. Mother Cluster: Self-Confident, Secure, Potent Parental Behavior. Items in this cluster were designed to assess Flexibility and Clarity versus Inflexibility and Lack of Clarity of the Parent's Views, and Confidence versus Lack of Confidence in Self as a Parent. h. Five Joint PBR Clusters. An additional cluster analysis was performed on the twenty-five items devised to define five constructs describing the parents' joint conduct. The items intended to define operationally the five theoretical constructs emerged almost intact in the empirical clus12

DIANA BAUMRIND

ter analyses, except that Discourage Emotional Dependency was defined by only two of the items and was therefore quite limited in its meaning. These five joint clusters were designated Expect Participation in Household Chores; Enrichment of Child's Environment; Directive; Discourage Emotional Dependency; and Discourage Infantile Behavior. Pattern Definitions. Families were typed on the basis of their patterns of scores on these Parent Behavior Rating clusters: 54 families of white boys and 48 families of white girls were assigned to patterns. Most of the families not assigned had cluster scores which resembled one or another pattern in shape, but failed to meet the criteria for magnitude of scores. These patterns of parental authority were defined to produce contrast groups of families corresponding to more refined definitions of the patterns described in Study I. The focus here is on four theoretically interesting patterns, two high on control and two low on control. The two high on control are designated Authoritarian and Authoritative. The two low on control are designated Permissive and Nonconforming. Conceptual and operational definitions of the terms Authoritarian, Authoritative, and Permissive follow. The fourth term, Nonconforming, is defined operationally but not conceptually because it was originally conceived as a variant of Permissive upbringing. The Authoritarian parent values obedience as a virtue and believes in restricting the child's autonomy. This parent values the preservation of order and traditional structure as an end in itself. He or she does not encourage verbal give and take, believing that the child should accept the parent's word for what is right. In defining this pattern operationally, it was required that parents have scores above the median on Firm Enforcement, below the median in Passive Acceptant, and below the median in Encourages Independence and Individuality, and that the father score in the bottom third on Promotes Nonconformity or the top third on Authoritarianism. The Authoritative parent attempts to direct the child's activities in a rational, issue-oriented manner. Both autonomous self-will and disciplined conformity are valued by Authoritative parents; they affirm the child's present qualities but also set standards for future conduct. They use reason, power, and shaping by regime and reinforcement to achieve objectives. In defining this pattern operationally, it was required that Authoritative parents have scores above the median in Firm Enforcement, below the 13

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY median in Passive-Acceptant, and above the median in Encourages Independence and Individuality. The Permissive parent behaves in an affirmative, acceptant, and benign manner toward the child's impulses and actions. The aim of the ideologically aware Permissive parent is to give the child as much freedom as is consistent with the child's physical survival. Freedom to the Permissive parent means absence of restraint. As in the previous study, it was not possible to find a group of parents who corresponded precisely to the prototypic Permissive parent described above because many noncontrolling, passive-acceptant parents were also cool or uninvolved. The criteria were: scores below the median on Firm Enforcement, above the median on Passive-Acceptant, below the median on Expect Participation in Household Chores, and below the median on Directive. The criteria for assignment to the category Nonconforming were that both parents score in the top third of the distribution on Encourages Independence and Individuality or father scores in the top third on Promotes Nonconformity and in the bottom third on Authoritarianism. These parents were anti-authoritarian and some were anti-authority, but they did make demands on their children. The data for these parents are presented separately because Nonconforming parents differed from Permissive parents in that the former were less passive and exerted firmer control than the latter. Results. Comparisons among the four patterns on the fifteen constructs are illustrated in Table 1 — in Part A for parents of girls, and in Part B for parents of boys. Numbers refer to mean scores with a possible range of from 1 to 5 (see Baumrind [197la], pp. 82-83, for Ns and SDs). In Table 2, relations between patterns of parent behavior and child behavior are summarized — in Part A for girls and in Part B for boys. Numbers refer to cluster scores obtained from a sample with a mean of 50 and a standard deviation of 10 (see Baumrind [1971a], pp. 56-58, for Ns and SZ)s).Six comparisons of child-rearing practices and the related differences in child behavior for girls and boys will be discussed: (a) Authoritarian versus Authoritative; (b) Authoritarian versus Permissive; (c) Authoritarian versus Nonconforming; (d) Authoritative versus Permissive; (e) Authoritative versus Nonconforming; and (/) Permissive versus Nonconforming. a. A uthoritarian versus A uthoritative. It can be seen in Part A of Table 14

DIANA B A U M R I N D 1 (A versus B) that Authoritarian parents of daughters in comparison with Authoritative parents did not enrich their child's environment or present her with a flexible, clear position. The Authoritarian parents discouraged independence and verbal exchange, and did not promote individuality. They were relatively punitive. It can be seen in Part A of Table 2 (A versus B) that daughters of Authoritarian parents were significantly less Dominant and Independent, and were somewhat less Domineering, Purposive, and Achievement Oriented than were' the daughters of Authoritative parents. It can be seen in Part B of Table 1 (A versus B) that Authoritarian parents of sons by comparison with Authoritative parents did not enrich their child's environment or present him with a flexible, clear position. The Authoritarian parents were less firm in enforcing directives, and they promoted greater respect for established authority. They lacked confidence in themselves as parents, discouraged independence and verbal exchange in their sons, and did not promote individuality. They were relatively more punitive. In Part B of Table 2 (A versus B) it is seen that the sons of Authoritarian parents were significantly more Hostile, Resistive, and less Achievement Oriented than were sons of Authoritative parents. (Because sons of Authoritarian parents had significantly lower, although also above average IQ'S, and IQ scores were positively correlated with Achievement Orientation in this sample, intelligence cannot be ruled out as a significant source of the differences in these attributes defining competence.) b. Authoritarian versus Permissive. In Part A of Table 1 (A versus C) it is seen that Authoritarian parents of girls by comparison with Permissive parents were more directive, more likely to discourage emotional dependency, firmer in enforcing directives, valued obedience more, and promoted respect for established authority. The Authoritarian parents discouraged independence and verbal exchange. They were more willing to express anger, did not promote individuality, and demonstrated relatively punitive behavior. From Part A of Table 2 it is seen that these numerous differences in child-rearing practices showed no reliable relations with behaviors of the preschool girls studied. Authoritarian parents of boys by comparison with Permissive parents (Table 1, Part B, A versus C) expected more participation in household chores, were more directive, and were firmer in enforcing directives. They valued obedience more and promoted respect for established authority. They discouraged independence and verbal exchange. They were more 15

Table 1. Pattern Comparisons for Fifteen Parent Behavior Rating (PER) Constructs Pattern Mean Scores A. Authoritarian B. Authoritative C. Permissive IV. Nonconforming H. Harmonious A vs. others" A vs. B A vs. C A vs. IV A vs. H B vs. others0 B vs. C B vs. IV B vs. H C vs. others0 C vs. IV C vs. H IV vs. others0 H vs. others0

I

II

III

IV

V

34 3.6 2.6 3.4 3.2

1.6 2.9 2.0 3.2 2.6

3.1 3.2 2.1 2.8 2.0

3.1 3.1 2.6 3.0 3.0

2.8 3.1 2.1 2.5 3.0

10

.05

. .

.05

.01 .01 .01 .05 .05 .05

.05 .05

.10 .01

.05

.05 .05 .01

.10

.01 .05

.05

.05

VI

PER Constructs" VII VIII IX PART A. GIRLS

X

2.4 3.6 3.9 2.9 3.3 3.5 3.6 3.4 2.7 3.6 2.6 1.9 2.4 1.6 2.7 4.0 2.6 2.6 1.2 3.6 3.3 1.8 1.2 3.5 Significant Pattern Differences* .05 .05 .01 .01 .05 .01 .01 .01 .05 .05 .01 .01 .10 .01 .01 .01 .05 .10 .10 .01 .01 .01 .01 .01 .05 .01 .01 .05 .05 .01 .05 .05 .01 .10 .10 .10 .10 .05 .01 .05

XI

XII

XIII XIV

XV

2.1 3.6 3.7 3.8 3.7

2.3 3.8 3.8 4.4 4.3

3.3 3.1 2.1 3.3 2.5

3.3 2.5 2.5 2.0 2.3

.01 .01 .01 .01 .01

.01 .01 .01 .01 .01

.05 .01 .10

2.1 3.1 3.1 3.4 4.0 .01 .01 .05 .05 .01

.01 .01 .01 .01 .01

.01 .01 .05 .10 .10

.05 .05 .01

.10

Table 1 — Continued PER Items'1 I

Pattern

II

III

IV

V

VI

A. Authoritarian B. Authoritative C. Permissive IV Nonconforming

3.9 4.4 2.7 3.0

0

A vs. others A vs. B A vs. C A vs. IV B vs. others0 B vs C B vs. IV C vs others0 C vs. IV IV vs others0

VII

VIII

IX

X

XI

XII

XIII XIV

XV

3.3 3.9 2.6 3.5

2.4 3.2 3.3 3.6

2.6 3.8 3.5 3.6

3.1 3.4 2.2 3.0

3.6 2.4 3.0 2.0

PART B. BOYS

Mean Scores

.01 .10 .01 .01 .05 .01

1.6 2.8 1.9 2.7 .01 .01 .01 .01 .01 .05 .10

3.0 3.3 2.5 2.8

3.1 3.1 2.9 3.0

2.7 3.2 2.0 2.9

.10 .05 .01 .01 .05 .01 .10

.05 .01 .01 .05

2.5 3.6 2.5 3.5

3.9 4.6 2.1 3.0

3.8 3.6 2.5 1.8

3.2 2.8 2.1 1.7

Significant Pattern Differences13 .05 .05 .01 .01 .01 .05 .05 .01 .01 .01 .10 .05 .01 .01 .01 .01 .01 .01 .01 .01 .05 .05 .01 .01 .01 .01 .10 .01 .05 .05 .01 .05 .10 .05 .01 .01

.01 .05 .05 .01

.01 .01 .01 .01 .05

.05 .01 .01 .10 .01 .05

2.2 3.2 3.5 3.8 .01 .01 .01 .01

.01 .01 .10 .01 .05 .10 .10 .01

"PBR items are: I, expect participation in household chores; II, enrichment of child's environment; III, directive; IV, discourage emotional dependency; V, discourage infantile behavior; VI, flexibility and clarity of the parent's views; VII, firm enforcement; VIII, obedience as salient positive value; IX, promotes respect for established authority; X, confidence in self as parent; XI, encourages independence; XII, encourages verbal exchange; XIII, willing to express anger; XIV, promotes individuality; XV, expresses punitive behavior. b There were no significant differences for IV vs. H. c Others refers to all subjects of the same sex whose families were visited except those in the pattern under consideration.

Table 2. Pattern Comparisons for Q-Sorl Clusters and IQ Scores"

IQ

Q-Sort Cluster Pattern

I

III

II

IV

V

VI

Score VII Means

PART A. GIRLS A. Authoritarian (N = 10) . B. Authoritative (N = 11) . C. Permissive (N = 11) ... IV., Nonconforming (AT = 5) H. Harmonious (N = 6)

A vs. others" A vs. B A vs. H B vs. others'1 B vs. C B vs. IV B vs. H C vs. othersb C vs. H IV vs. others" IV vs. H H vs. others'5

-

46.8 50.2 50.4 52.5 40.2

47.6 52.4 50.6 45.7 44.5

46.0 48.,6 49.1 47.6 44.7 54.8 56..6 55.0 54.3 56.7 48.1 46.,2 45.0 50.4 48.5 42.3 40,.2 44.3 47.1 41.0 48.1 54.,1 56.2 58.6 56.3 Significant Pattern Differences

10

.10

.05

.10

.05

.01

.05

.01

.05 .05 .05

.05

01 .05 .01

05

.05

.10 .05

.10 .05

.10

.10 .10

.10 .05

.10 .01 .01 .05 .10 .01

122.6 132.7 128.3 129.8 135.8 .10 .10

.05 .01

PART B. BOYS A. Authoritarian (N ~ 16) . B. Authoritative (N =14) . C. Permissive (N = 11) . . . IV. Nonconforming (N = 8) A vs. others" A vs. B A vs. IV B vs. others" B vs. C B vs. IV C vs. others" C vs. IV IV vs. others

52.7 43.1 52.4 50.8

50.6 49,.4 50.7 47..6 47.4 46.0 51..9 54.2 56.,1 50.3 51.3 48..0 47.1 44..3 45.6 48.8 52,.8 51.1 57.,3 57.2 Significant Pattern Differences .10

5.1..1 43.,7 52,.1 46.,3

05

.05

01 05 10

.05 .05

.10

.01 .05 .05

.01

115.9 128.9 121.0 131.4 .05 .05 .05

.01

.01 .01 .05

.05 .05

.10

NOTE: 0-Sort Clusters are: I, Hostile-Friendly; II, Resistive-Cooperative; III, Domineering-Tractable; IV, Dominant-Submissive; V, Purposive-Aimless; VI, Achievement Oriented; VII, Independent-Suggestible. " Pattern comparisons which were not significant have been omitted. b Others refers to all subjects of the same sex whose families were visited except those in the pattern under consideration.

18

DIANA BAUMRIND

willing to express anger and did not promote individuality. As with the girls, these numerous differences in child-rearing practices showed no significant relations with the boys' behavior. Such differences for boys as did exist were in the opposite direction from that for girls — that is, sons of Permissive parents were less Achievement Oriented and Independent (despite their higher IQ'S) than were sons of Authoritarian parents. c. Authoritarian versus Nonconforming. Authoritarian parents of girls by comparison with Nonconforming parents (Table 1, Part A, A versus IV) did not enrich the child's environment or present their daughters with a flexible, clear position. The Authoritarian parents were firmer in enforcing directives, valued obedience more, and promoted respect for established authority. They discouraged independence and verbal exchange, and did not promote individuality. They were more punitive. There were no significant differences between the two groups in their daughters' behavior at this age. However, it is of interest that the girls reared by Nonconforming parents were actually less Dominant and Purposive than daughters of Authoritarian parents. If the aim of Nonconforming parents was to encourage independence and individuality by their methods of upbringing, they apparently failed to achieve this objective. As can be noted from Table 1 (Part B, A versus IV), Authoritarian parents compared with Nonconforming parents of boys differed significantly along the same dimensions of child rearing as did the respective parents of girls. Unlike daughters, sons of Authoritarian when compared with Nonconforming parents (Table 2, Part B, A versus IV) were significantly less Achievement Oriented and Independent. (Again, IQ differences cannot be ruled out as a source of these differences.) d. Authoritative versus Permissive. Authoritative parents of girls by comparison with Permissive parents (Table 1, Part A, B versus C) expected more participation in household chores, enriched the child's environment, and were more directive. They were firmer in enforcing directives, valued obedience more, were more likely to promote respect for established authority, and had more confidence in themselves as parents. They were more willing to express anger. The daughters of Authoritative parents (Table 2, Part A, B versus C) were significantly more Dominant and Purposive and somewhat more Independent than daughters of Permissive parents. Authoritative parents of boys differed significantly from Permissive 19

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY parents of boys (Table 1, Part B, B versus C) in the same way as the respective parents of girls. In addition, the Authoritative parents of boys discouraged infantile behavior and presented the child with a position which was clear and flexible — both to a greater degree than Permissive parents of boys. Sons of Authoritative families (Table 2, Part B, B versus C) were significantly less Hostile and Resistive and were more Achievement Oriented than sons of Permissive parents. Also, the sons of Authoritative parents were somewhat more Purposive than sons of Permissive parents. e. Authoritative versus Nonconforming. Authoritative parents of girls by comparison with Nonconforming parents (Table 1, Part A, B versus IV) were firmer in enforcing directives and were more likely to value obedience and to promote respect for established authority. Daughters of Authoritative parents (Table 2, Part A, B versus IV) were significantly more Domineering, Dominant, Purposive, and Independent than were daughters of Nonconforming parents. Authoritative parents of boys differed significantly from Nonconforming parents of boys (Table 1, Part B, B versus IV) in all the ways indicated for girls. In addition, Authoritative parents expected more participation in household chores and were more directive. The sons of Nonconforming parents were somewhat less Friendly than sons of Authoritative parents (Table 2, Part B, B versus IV), but their scores were near the mean for the total sample. The differences were due to the very low scores of the sons of Authoritative parents on the Hostile-Friendly cluster. /. Permissive versus Nonconforming. Permissive parents of girls by comparison with Nonconforming parents (Table 1, Part A, C versus IV) did not enrich their daughters' environment and were less willing to express anger. These differences between the two groups of parents both low in control were not related to differences in their daughters' behaviors. Permissive parents of boys, by comparison with Nonconforming parents (Table 1, Part B, C versus IV) did not enrich their sons' environment or discourage infantile behavior. The Permissive parents failed to present their sons with a flexible, clear position; lacked confidence in themselves as parents; and were less willing to express anger, although somewhat more punitive. The sons of Permissive parents, in comparison with sons of Nonconforming parents, were significantly less Achievement Oriented and Independent (Table 2, Part B, C versus IV). 20

DIANA B A U M R I N D EFFECTS OF AUTHORITARIAN, AUTHORITATIVE, PERMISSIVE, AND NONCONFORMING UPBRINGING

In both studies, the direction of the effects of Authoritative child rearing when compared with Permissive or Authoritarian upbringing on the development of competence denned by social responsibility and independence was generally positive. In Study I, boys and girls who were socially responsible (i.e., nondisruptive with adults and affiliative with peers) and independent (i.e., approach-oriented, self-reliant, and buoyant) had parents who were classified as Authoritative. In Study II, Authoritative parental behavior was associated with independent, purposive, dominant, and achievement-oriented behavior in girls and with all indices of social responsibility in boys. There was some indication that the extremely firm control of Authoritative parents, even when compared with that of Authoritarian parents, impaired the development of independence in otherwise highly competent boys — namely, sons of Authoritative parents did not have above average scores on Independence. In both studies, the differences were more marked between children reared authoritatively and those reared either permissively or in an authoritarian manner than were the differences between children reared in a permissive and those reared in an authoritarian manner. Permissive and Authoritarian parents differed in the control-nurturance ratio (with Permissive parents higher in nurturance and Authoritarian parents higher in control). But both differed significantly from Authoritative parents in that the Permissive and Authoritarian parents lacked confidence in their childrearing practices, did not enrich their children's environment, and, for boys, did not have a clearly defined child-rearing policy. Both Permissive and Authoritarian parents lacked balance between what was offered to the child in the way of support and what was demanded of him in terms of obedience. True, the imbalance was in opposite directions, but it was difficult to distinguish between the effects of these two kinds of imbalance, at least in the preschool years. Authoritative parents, by contrast, balanced what they offered with what they demanded. They balanced high control with high independence-granting, high standards for maturity with much support and nurturance. It will be of interest to see whether this early experience with reciprocity in human relations is associated at later time periods with accelerated attainment of post-conventional morality. The distinction made in Study II, but not in Study I, between Permissive and Nonconforming parents, both relatively noncontrolling, proved pro21

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ductive with boys. Sons of Nonconforming parents were significantly more achievement oriented and independent than were sons of Permissive parents. For boys, the firmer control, higher maturity demands, and clearer philosophy of the Nonconforming parents resulted in achievement orientation and independence. By contrast, Nonconforming parents produced daughters who were less rather than more achievement oriented and independent than daughters of Permissive parents, although not to a significant degree. While Nonconforming mothers of boys did not see themselves as victimized by society, these particular Nonconforming mothers of girls saw themselves as pawns of forces they could not control (attitudes expressed during interview about their sex roles). Because of their considerable reluctance to either punish or place high demands upon their daughters, it is possible that Nonconforming mothers did not provide a strong model or train their daughters in instrumental functions and that these were reasons for their daughters' relative incompetence.

Socialization Practices Associated with Dimensions of Competence in Preschool Girls and Boys The objective of Study III was to identify parent dimensions and variables associated with instrumental competence in preschool children. In Studies I and II, a typological approach to data reduction was used in order to reveal the relations between contrasting configurations of parent variables and child behaviors. That the relation with child behavior of a particular parent behavior depends upon the total configuration of variables has been demonstrated. For example, it was shown that the relation of high parental control to social responsibility and independence in the child depends upon the extent to which the parent also encourages individuality and independence; thus, the distinction between the effects on the child of Authoritarian versus Authoritative control. A correlational study can specify the general effect on the child of a theoretically interesting parent dimension. If there is no effect seen for a parent variable thought to have a universal linear effect, or if an effect is seen but the direction is opposite to that which would be postulated as universal by a currently popular child-rearing belief or theory, the validity of that belief or theory may properly be questioned. In Study III, subjects were 95 new sets of parents and their preschool children. Data were analyzed separately for boys and girls. Behavioral 22

DIANA BAUMRIND

and interview data on parents were derived first from home visits and are in the form of discrete HVSA variables as in Study I (only 9 of the 18 variables used in Study I were computed for Study III) and second from PARENT INTERVIEW DIMENSIONS arrived at through cluster-analytic techniques. Data on children were derived from prolonged observations in nursery school and structured settings and consist of g-sort ratings. A child-behavior model (similar in structure to models presented by Schaefer [1961] and Becker & Krug [1964]) was developed and its relation to these parent measures was assessed. The Q-sort clusters in Figure 1 to which they correspond will be listed in parentheses after the name. The names given the Q-sort clusters in the four-cluster solution described in Baumrind and Black (1967, p. 297) were Disafnliative-Affiliative (I), Resistive-Cooperative (II, III), Independent-Dependent (IV, VII), and Assertive-Withdrawn (V, VI). The following relationships (significant at <.05 unless otherwise indicated) between child-rearing dimensions and dimensions measuring competence in boys and girls were found (see Table 3): Table 3. Correlation Coefficients between Parent Variables and Child-Behavior Clusters (Four-Cluster Solution)

Parent Variables Father-interview clusters Warmth Consistent Discipline Strictness Concerning Orderliness Punitiveness Mother-interview clusters Warmth Consistent Discipline Maturity Demands Punitiveness Socialization Demands HVSA variables a. Positive Outcome c. Accepts Power Conflict with Child . . . f. Independence Training h. Respects Child's Decision j. Uses Reason to Obtain Compliance . .

23

PART A. GIRLS"

-.18 -.35 .09 .10

-.10 .10 .20

-.22

- .25 .27 .25 .04 .08 .24

-.26 -.12

-.12

-.22 -.21 -.07 -.05 -.22

- .04 -.32 .17 .01

.05

-.18

-.13 .05 .25 .34

.03 .01 .15 .09 .29 - .09

-.32 .15 .17

-.0

.02 .28 .08 .08 .24 .12 .09

-.05 .38

.16 .04 .32 .12 .30

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Table 3 — Continued

Parent Variables k. m q r.

Encourages Verbal Give and Take . . Satisfies Child Uses Coercive Power Takes Initiative in Control Sequences .

Father-interview clusters Warmth Consistent Discipline Strictness Concerning Orderliness Punitiveness Mother-interview clusters Warmth Consistent Discipline Maturity Demands Punitiveness Restrictiveness Encourages Independent Contacts HVSA variables a. Positive Outcome c. Accepts Power Conflict with Child. . . f. Independence Training Control . . . . h. Respects Child's Decision j. Uses Reason to Obtain Compliance . . k. Encourages Verbal Give and Take . . m. Satisfies Child q. Uses Coercive Power r. Takes Initiative in Control Sequences .

-.30 -.06 .12

-.05

-.29

-.04 -.11 .17

-.06 -.19 -.28 .28

.21 - .20

-.42 .19

PART B. BOYS"

-.08 -.25 -.21 .24

.07

.12 .06

.11 .30

-.13 -.02

-.11

-.09

.17

.12 .04 .34 .10

.13 .05 .40

-.10

-.05

.14 .06 .03

-.07

-.17

.18 .42

.20

.02

.01

-.13

-.09

-.29

-.08

-.18 -.07

-.06

.02 .20

.12

.26 .03

.02 .22 .01 .40 .16 .16 .22

-.02 -.32

-.05 -.17

.00 - .05

.21

.05

-.11 -.28 -.18 -.09

- .11

.42

-.11 .14 .36 .13 .01

-.27

- .09

.15

-.03 .03

SOURCE: Reprinted by permission from D. Baumrind and A. E. Black, "Socialization Practices Associated with Dimensions of Competence in Preschool Boys and Girls," Child Development, 1967, 38, 321, Table 11. a p<.05 for ;-^.30 for independent tests. b /?<.05 for ;-^.28 for independent tests.

The Effects of Maturity Demands for girls were measured by Father interview cluster designated Strictness Concerning Orderliness and the Mother interview clusters designated Maturity Demands and Socialization Demands. (The cluster designated Maturity Demands contains items measuring expectations that the young child will help with housework and become more self-sufficient; the cluster designated Socialization Demands contains items measuring control of child's aggression and demands for 24

DIANA BAUMRIND intellectual achievement.) For boys, the relevant parent interview clusters were the same except that a Mother cluster equivalent to Socialization Demands for girls did not appear. HVSA variable /, designated Independence Training, was the behavioral measure of Maturity Demands for both sexes. Maternal Socialization Demands and HVSA variable / were both associated with Assertive behavior in girls. Maternal Maturity Demands was associated with Independence and Assertiveness in boys. For both boys and girls, then, significant mother-child effects appeared between measures of Maturity Demands and Assertive behavior in the child. The Effects of Firm Control were measured for both boys and girls by Father and Mother interview clusters designated Consistent Discipline; HVSA variables a, Positive Outcome; c, Accepts Power Conflict; and r, Takes Initiative in Control Sequences. Paternal Consistent Discipline was associated in girls with Affiliative behavior toward peers. HVSA variable c (which measures the instances in which the parent confronts the child and precipitates conflict) was associated in girls with Cooperative behavior with adults but also with Dependent behavior. Paternal Consistent Discipline was associated in boys with Independent and Assertive behavior. HVSA variable c was associated positively (but insignificantly) with Independent behavior in boys rather than negatively as with girls. The Effects of Rational Methods of Discipline were measured by HVSA variables h, Respects Child's Decision; /, Uses Reason to Obtain Compliance; k, Encourages Verbal Give and Take; and q, Uses Coercive Power (negative). Assertive behavior in girls was associated positively with Use of Reason to Obtain Compliance (/') and negatively with the Use of Coercive Power (q). Girls whose parents Encourage Verbal Give and Take (k) were Affiliative. For boys, Respects Child's Decision (h) was associated with Affiliative, Cooperative, and Assertive behavior while Uses Reason to Obtain Compliance (/) was associated with Independent behavior. The Effects of Parental Warmth for both boys and girls were measured by Father and Mother interview clusters designated Warmth and by HVSA variable m, Satisfies Child. Warmth, as measured by these variables, was not a significant predictor of child behavior for either sex in this study. The Effects of Parental Punitiveness for both sexes were measured by Father and Mother interview clusters designated Punitiveness. That cluster for this sample contained items which measured willingness to express anger and use of negative sanctions such as corporal punishment and frightening the child. Punitiveness in the father correlated negatively with 25

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY measures of consistency and positively with low self-confidence, suggesting that the punitive father was weak rather than strong. Punitiveness in the father was associated positively with independent behavior in girls and not at all for boys. The Effects of Maternal Restrictiveness for Boys were measured by two interview clusters for mothers of boys which emerged in no other parentchild combination; these were designated Restrictiveness (measuring the extent to which the boy was prevented from exploring on his own) and Encourages Independent Contacts (a related, but somewhat different cluster measuring the extent to which the boy is actively introduced to new adults and new experiences). Both were related significantly to Independence in the son, the former negatively and the latter positively. To summarize the results of Study III: Warmth was not linearly related to indices of either social responsibility or independence in boys or girls. The belief that unconditional acceptance promotes competence defined as self-assertiveness and independence, or affiliativeness and cooperativeness, was not supported by these data. In fact, parental practices which were stimulating and to some extent tension-producing (e.g., maternal socialization and maturity demands, and paternal punitiveness for girls) were associated in the young child with assertiveness. Firm paternal discipline was associated for girls with socially responsible behavior and for boys with independence and self-assertiveness. Consistent discipline and high socialization demands were not characteristics of restrictive or punitive parents. In fact, the opposite was true. Restrictive, nonrational discipline (by contrast with consistent discipline and high maturity demands) was associated with withdrawn, dependent, and disaffiliative behavior in both boys and girls. The use of reason to obtain compliance was significantly associated in the parent with a complex of other parent variables similar to those which define authoritative discipline (i.e., high maturity demands, encourages independent contact, and lack of punitiveness), and this may contribute to its associations in the child with independent and socially responsible behavior. Nonauthoritative Patterns of Control Associated with Instrumental Competence In this section, nonauthoritative patterns of control associated with instrumental competence, particularly in girls, will be described. In Study II, two additional groups of parents were identified whose 26

DIANA BAUMRIND

daughters were unusually independent (neither group appeared in sufficient numbers among parents of boys to permit statistical analysis). The first pattern was designated Harmonious, and its members were found among highly advantaged white families of nursery school children, and also of adolescents in a separate pilot study. The second pattern emerged from lower-middle-class black families and can be described as quasiAuthoritarian. HARMONIOUS PATTERN

Although in the study proper, pattern membership was determined by multiple criteria, the eight families of preschool children placed in the Harmonious pattern had but one common identifying characteristic — the observer assigned to study the family would not rate the family for the construct designated Firm Enforcement. In each case, the observer stated that any rating on these items would be misleading since the parents almost never exercised control, but seemed to have control in the sense that the child generally took pains to intuit what the parent wanted and to do it. The atmosphere in these families was characterized by harmony, equanimity, and rationality. Harmonious parents were equalitarian in that they recognized differences based upon knowledge and personality, and tried to create an environment in which all family members could operate from the same vantage point, one in which the recognized differences in power did not put the child at a disadvantage. In their hierarchy of values honesty, harmony, justice, and rationality in human relations took precedence over power, achievement, control, and order, although they also saw the practical importance of the latter values. Parents were generally drawn from the highest educational levels and were either very well to do or, as was true of two families, had "dropped out" of that class. The effects of Harmonious child-rearing patterns on children appeared sex related. The six daughters of Harmonious parents were extraordinarily competent and very similar in their scores on the child behavior measures. Their average Stanford-Binet IQ was 136 (that of the entire sample was also high, 128). On clusters derived from scores on the Preschool Behavior Q sort, when compared with the girls in all other categories combined, these girls were achievement oriented (.05), friendly (.05), and independent (n.s.). These characteristics were not apparent for the two boys whose parents were classified as Harmonious (Baumrind, 1971b). In 1968, 103 tenth-grade students from a high school in Berkeley were 27

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY interviewed about their attitudes toward authority and toward relevant social issues, their feelings about their parents, their use of marijuana and the psychedelic agents, and their views on sexual mores. These interviews were rated on several hundred variables. Sixty-six variables with a response of 95 per cent or better were intercorrelated so that clusters of defining variables could be formed. An interesting adolescent profile analogous to the preschool girls from Harmonious families emerged. These subjects, compared with other subjects interviewed, scored significantly higher on clusters designated Concordance with Parents, Obeys Rules Willingly, and Denies Generation Gap, but scored low on Achievement Oriented. (The preschool girls scored high on a cluster also named Achievement Oriented. However, the cluster name, while identical, measures behavior in the preschool girls and in the adolescent girls measures attitudes toward achievement as expressed during an intensive interview. The behavior of the adolescents was in fact also achievement-oriented, as indicated by high grade-point averages and honors received.) These adolescents described their families as significantly less pathological and claimed significantly less sexual experience, considering chastity a virtue. They had significantly higher grade-point averages and slightly more honors than adolescents in the other groups. Proportionately, fewer of them applied to college, but, consistent with their excellent academic performance, all their applications were to universities rather than to junior colleges (ratio of applications to universities compared with junior colleges are significantly different from other groups) and they obtained higher scores on their Scholastic Aptitude Tests. This group of adolescents were extremely competent both socially and intellectually, but unlike their wealthy professional parents were predominantly interested in pursuing artistic and nonintellectual vocations. They felt, however, that their parents approved of their life style and future vocational interests. It would appear, then, that Harmonious upbringing is associated in preschool girls with intellectual competence and with friendliness, although not necessarily with motivation to excel or. aggressiveness as in daughters of Authoritative parents. It will be of interest to see whether the preschool girls from Harmonious families will develop characteristics and vocational interests similar to the "Harmonious" adolescents just described. BLACK AUTHORITARIAN PATTERN

In Study II, the data for the sixteen black children and their families were analyzed separately from the data for white families since it was 28

DIANA B A U M R I N D thought that the effects of a given pattern of parental variables might depend on the social context in which the family operates. Five of the nine black girls in this study came from families who met the criteria for Authoritarian parents. Thus, it was possible to contrast the effects of authoritarian upbringing on black and on white girls. In Table 4, it can be seen that black daughters of Authoritarian parents, when compared with white girls, were significantly more Domineering and Independent, and somewhat more Resistive and Dominant. These differences are striking because there were no significant black-white differences in child behavior in the total sample. The socialization practices which characterize black families are authoritarian by white standards, and as such would be regarded as changeworthy by many child-rearing experts. The preliminary findings reported here suggest the opposite possibility — that the child-rearing practices which characterize Authoritarian black families benefit the social development of their daughters. (It is important to note, however, that Authoritarian child-rearing practices were not associated with intellectual achievement in young children, in either black or white homes.) Obedience to authority seen as justified by the child and not accompanied by parental rejection should not produce passivity or repressed hostility in the child and indeed did not appear to do so in the black girls obTable 4. Comparison of <2-Sort Clusters for Daughters of Authoritarian Parents, Black versus White Blacka Group

Mean

54.8 I. Hostile-Friendly II. Resistive-Cooperative . 59.3 III. DomineeringTractable 58.2 IV. Dominant-Submissive . 57.5 V. Purposive-Aimless . . . . 54.7 VI. Achievement OrientedNot Achievement Ori48.2 ented VII. Independent54.9 Suggestible

White"

Significant Differences, SD B lack vs. White

SD

Mean

7.6 11.8

46.8 47.6

10.5 11.8

7.7 4.9 6.6

46.0 48.6 49.1

11.2 9.2 9.2

9.3

47.6

9.9

4.8

44.8

8.7

B>W* B>W** B>W*

B>W**

SOURCE: Reprinted by permission from D. Baumrind, "An Exploratory Study of Socialization Effects on Black Children: Some Black-White Comparisons," Child Development, 1972, 43, 264 (Table 1). b *N = 5. N=10. *p<.10. **p<.05.

29

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY served. In this study, the black daughters of Authoritarian parents were exceptionally independent and at ease, especially in view of the novel, perhaps alien, nursery school setting in which they were functioning. The association of authoritarian practices with independent, aggressive behavior in the five black compared with the ten white girls observed may be understood by viewing authoritarian practices as consistent with the social norms of middle-class black families but at variance with the social norms of middle-class white families in Berkeley. Within each ethnic group, the social norms should reflect the actual conditions needed for optimum survival in that cultural setting. In the probable event that strict parental control requiring immediate obedience furthers the black girl's survival and is viewed as having that effect within her own culture, it is likely that these authoritarian practices will be experienced by the child as supportive and reassuring. These practices may reflect caretaking and nurturant attitudes in the parent by contrast with the dogmatic attitudes motivated by repressed anger and emotional coldness which authoritarian practices may reflect in upper middle-class white homes where such practices are neither necessary to the child's survival nor normative.

Discussion and Generalizations from This Program of Research The following issues will be considered in this section: (a) the effects of firm control; (b) the effects of passive-acceptance and high warmth; (c) the effects of punishment; (d) the effects of restrictive control and authoritarian practices; and (e) the development of independence in girls. The section will conclude with a brief interpretive commentary. a. The effects of firm control. Firm control is measured in the three studies discussed by attitudes supporting the intentional use of contingent reinforcement to shape the child's behavior and by actions which firmly oppose the child's intentional disobedience or his coercive demands upon the parent. Where firm control was part of a pattern of restrictiveness and arbitrary rule, the term authoritarian (or restrictive) control was used. Where firm control was part of a pattern in which independence and individuality were also encouraged, the term authoritative (or firm) control was used. The main thrust of the results of the three studies is that, in the preschool years, authoritative control, by comparison with authoritarian control or permissive noncontrol, is associated with social responsibility 30

DIANA BAUMRIND

(achievement orientation, friendliness toward peers, and cooperativeness toward adults) and independence (social dominance, nonconforming behavior, and purposiveness) whereas authoritarian control is not associated with social responsibility (its stated aim in the authoritarian middleclass white families studied) and permissive noncontrol is not associated with independence (its stated aim). Several qualifications to this generalization should be mentioned: (a) in one study (Baumrind, 1971), sons of Authoritative parents (who were exceptionally high on the measures of firm control) were not above average in independence; ( b ) in another study (Baumrind & Black, 1967), HVSA variable c, which measures a heavy-handed and effective use of power, was associated with dependence in girls (although not in boys); (c) the daughters (but not the sons) of Authoritarian black families were highly assertive and independent. Very high control, when at variance with the social norms (as in [a] and [b] above) may indeed interfere with the development of independence in young children. There are two converging lines of evidence which could explain why in the young child obedience can be achieved by firm, rational control rather easily and generally does not inhibit independence or self-assertion in the child. First, it seems likely that obedience in young children is consistent with their nature and facilitates their physical well-being. Second, at this very early stage of moral development, authority in the minds of children is legitimated by power, so that the self-esteem or sense of autonomy of the young child, unlike the adolescent, is not threatened by obeying a powerful authority. Obedience to the norms of the species facilitates the survival of the individual of that species. Obedience has its earliest expression in attachment behavior. Like other baby animals, baby humans show a propensity for attachment to caretakers, accompanied by compliance with maternal signals. Since infant animals, including infant humans, can physically explore their environment before they can understand the hazards attendant upon such exploration, as soon as the infant can locomote, it becomes advantageous to its survival that it comply with maternal signals. As Stayton, Hogan, and Ainsworth (1971) have shown, the great majority of maternal commands are in fact obeyed by young infants and, as they suggest, human infants appear to have an inborn propensity to comprehend and comply with a complex of tone, voice, gesture, and facial expression. This propensity of human infants to comply with the demands of adults 31

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY whom they trust to protect them is shared by all primate young (Bleibtreu, 1969, p. 199). While the social structure of man, unlike that of other animals, is not a result primarily of instinctual forces, it must be emphasized that social structures not based on a hierarchy of authority would be truly novel among primates. Despite the fact that most preschool children (as well as toddlers) comply with most parental directives, findings from this program of research indicate the importance of the control dimension as a predictor of competence in early childhood. What the parent does on those minority of occasions when the child does not comply seems to affect in important ways the child's ability to adapt and his motivation to master his environment. It is consistent with the work of Piaget (1932) and Kohlberg (1969) that the toddler and preschool child readily accept as a legitimate authority the person who has power to contingently reinforce his actions as well as to provide for his needs, and believe that the standards which that authority can enforce are "right." The use of punishment and reward serves to establish the potency of the reinforcing agent. The use of contingent reinforcement makes a clear statement to the child that rules are there to be followed and that to disobey is to break a known rule. Punishment itself signifies to the young child that an act is "wrong" and that he is "bad" when he commits that act so that children of this age demonstrate anxiety reducing maneuvers in reaction to committed transgressions which previously have been punished (Aronfreed, 1968). The moral reasoning of the young child is qualitatively different from that used by most adolescents or adults. Behavior which has been punished, especially by a loved adult, is experienced by the child as wrong precisely because it has been punished. In their early years, children cannot be taught rules by processes which depend upon the ability to grasp a rationale. The adult cannot expect the child to understand a justification based on post-conventional reasoning, nor can he attribute post-conventional (that is, levels 5 and 6) morality to the young child who objects to parental discipline. (However, by providing the young child with reasons for rules, the parent should sensitize the child to considerations of convention and principle. This may, in turn, facilitate the development in the child of later ability to distinguish between instances of legitimate and illegitimate authority and to take the role of the other at times when altruism is called for.) An investigator who views obedience as a habit or an invariant trait might predict that an obedient preschooler will (if exposed to consistent 32

DIANA BAUMRIND firm control) become increasingly compliant as he grows older. However, a developmental psychologist might reason differently. Parent behavior which appears similar will have different meaning to the child at different ages, and even if exposed to the same stimuli, the child as he matures will interpret these stimuli differently. Actions motivated by self-interest or conformity at early stages can be motivated by consideration at a later stage of development. For example, an act of kindness is often seen as kinder by young children if done as conformity to obligation rather than from desire to be helpful. As the Baldwins conclude in their study of compassion: ". . . it appears that younger children may fail to discriminate between the desire to benefit another person and conformity to a moral obligation, but that is not the whole story. Some children — perhaps all children at some age — say it is kinder when one conforms to the obligation and benefits the other person. They answer opposite to adults and justify their response by the opposite reasoning of adults" (Baldwin, 1969, p. 342). If compliance is a sign of maturity and competence at ages 3 and 4, and independence and internalized conscience are signs of maturity and competence at later stages of development, then perhaps compliance at this early age should predict independence and internalized conscience (rather than slavish conformity) at later stages. In any case, there is little empirical evidence for boys (although there is some for girls) that ability to conform is associated with inability to make independent decisions or to take autonomous actions. At ages 3 and 4 (in Studies I and II) there were groups of children who were both outstandingly obedient and outstandingly independent (e.g., Pattern I children in Study I and Authoritativeconforming and Harmonious girls in Study II). Whether children high both in measures of obedience and independence will be paired with Authoritative parents at later stages is now being investigated. b. The effects of passive-acceptance and high warmth. In none of the three studies reported here was warmth by itself a significant predictor of child behavior, although in both studies based upon configurational analyses, parents of the most competent children were warm as well as controlling. There are two conditions which may account for the unimportance of parental warmth as a predictor of child behavior in this study. First, the range of parents studied was restricted as to warmth. Second, warmth, punitiveness, and coerciveness were measured by separate variables. In most correlational studies (e.g., Becker & Krug, 1964), the warmth factor 33

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY tends to be a rather global construct including such diverse variables as use of reasoning, success of enforcement policy, and nonpunitive attitudes. When variables measuring warmth are limited to expression of compassion and tenderness, associations with indices of competence in the child appear to be curvilinear rather than positive. A certain amount of warmth is necessary, but if there is much more, its effects become negative, perhaps because very high warmth is often accompanied (although it need not be) by lax control and overprotection from stress and frustration. Passive-acceptant and overprotective parental practices were associated with dependence in the children studied. In Studies I and II, parents whose children were independent were rated low on measures of passive-acceptance and high on measures of maturity demands. In Study III, paternal punitiveness was associated positively with indices of independence in girls. These findings concerning the negative effects on children (girls in particular) of a high degree of passive-acceptance are supported by observations of other investigators. Thus, in a recent progress report to the National Institute of Mental Health, Crandall notes: Perhaps internality at later developmental stages is best predicted by some degree of maternal "coolness," criticality, and stress, so that offspring were not allowed to rely on overly indulgent affective relationships with parents but were forced to learn objective cause-effect contingencies, adjust to them, and recognize their own instrumentality in causing those outcomes. . . . The data further suggested that internals' mothers had allowed room for, and provided opportunity for, more of this contingency learning. They participated more frequently with their children in achievement activities, but were less protective, had lower intensity and frequency of contact with them, and rejected their dependency overtures. In fact, it is possible that this cluster of behaviors, together with the criticality and lack of affection cited above may have served to thrust the child "out of the nest" and into more active contact with his physical and social world. (Crandall, 1972, pp. 38-39.) Similarly, Rosen and D'Andrade (1959) found that high achievement motivation was facilitated by high maternal hositility and rejection when the child was displeasing to the parent as well as by high maternal warmth when the child pleased the parent. Hoffman, Rosen, and Lippitt (1960) found that mothers of achieving boys were more coercive than those who performed poorly. Kagan and Moss (1962) reported that achieving adult women had mothers who in early childhood were unaffectionate, "pushy," and not overprotective. Probably the parent, by being self-assertive and 34

DIANA B A U M R I N D self-confident, provides a model of similar behavior for the child. Thus, Norman (1966) found that the same-sex parents of gifted achieving children had significantly lower conformity scores and higher independence scores than did the same-sex parents of gifted children who were not achieving. c. Punishment, its side effects and effectiveness. Most parents would in theory prefer to use reward rather than punishment. However, in practice, parents in these studies frequently used some form of punishment (often corporal) to achieve compliance. Is punishment ineffective and harmful? The data in the three studies reported above which bear upon these questions are as follows: Study I : Parents of mature children (Pattern I) used more positive and negative reinforcement than parents of other children. Their ratio of positive to negative reinforcement was significantly higher than that of parents of dysphoric and disaffiliated children (Pattern II). These parents of mature children used corporal punishment more frequently than parents of immature children (Pattern III), but they used withdrawal of love and ridicule less frequently (the mothers to a significant degree). Compared with mothers of dysphoric and disaffiliated children, mothers of mature children resorted less frequently to frightening the child as a means of behavior control. Study II: All but 2 of the 150 families stated that they used corporal punishment. Authoritative parents expressed anger easily when the child disobeyed but were not rated as rejecting or punitive. Permissive parents, by contrast, were, according to ratings, not willing to express anger when the child disobeyed, but were rated as more punitive in attitude. Moreover, Permissive parents (by contrast with Authoritative or Harmonious families) admitted during the interview to explosive attacks of rage in which they inflicted more pain or injury upon the child than they had intended. In the self-report measure (Parent Attitude Inquiry), an empirical cluster (designated Angered over Loss of Control) appeared for both mothers and fathers, defined by items measuring both frequency of anger and inability to control the child. Permissive fathers of boys had scores on this cluster significantly higher than those of other parents. Permissive parents apparently became violent because they felt that they could neither control the child's behavior nor tolerate its effect upon themselves. Study HI: Punitiveness in fathers was correlated positively with inde35

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY pendence in girls. Punitiveness (or lack of it) in neither parent was correlated significantly with independence or social responsibility in boys. The evidence from the above studies does not indicate that negative reinforcement or corporal punishment per se were harmful or ineffective procedures, but rather that the total pattern of parental control determined the effects on the child of these procedures. Punishment of many kinds was in fact used by effective parents (Authoritative parents, Nonconforming parents of boys, Authoritarian parents of black girls) as well as by less effective parents (Authoritarian, Permissive) in the studies reported above. However, Authoritative parents, who were particularly effective, favored corporal punishment over other negative sanctions. Many studies demonstrate that brutal, arbitrary punishment (by contrast with nonbrutal, contingent punishment) is associated with various kinds of undesirable behavior in children. It is associated with antisocial aggression (e.g., Glueck & Glueck, 1950; Hetherington, Stouwie, & Ridberg, 1971) and also with passivity, dependence, and social withdrawal (Becker et al., 1962; Kagan & Moss, 1962). There is considerable evidence, however, that the use of punishment and coercion by parents is associated reliably with antisocial and non-achievement-oriented behavior in children when the parent is also repressive, hostile, and restrictive, but not reliably when he is not. Thus, in one study of 211 third-graders' attitudes (Hoffman et al., 1960), the children who described their parents as coercive but also permissive of high autonomy, compared with the remainder of the sample, were higher in academic success, use of directives, successful influence of peers, group leadership, friendliness, and also conscious experience of hostility. They were striving and aggressive but not rebellious. Sears (1961) found that the antecedents at age 12 of prosocial aggression scores, in maternal interview data obtained when the child was age 5, were high permissiveness for aggression and high punishment. In the Sears study, punishment for aggression appeared to reduce antisocial but increase prosocial aggression, indicating once again that parental authority which is not accompanied by repressive, arbitrary punitiveness may stimulate non-hostile self-assertiveness. In Studies I and II, Authoritative parents whose children demonstrated to an unusual extent prosocial but not antisocial aggression resorted to corporal punishment as well as positive sanctions (using each contingent upon the child's behavior). In this instance, as in the Sears study, use of (in this case corporal) punishment was not associated with hostile behavior in the child. 36

DIANA B A U M R I N D It is reasonable to postulate that nonbrutal punishment, including physical expressions, by a loved and respected parent, if consistently contingent upon the child's behavior, should have, in addition to its effectiveness as a means of behavior control, side effects which are also beneficial, such as the following: (a) more rapid re-establishment of affectional involvement between participants following emotional release; ( b ) high resistance to similar deviation by siblings who vicariously experience punishment; (c) lessened guilt reactions to transgression since the parent inflicts an unpleasant consequence; (d) increased ability of the child to endure punishment in the service of a desired end if he should decide to persist; (f) willingness by the child to openly confront another within a power relationship rather than to deny or disguise anger, brought about by emulation of the aggressive parent; (g) increased feeling of internal control within defined limits since reinforcement is made consistently contingent upon the child's own action; (/z) reduced dependency upon the parent as a source of gratification since the parent, by punishing the child, arouses ambivalent feelings toward himself. The proposition that punishment is an ineffective means of controlling human behavior is probably a "legend," as shown by the work of Solomon (1964); Walters, Parke, and Cane (1965); and Azrin and Holz (1966). Under conditions prevailing in the home setting, punishment may be quite effective in helping to accomplish particular objectives, as it appears to be for Authoritative parents in the studies cited. Aronfreed (1968) points out: The power and extent of aversive behavior-contingent learning appear to have been badly underestimated in some conceptions of the control of conduct (see, for example, Skinner, 1953, Ch. 12). . . . This lack of attention to punishment may have reflected widely held assumptions about its "traumatic" emotional impact on the child; it may also have been influenced by certain social and educational philosophies which implied that punishment was undesirable and not required where the informed socializing agent could elicit the child's natural inclinations toward social growth through affection and permissiveness. Some early observations of punishment learning in animals did suggest that the behavioral effects of punishment were sometimes neither very marked nor durable (Estes, 1944; Skinner, 1938; Thorndike, 1932). . . . However, these studies characteristically examined the efficacy of punishment in eliminating behavior that was instrumental to the reduction of a very strong motivational state, in situations where alternative forms of relevant behavior were not available to the animal. In some of the studies, the animals were placed in essential37

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ly insoluble learning situations. The paradigms which were used in these studies did not, therefore, represent the conditions under which much of human socialization typically takes place. (Aronfreed, 1968, pp. 61-62.) Aronfreed then goes on to cite literally scores of studies in which punishment is shown to modify behavior in the desired direction. Parents not opposed in principle to the use of aversive stimuli may be taught how to use punishment effectively and humanely rather than be advised incorrectly that punishment is intrinsically harmful to the child and ineffective. Specifically, (a) timing punishment so that it occurs as closely as possible to the changeworthy response increases the long-range effectiveness of punishment as a deterrent (Aronfreed & Reber, 1965; Walters, Parke, & Cane, 1965); (b) consistent punishment in which no avenue of escape from contact with the stimulus is permitted maximizes the effectiveness of punishment procedures (Azrin & Holz, 1966; Parke & Deur, 1972); (c) as Parke's later work has shown, punishment accompanied by an explanation in which both the changeworthy act and a more acceptable act are specified is more effective than punishment alone; (d) consistent with Aronfreed's observations, parents should be cautioned not to use punishment to eliminate behavior which the child is highly motivated to perform without offering alternative forms of behavior by which the drive can be redirected. d. The effects of restrictive control and authoritarian practices. In Study I, children who were notably responsible, assertive, and self-reliant were paired with parents whose disciplinary interventions were firm but not restrictive. These Authoritative parents gave considerable leeway to the child to make his own decisions and accompanied their directives with a reason. However, they also exerted very firm control. In Study II, control exerted by Authoritative parents of boys was more restrictive than control exerted by Authoritative parents of girls. The sons of Authoritative parents in that study did not show the very high levels of independence demonstrated by the daughters of Authoritative parents. In Study III, parents' willingness to offer justification for their directives and to listen to the child were associated with indices of independence and social responsibility in both boys and girls. Parental restrictiveness and refusal to grant sufficient independence were associated in boys with dependent and passive behavior. It is important to distinguish between the effects on the child of firm control and of restrictive control (in which extensive proscriptions and 38

DIANA BAUMRIND prescriptions cover many areas of the child's life and need systems, and arbitrary limits are placed upon his autonomous strivings to try out new skills and make decisions for himself). In general, as Becker (1964) indicates, restrictive discipline does appear to lead to "fearful, dependent and submissive behaviors, a dulling of intellectual strivings and inhibited hostility." Whether the child perceives the ways in which parental authority is expressed to be arbitrary and overprotective or instead justified in terms of his own welfare may determine how well that authority is accepted and whether behavioral compliance is accompanied by immaturity and rebellion, or by independence in the child. Black families of girls (Baumrind, 1972) who (by criteria set for white families) exerted authoritarian control contained daughters who were independent and self-assertive. But restrictive control in these black families was accompanied neither by rejection nor by overprotection as was true for the white families who exerted similar control, and the girls neither rebelled at home nor were submissive or passive in the nursery school setting. The contrasting effects of authority viewed as justified versus authority viewed as illegitimate should become particularly apparent at adolescence. Thus, Pikas (1961), in his survey of 656 Swedish adolescents, showed that significant differences occurred in their acceptance of parental authority, depending on the reason for the directive. Authority which was based on rational concern for the child's welfare was accepted well by the child, whereas authority which was based on the adult's desire to dominate or exploit the child was rejected. The former, which Pikas calls rational authority, is similar to what we have designated authoritative or firm control and the latter, which he calls inhibiting authority, is similar to what we have designated authoritarian or restrictive control. His results are supported by Middleton and Snell (1963), who found that parental discipline regarded by the child as either very strict or very permissive was associated with lack of closeness between parent and child and with rebellion against the parent's political viewpoints. In an interesting study of social norms and authoritarianism, Kagitcibasi (1970) showed that her sub jects in the United States who scored high on content areas of authoritarianism (i.e., respect for authority and high value placed on obedience) were more likely to suffer from the authoritarian personality syndrome (i.e, dogmatic and intolerant attitudes, motivated 'by repressed anger, emotional coldness, and a sense of impotence) than their Turkish counterparts. It is supportive of our previous distinction between the personality 39

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY concomitants of normative versus non-normative authoritarian practices that she notes: "Obedience to justified authority is a basic code of decency and morality in Turkey, and a valued historical tradition. However, this obedience, coming from social norms, seems to carry different overtones than the blind obedience and submission, mixed with repressed hostility, that the authoritarian personality postulates on a psychodynamic basis" (Kagitcibasi, 1970, p. 445). e. The development of independence in girls. Resistiveness to adult authority is not related to low achievement in girls, although it may be for boys. It was found that girls in the preschool years were rated as less hostile and resistive than boys, but that resistiveness toward adults, while highly correlated negatively with achievement oriented behavior for boys (—.40), was not at all for girls (—.06). Moreover, for girls as compared with boys, domineering behavior toward peers was related to dominant and purposive behavior and less related to hostile behavior. It may be postulated that resistiveness to adult authority and a domineering attitude toward peers will facilitate the development of independence and achievement at later ages in girls, but may not in boys, perhaps because "high" resistiveness and hostility in boys compared with girls is sufficiently extreme to disrupt social relations. Dominance and stridency may signify in girls that they are resisting sex role pressure to undercompete and underachieve. The positive association shown in these studies between independent behavior for girls and paternal punitive attitudes supports Bronfenbrenner's observation (1961) that among educationally advantaged subgroups too much warmth and support seem to have a "debilitating" effect on girls. Disciplinary techniques which foster self-reliance, whether by placing demands on the girl for high-level performance or by encouraging her to take independent action seem to facilitate assertive, independent behavior in girls. In attempting to understand the "debilitating" effects of passive-acceptance by parents on girls, the hypothesis (advanced by Wolpe [1958] among others) deserves consideration that avoidance and self-assertion are reciprocally inhibiting responses to threatening or frustrating experiences. The aspect of the fundamental fight-flight reaction to such experiences that will predominate for a given individual will be a function largely of prior experience. Vigorous, abrasive interaction in which assertive responses are stimulated and either not punished or rewarded should increase the likelihood that the individual so stimulated will react 40

DIANA BAUMRIND assertively rather than with avoidance to threatening or frustrating stimuli. Daughters reared by black Authoritarian, Harmonious, and Authoritative parents were outstandingly assertive and independent. Both the Authoritative and the Harmonious parents of these girls were self-confident, firm as well as flexible in their child-rearing attitudes, and willing to express angry feelings openly. They avoided the use of guilt-producing techniques of discipline. Authoritarian parents of black girls also demonstrated willingness to express positive and negative feelings openly and a forthright, brisk treatment of their daughters. By being demanding and not overly protective, Authoritarian parents of black girls and Authoritative parents permitted their daughters to extricate themselves from stressful situations and encouraged them to place a high value on tolerance of frustration and courage. The passive-acceptant and overprotective attitudes of many parents of girls when compared with boys may explain in part why girls and women are found to be more affiliative, feel dependent, suggestible, and obedient than men (Janis & Field, 1959; Walker & Heyns, 1962; Witkin et al., 1954). In summarizing some of Douvan's findings (with Adelson) on the adolescent experience, Bardwick and Douvan observed: "The overwhelming majority of adolescent girls remain dependent upon others for feelings of affirmation. Unless in early life the girl exhibited the activity, aggression, or sexuality usually displayed by boys, and thereby experienced significant parental prohibitions, there is little likelihood that she will develop independent sources of esteem that refer back to herself. Instead, the loss of love remains for her the gravest source of injury to the self and, predictably, she will not gamble with that critical source of esteem" (Bardwick & Douvan, 1971, p. 151). Parental values which stress individuality, self-expression, initiative, divergent thinking, and aggressiveness appear to facilitate the development of independence in the girl, provided that these qualities in the parent are not accompanied by lax and inconsistent discipline and unwillingness to make demands upon the child. Authoritative and Harmonious parents, in addition to their relatively high maturity demands, encouraged their daughters to ask for, even to demand, what they desired. They themselves acquiesced in the face of such demands provided that the demands were not at variance with parental policy. Thus, their children were positively reinforced for autonomous self-expression. In contrast to these results, the Authoritarian parents of white (but not black) girls did not value willful41

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ness or independence in their daughters and therefore did nothing to promote such behavior, and the Permissive parents did not contingently reward mature or praiseworthy demands by the child and ignore or punish regressive or deviant demands. COMMENTARY ON PARENTS' CONCEPTIONS OF THE CHILD

Most parents could articulate to some extent during their interviews an image of man, a theory of how children learn, a theory of knowledge, and so on. Generally speaking, Authoritarian and Permissive parents articulated an image of what their child was like which was not realistic or flexibly responsive to the developing competencies in the child. Both Authoritarian and Permissive parents saw their child as dominated by egoistic and primitive forces. Authoritarian parents saw these characteristics as in need of constraint, and Permissive parents (many citing Rousseau) tended to glorify these same characteristics. Few parents from either pattern took very much into account, on the one hand, the child's stage-appropriate desire to be good and to conform to parental expectations nor, on the other hand, the child's impulsivity and use of concrete reasoning which often interfered with his efforts to behave maturely. Thus, these parents appeared to construct a fiction about what their child was like and to relate to that fiction. Authoritative and Harmonious parents in the studies discussed were more inclined to see the rights and duties of parents and young children as complementary rather than identical. They tended to believe that as parents they should be receptive to and aware of the child's needs and views before making any attempt to alter the child's actions and to see the child as maturing through stages with qualitatively different features. However, they did not describe this maturational process as an automatic unfolding but rather as subject to .modifications resulting from interactions occurring between the child and socializing agents. Authoritative and Harmonious parents by comparison with other parents interviewed made most frequent reference to the stage the child was at to sustain their current practices. These parents were particularly aware that children's concepts of right and wrong and their ability to handle absence of restraint changed with age. They were, relative to other parents, free of ideologies which would deter them from changing as the child changed. They tended to look upon the control aspects of their parental role as facilitative of competence, and therefore independence, in the child. While Authoritarian parents tended to view children as having respon42

DIANA BAUMRIND

sibilities similar to those of adults, and Permissive parents tended to view children as having rights similar to those of adults, Authoritative and Harmonious parents saw the balance between the responsibilities and rights of parents and the responsibilities and rights of children as a changing function of stage of development. Authoritative parents in their interviews tended to refer to the norm of reciprocity in Judeo-Christian terms, "Do unto others as you would that they do unto you," while Harmonious parents referred to the norm of reciprocity within the metaphor of Karmic consequences. But in each case the relationship between parent and child was viewed as a manifestation of this law and the law itself as an important principle to teach the child. Both exploitation and indulgence of the child were seen as a violation of this law and as interfering with the development in the child of a firm understanding of the norm of reciprocity and the willingness to abide by its dictates.

Future Directions The study in progress has as subjects eight- and nine-year-olds who, it is hoped, will be seen through adolescence. In the early years, parental functions center around providing the child with a nourishing environment and shaping his behavior and early attitudes. The functions which define the parental role in late childhood and adolescence are different, and may include enhancing meaning — for example, maintaining in the child a high level of motivation, sustaining the belief that effort is worthwhile and will bear fruit, teaching the child how to differentiate between trivial and fundamental goals; testing reality — for example, showing the child how to exert power successfully, to decide which limits to test and which to accept, and to differentiate between visionary ideals which can be reached for but never grasped and hard-to-achieve goals which with great effort can be grasped; and constructing a model of reality — for example, helping the child to develop an image of man, a theory of knowledge, and a workable reality. These parental functions are more difficult to define operationally than the functions designated by the terms nurturance and control, but they are very real all the same and deserve to be investigated. As the child matures, new abilities and attributes emerge, both in the child and in the parent-child relationship. Because subjects in this investigation are older, we have included measures of moral judgment, internalexternal locus of control, field independence-dependence, creativity, roleplaying ability, processes by which mother teaches the child stage-related 43

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY cognitive tasks, and conflict-resolution among family members whose task is to draw conclusions concerning moral judgment. These measures will permit investigation of: the relation of parents' level of moral reasoning to indices in their children of social sensitivity and level of moral reasoning, the development in children of sex-role typing as a function of parents' values concerning appropriate sex role behavior and of their level of moral reasoning, and the effect of parent's locus of control on whether the child perceives himself as originator or pawn. The extent to which trait measures exhibit structural continuity across time will be investigated, as will the stability of the individual's score on traits whose structure remains relatively invariant. Cause-and-effect relations hypothesized to exist as a result of findings from this program of research can be tested most meaningfully in an intervention study. An attempt should be made to assist parents to achieve their objectives by appropriate child-rearing practices. Through parent consultation, these findings have been applied with apparent success in family settings. In the future, efforts will be made to design an intervention study in which these findings are systematically applied to parent consultation, and the results evaluated. References Aronfreed, F. Conduct and conscience. New York: Academic Press, 1968. & Reber, A. Internalized behavioral suppression and the timing of social punishment. Journal of Personality and Social Psychology, 1965, 1, 3-16. Azrin, N. H., & Holz, W. C. Punishment. In W. K. Honig (Ed.), Operant behavior, New York: Appleton, 1966. Baldwin, A. L. A cognitive theory of socialization. In D. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally, 1969. Kalhorn, I., & Breese, F. H. The appraisal of parent behavior. Psychological Monographs, 1949, 63 (4, Whole No. 299). Bardwick, I. M., & Douvan, E. Ambivalence: The socialization of women. In V. Gornick & B. K. Moran (Eds.), Women in sexist society. New York: Basic Books, 1971. Baumrind, D. Effects of authoritative parental control on child behavior. Child Development. 1966. 37. 887-907. Child care practices anteceding three patterns of preschool behavior. Genetic Psychology Monographs, 1967, 75, 43-88. Authoritarian versus authoritative parental control. Adolescence, 1968, 3, 255-272. Current patterns of parental authority. Development Psychology Monograph, 1971,4, No. l,Pt.2. (a) Harmonious parents and their preschool children. Developmental Psychology, 1971,4,99-102. (b) From each according to her ability. School Review, 1972, 80, 161-197. 44

DIANA B A U M R I N D & Black, A. E. Socialization practices associated with dimensions of competence in preschool boys and girls. Child Development, 1967, 38, 291-327. Becker, W. C. Consequences of different kinds of parental discipline. In M. L. Hoffman & L. W. Hoffman (Eds.), Review of child development research. Vol. 1. New York: Russell Sage Foundation, 1964. & Krug, R. S. A circumplex model for social behavior in children. Child Development, 1964, 35, 371-396. Peterson, D. C., Luria, Z., Shoemaker, D. L, & Hellmer, L. A. Relations of factors derived from parent-interview ratings to behavior problems of five-yearolds. Child Development, 1962, 33, 509-535. Bleibtreu, J. N. The parable of the beast. New York: Collier, 1969. Bronfenbrenner, U. Two worlds of childhood. New York: Russell Sage, 1970. Crandall, V. Progress report, NIMH Grant No. MH-02238. Pels Institute, Yellow Springs, Ohio, 1972. Estes, W. K. An experimental study of punishment. Psychological Monographs, 1944,57 (3, Whole No. 263). Glueck, S., & Glueck, E. Unraveling juvenile delinquency. New York: Commonwealth Fund, 1950. Goodman, P. Compulsory mis-education. New York: Horizon, 1964. Hetherington, E. M., Stouwie, R. J., & Ridberg, E. H. Patterns of family interaction and child-bearing attitudes related to three dimensions of juvenile delinquency. Journal of Abnormal Psychology, 1971, 78, 160-176. Hoffman, L., Rosen, S., & Lippitt, R. Parental coerciveness, child autonomy, and child's role at school. Sociometry, 1960, 23, 15-22. Janis, I. L., & Field, P. B. Sex differences and personality factors related to persuadability. In I. L. Janis et al. (Eds.), Personality and persuadability. New Haven: Yale University Press, 1959. Kagan, J., & Moss, H. A. Birth to maturity: A study in psychological development. New York: Wiley, 1962. Kagitcibasi, C. Social norms and authoritarianism: A Turkish-American comparison. Journal of Personality and Social Psychology, 1970, 16, 444-451. Kohlberg, L. Stage and sequence: The cognitive-developmental approach to socialization. In D. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally, 1969. Maslow, A. H. Motivation and personality. New York: Harper, 1954. Middleton, R., & Snell, P. Political expression of adolescent rebellion. American Journal of Sociology, 1963, 68, 527-535. Neill, A. S. Summerhill. New York: Hart, 1964. Norman, R. D. Interpersonal values of parents of achieving and nonachieving gifted children. Journal of Psychology, 1966, 64, 49-57. Parke, R. D., & Deur, J. L. Punishment and inhibition of aggression in children. Developmental Psychology, 1972, 7, 266-269. Piaget, J. The moral judgement of the child. London: Routledge, 1932. Pikas, A. Children's attitudes toward rational versus inhibiting parental authority. Journal of Abnormal Social Psychology, 1961, 62, 315-321. Rogers, C. R. A therapist's view of personal goals. Pendle Hill Pamphlet 108. Wallingford, Pa.: Pendle Hill, 1960. Rosen, B. C., & D'Andrade, R. The psychological origins of achievement motivation. Sociometry, 1959,22, 185-218. Schaefer, E. S. Converging conceptual models for maternal behavior and for child behavior. In J. C. Glidewell (Ed.), Parental attitudes and child behavior. Springfield, 111.: Thomas, 1961. 45

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Sears, R. R. Relation of early socialization experiences to aggression in middle childhood. Journal of Abnormal Social Psychology, 1961, 63, 466-492. Maccoby, E. E., & Levin, H. Patterns of child rearing. Evanston, 111.: Row, Peterson, 1957. Skinner, B. F. The behavior of organisms: An experimental analysis. New York: Appleton, 1938. Science and human behavior. New York: Macmillan, 1953. Solomon, R. L. Punishment. American Psychologist, 1964, 19, 239-253. Stayton, D., Hogan, R., & Ainsworth, M. Infant obedience and maternal behavior: The origins of socialization reconsidered. Child Development, 1971, 42, 10571069. Thorndike, E. L. Reward and punishment in animal learning. Comparative Psychology Monographs, 1932, 8, No. 39. Tryon, R. C., & Bailey, D. E. The BC TRY computer system of cluster and factor analysis. Multivariate Behavior Research, 1966, 1, 95-111. Walker, E. L., & Heyns, R. An anatomy for conformity. Englewood Cliffs, N.J.: Prentice-Hall, 1962. Walters, R. H., Parke, R. D., & Cane, V. A. Timing of punishment and observation of consequences to others as determinants of response inhibition. Journal of Experimental Child Psychology, 1965,2, 10-30. Witkin, H. A., et al. Personality through perception. New York: Harper, 1954. Wolpe, J. Psychotherapy by reciprocal inhibition. Stanford, Calif.: Stanford Univ. Press, 1958.

46

<8> C. G. B E E R <£>

A View of Birds

C O N T R I B U T I N G to a volume on child psychology presents a temptation to someone who, like myself, is an ethologist. Ethology purports to be a pure science of animal behavior, yet ethologists are frequently prone to visions of their science in use — applied to understanding of the human condition and the remedying of its ills. Consider the number of ethologists who began their research careers with studies of fish or birds and later moved on, if not to people, at least to primates: Hinde, Morris, Andrew, Eibl-Eibesfeldt, Blurton-Jones, and many others. Lorenz, of course, has always included the human case within the scope of his thought, but it has loomed more and more prominently there in recent years; and Tinbergen has now joined the company of "human ethologists" with a study in which he and his wife have brought an ethological perspective to bear on autism in children (Tinbergen & Tinbergen, 1972). This increasing preoccupation of ethologists with the relevance of ethology to the study of humans has been encouraged by increasing and extensive references to ethological concepts and research in the writings of scholars in other fields, such as NOTE: Research reported in this paper was supported by U.S. Public Health Service Research Grants GM 12774 and MH 16727, and a grant to the Institute of Animal Behavior, Rutgers University, from the Sloan Foundation, New York. Permission to work in the Brigantine National Wildlife Refuge and to use a federally owned building there as a field station was granted by the U.S. Fish and Wildlife Service. I am grateful to the refuge manager and personnel for their hospitality and cooperation. I should also like to express gratitude to all those who contributed guidance, companionship, and cooperation during my research career, in particular Professor B. J. Marples, Professor N. Tinbergen, F. R. S., Dr. J. P. Mailman, and Dr. Monica Impekoven. Finally I acknowledge my profound debt to the late director of the Institute of Animal Behavior, Professor D. S. Lehrman. I should like to dedicate this paper to his memory.

47

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Piaget (1971) and Bowlby (1969) in the field of child psychology. There have also been conferences devoted to discussion of the relevance of ethology and comparative psychology to human psychology. Without doubt, this involvement of ethology with human matters has enriched both ethology itself and the sciences of man to which it has contributed and from which it has borrowed. Such enrichment is particularly evident where ethological ways of looking at behavior have been employed in the observation and description of people. However, there has also been a flood of tender-minded speculation pouring from the attempts to establish an ethology of man, particularly in the popular press in recent years. The temptation presented by a new and unsophisticated audience, eager to be told the biological truth about itself, has tended to lead writers away from sound science. Ideas have been resuscitated that most toughminded ethologists had discarded as defunct. Arguments have been developed out of a tissue of loose analogies held together by ingenious concoctions of selected evidence. This kind of speculation, in which facts are fitted together to form a plausible pattern, continues a tradition that ethology inherited from its parent science of comparative morphology, and it is still the only resort in the case of those questions about behavior that pertain to its evolutionary history. But just as Darwinism was discredited for a time by the armchair speculations about phylogeny and adaptation propounded by biologists in this tradition toward the end of the nineteenth century, so, I fear, ethology is likely to be discredited by the story-spinning done in its name — at least with those who judge a science by its respect for evidence and rigor in argument. It seems to me that the involvement of ethology with human nature has created a "tumbling ground for whimseys" that tempts even those ethologists who, like myself, have only a layman's knowledge of the sciences of humanity. Therefore I shall try to resist the temptation to exercise my imagination on the human condition, or to pose as a bringer of fundamental truths about the affairs of men. I own up to being a mere bird watcher. I am aware, of course, that to refer to someone as a mere bird watcher is often to imply that the person is not really a scientist. Bird watchers tend to be thought of as passive in the face of nature, either gazing openmouthed with wonder which they try to express in sentimental description, or ticking off lists in an effort to identify everything that can be seen or to see everything that can be identified. Scientists, in contrast, actively en48

C. G. BEER gage nature in experiment and measurement in which ingenuity, precision, control, and logic combine to advance knowledge and understanding beyond surface appearances and the what and where of field guides. However, there is a view of science that sees the bird watcher's kind of activity as the necessary first step in any field of scientific endeavor. According to Lorenz (1950, p. 232), "It is an inviolable law of inductive natural science that it has to begin with pure observation, totally devoid of any preconceived theory and even working hypothesis." This view has come under attack from philosophers of science such as Karl Popper (1959, 1963), who have argued that preconceived theories or working hypotheses must always be involved in scientific observation to enable the scientist to decide what is to count as a fact of relevance to his investigation. I, myself, have been a critic of this "doctrine of immaculate perception." Each year my students hear why, for both logical and practical reasons, there can be no such thing as pure observation, even for a bird watcher. I found it ironical, therefore, that when I was called on to make my own preconceived theories and working hypotheses explicit, as I was for this paper, the job proved to be so difficult that I had fears of coming up empty-handed. Part of the problem is that one's ideas evolve with one's research, reading, and thinking—assimilation eventuates in accommodation. More difficult still is the fact that one is so close to one's own thought that the distinction between schema and content is blurred; trying to put oneself at sufficient distance for clear vision is like trying to leap over one's shadow. However, such distance is perhaps attainable for one's past work. By stalking one's present thinking via its past a way may be found to catch it unawares. I therefore decided to lead up to my account of my current research through some reflections on my earlier work, to try to bring out the kind of continuous interplay between ideas and facts, methods and tastes, that has shaped what I am doing now. In following this course my aim will not be to indulge in autobiography for its own sake, but rather to give substance to some general reflections about ethology with which I shall conclude.

The Framing of a View I have alluded to the stereotype of the bird watcher enraptured by natural beauty. It would be a mistake to dismiss the taste and sentiment in49

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY volved in this sort of experience as irrelevant to scientific research. Lorenz (1950) has argued that the long hours of careful observation necessary to obtain a detailed description of the natural behavior of a species will be undertaken only by someone who has a passion for the kind of animals concerned. It is probably no mere coincidence that many ethologists, particularly among those who deal with the "surface structure" of behavior, have artistic ability or at least an interest in art. At any rate my earliest fascination with biology, and, hence, later with animal behavior, was as much aesthetic as scientific. As a schoolboy I was taken deer stalking and trout fishing by my father, who thus nurtured in me a taste for mountain and river scenery, and a curiosity about the forms of animals and plants. As an undergraduate my major subject was zoology, and I received my instruction in a department in which the teaching was still mainly in the Huxley-Goodrich tradition of comparative morphology. This training in the perception of pattern in animal form involved appeal to aesthetic sensibility to a degree that was quite inconsistent with the prevalent conception of the contrast between science and the humanities. The comparative tradition in zoology has been a major influence in the shaping of ethology since its inception. Both Lorenz and Tinbergen received their training in settings in which comparative anatomy, informed by the theory of evolution, was still dominating zoological teaching. My conversion to the study of animal behavior was effected by my reading of some of the now classical writings in ethology, such as Lorenz's (1952b) King Solomon's Ring and Tinbergen's (1951) The Study of Instinct. Tinbergen was the supervisor of my doctoral research. This research was done in England, on the nest-building and incubation behavior of Black-headed Gulls — the common small gulls of Europe and the British Isles. In addition to confirming my taste for the observation of animal behavior in the wild, and providing an apprenticeship in the conduct of such study, the investigation was of special significance for my later work in three particulars. The first of these was an involvement in the use of quantitative methods and statistical analysis. In a paper delivered at the Zoological Convention at Freiburg in 1952 (Lorenz, 1952a), Lorenz had lamented that in coming to grips with the full complexity of animal behavior, ethology was bound to become increasingly statistical. When reporting on my work to a group in Oxford in 1960 I lamented that instead of realizing an ambition 50

C. G. BEER to write another King Solomon's Ring I had produced something much more like an issue of the British Railways Timetables. The resort to statistical methods that was typical of the ethologists of my generation was due to more than the realization that it was called for by the complexities of the behavioral problems with which we found ourselves confronted, however. Indeed, the awareness of the complexity was as much a consequence of the statistical approaches as a reason for their adoption. The emphasis on statistics and quantitative methods was part of a reaction to the older style of ethological theorizing, which appeared to many critics, particularly in America, to be altogether too cavalier about the treatment of evidence. Added to this was the arrival of the electronic computer, which enlarged the universe of feasible quantitative analysis and also brought concepts from computer technology and communications theory into ethological thinking. By the time I came to write my doctoral thesis ethologists were beginning to seek training in the use and theory of computers; statistical handbooks such as Siegel's Nonpar-ametric Statistics (1956) had become as vital a part of the ethologist's equipment as his binoculars, and no conclusion with pretensions to respectability went abroad unattended by appropriate control groups and a computation of the confidence with which belief in it could be held. The second of these formative influences was a discovery by G. H. Manley, whose companionship I had the good fortune to enjoy during the periods of field study of the Black-headed Gulls. Manley investigated the agonistic and courtship displays of the gulls and found evidence that what is conveyed by performance of such a display depends upon the context in which it occurs as well as the form of the display itself. The displays used in pair formation are of the same types as those used in hostile interactions, but in pair formation there are features of sequential patterning, orientation, and other details that distinguish the displays from their occurrences in situations in which they express hostility. This discovery of a syntactic component in the functioning of displays was contrary to the then prevalent assumption that each type of display has but one signal function — namely, to elicit but one type of reaction from the recipient animal — social communication being a more or less one-to-one correspondence between signals and responses. Unfortunately, this work of Manley's, which was completed in 1960, has not been published. However, the importance of context in the social communication of birds was discovered independently by John Smith in his studies of North American Flycatchers, and the 51

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY observation has consequently found its way into the literature (Smith, 1963,1965,1966,1968). The third point to which I attach significance for my later work arose from my own studies of incubation and nest-building of the gulls. I shall have to abbreviate and simplify what was really a rather complex matter. The gulls of a pair build their nest before the eggs are laid, but they continue to perform nest-building behavior throughout the incubation period. Moynihan (1953) had concluded that this nest-building performed during incubation was not caused by the kinds of factors that induce building before laying, but, instead, by frustration of the drive to incubate, either by a gull's being denied access to its nest or through deficiency in the quality of the stimulation effected by the contents of the nest. In ethologese, nest-building during incubation was regarded as a "displacement activity." Implicit in this conclusion was the assumption that underlying and determining the occurrences of the various movements constituting nestbuilding and incubation behavior are two causal systems or drives: a nestbuilding drive and an incubation drive. I attempted to test this assumption (Beer, 1963). A direct approach to the physiology of the matter was not a practical possibility in the circumstances, but it was possible to select and manipulate various factors to see how they might affect the behavior. I argued that if the two-drive assumption were correct, then movements supposed to belong to one of the drives should manifest a closer causal affinity to one another than to any movements supposed to belong to the other drive, and vice versa. Causal affinity was to be judged according to correlations among variations in frequencies of occurrence of the movements, temporal association between the different types of movements in different situations and at different phases of the reproductive cycle, the degrees to which the different types of movements were affected similarly by manipulation of external factors (e.g., the number of eggs in the nest), and other comparisons. The evidence I obtained failed to confirm the twodrive assumption. Instead it appeared that some of the nest-building movements were more closely bound to incubation movements than to other nest-building movements, and vice versa. A classification of the movements according to the evidence of causal affinity would thus cut across the approved classification which was, in contrast, based on the functions served by the movements: nest-building and incubation. A close look at the forms of the different kinds of movements — their motor patterns — also revealed a lack of conformity with the functional categoriza52

C. G. BEER tion: where the functional categorization recognized two kinds of lowering into the nest, a categorization based solely on motor pattern would recognize only one. The general point impressed upon me by the results of this study was that behavior can be divided up in a variety of different ways for purposes of description: according to functional criteria, causal criteria, criteria of pure form, and in other ways. To assume that the divisions in one type of description will match those in another is to beg a question that could be crucial to the matter at issue. The point was reinforced for me by my reading of some of the contemporary British philosophers, such as Stuart Hampshire. In his book Thought and Action (1959) Hampshire argues, in effect, that the notion of a complete description of anything is a contradiction in terms, since description involves division and classification which exclude other possible divisions and classifications and hence other possible descriptions. This point was brought home to me again during my postdoctoral fellowship at Rutgers University. I undertook to collaborate in the development of an elaborate computer-based system for the processing and analysis of behavioral data. A keyboard device had been constructed by means of which behavioral data could be collected in a form that could be fed directly into the computer. The computer was to be instructed to generate tables showing frequency distributions for all the items recorded, transition probabilities between items, correlation coefficients, and so forth to reveal the features of surface organization of the behavior which might provide clues to the features of the deep structure underlying it. My assignment was to use the keyboard to accumulate a body of data which would serve as raw material for construction of the computer programs. I attempted to do this with the pre-laying behavior of pairs of caged ring doves. A flaw in the enterprise soon became apparent. The number of keys on the keyboard set a limit of about fifty to the number of different items that could be included in the records. I found that this was incompatible with the aim of obtaining a record that was both comprehensive and free of ambiguity. For example, the doves frequently adopt a "stooped" posture — body inclined with the head down and the tail up — and this posture gives the impression of being a display or part of a display having to do with selection of the nest site and nest-building: it is performed repeatedly on prospective nest sites and on the nest, and is often accompanied by other 53

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY behavior associated with nest-site selection and nest-building. However, there are also many occasions when the doves get into this same attitude in places other than the nest site or nest, and in contexts that seem to have nothing to do with nesting — for example, while feeding or drinking, just before flight from a height or just after landing. To leave all these latter occurrences of the posture out in an investigation of the connection of stooping with nesting would be to court circular argument, since association with the nest site or with nesting activities would be among the criteria for identifying an occurrence of the posture as stooping. On the other hand, to include all occurrences of the posture would involve loss of important distinctions unless sufficient information about places and contexts of occurrence were also included. To cover the full range of situations in which the posture can occur, however, would use up all the places on the keyboard and thus leave no room for any of the other items of behavior, many of which involve the same kind of problem about specification that affects stooping. In short, I was confronted again by the general principle that, of necessity, all description is limited. In spite of the technical advances in data collection and data processing that have been effected by the application of electronics to the study of behavior, one still has to start out with selection of one out of an infinite number of possible descriptive strategies, in accordance with whatever one's wits and experience offer as the best bet. Following my postdoctoral term at Rutgers, I spent two years in New Zealand as a lecturer in zoology at the university where I had been an undergraduate. During this period, I made some field study of Black-billed Gulls, a species closely related to the Black-headed Gulls of Europe. Comparison of the two species revealed a number of differences between them in various features of breeding behavior, and these appeared to be related to differences in their breeding habitats. Black-billed Gulls nest on the shingle of riverbeds that are subject to irregular flooding. This source of danger to the survival of broods appears to have resulted in the natural selection of a compression of the period of brood vulnerability to flooding, which has involved a complex of changes including effects on reproductive physiology, social behavior, and the rate of development of the young (Beer, 1966). In addition to illustrating for me at first hand how interspecies differences in reproductive biology can be interpreted in terms of evolutionary adaptation, the observations made the point that the reproductive biology of a species can be viewed as an 54

C. G. BEER integrated whole in which physiology, behavior, and ecology are interrelated so intimately that contemplation of any one facet in isolation from the rest is likely to distort it. In 1964 I returned to the United States and to Rutgers and began the program on which I am engaged at present: an ethological study of Laughing Gulls. Laughing Gulls and Recognition The study of Laughing Gulls in which I and my associates have been engaged during the last several years has included a number of aspects of the reproductive and social behavior of the species. I shall confine myself to only that part of this work which pertains most closely to developmental matters: investigation of vocal communication between parents and young, with particular attention to individual recognition by ear. Before turning to the experiments however, I must sketch some of the general features of the species and its breeding cycle. The Laughing Gull (Lams atricilld) is the common small gull of the eastern coast of the United States, where it typically breeds in colonies on salt marsh islands such as those in the Brigantine National Wildlife Refuge, New Jersey, the location of our studies. The gulls winter in the tropics and arrive on the breeding grounds during April. Egg laying begins in mid-May, hatching a little more than three weeks later, and the first fledged young can be observed toward the end of July,-unless the breeding is set back as a consequence of tidal flooding. The birds of a pair build a shallow, grass-lined nest, on the ground, in which, as a rule, three eggs are laid. Both sexes incubate and share in the care of the chicks when they hatch. The chicks of Laughing Gulls have been described as "semi-precocial" (Nice, 1962), which means that at hatching they have down feathers, open eyes, and some modest powers of locomotion, but are dependent on their parents for their food and a certain amount of warmth. The parents feed their chicks by regurgitating food, holding the food in the tips of their bills and offering it close to the heads of the chicks. The chicks obtain the food by pecking at the bill of the parent, a movement which can also induce the parent to regurgitate (Hailman, 1967). During the first few days after hatching the chicks spend almost all of their time in the nest with one or other of their parents standing or squatting over them. On about the fourth day the chicks begin to make excur55

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY sions from the nest, often, apparently, being induced to do so by one of the parents — either because they follow the parent as it walks from the nest or because they approach the parent as it stands at a distance from the nest and calls to them. At first these excursions extend no more than a meter or so from the nest, but within a day or two they can range ten meters or more. Until about the tenth day after hatching one or the other parent is still continuously in attendance on the chicks, although the chicks wander from the immediate vicinity of the parents from time to time. From about the tenth day onward, however, the parents begin to leave the chicks unattended for spells which range up to two hours or more by the time the chicks are three weeks old. A three-week-old chick may not see its parents for most of the day, except for the periodic visits, which may be limited to a minute or so, during which it is fed. In this period, when the chick's activities are becoming progressively independent of those of the parents, the range of the chick's wanderings increases, as a consequence of which it comes to experience frequent encounters with gulls other than its parents and chicks other than its siblings. The early social development of a Laughing Gull chick can thus be divided into three stages: a nestling stage in which the chick's world is restricted to the immediate vicinity of the nest and continuous close association with the parents; a stage in which the chick's world extends to include an area the margins of which are some distance from the nest, but in which parental presence is still virtually a constant feature; and a stage in which the chick's world enlarges to include space that has to be shared with other gulls and from which parental presence is often missing. These three stages were selected for comparison in experiments that I shall describe below. Colonial breeding of the sort one finds in Laughing Gulls has a number of features that one expects to be reflected in the social behavior of the species. For example, the close proximity of families entails a high probability that members of different families will meet from time to time and, hence, that either social interactions are without regard to family affiliations or there must be some means by which the members of a family distinguish one another from other individuals and distribute their social behavior accordingly. Thorpe (1968) has asserted that efficiency and success in breeding in colonial birds necessitate that individuals of a family be able to recognize one another. Although individual recognition is not the only feasible solution to the problem of achieving productive harmony in a crowd, it is the case that evidence of such recognition has been ob56

C. G. BEER tained in all the species of colonial birds in which it has been sought, with the exceptions of the Kittiwake (Cullen, 1957) and Franklin's Gull (Burger, personal communication), both of which have special features which, it has been argued, make them exceptions that prove the rule (Beer, 1970c; Burger, personal communication). The existence of individual recognition in avian social behavior raises interesting questions about social dynamics and social development, some of which I have addressed in my studies of Laughing Gulls. EXPLORATORY TESTS

My active interest in individual recognition in Laughing Gulls was aroused initially when I was looking for something else. From field observation of social interactions between gulls I had derived some ideas about the signal functions of various of the call and posture types. What seemed to me to be the clearest cases were two kinds of call by means of which the parents appeared to influence the behavior of their chicks: the "crooning call" and the "alarm call." Crooning, as the label suggests, is a relatively soft, smooth, drawn-out note which sometimes includes a single yodel-like step up in pitch. A parent utters the call when feeding its chicks, and field observation suggested that its utterance stimulates a chick to approach, to peck at the parent's bill, and to take food when it is presented there. The call is used in a number of other contexts as well, but these can be ignored for now. The alarm call consists of two or three short and not very loud notes uttered in quick succession — something like "kek-kek" or "kekkek-kek." This call is uttered when the gulls are disturbed in the gullery by an event such as the approach of a person, and is particularly prominent at the beginning of such a disturbance, when it occurs almost to the exclusion of any other type of call. The call apparently alerts all gulls within earshot to the imminence of a situation of potential danger, and puts them to flight if the situation persists or intensifies. For chicks older than a day or so post-hatching the call appears to act as a stimulus to hide. When the call occurs, the chicks run to the edge of the nest or into the marginal vegetation and crouch in silence until the situation returns to normal. To test the apparent effects of these vocal signals on the behavior of the chicks, I attempted an experiment in which recordings of the calls were played to chicks indoors, and their responses observed. The apparatus for this experiment, and the method for scoring the behavior, were used with minor modifications where necessary in later experiments as well, so the

57

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY description of apparatus and procedure here will serve for the whole series of experiments. A wooden box was constructed to furnish a test arena (we christened it the "chickarena"). Its dimensions were 122 by 30.5 by 30.5 cm. It was open at the top, and the end walls consisted of cheesecloth mounted on wooden frames. The inside of the box was painted a flat grey, and the floor was marked out in a grid to enable me to score a chick's position with reference to the source of sound, which was a loudspeaker placed against one or other of the end screens. A light was fixed above the box to give even illumination over the floor, and a tilted mirror enabled me to view the floor from behind a 120 cm-high plywood screen which concealed me from a chick in the box. From behind this screen, I made my observations and controlled the tape recorder connected to the loudspeaker. The recordings used in the tests were obtained with this same tape recorder (Nagra III) and a directional microphone (Sennheiser MKH 404) during observation sessions in the gullery. Examples of the types of call in question were selected for good quality from the original tapes and transcribed to loops of tape — one call repeated three times at intervals of five seconds on each loop — and these loops were used as the test tapes. A chick to be tested was placed at the central position of the floor grid of the arena. It was observed for five minutes without sound, and then for a further five minutes during playback of a test tape. In each of these fiveminute periods I noted the chick's position on the grid, its orientation with respect to the speaker, and its posture at the end of each fifteen-second interval, and I made a mark for each vocalization it uttered. Chicks of various ages were captured in the field and tested in this way with recordings of crooning calls and alarm calls, the testing taking place about one hour after capture. The results of these initial tests were disappointing. None of the chicks showed any clear evidence of response to any of the calls played to them. This outcome was inconsistent with what I had observed in the field, so I sought a flaw in the assumptions I had made in setting up the experiment. In conformity with classical ethological doctrine and the principle of parsimony, I had taken it for granted that the effects of these vocal signals were species-characteristic — that is, universal with respect to individuals of the species, at least in a particular age or sex class. This assumption was not contradicted by the results of my tests, for it remained possible that the calls by themselves form only part of the effective signals, which, for 58

C. G. BEER completion, require the accompaniment of specific visual stimuli or some other features to give them a necessary context. But another possibility was that the calls produce their effects only if the chicks recognize them as utterances of their own parents. This latter possibility I immediately attempted to test by selecting recordings from a chick's own parents for the test-tape loops and using these in a repeat of the experiment. Again, the chicks tested with these recordings failed to show any of the expected responses. I was about to abandon this line of investigation when, more or less as a last resort, I played the whole of an unedited recording to a chick which, like all of the others, had just ignored the specially prepared loops of its own parent's crooning and alarm calls. The unedited recording had been made that morning in the vicinity of the place where the chick had been captured, and it included a variety of calls from a variety of individuals including the chick's own parents, and also the occurrence of alarm calls caused by the approach of my assistant at the end of the recording session. The first half of the tape contained calls of gulls other than the chick's parents, and the chick ignored them in the way it had the test loops. Near the middle of the tape I had recorded the arrival calls of one of the chick's parents. As soon as the sequence of calls by the parent began, the chick started calling; it turned and ran toward the speaker and attempted to scramble over the screen at that end of the box. From then on it continued to call and to try to get at the source of the sound, in response to the voice of its parent, until, toward the end of the tape, the alarm calls began. Then the chick ceased calling, turned, and fled to the opposite end of the arena, where it crouched, motionless, its head pushed into a corner. I had obtained both of the patterns of response for which I had been looking. As I had hoped, calls of adult gulls do not require accompaniment by the usual visual stimuli to produce their effects on a chick. But why did the unedited recording work when the test loops did not? The arrival sequence of the chick's parent included a number of calls other than and in addition to crooning. Most prominent among these other calls were "long-calls" and "ke-hah" calls. The long-call is the most striking and elaborate of Laughing Gull vocalizations. Indeed, it gives the gull its name. It consists of a string of notes divided into three parts which I refer to as the short notes, the long notes, and the head-toss notes (see Beer, 1970b, fig. 3, for an illustration of sonagrams of this call). It occurs in almost the whole range of contexts of Laughing Gull social behavior during

59

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY the reproductive season. Ke-hah calls are disyllabic calls that are often uttered in strings (again, see Beer, 1970b, fig. 2). They occur, also, in a variety of social contexts, often in association with long-calls. Now both of these calls, but particularly the long-call, have individual characteristics that enable one to identify individual gulls by ear in the field. I was unreflectingly aware of this fact before I began the arena tests. The evidence of the ear was confirmed by the eye when I was able to view sonagrams of the calls. In sonagrams of long-calls, for example, it was apparent that the short-note phases of the calls of an individual are close to uniform in the number of notes they contain, the durations of the notes, and the intervals between them, and in other features, but differ considerably in these respects from the long-calls of other gulls (Beer, 1970c, fig. 1). No such distinct individual characteristics are evident in crooning calls, either to the ear or in sonagrams. The explanation that suggested itself, therefore, for the fact that the chick responded positively to an unedited recording but not to the crooning on the test loop, was that the long-calls and ke-hah calls in the unedited recording identified the calls to the chick as those of one of its parents, whereas crooning by itself was not so identified, such identification being necessary for positive response to occur. This hypothesis became the basis for the further experiments that I shall describe. But first I have to deal with the question of why the recordings of alarm calls worked in the one case but not in the other. The most obvious difference between the alarm calls on the test loops and those on the unedited tape was that a test loop consisted of repetitions of a single call, the intervals between repetitions being about five seconds, whereas the unedited tape presented a chorus of alarm calls which continued without interruption until the end of the tape. The possibility suggested itself that alarm calls have to occur in a chorus to elicit the hiding response of a chick. To examine this possibility playback tests were run in which tapes of such alarm call choruses were used. In every case the chick responded by turning away, retreating from the source of sound, and crouching, motionless and silent, until the end of the test — whether or not the recording included alarm calls of the chick's own parents. Again, the results of these tests were consistent with what I had observed but not sufficiently remarked in the field. As a rule chicks ignore isolated alarm calls in the field, even when they are uttered by their own parents. It is only when several gulls have been put to flight, and hover, alarm calling 60

C. G. BEER together overhead, that the chicks break off what they have been doing and flee and hide. Thus, in the cases of both crooning and the alarm call, my initial assumptions had led me to design an experiment that was too removed from nature to retain the phenomenon I was seeking to study. The failure of the experiment forced me to resort to an experimental situation closer to the natural situation, and thus drew my attention to things that I had seen but not noticed in the field. I was reminded of some comments about research strategy that Philip Teitelbaum made in some lectures that I attended and which he summarized briefly in his little book Physiological Psychology (1967). In pursuing the analytical investigation of physiological mechanisms, Teitelbaum said, one tries to pare away everything that is inessential to whatever function is in question to isolate what is essential. In doing so, however, it is wise to proceed in small steps to retain the function as long as possible and to pinpoint the stage beyond which it is lost. Only so long as one still has the function intact can one continue to investigate it, and if analysis has proceeded so far as to lose contact with it, the only thing to do is to back up until contact is remade. By resorting to a natural sequence of calls I had remade contact in the laboratory with communication functions I had observed in the field. For my next step in the investigation of these functions I therefore used similar recordings of natural sequences of calls in tests of the effects of the calls on chicks of the same age as the one that had responded to the playback of the unedited tape, 6-8 days post-hatching. TESTS OF INDIVIDUAL RECOGNITION BY EAR

To begin with it needed to be established that Laughing Gull chicks can recognize the voices of their parents. Chicks from two parts of the gullery were each tested with playback of a sequence recorded from one of their own parents and a sequence recorded from one of the adults from the other part of the gullery, the chicks being selected in pairs so that what was the parental sequence for one was the foreign sequence for the other, and vice versa. Each test recording contained crooning, long-calls, and ke-hah calls, and lasted five minutes. A chick heard both of its test recordings twice — once from each end of the arena — and in alternation. The chick was observed for five minutes without playback before each five minutes with playback. The testing sequence was shuffled between chicks to cancel possible order effects. 61

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY The results of this experiment were unequivocal. The chicks responded to the recordings of the calls of their own parents by orienting toward the sound, approach, ceaseless locomotion, and vocalization. When recordings of the same kinds of calls from the foreign gull were played, most of the chicks turned away from the sound and retreated from it; they were silent, and some of them crouched (Beer, 1969). These results were consistent with the belief that Laughing Gull chicks can recognize the voices of their parents, but they left open another possibility. The two parts of the gullery in which the recordings of the test calls were made were so distant from each other that a chick may well never have heard the voice of the foreign gull before being tested. The discrimination shown between the voice of the parent and the voice of the foreign gull could thus have been no more than discrimination between the familiar and the strange. To examine this possibility I repeated the experiment but with the difference that the calls to be distinguished from those of the parent were recorded from a close neighbor in the gullery. The results were the same as in the previous tests: the chicks reacted positively to the calls of their parents and negatively to the calls of their neighbors. The case for individual recognition by ear was thus supported (Beer, 1970a). However, an effect of previous experience of hearing adult calls, and hence of familiarity, did emerge in a comparison of the behavior of these chicks that had been in the care of their parents before testing, and the behavior of their siblings that had been incubator-hatched, from eggs collected during the second half of the incubation period, and hand raised indoors before testing. The hand-raised chicks were tested in the same way, with the same recordings, and at the same age (6-8 days post-hatching) as were their parent-raised siblings. The hand-raised chicks were more active and vocal than the parent-raised chicks in the five-minute silent periods, but, without exception, they reacted to the playback by fleeing and crouching in silence. Thus, in addition to showing no discrimination of the voices of their parents, they failed to show anything like the usual response to the kinds of call played to them and reacted instead in the way parentraised chicks of the same age react only to alarm calls (Beer, 1970a). This last impression was confirmed by another series of tests. A threeminute test tape was prepared in which the first two minutes consisted of a mixture of long-calls, ke-hah calls, and crooning, and the third minute consisted of a chorus of alarm calls. Six parent-raised chicks and six hand62

C. G. BEER raised chicks were tested with this tape. The results are summarized in Table 1. The hand-raised chicks showed the fear response of fleeing and crouching from the beginning of the three-minute test, whereas the parentraised chicks showed this response only during the last minute of the test. Wilcoxon matched-pairs comparisons of the scores in the second and third minutes of the tests showed that for parent-raised chicks there was more orientation away from the speaker (p<.01), the positions taken up were farther from the speaker (p<.01), more time was spent crouching (p< .01), and calling by the chick was less frequent (p< .05) in the third minute than in the second minute; but no such differences were found for the hand-raised chicks. The fear aroused in the hand-raised chicks by playback of any calls may have been due in large part to the novelty of the sounds to them. My colleague Dr. Monica Impekoven has found that the indiscriminate fear of playback calls soon wanes if the chicks are given repeated exposure to them (personal communication). I placed my hand-raised chicks with their parents in the field after the initial testing. Six of the thirteen survived Table 1. Results of Three-Minute Playback Tests in Which Alarm Calls Occurred in the Third Minute for Parent- and Hand-Raised Chicks 6-8 Days Post-Hatching" Median Scores of Parent-Raised Chicks

Response

Median Scores of Hand-Raised Chicks

p" Pb Pc (2nd (Parent(2nd vs. vs. vs. Whole 2nd 3rd 3rd Whole 2nd 3rd 3rd HandTest Min. Min. Min.) Test Min. Min. Min.) Raised)

Orientation 0 Speaker end . . . . 4 Other end Position Speaker side . . . . 0 36 Other side 2 Change 1.5 Calling 4 Crouching

0 0 0 0.5 0 0.5 0

0 n.s. 4 <.01

0 8

0 2.5

0 3

0 0 0 0 n.s. 32 <.01 39.5 13.5 14.5 1 n.s. 0 0 0 0 .05 0 0 0 4 4 4 .01 12

n.s. n.s.

n.s. .05

n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. .05 .05

a Except for position, the maximum possible score for a whole test was 12, and for each minute was 4. For position, the maximum score for a whole test was 144, and for each minute was 48. b Probability of difference according to Wilcoxon matched-pairs comparisons, two-tailed, N = 6. c Probability of difference according to median tests; except for calling, for which Fisher tests were run on the number of chicks that called and the number of chicks that did not call in the two groups.

63

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY and I recaptured and retested them after they had been with their parents for six days. They no longer showed fear of long-calls, ke-hah calls, or crooning calls (they were not retested with alarm calls), but neither did they show any positive response of the kind shown by their siblings that had been in the care of their parents since hatching: they simply ignored the sounds (Beer, 1972). This result suggested that experience of interaction with the parents before the sixth day post-hatching is involved in the development of the normal responsiveness of a chick to the calls of adults. I turned next to this question of the development of a chick's responses to the calls of adults and, in particular, the emergence of individual recognition by ear. Still using test tapes consisting of mixtures of crooning, ke-hah, and long-calls in sequential preference tests of the voices of parents and the voices of neighbors, I extended my investigation of individual recognition to chicks younger and older than the 6-8 day old chicks of the previous tests. I selected three groups corresponding to the three stages of chick social development described earlier: an early group consisting of chicks 13 days post-hatching; a middle group, 6-8 days post-hatching; and a late group, 12-28 days post-hatching. The results of the tests of these three groups are summarized in Table 2 (see also Beer, 1970b). Discrimination of the voices of the parents was shown in all three groups — even by chicks as young as 24 hours posthatching. In spite of this discrimination, however, the chicks of the early group responded strongly with approach and vocalization to the calls of the neighbor as well as to the calls of the parents. In the two older groups, on the other hand, approach and vocalization by the chicks were virtually reserved for the calls of the parents. The question of exactly when and how a chick first learns to distinguish the voices of its parents from those of other gulls still remains open. There is support for the possibility that the learning of the discrimination may begin before hatching, as Tschanz (1968) found to be the case for Guillemot chicks (Uria aalge). Incubator-hatched chicks to which crooning calls were played during the last two days or so before hatching increased their rate of pecking at a red rod when presentation of the rod was accompanied by crooning in tests carried out after the chicks had hatched. Chicks with no pre-hatching experience of the calls showed no such augmentation of the pecking response in the presence of crooning (Impekoven, 1971). There is, thus, good reason to believe that the development of a chick's 64

C. G. BEER Table 2. Results of Playback Tests of Three Age Groups of Laughing Gull Chicks" Early Group1' Response

Middle Group 0 Neighbor Parent Tests Tests

Neighbor Parent Tests Tests

Orientation . . 41.3 Speaker end 20.0 Other end Position 35.1 Speaker side . . . 1.6 Other side 52.5 Change Calling . . . . 61.3 107.0 Total calls . . . . 8.8 Crouching

Late Group'1 Neighbor Parent Tests Tests

63.8 2.5

10.0 12.5

55.0 5.0

7.5 17.5

32.5 2.5

44.2 1.1 53.8 87.5 180.0 1.3

0 7.9 5.0 0 0 30.0

34.8 2.1 45.0 51.3 98.0 0

3.3 9.0 10.0 2.5 1.0 2.5

26.3 3.1 30.0 40.0 40.0 0

" Test recordings consisted of natural sequences of crooning, ke-hah, and longcalls of a chick's neighbor in Neighbor Tests and of one of the chick's parents in Parent Tests. They were five minutes in duration and were presented twice to each chick — once from each end of the arena. Scores in the table are medians converted to percentages of the maximum possible score for each behavior category, except Total Calls, which are medians. b Chicks 1-3 days post-hatching; N = 8. e Chicks 6-8 days post-hatching; N = 15. a Chicks 12-28 days post-hatching; N - 11.

responsiveness to the calls of adults begins before hatching and is affected by pre-hatching auditory experience. In a more recent experiment, Dr. Impekoven and Dr. Peter Gold have played long-calls to chicks before hatching. They played one such call to one group of chicks and another to a second group, noted the responses of the unhatched chicks to these calls, and then, after the chicks had hatched, tested the chicks for discrimination of the two calls in a version of the kind of test that I had used to investigate individual recognition. Although discrimination of the tutor call from the other was not consistently demonstrated in this experiment, the trend of the results was sufficiently in this direction to encourage further pursuit of the possibility that chicks begin to learn the individual characteristics of their parents' voices from auditory experience before hatching (Impekoven & Gold, in press). There is also some less direct evidence that is consistent with this possibility. From field observation we have the impression that the frequency of calling by incubating gulls increases during the last few days before hatching, and there is reason to believe that this increase may be due to tactile and auditory stimuli coming from the pipping eggs. Impekoven (in press) found that when she played recordings of the sounds of a hatching chick 65

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY through a small loudspeaker placed in the nest of an incubating gull, the gull usually responded by rising and re-settling and also by calling. She has also found that young at the pipping stage respond to playback of adult calls, such as crooning, by moving and "bill-clapping." The parents and unhatched young thus apparently provide mutual stimulation before hatching. The consequent contiguity between the activity of the parents and activity of the young could be involved in the inception of the discrimination learning, since such contiguity does not obtain between the behavior of the young and the behavior of any adult other than its sitting parent. Another feature that may be relevant is the loudness of the calls of the bird on an egg to the chick inside the egg, compared with the calls of all other birds. Finally, there is evidence from studies of species other than gulls that pre-hatching auditory experience can affect behavioral development, including the development of auditory discriminations and preferences (Gottlieb, 1971;Tschanz, 1968). AGE CHANGES IN THE RESPONSES OF CHICKS TO THE CALLS OF ADULTS

However the learning of the distinguishing features of the parents' voices begins in a chick's life, it was evident from comparison of my tests of the three age groups of chicks that the behavior in which this discrimination is involved undergoes changes in the first two weeks post-hatching. Preference for the voices of the parents progresses from a slight inclination in favor of the parents to an absolute and exclusive bias. Some indications of how this change comes about were obtained from further analysis of the results of the tests. The recordings used in these tests consisted of natural sequences of crooning, ke-hah, and long-calls. Although I attempted to select sequences as similar to one another as possible, some differences were unavoidable, particularly in the numbers of occurrences of the different types of call. However, this variation in the test recordings provided an indirect way of asking how the three types of call compare in their effects on a chick's behavior, and whether these effects change with the age of the chicks. If variation in the numbers of occurrences of a call on the test tapes and variation in the behavioral measures obtained in the corresponding tests were correlated, the direction of the correlation and the specific behavioral measures so related to the call would indicate the effect associated with the call. For each of the three types of call coupled with each of the parameters of chick 66

C. G. BEER behavior, I worked out the rank correlation coefficient (Spearman rs, Siegel, 1956) within each test category (parent tests and neighbor tests) in each age group to find how variation in number of calls was associated with variation in chick behavior. I also analyzed the temporal distributions of occurrences of the test calls and chick responses, for each call-response pair, to see whether temporal contiguity was greater or less than chance expectancy. Finally, in those tests in which vocalization by the chick occurred, I sorted the tests according to the kind of call that had immediately preceded (within 15 sec) the onset of vocalization by the chick — vocalization nearly always being the first sign of positive response to the playback calls. The details of these analyses have been published (Beer, 1970b), so here I shall simply summarize the pattern that emerged (Table 3). The results of the three analyses agreed with one another so I shall not deal with them separately. Also, I shall combine orientation toward the sound, approach, locomotion, and vocalization as "positive response," since they tended to go together; and likewise I shall combine orientation away from the sound, withdrawal, and crouching as "negative response." In the tests of the early group, crooning was associated with positive response whether it was crooning by the parent or crooning by the neighbor, but ke-hah and long-calls tended to be associated either with depression of positive response or with negative response; the trends were similar for parental and neighbor calls. In the tests of the middle group, crooning and ke-hah by the parent were associated with positive response; crooning by the neighbor was equivocal; ke-hah and long-calls by the neighbor and long-calls by the parent were associated with depression of positive response or with negative response. In the tests of the late group crooning by either the parent or the neighbor was equivocal; ke-hah and long-calls by the parent were associated with positive response; ke-hah and long-calls by the neighTable 3. Summary of Laughing Gull Chick Responses to Different Calls in Playback Tests'* Early Group Middle Group Late Group Long- KeLong- KeLong- KeCall hah Croon Call hah Croon Call hah Croon

Tests Parent Neighbor

....

—

—

+

—

+

+

+

+

±

a + = positive response; — = negative response; ± = some positive and some negative response.

67

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY bor were associated with depression of positive response or with negative response. The results of the analyses thus suggested two trends in the development of a chick's responses to these calls of adults: The first trend is a weakening of the indiscriminate and highly stimulating effect of crooning on the nestling's positive filial responses. Second, the ke-hah and long-calls of the parents supplant crooning as the effective stimuli for eliciting the chick's filial responses. There is a simultaneous increase in the suppressing effect of ke-hah and long-calls of adults other than parents on the nestling's filial responses. This evidence of change in the growing chick's responsiveness to the calls of adults was complemented by observations of parent-chick interactions in the field. During the nestling stage it appeared that the only call used by a parent to stimulate the filial responses of its chicks was crooning. Crooning is usually part of the prelude to regurgitation feeding of the chicks, and it occurs during the feeding as well, particularly if the chicks' interest in the food presented to them starts to wane before the food is finished. In this case, the crooning appears to rearouse the chicks to peck and eat further. Ke-hah, long-calls, and other calls occur from time to time during this stage when a parent is in close proximity to its chicks, but these calls appear to be directed at the mate or at other adults and not at the chicks. During the middle stage, when the chicks embark on excursions and the parents often call to them from a distance, the parents use ke-hah calls as well as crooning to induce their chicks to approach them; the kehah calls predominate until the chicks come close. The transition from the early to the middle stage is marked by excursions that are led by the parents. Crooning is the main call used in these excursions but one also hears a combination of ke-hah and crooning, a call that begins as a typical kehah but the second syllable of which is drawn out into the plaintive thread of a croon. Such tagging of ke-hah to crooning at this stage may facilitate the process by which the ke-hah calls of the parents acquire positive significance for the chicks. In a similar way long-calls come to be included in more and more of the sequences that lead to parental feeding of the chicks, as the middle stage passes into the late stage. By the time the chicks are being left to themselves for spells, long-calls are introductory to almost all such sequences; a parent returning to the vicinity of its chicks almost always announces its arrival by long-calling, and it is just after such arrivals that most of the feeding sequences take place. There is also a functional fit between the increased selectivity of a 68

C. G. BEER chick's responsiveness to the calls of adults, as evidenced in my tests, and the demands of the social situations in which the chicks find themselves. In the nestling stage a chick is in close proximity of one of its parents virtually all of the time, and the occasion does not arise when a chick would be likely to direct its filial behavior to gulls other than its own parents. By the late stage, on the other hand, a chick does encounter adults other than its parents on its wanderings and also on the home ground, particularly when the parents are away. Now the chick has to be able to tell its parents from other gulls if filial responses are to stay within the family, and if the chick is to avoid the punishment that would be consequent on indiscriminate approach to adults. As a rule, adults are vicious toward chicks other than their own. They attack and peck the chicks they encounter, sometimes with such severity as to kill. That the chicks should be able to make the required discrimination by ear is consistent with the fact that a chick's vision is often impeded by the vegetation in which it spends much of its time so that it has to recognize and locate its parent when the parent is out of sight. Thus, when combined with field observation of natural social behavior, the pattern of age changes in the responses of Laughing Gull chicks to the calls of adults that I had teased out of the results of my tests seemed sufficiently well supported to command belief. However, my experimental evidence on the question of the effects of the different types of call was so indirect, so reliant on statistics that might be considered devious, that I decided that my conclusions required testing by a more direct approach. Moreover, there was an unsettled question about the basis of the individual recognition, Do all three types of call used in the tests convey individual identity, or does this function reside in only one or two of them? I have mentioned that both ke-hah and long-calls possess individual characteristics distinct enough to enable one to identify particular birds by ear, but that I had found no such characteristics attaching to crooning calls. Could it be that the nestlings in my tests were positively stimulated by calls of both the parent and the neighbor because of indiscriminate response to the crooning, and that the evidence of individual recognition was due to the ke-hah and long-calls, the effects of which were largely overridden by crooning? Then the evidence would suggest that the dominance of ke-hah and long-calls by crooning is reversed as the chick grows older so that the chick comes to respond positively only to the voices of its parent.

69

M I N N E S O T A SYMPOSIA ON CHILD PSYCHOLOGY RECOGNITION AND MEANING

To confirm the evidence of my previous analyses, and to test the hypotheses suggested by it, I reverted to my original strategy of testing chicks with one type of call at a time. The testing procedure I adopted was similar to that already described, except that the testing sequence consisted of a successive preference test followed by a simultaneous preference test. A chick was observed in the arena for five minutes without sound, then it was played a recording of a call of either its parent or a neighbor for five minutes, followed by another five minutes without sound, then five minutes of playback of the other test call, five minutes without sound, and finally five minutes of playback of both test calls together from opposite ends of the arena. The location of the sounds was switched from one end to the other between the successive tests, and then switched again before the simultaneous test. Order of testing in the successive tests was alternated between chicks to allow for possible order effects. I added the simultaneous tests because animals have sometimes been found to perform differently in such tests from the way in which they perform in successive tests, and because the two kinds of situations have counterparts in nature. A chick separated from its parents may well be more inclined to approach another adult than it would be in the presence of one of its parents, and it is plausible that such a difference in social selectivity would have survival value for the chick (cf. Evans, 1970). The test tapes consisted of a single call repeated at regular intervals. The full series of tests has yet to be completed, but already enough has been done to add to our story. I shall report on tests with crooning calls and tests with long-calls run on two age-groups of chicks: nestlings less than 48 hours post-hatching, and older chicks between 14 and 21 days posthatching. The results of the successive tests are summarized in Tables 4 and 5. These results confirm the pattern of the previous tests with different agegroups and they also confirm the hypothesis about which calls are involved in individual recognition. The nestlings responded positively to the crooning calls (Table 4, sound compared with no sound) of both the parent and the neighbor, but they showed no evidence of distinguishing the call of the parent from that of the neighbor (Table 4, parent sound compared with neighbor sound). In the long-call series the nestlings were significantly less vocal and spent significantly more time crouched in the neighbor tests than in the parent tests, and thus demonstrated individual recognition of the 70

Table 4. Individual Recognition Tests with Crooning and Long-Calls in Parent-Raised Laughing Gull Chicks less than 48 Hours Post-Hatching3

Response Orientation To speaker . . . To other end . . Position Speaker side . . Other side . . . . Change . . . Calling Total calls Crouching

nns b

nns vs. ns c

nps vs. psc

n.s. n.s.

<.01 n.s.

<.01 n.s.

n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

<.01 n.s. <.01 n.s. n.s. n.s.

<.01 n.s. <.01 n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s.

.01 n.s.

<.05 n.s.

n.s. n.s.

0 0 17.0 n.s. 0 0 0 n.s. 1.0 1.0 3.0 n.s. 16.0 17.0 20.0 <.05 60.0 56.0 161.0 <.05 20.0 20.0 11.0 n.s. Crooning vs. Long-Call"1

n.s. n.s. n.s. n.s. n.s. n.s.

.02 n.s. <.01 <.02 <.01 <.01

n.s. n.s. n.s. <.02 <.02 <.05

ns"

npsb

ps"

ns vs. psc

Crooning (N - 22) 0 0 0 0

0

14.5 59.5

1.5

Orientation To speaker . . . 0 To other end . . 6.0 Position Speaker side . . 0 Other side . . . . 0 Change 1.0 Calling . 20.0 102.0 Total calls Crouching 20.0 Orientation To speaker . . . To other end . . Position Speaker side . . Other side Change Calling Total calls . . . . .

nns vs. nps c

12.5 1.0

0 0.5

11.0 0.5

141.5 0 93.0 0 0 0 8.5 0 12.5 19.5 20.0 20.0 106.0 118.5 106.0 0 0 0 Long-Call (N = 17) 4.0 0

4.0 1.0

n.s. n.s.

.006 n.s.

n.s. n.s. <.001 n.s. n.s. n.s. .008 n.s. n.s. ns n.s. n.s. <.001 <.001 <.001

.004 n.s. .008 n.s. <.001

n.s. n.s.

n.s. n.s.

0 0

" The scores are medians from 5-min observation periods. Position was scored according to a grid on the floor of the arena, the cells of which were numbered from 0 to 12 reading from the center to either end. Hence, a chick that moved to one end of the arena at the start of a 5-min observation and stayed there would score 240 (12 x 20). Scores for the speaker side of the arena and the other side were tallied separately. If a chick occupied different cells in successive 15-sec intervals, it was scored as having changed position, and the tally is an index of locomotion (maximum possible score = 20). Orientation, crouching, and calling are, similarly, the numbers of 15-sec intervals in which these were noted (maximum possible score = 20 for each). Total calls is a count of all vocalizations given by a chick in a 5-min observation period. b nns = neighbor tests, no sound, ns = neighbor test, sound, nps = parent test, no sound, ps = parent test, sound. c Significance according to Wilcoxon matched-pairs comparisons (two-tailed). d Significance according to Mann-Whitney U tests (two-tailed).

71

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY long-calls of their parent. Comparison of the results of the crooning and the long-call tests of the nestlings showed that crooning produced significantly higher scores for approach (position near the speaker) and for locomotion (position change), and long-calls produced significantly higher scores for crouching. The tendency for long-calls to induce crouching carTable 5. Individual Recognition Tests with Crooning and Long-Calls1 in Parent-Raised Laughing Gull Chicks 14-21 Days Post-Hatching'

Response Orientation To speaker To other end . . , Position Speaker side . . . Other side . . . . Change .... Calling Total calls Crouching Orientation To speaker To other end . . . Position Speaker side . . . Other side Change Calling Total calls Crouching Orientation To speaker To other end . . . Position Speaker side . . . Other side Change Calling Total calls Crouching

nns

b

b

b

ns

nps

b

ps

nns vs. npsc

nns vs. nsc

nps vs. psc

ns vs. ps°

n.s. n.s.

n.s. n.s.

n.s. n.s.

n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

Crooning (N — 22)

0 0

0 0

0 0

1.5 0

0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Long-Call (N-18)

0 0

0 0

0 0

13.5 0.5

n.s. n.s.

n.s. n.s.

<.01 n.s.

<.01 n.s.

0 0 0 0 0 0

0 0 0 0 0 0

0 0 0 0 0 0

81.5 0 15.0 20.0 146.5 0

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

<.01 n.s. <.01 <.01 <.01 n.s.

<.01 n.s. <.01 <.01 <.01 n.s.

Crooning vs. Long-Call'1

n.s. n.s.

n.s. n.s.

n.s. n.s.

<.02 n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

<.01 n.s. <.01 <.01 <.01 n.s.

a

Scoring done as in Table 4, note a. nns = neighbor tests, no sound, ns = neighbor test, sound, nps = parent test, no sound, ps = parent test, sound. c Significance according to Wilcoxon matched-pairs comparisons, (two-tailed). d Significance according to Mann-Whitney U tests (two-tailed). b

72

C. G. BEER ried over into the period without sound that intervened between the periods with sound; hence, there was a difference between the crooning series and the long-call series in the scores for crouching in the periods without sound. In the main, then, crooning induced positive response; long-calls did not and instead induced the negative response of crouching without movement. However, the scores for vocalization did not conform to this pattern; total calls by the chicks were significantly higher in response to parental long-calls than to parental croons. This last result warns against the oversimplification that can attend the combining of several behavioral measures in a category such as positive or negative response, and the consequent tendency to assume that the measures will always vary together in the same direction, and to the same degree. However, some such simplification is necessary at every level of description and measurement. My category of chick vocalizations was heterogeneous in ways that I was unable to accommodate with my pencil-andpaper method of recording the behavior in the tests. Had I consistently been able to distinguish the several different types of chick vocalizations in my notes of the behavior, I suspect that I might have found that the composition of the calls uttered in the long-call tests was different from that of the calls in the crooning tests in ways consistent with the positivenegative response dichotomy. In any case this particular complication did not enter into interpretation of the results of the tests of the older chicks, although, as we shall see, another complication did. In the tests of the older chicks (Table 5), crooning, whether by the parent or by the neighbor, produced no response whatever — either positive or negative. Likewise, the chicks were completely unresponsive to longcalls by the neighbor. To long-calls of the parent, they showed strong positive response: turning toward the sound, approach, persistent locomotion, and calling. Loss of the power of crooning calls by themselves to command filial response in a chick older than a certain age and the passing of that power to the long-calls of the parents, which are sharply discriminated from the long-calls of other gulls, were thus clearly demonstrated. The scores for the simultaneous tests were fully in harmony with the results of the successive tests. The series for orientation and approach for the simultaneous test are shown in Table 6. The failure of my initial efforts to elicit filial responses by playing recordings of crooning calls to chicks would thus appear to have occurred not only because the crooning lacked individually identifying characteris73

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Table 6. Comparisons of Scores for Orientation and Position in Simultaneous Presentations of Parents' and Neighbors' Calls" Orientation Age Group and Test

To to Parent Neighbor Call Call

Chicks <48 hr. post-hatching Crooning tests (N = 22) . . 7.5 Long-call tests (N = 17) . . 9.0 Chicks 14-21 days post-hatching Crooning tests (N = 12) . . 0 Long-call tests (N = 18) . . 13.5

Position Parent Neighbor P* Call Side Call Side

Pb

4.5 0

n.s. <.05

51.0 19.0

33.5 0

n.s. <.05

0 0

n.s. <.01

0 84.5

0 0

n.s. <.01

" The scores are medians from 5-minute tests. According to within-test comparisons of scores using the Wilcoxon matchedpairs test (two-tailed). b

tics and the context that ordinarily provides such characteristics, but also because the chicks were beyond the age of indiscriminate response to crooning. Some of the results of the most recent series of tests indicate that, though recognition of the parent's voice may be necessary for a call to elicit filial response, it is not sufficient, for chicks beyond the nestling stage. In the series of long-call tests on older chicks, the first eight chicks responded strongly to the call of the parent, but the next two did not. The contrast between these two chicks and the others was so marked that I could not dismiss it merely as a manifestation of variation in responsiveness among chicks and the chance of sampling. It seemed more likely that the contrast might be due to differing origins of the test recordings. Whereas the first eight chicks had been tested with recordings made on the day of testing or on the day before, the two unresponding chicks had been tested with recordings made when they were nestlings — more than two weeks earlier. To investigate the possibility suggested by this conjunction, I began a series of tests in which chicks between 14 and 21 days post-hatching were given successive exposures to two recordings of long-calls of their parents: one made when they were nestlings, and one made within two days of testing. So far, only seven chicks have been tested in this series, but even with that small number the results clearly point to a difference in the effects of the two sets of recordings (Table 7). The recordings made during the nestling phase failed to evoke positive response, except in one case; the re74

C. G. BEER cordings made close to the time of testing did evoke positive response in all cases. Now the question arises, why do recordings of long-calls made during the nestling phase fail to produce the same kind of response from 14-21-day-old chicks as do recordings made during the later period? Three answers suggest themselves. First, it could be that the recordings deteriorate so much during two weeks or so that too much is lost from the original signals for the chicks to recognize the sounds. Second, it could be that the voices of the parents change during the two weeks or so following hatching, particularly in those features that convey individual identity, and the chicks perceive the differences, recognizing only the most recently recorded calls as those of their parents. Third, it could be that there are different versions of the long-call, which convey different meanings to older chicks, and that the version predominating during the nestling phase is one that does not command a chick to approach, even though the chick recognizes the call as that of one of its parents. The first possibility I take least seriously. There is no evidence independent of the chicks' behavior to support it. Both Impekoven and I have used recordings a year or more old in experiments, such as those testing effects of crooning on very young chicks, and we have found no indication Table 7. Responses of Chicks 14-21 Days Post-Hatching to Parents' Long-Calls Recorded in Nestling and Late Periods"

Response Orientation To speaker To other end . . , Position Speaker side . . , Other side Change Calling Total calls . . . . Crouching

Call from Nestling Period1' esd ensd

Call from Late Period0 lnsd lsd

ens

ens

Ins

es

vs.

vs

vs

vs

lnse

ese

lse

lse

0 0

0 0

11.0

7.0

14.0

n.s. n.s.

n.s. n.s.

<.01 n.s.

<.01 n.s.

0 0 0 0 0 0

0 0 0 0 0 0

0 5.0 3.0 0 0 0

134.0

n.s. n.s. n.s. n.s. n.s. n.s.

n.s. n.s. n.s. n.s. n.s. n.s.

<.01 n.s. <.01 <.01 <.01 n.s.

<.01 n.s. <.01 <.03 <.01 n.s.

a

0 0

16.0 19.0 149.0 0

Scoring done as in Table 4, note a. N = 7. Less than 48 hours post-hatching. c Within 2 days of testing. d ens = early (nestling) period, no sound, es = early period, sound. Ins = late period, no sound. Is = late period, sound. e Significance according to Wilcoxon matched-pairs comparisons (two-tailed). b

75

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY of difference between response to these recordings and response to recordings made a short time before testing. The possibility of change in the voices of the parents cannot be so easily dismissed. Change in the tonal quality of the calls of many gulls becomes discernible as the breeding season draws toward its close, along with loss of black from the head, disappearance of red from the bill, and other signs of reversion to nonbreeding condition. However, none of these signs is apparent until well after the first three weeks post-hatching, and the features that, to me at least, constitute a long-caller's signature — the number and durations of the short notes — remain unchanged throughout this period, according to evidence we have at present. Furthermore, the chicks in the test series in question gave indications (which I cannot substantiate statistically) that they recognized the calls of their parents even when they did not react positively to them. Some of the chicks reacted negatively to the long-calls of neighbors, but in no case did they do so to the long-calls of the parent. The third possibility is the most interesting and the one that I favor. Long-calls vary in a number of features other than those that apparently convey individual identity, and these calls also occur in a variety of contexts. It is already apparent that some of the variation in long-call features is associated with differences of context. For example, long-calls exchanged between mates or prospective mates immediately after a spell apart contain more long notes than long-calls uttered in most other situations. It is therefore plausible that different information or commands are encoded in the different versions of the call. The long-calls uttered by parents during the nestling phase appear to be directed at other adults — the mate, neighbors, or foreigners — and not at the chicks. However, many of the long-calls uttered by parents in the presence of their older chicks appear to be directed at the chicks, and because these are the calls for which it is easiest to get clean recordings at this stage — since they tend not to be accompanied by long-calls of other gulls in the vicinity — I unintentionally selected such calls for my test tapes. If the version of the long-call that a parent uses to command the approach of its chicks differs from the versions used for other purposes, the distinctive features of this version should be discernible in sonagrams of the recordings. An investigation of this is in process but it has not proceeded far enough for me to be able to say anything definite. In the meantime I have continued the experimental approach to whether different versions of the long-call convey different meanings. 76

C. G. BEER I have alluded to the fact that during the time that a parent spends in the vicinity of its chicks, after they have reached the age of discretion, only some of the parent's long-calls are directed at the chicks and the rest are directed at adults, As a rule, the chicks ignore any but the long-calls that appear to be directed toward them. Since the parent is often out of sight of the chicks when it long-calls, whether or not a chick responds positively to the call — assuming the chick to be normally responsive — must be determined by the call itself, and not by the accompanying visual signs that might qualify the meaning of the call. To test this reasoning, and hence the more general thesis that different versions of the long-call are used in different situations to convey different meanings, I began another series of playback tests. For this series I again selected two classes of long-call recordings. One class consisted of recordings of calls that were apparently directed by parents toward their chicks and to which the chicks had responded positively in the field. The other class consisted of calls that were apparently directed at adults and to which the chicks had not responded positively in the field. Chicks between 14 and 21 days post-hatching were given successive tests with each of the recordings of the parent and a recording of a call directed by a neighbor at its chicks. Table 8. Responses of Chicks 14-21 Days Post-Hatching to Parents' Long-Calls Directed and Not Directed to Their Chicks"

Response Orientation To speaker . . . . To other end Position Speaker side . . Other side Change Calling , Total calls . Crouching

Call Directed at Chick nsab sab

Call Not Directed at Chick nsb" sb"

nsbc

sac

sbc

sbc

0 1.0

n.s. n.s.

.03 n.s.

n.s. n.s.

n.s. n.s.

1.5 0 2.0 5.0 13.0 0

n.s. n.s. n.s. n.s. n.s. n.s.

.05 n.s. n.s. .03 .03 n.s.

n.s. n.s. n.s. n.s. n.s.

.03 n.s. <.03 n.s. n.s. n.s.

0 0.5

10.5 0

0 4.0

9.5 0.5 1.0 0 0 0

88.5 0 9.5 9.0 41.0 0

0 0 0 0 0 0

nsa

nsa

nsb

vs

vs

vs

n.s.

sa

vs

* All long-calls recorded within 2 days of testing. Scoring done as in Table 4, note a. AT = 8. b nsa = call directed at chick, no sound, sa = call directed at chick, sound, nsb = call not directed at chick, no sound, sb = call not directed at chick, sound. c Significance according to Wilcoxon matched-pairs comparisons (two-tailed).

77

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY So far only eight chicks have been tested in this series, and they were less consistent in their reactions to the recordings than were the chicks tested with previous and recent recordings. However, the general trend of the results is in the expected direction: the chicks tended to respond positively to the calls that their parents had directed at them, but not to the calls that their parents had directed at adults or to the calls of the neighbor. Responses to the two calls of the parent are compared in Table 8. The sonagraphic analysis, which I hope will bring to light the respects in which the two classes of call differ, has yet to be completed.

Reflections The Laughing Gull chicks in my tests revealed by their behavior that they become increasingly discerning in their responses to what they hear as they grow older. In so doing they increased my discernment of what constitutes the social behavior of Laughing Gulls. For example, they brought me to perceive that individual recognition is involved in communication that I had assumed to consist of signals and responses tuned to one another on the bandwidth of the species. Most recently, they have taught me that distinctions will probably have to be drawn within what I and others have regarded as one category of call, comprising calls that convey individual identity, since it appears that a bird may make its recognition manifest by overt response to one version of the call but not to others. In the light of these considerations, the study of the communication functions of signals and the study of individual recognition are revealed in a relationship of intimacy that I think should be a matter of concern to people involved in these kinds of study. If the meaning of a signal is given only by a recipient that recognizes the signaler as a particular individual — parent, mate, offspring, neighbor, and so forth — then failure to take this fact into account could lead either to frustration of the attempt to elucidate the function of the signal or to misinterpretation of the way it works — for example, if it were concluded that individuals are assorted randomly as responders or as non-responders. On the other hand, if the focus of attention is on individual recognition, one could be led to false conclusions if one were unaware that the signal in question occurs in several versions, not all of which elicit overt response in the kinds of individuals or in the context being observed. In short, whenever it is suspected that individual recognition might be involved in the communication function of a signal, 78

C. G. BEER the individual recognition and the other respects in which the signal has meaning need to be investigated together. This conclusion has practical implications that are not always easy to satisfy, however, the most obvious being that the investigator must be able to identify his animals individually and treat the data from each separately. But in addition to the intrinsic practical difficulties that often hinder or discourage such an approach to the study of communication, I think there have been general conceptual and methodological presuppositions contending against it in ethology. By and large, ethologists have concerned themselves about the behavioral characteristics and mechanisms of species, individuals being of interest only insofar as they contributed to knowledge of what is general to animals of their kind. Implicit in the concepts of instinct of the classical ethologists was the assumption that what applies to one must apply to all as far as the instinctive social communication of the species is concerned. With the advent of quantitative approaches to the analysis of social interactions, it became apparent that the sequential patterning of such interactions is usually a matter of probabilities rather than of obligatory and universal stimulus-response couplings. Nevertheless, the focus was still on the general rather than the particular, for the statistics dealt with the quantitative properties of populations, and the particular occurrence or the particular individual was significant only as a member of a population of such occurrences or individuals. Neglect of the individual in its particularity may have led to some odd conclusions, however. Suppose, for instance, that a stochastic analysis is made of transition probabilities derived from behavioral observations of a number of animals, and that the data are treated without regard to the individual contributions of the several animals. Suppose further that the results of this analysis are made the basis of a model purporting to represent how the behavior in question is controlled or organized. To what would this model refer? For it to be of any use to someone interested in physiological mechanisms of behavior, the model would have to refer to what obtains in each individual animal. But since the model was based on data from a population, it must refer to the rather nebulous notions of control or organization at the level of the population. It would be an open question what relationships these notions might bear to individual mechanisms. It could be that each individual in a population is extremely stereotyped in its sequential patterning of the behavior in question, but that these patterns differ from one individual to another. Then a stochastic analysis of 79

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY pooled data would give a picture of probabilistic order that would not be an accurate representation of any of the individuals contributing to the data. A similar situation could arise with the use of factor analysis on data from a number of individuals. One reason why analyses such as these have been done on pooled data rather than on individual data is that these analyses require a large quantity of information to cover the full range of behavioral permutations and for comparisons to reach the level of statistical significance. In their analysis of aggressive interactions in eight species of hermit crabs, Hazlett and Bossert (1965) collected data from 1,000 interactions for each species. They found that this was an underestimate of the amount of information needed to cover the full range of transition patterns in even two act sequences adequately, and that at least 10,000 interactions per species would be necessary for a thorough investigation of the sequential patterning in this behavior. Even this quantity of data, they said, would probably not provide enough for any sort of time analysis. I have come to feel uneasy about an approach that entails such an accumulation of evidence to make a point, especially when the point turns on measured differences so slight that only the statistical tables can make them seem real. Like Liam Hudson (1966) in his comments on the use of complex statistics such as factor analysis in the investigation of human intelligence, I am inclined to suspect that a fact needing a mountain of measurements and hair-splitting statistical differences to make itself apparent is unlikely to be of any great importance — particularly in animal communication, where too much subtlety would, in all probability, be counterfunctional. At any rate, it is my impression that the returns from such statistical approaches to the investigation of animal communication have been considerably less than would be commensurate with the investments in time, effort, and ingenuity that they have entailed. Lorenz (1952) was perhaps wiser than he fully realized to regard quantitative methods as a mixed blessing for ethology. In spite of the increased rigor and sophistication in experimental design, measurement, and inference that these methods have brought to ethology, ethology may have suffered because of the extent to which servitude to statistical style has tended to restrict perception and imagination. I am now more open than I was formerly to the positive side of Lorenz's injunction about the primacy of observation in the study of behavior. The relatively innocent but sharp eye of the bird watcher may be open to the existence of phenomena to which the statistically 80

C. G. BEER minded scientist, with all his sophistication, turns a blind eye. The remarking of exceptions to expectation, and the reporting of anecdotes, can still feed new life into our science. However, a quantitative and experimental approach to animal behavior is not, of course, incompatible with observation and description of natural phenomena, even when the focus is on individuals in their particularity. Indeed, the two approaches need each other and should be complementary. At least in my own case, research has consisted of an oscillation between the field and the laboratory, observations in the field setting questions for experiments, the results of which have led to a sharpening of the original observations and the making of new observations. Measurement and statistical analysis have provided means but not an end in the pursuit of this endeavor. Preconceptions have exerted their influence throughout this research, of course, and one of these has been a conviction of the importance of context in the functioning of signals that was first impressed upon me by Gilbert Manley's work. However, my conception of context has not remained static since it has had to accommodate to the assimilation of the results of the research. At first, I recognized only a few coarse contextual-categories, such as agonistic situations, courtship situations, situations involving parents and young. Then I saw the need to distinguish within these categories the contextual differences associated with the different dramatis personae of each, in consequence of which, for example, the same signal will evoke different response in an agonistic situation depending upon whether the recipient is at home or abroad, an ally or a rival, a male or a female. The observations and experiments on vocal communication required the distinctions to be drawn even finer. It became apparent that the meanings of certain calls — their influences on the behavior of recipients — are affected by what other sounds precede and follow them, that is, by differences of context within the communication channel. Even within the fabric of a call itself (at least for some kinds of call), the carrying of more than one kind of information gives rise to relationships that can be construed as contextual. For example, in the case of calls that convey both individual identity and a command, whether and how the command is obeyed may depend upon whether the voice is recognized; moreover, whether recognition of the voice is made overt may depend upon the nature of the command. However, to retain the notion of context through all of these transitions threatens to strain the concept beyond usefulness and runs the risk of blur81

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ring distinctions that probably should be kept clear. The concept of syntax is perhaps more apt for the kinds of relationship that affect meaning from within the communication medium, and there is precedent for this usage in the terminological recommendations of Charles Morris (1938) and in the literature on semiotics that has grown from them. Moreover, by describing these relationships as syntactical, we can reserve the concept of context for those influences that affect the meaning of a signal from without the communication channel, such as the environmental setting in which communication occurs. However, to use the term syntax in this way may also be to encourage blurred distinctions and false analogies. The term originally pertained only to the structure of human language. Extending It to animal communication in general is perhaps likely to be taken as implying that language differs from animal communication only in degree, not in kind. This would be unfortunate, because this question of the relationships between language and animal communication is still a very live issue, particularly so since a chimpanzee began making contributions to it (Gardner & Gardner, 1971). However, one does not have to beg the question to find in the structure of language a source of concepts or ideas that might prove useful in ordering thought about communication in animals. I am attracted to this possibility because it seems to me now that neither the venerable stimulus-response conception of animal communication, nor the more modish conception of animal communication as a finite state process, can do full justice to all that gulls give evidence of doing in their social interactions. It is at least conceivable that rules, analogous to the syntactic and semantic rules of language, govern the use of at least some of the signals of these animals. If so, discovery and formulation of these rules would be a necessary first step to full understanding of gull communication in all its aspects. This step has yet to be attempted in earnest. But already the linguistic analogy has suggested that the categories that have been used in describing the social behavior of gulls might provide a starting point. By and large I have worked with the same categories of calls and postures that most other observers of gull behavior have used since they were distinguished in the pioneering studies of Kirkman (1937) on Blackheaded Gulls and Noble and Wurm (1943) on Laughing Gulls, and refined and supplemented in the later work of Tinbergen (1953, 1959) and Moynihan (1955, 1956, 1958, 1962). These categories were based mainly on features of sound or form that strike the human ear or eye as discrete 82

C. G. BEER and recurring patterns, in some cases qualified or further specified by association with a particular context (e.g., Noble and Wurm qualified "choking" as "hostile" or "friendly" depending upon whether it occurred in an agonistic or some other context, although it was claimed that formal features also distinguished the two types). It is possible, however, that in some respects the ethologist's anatomy of gull social signals does not match that of the gulls themselves. In fact, if the calls and postures were to be sorted out solely on the basis of the effects they have in natural interactions and in playback tests, a set of categories rather different from that generally assumed would result. I suspect that these seemingly conflicting criteria for categorizing the communication behavior of the gulls might be at least partly reconciled if something like the linguist's distinction between phonemes and morphemes were observed: formal elements of utterance and posture on the one hand, and the combinations of them that carry information on the other. For example, the notes of different types that can be distinguished in the long-call of the Laughing Gull also occur in other combinations, and even by themselves in the cases of long notes and head-toss notes; but I have been able to find no common thread of meaning linking all the occurrences of such a type of note. The same applies to the long-call itself, apart from the function of conveying individual identity, but the reason in this case may be that what has been regarded as a single and discrete type of call is, in effect, a syntactic structure, the semantic content of which can be varied. This kind of preoccupation with linguistic analogies is not peculiar to me among students of animal communication. Indeed, I suspect that a general swing in this direction is under way. For example, the possible usefulness of the phoneme/morpheme distinction has also occurred to Millicent Ficken (personal communication) in her investigation of the song of the Black-capped Chickadee. Whether such comparison with language will profit the study of animal communication remains to be seen. In the meantime it serves my purpose by illustrating how one's conception of a subject can be affected by the models or metaphors (cf. Black, 1962) one brings to it, and how these tend to change as knowledge of the subject advances. Ethology as a whole has undergone evolution — both linear transformation and adaptive radiation — as guiding sets of ideas or ideals have vied with and succeeded one another. Consequently, an outsider's conception of the science is likely to be last year's model. 83

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY One of the best recent reviews of ethological thought was written by a psychiatrist, John Bowlby (1969), but it is already dated in some respects because of the way in which socio-ecology of the type introduced by John Crook (1970a&b) has since swept into large areas of the science. I suspect that we may be in for some revisions of our ideas about evolution and development when ethology assimilates some of the new thinking that is being directed at these matters (e.g. Piaget, 1971; Waddington, 1968, 1969). Consequently, Bowlby's argument that "protection from predators is by far the most likely function of attachment behavior" (p. 226), in humans, would probably no longer be debated along the lines he assumed. Instead of regarding selection pressures and functions in this simple atomistic way, the trend now is to view the adaptations of an organism as an integrated whole in which each feature may be the outcome of interaction between multiple selection pressures. Further, each feature may relate to several functions — much in the way that genotypes and gene pools are now regarded as integrated wholes in which the rule is that genes have multiple effects and conspire together in the production of characters. To sum up my present opinion, I think ethology is still too much in a state of flux to be a reliable source of generalizations or doctrine applicable to human behavior in general or child psychology in particular. Some of its methods and concepts may be, and indeed have been, useful in suggesting new approaches or casting old facts in new light when applied to the human case, as the books by Bowlby and by McGrew (1972) bear witness. But the study of animal behavior is as much in need of new ideas as is the study of people and children. Indeed, since, in some respects, we know the behavior of our own species considerably better than we know that of any other, one might expect the traffic in ideas to be stronger from human psychology to ethology than in the other direction. I am inclined to think that the campaign against anthropomorphism and anecdotalism, waged since the days of Romanes and Lloyd Morgan, has been overdone, and may have kept the ethologist from a useful source of ideas. Be that as it may, it is my belief that, at present, the best that a bird watcher can offer the student of child psychology is a view of birds. References Beer, C. G. Incubation and nest-building behaviour of Black-headed Gulls IV: Nest-building in the laying and incubation periods. Behaviour, 1963, 21, 155-176. Adaptations to nesting habitat in the reproductive behaviour of the Blackbilled Gull Lams bulleri. Ibis, 1966, 108, 394-410. 84

C. G. BEER . Laughing Gull chicks: Recognition of their parents' voices. Science, 1969, 166,1030-1032. . On the responses of Laughing Gull chicks (Larus atricilla) to the calls of adults I. Recognition of the voices of the parents. Animal Behaviour, 19JO, 18, 652-660.(a) . On the response's of Laughing Gull chicks (Larus atricilla) to the calls of adults II. Age changes and responses to different types of call. Animal Behaviour, 1970, 18,661-677. (b) . Individual recognition of voice in the social behavior of birds. Advances in the Study of Behavior, 1970, 3, 27-74. (c) . Individual recognition of voice and its development in birds. Proceedings of the XVth International Ornithological Congress (The Hague, 1970), 1972. Pp. 339-356. Black, M. Models and metaphors. Ithaca, N.Y.: Cornell University Press, 1962. Bowlby, J. Attachment and loss. Vol. 1. Attachment. London: The Hogarth Press and The Institute of Psycho-analysis, 1969. Crook, J. H. (Ed.) Social behaviour in birds and mammals. London & New York: Academic Press, 1970. (a) . Social organization and the environment: Aspects of contemporary social ethology. Animal Behaviour, 1970, 18, 197-209. Cullen, E. Adaptations in the Kittiwake to cliff-nesting. Ibis, 1957, 99, 275-302. Evans, R. M. Imprinting and mobility in young Ring-billed Gulls Larus delawarensis. Animal Behaviour Monographs, 1970,3, 193-248. Gardner, B. T. & Gardner, R. A. Two-way communication with an infant chimpanzee. In A. M. Schrier and F. Stollnitz (Eds.), Behavior of non-human primates. Vol. 4. New York & London: Academic Press, 1971. Gottlieb, G. Development of species identification in birds. Chicago: University of Chicago Press, 1971. Hailman, J. P. The ontogeny of an instinct. Behaviour Supplement, 1967, 15, 1-159. Hampshire, S. Thought and action. London: Chatto & Windus, 1959. Hazlett, B. A. & Bossert, W. H. A statistical analysis of the aggressive communications systems of some hermit crabs. Animal Behaviour, 1965, 13, 357-373. Hudson, L. Contrary imaginations. London: Methuen, 1966. Impekoven, M. Prenatal experience of parental calls and pecking in the Laughing Gull (Larus atricilla). Animal Behaviour, 1971, 19, 471-476. . The response of incubating Laughing Gulls (Larus atricilla) to calls of hatching chicks. Behaviour, in press. & Gold, P. S. Prenatal origins of parent-young interactions in birds — a naturalistic approach. In G. Gottlieb (Ed.), Developmental studies of behavior and the nervous system. Vol. 1. New York & London: Academic Press, in press. Kirkman, F. B. Bird behaviour. London: Nelson, 1937. Lorenz, K. Z. 1950. The comparative method in studying innate behaviour patterns. Symposia of the Society for Experimental Biology, 1950, 4, 221-268. . Die Entwicklung der vergleichended Verhaltensforschung in den letzten 12 Jahren. Verhandlungen der Deutschen Zoologischen Gesellschaft in Freiburg, 1952, 36-58. (a) . King Solomon's ring. New York: Crowell, 1952. (b) McGrew, W. C. An ethological study of children's behavior. New York & London: Academic Press, 1972. Morris, C. Foundations of the Theory of Signs. Chicago: University of Chicago Press, 1938. Moynihan, M. Some displacement activities of Black-headed Gulls, Behaviour, 1953, 5, 58-80.

85

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY . Some aspects of reproductive behaviour in the Black-headed Gull (Lams ridibundus L.) and related species. Behaviour Supplement, 1955, 4, 1-201. . Notes on the behavior of some North American gulls. I. Aerial hostile behavior. Behaviour, 1956, 10, 126-178. . Notes on the behavior of some North American Gulls. III. Pairing behavior. Behaviour, 1958, 13, 112-130. . Hostile and sexual behavior patterns of South American and Pacific Laridae. Behaviour Supplement, 1962, 8, 1-365. Nice, M. M. Development of behavior in precocial birds. Transactions of (lie Linnaean Society of New York, 1962, 8, 1-211. Noble, G. K., & Wurm, M. The social behavior of the Laughing Gull. Annals of the New York Academy of Sciences, 1943, 45, 179-220. Piaget, J. Biology and knowledge. Chicago: University of Chicago Press, 1971. Popper, K. R. The logic of scientific discovery. New York: Basic Books, 1959. . Conjectures and refutations. London: Routledge & Kegan Paul, 1963. Siegel, S. Nonparametric statistics. New York: McGraw-Hill, 1956. Smith, W. J. Vocal communication of information in birds. American Naturalist, 1963,97, 117-125. . Message, meaning and context in ethology. American Naturalist, 1965, 99, 404-409. . Communication and relationships in the genus Tyrannus. Publications of the Nuttall Ornithological Club, 1968, No. 6. Teitelbaum, P. Physiological psychology. Englewood Cliffs, N.J.: Prentice-Hall, 1967. Thorpe, W. H. Perceptual basis for group organization in social vertebrates, especially birds. Nature (London), 1968, 220, 124-128. Tinbergen, E. A., & Tinbergen, N. Early childhood autism — an ethological study. Zeitschrift fi/r Tierpsychologie, sup. 10, 1972, 1-53. Tinbergen, N. The study of instinct. Oxford: Oxford University Press, 1951. . The Herring Gull's world. London: Collins, 1953. . Comparative studies of the behavior of gulls (Laridae): A progress report. Behaviour, 1959, 15, 1-70. Tschanz, B. Trottellummen. Beiheft zur Zeitschrift fiir Tierpsychologie, 1968, 4, 1-103. Waddington, C. H. (Ed.) Towards a theoretical biology 1. Prolegomena. Edinburgh: Edinburgh University Press, 1968. . Towards a theoretical biology 2. Sketches. Edinburgh: Edinburgh University Press, 1969.

86

3> N O R M A D. F E S H B A C H <S>

Cross-Cultural Studies of Teaching Styles in Four-Year-Olds and Their Mothers

THE series of investigations presented in this paper stem from a clinical and research interest in the relationship between parental socialization practices and the child's functioning in school, and the processes by which the school functions as a socializing agent. A related interest is the role of individual differences in the use of and response to variation in types of reinforcement. The common core of this interest pattern is the interaction between the socialization practices of the home and of the school, with the child being the active mediator between the two. Individual differences in response dispositions among children, resulting from differences in early socialization experiences, are presumed to interact with variations in teachers' programs and instructional styles. The studies described here represent efforts to delineate dimensions of parent, student, and teacher behaviors relevant to this interactional model and to specify relationships among these behaviors. These relationships are examined in the context of social class and ethnic differences within our own culture and in two other national settings. But first, it may be helpful to explicate several theoretical assumptions and methodological strategies guiding our approach to this research probNOTE: The early studies reported in this paper were supported by Contract 4-6-061646-1909 from the Office of Education, HEW, to the Research and Development Center, School of Education, University of California, Los Angeles, and the cross cultural research and the studies with parents of slow and problem readers was facilitated by UCLA faculty grants. I would like to thank Seymour Feshbach, Anne Singer, and Robert Singer for their critical and helpful reading of the manuscript.

87

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY lem. The most important initial decision was the choice of the response dimension — reinforcement style. Parents and children influence each other in many ways, children vary in a large number of personality traits, and teacher behaviors vary on many dimensions. The use of positive and negative reinforcement was seen as a response dimension that could be applied meaningfully to parent, teacher, and child behavior and that also would have explanatory value. A number of considerations went into the choice of this dimension. First, reinforcement is central in the learning process, having both motivational and informational functions. Second, there is much clinical and experimental evidence suggesting that the effects of positive reinforcement upon learning and performance are qualitatively and quantitatively different from the effects of negative reinforcement. These studies range from experimental analyses of achievement motivation (Weiner, 1972) to investigations of schizophrenics' avoidance responses to signs of maternal censure (Rodnick & Garmezy, 1957). The child who is motivated by negative reinforcement runs on a different course than the child who is motivated by positive reinforcement. A third consideration in selecting the response dimension is the pervasiveness of reinforcement in social interaction. Teachers provide rewards and punishments, praise, and criticism in diverse academic contexts. Similarly, parents use rewards and punishments in a variety of social contexts; when they are training independence, when they are disciplining aggression, and when they are teaching a child how to button his clothes. Parental socialization practices and teacher instructional styles share important functions. The process of education and the process of socialization have some common characteristics and, at times, the processes are indistinguishable. Although society ostensibly assigns the teacher specific responsibility for the cognitive domain, and the parent specific responsibility for the social and affective domain, there is obviously considerable overlap in function. Teachers are concerned with the development of social skills, with the strengthening of behavioral controls, and with the acquisition of moral values. Parents shape their child's language performance and, directly or indirectly, teach their child to make appropriate discriminations and causal inferences. Many parents spend hours reading to their children and directly instructing them in matters clearly cognitive. This commonality between education and socialization makes the teacher a parent and the parent a teacher. The fact that teachers and parents share functions is not meant to imply 88

N O R M A D. FESHBACH that the parent is a teacher in the narrow sense of providing instruction in academic skills pertinent to school achievement, although there are a number of current formulations of "parent as teacher" in this specific sense (e.g., Gordon, 1970; Schaefer, 1971). However, from a broader perspective, one can view almost all of socialization as education. Parents are teaching their child when they provide guidance, when they train for selfreliance and aggression control, when they instruct the child in proper handling of fork and spoon — no less than when they teach the child to count, to solve a simple puzzle, or to discriminate among the letters of the alphabet. In the context of this research program, all of these behaviors are seen as being able to be taught through the application of positive or negative reinforcements. The type of reinforcement employed by the parent is assumed to reflect a personality disposition or response style which characterizes many different teaching interactions. The validity of this assumption is, of course, an empirical question. In addition to reinforcement style, social class and ethnicity were assessed in most of the studies to be reported. Social class was of interest because of its association with both cognitive performance and socialization practices as documented in the extensive literature relating economic level and academic functioning (Deutsch, Katz, & Jensen, 1968; Jensen, 1969) and in the substantial literature reporting differences in behavior patterns and socialization practices in parents of different social classes (Becker, 1964; Bronfenbrenner, 1958; Kohn, 1963; Miller & Swanson, 1960; Sears, Maccoby, & Levin, 1957). However, a considerable theoretical gap exists between the socialization practices used by parents of different social classes and the cognitive performance of children from these social groups. Loosely defined constructs such as "cultural deprivation" which have been used to fill the gap have proved to have limited utility and are even considered by some investigators to be more misleading than helpful (e.g., Baratz & Baratz, 1970; Cole & Bruner, 1972). In my judgment, studies such as those stimulated by Bernstein (1961), in which an attempt is made to link directly specific patterns of parental behavior to the child's level and type of cognitive skill, are necessary for the identification of critical intervening processes. Bernstein's work (1961), which emphasized the importance of maternal speech as a reflection of social context shaping the child's cognitive development, provided an impetus to the investigation of mechanisms mediating the effects of social class upon the child's cognitive performance. 89

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Bernstein identified two modes of linguistic style, the elaborate formal code found to be more characteristic of the middle-class parent, and the restricted code (manifested in a rigid and limited grammatical usage) found to be more typical of the lower-class parent. Bernstein's formulations regarding the relation between language modes and the social class structure in England have been supported empirically in American families by the research of Hess and Shipman (1965, 1967) and of Bee et al. (1969). More recently, however, linguists have challenged the cognitive implications of the two modes of linguistic style by questioning whether the two linguistic modes actually differ in communicative value. In addition, observed variations in linguistic behavior of black children, as a function of the social context in which their speech is assessed, raise a further question as to the generality of these linguistic modes (Labov, 1970). Nevertheless, the research stimulated by Bernstein has made an important contribution by specifying processes by which social class membership can influence cognitive functioning. The importance of linguistic styles as mediators can only be assessed by additional research. The same can be said of the behavioral parameter of reinforcement style which has been the object of my own research. However, I shall try to demonstrate that the concept of reinforcement style has generality across cultures and is a useful link between cognitive performance and social class or ethnic group. IMITATION AND TEACHERS' REINFORCEMENT

STYLES

The initial investigations of reinforcement styles in our laboratory entailed the experimental manipulation of positive and negative reinforcement and the assessment of the effects of this variable upon various indices of learning. The concern in these early studies was with the effects of differences in teacher's reinforcement style and in children's socioethnic background upon the imitative behavior of elementary school age children. The paradigm used in these studies (Feshbach, 1967a&b; Portuges & Feshbach, 1972) was children's observation of teachers who used different modes of reinforcement and who manifested distinctive mannerisms incidental to the lesson presentation. The children's subsequent display of these incidental mannerisms was the measure of imitation. The teacher models were presented in two four-minute films. One film featured a positive reinforcing teacher, and the second showed a negative reinforcing 90

N O R M A D. FESHBACH teacher, each giving a geography lesson on Africa. After the child observed the two experimental films, counterbalanced for order of presentation, he or she was then required to teach a lesson like that observed in the films. In each of two experiments in which this procedure was used, a control group, consisting of children who were not shown the film but required to perform the same teaching task, was included to provide a base line for assessing degree of imitation. In the first study, the participants were nine- to ten-year-old boys with learning problems; in the second study, the subjects were a more representative group of both boys and girls. In each study, there was a middle-class sample which was predominantly white and a less economically advantaged sample which was predominantly black. The results of the two studies were highly consistent. There was practically no imitation of the negative reinforcing teacher but considerable imitation of the positive reinforcing teacher. However, only the middle-class children imitated the positive teacher more than they imitated the negative teacher; the lower-class children showed little imitation of either teacher. The issues of ethnic differences in children's imitation is intriguing (Feshbach & Feshbach, 1972) but peripheral to the principal focus of the present paper. Of greater relevance here is the finding that the teachers' reinforcement style exerted a strong influence on one index of learning. The fact that we found little or no observational learning of a negative reinforcing teacher could have important educational implications. A number of reports of teacher attitudes and behaviors in ghetto schools suggest that the learning environment in these schools is permeated with negative reinforcement (e.g., Kozol, 1965). Frequent resort to negative reinforcement by teachers and administrators may exaggerate existing avoidance patterns in the children, patterns which are incompatible with the assimilation of academic content and desired behavioral norms. There is evidence that children from economically disadvantaged backgrounds, before school attendance, have already been exposed to a negative reinforcing environment conducive to the development of avoidance behaviors. Reports from a number of studies of child-rearing practices have indicated that working-class parents tend to make greater use of physical punishment and related negative reinforcement than do middle-class parents (Becker, 1964; Bronfenbrenner, 1958). It seemed also possible, if these negative reinforcing behaviors are as pervasive as is indicated, that 91

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY through modeling and imitation, these negative behaviors might even extend to transactions among the children themselves. REINFORCEMENT STYLES IN FOUR-YEAR-OLDS

Our next step, then, was to investigate patterns of reinforcement used by lower- and middle-class children in interaction situations involving cognitive learning concomitants (Feshbach & Devor, 1969). For these studies we required an age group sufficiently young that evidence of social class variations in behavior could not be attributed to school influences. Furthermore, the earlier the age at which social class and ethnic differences in reinforcement styles could be established, the greater could be the confidence in this variable as a contributor to cognitive differences between children of different social classes. Thus, there were methodological and theoretical constraints limiting the choice of the age of the sample. The children had to be at a level of social maturity and linguistic competence sufficiently advanced to permit the expression and assessment of positive and negative reinforcement. In addition, since imitation was a possible mechanism for the acquisition of reinforcement styles, the age level had to be one at which modeling effects had been demonstrated. For these various reasons, we decided to study four-year-old children to see if they demonstrated differences in patterns of reinforcements as a function of socioeconomic background. In order to observe the children's reinforcement behavior, each fouryear-old was asked to teach a three-year-old a simple puzzle. Previously, the experimenter demonstrated the puzzle to the four-year-old. During this familiarization period, the four-year-old teacher-child was given three trials in which to assemble the puzzle. During the first trial, the child was given active verbal assistance by the experimenter. During the second trial, the experimenter made one positive verbal remark, "that's very good," and one critical remark, "that's not right," about the child's performance. At the beginning of the third trial, the child completed the puzzle while the experimenter left to get the three-year-old pupil. After being introduced to the younger child, the four-year-old was given the following instructions. "Now it's your turn to be the teacher. You're going to teach Andy how to do this puzzle. What does teach mean? It means to help. You may help Andy by using words but not your hands. Now you may begin by telling Andy how to do the puzzle." The four-year-olds had no difficulty in understanding the teaching task 92

NORMA D. FESHBACH they were being asked to perform, and most of them entered into the teaching role with great seriousness of purpose. Like their parents and their teachers, they wanted their three-year-old pupils to learn. One could see in these four-year-old teachers reactions of frustration, pleasure, disappointment, and excitement as their pupils shifted back and forth from persistent errors to flashes of insight. The children's active engagement in the teaching task resulted in a range of verbal reinforcement behaviors which were recorded verbatim and which could readily be scored as positive or negative. Positive and negative reinforcing statements were defined in terms of their encouraging and discouraging connotations, rather than in the more formal, restricted sense of increasing or decreasing a specific response. The positive category included statements of praise, encouragement, and affirmation; the negative category included criticism, negations, and derogatory comments. Typical examples of positive comments were: "See, she can put it together." "That's better." "Yeah, like that. He did it." Frequent negative comments were: "Wrong way." "Not that way." "No, don't bang it." "You stupid." To determine the reliability of this dichotomous classification, eighty randomly selected statements were scored by two independent raters. There was only one instance in which the raters disagreed. The total number of positive and total number of negative statements were determined for each child and constituted the basic dependent measure. In addition, the performance of the three-year-olds was assessed in most of the studies by the number of errors made and time taken to complete the puzzle. The four-year-olds in our first study were 50 boys and 52 girls, approximately equally distributed by ethnicity and social class. There were four combinations of race and social class: middle-class and lower-class whites and middle-class and lower-class blacks. The middle-class groups attended private nursery schools, lived in neighborhoods identified as middle to upper-middle class, and their fathers' occupations were professional and managerial. The lower-class children were enrolled in children's centers, lived primarily in neighborhoods identified as dis advantaged, and their parents were engaged in unskilled or semiskilled occupations. In all instances, the ethnicity and social class of the three-year-old pupil was the same as that of the four-year-old teacher. In this first study, the sexes of the two children were varied so that within each grouping, half the pupils were of the same sex and half were of the opposite sex as their 93

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY four-year-old teachers. However, the variation in sex of pupils did not affect the outcome and consequently will not be elaborated in the presentation of the results. The mean frequencies of positive reinforcements used by each group were as follows: middle-class white children, 2.3; middle-class black children, .2; lower-class white children, .8; and lower-class black children, .7. Middle-class white children used a significantly greater number of positive reinforcements than did any of the other three groups. In contrast, the middle-class black children used significantly fewer positive reinforcements than either of the two lower-class groups. The overall pattern of social class and race differences was reflected by both sexes, and the strikingly greater use of positive reinforcement by the middle-class white boys and girls was consistent with expectation. The results for the middle-class blacks are an interesting deviation from the hypothesized relationship between social class and use of positive reinforcement. The mean frequencies of negative reinforcement used by each group were as follows: middle-class whites, 1.6; middle-class blacks, 1.3; lowerclass whites, 1.7; and lower-class blacks, 2.1. None of the group differences were statistically significant. However, if the two ethnic groups are combined and the children are categorized by those who did and did not use negative reinforcement, an interesting social class difference emerges. Of the 49 middle-class four-year-old teachers, 21 in comparison with only 10 of the 46 lower-class teachers did not use any negative reinforcement. Another way of viewing the data is to examine the relative frequency with which positive and negative reinforcement is used by each group. Here we find the greatest difference to be between middle-class whites and lowerclass blacks. The middle-class white sample was the only group to use more positive than negative reinforcements when instructing the threeyear-olds, whereas the lower-class black children displayed the largest preponderance of negative over positive reinforcements. It is reasonable to ask whether these differences in the children's teaching behaviors could be attributed to differences in pupil performance. An analysis of the mean errors and of mean times to solution failed to reveal any significant performance differences among the groups. In addition, neither pupil performance measure was significantly correlated with either the number of positive or the number of negative reinforcements administered by the teacher-child. The possible role of linguistic factors in the pattern of results is also of 94

NORMA D. FESHBACH interest. Here, too, the performance of the various groups was comparable. There were no significant differences among the four-year-old groups in number of words used to make a positive or a negative reinforcing statement or in the grammatical complexity of the statements that were made. A possible explanation of the findings in terms of greater verbal facility on the part of the middle-class white child is rendered particularly unlikely by the fact that the mean number of words (3.5) used to make a negative comment was greater than the mean number of words (2.3) used to make a positive reinforcing statement. The results of this first study of children's teaching styles suggested that as early as age four, children display different reinforcement patterns as a function of their socioethnic background. The findings were generally consistent with the assumption that the styles of reinforcement used by these four-year-olds are modeled after the socialization practices which they experience at home. The results also support the selection of reinforcement style as a fruitful dimension of individual difference which relates variations of socialization practices in the home and in the school with variations in cognitive performance. There have been studies of social class differences in learning in which the differential responsiveness of lower- and of middle-class children to various types of positive reinforcement and to praise and reproof has been emphasized (Kennedy & Willicut, 1964; Stevenson, 1967). The differences found here in reinforcement styles in fouryear-old children suggested that a fuller understanding of social class and ethnic differences in learning required an assessment of the typical reinforcement contingencies present in the child's home and peer environments. REINFORCEMENT STYLES IN FOUR-YEAR-OLDS AND THEIR MOTHERS

We next turned to an examination of social class and ethnic differences in parental reinforcement styles, specifically, in the type of reinforcement used by mothers in instructing their own four-year-old children (Feshbach, 1972). In addition, we wanted to replicate the results of the first experiment so that we might relate type of maternal reinforcement with the type used by the child. The second study included 109 four-year-olds and their mothers. An equal number of three-year-olds served as pupils. The subjects were again divided into four groups, based on social class and ethnicity: middle-class white, lower-class white, middle-class black, 95

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY and lower-class black. The mothers and children were drawn from eleven parent educational preschool centers distributed throughout Los Angeles. These are preschools run by the city which are attended one day a week by the child and his mother. The middle-class sample in this second study was less privileged and the lower-class sample less disadvantaged than their counterparts in the first study. For the initial instructional sequence, we followed the same procedure as before, the four-year-olds' being asked to teach a puzzle to three-yearolds. A second instructional sequence occurred about an hour after the first. During this second session each mother was asked to teach her own four-year-old a similar but more complex puzzle than the puzzle used for the three-year-olds. The same measures used to analyze the child-child interactions were used to analyze the mother-child interactions. The reliability of the categorization of reinforcements was again very high, there being only one disagreement for sixty randomly selected statements. The mean frequency of positive reinforcements used by each of the four-year-old groups while teaching the three-year-olds was as follows: middle-class white children, 2.4; middle-class black children, .08; lowerclass white children, 1.9; and lower-class black children, .08. The pattern of these data, except for the relatively high number of positive reinforcements employed by lower-class whites, is similar to the pattern observed in the earlier study (Feshbach & Devor, 1969). Both middle- and lowerclass white four-year-olds used a significantly greater number of positive reinforcements than did either middle- or lower-class black children. There is less consistency between the results of the first and second study in the use of negative reinforcements than in the use of positive reinforcements by the four-year-old children. The mean frequencies of negative reinforcements for the four groups were as follows: middle-class white children, 2.3; middle-class black children, .4; lower-class white children, 2.6; lower-class black children, 1.6. Although in the first study there were no significant differences in the frequency of negative reinforcements, the white children in the second study used significantly more negative reinforcements as well as positive reinforcements. However, the patterns of reinforcement — that is, the relative frequency of positive as compared with negative reinforcement — is similar in the results of both studies. The preponderance of negative over positive reinforcement was greatest for the lower-class black children and was least for the middle-class whites. 96

N O R M A D. FESHBACH Some insight into the functional significance of the greater use of positive and negative reinforcements by the white four-year-olds is provided by an appraisal of the performance of the three-year-old pupils. Differences between middle- and lower-class white and black three-year-olds, both in time taken to complete the puzzle and in number of errors made, were small and statistically insignificant. However, pupil performance was not unimportant. For both the middle-class and lower-class white samples identical correlations of +.71 were obtained between the four-year-old teacher-child's negative reinforcement score and the number of errors made by his pupil. The corresponding correlations for the black samples were zero. The mean frequencies of mothers' use of positive reinforcements in teaching the four-year-olds were as follows: middle-class white mothers, 6.5; middle-class black mothers, 4.7; lower-class white mothers, 4.7; lower-class black mothers, 4.8. The mean frequencies of mothers' use of negative reinforcements were: middle-class white mothers, 1.4; middleclass black mothers, 2.0; lower-class white mothers, 2.2; and lower-class black mothers, 5.4. Although middle-class white mothers tended to make greater use of positive reinforcements than did the other groups of mothers, none of the mean differences is significant. However, large and significant social class and ethnic differences are found in the frequency of mothers' use of negative reinforcement. Lower-class mothers use significantly more negative reinforcement than do middle-class mothers, and black mothers use significantly more negative reinforcements than do white mothers. The ethnic differences appear to be due largely to the high frequency with which lower-class black mothers use negative reinforcements in comparison with the other groups. In contrast to the performance of the three-year-olds, ethnicity was a factor in the performance of four-year-olds. Middle- and lower-class black four-year-olds took significantly longer to complete the puzzle and made significantly more errors than did the white four-year-old groups. The mean error and time scores for lower- and middle-class groups within each ethnic category were quite similar. The frequency of the mothers' use of negative and also of positive reinforcement were significantly correlated with their children's time and error scores. Consequently, the ethnic differences found in the maternal use of negative reinforcement could be a function of the child's performance rather than a reflection of a stable maternal reinforcement style. However, an explanation of reinforcement styles in 97

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY

terms of the child's performance still would not account for the difference between lower- and middle-class black mothers in their use of negative reinforcement. To summarize, the patterns of reinforcement used by four-year-old children and by their mothers vary as a function of ethnicity and, to a lesser extent, as a function of social class. In general, middle-class white children and their mothers used relatively more positive than negative reinforcement, lower-class black children and their mothers used more negative than positive reinforcement, with the other two groups falling between these two extremes. The patterns of reinforcement used by children and by their mothers in each social class and ethnic group is but one index of degree of similarity. A more precise measure is provided by the correlations between mother's and child's use of positive and of negative reinforcement. For the total sample only, there is a small, although statistically reliable, correlation (+.22) between the child's and the mother's use of positive reinforcement. There is a correlation of similar magnitude between the child's use of negative reinforcement and the mother's use of positive reinforcement. However, neither correlation is readily interpretable since correlations for the total sample are influenced by social class and by ethnic differences in reinforcement frequencies. The meaningful mother-child correlations are those within each social class and ethnic group. These correlations are all small and statistically insignificant except for the correlation of +.65 (p<.01) between mother's and child's use of negative reinforcement in the middle-class white sample. The findings concerning mothers' and children's reinforcement style suggest that the learning environment of the lower-class black child is more stressful than the environment of his white advantaged counterpart. On the basis of the mother-child interaction data, we can infer that the lower-class black child receives more negative reinforcements from his mother than do other children; and on the basis of the child-child interaction data, we can infer that the same child receives fewer positive reinforcements from his peers than does the middle-class white child. By the age of four, the economically and socially disadvantaged child has already internalized aspects of this learning environment into his own behavioral repertoire. This child's mother is part of the environment but she is no less affected by it. Both have to cope in an environment replete with "no," "not that," "wrong," "can't you do anything right," and "what's 98

N O R M A D. FESHBACH the matter with you." It is reasonable to hypothesize that exposure to a primarily negative reinforcing environment can be disruptive of learning and can depress cognitive performance. Yet, a tenable alternative hypothesis is that these reinforcement styles merely reflect language dialect. They make it possible to distinguish between cultural groups, but may not contribute toward an explanation of group differences in cognitive skills and performance. Our understanding of the role of reinforcement style in cognitive development would be enhanced if reinforcement styles were shown to be more directly related to cognitive performance or if the relationship were evaluated in the context of other cultures. Clearly, reinforcement style is only one of many behavioral dimensions of socialization which would allow social class and ethnic differences to exert an influence on cognitive functioning and achievement. However, if the concept of reinforcement style is to have power, then there should be differences in patterns of reinforcement in other social groups who differ in cognitive achievements. Thus, our next study was carried out in Israel where one can find ethnic and economic divisions within the society associated with differences in children's academic levels. ISRAELI REINFORCEMENT STYLES

The social division that exists in Israel between Israelis of Western origin and Israelis of Middle Eastern origin is somewhat comparable to socioethnic divisions in our own society. Middle Eastern Jews, whose dominant culture has been Arabic, are economically and socially disadvantaged in comparison with Israelis of Western origin. The former have less access to the resources of the society, are less well represented in positions of status and power, have lower mean incomes, more rarely attend institutions of higher learning, and, most germane to the hypothesis under investigation, perform more poorly on various measures of academic achievement (Smilansky & Yam, 1969). The discrepancy in academic achievement between Jews of Eastern and of Western origins has been of great concern to the Israeli educators and governmental authorities. The special educational programs that have been implemented for disadvantaged Israeli children and their families (Lombard, 1971; Smilansky, 1968) have objectives and goals similar to compensatory and other intervention programs for disadvantaged groups in the United States. Given these two economically and educationally disparate Israeli ethnic groups, there was an intriguing possibility that the reinforcement styles 99

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY of four-year-olds and their mothers in the two groups would also be disparate. Consequently, it was decided to replicate in Israel the study of reinforcement styles that we had carried out with American four-year-olds and their mothers (Feshbach, 1973a). The procedures developed for the American samples were followed as closely as possible in Israel. The principal difference was that it took longer to collect the data in Israel. The complexity of collecting parental data is compounded in another culture. Also, although Israel generally supports educational research, the government exercises thorough administrative control over experimentation in the schools. We began the study after receiving local approval but subsequently had to halt the procedures while awaiting separate administrative clearance for each of the facilities from which we drew our sample. About the time the final administrative clearance was obtained, the stopwatch broke; the decision then had to be made to complete the sample while administrative approval was in effect rather than to risk waiting for the stopwatch to be repaired or for a new one to be shipped from the United States. Hence, performance measures were not recorded for the children. So much for cross-cultural research on a shoestring. Our final Israeli sample included 60 four-year-olds, their mothers, and an equal number of three-year-olds. The children were selected from eight preschools in Jerusalem and its environs. The sample was equally divided by sex and social class, the latter being determined by father's occupation. The lower-class children were predominantly from Yemenite and related ethnic backgrounds, whereas the middle-class sample were largely of Western origin. Again, two teaching interaction situations were used: in the first, a four-year-old taught a three-year-old of the same sex and social class; in the second instructional sequence, a mother taught her own fouryear-old. The mean frequencies of positive reinforcements used by the Israeli middle-class four-year-olds was 1.8 and that for the lower-class four-yearolds was .6. The mean frequencies of negative reinforcements for the middle-class children was 1.2, and the mean for the lower-class group was 1.0. The difference between the two groups in their use of positive reinforcement was highly significant; the middle-class Israeli child displayed about three times the frequency of positive reinforcements displayed by the lower-class child. On the other hand, the Israeli children made little use of negative reinforcement; the difference between the middle- and lower100

N O R M A D. FESHBACH class children was slight and insignificant. The differences between the two Israeli four-year-old groups are quite similar to the differences found between middle-class white and lower-class black American children in the first study of reinforcement styles (Feshbach & Devor, 1969). The Israeli middle-class mothers displayed a significantly greater frequency (mean = 6.7) of positive reinforcement than Israeli lower-class mothers (mean = 4.3) when instructing their four-year-olds. This finding is of special interest since the differences in maternal use of reinforcement among the various American socioethnic groups occurred primarily with negative rather than positive reinforcement. When the frequencies of negative reinforcement are compared, the results correspond in part to the American results. Israeli middle- and lower-class mothers use comparable amounts of negative reinforcement when instructing their four-year-old daughters (means = 3.5, 3.0, respectively), but they behave differently toward their four-year-old sons. Lower-class Israeli mothers used about twice as much negative reinforcement (mean = 4.6) as did the middleclass Israeli mothers (mean = 2.2) when teaching their sons. We cannot ascertain whether this difference is related to sex differences in performance since, unfortunately, performance measures are not available for the Israeli sample. However, in view of the absence of prior sex differences on these tasks, sex differences in the performance of the Israeli four-yearolds seem unlikely. In terms of the observed patterns of reinforcements, the environment of the Israeli lower-class child resembles that of the American lower-class black child, and the Israeli middle-class child and the American middleclass white child appear to share common experiences. Israeli lower-class children received fewer positive reinforcements from their peers, fewer positive reinforcements from their mothers, and, if they are male, more negative reinforcements from their mothers as compared with the middleclass Israeli children. The differences in reinforcement style between lower- and middle-class Israeli mothers and between lower- and middleclass Israeli children provide additional support for the hypothesis that reinforcement styles may be an important behavioral dimension linking socioethnic differences in socialization practices with socioethnic differences in cognitive performance. Correlations were also obtained between the frequency of the mother's and of the child's use of positive and negative reinforcement in the two Israeli groups. Again, most of the correlations were small and statistically 101

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY insignificant except for that found for Israeli middle-class mothers' and their four-year-olds' use of negative reinforcement (r = + .39; p < .05). In view of the restricted range of negative reinforcements used by the Israeli children, this finding was unexpected, although consistent with the similar relationship found for the American middle-class white children and mothers. ENGLISH TEACHING STYLES

England was selected as a third culture in which to study socioeconomic group differences in reinforcement styles. I selected England partly because I was already involved in a study of English preschool education (Feshbach, Goodlad, & Lombard, 1973). In addition to having ready access to English nursery schools, I, like others, had been impressed by the striking difference in linguistic style between English middle- and workingclass children, differences which had been extensively analyzed and documented by Bernstein (1961, 1962). Bernstein had proposed that social class differences in language exposure and usage are an important, if not the most important, determinant of social class differences in English children's cognitive performance and achievement. The question could be asked whether social class differences in reinforcement style could also be mediating the differential response of English middle- and lower-class children to the academic tasks they were required to master. And so the same procedures used in the United States and in Israel to assess reinforcement styles in four-year-olds and their mothers were also applied to an English sample (Feshbach, 1973b). The English sample comprised 50 middle- and working-class mothers and their four-year-old children and 50 three-year-old pupils. The children were drawn from seven preschools in the greater London area; five were private nursery schools attended primarily by middle-class youngsters, and two were State-supported preschools attended primarily by working-class children. Father's occupation in conjunction with residential area were used as criteria for assigning a child to the middle-class or working-class group. Practically all of the English middle- and workingclass samples were white and were native to England. Thus, in contrast to the Israeli sample, there were no ethnic differences between the English middle- and working-class groups; the distinctions were economic and social. Before considering the results, it should be noted that, in contrast to the American and Israeli children, a number of children in both of the 102

N O R M A D. FESHBACH English samples, apparently anxious about being left with the experimenter, had difficulty conforming to the experimental instructions. The mean frequencies of positive reinforcement used by the children were 1.8 for the middle-class group and .7 for the lower-class group. The mean frequencies of negative reinforcement were 2.4 for the middle-class group and 2.5 for the lower-class group. The middle-class four-year-olds again used significantly more positive reinforcements than did the lowerclass group. The means for the middle- and working-class English children for frequency of negative reinforcement were almost identical. Thus, the social class differences in reinforcement patterns observed in English four-year-olds are comparable to the socioethnic differences in children's reinforcement styles found in the Israeli and American samples. The results for maternal use of reinforcements differ in several interesting respects from those obtained for the American and Israeli mothers, but the overall implications of the data are similar for all the groups. The middle-class English mothers teaching their sons made significantly greater use of positive reinforcement (mean = 6.1) than did the working-class English mothers teaching their sons (mean = 2.7). There was no difference in the frequency of use of positive reinforcement by the two groups of English mothers teaching their daughters (mean = 3.2 for both groups). It appears that middle-class English mothers shower their sons with positive reinforcement even though their sons perform no better than their daughters. In contrast to the findings for the Israeli and American mothers, there was no difference in the frequency with which the middle- and lower-class English mothers used negative reinforcement (means —1.5, 2.0, respectively). The performance times of the English children were longer than those obtained by the American children. This difference occurred for the entire three-year-old pupil sample and also for the middle-class four-year-olds. Since number of negative reinforcements is related to puzzle completion time, the greater time taken by these English children undoubtedly had some effect on the negative reinforcement scores. Perhaps for this reason the correlation between mother's and child's use of negative reinforcement, significant for both the middle-class Israeli and American white samples, was not significant for the English sample. In addition, none of the other correlations between mother's and child's reinforcement style were significant in the English sample. 103

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY CROSS-CULTURAL COMPARISONS

In describing reinforcement styles of four-year-olds and their mothers in the United States, Israel, and England, the focus has been on the similarity between cultures of socioethnic differences within each culture. The question of cultural differences in reinforcement styles has not been addressed. This approach was taken for primarily methodological reasons. We have much more confidence in the within-culture comparisons than in the between-culture comparisons. Within each culture, the same experimenters carried out the procedure with all socioethnic groups of mothers and children. However, different experimenters were used in different cultures. Although every effort was made to ensure that the same standardized procedures were followed in each cultural setting, experimenter effects may be included in the results. The degree of support provided by the central educational authority within each culture is another factor that varied between cultures and which could affect the degree of cooperation afforded by the mothers and children in the preschools. The point is that caution must be exercised in drawing inferences from direct cross-cultural comparisons. In Table 1 the data are summarized for frequency of positive and of negative reinforcements by American, Israeli, and English four-year-olds when teaching their three-year-old pupils. The number of positive reinforcements used by middle-class white American, Israeli, and English children are quite similar. The means for use of positive reinforcement for the lower-class children are also similar, with the exception of the elevated score found for lower-class white children in the second American study. The most deviant children are the Table 1. Mean Frequencies of Reinforcement Administered by American, Israeli, and English Four-Year-Old Teachers Culture American (Study 1 ) . . . . American (Study 2) . . . . Israeli English American (Study 1 ) . . . American (Study 2) . . . Israeli English

Middle-Class Middle-Class Lower-Class Lower-Class White Black White Black Positive Reinforcement 2.3 .2 2.4 .08 1.8 1.8 Negative Reinforcement 1.6 1.3 2.3 .4 1.2 24

104

.8 1.9 .6 .7 1.7 2.6 1.0 2.5

.7 .08

2.1 1.6

N O R M A D. FESHBACH American middle-class black children, who consistently display less positive reinforcing behavior than do children in any of the other middle-class groups. The frequencies of negative reinforcements reflect surprisingly little variation among cultures in the extent to which four-year-olds use negative reinforcements in this teaching interaction situation. On the whole, the reinforcement vstyles characterizing the children within a socioethnic group are similar from culture to culture. In Table 2 the results are summarized for the reinforcement styles of the mothers teaching their four-year-old children. The differences between cultures in mothers' use of positive reinforcement are relatively small. Although both samples of English mothers appear more conservative in their use of positive reinforcers, there is considerable variability within the English sample and the differences between cultures are not significant. When the mean frequencies of negative reinforcement are compared, one sees that the relative differences among groups are much larger than the relative differences among groups in the use of positive reinforcement. The American lower-class black mothers used much more negative reinforcement than did any other cultural subgroup. The scores of the Israeli mothers, both lower class and middle class, are also high relative to the English and American white subgroups. The possible contribution of differences in the children's performance to these cultural differences in maternal reinforcement frequency cannot be assessed inasmuch as performance measures were not obtained for the Israeli children. It will be recalled that English three-year-olds took more time to complete the puzzle than any of the American three-year-olds. The English Table 2. Mean Frequencies of Reinforcements Administered by American, Israeli, and English Mothers Culture

Middle-Class Middle-Class Lower-Class Lower-Class White White Black Black

American Israeli English

Positive Reinforcement 6.4 4.6 6.7 4.4

4.7 4.3 3.0

4.7

American Israeli English

Negative Reinforcement 1.4 1.8 2.9 1.5

2.2 3.8 2.0

5.4

105

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY four-year-olds also took more time than American white four-year-olds. The data from this study do not help explain this difference. However, the finding is consistent with other observations made in connection with my study on early schooling in England. The philosophy and programs of English preschools were less oriented toward early cognitive stimulation and training than is the case for either the United States or Israel. The English curricula and materials reflected this orientation. Consequently, the English children's poorer performance on the puzzles could be due to a relative lack of familiarity and experience with this type of task. Despite the longer times taken to complete the task by both middle- and lower-class English children, social class differences in types of reinforcement are still obtained for the English sample. Middle-class English fouryear-olds and their mothers use more positive reinforcements than do lower-class English four-year-olds and their mothers. The finding for the mothers appears to differ from that obtained for the other cultures where the observed social class difference was in the frequency of negative reinforcement. However, the reinforcement pattern — that is, the relative use of positive and negative reinforcement — is similar. In order to demonstrate the similarity in reinforcement pattern more directly, the proportion of positive reinforcements in relation to the total number of reinforcements was determined for each subject. This statistic consisted of the number of positive reinforcements divided by the number of positive plus negative reinforcements. In addition to providing a direct index of reinforcement pattern, the proportional analysis also controls for differences in absolute number of reinforcements. A 3 X 2, culture by social class, analysis of variance was carried out for the entire sample of mothers and a comparable analysis was made for the four-year-old teacher-children. For each analysis, a significant main effect was obtained for social class, the middle-class samples displaying consistently higher proportions of positive reinforcement than the lower-class samples. Analyses of the mother data reflected significant social class differences for each of three countries. For the four-year-olds, the difference was significant for the Israeli and English teacher-children. When a further analysis for the American sample was carried out, a significant difference in the relative use of positive reinforcement was obtained between middleclass white children and lower-class black children. In summary, the most consistent and least ambiguous aspect of the cross-cultural data are the socioethnic differences found in children's and 106

NORMA D. FESHBACH mother's reinforcement styles within each culture. Middle-class children and middle-class mothers use more positive reinforcements than do lowerclass children and their mothers, regardless of whether the culture is American, English, or Israeli. The socioeconomic differences that we have found in reinforcement styles are consistent with the socioethnic differences that have been observed in academic achievement. It seems reasonable then, to conclude that reinforcement style may be one factor — not the only factor or the most significant factor, but an important factor — mediating socioethnic differences in cognitive performance and academic achievement. Reinforcement styles employed by parents and by children can be viewed as reflectors of more pervasive features of middle- and lower-class environments. The lower-class family, by reason of its impoverished economic status, is subject to more privation, frustration, illness, and in general, to more stressful events than is the middle-class family. The lowerclass parent, who is trying to maintain at least a marginal level of social and economic adjustment, is confronted with more daily pressures and demands (including more pregnancies and children) than is the middle-class parent. Under these circumstances, we might expect the lower-class parent to be less tolerant and more critical of their child's errors and other deviant behaviors. It is difficult for a parent to be patient and encouraging when the family lives in crowded quarters and is beset with many, often conflicting demands. In brief, the economic circumstances under which lowerand middle-class families live render it likely that the lower-class family will make greater use of negative reinforcement than the middle-class family. I am proposing that frequent resort to negative reinforcement by parents is a reflection of impatience, environmental pressure, and frustrations. These are conditions which are hardly conducive to the learning process. Rather, they are likely to discourage exploratory activities in the child and foster avoidance behaviors incompatible with effective learning. Perhaps the most unfortunate consequence of these socialization experiences is the internalization by the child of these very response modes, which we have suggested interfere with effective learning. By age four, the genetics of poverty has already taken place, the process of identification and modeling being its DNA. The child is now prepared to duplicate the patterns of reinforcement of the parent in transactions with peers today and perhaps with his or her own offspring tomorrow. 107

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY I wish to re-emphasize that negative reinforcement and the environmental presses with which it is associated are seen as only one of a number of factors contributing to socioethnic group differences in cognitive performance. Also, I do not maintain that economic poverty inevitably leads to parental impatience, punishment, criticism, and other avoidance-producing behaviors detrimental to the child's cognitive development. These potential psychological effects of poverty may be mitigated or overridden by other factors operative in the family and in the culture. However, all other things being equal, one can expect a higher frequency of negative reinforcement and cognitive deficits among poor families than among middle-class families. Poverty, then, fosters psychological dispositions which place the poor child at an even further disadvantage in a society in which most avenues of economic reward have academic and related cognitive competencies as entrance requirements. MATERNAL REINFORCEMENT STYLES AND LEARNING PROBLEMS

Although the major focus of this paper has been on cross-cultural similarities in patterns of reinforcements used by different socioethnic groups, the theoretical implications of positive and negative reinforcement usage are by no means restricted to cognitive differences between poor and more advantaged children. Differences in environmental stresses produced by variations in socioeconomic levels constitute but one possible source of differences in patterns of reinforcements employed by mothers and by their children. Differences in school economic resources, class size, school problems, and so forth should also contribute to differences in the patterns of reinforcements employed by teachers. Other possible antecedents of reinforcement style are variations in personality, in situational stresses, and in cultural mores. Whatever the source of reinforcement style, we should expect variations in reinforcement patterns to be associated with variations in cognitive performance. Thus, there should be a relationship between reinforcement styles and cognitive competencies within a particular socioethnic group. In the final study to be reviewed here the relationship between one facet of a child's cognitive competence, level of success in learning to read, and various maternal behaviors including reinforcement styles is investigated. In the previous studies, the principal independent variable had been socioeconomic and ethnic differences; the independent variable in the present study was the reading competence of the child. Successful and problem 108

NORMA D. FESHBACH readers were drawn from equally comfortable middle-class, advantaged family backgrounds. In a recent study undertaken in our laboratory (Bercovici & Feshbach, 1973), mothers of successful readers and of problem readers were observed while instructing children in several cognitive tasks. A reliable behavioral scoring scheme was developed which permitted the assessment of such maternal behaviors as controlling and directive statements, autonomy fostering statements, manual guidance, and verbal and nonverbal organization of task materials, in addition to the frequency of positive and negative reinforcing behaviors. An added methodological feature of this study was the broadening of the sample of children taught by individual mothers. Besides instructing their own child on two cognitive tasks, each mother also individually instructed two other children — one a problem reader, and the other a successful reader, on these same two tasks. This additional variation enabled us to assess whether a particular maternal reinforcement pattern was specific to the mother's own child or reflected a more general mode of interaction with children. By having the mother teach both problem and successful readers, further information was obtained about the relevance of the child's performance to the mother's behavior. Finally, we used the opportunity provided by this project, to investigate child-rearing attitudes and values that might relate to maternal reinforcement behavior and to children's reading competencies (Feshbach & Bercovici, in preparation) . For this purpose the Child Rearing Practices Report measure, developed by Jean Block (1969), was completed by the participating mothers. The total sample consisted of 40 first-grade children and their mothers, and 80 additional first-graders. Half the mothers had children categorized as problem readers, and the other half, successful readers. The designation of reading competence was made on the basis of both test data and teacher evaluation.* The successful and problem readers were matched for sex and IQ, and all children were of at least average intelligence with no manifest neurological impairment. Three-fourths of the sample in each group were male, and the total sample was drawn from schools in predominantly middle-class areas. * The author wishes to thank Seymour Feshbach, Howard Adelman, and Will Fuller of the UCLA Early Prediction and Prevention of Reading Disability Project who made available data on children's IQ, tests of reading ability, and teachers' evaluation of class performance and competence.

109

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY First, each mother was asked to instruct her own child on a task which required the matching of 12 simple line-drawn faces. Then the mother was asked to show her child how to fit pegs of different lengths into holes of different depths so that all the pegs would be level across the top. In the second experimental situation the mother taught the same two tasks to another child whose reading level was comparable with that of her own child. In the third experimental situation the mother instructed a third child, whose reading level was divergent from that of her own child. During the instruction of her own and the two other children a minute-by-minute time sampling of the mother's behavior was made by two observers. Finally, the mother was administered the Block Child Rearing Scale. There were no significant differences among the groups of children — problem readers or successful readers, instructed by their own or another mother — in task completion time. The only significant effect was the greater amount of time required to complete the more difficult second task as compared with the first task. There were only slight, nonsignificant differences between the mothers of problem readers and the mothers of successful readers in the mean frequencies of positive reinforcement. A very different picture emerged when the mean frequencies of negative reinforcing behaviors were examined for tasks 1 and 2 combined (see the tabulation on p. 111).There were highly reliable and consistent differences between mothers of the problem readers and mothers of the successful readers. Mothers of the problem readers used much more negative reinforcement than did the mothers of the successful readers, both when instructing their own children and when instructing other children. The difference between amount of negative reinforcement used by the two sets of mothers occurred for both tasks, but it was greater for the second, more difficult task than it was for the first task. These data support the suggested theoretical relationship between children's cognitive performance and the reinforcement style employed by their parents. The socioethnic differences obtained for the American, Israeli, and English samples demonstrated systematic differences in reinforcement styles between middle- and lower-socioethnic groups. There remains, however, a gap between these cross-cultural findings and the proposition that the social class differences in reinforcement styles are functionally related to social class differences in cognitive performance. The data of this last study, demonstrating greater use of negative reinforcement by mothers of problem readers as compared with mothers of successful readers, 110

NORMA D. FESHBACH help fill this gap. The correlation between mother's behavior and children's reading competence is, of course, subject to alternative interpretations. However, it does provide a link in the chain of relationships that we believe exist between social class and socialization practices, as reflected in reinforcement styles, and between reinforcement styles and cognitive performance.

Own child Other problem reader Other successful reader Mean

Mother of Problem Reader 1.7 1.0 5 1.1

Mother of Successful Reader .8 .4 .3 .4

The detailed analyses made of the mothers' instructional behaviors provide further insight into the behavior patterns denoted by a particular reinforcement style. For example, there was a greater frequency of controlling and directive statements made by the mothers of problem readers than by mothers of successful readers. The mothers of the problem readers were also much more likely to intervene verbally or manually when the child made an error or encountered some difficulty. Children taught by mothers of problem readers would sometimes complain about their instructor's intrusiveness and ask that they be allowed to solve the task themselves. The intrusiveness or impatience of these mothers was reflected particularly in the data for manual guidance. Manual guidance was scored whenever the mother manipulated the task materials after the child began the task or whenever she guided the child's hand. The difference was large between the frequency of manual guidance by the mothers of problem readers and by the mothers of successful readers. The mothers of problem readers had great difficulty in permitting the child to proceed at his or her own pace. In summary, mothers of problem readers used more negative reinforcement, were more directive and intrusive, and appeared to be less patient than mothers of successful readers when instructing their own and other children. A thesis of this paper is that the behavior patterns observed are not idiosyncratic to the teaching of cognitive tasks but are characteristic of the mother-child interaction in many child-training situations. Thus, it is suggested that teaching styles displayed by mothers in these experimental situations constitute a dimension of socialization which has relevance for other child-rearing situations. To investigate reinforcement styles in these 111

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY varied socialization situations, it would have been desirable to employ a methodology, comparable to that used in all of our studies, which allows for the direct observation of parental behaviors. We might have observed parents instructing their children on tasks ranging from eating routines to coping with frustration and conformity to adult controls. Our resources did not permit such an undertaking, and since our interest at this point was exploratory, we used instead a structured variation of an interview procedure, the Block Q sort measure. The mothers sorted 91 child-rearing items into seven piles ranging from least characteristic of their behavior to most characteristic. Of the 91 items in the Q sort, 16 were considered applicable to the negative and positive reinforcement style dimension. The responses to these items were combined such that a higher score was indicative of more negative reinforcing attitudes and behaviors. Examples of items selected for inclusion in this reinforcement dimension are "I believe physical punishment to be the best way of disciplining," "I believe in praising the child when he is good and think it gets better results than punishing him when he is bad," and "I make sure my child knows that I appreciate what he tries or accomplishes." The variance in response to the individual items was small, which probably reflected the influence of social desirability norms and the homogeneity of the parent population. Nevertheless, this response dimension did discriminate between the two groups of mothers; the mean score obtained by the mothers of problem readers was significantly higher than the mean score obtained by mothers of successful readers. Although the inventory difference was not large and certainly was much less dramatic than the behavioral differences manifested in the experimental tasks, the data support the proposition that the behaviors displayed by mothers in the teaching situations are indicative of more generalized reinforcement styles. The mothers of problem readers made greater use of negative reinforcement in the teaching situation and they also revealed more general punitive attitudes on the Q sort than did the mothers of successful readers.

Some Implications These findings, in conjunction with the cross-cultural data, increase our confidence in the utility of the construct of reinforcement style as a dimension of socialization with developmental implications for the cognitive growth of the child. One of the advantages of this construct is that as a 112

N O R M A D. FESHBACH structure or style variable, it cuts across many content areas. It can be applied with equal relevance to the interaction between parent and child, between teacher and child, and between child and peer. It provides a framework for the analysis of the developmental consequences of some of the behavioral practices of the socializing agencies of family, school, and peer culture. Finally, it suggests another promising approach to early intervention efforts designed to eliminate social class and ethnic differences in cognitive competency. It suggests that the way in which a child is taught may often be at least as important as what the child is taught. Many current intervention efforts have moved in the direction of parent-training programs as a way of implementing cognitive and other developmental objectives (e.g., Gordon, 1969, 1970; Gray & Klaus, 1970; Karnes, 1970; Schaefer, 1969). Partly on the basis of the theoretical rationale for early intervention (Bloom, 1964; Hunt, 1961) and partly as a reaction to discouraging Head Start results (Westinghouse Learning Corporation, 1969) a number of these programs are attempting to reach the child under two. In general, the focus of the parent education programs has been primarily cognitive. Parents are trained and encouraged to instruct their young children before the child begins formal schooling, with cognitively related materials that are provided them. Occasionally, some emphasis is also placed on parental attitudes (Gordon, 1970). We can expect that the development and implementation of parenttraining programs will continue to expand and will constitute a major mode of early intervention for the very young child. However, these programs may prove to be ineffective if, in instructing the child, the parent teachers are impatient, intrusive, and use a high degree of negative reinforcement. These factors may be especially important when one considers that the target population of these programs are children in the first two years of life. Before a new Jensen report (1969) or a new Westinghouse report (1969) or a new Hernstein report (1971) appears on the lack of success of these future infant/parent intervention programs, a re-examination of the critical dimensions of the curricula employed in these projects is in order. The thesis of this report suggests that modes of instruction as well as the content of instruction require attention and analysis and, further, that parental influences on the cognitive development of the child are not limited to cognitive training situations but encompass the full spectrum of parent-child interactions. The child's schooling may begin with day care, 113

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY preschool, or kindergarten, but the child's education begins in the family. And the studies that have been reviewed suggest that the patterns of reinforcement and related modes of socialization used by the family are an integral part of this educational process. References Baratz, S., & Baratz, J. Early childhood intervention: The social science base of institutional racism. Harvard Educational Review, 1970, 40, 29-50. Becker, W. C. Consequences of different kinds of parental discipline. In M. L. Hoffman & L. W. Hoffman (Eds.), Review of child development research. New York: Russell Sage, 1964. Pp. 169-208. Bee, H. L., Egeren, L. F., Streisaguth, A. P., Nyman, B. A., & Leckie, M. S. Social class differences in maternal teaching strategies and speech patterns. Developmental Psychology, 1969, 1, 726-734. Bercovici, A., & Feshbach, N. D. Teaching styles of mothers of successful and problem readers. Paper presented at the meeting of the American Educational Research Association, New Orleans, February 1973. Bernstein, B. Social class and linguistic development: A theory of social learning. In A. H. Halsey, J. Floud, & C. A. Anderson (Eds.), Education, economy, and society. New York: Free Press, 1961. . Social class, linguistic codes, and grammatical elements. Language and Speech, 1962,5,221-240. Block, J. Q-sort: Child rearing attitudes. MS., University of California, Berkeley, Department of Psychology, 1969. Bloom, B. S. Stability and change in human characteristics. New York: Wiley, 1964. Bronfenbrenner, U. Socialization and social class through time and space. In E. E. Maccoby, T. M. Newcomb, & E. L. Hartley (Eds.), Readings in social psychology (3rd ed.) New York: Holt, 1958. Cole, M., & Bruner, J. S. Preliminaries to a theory of cultural differences. In I. J. Gordon (Ed.), Early childhood education: The seventy-first yearbook of the. National Society for the Study of Education. Chicago: National Society, for the Study of Education, University of Chicago Press, 1972. Deutsch, M., Katz, I., & Jensen, A. Social class, race, and psychological development. New York: Holt, 1968. Feshbach, N. D. The effects of teachers' reinforcement style upon children's imitation and preferences. Proceedings of the 75th Annual Convention of the American Psychological Association, 1967, 2, 281-282. (a) . Variations in teachers' reinforcement styles and imitative behavior of children differing in personality characteristics and social background. University of California, Los Angeles, Center for the Study of Evaluation, Technical Report No. 2, 1967. (b) .Teaching styles in four year olds and their mothers. Paper presented at the Meeting of the Western Psychological Research Association, Los Angeles, 1970. Reprinted in J. F. Rosenblith & W. Allinsmith (Ed.), The causes of behavior: Readings in child development and educational psychology (3rd ed.) New York: Allyn& Bacon, 1972. . Teaching styles of Israeli four-year-olds and their mothers: A cross-cultural comparison. Paper presented at the meeting of the American Educational Research Association, New Orleans, February 1973. (a) 114

NORMA D. FESHBACH . A comparative study of teaching styles in American, Israeli, and English four-year-olds and their mothers. In preparation, (b) & Bercovici, A. Maternal Reinforcement Behavior and First Grade Children's Reading Competence. In preparation. Feshbach, N. D., & Devor, G. Teaching styles in four year olds. Child Development, 1969,40,183-190. Feshbach, N. D., & Feshbach, S. Imitation of teacher preferences in a field setting. Developmental Psychology, 1972, 7, 84. Feshbach, N. D., Goodlad, L, & Lombard, A. Early schooling in England and Israel. New York: McGraw-Hill, 1973. Gordon, I. J. Early child stimulation through parent education. Final Report, Children's Bureau, Social and Rehabilitation Service, HEW, Grant No. PHS-R-306. June 30, 1969. — . Parent participation in compensatory education. Urbana: University of Illinois Press, 1970. Gray, S., & Klaus, R. Early training project: The seventh year report. John F. Kennedy Center for Research on Education and Human Development, George Peabody College for Teachers, Nashville, 1970. Hernstein, R. J. I.Q. Atlantic Monthly, 1971, 228, 43-64. Hess, R. D., & Shipman, V. C. Early experience and the socialization of cognitive modes in children. Child Development, 1965, 34, 869-886. . Cognitive elements in maternal behavior. In J. P. Hill (Ed.), Minnesota symposia on child psychology. Vol. 1. Minneapolis: University of Minnesota Press, 1967. Pp. 57-81. Hunt, J. McV. Intelligence and experience. New York: Ronald Press, 1961. Jensen, A. How much can we boost IQ and scholastic achievement? Harvard Educational Review, 1969, 39, 1-123. Karnes, M., Reska, J., Hodgins, A., & Badger, E. Educational intervention at home by mothers of disadvantaged infants. Child Development, 1970, 41, 925-935. Kennedy, W. A., & Willicut, W. C. Praise and blame in incentive. Psychological Bulletin, 1964, 62, 323-332. Kohl, H. 36 Children. New York: New American Library, 1967. Kohn, M. L. Social class and parent-child relationship: An interpretation. American Journal of Sociology, 1963, 68, 471-480. Kozol, J. Death at an early age. New York: Houghton, 1965. Labov, W. The logical non-standard English. In F. Williams (Ed.), Language and Poverty. Chicago: Markham Press, 1970. Pp. 153-189. Lombard, A. D. Home instruction program for preschool children (Hippy). Interim Report 1969-1970, Center for Research in Education of the Disadvantaged, Hebrew University of Jerusalem, Israel, March 1971. Miller, D. R., & Swanson, J. E. Inner conflict and defense. New York: Holt, 1960. Portuges, S. H., & Feshbach, N. D. The influence of sex and socio-ethnic factors upon imitation of teachers by elementary school children. Child Development, 1972, 43,981-989. Rodnick, E. H., & Garmezy, N. An experimental approach to the study of motivation in schizophrenia. In M. R. Jones (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press, 1957. Pp. 109-184. Schaefer, E. S. Development of hierarchical, configurational models for parent behavior, and child behavior. In J. P. Hill (Ed.), Minnesota symposia on child psychology. Vol. 5. Minneapolis: University of Minnesota Press, 1971. Pp. 130-161. . Home tutoring program. Children, 1969, 16, 59-61. Sears, R. R., Maccoby, E. E., & Levin, H. Patterns of child rearing. Evanston, 111.: Row, 1957. 115

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Smilansky, S. The effect of certain learning conditions on the progress of disadvantaged children of kindergarten age. Journal of School Psychology, 1968, 4, 68-81. & Yam, Y. The relationship between family size, ethnic origin, father's education, and student's achievement. (Hebrew T. Megamot.) Behavioral Sciences Quarterly, 1969, 16,248-273. Stevenson, H. W. Developmental psychology. In P. R. Farnsworth, O. McNemar, & Q. McNemar (Eds.), Annual review of psychology. Palo Alto, Calif.: Annual Reviews, 1967. Pp. 87-128. Weiner, B. Attribution theory, achievement motivation, and the educational process. Review of Educational Research, 1972, 42, 203-215. Westinghouse Learning Corporation, Ohio University. The impact of Head Start: An evaluation of the effects of Head Start on children's cognitive and affective development. June 1969.

116

«> JOHN W. H A G E N AND G O R D O N H. H A L E 3>

The Development of Attention in Children

THE ability to attend selectively to critical stimulus features and to ignore others is an integral part of the learning process, and it is necessary to understand the development of this ability in order to establish an adequate model of children's learning and thinking. We have examined the development of selective attention through research on children's incidental learning — that is, the acquisition of information that is extraneous or irrelevant to task performance. The original research paradigm was derived from Broadbent's model (1958), which states that a filtering mechanism causes certain information in a subject's environment to be attended to while other information is ignored. The former is held in memory briefly before being passed through filters for further processing, while the latter does not pass through the filters and fades from memory. More recent analyses by Neisser (1967), Treisman (1969), and others have expanded upon Broadbent's relatively simple filtering concept, but the essential aspect of the model, the principle of attention to selected stimulus features at the expense of others, remains useful. We have employed a paradigm in which certain features of the stimulus are designated as relevant for task performance and others are defined as incidental. Performance on this central task is assessed as well as later recall of information about the incidental stimuli, and these two measures together provide a basis for inferring selective attention. High incidental learning is assumed to reflect a high degree of attention to incidental cues; on the other hand, low incidental learning, in combination with high central performance, indicates selectivity in attention — that is, attention directed primarily to task-relevant rather than irrelevant stimuli. We made 117

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY two developmental predictions based on the model of selective attention. First, improvement in memory with increasing age will occur at least in part because of increasing ability to attend to specific cues and to ignore others. Second, under information overload conditions, incidental information will be "given up" to maintain adequate performance on the central task, and this "trade-off" will become more evident as children grow older. INITIAL STUDIES

In a study in collaboration with Eleanor Maccoby (Maccoby & Hagen, 1965), arrays of picture cards were used in the central-incidental task. Each card depicted a common object such as a toy train or a scooter with a background of a distinctive color. There were fourteen arrays, varying in length from four to six cards. Each array was shown briefly, and was followed by presentation of a cue card in a solid color, identical with the background color of one of the cards in the array. The child's central task was to locate the position in the array of the card that matched the color of the cue card. After the fourteen picture arrays were presented and the number of correct matches was recorded, the incidental task was presented. In this task the child was asked to match the pictures which had appeared on the previous trials with the appropriate color of background. Each picture had always appeared on the same background color. The number of correct matches constituted the subject's incidental learning score. Information overload was produced by including a distractor task. At each age level half the children performed the task in the distraction condition, which consisted of a tape recording of piano notes. Whenever a note occurred which was obviously lower in pitch than the others, the child was required to tap the table. The subjects were 7,9, 11, and 13 years of age. The results are easily summarized. The central memory task scores increased regularly as a function of age, but the incidental scores did not; they actually declined at the oldest age level. Thus, the hypothesized developmental improvement in selective attention was found: with increasing age, the children devoted more attention to the task-relevant than to the incidental information. The second prediction, concerning the effects of information overload, did not fare as well. This manipulation — requiring the subject to listen for an auditory stimulus — affected mainly the cen118

JOHN W. HAGEN tral scores, which were reduced by about the same amount at all ages. Incidental learning was impaired by distraction at age 13 but not at the other age levels, so that only for the oldest children was there any evidence for "giving up" of incidental information in the face of overload conditions. In a second study (Hagen, 1967) two modifications were made to eliminate certain problems with the first study and to provide further evidence regarding the hypotheses. New stimulus materials were used, and these have served as the prototypic materials for much of the subsequent research. Each card pictured two objects, an animal and a household object (see Fig. 1). Pretesting had revealed that with the original stimuli incidental learning of the background colors did not occur if the central task was to recall the objects themselves. The new stimuli permitted counterbalancing of central and incidental picture sets. For half the subjects, the central task was to recall locations of animals; for the other half, the task

Figure 1. Stimulus materials for the central-incidental task. (Reprinted by permission from J. W. Hagen, "Strategies for Remembering," in S. Farnham-Diggory [ed.], Information Processing in Children. New York: Academic Press, 1972.) 119

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY was to recall locations of household objects. In both cases, incidental learning was measured by asking the child to indicate the household object that had been paired with each animal during the central task. The second modification was the inclusion of a series of trials in which only one picture appeared on a card. This condition was introduced to obtain developmental norms of task performance in the serial-position recall task, and to discover if the mere presence of the incidental pictures on the cards affected central task performance. As before, the subjects were 7-, 9-, 11-, and 13-year-old children. As in the previous study, central task performance improved with age but incidental performance did not (see Figs. 2 and 3). The effects of distraction were also similar to those observed in the previous study. Central performance was lower when distraction was present than when it was absent, and this effect was about equal across age levels. However, under distraction, incidental performance was impaired only at the oldest age level. The task with one picture per card produced higher central performance than did the standard condition at all ages, demonstrating that the presence of the incidental picture impaired central task performance. Thus, regardless of the degree to which the incidental features are processed, their mere presence makes the central recall task more difficult for children in this age range. Several conclusions can be drawn from the results of these studies. First, the improvement in central recall with age without improvement in incidental recall indicates a developmental increase in efficiency of selective attention, as hypothesized. As children approach adolescence, they tend to focus on aspects of stimuli that are critical for task performance at the expense of processing extraneous information. Second, the hypothesis about the effects of information overload may have to be re-examined, since the auditory monitoring task did not produce a greater impairment of incidental than of central performance. Thus the effects of the distractor cannot, strictly speaking, be interpreted in terms of information trade-off, a giving up of incidental information in favor of central information. Still, there is evidence congruent with the developmental prediction since only the oldest children gave up irrelevant as well as task-relevant information when the distractor was present. In the second study, correlations were obtained between central task and incidental task scores, which indicate differences between the younger 120

JOHN W. HAGEN and older children's task performance. At the younger age levels, the correlations were positive, but at the oldest age level the correlation was negative. Among the younger children, then, those who performed well on the central task also showed a high degree of incidental learning, but among the oldest children, those who did well on the central task did poorly on the incidental measure. We shall return to these correlations, but at this point we note that by age 12-13 years children's performance on the cen-

Figure 2. Central task performance at grades 1, 3, 5, and 7. (Reprinted by permission from J. W. Hagen, "The Effect of Distraction on Selective Attention," Child Development, 1967, 38, 686-694.) 121

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY

Figure 3. Incidental task performance at grades 1, 3, 5, and 7. (Reprinted by permission from J. W. Hagen, "The Effect of Distraction on Selective Attention," Child Development, 1967, 38, 686-694.)

tral memory task appears to be maintained in part by excluding incidental or task-irrelevant information. Related Evidence. The results of two other studies that appeared in the literature about the same time as these provide further evidence for a developmental change in children's selective attention. Crane and Ross (1967) presented children with a visual discrimination learning task that contained both relevant and irrelevant dimensions. After the initial discrimination had been acquired, additional practice was given during which both dimensions were usable. As measured by a subsequent transfer task, second-graders were found to be using both dimensions whereas sixth-graders attended primarily to the originally relevant dimension. Siegel and Stevenson (1966), also using a discrimination task, found incidental learning to increase between ages 7 and 12 years but to decline between ages1 12 and 14 years. It would seem that a developmental pattern can be discerned: incidental learning does not improve monotonically with increasing chronological age; rather, incidental learning either increases or remains stable up to about 12-13 years and then it declines. The initial hypothesis — that improvement with age in central task performance occurs in part because of improved skill in ignoring irrelevant information — is consistent with these findings. 122

JOHN W. HAGEN THE ROLE OF STIMULUS FACTORS

Integration of Pictorial Components. Having found that children's efficient use of selective attention increases with age, we began to look for the reasons behind younger children's inefficiency in deployment of attention. One hypothesis is that younger children have difficulty analyzing stimuli into components, and thus they maintain attention to all features as a global unit. In the studies discussed thus far the central and incidental features of the stimuli have been depicted together. Under such conditions young children may attend to both features together as a single unit while older children attend to the components separately. In the next studies to be considered, the relation between the central and incidental features was varied — toward lesser or toward greater integration of components. These manipulations were intended to affect the degree to which the stimuli were amenable to analysis into components. By observing variations in performance with these several types of material, it was possible to determine whether younger children's nonselective approach is induced by specific types of stimuli or whether it is a general characteristic of children's orientation to multif aceted stimuli. In a study by Druker and Hagen (1969) the animal-and-object cards were used, but the arrangement of the central and incidental pictures was changed from that of previous studies in two ways. First, the two pictures on each card were presented spatially separated from each other, and this arrangement was compared with the usual contiguous arrangement. Second, the pictures were presented in a nonalternating fashion, such that the central picture always appeared above the incidental picture, and these stimuli were compared with the standard materials in which the central picture appeared above the incidental picture in only half the stimulus pairs. Both of these changes were intended to facilitate discrimination of the two features for the young children and to allow them to focus more exclusively on the task-relevant information, thereby reducing their level of incidental learning more than older children's. The results, while indicating an overall effect of stimulus spacing on amount of incidental learning (but no effect of nonalternation), did not show differential effects for children of ages 9, 11, and 13. The basic developmental results therefore were not altered by these attempts to facilitate identification of the taskrelevant features. Sabo and Hagen (1973) also tried to assist younger children in identifying the relevant information by presenting the central and incidental 123

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY pictures in different colors. In comparison with the standard material, the presence of color did improve the children's simple recognition of the stimuli, but the facilitation was no greater for younger than older children. Also, the presence of color did not affect the amount of incidental learning. Thus far we have found that attempts to reduce the integration of components and make the stimuli more amenable to analysis have had little discernible effect on the basic developmental results. Hale and Piper (1973) used stimuli in which the integration of pictorial components was increased, and they also found little developmental effect. The animal and object pictures were similar to those of the studies just discussed, but in three conditions, these pictures were shown in various action relations. Performance in these conditions was to be compared with performance in the standard condition in which the animals and objects were pictured separately. It was reasoned that if the animal and object in each stimulus were presented together to form a unitary scene, then older as well as younger children would view the stimuli as integral wholes and would maintain attention to both features of the stimuli. Thus, the degree of incidental learning should increase with age along with the degree of central task learning. This expectation was not borne out, however. Children shown the action stimuli exhibited more incidental learning than those shown the standard materials, but this effect was more pronounced at age 8 than at ages 11 and 14 in one experiment, and was approximately equal at ages 8 and 14 in a replication experiment. Subsequent analyses have suggested (Hagen, 1972; Maccoby, 1969) that incidental learning is determined by a two-stage sequence of information processing, such as that proposed by Neisser (1967). In the present context the first stage may be regarded as the initial discrimination of relevant and irrelevant material. Certain information is then selected for further processing, and only that material which becomes the object of one's attention is stored in memory for later retrieval. According to this model, the inefficiency of information processing attributed to young children could reflect either a failure at the initial discrimination stage or a deficiency in maintaining attention to relevant information. Although there actually may be developmental improvement at both stages, we believe that the primary changes in attention observed in the research on incidental learning reflect age differences in performance at the second stage, after the subject has performed the initial discrimination of relevant and irrelevant information. Support for this conclusion is provided by the studies on 124

JOHN W. HAGEN stimulus factors just discussed. The developmental trend toward greater use of selective attention remained clearly evident despite all of the attempts to increase or decrease the degree to which the stimuli were amenable to analysis into components. Thus, it is unlikely that the inefficient performance of younger children merely reflects a deficiency in initial discrimination of components. Presence versus Absence of Incidental Features. A second piece of evidence for this conclusion involves the effects on performance of the presence versus the absence of incidental cues. If the younger child's inefficient performance were the result of difficulty in initial discrimination of relevant and incidental features, then removing the incidental cues should improve performance to a greater extent for younger than for older children. In the Hagen (1967) study, removing incidental cues did result in improved performance, but not to a greater degree at one age level than any other. Apparently, children require some time for separation of relevant and extraneous features, but the effort expended at this initial stage of information processing may not differ markedly across age levels. Why is incidental information not ignored completely? Are incidental features noticed because they have some functional relation to central features, or are they simply picked up because they are there? Evidence bearing on this question is provided by Hagen and Frisch (1968), who examined central task performance as a function of the way in which the central and incidental pictures were paired. In the standard task, each incidental picture was paired with the same central picture across trials. In a second condition, each incidental picture was paired with a different central picture on different trials. In a third condition, the incidental pictures presented on a given trial were all identical. Thus, only in the first condition was there a consistent relation between the central and incidental pictures. No differences in central task performance were observed among these three conditions for any age group. The findings from these two studies, then, suggest that the distracting effect of incidental features can be attributed to their mere presence, and the incidental information need not have any functional relation to the central stimuli. An Anomalous Result. The data discussed thus far are from tasks using pictorial materials, and the developmental results show remarkable consistency in spite of variation in the nature of the stimuli used. One type of material, however, has been found to produce different results. Hale and Piper (in press) used the central-incidental task with colored shapes as 125

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY stimuli. The shape of each stimulus constituted the central feature and color the incidental feature. With these materials, both central and incidental scores were found to increase markedly between ages 8 and 12. When the stimuli were line drawings of animals and objects, however, incidental learning did not increase with age but remained relatively constant as found in previous experiments. In another experiment, also with children of ages 8 and 12, the developmental increase in incidental learning was demonstrated again with colored shapes; however, no age differences were observed when the shape and color formed a figure-ground relation (Hale & Piper, in press, Exp. 2). In the latter case, the stimuli were shape outlines on colored backgrounds, with the color visible both within and surrounding the shape. The locus of the incidental information was thus roughly equated in these two tasks, so that the differences in results cannot be attributed to factors related to orientation of sense receptors. That is, in both cases, as the children viewed the shapes their gaze necessarily was directed to the color also. The differences in results, therefore, appear to be a function of the relation between the central and the incidental information. For the colored shapes, the incidental information was integrally contained within the central stimulus elements, while for the shapes on colored backgrounds, the incidental information was independent of the central feature. These unusual results may be interpreted with reference to the twostage model discussed previously. It is assumed that pictorial stimuli of the type used in earlier studies are readily analyzable into components. That is, even when the central and incidental elements are depicted together, each still may be recognized as an entity independent of the other. The same is true of materials whose components form a figure-ground relation. When stimuli are thus readily analyzable, the initial process of discriminating task-relevant and incidental components is facilitated. The effort required at this first stage is minimized, and the subject can proceed easily to the next stage and focus his attention on the relevant information. However, when the components are attributes that are not regarded as separate entities, such as the shape and color of an object, then considerable effort must be expended in the initial discrimination process. Under such conditions it actually may be more efficient to maintain attention to all features of the stimulus, whether relevant or not, than to try to discriminate the relevant and irrelevant features. Apparently the older 126

JOHN W. HAGEN subjects did the former, as indicated by their relatively high level of incidental performance as well as of central performance with the colored shape stimuli. In summary, it is believed that the developmental trend toward greater use of selective attention involves a stage of information processing beyond the initial discrimination of components, and that this developmental change is most evident when the effort required at the discrimination stage is minimized. If this condition is not met, as when the central and incidental features are more naturally viewed as integral parts of a unit, then even older children may find it too difficult, or perhaps too inefficient, to employ selective attention. DEGREE OF TRAINING

We have shown that, with certain types of material at least, early adolescents exercise selective attention to a greater degree than younger children. Do older children exercise it maximally from the outset, or do they attend more selectively as their initial experience with the task indicates such an approach to be most efficient? Baker (1970) presents evidence in support of the latter alternative. She assessed incidental learning with either an eight- or a sixteen-trial task and found that, for children of ages 8 and 10 years, the incidental learning scores were greater following sixteen than following eight trials, but she found no difference for the 12year-olds. The younger children thus maintained attention to the incidental features of the stimuli and continued to acquire information about them. The oldest children, on the other hand, acquired incidental information primarily in the first eight trials. For these early adolescent subjects, apparently, attention was least selective at the outset of the task, permitting some incidental learning to occur over the initial trials; then attention became more selective as the task progressed, allowing little learning of incidental stimulus features to take place during the latter portion of the task. It is clear, then, that these subjects do not enter a learning situation with a predisposition to attend selectively. Rather, their approach is efficient in a more general sense; they are flexible and can adapt their strategy after experience with the task dictates the most effective means of attention deployment. Degree of training can also be defined in terms of the level of learning a subject has reached — that is, the relation between his performance and a specified criterion of learning. Defined in this manner, degree of training has received considerable theoretical emphasis (see Lovejoy, 1965; 127

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Mackintosh, 1965; Trabasso & Bower, 1968), and various positions have been taken regarding changes in attention before mastery of a learning task, and regarding the effects of overtraining (e.g., Houston, 1967; James & Greeno, 1967). Hale and Taweel (in press) examined the effects of degree of training on children's performance in a component selection task — a measure related to the incidental learning problem. It consisted of two parts, an initial learning phase and a posttest. In the initial phase the children were required to learn the spatial positions of stimuli that differed on two redundant dimensions, color and shape.* In the posttest, the child was shown a number of cards, each containing only a shape or only a color, and was asked to identify the position in which each had appeared. All of the shapes and colors were presented in the test, and scores indicating the number correct for each of these two components were obtained. It was assumed that the amount of information retained about each stimulus component reflected the degree to which attention had been directed to that feature during learning. The paradigm was thus similar to the incidental learning task but with two critical differences. First, neither feature of the stimuli in the component selection task was defined as central or incidental, since the task was intended to measure a subject's natural disposition to attend selectively rather than his ability to attend to externally defined relevant information. Secondly, the stimuli remained in the same positions throughout the initial phase of the task, so that the subject could be trained to a specified criterion of performance before administration of the posttest. Hale and Taweel's subjects were 4, 8, and 12 years of age; and at each. age level subjects were assigned to one of six different groups. These groups were given six different amounts of training, ranging from undertraining to overtraining, on the initial phase of the task. Performance on the posttest was compared for the six groups, and two effects were of interest. First, the scores for both the shape and color components increased markedly across all degrees of undertraining, suggesting that attention was directed to both components of the stimuli as the task was learned. Second, there was little increase in these scores with overtraining, indicat* This research was conducted independently of the Hale and Piper (in press) study demonstrating a functional difference between colored shapes and pictorial stimuli; given the latter results, caution is warranted in generalizing from the present findings to the more analyzable materials typically used in studies of incidental learning.

128

JOHN W. HAGEN ing that the post-criterion exposure produced a negligible amount of additional stimulus learning. This last result contradicts those models which predict that overtraining will "broaden" attention and produce increased acquisition of stimulus information (e.g., James & Greeno, 1967). Rather, the results are consistent with a model such as that of Trabasso and Bower (1968), which assumes that attention is least selective during the premastery stages of learning and becomes most selective during a period of overtraining. It is particularly interesting from a developmental standpoint that no marked age differences were observed in the pattern of results. For learning tasks of this type, then, a model which assumes that attention becomes maximally selective following mastery of the task is appropriate for children throughout the range from preschool age to early adolescence. Although methodological differences preclude a direct comparison of the results of the two studies just discussed, some integrative remarks may be made. In Baker's study, only the oldest subjects adopted a more selective approach with increased training, in that only these subjects failed to show much incidental learning during the latter portion of the task. For all age levels in Hale and Taweel's study, however, attention was most selective during the final trials of the task. The most reasonable explanation for this difference lies in the tasks used. The latter study used a task in which a criterion of performance could be specified, and continued trials beyond that point constituted overtraining. The nature of the task changed when criterion was attained, in that the subject no longer needed to learn the correct responses (positions of the stimuli) but only to continue responding correctly. This change in the task may have been partly responsible for the children's assuming a more selective approach with extended training. The central-incidental task, on the other hand, is not actually a learning problem but a series of short-term memory measures. Since each trial is independent of the next (the stimulus arrangement is altered each time), theoretically the task could be continued indefinitely with no change in the nature of the task analogous to that associated with the attainment of criterion in a learning problem. In the absence of such changes, younger children apparently perseverate in a nonselective approach to the stimuli. Older children, however, modify their method of attention deployment on their own initiative as they determine that a selective approach is most efficient in an incidental learning task. 129

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY COMPONENT SELECTION VERSUS INCIDENTAL LEARNING

We have stressed that employment of selective attention is the most efficient approach to use in an incidental learning task. This, of course, derives from the fact that one component of the stimuli is defined as relevant in this situation, and attention to other features is nonfunctional; children become increasingly proficient in attending selectively under these conditions as they grow older. It is of interest to ask, then, whether a similar developmental trend will be observed if children are allowed to discriminate among stimuli in whatever way they choose, rather than being required to focus on a single feature. To phrase the question in another way, Is there an increase with age in children's natural inclination to exercise selective attention, or is this simply a strategy that older children employ in situations such as an incidental learning task where selective attention is functional? In addressing this issue, Hale and Morgan (in press) used a component selection task similar to that used in the study on degree of training. Children's performance on this task was compared with their performance on two variant conditions in which a single stimulus feature was designated as relevant. Colored shapes were used, and subjects in one variant condition were told at the outset to attend to the shapes of the stimuli in preparation for a subsequent test. In the second variant condition, the subjects were required to attend to shape in order to learn the initial phase of the task. In the standard task, of course, no reference was made to the dimensions of the stimuli during this learning phase. The posttest was identical for all groups and produced two scores indicating recall for the positions of the shapes and colors, respectively. In one portion of the study, involving 4- and 8-year-olds, the results, were found to differ across the three tasks. Recall for information about the shapes was uniformly high; recall for color information increased with age for the component selection task but not for either of the variants. Thus, the developmental trend in attention to the secondary color component depended on whether this component was defined as incidental, as in the variant conditions, or was a redundant feature whose status was undefined, as in the component selection task. The results indicate that it is appropriate to view these two situations as tapping different processes. That is, it is necessary to distinguish between the process of attention to experimenter-defined relevant information, on the one hand, and the natural inclination to attend selectively, on the other. When 130

JOHN W. HAGEN 8- and 12-year-olds were compared, no age difference was observed in the shape or color scores for either the standard or the variant conditions. Thus, this distinction appears most applicable to children in the years before middle childhood. To determine the reliability of these results, Hale and Taweel (unpublished study) focused on the 5- to 8-year age range. Using a variety of stimulus materials, they manipulated the relevance of stimulus components in a manner analogous to that of the previous study. The earlier results were essentially replicated. There was an increase with age in recall for secondary stimulus information when such information was redundant but not when it was designated as incidental. Although the research on component selection is still in its early stages, two conclusions seem warranted. The first, as already noted, is that it is useful to distinguish between attention to externally defined critical features and the natural disposition to attend selectively. The other conclusion is that this distinction is applicable to children between preschool age and middle childhood, well below the ages of children focused upon in the incidental learning research. Thus far, we have emphasized the attentional inabilities of children in middle childhood relative to adolescence, but there is considerable development of attentional capabilities before this age as well. The results just discussed suggest an increase before middle childhood in the capacity to accommodate to the attentional demands of the situation. In the component selection task, a redundant secondary feature can serve as a cue for discriminating among stimuli; attention to this cue can be advantageous. When this component is defined as incidental, however, to ignore it in favor of attention to other stimulus features is more adaptive. With increasing age children apparently become better able to differentiate between these situations and respond accordingly. Thus, older children are more likely to employ selective attention and to ignore secondary stimulus features when these features are defined as incidental than when they constitute useful redundant information. Clearly, there are developmental increases not only in children's ability to attend selectively but in their ability to determine when it is most appropriate to employ selective attention. THE DEVELOPMENT OF TASK STRATEGIES

We have considered how children react to variations of stimuli and procedures in the incidental learning task. It is apparent that responses vary with age and with specific stimulus properties. Also, the develop131

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY mental results which emerge from these studies occur across many variations of the stimulus properties. A possible explanation for certain cues being learned and remembered at particular age levels when others are not might be that older children use particular strategies for stimulus encoding and storage which account for their better central task performance at the expense of incidental learning. Verbal Rehearsal. The use of verbal rehearsal as a mnemonic strategy was studied by Hagen, Meacham, and Mesibov (1970). Previously, Hagen and Kingsley (1968) had found that requiring children to say aloud the names of the pictures affected recall at some age levels but not at others. Such naming facilitated recall for 6- to 8-year-old children, but not for 10-year-old children. In Hagen et al.'s (1970) study, the children were 9 through 14 years of age, a range in which changes in selective processing had been found to occur (e.g., Druker & Hagen, 1969; Hagen, 1967; Maccoby & Hagen, 1965). Hagen and Kingsley (1968) had concluded that by the age of 10, children were able to use verbal rehearsal to facilitate recall and that simply labeling the stimuli interfered with rehearsal. Thus, the children in Hagen et al.'s study were all old enough to employ verbal rehearsal; the purpose of the study was to discover if such an encoding strategy played a role in the observed age differences in selective attention. At each age level, overt labeling was required for half the subjects. The results were as follows: Labeling did not affect either central or incidental performance overall. However, the serial position curves for the central scores showed that, at all age levels, naming lowered primacy recall and increased recency recall, a pattern similar to that found for 10year-olds by Hagen and Kingsley. It appeared that the required overt naming of the pictures interfered with spontaneous rehearsal of the to-beremembered items; hence, recall for primacy items, those presumably most facilitated by rehearsal, was impaired. Thus, since children across the age range 9 to 14 years uniformly use rehearsal, the central-incidental interaction with age, once again replicated in this study, cannot be attributed simply to developmental changes in the use of rehearsal strategies. Correlational Evidence. If the developmental trend toward greater use of selective attention is not attributable to increasing use of verbal encoding strategies, then are there other strategies which older children are more likely to employ? We have noted that older children tend to adjust 132

JOHN W. HAGEN their responses to task demands more than younger children, reflecting an adaptability or flexibility in their approach to the stimuli. Further evidence that older children are employing a task-appropriate strategy derives from available correlations. There are two types of correlations to consider: that between central and incidental scores, and correlations of these task measures with other indices of cognitive aptitude. We have mentioned one study (Hagen, 1967) in which central and incidental scores correlated positively at younger ages and negatively at older ages. Similar relations were found in other studies as well (Druker & Hagen, 1969; Hagen et al., 1970). At the younger age levels, then, those children who perform better in one task also perform better in the other task. Beyond a certain age, however, those who perform well on the central task do poorly on incidental learning and vice versa. It would appear that, for older subjects, success in task performance is accomplished partly through inhibition of attention to the incidental cues —clearly an efficient strategy to employ in this task. It is interesting to note that the negative correlation between central and incidental learning for the oldest subjects was largest in the nondistraction conditions in the Hagen study and in the no-label condition in the Hagen, Meacham, and Mesibov study. Labeling may be a type of distractor in the latter case, insofar as it impaired primacy recall. Then, in general, the trade-off of central for incidental information is less evident in the presence of distraction than in its absence. In other words, older children typically ignore incidental features in order to facilitate performance, but this strategy is disrupted when external factors such as noise or imposed labeling are included. The central and incidental scores have also been correlated with standardized measures of intelligence (e.g., Druker & Hagen, 1969; Hagen, 1967; Hagen et al., 1970). With increasing age, the correlations between central performance and intelligence test scores have generally increased in magnitude. Incidental scores, however, have shown only very low correlations with intelligence test scores and there is no discernible pattern in the correlations. In Hagen et al.'s (1970) study of verbal labeling, a second experiment was conducted with college students for whom scores from the Scholastic Aptitude Test (SAT) were available. For the condition in which no verbal labeling was required, both the mathematical and verbal scales of the SAT correlated positively with central recall (Verbal, .38; Mathematical, .51) but not with incidental learning. For the labeling 133

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY condition, the correlations were near zero, so that the relation between measured mental ability and central task performance was not apparent when verbal labeling was required. Externally imposed conditions seem to diminish whatever advantage is gained from high mental ability. In general, where significant correlations were found between task performance and measures of mental ability, they were found for the central and not for the incidental task; such correlations were more likely to be found among older than among younger subjects. Thus, there is additional evidence that central task performance and incidental task performance, especially for older children, involve independent processes. ATTENTION IN RETARDED CHILDREN

The discussion thus far has focused on age differences in incidental learning and implications of these differences for the development of selective attention in normal children. We have also studied variation in selective attention as a function of mental age in retarded children, a population which has been described as deficient in attention (Zeaman & House, 1963). In one study (Hagen & Huntsman, 1971), the centralincidental task was administered in its standard form. The pattern of results observed for the retarded children was similar to that found for the normal children in that central task performance increased across MA levels while incidental learning remained relatively constant. Further, when scores of the retarded children were compared with scores of the normal children at equivalent MA levels, no differences in performance were seen. Only when comparisons were made of equivalent CA groups did the retarded subjects perform more poorly than the normal subjects. It was then decided to test another sample of retarded children, those living in institutions. For this sample, evidence of an attentional deficiency was found; the institutionalized retarded children obtained generally lower central task scores and higher incidental task scores than either the normal children at the same MA levels or the noninstitutionalized retarded children. We now wonder whether the institutional environment itself may be responsible for the poor attention performance of its residents. Zigler (1966) has argued persuasively that deficits in retarded youngsters in institutions are more often associated with motivational and emotional factors than with retardation per se, since very few institutional environments are conducive to maximal development. At present, we must recognize that the differences found might also be due to characteristics of 134

JOHN W. HAGEN retarded children who get placed in institutions as compared to those who do not. However, an implication of the results is that, when an attentional deficiency is found in the retarded child, that deficiency may be associated with his environmental conditions rather than with his intelligence level. Can institutionalized retarded youngsters be trained to improve in performance in the central-incidental task? Hagen and West (1970) modified the task to explore this possibility, using a primary and a secondary dimension in place of the central and incidental dimensions. Pennies could be earned for recalling the pictures of either dimension, but the payoff was five times as great for recalling pictures of the primary dimension. The stimuli were simple geometric shapes and colors. As expected, recall was better for primary than for secondary pictures. At the younger MA level (8 years), the difference between recall of primary and recall of secondary pictures increased over trials; for the older MA level (10.6 years), very little change occurred. Since the older children performed better initially, they may have already been operating near their maximum level, and thus differential reward could not help. Or perhaps there is less ability to profit from such reinforcement among older retarded children. It does appear, though, that there are conditions under which retarded children are able to improve in selective attention. Hagen and Hallahan.(1972) tested severely retarded institutionalized children both on the central-incidental task and on the discrimination learning task used by Zeaman and House (1963), a task in which a modified version of the Wisconsin General Test apparatus is used. A finding of major interest is that performance on the Zeaman-House task was positively related to performance on the central task. It appears that similar abilities are being tapped by these tasks, abilities relating to efficiency in deployment of attention. A useful approach in future studies of attention would be to incorporate into a single battery these and other tasks purporting to measure various aspects of selective attention in order to determine the interrelations among the measures. CROSS-CULTURAL EVIDENCE

Cultural differences in attention and memory processes have been explored by Daniel Wagner. In a study in Yucatan, Mexico, he used a modified version of the central-incidental task in which the pictures were taken from a game well known to the children and adults of the area. There were about 400 subjects from both urban and rural backgrounds 135

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ranging in age from 7 to 27 years. Although the data analyses are not complete, some of the more striking findings may be mentioned here. The performance of the urban sample was similar to that of the American samples already described. Central task performance increased with age; incidental task performance increased from 7-9 years of age until 13-16 years of age and then declined. Thus, the interaction between age and central versus incidental performance was replicated, although in this case the incidental learning scores reached a maximum level at a later age than in the American samples. For the rural sample, however, a different picture emerged. There was no overall increase with age in central task performance, even though the age span covered 20 years. Incidental scores increased with age up to 20-21 years of age and then declined at 27 years; the decline thus occurred almost 6 years later in this rural group than in the urban groups. Overall, performance was lower for the rural than for the urban groups. We do not know what aspects of cultural difference may be responsible for these findings. Possibly school experience is an important factor. School-age subjects in both the rural and urban Yucatan samples were attending classes, but the nature of the school experience was vastly different for children in the two settings. Furthermore, most of the adults in the rural sample had had little or no formal schooling. We have seen that two types of environmental variation, institutionalization versus noninstitutionalization for retarded children and urban versus rural cultural setting in a Mexican sample, are related to our indices of attention. Although this evidence raises many questions, it suggests that environmental factors play a critical role in determining how attention is deployed. SUMMARY AND CONCLUSIONS

We have presented evidence relating children's selective attention to a variety of factors, and although the general picture emerging from this research is complex, certain conclusions can be drawn with reasonable confidence. A reappearing theme is that of a developmental improvement in efficiency of attention deployment. To recapitulate the evidence bearing on this theme, children's incidental learning undergoes little change from middle childhood to early adolescence, whereas central performance increases markedly over this period. It is concluded that children's ability to exercise selective attention improves with age; increasingly, 136

JOHN W. HAGEN they concentrate on task-relevant stimuli and ignore extraneous information. This conclusion came from the results of early studies; more recent evidence has expanded our view of how children's use of attention becomes more efficient with increasing age. For example, older children do not simply enter a learning situation with a predisposition to employ selective attention, but rather, in performing an incidental learning task, they adopt a selective approach only as the task proceeds. By early adolescence, children are apparently quite flexible in their attention deployment; they modify their approach upon realizing what strategy will maximize their performance. The most efficient strategy in the central-incidental task, of course, is to focus upon relevant features at the expense of extraneous information. The correlational data suggest that such a strategy is more characteristic of older than of younger children. Although the relation between central and incidental learning was positive for young children, it was negative for subjects beyond early adolescence. Thus, only at the upper age levels was successful performance on the central task accompanied by an inhibition of attention to incidental features. Another way in which children become more flexible in attention deployment is indicated in the results of the studies on component selection. With development, children increasingly tend to distinguish between situations in which it is useful to attend selectively and conditions under which attention to several stimulus features can be more advantageous. The incidental learning task, of course, demands a selective approach, and thus the developmental increase in selectivity observed in this task indicates an increasing accommodation to task demands. When selective attention is not required, however, as in a component selection task where two or more redundant features may define the effective stimulus, a selective orientation is not used. In general, children not only improve in ability to exercise selective attention as they grow older, but they also become better able to determine when it is appropriate to attend selectively. We have considered Neisser's (1967) two-stage sequence of information processing and have suggested that, in the present context, the sequence consists of an initial identification of relevant cues followed by maintaining attention to those cues while ignoring irrelevant cues. It has been argued that the age differences in attention observed here reflect developmental changes in performance at the second stage, after the child discriminates the relevant from the incidental information. Evidence for 137

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY this conclusion is that the younger children maintained a relatively nonselective approach despite variations in the pictorial materials designed to facilitate the initial discrimination. Further, when the incidental cues were removed, thereby obviating the need for the initial discrimination, younger children's performance on the central task did not improve to a greater degree than that of the older children. Apparently the developmental differences observed involve an ability to maintain attention to relevant material and to ignore extraneous features after the two types of information have been identified. A two-stage model of this type can also account for the anomalous finding of a developmental increase in incidental learning with colored shapes. In this case, the relevant and incidental components were attributes that are not naturally viewed as separate entities. Thus, the initial discrimination of components was presumably difficult enough that even the oldest subjects were forced to maintain attention to both features of the stimuli. In general, while the use of selective attention may be the characteristic approach of an older child to an incidental learning task, this will be most clearly evident when the stimuli are readily analyzable and the effort required to separate the relevant and extraneous information is minimized. We have identified some of the ways in which children improve with age in efficiency of attention deployment. These changes reflect the patterns of growth in the environment to which we are accustomed, and it remains to be determined whether there are particular aspects of the environment, or specific characteristics of the children in it, that are responsible for the observed results. Work on cultural differences and mental retardation has provided some initial evidence, but continued effort is needed to identify the subject factors and the situational variables that determine how children process information from stimuli. References Baker, S. L. A developmental study of variables affecting the processing of taskrelevant and task-irrelevant information. Ph.D. thesis, University of Michigan, 1970. Broadbent, D. E. Perception and communication. London: Pergamon Press, 1958. Crane, N. L., & Ross, L. E. A developmental study of attention to cue redundancy introduced following discrimination learning. Journal of Experimental Child Psychology, 1967,5, 1-15. Druker, J. F., & Hagen, J. W. Developmental trends in the processing of task-relevant and task-irrelevant information. Child Development, 1969, 40, 371-382. 138

JOHN W. HAGEN Hagen, J. W. The effect of distraction on selective attention. Child Development, 1967,38,685-694. . Strategies for remembering. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 65-79. Hagen, J. W., & Frisch, S. R. The effect of incidental cues on selective attention. The University of Michigan Center for Human Growth and Development, Report No. 57, 1968. Hagen, J. W., & Hallahan, D. Selective attention in retardates: A validation study. Development Program, Department of Psychology, University of Michigan, Report No. 17, 1972. Hagen, J. W., & Huntsman, N. J. Selective attention in mental retardates. Developmental Psychology, 1971, 5, 151-160. Hagen, J. W., & Kingsley, P. R. Labeling effects in short-term memory. Child Development, 1968, 39, 113-121. Hagen, J. W., Meacham, J. A., and Mesibov, G. Verbal labeling, rehearsal, and short-term memory. Cognitive Psychology, 1970, 1, 47-58. Hagen, J. W., & West, R. F. The effects of a pay-off matrix on selective attention. Human Development, 1970, 13, 43-52. Hale, G. A., & Morgan, J. S. Developmental trends in children's component selection. Journal of Experimental Child Psychology, in press. Hale, G. A., & Piper, R. A. Developmental trends in children's incidental learning: Some critical stimulus differences. Developmental Psychology, in press. . The effect of pictorial integration on children's incidental learning. Research Bulletin 73-26. Princeton, N.J.: Educational Testing Service, 1973. Hale, G. A., & Taweel, S. S. Children's component selection with varying degrees of training. Journal of Experimental Child Psychology, in press. . Age differences in children's performance on measures of component selection and incidental learning. Unpublished MS. Houston, J. P. Stimulus selection as influenced by degrees of learning, attention, prior associations, and experience with the stimulus components. Journal of Experimental Psychology, 1967, 73, 509-516. James, C. T., & Greeno, J. G. Stimulus selection at different stages of pairedassociate learning. Journal of Experimental Psychology, 1967, 74, 75-83. Lovejoy, E. An attention theory of discrimination learning. Journal of Mathematical Psychology, 1965, 2, 342-362. Maccoby, E. E. The development of stimulus selection. In J. P. Hill (Ed.), Minnesota symposia on child psychology. Vol. 3. Minneapolis: University of Minnesota Press, 1969. Pp. 68-96. Maccoby, E. E., & Hagen, I. W. Effects of distraction upon central versus incidental recall: Developmental trends. Journal of Experimental Child Psychology, 1965, 2,280-289. Mackintosh, N. J. Selective attention in animal discrimination learning. Psychological Bulletin, 1965, 64, 124-150. Neisser, U. Cognitive psychology, New York: Appleton, 1967. Sabo, R. A., & Hagen, J. W. Color cues and rehearsal in short-term memory. Child Development, 1973, 44, 77-82. Siegel, A. W., & Stevenson, H. W. Incidental learning: A developmental study. Child Development, 1966, 37, 811-817. Trabasso, T., & Bower, G. H. Attention in learning: Theory and research. New York: Wiley, 1968. Treisman, A. M. Strategies and models of selective attention. Psychological Review, 1969,76,282-299.

139

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Zeaman, D., & House, B. J. The role of attention in retardate discrimination learning. In N. R. Ellis (Ed.), Handbook of mental deficiency. New York: McGrawHill, 1963. Pp. 159-223. Zigler, E. Mental retardation: Current issues and approaches. In L. Hoffman and M. Hoffman (Eds.), Review of child development research. Vol. 2. New York: Russell Sage, 1966. Pp. 107-168.

140

<» D A V I D K L A H R <$>

An Information-Processing Approach to the Study of Cognitive Development

THE purpose of this paper is to describe both the general features and some specific examples of an information-processing approach to the study of cognitive development. Our work represents an attempt to apply the theoretical and methodological approach of Newell and Simon to the complex problems raised by empirical work in the Piagetian tradition. The general paradigm is to formulate precise models of performance of the organism at two different levels of development, and then to formulate a model for the transition or developmental mechanisms. The guidelines for this sort of research were sketched by Simon (1962, pp. 154-155): "If we can construct an information processing system with rules of behavior that lead it to behave like the dynamic system we are trying to describe, then this system is a theory of the child at one stage of the development. Having described a particular stage by a program, we would then face the task of discovering what additional information processing mechanisms are needed to simulate developmental change — the transition from one stage to the next. That is, we would need to discover how the system could modify its own structure. Thus, the theory would have two parts — a program to describe performance at a particular stage and a learning program governing the transitions from stage to stage." NOTE: The use of first person plural in this paper is intended to have a literal interpretation. All of the work to be described is the result of my close collaboration over the past several years with J. G. Wallace, of the University of Warwick. We thank Sylvia Farnham-Diggory, Paul Goodman, and Guy Groen for their comments on an earlier draft. 141

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Looking back over a twenty-year period, Brown (1970) lists several of the forces that revitalized research in cognitive development in the late 1950's, among them: (a) computer simulation of cognitive processes ("Since machines — hardware — could accomplish information processing of great complexity, it was obviously perfectly scientific and objective to attribute such processing to the human brain. Why limit the mind to association by contiguity and reinforcement when the computer, admittedly a lesser mechanism, could do so much more? Computers freed psychologists to invent mental processes as complex as they liked.") and (b) America's discovery of Jean Piaget ("Computer simulation, psycholinguists, curriculum reform, and mathematical models altered our notions of the scientific enterprise in such a way to cause us to see Piaget as a very modern psychologist. To see that he was, in fact, the great psychologist of cognitive development." * However exciting the promised merger of these two areas may have appeared at the time, it is only quite recently that studies of the type suggested have been carried out (Baylor et al., 1972; Klahr & Wallace, 1970a&b; Young, 1971). Several factors probably contributed to the delay. The information-processing methodology in its most elaborate form (Newell & Simon, 1972) was devoted to the creation of models derived from intensive analysis of verbal protocols generated by adult subjects. It is now clear that such an empirical base is not the only one that can be used to support process models. Response latencies, nonverbal protocols, and error patterns are all amenable to process modeling. Early formulations of languages for describing models were difficult to deal with from the viewpoint of the developmental psychologist. Perhaps it seemed that the language of Piaget, whether at the level of either compre* In this same passage, Brown (1970, pp. ix-x) notes the Social Science Research Council's "wonderful ability to detect intellectual developments in an early critical period when they can benefit maximally from a little enrichment." He cites their establishment in 1959 of the Committee on Intellective Processes Research, which in turn produced the now classic series of five Monographs of the Society for Research in Child Development. Our own work has been influenced by the SSRC'S on both sides of the Atlantic. In 1965 SSRC sponsored a research conference on learning and the educational process at Stanford, where I. G. Wallace and I first met and began to consider the application of the information-processing paradigm to the area of cognitive development. The British SSRC funded Wallace's initial proposal to do research in this vein and has supported his research ever since. In the area of computer simulation, SSRC sponsored several intensive summer seminars at the RAND Corporation in the late fifties and early sixties whose purpose was to stimulate the use of simulation programs and list processing languages.

.

142

DAVID KLAHR hensive and general verbal statements or formal logico-mathematical notation, bore no direct relation to the forms of dynamic languages used by the information-processing theorists. Nevertheless, Piaget's steadfast insistence on the characterization of the child as an organism functioning under the control of a developing set of central processes led to an everwidening search for an appropriate descriptive language. At the same time, increasingly sophisticated experimental work was generating a complex web of interrelated empirical results that was inadequately accounted for by extant theoretical formulations. We are dealing with an organism whose behavior is "curiously elusive," as Halford (1970, p. 315) has nicely put it: "It is in its nature to compensate for any condition which is imposed." The modeling of an organism that yields so grudgingly to systematic analysis, whose failures are as important as its successes, and for which the most comprehensive theories are couched in subtle Gallic nuance, poses a formidable challenge. Before describing our approach to these problems, we shall briefly summarize our view of the methodology and the problems to which it is addressed. The Information-Processing Paradigm. Faced with a piece of behavior of a child performing a task, we pose the question, What routines for processing information and what kind of internally stored information would a child need in order to perform successfully? The answer takes the form of a set of instructions for an information-processing device, a computer program. The program thus constitutes a model of the child. It is constrained by four criteria (Simon, 1972) : (a) consistency with what we know of the physiology of the nervous system, (b) consistency with what we know of behavior in tasks other than the one under consideration, (c) sufficiency to produce the behavior under consideration, and (d) definiteness and concreteness. With respect to the last criterion, one can view three levels of explanation in information-processing terms. At level I are task-specific models. They are designed to explain cognitive behavior for specifically, narrowly defined tasks — for example, playing games such as chess and checkers and solving problems such as series completion. Such models are usually stated as running programs that are sufficient to meet minimal performance criteria. At level II, the information-processing models are aimed at a general reformulation of a wider range of problems in terms of the requirements these problems impose on the organism. Such models also 143

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY postulate some basic mechanisms that could explain a wide range of cognitive activities, for example an EPAM discrimination net (Feigenbaum, 1963) or a model of semantic memory (Quillian, 1967). The emphasis at this level is not so much upon a running program as it is upon the analysis of cognitive activity in terms of internal symbolic representations and processes that operate upon these representations. An excellent example of this level of theorizing can be found in a recent reformulation of some stochastic learning theories in information-processing terms (Gregg & Simon, 1967). At level III, the metaphorical level, the unique advantages of information-processing models begin to disappear. The notion of informationprocessing models has long had an appeal to psychologists, although it has only been with the recent invention of suitable languages to express ideas that any progress beyond vague verbal models has been possible. The best example of the metaphorical level is the book by Miller, Galanter, and Pribram (1960), in which a wide range of cognitive activities are described in terms of a hypothetical information-processing device called a TOTE unit. Examples of less extensive views of the process metaphor can be found in many discussions of cognitive development. For example, "If one does what is possible to do at the purely intensive level with logical operations of class and relation, one is behaving rather like a computer with a grouping program" (Flavell, 1963, p. 188). Or, "The difference in an error pattern from subject to subject indicates that the individual variability stems partly from differences in processing routines" (Smedslund, 1966, p. 166). It is useful to keep these three levels in mind when evaluating information-processing theories, particularly those that purport to be models of cognitive development. When stated only at the highest level of generality and metaphor, information-processing theories provide merely a different, rather than an improved, mode of theorizing. Ultimately, one hopes to push all such theorizing down to the concrete level of a running program. In our own work, this desideratum has been met only with respect to models of different performance levels. The model of the transition mechanism itself still remains at a metaphorical level. However, we believe that the precise formulation of the various stages of cognitive development makes it much more likely that a similarly precise model of the transition mechanism itself will soon be obtainable. 144

DAVID KLAHR Some Problems in Cognitive Developmental Research. During the last decade two fundamental problems have dominated the area of cognitive development. The first of these is the determination of the developmental relationships among particular cognitive skills. The intractable nature of this problem is exemplified in the steady stream of studies dealing with the Piagetian concept of a stage in intellectual development and with the stage of concrete operations in particular (Flavell & Wohlwill, 1969; Pinard & Laurendeau, 1969). That the concept of stage itself is ill defined has been clearly demonstrated by Flavell's (1971) presentation of alternative interpretations of the notion that "item A precedes item B." A second and equally complex problem is the determination of the transition rule in cognitive development. The characteristic difficulties of research upon this theme are illustrated in the conclusion of Laurendeau and Pinard (1966) that an overview of the studies to date provides grounds for pessimism about the outcome of the experimental approach to the transition problem. These two problems are interrelated. The cognitive skills whose mastery defines a stage are operationally defined in terms of a set of tasks that have become classics in Piagetian literature (e.g., class inclusion, transitivity, conservation). However, a careful analysis of the informationprocessing demands of the task variants indicates that we are dealing with a far from homogeneous entity, even when tasks are supposedly identical. The difficulty compounds itself when procedural variations are introduced; although we can record the gross effects of systematic procedural variations, we have no theoretical basis upon which to explain the effects of those variations or from which to derive new procedural variations. What we proporse here is an analysis that takes the point of view of an information-processing device and which posits underlying mechanisms that could lead to success or failure on a given task. With respect to transition, until we have more precise models of what it is that is going through transition, our theories must remain vague and nonoperational. The absence of a precise process-performance link, noted by several writers (e.g., Flavell, 1972), contributes to the difficulty of putting Piaget's account of transition to an experimental test, since the lack of specificity makes it all too easy to argue that a wide range of experimental results are compatible with his position. 145

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY

Process Models for Piagetian Tasks: An Example Our procedure dictates that we pose questions in terms of specific tasks or classes of tasks. In the developmental literature, certain tasks have assumed such prominence as to become the foci of major theoretical questions — for example, conservation, transitivity, class inclusion. In this section we shall provide an example of the application of informationprocessing analysis to a specific task. Based upon the questions and problems raised by this procedure, we shall, in the next section, move on to a modified methodology applied to a related class of tasks. Exhaustive Classification. Consider the processing requirements of Exhaustive Classification (EC) — one of a series employed by Kofsky (1966, p. 195) in her study of classification skills of children aged 4 to 9 years: A collection of blocks, including a red and a blue circle, one green and two blue squares, two red and two green triangles, was shown to S. He was to choose a block and put it in a box along with all the others that were "like it." After the first box was filled, the procedure was repeated with the remaining blocks until all the blocks had been chosen. In an exhaustive [classification], Ss consistently used an attribute to select the contents of each box and filled the box with all the blocks that possessed the criterial attribute. Note that in order to pass this task, the child must use the same attribute for each sort. If, for example, he selects objects to place in the first box on the basis of their color he must select other colors as the basis of subsequent sorts. He is not allowed to sort first all blue things, then all remaining triangular things, and so on (Kofsky, 1963, pp. 90-91). We can write a set of step-by-step instructions for performing EC : 1. Select a block from the collection and place it in the box (e.g., the red triangle) .1.1. Select a value of the block and remember it (e.g., red). 2. Find all the blocks remaining in the collection that have the value selected in step 1.1. Place them in the box. 3. Notice the attribute of the value selected in step 1.1. and remember it (e.g., color). 4. Select a block from the remaining collection and place it in an empty box. 4.1. If none are left exit: output is content of boxes. 4.2. If a block is found, go to step 5. 5. Find the value of the block just selected on the attribute determined in step 3 (e.g., find its color). 6. Find all the blocks remaining in the collection that have the value determined in step 5. 146

DAVID K L A H R 7. Go to step 4. This list of instructions can be viewed as a theory of task performance which implicitly assumes that, in addition to the instructions, the organism has an interpretive mechanism that can understand and obey them. In this form, the model of exhaustive sorting satisfies the criteria of sufficiency and, to some extent, of definiteness; however, it tells little about the demands it makes upon the processor or about its relation to other tasks. The remedy for this is, first, to move to a more detailed level of model building and, second, to postulate subprocesses that could function in tasks other than the one under consideration. In order to execute the model for exhaustive sorting, the child would need four subprocesses: (a) A subprocess to produce an element from a list of elements, either a block from a list representing a collection of blocks, or a value from the list of values representing a block. We call this process NOTICE, (b) A subprocess to attend to all items possessing a given value. This is FINDALL. (c) A subprocess to produce the name of an attribute, given the name of a value (given red, produce color; given square, produce shape). We call this ATTOF. (d) A subprocess to determine what the value of a given item is on a given attribute (what is the color of block xl). This is VALOF. Primary is (object, value) -» truth value NOTICE (object) -> value VALOF (object, attribute) -» value/nil ATTOF (value) -> attribute Secondary SAMEANY (object x, object y) -» value/nil DIFANY (object x, object y) -> attribute/nil FINDONE (value, set) -> object/nil Tertiary FINDALL (value, set) -» list of objects PAIRSAME (set) -> object x, object y, value/nil PAIRDIF (set) -> object x, object y, attribute/nil COUNT (set, value) -> number NOTE: Reprinted by permission of Academic Press, Inc., from D. Klahr and I. G. Wallace, "An Information Processing Analysis of Some Piagetian Experimental Tasks," Cognitive Psychology, 1970, 1, 367.

We postulated a basic set of these fundamental processes in a paper dealing with several of Kofsky's classification tasks (Klahr & Wallace, 1970). They are listed below in terms of their input-output relationships. These fundamental processes can be used as the building blocks in writing the task-specific routines as computer programs. The verbal model 147

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY for exhaustive sorting can be more precisely stated in the form shown in the second accompanying list. The routine, called EC, is written in an ALGOL-like language called POP-2 (Burstall, Collins, & Popplestone, 1968), and consists almost entirely of calls upon the fundamental process. All that remains indigenous to EC are the control statements that determine the sequence of events and the variables needed to store information about the results of the subprocesses. function EC set = > olst; vars obx valx attl; NOTICE (set) -» obx; PLACE (obx, olst); NOTICE (obx) -»valx; FINDALL (valx, set) -» olst; ATTOF (valx) -> attl; loop: NOTICE (set) -* obx; PLACE (obx, olst); //obx = nil then exit; VALOF (obx, attl) -> valx; FINDALL (valx, S6t) -> olst;

goto loop end; NOTE: Reprinted by permission of Academic Press, Inc., from D. Klahr and J. G. Wallace, "An Information Processing Analysis of Some Piagetian Experimental Tasks," Cognitive Psychology, 1970, 1, 374.

There are three classes of fundamental processes shown in the first list. The primary processes represent the lowest level of explanation: they are defined in terms of the functional relations between input and output, but no model for their actual operation is presented. These are processes which are not yet analyzed but which, presumably, are analyzable. The other fundamental processes are defined in terms of the primary processes. For example, in the accompanying list FINDALL consists of repeated calls upon FINDONE, which in turn calls NOTICE and is. The list also illustrates a crude attempt to include motivational mechanisms in our performance models. The DRIVE and LIMIT represent an extremely simple means of accounting for the fact that children's performance is partly a function of their motivation to do the task presented to them. We have described the specifics of their operation (Klahr & Wallace, 1970a); here we simply note the fact that information-processing theories can include such variables (see also Simon, 1967). function FINDONE val set -» obj; loop: NOTICE (set) -» obj; /'/ is (obj, val) then exit; if. DRIVE then goto loop else LIMIT (obj) exit end; function FINDALL val set —> olst; vars ob; loop: FINDONE (val, set) -» ob; PLACE (ob, olst); // ob -> nil then exit; if. DRIVE then goto loop else exit end; NOTE: Reprinted by permission of Academic Press, Inc., from D. Klahr and J. G. 148

DAVID KLAHR Wallace, "An Information Processing Analysis of Some Piagetian Experimental Tasks," Cognitive Psychology, 1970, 1, 370.

A Basic Model. Although it seems implausible to assume that children have stored in long-term memory task-specific processes such as EC, it does seem plausible to assume that by the time they reach the age range under consideration, they have some functional equivalent of at least the primary processes. The child has a twofold problem when novel tasks are presented. First he must assemble, from his repertoire of fundamental processes, a task-specific routine that is sufficient to pass the task at hand; then he must execute that routine. A schematic view of this basic model is presented in Figure 1. Failure can be attributed to any of three causes: the absence of the appropriate fundamental process, the inability of the assembler to put together the appropriate fundamental processes, or failure to correctly execute the task-specific routine. Determination of the adequacy of the fundamental process repertoire suggests a logical and empirical analysis similar to the tradition of task analysis (Gagne, 1970; Glaser & Resnick, 1972; Resnick, Wang, & Kaplan, 1970). However, our approach places greater emphasis upon the dynamic processing demands of task components than upon the purely

Figure 1. An assembler model of task performance. (Reprinted by permission from D. Klahr and I. G. Wallace, "An Information Processing Analysis of Some Piagetian Experimental Tasks," Cognitive Psychology, 1970, 1, 362.) 149

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY logical components. For example, the analysis yields predictions about the effects upon the response times of the number of objects in the stimulus display. Failures in cognitive synthesis (Farnham-Diggory, 1970) provide striking examples of situations where, although requisite subprocesses exist, the full task cannot be successfully performed. Whether to attribute this to assembler failures or to the limited capacity of the system to execute the (correctly assembled) routine is an unresolved issue (FarnhamDiggory, 1972; Simon, 1972). Furthermore, the level of generality of both the assembler and the fundamental processes with which the assembler works is itself a complex issue. Several possible dichotomies between what the subject brings to the experiment and what he constructs at "runtime" have been suggested by Newell (1972a). Failures to execute appropriately assembled task-specific routines can be attributed to at least two sources. The inclusion of some simple motivational mechanisms allows us to sweep many performance failures under this general heading of motivation. The unsatisfactory nature of this category can be improved only by elaborating the current mechanisms into an information-processing motivational model (Simon, 1967). The second source of failures of execution lies in the processing demands made by the routines. It is possible for a computer program that controls a real-time device to be logically correct, but to fail to run because of speed or storage limitations of the machine. Similarly, the programs in the human information processor, in conjunction with certain task environments, may exceed his processing capacity. Assumptions about longterm memory, short-term memory, and the control of attention must be made more explicit if we are to determine the effects of these kinds of reasons for task failure. The "closed subroutine" representation described thus far has some severe shortcomings as a means for attacking these problems. First, there is an overly restrictive distinction between control and computation, a distinction that seems to have no psychological basis. Second, there is no straightforward manner in which relatively local and minor changes in the system can produce apparently qualitative improvements in performance. A third problem with the subroutine approach is the lack of constraint it imposes on short-term memory. Finally, the model requires that task-specific routines be assembled from the fundamental processes by an assembler of considerable power and scope. In the next section we intro150

DAVID KLAHR duce a form of process model that provides a partial resolution for each of these difficulties.

Production Systems: A Language for Process Models In the past few years a new form for describing information-processing models of cognition has been proposed and successfully applied to a modest range of complex problem-solving activities (Newell, 1966; Newell & Simon, 1972). The models are posed in the form of a collection of independent rules, called productions, that together form a production system. The rules are stated in the form of a condition and an action: C—> A. The condition refers to the symbols in short-term memory (STM) that represent goals and knowledge elements existing in the system's knowledge state; the action consists of transformations on STM including the generation, interruption, and satisfaction of goals; modification of existing elements; and addition of new elements. A production system obeys simple operating rules. a. The productions are considered in sequence, starting with the first. b. Each condition is compared with the current state of knowledge in the system, as represented by the symbols in STM. If all of the elements in a condition can be matched with elements (in any order) in STM, then the condition is satisfied. c. If a condition is not satisfied, the next production rule in the ordered list of production rules is considered. d. If a condition is satisfied, the actions to the right of the arrow are taken. Then the production system is re-entered from the top (Step a). e. When a condition is satisfied, all those STM elements that were matched are moved to the front of STM. This provides a form of automatic rehearsal. f. Actions can change the state of goals, replace elements, apply operators, or add elements to STM. g. The STM is a stack in which a new element appears at the top, pushing all else in the stack down one position. Since STM is limited in size, elements may be lost. These rules operate within the framework of a theory of human problem solving described in Newell and Simon (1972) and briefly summarized by Newell (1972b, pp. 375-376). Structurally, the subject is an information processing system (IPS) consisting of a processor containing a short-term memory which has access to a long-term memory (LTM). The processor also has access to the 151

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY external environment, which may be viewed as an external memory (EM). All action of the system takes place via the execution of elementary processes, which take their operands in STM. The only information available upon which to base behavior is that in STM; other information (either in LTM or EM) must be brought into STM before it can effect behavior. At this level the system is serial in nature: only one elementary information process is executed at a time and has available to it the contents of STM as produced by prior elementary processes. Seriality here does not imply seriality either of perception or of accessing in LTM. This general model and the verbal rules listed above have been precisely specified in a special programing language created by Newell (1972b) called PSG (for Production System, version G). Production systems written in PSG can be run on the PSG interpreter in a time-sharing mode under user (i.e., model builder) control. The user can observe the sequential consequences of his model, and can provide inputs for as yet unspecified subprocesses. An Example of PSG. The example to be presented in this section is intended not as a psychological model, but rather as a demonstration of some of the features of PSG qua programing language. Figure 2 shows a production system that represents the hammering of a nail into a board (a) and a trace of the system as it is executed (b). First, we shall list some PSG conventions. Everything to the right of a semicolon is ignored: line 100 is merely a title. The last line (1000) contains information used by the loading routines, and is not part of the model, per se. Everything else is in the form of a label, followed by a colon, followed by a structure enclosed in parentheses. The first four of these structures (lines 200-500) are production rules. These rules consist of a condition written as a conjunction of elements, an arrow ( >) and a series of actions. The details of these productions differ, but they are all of this same general form. We shall return to the meaning of their particulars when we describe their execution. Line 600 contains a special structure, named LOOK, which requests input from the terminal. That is, LOOK represents a part of the model that has not yet been programed, and whose operation must therefore be simulated by the model builder. When LOOK is evoked as an action, the model builder must decide what information to provide before the system can continue to function. Line 700 contains the information that the system needs to get started: when PSG initializes the interpretation of a production system, it inserts TOP.152

DAVID KLAHR GOAL into the top of STM. The structure named BUILD (line 800) determines the sequence in which the productions will be considered: thus, P4 will be tried first, then PI, etc. Finally, the size of STM is determined by the number of NIL'S allocated to it. As for the dynamic behavior of the system (Fig. 2b), everything in lower-case letters is typed by the model builder, everything in upper case is typed out by the system. The command build start! initializes the production system interpreter: STM is cleared to all NIL'S, the count of ac00100 00200 00300 OOUOO 00500 00600 00700 00800 00900 01000

;DEMO OF PSG USING SIMPLE EXAMPLE PI:((GOAL * JOIN) AND (UP HAMMER) AND (NAIL UP) --> (UP ===> DOWN)) P2:((GOAL * JOIN) AND (DOWN HAMMER) —> (DOWN ===> UP) LOOK) P3:((GOAL * JOIN) —> (DOWN HAMMER)) PU:((GOAL * JOIN) AND (NAIL FLUSH) —> (* ==> +)) LOOK:(OPR CALL) TOP.GOAL:(GOAL * JOIN) BUILD:(PI* PI P2 P3) STM:(NIL NIL NIL) ("BUILD LOADED. DO BUILD START!") RETURN.TO.TTY!

•build start! 0. STM: ((GOAL * JOIN) NIL NIL) P3 TRUE 1. STM: ((DOWN HAMMER) (GOAL * JOIN) N I L ) P2 TRUE (GOAL * JOIN) () OUTPUT FOR LOOK = *(naM up) *|z ,z U. STM: ((NAIL UP) (GOAL * JOIN) (UP HAMMER)) PI TRUE 5. STM: ((GOAL * JOIN) (DOWN HAMMER) (NAIL UP)) P2 TRUE (GOAL * JOIN) () OUTPUT FOR LOOK = **(nail up) *lz ,z 8. STM: ((NAIL UP) (GOAL * JOIN) (UP HAMMER)) PI TRUE 9. STM: ((GOAL * JOIN) (DOWN HAMMER) (NAIL UP)) P2 TRUE (GOAL * JOIN) () OUTPUT FOR LOOK = **(nail flush) *|z /z 12. STM: ((NAIL FLUSH) (GOAL * JOIN) (UP HAMMER)) PI* TRUE 13. STM: ((GOAL + JOIN) (NAIL FLUSH) (UP HAMMER)) END: NO PO TRUE

**

Figure 2. A simple production system and its trace. 153

MINNESOTA S Y M P O S I A ON CHILD P S Y C H O L O G Y

tions is set to zero, the structure named TOP.GOAL is put into STM, and a scan of the list of productions in BUILD commences. The first production to be considered is PI. It scans STM for the three elements: (GOAL * JOIN), (UP HAMMER), and (NAIL UP). Only the first of these is currently in STM, so P4 is not satisfied. Next PI looks for the two elements in its condition, and fails. P2 also fails. Finally, P3 finds a match between its condition and the elements in STM. Since P3's condition is satisfied, the actions associated with it are taken. In this case the action consists of simply adding a new element (DOWN HAMMER) to the front of STM, and pushing everything else in STM down one notch. The current state of STM is printed out in Figure 2b after 1 (to indicate that one action has been taken since the system started). After a production has "fired" — that is, after its condition has been matched and its associated actions have been taken, the production system is re-entered at the top. Thus, after P3 has completed its actions, the sequence of productions, as listed in BUILD, is tested sequentially once again. P4 and PI fail again, but this time P2 is satisfied. The two actions in P2 are (DOWN = — = > UP) and LOOK. The first of these is an action that modifies an element already in STM: the second element in STM is scanned for DOWN, and it is changed to UP. The action LOOK is, as described above, a call upon the terminal for the output of an as yet unprogramed routine that is supposed to determine whether the nail is up or flush. At this point, the modeler decided that it was still up, entered (nail up), and returned control to the system. The state of STM after four actions is now printed out. The same basic control cycle is repeated, with the determination of which production will fire being entirely dependent upon the contents of STM. Further explanations of the details of PSG will be given in the following descriptions of some psychological models. A Model for Quantitative Comparison: PSQC1. In experimental studies of class inclusion, conservation, transitivity, and many other tasks, the child is asked to make judgments of the relative amounts of two or more quantities. In this section we will describe a model, written in PSG, for quantitative comparison. Once the model has been presented, it will be used as the point of departure for extensions in three directions: upward, into two contexts in which quantitative comparison takes place: class inclusion and conservation; downward, to more specific models of the subprocesses utilized by quantitative comparison; and forward, to models of development ally later versions of quantitative comparison. 154

DAVID KLAHR PSQC1 is a running program for quantitative comparison written in PSG. Figure 3 shows the complete set of productions for PSQC1. The system consists mainly of goal manipulations and operator calls that provide a precise model of the following view of quantitative comparison. The question More x or more y? is intended to mean Are there more things with value x than there are things with value y? In order to determine the answer to this question the system must compare the quantitative symbols corresponding to the amount of the things that are Jt, and the things that are y. In order to compare the symbols, it must first generate them, using some quantification operator. PSQC1 works in a relatively simple problem space. In addition to the various goals and their states, it needs only the information listed in lines 600-800 (Fig. 3) concerning the classes of elements for values, relations, and quantitative symbols. The productions on lines 2100-3500 constitute the core of the psychological model. The order in which they are scanned is determined by PSQC1 (4400), in this case much the order in which they are listed, with the exception of PA, PSTART, and PZ. PA ensures that only one goal at a time is active (indicated by *). Whenever it finds two active goals, it sets the older of the two to interrupted (% ). Since it has top priority (being the first production in PSQC1), it will repeatedly fire until only a single goal is active. PZ serves the opposite function: reactivating interrupted goals in the absence ( ABS) of an active one. The main productions will be described in the order in which they are actually satisfied. P3 converts the word MORE into an active goal. It represents the process whereby the semantics of the term more establish, in this context, an active goal that will trigger further actions. The action in P3 contains a special symbol, = = >, that scans the first element in STM for the symbol on the left of the arrow, and replaces it with the symbols on the right: thus (WORD MORE) gets transformed to (GOAL*MORE). P4 simply adds the active goal COMPARE to STM whenever it finds an active goal MORE. P9 says, in effect, if the goal COMPARE is active and a value exists, then create the goal of quantifying that value. The condition element (VAL) searches for an element that has two components: the tag VAL and a member of the class . As defined in line 600, any of the terms RED, BLUE, etc., could satisfy the search for a . Within the local domain of P9, is assigned whatever member of the class it finds, and retains this value for the action side of 155

00100 00200 00300 00*00 00500 00600 00700

;PSQC1 ; PSG VERSION OF PSMORE, KLAHR AND WALLACE (1972). ;CALLS TERMINAL FOR QUANTIFICATION OPERATORS AND ;FOR COMPARISON OF MAGNITUDES ; :(CLASS RED BLUE SQUARE ROUND WOOD) :(CLASS < > =)

00800 :(CLASS QS) 00900 X1:(VAR) 01000 X2:(VAR) 01100

;

01300 01*00

QUANT I FY:(OPR CALL); INPUT QUANTITATIVE VALUE: ( Q S N < V A L » . RELATE:(OPR CALL); RELATE TWO SYMBOLS. «VAL>
01200 ;

01500 01600 01700 01800 01900 02000 02100 02200 02300 02*00

; ;GOAL MANIPULATION PA:((QOAL *) AND (GOAL *) — > (* =«> *)) PZ:((GOAL *) ABS AND (GOAL %) ~> (% <==> *)) ; ;MAIN PRODUCTIONS PI:((GOAL + MORE) AND (XI X2) —> SAY) P3:((WORD MORE) —> (WORD ==> GOAL *)) P*:((GOAL * MORE) —> (GOAL * COMPARE)) P5:((GOAL + COMPARE) AND (GOAL % MORE) —> (? ===> +))

02800 02900 03000 03100 03200 03300 03UOO 03500 03600

; P9:((GOAL * COMPARE) AND (VAL (VAL ===> OLD VAL) (GOAL * QUANTIFY (* ==> + )) ; P11:((GOAL * QUANTIFY QUANTIFY) ; PDVO:((WORD (WORD «> VAL)) ;

03800 03900

; ;

02500 02600 02700

03700

P7:(X1 « «QS» AND X2 =- «QS» AND (GCftL * COMPARE) —> «QS> «*=> OLD «=> OLD (* ==> +))

SAY:(ACTION (OPR XI PRVL) (OPR PRVL) (OPR X2 PRVD)

01(000 OUOO 0*200

TOP.GOAL: (GOAL * ATTEND) PSTART:((GOAL * ATTEND) --> PROBLEM) PROBLEM:(OPR CALL); INPUT (MORE) (VAL1) (VAL2).

0**00

PSQC1:(PA PDVO PI P3 P* P5 P7 P6 P9 P10 Pll PZ PSTART)

0*300

;

0*500 ; 0*600 STM:(NIL NIL NIL NIL NIL NIL NIL NIL NIL) 0*700 ; 0*800

("MOR.AP1 LOADED. DO PSQC1 START!") RETURN.TO.TTY!

Figure 3. A production system for quantitative comparison: PSQC1.

156

DAVID KLAHR the production. P9 also ensures that it does not fire repeatedly, by changing the tag VAL to OLD VAL. Figure 4 shows the trace of PSQC1 operations for the task of subclass comparison, to determine whether in an array of red and blue squares, there are more reds or more blues. The trace consists of a printout of STM each time that a production is satisfied and its associated actions have been •psqcl start I

0. STM: ((GOAL • ATTEND) NIL NIL NIL NIL NIL NIL NIL NIL) PSTAKT TRUE (GOAL • ATTEND)

()

•|z

OUTPUT FOR PROBLEM - '(word noreKword blue)(word red)

h. STM: ((WORD RED) (WORD BLUE) (WORD MORE) (GOAL • ATTEND) NIL NIL NIL NIL NIL) PDVO TRUE 5. STM: ((VAL RED) (WORD BLUE) (WORD KDRE) (GOAL * ATTEND) NIL NIL NIL NIL NIL) POVO TRUE 6. STM: ((VAL BLUE) (VAL RED) (WORD MORE) (GOAL * ATTEND) NIL NIL NIL NIL NIL)

n TRUE

7. STM: ((GOAL • MORE) (VAL BLUE) (VAL RED) (GOAL * ATTEND) NIL NIL NIL NIL NIL) PA TRUE 8. STM: «fiOAL • MORE) (GOAL * ATTEND) (VAL BLUE) (VAL RED) NIL NIL NIL NIL NIL) Pk TRUE 9. STM: ((GOAL • COMPARE) (GOAL • MORE) (GOAL t ATTEND) (VAL BLUE) (VAL RED) NIL NIL NIL NIL) PA TRUE 10. STM: ((GOAL • COMPARE) (GOAL 4 MORE) (GOAL 1 ATTEND) (VAL BLUE) (VAL RED) NIL ML NIL NIL) P9 TRUE 12. STM: ((GOAL * QUANTIFY BLUE) (GOAL • COMPARE) (OLD VAL BLUE) (GOAL : MORE) (GOAL 4 ATTEND) (VAL RED) NIL NIL NIL) PA TRUE 13. STM: ((GOAL • QUANTIFY BLUE) (GOAL % COMPARE) (OLD VAL BLUE) (GOAL i MORE) (GOAL % ATTEND) (VAL RED) NIL NIL NIL) Pll TRUE (GOAL • QUANTIFY BLUE) «VAL> BLUE) OUTPUT FOR QUANTIFY • *«(qs 2 blue) •U ,z 15. STM: ((OS 2 BLUE) (GOAL • QUANTIFY BLUE) (GOAL ". COMPARE) (OLD VAL BLUE) (GOAL » MORE) (GOAL * ATTEND) (VAL RED) NIL NIL)

P10 TRUE 16. STM: ((GOAL * QUANTIFY BLUE) (OS 2 BLUE) (GOAL % COMPARE) (OLD VAL BLUE) (GOAL } MORE) (GOAL % ATTEND) (VAL RED) NIL NIL)

n TRUE

17.

STM: {(GOAL • COMPARE) (GOAL » QUANTIFY BLUE) (QS 2 BLUE) (OLD VAL BLUE) (GOAL » MORE) (GOAL \ ATTEND) (VAL RED) NIL

NIL)

P9 TRUE 19. STM: ((GOAL • QUANTIFY RED) (GOAL • COMPARE) (OLD VAL RED) (GOAL * QUANTIFY BLUE) (OS 2 BLUE) (OLD VAL BLUE) (GOAL * MORE) (GOAL \ ATTEND) N I L ) PA TRUE 20. STM: ((GOAL * QUANTIFY RED) (GOAL t COMPARE) (OLD VAL RED) (GOAL + QUANTIFY BLUE) (QS 2 BLUE) (OLD VAL BLUE) (GOAL I MORE) (GOAL i ATTEND) NIL) Pll TRUE (GOAL • QUANTIFY RED) «VAL> RED) OUTPJT FOR QUANTIFY • *«(qs 1 red) •U ,z 22. STM: ((OS 1 RED) (GOAL * QUANTIFY RED) (GOAL * COMPARE) (OLD VAL RED) (GOAL * QUANTIFY BLUE) (QS 2 BLUE) (OLD VAL BL E) (GOAL t MORE) (GOAL t ATTEND)) P10 TRUE 23. STM: ((GOAL * QUANTIFY RED) (QS 1 RED) (GOAL t COMPARE) (OLD VAL RED) (GOAL * QUANTIFY BLUE) (QS 2 BLUE) (OLD VAL BLU E) (GOAL * MORE) (GOAL t ATTEND))

n TRUE

2k. STM: ((GOAL • COMPARE) (GOAL + QUANTIFY RED) (QS 1 RED) (OLD VAL RED) (GOAL » QUANTIFY BLUE) (QS 2 BLUE) (OLD VAL BLU E) (GOAL t MORE) (GOAL I ATTEND)) P7 TRUE (OLD OS 1 RED) (NIL NIL X2 (OLD OS 2 BLUE) XI (OLD QS 1 RED)) OUTPUT FOR RELATE • "(blue > red) •U 28. STM: ((BLUE > RED) (OLD QS 1 RED) (OLD QS 2 BLUE) (GOAL « COMPARE) (GOAL » QUANTIFY RED) (OLD VAL RED) (GOAL * QUANT FY BLUE) (OLD VAL BLUE) (GOAL ? MORE)) PE TRUE 29. STM: ((GOAL « COMPARE) (BLUE > RED) (OLD QS 1 RED) (OLD QS 2 BLUE) (GOAL « QUANTIFY RED) (OLD VAL RED) (GOAL * QUANTI FY BLUE) (OLD VAL BLUE) (GOAL % MORE)) P5 TRUE 30. STM: ((GOAL » COMPARE) (GOAL « MORE) (BLUE > RED) (OLD QS 1 RED) (OLD QS 2 BLUE) (GOAL « QUANTIFY RED) (OLD VAL RED) (GOAL » QUANTIFY BLUE) (OLD VAL BLUE)) PI TRUE XI: BLUE : >

X2: RED

3U. STM: ((GOAL * MORE) (BLUE > RED) (GOAL » COMPARE) (OLD QS 1 RED) (OLD QS 2 BLUE) (GOAL » QUANTIFY RED) (OLD VAL RED) (GOAL « QUANTIFY BLUE) (OLD VAL BLUE))

Figure 4. A trace of PSQC1 on subclass comparison. 157

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY taken. The numbers that precede each STM display indicate the total number of actions that have been taken since the system started. As described above, in several cases the system calls upon the terminal for the output of operators that are not included, in programed form, in this production system. These outputs, indicated by lower-case letters, are typed in by the model builder. The command PSQC1 start! inserts the TOP.GOAL into STM and initiates PSQC1. The only production whose condition side is matched by STM elements is PSTART. PSTART has an action named PROBLEM which calls upon the terminal for an input (in Fig. 3, see lines 4000-4200). The input for quantitative comparison is then provided (in lower case). This input can be viewed as the result of preliminary linguistic processing of the experimental instructions. The state of STM after PSTART has completed its actions is shown in Figure 4 at (4.). Then the simple verbal encoder, PDVO, changes the WORD tag, having detected that both RED and BLUE are members of the class . Next, P3 converts the verbal term (MORE) to the active goal (GOAL*MORE). Notice that this goal appears (8.) at the front of STM. PSG includes a mechanism that automatically moves all the matched elements in a condition up to the front of STM. This represents a psychological assertion about automatic rehearsal in STM for those elements that evoke a production. As new elements are added, any elements in STM that are not matched by condition elements get pushed farther down until they eventually drop out of STM. For example, a scan of Figure 4 shows the element (GOAL % ATTEND) being pushed down and out of STM. The interrupted goal (GOAL % MORE) is just about to drop out of STM (29.) whenPS matches (and modifies) it (30.). Returning to (8.), P3 moves (WORD MORE) to the front of STM and modifies it to (GOAL*MORE). Since this is a modification of an existing element, there are as many elements in STM at (7.) as at (6.). P4 adds a new element to STM, (GOAL*COMPARE), and PSG operates such that new elements are added to the front of STM, pushing all else down one notch. A t ( 1 3 . ) P l l i s satisfied, resulting in the action QUANTIFY, an operator that calls on the terminal for the hypothesized output from a quantification operator. The call is made in a local context, (listed after PI 1 TRUE) just as it would be if QUANTIFY were itself a production system. The element (QS 2 BLUE) is the result of a human simulation of a quantifier that scans the hypothetical display for blue things and creates a quantitative 158

DAVID KLAHR symbol for two blue things. (Notice that at this level we have no model of quantification per se, although we have indicated the necessity for quantification operators of some sort. We shall return to this issue in detail at a later point.) The tag QS is the notation for a quantitative symbol, and the numeral 2 is an arbitrary (but obviously convenient) symbol for a specific value. The tag will be used directly by other productions, while the value will be used by another terminal call for a quantitative relational operator. The behavior of PSQC1 between (8.) and (16.) can be summarized: Given a goal of determining which of two values is MORE, a subgoal of COMPARE is generated; this, together with a VAL, generates a subsub goal of QUANTIFY for that value; this fires the quantification operator, which produces a quantitative symbol, which in turn satisfies the quantification goal. But two quantitative symbols are required for a comparison, and at this point only one exists, thus the sequence is repeated (17.)-(24.). Now P7 detects that the relationship between the two quantitative symbols can be determined by the operator RELATE. Since this is also a terminal call, the local context is provided and the system asks for the result of the relational operator. Once this relationship is provided, the comparison goal is satisfied (29.), and then the goal of determining more is satisfied (30.). Finally the system produces the result (following PI TRUE), and stops (34.). Class Inclusion. A slight alteration in the question which stimulated the trace just presented yields the familiar Class Inclusion (ci) question, More blues or more squares? (Let us assume for the remainder of this discussion that the array consists of two blue squares and one red square.) The prominence accorded to class inclusion by Piaget (1952) has stimulated a wide variety of ci studies. However, an analysis of the empirical results generated by these studies yields widely disparate pictures of the developmental course of ci (Klahr & Wallace, 1972a). The age of 50 per cent failure has been found to be as high as 11 years (Inhelder & Piaget, 1964) and as low as 6 years (Smedslund, 1964). The explanations offered for failure on ci can be divided into two categories: perceptual encoding failures and processing failures. An example of the former is Wohlwill's (1968) hypothesis that "the perception of two contrasting subclasses, unbalanced as to number, creates a strong tendency to translate a class inclusion question into a subclass comparison question" (p. 462). Inhelder and Piaget (1964), in contrast, attrib159

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY lite failure to the child's inability to think of the superset B while thinking of one of its parts A. It is only later that the child develops the requisite representations and processing capacity such that "the whole B continues to exist even while its components A and A' are separated in thought" (pp. 103-104). How can the model of quantitative comparison further our understanding of these issues? Initially, we (Klahr & Wallace, 1972a) placed the burden of explanation for failure on ci upon the unspecified (at that time) quantification operators. We argued that when the quantification operator was evoked for the first value to be quantified, it marked the internal representation of the collection; upon the second attempt the marked items were bypassed, thus giving the classic erroneous response. (For example, during the "blue" pass for "more blue or more square" — where squares are the superset — all blue squares are marked, so that on the "square" pass only the red squares are quantified.) In the context of PSQC1 such a mechanism could account for a system that can pass the subclass comparison task (more reds or blues), but fail ci. With development, the quantification operators were postulated either to remove old marks before processing or to make a copy of the representation for the collection. One of the benefits of specifying a process model in an unambiguous form is that it becomes highly falsifiable (Popper, 1959), both logically and empirically, and thus the model is amenable to extension, to modification, and if necessary, to abandonment. Hayes (1972) noted what he believed to be a logical flaw in our model, based upon the observation that if the superset happens to be quantified first, then on the second pass the system would find nothing at all to quantify. We subsequently demonstrated (Klahr & Wallace, 1972b) that this was not a logical error, in the sense of inconsistency, contradiction, or incompleteness. The system would in fact produce the correct answer, but for the "wrong" reason. However, the predicted frequency of such serendipity far exceeded the empirical facts, and on this basis PSQC1 can be considered only as a partial explanation of the ci phenomena.

Quantification Operators Three features of PSQC1 are candidates for revision, two that lie on the system-environment interface and one that lies totally within the system. The latter concerns changes in the existing production rules or their 160

DAVID K L A H R sequence. Such modifications are, as we shall demonstrate, unnecessary. The two areas where changes will be made are in the verbal encoding rules — about which we have said very little thus far — and in the quantification operators. In this section we shall summarize the empirical evidence concerning the nature of quantification operators, and then describe the form of the PSG models for such operators. These models will then be linked to the quantitative comparison model presented earlier. What was initially an undefined process "simulated" by the model builder, will become a programed model evoked in the context of a higher level task. A quantification operator is an organized collection of elementary processes that takes, as input, the stimulus to be quantified (e.g., a collection of blocks) as well as specified constraints (e.g., red only) and produces, as output, a quantitative symbol. Quantitative symbols are labeled internal representations (e.g., "two," "long," "tiny") that can be used in quantitative comparisons. Given two such symbols, the system can determine their relative magnitudes, whereas given two nonquantitative symbols, it can determine only whether or not they are identical. Evidence suggesting the existence of three quantification operators — subitizing, counting, and estimation — comes from analyses of reaction times (RT) and of errors in tasks requiring adult subjects to report the number of items in a display (Jensen, Reese, & Reese, 1950; Kaufman, Lord, Reese, & Volkman, 1949; Saltzman & Garner, 1948; Taves, 1941; Woodworth & Schlosberg, 1954, pp. 90-105). A reanalysis of the results of these studies and some new evidence is provided in Klahr (1973a). The earlier investigations were addressed to the question of whether the time required to quantify a collection is independent of the number of items viewed (n). The answer appears to be negative. A plot of RT versus n yields a monotone increasing curve in the range from 1 to 30 items. However, around n = 5 the slope abruptly changes from approximately 40 to approximately 300 miliseconds (ms) per item. There are corresponding discontinuities in error rates and in reported self-confidence. Finally, there is a subjectively different experience for n above and below 5. We have retained the term subitizing (Kaufman et al., 1949) for the operator used to quantify small collections. The parameters that define subitizing are a slope of 40 ms and a maximum range of 4 or 5 items. These parameters are similar to short-term memory (STM) scanning 161

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY rates (Steinberg, 1969) and to capacity limits (Miller, 1956). Thus, the parameters suggest that subitizing occurs in an interaction between STM and the environment. Our initial hypothesis was that the quantification operators utilize a complete internal representation of the stimulus (Klahr & Wallace, 1972a). One of the models we proposed as an explanation for successful functioning in the ci context actually made a copy of this representation before processing it. However, two lines of evidence appear to refute this hypothesis. First, the results of our studies with adults (Klahr, 1973a) indicate that the rapid rate of quantification (at least for small amounts) precludes LTM storage of the entire display. Second, the results of studies of the response latencies of children on ci (Wallace, 1972a) indicate that they respond much too rapidly to have created such an internal representation. In addition, Wallace (1972a) found that providing an external "copy" of the display did not substantially reduce the frequency of ci failures. All of this weakens the case for both the plausibility and the efficacy of an internally created copy of the entire display. A Process Model for Subitizing. The foregoing analysis has led us to the creation of some plausible models of quantification. A series of models for subitizing and counting have been written in PSG (Klahr, 1972). In this section we shall describe only one such variant in order to provide an example of the fine-grained analysis that is possible with this type of model. We shall then utilize the quantification model in the context of the quantitative comparsion model. The subitizing subsystem, PSUBIT, is shown in Figure 5, and a trace of the system running on a display of two elements is shown in Figure 6. The productions serve three general functions: the PDT'S (1000-1300) are part of a template matching system, PSTMP; the PDB'S (2000-2200) correspond roughly to the notion of moving sequentially along a list of stored quantitative symbols; and the PDS'S provide the control for initiating and terminating the subitizing process. PSUBIT is initialized when the goal SUBIT becomes active. The first production to be fired is PDS1: it ATTENDS to the environment and then sets up the first possible response in STM. PDS2, the last production to be satisfied, detects a satisfied subitizing goal and creates a quantitative symbol (QS) by concatenating the target value with the symbol name. Subitizing is viewed as a combination of template matching and of a 162

DAVID KLAHR 00100 00200 00300 00UOO 00500 OOGOO 00700 00800 00900 01000 01100 01200 01300

;SUB.E03 ; SANE AS SUB.E02, MIDIFIED TO WORK WITH HOR. SERIES. ; :(CLASS ISA ELM) XO:(VAR) X1:(VAR) :(CLASS RED BLUE SOJARE CIRCLE) ATTEND: (ACT I ON (Kit (OMO) (MTC(«VAL>») NOTICE) ; POn:((G«L • SUBIT) AW) (ELM) AW (TSA) AND «TD) ABS —> (• =*> «)) PDT2:((QOAL • SUBIT) AW (ELM) AND (TSA) AND (ELM) AND (TSA) AND «TE» ABS —> (• >=> *)) PBT5:((GOAL • SUBIT) AND (ELM) AMD (TSA) AND (ELM) AMI (TSA) AND (ELM) AND (TSA) AND «TE» ABS --> (• '=> »)) POTIi:((GOAL • SUBIT) AND (ELM) AND (TSA) AND (ELM) AND (TSA) AND(ELM) AND (TSA) AND (ELM) AND (TSA) AiJ) «TL» A

BS --> (* «> »»

01UOO PSTMP:(PDT1 POT2 PDT3 TOTU ) 01500 ; 01600 PDS1:((GOAL • SUBIT) —> ATTEND (SUB 1HTSA )) 01700 POS2:((QOAL * SUBIT) AM) (SUB XO) —> (SUB -—> OLD SUB) 01100 (NTC («VAL»)) (XI ==•> OS XO XD) 01900 ; 02000 POB2:((GOAL • SUBIT) AND (SUB 1) --> (1 •>==> 2)(TSA)> 02100 B)B3:((GQAL • SUBIT) AND (SUB 2) —> (2 ===> 3HTSA)) 02200 POBli:((GOAL • SUBIT) AM) (SUB 3) —> (3 ===> UXTSA)) 02300 ; 021.00 PSNB:(PDB2 PDB3 POB
Figure 5. A production system for subitizing: PSUBIT.

sequential transfer from LTM to STM both of template pieces and of number names. The highest priority productions in PSUBIT are those in PSTMP; they seek a one-to-one match between stimulus elements (ELM'S) and template pieces (TSA'S) . As soon as one of the conditions in the PDT'S is satisfied, the subitizing goal is changed from active to satisfied. Template pieces are symbols representing tolerance space atoms (ISA'S), the elementary unit of countableness (Klahr & Wallace, 1973). The PSNB productions implicitly tie cardinality and ordinality together, since the failure to match n items is followed by the addition of a single additional TSA and the change of the quantitative symbol from n to n + 1. Other variants of quantification (see Klahr, 1972) do not include this assumption about the relation between cardinality and ordinality. In these other variants, quantitative symbols correspond to templates of appropriate size, but the entire template must be loaded into STM for each match. ~stm:((goal * sublt)((red)) nil nil nil nil nil) *psubit ps! O. STM: ((GOAL * SUBIT) ((RED)) NIL NIL NIL NIL NIL) PDS1 TRUE ((RED)) «VAL> RED) *lz ,z 8. PDB2 10. POT2 11. PDS2 I1K END:

**

OUTPUT FOR NOTICE = *(elm red squareXelm red square)

STM: ((TSA) TRUE STM: ((TSA) TRUE STM: ((GOAL TRUE STM: ((QS 2 NO TO TRUE

(SUB 1) (ELM RED SQUARE)

(ELM RED SQUARE) ((RED)) (GOAL * SUBIT) NIL)

(GOAL * SUBIT) (SUB 2) (TSA) (ELM RED SQUARE) (FLM RED SQUARE) ((RED))) + SUBIT) (ELM RED SQUARE) (TSA) (ELM RED SQUARE) (TSA) (SUB 2) (RED)) (GOAL + SUBIT) (OLD SUB 2)

((RED)))

(ELM RED SQUARE) (TSA) (ELM RED SQUARE) (TSA))

Figure 6. A trace of PSUBIT on a two-element display. 163

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Such models are attempts to account for certain developmental data about the onset of cardinality prior to ordinality (Gelman, 1972). The general flow of PSUBIT is: a. NOTICE the stimulus. This is a call upon the terminal for input of the entire stimulus as a set of independent elements. A local context is provided (RED in Fig. 6). Note that we have pushed down the level of undefined processes toward the perceptual end of the system. b. Transfer the name of the first possible response (SUB 1) and a single piece of the template (ISA) from LTM to STM. c. If there are as many (TSA'S) as (ELM'S), then satisfy the SUBIT goal and produce a quantitative symbol containing the name of the current response. d. Otherwise, modify the response name to its next value and add another template piece to STM. Figure 6 shows a trace of this general scheme on a two-element display. In the range l^n^.4, PSUBIT takes 3 additional actions per additional element. The actions are: the input of an additional ELM, the addition of an extra TSA, and the modification of a response symbol. One may equate this to the 40 ms slope for subitizing reported earlier, by assuming that these actions take on the average 15 ms each, and by assuming also that no other activity in the interpretation of the production system contributed appreciably to the time. The most important specific assumption is that all conditions are tested in parallel: the sequential nature of both the testing of conditions and the STM scan for a particular condition is to be viewed as a serial simulation of a parallel process (on this issue see Newell, 1972b).Notice that this is quite a different view of the function of STM than is to be found in most other contemporary STM scanning models. A PSG model of the results of typical STM experiments (Newell, 1973) suggests an interpretation that bears little resemblance to the well-known flow-chart models (e.g., Norman, 1969;Sternberg, 1969). A REVISED MODEL FOR QUANTITATIVE COMPARISON

In this section we will describe PSQC2, a modified version of PSQC1 that includes PSUBIT as its explicit quantifier, and a rudimentary verbal encoder that accounts for ci failures. Both of these extensions require programing modifications which, in turn, require additional psychological assumptions. The model for PSQC2 (Fig. 7) can be compared with PSQC1 (Fig. 3). The main productions for goal manipulation (lines 5200-6400) are 164

DAVID KLAHR essentially unchanged. The major changes are the addition of productions for chunking and unchunking (2200-3900), a modified verbal encoder (4500-4700), and the change in QUANTIFY from a terminal call to a sequence of actions that includes subitizing (1900). STM Capacity. One major distinction between production system models and most other forms of information-processing models is the presence or absence of unconstrained local memories. In models using conventional programing languages, memory space is created simply by naming variables, and distinctions between short- and long-term memory are uncommon. In PSG, all productions take their inputs from and place their outputs in a common workspace: STM. The model builder must decide what constitutes a reasonable size for this workspace. A lower limit appears to be Miller's (1956) well-known estimate of 7 ± 2. However, our models must include room for control information as well as for the chunks that are typically measured in experimental situations. The original model, PSQC1, requires only 9 slots in STM. PSUBIT requires about the same STM capacity. However, when PSUBIT operates in the context of PSQC2, enough new elements are introduced to cause crucial information to be lost from STM. Thus we face a design choice: either increase STM capacity, almost doubling it, or find some way to save information temporarily. The former alternative would, in this instance, violate only an aesthetic sense, an STM capacity of 20 being inelegant, but in general expansion is completely unsatisfactory, for if PSQC2 were itself operating in the context of a higher level routine, we would need an even larger STM capacity. Furthermore even a small increase in capacity is not always desirable: some systems actually depend upon the loss of information from STM. If such systems are imbedded in larger systems with associated STM increases, they cease to function properly. We have chosen the second alternative for dealing with a limited STM capacity. PSQC2 contains a set of productions for creating a single chunk of up to 5 elements from a class of items that are worth saving. Just before subitizing, the action (ps.l PSCHNK) in the action chain QUANTIFY (1900) ensures that one pass through PSCHNK will occur. If there are any elements of the class <SAV> (1100) — that is, if there are any active or interrupted goals, quantitative symbols, or value names — then the first five of them will be packed together into a single element by PSCHNK (3000-3900). It seems reasonable that before initiating an encoding process that will flood STM with temporary information, the system col165

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY 00100 00200 00300 OOltOO 00500 00600 00700 00800 00900 01000 01100 01200 01300 01UOO 01500 01600

;PSQC2 ; XO:(VAR) Xlr(VAR) X2:(VAR) X3:(VAR) XU:(VAR) X5:(VAR) ; ; <SAV>:(CLASS * % VAL :(CLASS RED BLUE SQUARE ROUND WOOD) : (CLASS «WVAL1> NOT <WVAL2» <WVAL» <WVAL1>:(CLASS <WVAL» <WVAL2>:(CLASS <WVAL» :(CLASS < > =)

01700 01800 01900 02000

:(CLASS QS) RELATE:(OPR CALL); RELATE TWO SYMBOLS. «VAL>
02200 02300 02WO 02500 02600 02700 02800

;UNCHUNKING PS. UNPACKS AND MARKS OLD. PDUC1:((CHNK XI) —> (CHNK ==> OLD CHNK) XI) PDUC2:((CHNK XI X2) —> (CHNK ==> OLD CHNK) X2 XI) PDUC3:(( CHNK XI X2 X3) —> (CHNK ==> OLD CHNK) X3 X2 XI) PDUCii:((CHNK XI X2 X3 X«0 —> (CHNK ==> OLD CHNK) Xl» X3 X2 XI) PDUC5:((CHNK XI X2 X3 XU X5) --> (CHNK ==> OLD CHNK) X5 Xi» X3 X2 XI) ;

03100 03200 03300 03UOO 03500 03600 03700 03800

PDCH2:(X1 == «SAV» AND X2 == «SAV» —> (CHNK XI X2)) PDCH3:(X1 = «SAV» AND X2 == «SAV» AND X3 == «SAV» — > (CHNK XI X2 X3)) PDCHI»:(X1 — «SAV» AND X2 «• «SAV» AND X3 == «SAV» AND XU == «SAV» —> (CHNK XI X2 X3 X«0) PDCH5:(X1 — «SAV» AND X2 =« «SAV» AND X3 =» «SAV» AND X«t « «SAV» AND X5 == «SAV» — > (CHNK XI X2 X3 X<4 X5)) ;

03900

PSCHNK:(PDCH5 PDCHU PDCH3 PDCH2 PDCH1)

02100

02900 03000

PSUC:(PDUC5 PDUCU PDUC3 PDUC2 PDUC1) PDCH1:(X1 == «SAV» --> (CHNK XI))

Figure 7. A production system for quantitative comparsion including a call on PSUBIT and an object-oriented verbal encoder: PSQC2.

lects a small number of items with high saliency and puts them together for reference even while performing the encoding subtask. The chunk itself is rehearsed by the quantifier. In PSUBIT (Fig. 5, line 800), just before the environmental input, CHNK is brought to the front of STM by NTC (a PSG command). Those rules that deal with the environment must keep from being swamped by incoming information, and one way to avoid this is to rehearse important items before admitting such inputs. In summary, the effect of the decision not to increase STM capacity is, first, to cause one of the chunking productions (Fig. 7, 3000-3700) to fire before quantification takes place; second, to rehearse the chunk before noticing the display (Fig. 5, line 800), and third, to unpack the 166

DAVID K L A H R OdOOO 01(100 OU200 01(300 OliUOO 01(500 OU600 OU700 01*800 OU900 05000 05100 05200 05300 05400 05500 05600 05700 05800 05900 06000 06100 06200 06300 061(00 06500 06600 06700 06800 06900 07000 07100 07200 07300 071(00 07500 07600 07700 07800

;GOAL MANIPULATION PA:((* GOAL) AND (* GOAL) --> (* ===> %)) PZ:((* GOAL) ABS AND (% GOAL) --> (% -=> *)) •VERBAL ENCODING PDV1:((MORE) AND «WVAL1» AND «WVAL2» --> (XO =«> OLD XO) (XI «—> OLD X1XVAL «WVAL1> NOT <WVAL2») (VAL «WVAL2> NOT <WVAL1»)) /

PSVERB:(PDV1) ;MAIN PRODUCTIONS Pl:((+ GOAL MORE) AND (XO XI ) —> ( + ==> OLD) SAY)

P3:((MORE) --> (MORE -«> * GOAL MORE)) PU:((* GOAL MORE) ~> (* GOAL COMPARE)) P5:((+ GOAL COMPARE) AND (% GOAL MORE) --> (% -«> + ) ) P7:(X1 — «QS» AND X2 « «QS» AND (* GOAL COMPARE) —> «QS> ==> OLD «=> OLD (* «=> +)) P9:((* GOAL COMPARE) AND (VAL XO) --> (VAL «-> OLD VAL) (* GOAL QUANTIFY XO)) P10:((* GOAL QUANTIFY XO) AND «OjS> XO) —> (* »-> -O) Pll:((* GOAL QUANTIFY XO) —> (XO) QUANTIFY) SAY: (ACTION (OPR XO PRVL) (OPR PRVL) (OPR XI PRVD) TOP.GOAL:(ACTION PROBLEM) PROBLEM:(OPR CALL); INPUT (MORE) (VAL1) (VAL2). PSM2:(P1 P3 Pi( P5 P7 P6 P9 P10 PSEXEC:(PA PSVERB PSQC2) PSQC2:(PSUC PSM2 PZ)

Pll)

STM:(NIL NIL NIL NIL NIL NIL NIL NIL NIL NIL) ("MOR.B02 LOADED.

DO 'PSEXEC START! 1 ") RETURN.TO.TTY!

chunk upon return to PSQC2 by firing one of the unchunking productions (Fig. 7, 2200-2700). The effects of these productions will be indicated in the discussion of the trace below (see Fig. 8). Verbal Encoding. In both PSQC1 and PSQC2 we assume that an unspecified linguistic processor operates upon the input question and leaves three items in STM, the word MORE and the names of two values. We have postulated a rudimentary verbal encoder. In PSQC1 it was a simple rule that identified the value names and gave them appropriate tags for subsequent processing (PDVO in Fig. 3). The rule also translated the word MORE into a goal (P3). In PSQC2 we have a verbal encoder, PDV1, that operates as follows. Given the input from a ci question (MORE, BLUE, SQUARE) the system generates symbols for objects, not values: (BLUE NOT SQUARE) and (SQUARE NOT BLUE). These symbols are maintained in STM, ultimately providing a context 167

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY pw«ec start I

NIL

«VAL> SQUARE) OUTPUT FOR PROBLEM - •(m>re)(blue)(square)

,1 0. STM: ((SQUARE) (SLUE) (MORE) NIL NIL NIL NIL NIL NIL N I L ) PDV1 TRUE k. STM: ((VAL (BLUE NOT SQUARE)) (VAL (SQUARE NOT BLUE)) (MORE) (OLD SQUARE) (OLO BLUF) ML ML NIL NIL ML) P3 TRUE 5. STM: ((• GOAL MORE) (VAL (BLUE NOT SQJARE)) (VAL (SQUARE NOT BLUE)) (OLD SQUARE) (OLD BLUE) NIL NIL ML NIL N I L ) P* TRIE 6. STM: ({• OOAL COMPARE) (• GOAL MORE) (VAL (BLUE NOT SQUARE)) (VAL (SQUARE NOT BLUE)) (OLD SQUARE) (OLD BLUE) NIL ML N IL NIL) PA TRUE 7. STM: ((• GOAL COMPARE) C GOAL MORE) (VAL (BLUE NOT SQUARE)) (VAL (SQUARE NOT BLUE)) (OLD SQUARE) (OLD BLUE) NIL NIL N IL NIL) P9 TRUE 9. STM: ((• GOAL QUANTIFY (BLUE NOT SQUARE)) (• GOAL COMPARE) (OLO VAL (BLUE NOT SQUARE)) (% GOAL MORE) (VAL (SQUARE NOT BLUE)) (OLD SQUARE) (OLD BLUE) NIL NIL NIL) PA TRUE 10. STM: ((• GOAL QUANTIFY (BLUE NOT SQUARE)) (i GOAL COMPARE) (OLD VAL (BLUE NOT SQUARE)) It GOAL MORE) (VAL (SQUARE NOT BLUE}) (OLD SQUARE) (OLD BLUE) NIL NIL NIL) Pll TRUE POO* TRUE 16. STM: ((GOAL • SUBIT) (CHMK (• GOAL QUANTIFY (BLUE NOT SQUARE)) (? GOAL COMPARE) (? GOAL MORE) (VAL (SQUARE NOT BLUE)) ) (• GOAL QUANTIFY (BLUE NOT SQUARE)) (t GOAL COMPARE) <>. GOAL MORE) (VAL (SQUARE NOT BLUE)) ((BLUE NOT SQUARE)) (OLD VAL ( BLUE NOT SQUARE)) (OLD SQUARE) (OLD BLUE)) PDS1 TRUE ((BLUE NOT SQUARE)) «VAL> BLUE) OUTPUT FOR NOTICE - ••(elm blue squareMelm blue square) •U 2*. STM: <(TSA) (SUB 1) (ELM BLUE SQUARE) (ELM BLUE SQUARE) ((BLUE NOT SQUARE)) (CHNK (• OOAL QUANTIFY (BLUE NOT SQUARE)) (t GOAL COMPARE) (! GOAL MORE) (VAL (SQUARE NOT BLUE))) (GOAL • SUBIT) (• GOAL QUANTIFY (BLUE NOT SQUARE)) (> GOAL COMPARE ) (t GOAL MORE)) PDB2 TRUE 26. STM: <(TSA) (GOAL • SUBIT) (SUB 2) (TSA) (ELM BLUE SQUARE) (ELM BLUE SQUARE) ((BLUE NOT SQUARE)) (CHNK (* GOAL QUANTI FY (BLUE NOT SQJARE)) (i GOAL COMPARE) (i GOAL MORE) (VAL (SQUARE NOT BLUE))) (• GOAL QUANTIFY (BLUE NOT SQUARE)) (>. GOAL C OMPARE)) PDT2 TRUE 27. STM: ((GOAL « SUBIT) (ELM BLUE SQUARE) (TSA) (ELM BLUE SQUARE) (TSA) (SUB 2) ((BLUE NOT SQUARE)) (CHNK (• GOAL QUANTI FY (BLUE NOT SQJARE)) (>. GOAL COMPARE) (t GOAL MORE) (VAL (SQUARE NOT BLUO» (• GOAL QUANTIFY (BLUE NOT SQUARE)) (t GOAL C OMPARE)) POS2 TRUE 30. SIM: ((OS 2 (BLUE NOT SQUARE)) (GOAL « SUBIT) (OLD SUB 2) (ELM BLUE SQUARE) (TSA) (ELM BLUE SQUARE) (TSA) (CHNK (• GO AL QUANTIFY (BLUE NOT SQUARE)) (i GOAL COMPARE) (t GOAL MORE) (VAL (SQUARE NOT BLUE))) (• GOAL QUANTIFY (BLUE NOT SQUARE)) (t OOAL COMPARE)) END: NO PD TRUE 30. STM: ((QS 2 (BLUE NOT SQUARE)) (GOAL * SUBIT) (OLD SUB 2) (ELM BLUE SQUARE) (TSA) (ELM BLUE SQUARE) (TSA) (CHNK (• GO AL QUANTIFY (BLUE NOT SOUARE)) (t GOAL COMPARE) (t GOAL MORE) (VAL (SQUARE NOT BLUE))) (• GOAL QUANTIFY (BLUE NOT SQUARE)) (t GOAL COMPARE)) POX* TRUE 35. STM: ((• GOAL QUANTIFY (BLUE NOT SQUARE)) (>, GOAL COMPARE)
n TRUE

37. STM: ((• GOAL COMPARE) (« GOAL QUANTIFY (BLUF. NOT SQUARE)) (QS 2 (BLUE NOT SQUARE)) (% GOAL MORE) (VAL (SQUARE NOT BL UE)) (OLD CHNK (•> GOAL QUANTIFY (BLUE NOT SQUARE)) (• GOAL COMPARE) (( GOAL MORE) (VAL (SQUARE NOT BLUE))) (GOAL » SUBIT) ( OLD SUB 2) (ELM-BLUE SQUARE) (TSA))

n TRUE

39. STM: ((• GOAL QUANTIFY (SQUARE NOT BLUE)) (• GOAL COMPARE) (OLD VAL (SQUARE NOT BLUE)) (•> GOAL QUANTIFY (BLUE NOT SOU ARE)) (QS 2 (BLUE NOT SQJARE)) (? GOAL MORE) (OLD CHNK (•> GOAL QUANTIFY (BLUE NOT SQUARE)) (• GOAL COMPARE) <* GOAL MORE) ( OLD VAL (SQUARE NOT BLUE))) (GOAL • SUBIT) (OLD SUB 2) (ELM BLUE SQUARE)) PA TRUE M. STM: ((• GOAL QUANTIFY (SQUARE MOT BLUE)) (t GOAL COMPARE) (OLD VAL (SQUARE NOT BLUE)) (» GOAL QUANTIFY (BLUE NOT SOU ARE)) (QS 2 (BLUE NOT SQUARE)) (} GOAL MORE) (OLO CHNK (» GOAL QUANTIFY (BLUE NOT SQUAP.F)) U GOAL CO I PARE) U GOAL MORE) ( OLO VAL (SOJARE NOT BLUE))) (GOAL » SIJ8IT) (OLD SUB 2) (ELM BLUE SQUARE)) Pll TRUE PDCHd TRUE W. STM: ((GOAL • SUBIT) (CHNK (• GOAL QUANTIFY (SOUARE NOT BLUE)) U GOAL COMPARE) (QS 2 (BLUE NOT SQUARE)) (i GOAL MORE » (• GOAL QUANTIFY (SQUARE NOT BLUE)) (» GOAL COMPARE) (OS 2 (BLUE NOT SQUARE)) (* GOAL MORE) ((SQUARE NOT BLUE)) (OLD VAL (SQUARE NOT BLUE)) (* GOAL QUANTIFY (BLUE NOT SQUARE)) (OLD CHNK (• GOAL QUANTIFY (BLUE NOT SQUARE)) (J GOAL COMPARE) (» G OAL MORE) (OLD VAL (SQUARE NOT BLUE)))) POS1 TRUE ((SQUARE NOT BLUE)) «VAL> SQUARE) OUTPUT FOR NOTICE - ••(elm red square)

Figure 8. A trace of PSQC2 on a class-inclusion question.

for NOTICE in PSUBIT. For subclass comparison this will produce the correct result, but for ci it yields the classic failure. We shall demonstrate this in the trace of PSQC2 on ci below, but at this point we shall offer a psychological rationale for the verbal encoding rule. The basic assumption is that the system does not attend to the external 168

•ii ,t 55.

DAVID KLAHR

STM: ((ISA) (SUB 1) (ELM RED SQUARE) ((SQUARE NOT BLUE)) (CHNK (• GOAL QUANTIFY (SQUARE NOT BLUE)) (? GOAL COMPARE) OS 2 (BLUE NOT SQUARE)) (», GOAL MORE)) (GOAL • SUB IT) (• GOAL QUANTIFY (SQUARE NOT BLUE)) (t GOAL COMPARE) (QS 2 (BLUE NOT SQUARE)) (\ GOAL MORE))

pon TRIE

5k. STM: ((GOAL » SUBIT) (ELM RED SQUARE) (TSA) (SUB 1) ((SQUARE NOT BLUE)) (CHNK (• GOAL QUANTIFY (SQUARE NOT BLUE)) (t OAL COMPARE) (QS 2 (BLUE NOT SQUARE)) U GOAL MORE)) (• GOAL QUANTIFY (SQUARE NOT BLUE)) square)

•U 6*. STM: ((BLUE > SQUARE) (OLD QS 1 (SQUARE NOT BLUE)) (OLD QS 2 (BLUE NOT SQUARE)) (• GOAL COMPARE) <• GOAL QUANTIFY (S UARE NOT BLUE)) (* GOAL MORE) (OLD CHNK (« GOAL QUANTIFY (SQUARE NOT BLUE)) .(• GOAL COMPARE) (OLD QS 2 (BLUE NOT SQUARE)) t GOAL MORE)) (GOAL » SUBIT) (OLD SUB 1) (ELM RED SQUARE)) PS TRUE 69. STM: ((• GOAL COMPARE) (BLUE > SQUARE) (OLD QS 1 (SQUARE NOT BLUE)) (OLD QS 2 (BLUE NOT SQUARE)) (« GOAL QUANTIFY (SQ UARE NOT BLUE)) (t GOAL MORE) (OLD O*K (* GOAL QUANTIFY (SQUARE NOT BLUE)) (« GOAL COMPARE) (OLD OS 2 (BLUE NOT SQUARE)) ( t GOAL MORE)) (GOAL « SUBIT) (OLD SUB 1) (ELM RED SQUARE)) P5 TRUE 70. STM: {(» GOAL COMPARE) (• GOAL MORE) (BLUE > SQUARE) (OLD OS 1 (SQUARE NOT BLUE)) (OLD QS 2 (BLUE NOT SQUARE)) (• GOA L QUANTIFY (SQUARE NOT BLUE)) (OLD CHNK (« GOAL QUANTIFY (SQUARE NOT BLUE)) (• GOAL COMPARE) (OLD QS 2 (BLUE NOT SQUARE)) ( » GOAL MORE)) (GOAL * SUBIT) (OLD SUB 1) (ELM RED SQUARE)) PI TRIE XO: BLUE : > XI: SQUARE 75. STM: ((OLD GOAL MORE) (BLUE > SQUARE) (• GOAL COMPARE) (OLD QS 1 (SQUARE NOT BLUE)) (OLD QS 2 (BLUE NOT SQUARE)) (« G OAL QUANTIFY (SQUARE NOT BLUE)) (OLD (M4K (* GOAL QUANTIFY (SQUARE NOT BLUE)) <•> GOAL COMPARE) (OLD QS 2 (BLUE NOT SQUARE)) (OU) GOAL MORE)) (GOAL * SUBIT) (OLD SUB 1) (ELM RED SQUARE))

environment at a finer level of detail than is necessary for effective interaction. Attending to and processing objects as global wholes are preferred, ceteris paribus, to considering separately the dimensions on which the objects vary. Striking evidence of variations in performance which are attributable to variations in the level of detail being utilized by the system can be obtained (Wallace, 1972b) if children between 5Vi and 6 years are confronted with a typical ci test collection (e.g., blue and red squares) and then asked, first, "Show me all of the squares" and, subsequently, "Count all of the squares." The vast majority of the children will have no difficulty in pointing to all of the objects in the first task, but a considerable number of them will count only the members of the major or minor subset when seeking an answer to the second question. The "show me" instruction apparently leads to the activation of a relatively simple production system which functions in a dimension-oriented fashion: the value square is concentrated upon to the exclusion of consideration of each object as a whole. In contrast, the "count" instruction activates a complex production system, in which a global object orientation is adopted. 169

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY The practical effect of the global object level of detail is to substitute red square or blue square for square as the value being quantified. If these two tasks are administered to children aged between 6V2 and 7 years, neither task will present any difficulty. In the interim the system has discovered that effective quantifiers must function at the finer level of detail represented by the dimensional orientation. Thereafter, when activated directly by verbal instructions, quantifiers will lead to dimensionoriented processing. The outcome may be very different, however, if quantifiers are not activated directly but are called upon to function in the service of another production system. This is precisely the situation when PSQC2 calls PSUBIT. Class Inclusion. Figure 8 shows a trace of PSQC2 on the class-inclusion question, More blues or more squares? At 4., PDV1 creates the two target values as global objects. Then the main productions are satisfied in the same order as in PSQC1: P3, P4, PA, P9, PA, and P l l . The action chain evoked by Pll fires one of the chunking productions (16.), inserts the subitizing goal in STM, and calls up PSUBIT. The first production in PSUBIT to be satisfied is PDS1, which calls the terminal for outputs from NOTICE. The global object providing the context at this point is (BLUE NOT SQUARE). There are no such objects in the display, but NOTICE is satisfied with a partial match, so it returns the two blue things as elements. PSUBIT operates from 16. to 30., finally producing a quantitative symbol. At this point PSUBIT returns control to PSQC2 with the following information in STM: the quantitative symbol, a satisfied subitizing goal, some "garbage" remaining from subitizing, the CHNK of things chunked before subitizing, and the quantification goal that led to the just completed pass through PSUBIT. Next the CHNK is unpacked (35.) and the items that PSOC2 needs to continue are made individually accessible in STM. From here on PSOC2 functions like PSQC1 except for a repetition of the additional operations just noted. The second quantification takes place in the context of (SQUARE NOT BLUE) and the single red square is noticed and quantified (46.-5V.). Once the system has both quantitative symbols it can RELATE them (64.). RELATE, the process that determines the ordinal relation of two symbols, is still written as a terminal call. The final steps (6S.-75.) are similar to those in the trace of PSQC1.* * The system used 75 actions to complete the task, of which about 25 were generated by subitizing. Earlier estimates of about 100 ms per action (Simon, 1972) for

170

DAVID KLAHR

Transition and Change In order to advance our understanding of the developmental process, we must have some representation of what it is that is changing. Our approach has been to concentrate on a highly specific model for a relatively narrow, but important, class of tasks. Our strategic orientation is summarized by Newell (1972b, p. 375): "I will, on balance, prefer to start with a grossly imperfect but complete model, hoping to improve it eventually; rather than start with an abstract but experimentally verified characterization, hoping to specify it further eventually. These may be looked at simply as different approximating sequences toward the same scientific end. They do dictate quite different approaches." In the models presented here, the assumptions are clear, explicit, and, in most cases, subject to empirical refutation. The models are also quite clear with respect to what is not well understood. Each place where a terminal call is made is a potential point of departure for further model building. Two of these situations involve the interface between the environment and the organism. The initial input is the result of an unspecified linguistic processor, and the actual admitting of elements into STM, via NOTICE, involves perceptual processes that remain unmodeled. However, there is enough substance to these models to permit us to consider the guiding question for all of our efforts, How does the system develop? The Development of Conservation Rules. Quantitative comparison and quantification operators play a role in conservation tasks. They are crucial to conservation performance in an obvious way: the initial relation between the two amounts in the equivalence conservation task (Elkind, 1967) is established via the quantitative comparison rules. However, the assessment of having conservation rests upon the ability of the child to assert the continuing quantitative equality of the two collections without resorting to a requantification and comparison after observing a perceptual transformation. The conserving child has a rule which says, in effect: if the quantities were initially equal (as determined by a quantitative comparison system) and if one of them underwent a perceptual transformation, then the quantities must still be equal. There have been some recent indications that one can separate empirically the effects of the 50 non-subitizing actions would yield a predicted latency of around 5 sec for ci. Wallace (1972a) found a mean latency of approximately 4 sec for ci with young children. Although estimates of the duration of actions are still quite speculative, the model is not inconsistent with what latency data there are.

171

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY quantification operators from the effects of conservation rules (Gelman, 1972a&b). Further analysis indicates that the development of conservation rules is dependent upon the emergence of the quantification operators (Klahr & Wallace, 1973). The development of one quantification operator facilitates the establishment of rules (i.e., productions) that, in turn, facilitate the development of another quantification operator. The transition mechanism suggested in our account of conservation (Klahr & Wallace, 1973) is stated only at the verbal level, but the entity which is undergoing change is relatively specific: a production system. The general argument is based upon a principle of redundancy elimination. The developing system constantly searches for consistent sequences which enable it to eliminate redundant processing. A consistent sequence is an internal representation of environmental inputs and system processes that always yield the same result. It is not simply an environmental regularity, but rather regularity arising from the interaction between the environment and the system. A typical sequence might consist of an initial knowledge state, an operation upon that state, and a final test of the result of the operation. If the environment were such that a specific state-operator-test sequence repeatedly produced the same result, then the final test would become redundant. The redundancy elimination principle ultimately leads the system to dispense with the final test, and instead to retrieve from LTM the result of previous test applications for that specific state-operator sequence. This retrieval is tantamount to prediction of what the outcome would have been if the test had actually been made. At some intermediate stage between the initial appearance of the LTM representation of the sequence and total reliance upon it, the system uses both modes of operation. The LTM representation is used as a predictor, and the actual test is carried out to verify that prediction. Thus, the system would "grow" production rules from short sequences of empirical regularities. Training Procedures. Another form of development appears to be involved in the results of most training studies. In many cases the child seems to have all the component parts for successful task performance, but still fails on the task itself. Some training procedures can be interpreted as modifications in the action side of certain productions, modifications that ensure the evocation of the appropriate related productions. Another effect of training appears to be a change in the internal repre172

DAVID KLAHR sentation generated by perceptual and verbal encoders. As noted earlier, the transition from PSQC2 to a successful performance on ci requires only that target values rather than target objects be generated by PDV1. How is this change effected? The only substantive evidence relevant to this question emanates from the results of training studies in which an attempt has been made to inculcate successful ci performance (Kohnstamm, 1967). The most effective procedure appears to be one in which children who initially fail ci are repeatedly instructed to count the superset, to count the major subset, and then to answer the ci question with feedback on the accuracy of their response. The success of this procedure can be explained as follows. Children whose performance improves as a result of this type of training are in a developmental state in which their counting quantifier, PSCOUNT, when activated directly by verbal instructions, engages in dimension oriented processing. Quantitative comparison, in contrast, takes place via PSQC2 at the global object level. The initial instructions to count lead to two passes through PSCOUNT on a dimensional basis, both passes culminating in knowledge that successful results have been obtained. The ci question which follows activates PSQC2, produces global object orientation, and ends with feedback that an inaccurate response has been produced. Repetition of this sequence of events which places dimensional processing, followed by success, in juxtaposition to global object processing, followed by failure, appears to maximize the likelihood of a decision by the system that a change in the level of detail is required. This line of argument is supported by the detailed protocols of the instructional sessions provided by Kohnstamm (1967). The feedback following erroneous ci responses was frequently of a type which emphasized the dimensional as opposed to the global object aspect of the test collection. A child who, when confronted with three red squares and two red circles, only quantified the two red circles would be asked by the experimenter, "Aren't these [the squares] red too?" This account of transition from failure to success on ci fares reasonably well when assessed on the criterion of generality. A change of attention from a global object to a dimensional orientation constitutes a plausible explanation of transition from failure to success in children's performance on two dimensional series completion problems (Klahr & Wallace, 1970b). Such a change of attention also constitutes an essential 173

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY feature of a theory of the acquisition of conservation of continuous quantity (Klahr& Wallace, 1973). Changes in the Processor. The changes discussed thus far are in various parts of production systems. Underlying the systems is the interpretive mechanism itself. Although one may speculate upon the language in which the interpreter is written, then speculate upon the interpreter for that language, and so forth, we prefer to make the simplifying assumption that the ability to interpret productions is the substrate of the organism's processing capacity: it is wired in. There are two consequences of this assumption. First, when the transition mechanism is ultimately specified, it too will take the form of a production system, interpretable by essentially the same rules as govern the operation of the performance systems. Second, the power of the production system interpreter must be equally available to both the learning and the performance production systems. Although the interpreter is assumed to be wired in, it may increase its capacity with development. It may run faster, it may increase STM, or it may develop more complex rules for scanning STM and matching elements with conditions. These are all issues that are as yet undecided, but which are not undecidable.

Conclusion In this paper I have attempted to demonstrate, through the presentation of concrete examples, some of the properties of our approach to an information-processing theory of cognitive development. The specific models presented here are deficient in several respects, and will undoubtedly yield to stronger and more complete versions. However, the intent, in this paper, has not been to justify the models per se, but rather to present them as exemplars of the general paradigm. Perhaps the most striking feature of these models is their apparent complexity. In one sense this impression of complexity is false, for the basic model of the organism as a production system interpreter is quite simple. It is the specific instances of the system operating in particular task environments that necessitate most of the complexities. In another sense, the complexity is an inevitable side effect of increased precision. Stated most unequivocally, "Simple models will just not do for human cognition" (Flavell & Wohlwill, 1969, p. 74). Although precision has its price — complexity — it also buys something 174

DAVID KLAHR quite valuable. It enables us substantially to close the gap between theory and data, in the manner described in the discussion of the series of models for quantitative comparison. Based upon a precise model of a complex task, we can begin to make strong predictions about the effects of procedural variations in our traditional experimental tasks. Such predictions will rest upon hypotheses about the utilization of long- and short-term memory, about forms of representation, and about basic information processes. The modest progress we have seen thus far in the description of models of different stages, makes us optimistic about the possibility that similar advances will be made in understanding and describing the development of the processor. References Baylor, G. W., Gascon, J., Lemoyne, G., & Pothier, N. An information processing model of some sedation tasks. Department of Psychology, University of Montreal, 1972, MCP No. 8. Brown, R. Introduction. In Cognitive development in children. Chicago: University of Chicago Press, 1970. Pp. ix-xii. Burstall, R. M., Collins, I. S., & Popplestone, R. S. POP - 2 papers. Edinburgh: Oliver & Boyd, 1968. Elkind, D. Piaget's conservation problems. Child Development, 1967, 38, 15-27. Farnham-Diggory, S. Cognitive synthesis in Negro and white children. Monographs of the Society for Research in Child Development, 1970, 35 (2, Serial No. 135). . The development of equivalence systems. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 43-64. Feigenbaum, E. A. The simulation of verbal learning behavior. In E. A. Feigenbaum & I. Feldman (Eds.), Computers and thought. New York: McGraw-Hill, 1963. Pp. 297-309. Flavell, I. H. The developmental psychology of Jean Piaget. Princeton: Van Nostrand, 1963. . Stage-related properties of cognitive development. Cognitive Psychology, 1971,2,421-453. . An analysis of cognitive-developmental sequences. Genetic Psychology Monographs, 1972, 86, 279-350. & Wohlwill, I. F. Formal and functional aspects of cognitive development. In D. Elkind & J. H. Flavell (Eds.), Studies in cognitive development. New York: Oxford University Press, 1969. Pp. 67-120. Gagne, R. The conditions of learning. New York: Holt, 1970. Glaser, R., & Resnick, L. B. Instructional psychology. Annual Review of Psychology, 1972, 23, 207-276. Gelman, R. Logical capacity of very young children: Number invariance rules. Child Development, 1972, 43, 75-90. (a). The nature and development of early number concepts. In H. Reese (Ed.), Advances in child development. Vol. 7. New York: Academic Press, 1972. (b). Gregg, L. W., & Simon, H. A. Process models and stochastic theories of simple concept formation. Journal of Mathematical Psychology, 1967, 4, 246-276. Halford, G. S. A theory of the acquisition of conservation. Psychological Review, 1970,77,302-316.

175

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Hayes, J. R. The child's conception of the experimenter. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 175-181. Inhelder, B., & Piaget, J. The early growth of logic in the child: Classification and seriation. London: Routledge, 1964. Jensen, E. M., Reese, E. P., & Reese, T. W. The subitizing and counting of visually presented fields of dots. Journal of Psychology, 1950, 30, 363-392. Kaufman, E. L., Lord, M. W., Reese, T. W., & Volkmann, J. The discrimination of visual number. American Journal of Psychology, 1949, 62, 498-525. Klahr, D. Some production systems for quantification tasks. Carnegie-Mellon University, GSIA Working Paper, 1972. . A production system for counting, subiti/ing and adding. In W. G. Chase (Ed.), Visual information processing. New York: Academic Press, 1973. (a) . Quantification processes. In W. G. Chase (Ed.), Visual information processing. New York: Academic Press, 1973. (b) & Wallace, J. G. The development of serial completion strategies: An information processing analysis. British Journal of Psychology, 1970, 61, 243257. (a) An information processing analysis of some Piagetian experimental tasks. Cognitive Psychology, 1970, 1, 358-387. (b) -. Class inclusion processes. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 143-172. (a) Reply to Hayes: On the value of theoretical precision. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 183-186. (b) The role of quantification operators in the development of conservation of quantity. Cognitive Psychology, 1973, 4. Kofsky, E. Developmental scalogram analysis of classificatory behavior. Ph.D. thesis, University of Rochester, 1963. . A scalogram study of classificatory development. Child Development, 1966, 37, 191-204. Kohnstamm, G. A. Piaget's analysis of class inclusion: Right or wrong? The Hague: Mouton, 1967. Laurendeau, M., & Pinard, A. Reflexions sur 1'apprentissage des structures logiques. In F. Bresson & M. de Montmollin (Eds.), Psychologic et epistemologie genet.iques. Paris: Dunod, 1966. Pp. 191-210. Miller, G. A. The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 1956, 63, 81-97. , Galanter, E., & Pribram, K. H. Plans and the structure of behavior. New York: Holt, 1960. Newell, A. Studies in problem solving: Subject 3 on the Crypt-arithmetic task DONALD + GERALD = ROBERT. Carnegie Institute of Technology, July 1966. . A note on process-structure distinctions in developmental psychology. In S. Farnham-Diggory (Ed.) Information processing in children. New York: Academic Press, 1972. Pp. 125-139. (a) A theoretical exploration of mechanisms for coding the stimulus. In A. W. Melton & E. Martin (Eds.) Coding processes in human memory. Washington, D.C.: Winston, 1972. (b) -. Production systems as models of control structures. In W. G. Chase (Ed.), Visual information processing. New York: Academic Press, 1973. & Simon, H. A. Human problem solving. Englewood Cliffs: Prentice-Hall, 1972. Norman, D. A. Memory and attention. New York: Wiley, 1969.

176

DAVID KLAHR Piaget, J. The child's conception of number. New York: Humanities Press, 1952. Pinard, A., & Laurendeau, M. Stage in Piaget's cognitive-developmental theory: Exegenesis of a concept. In D. Elkind & J. H. Flavell (Eds.), Studies in cognitive development, New York: Oxford University Press, 1969. Pp. 138-145. Popper, K. R. The logic of scientific discovery. London: Hutchinson, 1959. Quillian, R. W. Word concepts: A theory and simulation of some basic semantic capabilities. Behavioral Science, 1967, 5, 410-430. Reitman, W. Cognition and thought. New York: Wiley, 1965. Resnick, L. B., Wang, M. C., & Kaplan, J. Behavior analysis in curriculum design: A hierarchically sequenced introductory mathematics curriculum. Monograph 2, LRDC, University of Pittsburgh, 1970. Saltzman, I. J., & Garner, W. R. Reaction time as a measure of span of attention. Journal of Psychology, 1948, 25, 227-241. Simon, H. A. An information processing theory of intellectual development. In . W. Kessen & C. Kuhlman (Eds.), Thought in the young child. Monographs of the Society for Research in Child Development, 1962, 27 (2, Serial No. 82). Pp. 150-155. . Motivational emotional controls of cognition. Psychological Review, 1967, 74, 29-39. . On the development of the processor. In S. Farnham-Diggory (Ed.), Information processing in children. New York: Academic Press, 1972. Pp. 3-22. Smedslund, J. Concrete reasoning: A study of intellectual development. Monographs of the Society for Research in Child Development, 1964, 29 (2, Serial No. 93). . Microanalysis of concrete reasoning: III. Theoretical overview. Scandinavian Journal of Psychology, 1966,7, 164-167. Sternberg, S. Memory-scanning: Mental processes revealed by reaction-time experiments. American Scientist, 1969, 57, 421-457. Taves, E. H. Two mechanisms for the perception of visual numerousness. Archives of Psychology, 1941, 39, 1-47. Wallace, J. G. The adaptation of instruction to individual differences: An information processing approach. Paper presented at the meeting of the American Psychological Association, September 1972. (a) . Class inclusion performance in children: Information processing theories and experimental studies. Paper presented at the 20th International Congress of Psychology, Tokyo, August, 1972. (b) Wohlwill, J. F. Responses to class inclusion questions for verbally and pictorially presented items. Child Development, 1968, 39, 449-465. Woodworth, R. S., & Schlosberg, H. Experimental psychology. (Rev. ed.) New York: Holt, 1954. Young, R. Production systems for children's sedation behavior. Carnegie-Mellon University, CIP Working Paper No. 202, 1971.

177

«> H A R R I E T L. R H E I N G O L D <3>

Independent Behavior of the Human Infant

THIS paper places a new construction on current views of infant behavior, especially of the infant's social behavior. I propose that some of the infant's behavior heretofore treated as dependence qualifies as the opposite — independence. I hold no brief for the terms dependence and independence. Indeed, a salutary outcome of the evidence here presented and weighed would be the abandoning of these terms in favor of a more precise delineating of the behaviors that compose them. However, to draw attention to an important but ignored characteristic of the infant and young child's behavior, I shall present the argument within the framework of these global terms. It is easy to speak of independence but difficult to define the term with the precision required for scientific study. "All of us," said George Eliot, "get our thoughts entangled in metaphors . . ." Independence is indeed a resplendent metaphor in common usage and no less florid in much of the psychological literature. It helps, but only a little, to adopt our usual procedure of substituting an adjective for the noun, and to translate "independence" into "independent activities" or "independent behavior." The task of specifying independent behavior, however, still remains. Independent activities may be defined quite simply as self-directed, NOTE: Tfre work on this paper was supported by United States Public Health Service Grants HD-23620 and HD-01107. For assistance in the conduct of the main study reported here and the analysis of its data I thank David R. Colby, Kaye V. Cook, Thomas C. Guthrie, Dale F. Hay, Sara P. Sparrow, Heinz J. Whitefoot, Christine B. Wing, and Merilyn G. Yates. I am grateful to H. Rudolph Schaffer, who helped clarify some of the ideas, and especially to Carol O. Eckerman, a valued colleague in this and other studies.

178

HARRIET L. RHEINGOLD self-guided, and self-initiated activities. But almost all behavior is selfdetermined — in the new infant as in the adult. Similarly, all behavior may be conceived as dependent — dependent upon other persons, the environment, or past experience. Thus, when the child asks his parent for help, he is being at once dependent and independent, because the request is made of others but is also initiated by him. This complicated conceptual problem has not yet been faced by psychologists. The relevant literature on independent behavior during the early years of life is particularly sparse. In many textbooks independence receives some brief mention past the first year of life — usually limited to negativism and toilet training — but not until adolescence, when presumably it constitutes a problem, does it receive a larger measure of attention. In the years in between, the topic is more often related to socialization practices of the parents or to the concept of achievement (e.g., Parke, 1969). Independent behavior is intimately related to such other classes of behavior as competence, effectance, and adaptive coping; to play and exploratory behavior; and to socialization and learning. Each of these classes is important in its own right, but I propose that a separate class of behavior may be distinguished from them — in fact, that these very classes of behavior may be viewed as serving self-directed, self-initiated behavior, that is, as serving independent behavior. THE CONCEPT OF INDEPENDENCE IN THE LITERATURE

In many current textbooks of child development, the independent behavior of the year-old child — the age of major concern in this paper — receives scant attention in contrast to that accorded dependence and attachment. Conspicuous, too, is the dearth of research findings. For example, in Child Development and Personality (Mussen, Conger, & Kagan, 1969) independence is treated under the heading of socialization; the role of the parents in directing the child's behavior with particular respect to aggression and dependent behavior is emphasized. The section that comes closest to treating independence is labeled "socialization and autonomy," takes up less than a page, and once again focuses more on the effect of parental behavior than on the child's behavior. Yet more than three pages are devoted to toilet training, and seven pages under the heading of "Additional Motives of the 2-Year-Old" are devoted 179

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY to fear, anxiety, and aggression. Textbooks do reflect the areas of research; one can conclude, therefore, that the attention of investigators has been drawn more often to the conflicts of life than to its positive achievements. Independence receives more attention in the textbook Psychology of the Child (Watson, 1965). Although here, too, more space is accorded dependence and aggression than independence, five and a half pages are devoted to the behavior of interest undjer the headings of "Exploratory Externally Aroused Behavior Tendencies" and "The Development of Self and Social Awareness." The author's graceful statement early in the chapter that "with certain distinguished exceptions, observational and anecdotal material and plausible but unverified appeals to common experiences are sources of ... evidence" (p. 226) emphasizes the absence of controlled studies on these topics. In still another textbook, Childhood and Adolescence (Stone & Church, 1968), the chapter labeled "The Toddler" does present a detailed and lively account of the growth of selfinitiated behavior, much of it based on the authors' sensitive observations of everyday behavior rather than on controlled studies. In the psychoanalytic literature, independence in general has received considerable attention under different terms: for example, Hartmann's (1951) "primary autonomous functions of the ego" and Hendrick's (1942) "instinct to master." Independence in the very young child has been described by Erikson (1950) under the term autonomy and by Mahler (1968) under the term separation-individuation. Mittelmann's (1954) motility urge also is related to the behavior of interest here. I treat these contributions in so summary a fashion because a fair share of the psychoanalytic literature is more theoretical than empirical, and in general has a therapeutic intent. In the psychological literature, current considerations of dependence and independence take as a point of origin the formulations of Beller (1955) and Heathers (1955a, 1955b). On the basis of observations of the free play of preschool children they sought to establish independence and dependence as motivational constructs. Beller (1955) asked whether the relationships between the components of "dependency and independent behavior" satisfied the concept of drive and, further, whether dependence and independence were correlated. The components of dependency, "derived conceptually through an analysis of the child's early experience with parents and the physical environment" (p. 34), were seeking help, 180

HARRIET L. RHEINGOLD seeking physical contact, seeking proximity, seeking attention, and seeking recognition. The components of independence were taking initiative, trying to overcome obstacles in the environment, trying to carry activities to completion, getting satisfaction from work, and trying to do routine tasks by oneself. On the basis of ratings of the children's behavior, he obtained evidence of a dependence drive, an independence drive, and a moderately negative correlation between them. He interpreted the latter finding as casting doubt on the extent to which dependence and independence could be considered as the limiting points of a single dimension. Heathers (1955a), using running-account observational time samples of behavior, concluded that emotional dependence on adults shifted with age to dependence on peers, and that dependence was not consistently correlated, either positively or negatively, with independence. In a subsequent theoretical paper, Heathers (1955b) distinguished between four terms: instrumental and emotional dependence, and instrumental and emotional independence. The terms set forth by Beller and by Heathers to define the concepts of dependence and independence are still used; they may be found in recent reviews of the concepts by Ainsworth (1969), Gewirtz (1969), Hartup (1963), Hetherington (1970), Maccoby and Masters (1970), and Parke (1969). The conclusions of Beller and of Heathers that dependence and independence are not just opposite ends of a single continuum are also accepted in these reviews. The instrumental and emotional dependence and independence of Heathers (1955b) have been collapsed by Bandura and Walters (1963) into a distinction between task-oriented and person-oriented dependency (and presumably, independence). In these reviews, and their theoretical considerations, one finds, as in the textbooks, a greater emphasis on dependence than on independence. Independent behavior in the infant and young child receives a fuller treatment in two recent books on the development of personality. The opening lines of Wenar's (1971) chapter entitled "Initiative, Willfulness, and Negativism" state: "As love is central to the first year of life, initiative is central to the second year. In an upsurge of independence, the toddler literally and figuratively stands on his own two feet and turns his back on his mother. He avidly explores the tangible world of objects and the subtle world of human relations. As he does so, he begins to sense his capabilities and his power, to develop an embryonic notion of himself as an agent, a doer, a causer. No sooner has this happened than his inability 181

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY to moderate and modulate propels him into a stage of marked willfulness and negativism" (p. 52). The chapter is marked by the same sensitive reporting to be found in Stone and Church's (1968) textbook. White's (1972) treatment covers the same topic under the heading of "The Early Growth of Competence." Finally, two papers that report research findings about the second year •of life are especially appropriate for the present argument. In the first, Levy (1953) reported that "resistant behaviors" during physical examinations and intelligence tests reached their maximums during the second year of life. More recently, Wenar (1972) reported on the free play activities of children in their own homes in the presence of their mothers. He rated the duration, intensity, and level of complexity of the child's self-initiated locomotive, manipulative, and visually regarding activities. Wenar also rated the child's spontaneous responses to the mother, as well as expressions of positive and negative affect. Wenar found that the child engaged in transactions with the physical environment more often and for longer periods of time than in transactions with the mother. More affect, both positive and negative, accompanied interactions with the mother. The child's interactions with the mother took the form — in order of decreasing frequency — of approaching and observing her, bids for her attention, attempts to get her to participate in his transactions with the environment, and bids for physical contact. He found no sex differences in the behaviors he studied. Wenar speculated that the child's attentiongetting behavior shows "healthy pride," and that his bids for participation suggest that he is taking the initiative in making exploratory activities a social rather than a solitary pursuit. IMPLICATIONS OF PREVIOUS RESEARCH IN THIS LABORATORY

The idea that much of the infant's and the young child's behavior can be viewed as independent occurred to me only after a long series of studies. To trace the origin of the idea I went back many years to a study comparing the amounts and kinds of caretaking in homes and in an institution (Rheingold, 1960). During those long hours of observing infants who were 3 months old, I was struck by how busily and cheerfully the infants in the institution occupied their waking hours with so few toys available. Surely their behavior was self-initiated. Some eight years ago I began a series of studies designed to measure 182

HARRIET L. RHEINGOLD the effect of a strange environment on the behavior of infants at 10 months of age. Only after the last sentence of the discussion of the study was written did I realize that it was not so much the strange environment that caused the distress of the children placed in it without their mothers, nor even the absence of their mothers, as it was being placed and left alone (Rheingold, 1969a). That this was so was demonstrated in a later study in which infants the same age were given the opportunity to leave their mothers and enter that same strange environment by themselves (Rheingold & Eckerman, 1969). All the children did enter on their own initiative, even when the environment contained no toy. Not only did they enter, but they crept to places in the room from which they could not see the mother. They returned to the mother's room, left again, and returned again — some infants many times — but 'on a third of the returns they did not contact the mother. In that study we recorded the observation that when the strange environment contained a toy, some infants, as they grasped the toy, looked at the mother, smiled, and vocalized "as though sharing a pleasant experience"*(p. 280), an observation that foreshadowed the theme of the present paper. About this time we stopped calling our rooms in the laboratory a "strange" environment; "unfamiliar" or "new" seemed more fitting terms. In another study, carried out at about the same time, we found that although the mother's presence encouraged exploratory behavior in a relatively empty environment, her presence did not support such behavior for long, given that the mother's response to the child was restricted to observing and smiling (Rheingold & Samuels, 1969). When toys were provided midway in the 20-minute session for one group of 10-month-old infants, only one fussed; when none was provided for the other group, all but one infant fussed. We had of course measured the obvious, that the behavior of infants in an unstimulating environment deteriorates into fussing. It was becoming clear that in the process of investigating the effect of different environments, we were seeing infants move away from their mothers. Although such moving away is an everyday occurrence, the current emphasis on attachment made the infants' detachment seem like a new idea. Even today, visitors to our laboratory marvel at seeing a young infant crawl on his belly, often with great effort but with single-minded purpose, toward a toy in another room 5 meters away. The visitors say, "And look! He leaves without crying." When one considers that most of 183

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY these skeptics had children of their own, and saw this behavior again and again from the time their children first achieved locomotion, one can only stand in awe of the power of the concepts in our minds to erase what our eyes see. A review of the nonhuman primate literature (Rheingold & Eckerman, 1970) added support to the concept of the infant's detachment from his mother. The data showed that as the nonhuman primate infant grows older, he leaves his mother more frequently, goes farther, and stays away longer. To obtain evidence on how far from his mother a human child would go, we placed mother and child in a backyard; the mother was seated and the child was left free. We found a positive relationship between the age of the children (from 1 to 5 years) and the distance they traveled from their mothers. As instructive as that relationship, however, were certain incidental observations: first, the older children talked to their mothers — presumably about the environment and what they were doing — even as they moved away; second, the younger children often brought little things they found to their mothers — pebbles, leaves, and so on. These observations were no more than interesting at the time; now they have become worthy of study in their own right. It was these studies and others by Corter (Corter, Rheingold, & Eckerman, 1972) and Ross (Ross, Rheingold, & Eckerman, 1972), who were then graduate students, considered together with the proposal (Rheingold, 1969b) that the human infant during the first year of life socializes others more than he is socialized by them, that led to the next insight: that the infant did not have to wait to grow up to become independent; he was already behaving in an independent fashion. Once again, the insight came only slowly to what now seems obvious. But how to study independent behavior? What are its hallmarks? How can one define so fluid a concept?

Enterprising Activities in the Second Year of Life Candidates for demonstrating the nature of the child's independent behavior were his activities in a free play situation, his efforts to overcome obstacles placed between him and a desired goal, resisting distractions, opposing his "will" to that of others, and feeding and dressing himself. Of these, we selected for study the child's enterprising activities in a free play situation. By enterprising activities was meant allowing the child to 184

HARRIET L. RHEINGOLD take the initiative in determining his own ways of dealing with a relatively unstructured environment. Enterprising is defined as "marked by an independent energetic spirit and by readiness to undertake and experiment" (Webster's Dictionary, 1969), a definition that seemed to fit the behavior of the infants and toddlers we have seen so often in the laboratory. We measured the nature, frequency, sequence, and duration of the child's use of space, toys, and his mother. We sought to find how quickly he engaged himself with the toys, how he distributed his time between mother and toys, and what was the nature of his interactions with the mother. In addition we tried to get some recording of his affective state. The mother was an element in the situation for several reasons. Her presence ensured that the child would tolerate and use the rest of the environment. The child's responses to her could be contrasted with those to the toys. More important, the nature of his responses to her could be assessed for the extent to which they might qualify as dependent or independent. Attention was paid to his affective state — again, for several reasons. Is dependency called forth by stress or anxiety? Is independent behavior accompanied by good humor? The eagerness, gaiety — yes, happiness — we have often observed in the laboratory as the child plays there should not be ignored, any more than should his fussiness at other times. Yet, although as students of infant behavior we recognize fussing and crying as indexes of some unpleasant emotional state, we have no parallel vocal indexes of pleasurable emotions. The children of this study were 12 and 18 months of age. For three reasons, the second year of life — rather than the first, where most of our previous work had been concentrated — was chosen to study the next step "in the continuous process of achieving psychological independence" (Rheingold & Eckerman, 1970, p. 83). First, children during the second year of life are capable of a greater variety of responses, and locomotion is easier. Second, "negativism," however defined, appears during this period, and although not the topic of investigation here, this study could provide some base-line data against which negativism could later be studied. Third, we have little information about behavior at this age. But let me state firmly that the age was not chosen because we thought independent behavior could not have been demonstrated at an earlier age. The ages of 12 and 18 months were used on the assumption that the child becomes more independent with age, and therefore that the weigh185

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ing of similarities and differences would bring the relevant behaviors into sharper focus. The two sexes were also equally represented because of evidence that boys at 13 months of age exhibit some behaviors that might be construed as showing more independence than girls (e.g., Goldberg & Lewis, 1969 ;Messer& Lewis, 1972). OBTAINING THE DATA

The Children. The sample of children was drawn on the basis of age and sex from the population of infants born at the North Carolina Memorial Hospital. Mothers of potential subjects were telephoned; more than 90 per cent agreed to participate. The sample consisted of 40 children, 20 at 12 months of age (mean, 12.5; range, 11.6 to 13.0) and 20 at 18 months of age (mean, 18.4; range, 17.6 to 18.9), evenly divided by sex. The average number of years of education was 17.4 for the fathers and 15.0 for the mothers; half of the parents were students. Both age and sex groups appeared by inspection to be similar in these subject variables. The records of 7 other children were not used: 4 fussed during a major portion of the trial, 2 fell and hurt themselves enough to cry, and 1 child played with none of the toys. Each child was observed individually. The 40 children were scheduled in ten consecutive sets, each set consisting of a male and female at each age, to control for any possible variations in subjects, procedures, or the laboratory environment during the course of the study. Furthermore, the first 20 subjects comprised Cohort 1; the second 20, Cohort 2. Each cohort then was equally balanced by age and sex; and the second cohort was considered a replication of the first. Test Environment. The test environment is shown in Figure 1. It consisted of three adjoining rooms: the mother's room and two (smaller) toy rooms. A narrow strip of masking tape outlined a circle of 1-meter radius around the mother's chair. Each toy room contained a sink, a small table, and two toys. The toys in the left-hand room were a brightly colored plastic dump truck and a shape-sorting box with a hinged lid and two sets of five different shapes, each shape a different color. The right-hand room contained, in the near corner, a boat train of four interlocking parts, and in the far corner, a set of five nesting wooden boxes. (The last three toys are distributed by Creative Playthings.) The far walls had one-way windows for observing the children. Photo186

HARRIET L. RHEINGOLD

Figure 1. The experimental environment for studying the enterprising activities of year-old children.

cells at the thresholds of the toy rooms recorded each time a child entered or left. Microphones in the rooms transmitted the child's vocalizations to the observers outside the experimental rooms. Procedure. The experimenter met the mother and child in the reception room. While the child played with a few simple toys, the experimenter instructed the mother in the role she was to assume during the test. She was to remain seated. She could look toward and smile at her child, and say a few words to him, but only in response to his overtures. She was not to tell him what to do or to play games with him; instead he was to be left free to do as he pleased. She was to accept any toy the child handed her, and she could offer support if he persistently requested to get on her chair or lap. The mother then carried the child into the test environment. The 20minute trial started when she sat down and placed the child in front of her. Measures. From the photocell records we obtain the latency to enter each toy room, the duration of time spent in each of the three rooms, the frequency of transits between rooms, and the time when they occurred. 187

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Observers behind the one-way windows depressed silent keys to record the latency, duration, and time of occurrence of the following behaviors: a. Contact of a toy, any manual contact of a toy from simple holding to precise combining of parts. b. Contact of a toy while in mother's circle, the portion of contact of a toy that occurred in the mother's circle, whether in contact with the mother or not. c. Simple combining of toys, any combining of the toys or their parts, such as putting one in or on another, taking one toy out of another, or hitting two together; a portion of contact of a toy. d. Precise combining of toys, any inserting of a shape in the shapesorting box, placing of one nesting box in or precisely on another, or connecting the parts of the boat train; also a portion of contact of a toy. e. Contact of the mother, any physical contact of the mother in the absence of contact of a toy. f. Locomotion, movement by any means without holding a toy. In addition, another observer dictated a narrative account of the child's activities; the details will be presented later. For calculating the reliability with which the timed measures were taken, two observers independently recorded them on 33 of the 40 records. The sums obtained by both observers for the 20 minutes of the trial were calculated for each child on each measure; the smaller of the two sums divided by the larger yielded a series of percentages for each measure. The lowest median percentage agreement was 92 (range, 59 to 99) for locomotion; the highest was 100 (range, 36 to 100) for precise combining of toys. The median percentage agreement of all measures was 95. For the statistical analysis, the measures of latency and duration were computed as proportions of 1,200 seconds (the duration of the trial) and subjected to arcsin transformations. The transformed means of logically related measures were examined by multivariate analyses of variance for the effects of cohort, age, and sex. If the analyses yielded reliable differences, the individual measures were then examined by univariate analyses of variance (Cramer & Bock, 1966). THE FINDINGS OF THE TIMED MEASURES

The multivariate analyses of variance of the frequency of transits and of the timed measures yielded one significant main effect, for age (F = 10.21, d f = 10/23, p<.001). Furthermore, there was no indication of 188

HARRIET L. RHEINGOLD any interaction of age with cohort or sex. The results of the univariate analyses of the measures are therefore presented. Furthermore, because of the absence of a cohort effect, the data from the two cohorts were pooled, but for the reader's inspection the data are presented in the tables by sex, as well as by age. Promptness to Engage the Environment. The 12-month-old children entered a toy room at a mean latency of 46 seconds; the 18-month-olds, at a mean latency of 38 seconds (Table 1). One or two children in each group were much slower than the others; the median latencies (28 sec for the younger children and 10 sec for the older) were therefore much shorter, and better portray the typical behavior. The children contacted a toy very shortly after entry: at a mean latency of 11 seconds for the younger children, of 4 seconds for the older ones. Some of the children were struggling to get out of their mothers' arms as soon as they saw the toys; in fact, 7 of the older children had contacted a toy in less than 10 seconds. The means for latency to enter a toy room or to contact a toy did not differ reliably by age. Use of Space. Latency to enter the second toy room was much longer for the younger children than for the older ones (F = 7.76, df — 1/32, p<.01). The 12-month-old children took more than 6l/2 minutes to enter a second room; the 18-month-old children, only 2V2 minutes (Table 1). Only 17 of the younger children entered more than one room but, even for those who did, the mean latency was 5 minutes. Another measure of the child's use of space was the number of times he moved from one room to another. The average frequency of transits was 9 (range, 2 to 32) for the younger children and 26 (range, 8 to 57) for the older ones, a reliable difference (F = 25.38, df - 1/32, p<.001). Table 1. Latency Measures: Means and Ranges in Seconds Male Item

Mean

Range

12-month-old Ss 28.7 6-64 Entry toy room . . 43.8 19-30 Contact toy Entry other toy room . . . . 490.3 17-1194 18-month-old Ss . . 41.3 2-337 Entry toy room . . 44.0 4-343 Contact toy 18-451 Entry other toy room . . . . 144.1

189

Mean

Female Range

Mean

63.9 70.0 311.2

14-218 17-220 88-791

46.3 56.9 395.8

34.3 40.2 150.4

1-144 4-145 21-366

37.8 42.1 147.2

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY The children spent about a third of the time in the toy rooms (Table 2); the younger ones spent somewhat more time there than the older, but not enough to be reliable by test. The two reliable differences reported so far — the shorter latency of 18-month-olds to enter a second room and their greater frequency of transits — were undoubtedly related to the older children's greater proficiency of locomotion. The narrative accounts were scanned for all references to the ways in which the children locomoted. Eleven of the 12month-olds were still creeping, while all of the 18-month-olds walked. Most interesting was the recording of running in 31 per cent of all references to the locomotion of the older children. Even the casual observer notes how often little children run, but the behavior seemed especially impressive in the laboratory. Mittelmann's motility urge indeed! Distribution of Activities. The children of both ages spent more than half the trial contacting one or another of the toys (Table 2). The mean durations, 11.3 minutes for the 12-month-olds and 11.1 for the 18month-olds, did not differ reliably. The older children spent more time contacting a toy in the mother's circle (54 per cent) than did the younger children (30 per cent), as Table 2 shows (F = 4.62, df = 1/32, /?<.05). Contributing to this difference was the fact that only 13 of the younger children brought a toy into the mother's circle, whereas all the older ones did. All but one child contacted other objects in the environment besides the toys — the doorknobs, sinks, photocells, etc. — but for mean durations of less than a minute. Interesting as the child's manipulation of these incidental objects seemed to the observer, it is worth noting that the toys were much more attractive; in fact, the children often contacted these sundry objects with a toy in hand. The 12-month-old children spent approximately 4 minutes in contact with the mother, the 18-month-old children, approximately 2 minutes (Table 2). This difference, although large, was not statistically significant. The mother, then, was contacted less than the toys by both groups. Were these differences reliable? Because the two measures were correlated (r = —.85), the amount of time contacting the mother was calculated as a proportion of the total time contacting either the mother or the toy for each age; these proportions were then tested against the null hypothesis of equal duration for each measure. The tests showed that both the younger and the older children spent reliably less time contacting the 190

HARRIET L. RHEINGOLD Table 2. Duration Measures: Means and Ranges in Seconds Male Item

Mean

Range

12-month-old Ss 26-950 449.0 55-1157 718.1 0-909 184.8 30.0 0-109 0.0 0-802 228.7 11-220 94.1 18-month-oldSs 319.5 Time in toy rooms 180-633 608.7 111-981 Contact of toy Contact of toy in M's circle 319.8 17-708 Simple combining of toys . 59.5 7-148 Precise combining of toys . 2.0 0-12 Contact of mother 161.0 0-444 Locomotion 184.1 70-402 Time in toy rooms Contact of toy Contact of toy in M's circle Simple combining of toys . Precise combining of toys . Contact of mother Locomotion

Female Range Mean

Mean

454.4 632.8 224.2 21.8 0.1 281.0 85.7

125-857 323-1056 0-771 6-62 0-0.9 56-569 29-156

451.7 675.5 204.5 25.9 0.1 254.8 89.9

308.9 718.6 395.8 72.0 3.2 126.9 173.2

51-601 413-892 119-848 8-169 0-22 0-338 80-279

314.2 663.6 357.8 65.7 2.6 143.9 178.7

mother than contacting the toy (by binomial test, p<.05, p<.01, respectively) . In fact, the proportion was only 27 per cent for the younger children and 18 per cent for the older ones. The combined durations of time spent contacting the toys or the mother accounted for the major portion of the children's activities during the trial — for 77.5 per cent of the younger children's activities, for 67.3 per cent of the older children's. The amount of time spent locomoting accounted for another 7.5 per cent among the younger children and 14.9 per cent among the older (Table 2). This difference was reliable (F — 14.69, df = 1/32, p<.001), a rinding that agrees with the older children's greater frequency of transits between the rooms. The older children, it should be noted, spent more time locomoting than contacting the mother (Table 2). The nature of the children's activities with the toys was recorded, in part, by measures of simple and precise combining (Table 2). The older children spent more time in simple combining of the toys than did the younger children (F = 14.10, df = 1/32, p<.001). Because so few of the younger children engaged in any precise combining of the toys, it is appropriate to test the difference by the number of children in each age group who did. The data show that 15 of the older children, but only 2 of the younger children, combined the toys precisely; this difference was re-' liable (p<.001, Fisher Exact Probability Test). We may conclude that

191

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY combining the toys in both a simple and precise fashion was more common among the older children. THE FINDINGS OF THE NARRATIVE ACCOUNTS

One observer dictated a narrative account of the child's behavior. This account was analyzed by 10-second units for the occurrence of (a) the child's behavior in relation to both toy and mother, and (b) expressions of affect. The measure for each category was the frequency of units in which a behavior occurred at least once. To obtain measures of observer agreement, still another observer simultaneously dictated accounts of the behavior of 19 children, of the second cohort, during the middle 10 minutes of the trial. Both observers wore sound-attenuating headsets to ensure independent recording. Behavior in Relation to Both Toy and Mother. Two categories of the child's behavior with both the toy and the mother were analyzed, showing a toy and sharing a toy. The category showing a toy included the child's pointing to a toy or holding it up while looking toward the mother, and his smiling at or vocalizing to her, but without contacting her. This behavior often occurred when the child first contacted a toy, but at other times too. It also occurred more often when the child was at a distance from the mother, but it did occur occasionally when he was in her circle. The category sharing a toy included such activities as bringing a toy to the mother, or putting it in her hand or lap, or manipulating a toy in her lap. Behavior of this type always involved some active interplay of toy and mother; contact of toy and mother by itself was not sufficient for inclusion in this category. The median percentage of observer agreement between the two narrative accounts of the 19 records, summed over both categories, was 92 (range, 0 to 100). The median percentage of agreement between two coders of the main observer's narrative account, summed over both categories, and based on a sample of 16 records drawn at random from each age and sex group, was 95.7 (range, 80 to 100). The mean number of units of showing a toy was 11.9 for each age (range, 0 to 20 for the 12-month-olds; 0 to 32 for the 18-month-olds). The mean number of units of sharing a toy was 9.2 (range, 0 to 67) for the younger children and 26.3 (range, 0 to 86) for the older ones. Sharing thus occurred almost three times more frequently among the older children. Because 8 of the 20 12-month-old children did not bring a toy 192

HARRIET L. RHEINGOLD to the mother at any time during the trial, the two categories were combined for statistical test. The frequency of the units was treated as a proportion of the 120 units of the trial, and subjected to arcsin transformations. An analysis of variance showed a reliable main effect of age (F = 5.28, df = 1/32, p<.05). Combining both categories — "communicating" with the mother about the toys or engaging her in his activities with the toys — we see that the behaviors occurred in a fifth of the units among the younger children, and in about two-fifths among the older ones. The younger children on the average brought only one toy (or part of a toy) to the mother (range, 0 to 3). In contrast, the older children on the average brought five different toys (or parts of toys) to the mother (range, 0 to 10). Some showing behavior also occurred in relation to the child's contact of the non-toy objects in the environment, but the frequency was minimal. Expressions of Affect. A prominent feature of the children's behavior was the pleasure they seemed to experience in the situation. A majority seemed exhilarated by finding the toys; they played with zest, and moved at a fast pace. The observer included in the narrative account mention of such behaviors as smiles (both to toy and mother), laughs, shrieks, prancing gaits, talking, playing with an echo (a feature of the rooms some children discovered) — and fussing. The amount of observer agreement could not be calculated because of the absence of any recording of affect in more than a third of the records during the middle 10 minutes of the trial when both observers were making the narrative accounts. Two coders of the narrative accounts agreed in all instances, since the coding required only the counting of the main narrator's statements concerning expressions of affect. To compute a measure of pleasurable affect we tallied the number of units in which the accounts reported smiles, laughs, or happy sounding vocalizations. All children but one boy of 12 months and one girl of 18 months showed one or more of these behaviors. Two children showed such behavior during 27 of the units. The mean number of units was 6 (range, 0 to 23) for the younger children and 8 (range, 0 to 27) for the older children; the difference was not reliable. These tallies represent a minimum measure of the children's pleasure and delight for, in addition, 14 of the 18-month-olds exhibited exaggerated playful gaits at some point, and 15 children in all played with their echoes. Several of the older children talked to the mother almost contin193

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY uously, usually about the toys. With all these counts added to the amount of time spent contacting the toys — some of which may be thought of as play and to that extent a candidate for consideration as pleasurable affect — a picture of active pleasurable engagement with the environment emerges. Expressions of unpleasurable affect were shown by only 65 per cent of the children (by 14 of the 12-month-olds and 12 of the 18-month-olds) in contrast to the 95 per cent who gave expressions of pleasurable affect. Fussing occurred more frequently among, the younger children (F = 4.83, df = 1/32, p<.05); the mean number of units in which it occurred was 7.7 (range, 0 to 34) for the younger ones, and 2.3 (range, 0 to 13) for the older ones. Further analysis showed that 8 per cent of the fussing among the younger children were requests that they made of their mothers (to be picked up, to leave the room, etc.); such requests accounted for 36 per cent of the fussing among the older ones. THE TEMPORAL COURSE OF ACTIVITIES

Changes in the children's activities during the trial were analyzed by inspection of the means for the successive 5-minute (300-sec) periods of the trial. The frequency of transits between the three rooms, always greater among the older children, decreased steadily over time, although by the end of the trial some transits were still occurring. Similarly, the duration of time in the toy rooms, always greater for the younger children, also decreased over time, although by the last period the younger chil:dren were still spending a mean of 92 seconds there, the older children, a mean of 42 seconds. The amount of time spent locomoting showed a similar decrease over time. The duration of time spent contacting a toy or toys was maintained at a fairly steady rate by the older children; for the younger ones, after a slight increase in the second period, it declined gradually in subsequent periods (Fig. 2). Duration of contact of mother, in contrast, was low at the beginning for children of both ages — the children were then more often moving from room to room, spending time in the toy rooms, and contacting the toys; thereafter it increased for both groups, but at a faster rate for the younger children. As a corollary to the temporal course of these activities, Figure 2 shows that the older children were also increasingly contacting a toy in the mother's circle, an activity that was maintained at a steady but lower rate by the younger children. It should be re194

HARRIET L. RHEINGOLD

Figure 2. Temporal course of contact toy and contact mother, mutually exclusive measures (at left), and of contact toy and contact toy while in mother's circle (at right). At right, the second measure is a fraction of the first. Total duration of a period is 300 seconds.

called that contact of a toy in the mother's circle included some simultaneous contact of mother and toy, an activity that is treated in detail in the narrative accounts. The data in Figure 2 show that the contact of mother increased in the latter half of the trial for children of both ages, but at no sacrifice of contacting a toy among the older group, who, in contrast, were carrying out more of their contact of the toys in the mother's circle. Both groups, then, as the trial progressed spent increasing amounts of time in the mother's vicinity, the younger children contacting just the mother, the older ones more often contacting the toy there.

Figure 3. Temporal course of behavior in relation to both the toy and the mother, showing (at left) and sharing (at right). Total number of 10-second units in a period is 30.

195

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY The measure of pleasurable affect changed little over time but the measure of fussing increased for both ages, the increase being greater for the younger children. However, even for them, the mean frequency was only 4.0 units out of 30 in the last period. As Figure 2 shows, the younger children were by then more often in contact with the mother. Early in the trial, fussing was all but absent. The behavior in relation to both the mother and the toy showed interesting changes over time (Fig. 3). Showing a toy was more common at the beginning of the trial (as one might expect) and then declined, running almost identical courses in both age groups. Sharing, in contrast, increased over time and was always more frequent among the older children. Were the Children Enterprising? The findings show that the behavior of the children at both ages was "marked by an independent energetic spirit and by readiness to undertake and experiment," if we adjust the meaning of the term experiment to the age of the subjects. The general picture was one of active, vigorous, nondistressed children. They left their mothers, promptly entered new rooms, and contacted the toys directly. They made ample use of the different rooms and explored or played with the toys during a major portion of the trial; in contrast, they contacted the mother for a much smaller proportion of the time. As the trial progressed, older children shifted more of their activities with the toys into the mother's vicinity, whereas the younger children spent increasing amounts of time in contact with the mother without a toy. The children at both ages showed the toys to their mothers, smiling and vocalizing as they did so. The older children, however, more often brought toys to their mothers and engaged them in their activities with the toys. At both ages the children showed pleasure in the situation. Fussing was more characteristic of the younger children and tended to occur later in the trial, at which time they were more often in contact with the mother. On the whole, age differences were few, and girls and boys were equally enterprising.

Were the Children Independent? The children can indeed be characterized as independent because in all respects their behavior was self-guided, self-determined, and self-initiated. No behavior was forced by the situation. But, in another sense, that 196

HARRIET L. RHEINGOLD of the usual definitions of dependence and independence, can we characterize their behavior as independent — as independent rather than as dependent? In the first place, we recognize that much of the children's selfinitiated behavior depended on the mother's being present. Had she not been present or accessible, we know that in the unfamiliar environment of the laboratory the amount and variety of locomotion, exploratory behavior, and play would be diminished. To that extent the children's behavior was dependent. It was also dependent, of course, on prior experience with toys in general, with spatial arrangements, and with the stability or permanence of objects. To pursue the matter further, most definitions of dependence list as components "seeking proximity" and "seeking physical contact," as, for example, in the person-oriented dependence of Bandura and Walters (1963, p. 138). Certainly we saw both kinds of activity in the children. The 12-month-olds gave evidence of "seeking" more physical contact with the mother than the 18-month-olds, but only later in the trial; and the 18-month-olds showed more "proximity seeking," if play with a toy in the mother's circle can be so interpreted. Bear in mind, however, that the mother's circle — a circle with a radius of 1 meter — represents only an experimenter's arbitrary measure of proximity. The child's behavior with respect to the mother was not limited to maintaining proximity and physical contact. Two other significant classes of behavior were showing and sharing. In the first class, the child showed a toy to the mother — looking toward her as he reached it, pointing to it, or holding it aloft while looking toward her. This class of behavior was evenly distributed among children of the two ages. Looking toward the mother might conceivably be interpreted by some as the child's checking that she is still there. Previous work in this laboratory, however, has shown that the child knows his mother is still there, even when she walks out of his sight (Corter, Rheingold, & Eckerman, 1972). Recall that in the majority of instances the showing of a toy while looking at the mother was accompanied by smiles and vocalizations, often of a happy nature. The other related class of the child's behavior, sharing, was his bringing the toy to the mother, putting it in her lap or hand, and manipulating the toy in her lap as he leaned against her or sat in her lap, behavior more frequent among the older children but not absent in the younger. Shall we interpret these behaviors as indicating the seeking of attention, help, and recognition, praise, or approval, the other components us197

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY ually listed as indicators of dependence (Seller, 1955)? How can we know in the largely preverbal child? If we are reduced to reading intent into his behavior, then may we not consider an alternative reading? Both the showing and the sharing of the toys, as defined here, can qualify as the child's taking the initiative to include the mother in his discovery and manipulation of the toys, and in his pleasure. The child makes the mother a partner in his activities. In different contexts, these behaviors of showing and sharing have been remarked by others. Buhler (1930) wrote, "At an age of nine months the child shows objects to the grownup. . . . In the grasping and offering of objects there lies the beginning of an organized, social play activity with the grownups" (p. 59). Valentine (1950) reports instances of the child at one year giving things to people; and he calls that social play — handing them objects and taking them back. Kohlberg (1969) states, "the child in the second year of life clearly takes a delight in sharing through imitation, reciprocal play and communication (for example, pointing things out to the other)." He continues, "It is this type of experience, rather than clinging, which clearly indicates that other people are people to the infant, not security blankets or 'cloth mothers' to be clung to in unfamiliar situations" (p. 468). Bronson (1971) wrote of a 12-month-old girl who is playing, "She shares her delight with her mother" (p. 269). And Wenar (1972) spoke of the child's taking the initiative in making exploratory activities a social rather than a solitary pursuit. The child's taking the initiative in sharing with the mother his discoveries and play, as well as his pleasure therein, I now suggest, should be viewed as not just dealing with the environment, not just social play, but instead as evidence of his independence. It is not help the child is requesting, or yet approval; instead, he takes the initiative in making the other person a partner in his play. He is the giver; the mother, the receiver. The older children show more of this behavior — probably for many reasons. They are more able, and they have had more experience with people's bringing things to them and with people's responding to their overtures — by a smile or a word.

The Mother's Effect on the Child's Behavior To what can the wide individual differences among the children at each age be attributed? Not to sex, because sex proved not to have a reliable effect on any of our measures. Order of birth and developmental maturity 198

HARRIET L. RHEINGOLD also come readily to mind. But I wish to say a few words about the effect of the mother's behavior as a possible antecedent variable. In this study, home visits to measure the mother-child relationship were not made, and the mother was not questioned in the laboratory about her attitudes or behavior. But in the moment-by-moment narrative account of the child's activity, it was inevitable that the observer recorded the mother's behavior as well. Even though by design the mother's role was considerably restricted, marked differences in behavior were apparent. Some mothers responded very little to the child's overtures to her. Were they following instructions conscientiously or were they behaving as they customarily do at home? On the other hand, a few mothers made every bid imaginable for their children's attention, smiling broadly when their children smiled, and turning their heads to maintain eye contact with them. These mothers also behaved in ways that kept their children close, stroking their heads, patting backs, adjusting clothes, and holding the children tightly when they climbed into their laps. Were they trying to thwart the experimenter, or was this their customary behavior? It might be thought that the restrictions placed on the mother's responsivity to the child in the experimental situation would increase the child's dependent behavior; at least, the restrictions might increase his contact with her. Quite the opposite was the general impression of the observers: particularly among the older children, those who fussed and hung on their mothers the most had mothers who responded the most. It would be tempting indeed to search for relations between how the mother behaved in response to the child's behavior and how he behaved in response to hers. Still, we should resist the temptation, if only because some part of the mother's behavior is shaped by the child's behavior — at the moment of this study and in all the moments of his lifetime — as his is by hers. We tend often to ignore the fact that in relating the child's behavior to the parent-child, or even to the child-parent, relationship, the child is an element on both sides of the equation. One therefore cannot miss finding a relationship.

Other Components of Independence The present study of self-initiated behavior by children in a free-play situation containing their mothers was intended to elucidate only one component of independent behavior in year-old children. It is assumed that there are other components and that investigation of them will yield 199

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY additional evidence to support the proposition that a large portion of young children's behavior can qualify as independent. Mothers often report an apparently sudden change in a heretofore docile child, occurring around 16, 17, or 18 months. The child will not have his hand held when walking; he resists aid in going upstairs; he will not be held for more than a few seconds even though he asked to be picked up; he protests when put to bed. He will not try new foods and will not eat if his bib is put on wrong side to. He insists on carrying a toy when going outdoors or being put to bed; he demands a partner when "reading" a book. He also begins to do forbidden acts — he throws stones or spills his milk, and he grabs toys from other children. At the same time he spends more time playing by himself; he is more active; and physical contact with his parents decreases. Storms and tantrums make their first appearance. In a few days or a couple of weeks the resistance disappears, only to reappear. Classically, this behavior is called negativism or willfulness. The parents are bewildered because they are unprepared. We, not faced by the actual behavior but seeing it from Olympian heights, can applaud the child's behavior and view the parent as insensitive; the parent and not the child as willful and negativistic. How can we document the behaviors of interest? Parental report is always a possibility but a flawed one. Observation of the child's behavior in the home offers an alternative. A representative sample of his behavior may also be obtained under the more easily controlled conditions of the laboratory; the average child has so wide an experience with diverse environments in the ordinary course of events that the benign setting of a well-planned laboratory cannot qualify as even unusual. It is possible, however, that any situation we set up in the laboratory may be so interesting that negativism will not occur; such behavior may need accustomed settings. Other candidates for study remain, such as the child's setting difficult tasks for himself. A more discerning recording and analysis of the children's activities in the present study might have provided such evidence. Originally we had hoped to observe this behavior by having the children return to the laboratory the next day, reasoning that as they became more familiar with the setting and with the toys, they would settle down to carry out more complex combinations of the toys. The results of a pilot study of several children, however, showed that the types of play engaged in on 200

HARRIET L. RHEINGOLD the second day were very similar to those of the first day but — more damaging to the hypothesis — the children ran through everything faster, fussed sooner, and the trials had to be ended earlier. Perhaps after all, as the setting became more familiar, it was producing a kind of negativism. Another candidate for study is the child's efforts to overcome obstacles such as might occur during extinction of an instrumentally conditioned response. Still another would be the effect of training on the difficulty of the task a child chooses. We could then begin to speak of achievement striving and level of aspiration, and relate these behaviors to similar behavior in older children.

Conclusions In this paper I have called attention to the self-initiated social and nonsocial behavior of the young child. I have attempted to clarify the metaphor independence and to show that self-initiated behavior — independence — finds a place in the timetable of development earlier than adolescence. In part, I wish to correct the emphasis placed on dependence to the exclusion of independence, but only to correct the imbalance, not to rule out dependence. The infant, as any older organism, is both independent and dependent; both constructions can be placed on almost any activity. Dependence and independence appear together, and increase side by side. To be sure, the kinds of dependent and independent activities change with experience. Speculation about the origins, acquisition, and developmental history of independent behavior have been omitted in this account. I do not doubt that both dependent and independent behaviors are modified by experience. I saw enough in the short 20-minute periods of observation to convince me that the behavior of some mothers works to keep their children at their sides and interferes with the children's play. Still, my purpose here is simpler and more single-minded. Similarly, I have refrained from placing a value on independence; there is value also in dependence. Up to this point I have advanced the argument within the framework provided by the terms dependence and independence. I have run the risk foretold by George Eliot who said not only, as I quoted earlier, that, "All of us get our thoughts entangled in metaphors . . ." but who concluded the sentence with the words, "and act fatally on the strength of them." Thus, I may have fallen into a trap of my own making. To save myself from so grievous an outcome, I claim only the purpose of directing atten201

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY tion to an important characteristic of the infant's behavior; at the same time I call into question the usefulness of the terms and disclaim any wish to perpetuate them. It is the behavior I call attention to, not the terms, for terms like dependence and independence are only verbal substantives for the nonsubstantial. The infant and young organism, human or nonhuman, behave as they always have; what changes is only how psychologists view their behavior. Today we regard the infant as more responsive, more able, and more socially active than formerly. We have also begun now to accord him a position of strength in what we loosely designate the parent-child relationship. I propose now that we go one step further and see him as the enterprising and independent creature he is. Dependent, yes — as dependent as we all are — but also independent. References Ainsworth, M. D. S. Object relations, dependency, and attachment: A theoretical review of the infant-mother relationship. Child Development, 1969, 40, 969-1025. Bandura, A., & Walters, R. H. Social learning and personality development. New York: Holt, 1963. Beller, E. K. Dependency and independence in young children. Journal of Genetic Psychology, 1955, 87, 25-35. Bronson, W. C. The growth of competence: Issues of conceptualization and measurement. In H. R. Schaffer (Ed.), The origins of human social relations. New York: Academic Press, 1971. Biihler, C. The first year of life. Trans. P. Greenberg & R. Ripin. New York: Day, 1930. Corter, C. M., Rheingold, H. L., & Eckerman, C. O. Toys delay the infant's following of his mother. Developmental Psychology, 1972, 6, 138-145. Cramer, E. M., & Bock, R. D. Multivariate analysis. Review of Educational Research, 1966,36,604-617. Erikson, E. H. Childhood and society. New York: Norton, 1950. Gewirtz, J. L. Mechanisms of social learning: Some roles of stimulation and behavior in early human development. In D. A. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally, 1969. Goldberg, S., & Lewis, M. Play behavior in the year-old infant: Early sex differences. Child Development, 1969, 40, 21-31. Hartmann, H. Ego psychology and the problem of adaptation. In D. Rapaport (Ed.), Organization and pathology of thought. New York: Columbia University Press, 1951. Hartup, W. W. Dependence and independence. In H. W. Stevenson (Ed.), Child psychology: The sixty-second yearbook of the National Society for the Study of Education. Pt. 1. Chicago: University of Chicago Press, 1963. Heathers, G. Emotional dependence and independence in nursery school play. Journal of Genetic Psychology, 1955, 87, 37-57. (a) . Acquiring dependence and independence: A theoretical orientation. Journal of Genetic Psychology, 1955, 87, 277-291. (b)

202

H A R R I E T L. R H E I N G O L D Hendrick, I. Work and the pleasure principle. Psychoanalytic Quarterly, 1942, 11, 33-58. Hetherington, E. M. Sex typing, dependency, and aggression. In T. D. Spencer and N. Kass (Eds.), Perspectives in child psychology: Research and review. New York: McGraw-Hill, 1970. Kohlberg, L. Stage and sequence: The cognitive-developmental approach to socialization. In D. A. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally, 1969. Levy, D. M. Observational psychiatry: The early development of independent and oppositional behavior. In R. R. Grinker (Ed.), Mid-century psychiatry: An overview. Springfield, 111.: Thomas, 1953. Maccoby, E. E., & Masters, J. C. Attachment and dependency. In P. H. Mussen (Ed.), Carmichael's manual of child psychology. Vol. 2. (3rd ed.) New York: Wiley, 1970. Mahler, M. S. On human symbiosis and the vicissitudes of individuation. Vol. 1. Infantile psychosis. New York: International Universities Press, 1968. Messer, S. B., & Lewis, M. Social class and sex differences in the attachment and play behavior of the year-old infant. Merrill-Palmer Quarterly, 1972, 8, 295-306. Mittelmann, B. Motility in infants, children, and adults: Patterning and psychodynamics. In R. S. Eissler et al. (Eds.), The psychoanalytic study of the child. Vol. 9. New York: International Universities Press, 1954. Mussen, P. H., Conger, J. J., & Kagan, J. Child development and personality. (3rd ed.) New York: Harper, 1969. Parke, R. D. (Ed.) Readings in social development. New York: Holt, 1969. Rheingold, H. L. The measurement of maternal care. Child Development, 1960, 31, 565-575. The effect of a strange environment on the behavior of infants. In B. M. Foss (Ed.), Determinants of infant behaviour IV. London: Methuen, 1969. (a) The social and socializing infant. In D. A. Goslin (Ed.), Handbook of socialization theory and research. Chicago: Rand McNally, 1969. (b) & Eckerman, C. O. The infant's free entry into a new environment. Journal of Experimental Child Psychology, 1969, 8, 271-283. The infant separates himself from his mother. Science, 1970, 168, 78-83. Rheingold, H. L., & Samuels, H. R. Maintaining the positive behavior of infants by increased stimulation. Developmental Psychology, 1969, 1, 520-527. Ross, H. S., Rheingold, H. L., & Eckerman, C. O. Approach and exploration of a novel alternative by 12-month-old infants. Journal of Experimental Child Psychology, 1972, 13,85-93. Stone, L. J., & Church, J. Childhood and adolescence: A psychology of the growing person. (2nd ed.) New York: Random House, 1968. Valentine, C. W. The psychology of early childhood: A study of mental development in the first years of life. (4th ed.) London: Methuen, 1950. Watson, R. I. Psychology of the child. (2nd ed.) New York: Wiley, 1965. Webster's seventh new collegiate dictionary. (7th ed.) Springfield, Mass.: G. & C. Merriam, 1969. Wenar, C. Personality development: From infancy to adulthood. Boston: Houghton, 1971. Executive competence and spontaneous social behavior in one-year-olds. Child Development, 1972, 43, 256-260. White, R. W. (Ed.) The early growth of competence. In The enterprise of living: Growth and organization in personality. New York: Holt, 1972.

203

This page intentionally left blank

LIST OF CONTRIBUTORS

This page intentionally left blank

List of Contributors

DIANA BAUMRIND is research psychologist at the University of California at Berkeley where she has been since receiving her Ph.D. there in 1955. She has studied the effects of group structure on group productivity in addition to her present research concerned with socialization effects on cognitive, emotional, and social development for children of different ages. Her interest in moral philosophy is expressed both in her current work relating ethical judgments in parents and children and in her published critiques concerning ethical standards in the treatment of human subjects in psychological research. She is also a clinical psychologist applying the findings from her research on parent-child relations to the problems faced by parents and teachers bringing up children. C. G. BEER comes from Waipukurau, New Zealand. Following his undergraduate training in New Zealand, a Rhodes Scholarship took him to Oxford, where he studied animal behavior under Niko Tinbergen. He received his D.Phil, in 1960 and then took a postdoctoral fellowship at Rutgers University which enabled him to work with the late D. S. Lehrman. He spent two years teaching zoology at his old university in New Zealand and then joined the faculty at Rutgers, where he has been ever since except for an appointment which he held for a year at Oxford. He is now professor of psychology at Rutgers, and joint editor of Animal Behaviour Monographs. His main research activity has been field study of the behavior of gulls, but his interests trespass into many other subjects. NORMA FESHBACH is currently an associate professor in the Department of Education at UCLA and director of the Training Program in Early 207

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Childhood Development. She received her Ph.D. in Psychology in 1956 from the University of Pennsylvania and has held a number of teaching and research positions since then. Her research interests have been varied, including investigations of children's aggression, the development of empathy and social comprehension, and .the interaction of the socializing influences of home and school. JOHN W. HAGEN received his Ph.D. degree from Stanford University in 1965. Since that time, he has been at the University of Michigan, where he is currently professor of psychology and chairman of the Developmental Program. His research interests include the development of attention and the effects of verbal processes on memory. He has studied cognitive development in deviant and retarded as well as in normal children. He has served as a consultant to numerous projects and on the Commission on the Age of Majority, State of Michigan. GORDON A. HALE obtained his Ph.D. in 1969 at the Institute of Child Development, University of Minnesota. He is presently a research psychologist at the Educational Testing Service. His research is concerned with various aspects of attention and memory in children. DAVID KLAHR is an associate professor of psychology at Carnegie-Mellon University. After he received his Ph.D. from Carnegie Institute of Technology in 1968, he held an appointment at the University of Chicago. In 1968-1969, he was a research fellow and Fulbright Lecturer in the British Isles. During that period he began his collaboration with J. G. Wallace at the University of Stirling, Scotland. The collaboration has emphasized the application of information-processing analytical and simulation techniques to the study of cognitive development. Since 1969, in addition to continuing his work in cognitive development, he has been director of the Educational Research and Development Unit at Carnegie-Mellon University. HARRIET L. RHEINGOLD, a research professor in the Department of Psychology of the University of North Carolina at Chapel Hill, holds her doctorate degree from the University of Chicago. Previously, she was a research psychologist at the National Institute of Mental Health in Bethesda, Maryland. Earlier in her career she taught at Rockford College and worked in the Illinois Institute for Juvenile Research. Her research interests fall under the broad heading of the development of social and exploratory behavior. 208

INDEX

This page intentionally left blank

Index

Achievement. See also Cognitive performance defined, 4 and parent authority patterns, 15, 19, 20,21,22,27, 34-35 and socioethnic differences, 99, 102, 107 Affective behavior ages differences in, 193-194 in infancy, 193-194, 196 Aggression: and punishment, 36 Attachment: and obedience, 31-32 Attention age differences in, 118-119, 120, 122, 123-124, 127, 129, 130, 131, 136-138 and central task learning, 118-119, 120, 122, 127, 129, 133 cultural differences in, 135-136 effects of task definition on, 130-131 environmental effects, 135, 136 filtering concept of, 117 and incidental stimuli, 117-118 and information overload, 117, 118-119 and information processing, 126-127, 137-138 reinforcement effects on, 135 in retarded children, 134-135 stimulus effects on, 123-127 as a task strategy, 133, 137 training effects on, 127-129, 135 and type of task, 129 and verbal rehearsal, 132 Classification: and information processing, 146-148

Cognitive development and class inclusion, 169-170, 173 and conservation, 171-172 information-processing models of, 146152, 154-160, 162, 170, 172, 173-174 and language, 89-90 and parent-training programs, 113 problems in research, 145 role of reinforcement in, 99, 112 and socialization, 112-114 theory and information processing, 141144 Cognitive performance and reinforcement style, 99, 107, 108112 and social class, 89-90, 102 Communication, in birds age differences in response to, 66-68 chick response to adult, 58-61, 66-68, 69, 76-78 during hatching, 65-66 function of adult calls in, 68 and language, 82-83 meaning of adult calls in, 74, 76-78 situational context and, 51-52, 81-82 types of vocal signals in: alarm, 57; crooning, 57; ke-hah, 59; long-call, 59 Communication, in humans parent-child, 7 and parent control, 7, 8 Competence, instrumental: defined, 4. See also Achievement; Independence; Social Responsibility

211

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Conservation development of, 171-172 and information-processing models, 172 Cross-cultural comparison of attention, 136 of central/incidental learning, 135-136 of reinforcement styles, 104-108 of socioethnic differences, 99-103, 106108 of task performance, 105-106 Dependence correlation with independence, 181 as a drive, 180-181 maternal influence on, 197 Discipline and independence, 25, 26 and parental authority patterns, 41 in parent-child interactions, 6, 25 and social responsibility, 25-26 Educational intervention programs: parental role in, 113 Ethnicity and imitation, 91 in Israel, 99 and reinforcement techniques, 94-98 and task performance, 97-98 Ethology limits of description in, 53-55 and observation, 49-50, 80-81 relevance to human behavior study, 47— 48,84 roots of, 50 statistical analysis in, 50-51, 80, 81 and study of the individual, 79-80 Exploration age differences in, 189 in infancy, 182, 183, 189 Frustration and discipline, 40-41 and parental authority patterns, 41 Imitation effects of reinforcement on, 90-92 and social class, 91 Independence and affect, 185 age changes in, 185 and autonomy, 180 correlation with dependence, 181 defined, 4 and dominance, 40 212

as a drive, 180-181 effects of discipline on, 40 effects of firm control on, 25 effects of maturity demands on, 7, 25 effects of parental authority patterns on, 19-22,27,29,30,31,41-42 effects of permissiveness on, 9, 31,34-35 effects of restrictiveness on, 26, 38-39 and exploration, 182 in infants, 196-198 and initiative, 181-182 maternal influence on, 197 measures of, 187-193 and obedience, 33 and overcoming obstacles, 201 problem of definition, 178-179 and resistance, 40, 182, 220 sex differences in, 41 and sharing behavior, 197-198 and showing behavior, 197-198 and socialization, 179-180 and task setting, 200-201 Infancy affect in, 193-194, 196 autonomy in, 180 contact with mother in, 190-191, 194195 drives in, 180-181 effects of environment on, 183 effects of institutions on, 182 enterprising behavior in, 184—185, 196 exploration in, 182, 183, 189 independence in, 196-198 initiative in, 181-182 leaving mother in, 183-184 maternal influences in, 198-199 resistance in, 182 sharing behavior in, 192-193, 197-198 showing behavior in, 192-193, 197-198 socialization in, 179—180 toy contact in, 190, 191-192, 194 use of space in, 189-190, 194 Information processing and attention, 126-127, 137-138 and class inclusion, 159-160, 170, 173 and classification, 146-148 criteria for a model of cognitive behavior, 143 and incidental learning, 124-125, 126127, 137 levels for explanation of cognitive development, 143-144 and Piagetian stage theory, 145 production systems in, 151-160

INDEX Maturity: parental demands for, 7, 8, 21, 22,24-25 Memory: and information processing models, 150-158, 161-162, 163, 164, 165167,170

quantification operators in, 160-164 and quantitative comparison, 154-160, 164-165 and short-term memory capacity, 165167 and task failure, 150 task-performance model in, 149-150 verbal encoding in, 167-170 Intelligence test scores and central task learning, 133 and incidental learning, 133-134

Nurturance involvement and, 7 warmth and, 7

Language and achievement, 102 and cognitive development, 89-90 and communication in birds, 82-83 of English children, 89-90, 102 and social class, 89-90, 94-95, 102 Laughing Gull individual recognition in, 56ff. reproductive behavior of, 55 social development of, 55-56 Learning, central task age differences in, 118-119, 120-122, 125, 126, 127, 132, 133-134, 136-138 cultural differences in, 135-136 effects of verbal rehearsal on, 132 and information overload, 118-119,120 and intelligence test scores, 133-134 in retarded children, 134 and selective attention, 118, 120, 122, 129, 133 stimulus effects on, 123-127 training effects on, 127-129 Learning, incidental age differences in, 118-119, 120, 122, 123-124, 126, 127, 129, 133-134, 136-138 cultural differences in, 135-136 effects of task definition on, 130-131 and information overload, 118-119,120 and information processing, 124-125, 126-127, 137 and intelligence test scores, 133-134 in retarded children, 134-135 stimuli effects on, 123-127 training effects on, 127, 129 Maternal influences and dependence, 197 on independence, 197 on infant behavior, 198-199 on infant exploration, 183

Obedience and attachment, 31-32 developmental implications of, 32—33 effects of punishment, 32 function of, 31-32 and independence, 33 and reinforcement, 32 Parental authority patterns and achievement orientation, 15, 19 behavior clusters in, 10-13 and communication, 8, 15 and conceptions of child, 42-43 and discipline, 41 and dominance, 15, 19, 20, 21 effects on independence, 15, 19, 31, 4143 effects on social responsibility, 30-31 and friendliness, 20 and hostility, 15,20 and punishment, 8, 9, 19, 35 and purposiveness, 15, 19, 20, 21 and reinforcement, 8 resistance to, 15, 20, 29 types of Authoritarian, 8,13,14-19,21-22, 30 Authoritative, 8, 13-14, 19-20, 2122,30 Black Authoritarian, 28-30, 41 Harmonious, 27-28 Nonconforming, 13, 14, 20, 21-22 Permissive, 8-9, 14, 15, 19, 20, 2122,31 Parent influences and communication, 7, 8 and control, 6, 9, 25, 26, 31, 38-40 and discipline, 6, 25 on independence, 9, 25, 31, 38-39 in late childhood and adolescence, 43-44 and maturity demands, 7, 24 and nurturance, 7-8, 25, 33-34 and punishment, 25-26 use of reinforcement, 8-9, 32

213

MINNESOTA SYMPOSIA ON CHILD PSYCHOLOGY Piaget's concept of stage and class inclusion, 159-160 problems of definition, 145 Primate behavior: and leaving mother, 184 Problem-solving information-processing model of, 149150 production systems model of, 151-160 Punishment and aggression, 36 effects on child behavior, 25-26, 32, 36-38 effects on independence, 26 effects on obedience, 32 and parental authority patterns, 8, 9, 19,35 Reading and maternal instructional behavior, 111 and reinforcement style, 108-112 Recognition, Laughing Gull of adult calls, 57, 58-61 effects of age on, 64-68, 70-74 effects of crooning calls on, 70-71, 7374 effects of hand-raising on, 62-64 effects of long-calls on, 70-73, 74-76 effects of social context on, 69 of parent versus other, 61-62 Reinforcement children's style of, 92-95, 96-97, 98, 100-101, 103 cross-cultural comparisons of, 99-108 effects on imitation, 90-92 effects on learning, 88, 91-92, 107 effects of pupil performance on, 94, 97, 103 effects on task performance, 135 and ethnicity, 94-98 and language, 94-95, 98 maternal style of, 95, 97-98, 101, 103, 109

parental use of, 8-9, 32 and reading ability, 108-112 role in cognitive development, 99, 112 role in cognitive performance, 107,108112 and social class, 90, 91-92, 94-98, 99102, 103, 107-108 in social interactions, 88 and socialization, 89, 95, 98-99, 107, 112-113 Retarded children effects of environment on, 134-135 effects of reinforcement on, 135 effects of training on, 135 and selective attention, 134-135 Sharing age differences in, 192-193 in infancy, 192-193 Social class and achievement, 99, 102, 107 and cognitive performance, 89-90, 102 and imitation, 91 and language, 89-90, 94-95,99 and reinforcement style, 90, 91-92, 9498, 99-102, 102-103, 104-108 and socialization practices, 89, 111 Social responsibility defined, 4 effects of restrictiveness on, 26, 38 and parental authority patterns, 21, 27, 30-31 Socialization. See also Independence agents of, 3 and cognitive development, 112-114 and education, 88-89 and obedience, 31-33 parent's role in, 88-89 process of, 4 and reinforcement style, 89, 95, 98-99, 107, 112-113 and social class, 89, 111 teacher's role in, 88-89

214