Constraints on peer socialization: let me count the ways.

Children spend significant amounts of time with other children and, in so doing, have extensive opportunities to influence one another. The same situation exists for adolescents, suggesting that peer relations contribute substantially to socialization from early childhood through the second decade and beyond. Specifying the circumstances under which exposure to other children occurs and documenting the consequences have been major concerns within developmental psychology for decades, and there is little doubt that many developmental outcomes are traceable to these encounters.

Although most investigators believe that children's lives are affected by their encounters with other children--both in the short term and the long term--it is not easy to describe these processes. The socialization outcomes of child-child interaction are constrained (moderated) by numerous subject and situational conditions, that is, the characteristics of the children involved and the settings in which their interaction occurs. Whether children influence one another through modeling, talk, and social reinforcement is no longer in doubt. Critical issues, however, concern the manner in which subject and situational conditions interact with social contingencies in determining outcome.

The contributors to this special issue each address one or more of these interaction effects. Some examine child characteristics (e.g., personality factors) whereas others examine relationship conditions existing between the child being socialized and the child socializer (e.g., friendship stability). Results demonstrate convincingly that interaction effects involving subject and situation must be included in any developmental analysis bearing on peer socialization. On the basis of these studies and earlier ones, one can argue that a comprehensive theory of peer socialization requires attention to the following constraints:.

First, some children exert greater influence over their contemporaries than others. Characteristics of the child socializer are among the most important moderators of peer socialization. Characteristics of the relationship between the child socializer and the child being socialized must also be taken into account.

Second, the conditions associated with behavior change (e.g., contingencies of coercion, conflict, or reward) affect different children differently. Differences among children in susceptibility to social influence or conformity proneness have been recognized for many years and current evidence suggests that these constraints remain relevant.

Third, both cognitive and affective development constrain children's influences on one another. Much remains unknown about developmental constraints on learning and cognition, but most investigators do not assume that attention, memory, and other cognitive processes work in exactly the same way throughout the child and adolescent years. Children and adolescents also make different attributions to themselves and others on the basis of age, and these attributions figure prominently in social comparisons and other social experiences (France-Kaatrude & Smith, 1985).

Fourth, what paradigms best account for behavior change through peer socialization? One may determine how outcome variance is apportioned among the child's characteristics, the child's selection of associates, and the child's social experiences with these same children, but it is another matter to ask how behavior change comes about.

Fifth, we should be able to say whether the processes responsible for the transmission of social norms between children are similar across behavioral domains or whether they differ. Some investigators are convinced that child and adolescent development are domain specific rather than widely generalized (Bugental & Goodnow, 1997; Gelman & Williams, 1997).

Each of these constraints is discussed briefly in this commentary.

CHARACTERISTICS OF THE INFLUENCE SOURCESocial interaction between two children reflects characteristics of both children; both are being socialized simultaneously. Consequently, the variance deriving from characteristics of the socializer is usually conflated with variance deriving from characteristics of the child being socialized. Older tutors, for example, may be generally more effective in instigating behavior change than younger ones, but a more accurate statement is that children who are older than their tutees are generally more effective than tutors who are younger than theirs (Ludeke & Hartup, 1983). Similarly, the effects of having a delinquent friend are different for children who are or are not troublesome themselves (Poulin, Dishion, & Haas, this issue).

Characteristics of child socializers needing investigation include certain basic ones--gender, chronological age, and ethnicity. Although each has been studied extensively (see Rubin, Bukowski, & Parker, 1997), the manner in which these characteristics moderate child-child socialization is not well understood. Gender, for example, usually constrains conformity according to the normative activity in question: Girls exert greater influence on girls than on boys when feminine norms are involved; the reverse occurs when masculine norms are salient (Hartup, 1983). Most investigators contrast girl-girl with boy-boy socialization, and surprisingly little is known about gender of the socializer uncon-founded with gender of the child being socialized. One needs to be aware of these conditions when considering socialization models that are based on one sex only (Poulin et al.) or that show gender not to account for much outcome variance (Berndt, Hawkins, & Jiao; Schulenberg et al.; Pilgrim, Luo, & Urberg; Rose, Chassin, Preston, & Sherman; all this issue).

Other understudied characteristics (both in these studies and the literature generally) are the social competence and social status of the socializer, especially relative to the child being socialized. The socializer's expertise (e.g., whether this child smokes or not, uses drugs or not) has been studied on numerous occasions. One wishes, however, that other characteristics of socializers were as commonly entered into statistical modeling exercises as is the expertise variable in studies of social deviance (see Schulenberg et al.).

One needs to know something, too, about relationships existing between socializers and the socialized. The articles in this issue reflect a growing awareness that close relationships (especially friendships) must be included in causal modeling--whether the criterion behavior is deviant or not. Relationship dimensions associated with peer influence are not the same across behavioral domains. Friendship quality, for example, seems to have something to do with socialization into delinquency but these effects depend on base rates and other subject characteristics (Poulin et al.). Simultaneously, friendship quality seems to have relatively little to do with school adjustment across the transition to junior high school; stability of the child's friendships is more relevant to outcome (Berndt et al.).

WHAT CHARACTERISTICS OF THE CHILDREN BEING SOCIALIZED ARE IMPORTANT?These studies make clear that behavioral base rates related to the criterion variable must be entered into regression models as well as interactions involving them. Base-rate information is necessary, first, to properly apportion outcome variance between pre-existing child attributes and the socializing events that occur between Time 1 and Time 2. Second, base rates constrain the interaction that occurs between two children: Delinquency base rates, for example, at ages 13 and 14 are positively correlated with the duration of rule-breaking talk observed between subjects and their friends, which in turn contributes significant variance to delinquency rates at ages 15 and 16 (Poulin et al.). Other data are consistent with this notion: Vitaro, Tremblay, Kerr, Pagani, and Bukowski (1997), for example, found that socially disruptive friends bring about increases in delinquency base rates among children who are moderately disruptive themselves, but not among children who are either highly disruptive or never disruptive. Apparently, highly disruptive children are so deep into antisocial behavior that friends make little difference to outcome whereas nondisruptive children are so well protected that the behavior of their friends is irrelevant to their future actions.

Child assessments assist in distinguishing between direct and indirect influences in child-child socialization: Pilgrim et al., for example, report that sensation seeking--a characteristic correlated with substance abuse in certain adult studies--is related to adolescent drug use both directly and indirectly. Specifically, a direct negative impact is detectable from Time 1 to Time 2 as well as as an indirect effect that is mediated by Time 1 drug use. Without the base rate and personality assessments in this investigation, relatively little new information would have been available to report.(FN1).

Sometimes, child characteristics are included in longitudinal studies on the basis of very sketchy rationales. Clearly, studies of social deviance must include base rate information about troublesomeness and disruptiveness in order to tap those childhood precursors which may be most relevant to the prediction of deviant outcomes (see Poulin et al.) But which other characteristics should be measured? Sensation seeking? Susceptibility? Conformity?

There was a time was when interest in conformity and related constructs was great (see Hartup, 1983). Only one study included in this collection draws inspiration from this body of literature--an effort to identify the antecedents of adolescent drinking (Schulenberg et al.). Results show that susceptibility increases overindulgence both directly and indirectly through increasing exposure to alcohol-using peers. Four indicators were used to measure susceptibility in this instance: succumbing to peer pressure to destroy property, skipping school, avoiding studying for a test, and smoking. What is needed now is an intensive effort to refract this susceptibility construct and describe its role in the transactions through which one child's behavior changes another's. Are children who are susceptible to antisocial norms especially attentive to deviant models? Especially responsive to attention from deviant children? Especially likely to encode certain kinds of deviant information and retrieve it more readily in "appropriate" situations?

One can write much more about child characteristics as constraints on peer socialization. Many stable characteristics among children, ranging from temperament to socialization histories, may be relevant. Given the vast array of seemingly relevant child characteristics, a major effort must be made to draw together a more integrated account of these constraints and subject this account to empirical test.

AGE CONSTRAINTS ON PEER INFLUENCESDevelopmental status bears on socialization outcome in at least three major ways: (a) Chronological age is a social category used by children to construct both self- and other-attributions. As mentioned earlier, these attributions--both in absolute and relative terms--are closely related to socialization outcomes. One of my most vivid memories concerns two unacquainted 4-year olds in an experimental session who spent 10 minutes one day arguing about who was the older before they turned to the play materials (they even took off their shoes and compared the size of their feet). (b) Chronological age is related to children's gradually-changing theories of mind, social relationships, and social action. Peer socialization may have different outcomes depending, for example, on whether children regard friendships as one-way social contexts providing them with desired rewards, as preschool children do, or whether they regard these relationships as committed, mutual entities, as adolescents do (Selman, 1980). (c) Chronological age constrains both cognitive processing and its social applications in that young children are not the same kinds of cognizers or learners as are older ones (Flavell & Miller, 1997).

Older children and adolescents are better change agents than are younger ones: Not only are older socializers more sensitive to instances in which cognitive structuring is needed by the children being socialized, base rates of structuring and feedback are higher than among younger socializers (Ludeke & Hartup, 1983). Developmental changes underlying these transformations in socializing effectiveness are numerous (Flavell & Miller, 1997). One example: Tutors who are good role-takers are more likely than poor role-takers to respond to indirect indications that the tutee needs assistance but not to direct requests (Hudson, Forman, & Brion-Meisels, 1982).

Among children being socialized, older children engage in discussions about social issues in ways that differ from younger children and respond to social pressure differently (Hartup, 1983). Once again, however, developmental status interacts with situational characteristics in determining social outcomes. Berndt (1979) discovered, for example, that conformity does not change much across the transition to adolescence when peer pressure involves socially sanctioned norms but increases through early adolescence when conformity to anti-social norms is involved. Relatively little is actually known about age-related changes in cognition and age-related changes in children's responsiveness to one another in everyday socialization. But, as scientific thinking about thinking improves, our understanding of socialization pressures and their consequences should improve, too. Consensus remains that cognitive development is not completed during childhood; numerous refinements in reasoning occur afterward (Moshman, 1997). Social interaction may contribute to these refinements; reflection may contribute, too. Such refinements, in turn, may change the nature of the interaction between child socializers and the individuals they socialize.

Significantly, chronological age or age-related variables are absent in the causal models contained in this issue. One may argue that age was not relevant to these studies. Some contributors to this issue have failed to find age effects on influence in adolescents (Chassin, Presson, Sherman, Montello, & McGrew, 1986; Urberg, Degirmencioglu, Tolson, & Halliday-Scher, 1997). Given what is known about the origins of deviant behavior, however, one can register at least mild surprise in discovering that age-related considerations are missing in this collection of studies.

WHAT PARADIGMS ARE BEST?The construct most commonly used to describe the processes by which social experience brings about behavior change is internalization. That is, social interaction is conceived of as a channel through which something "external" (e.g., social norms) becomes "internal." The knowledge structures internalized are conceived in many different ways: These may be schemas, scripts, or representations (Mandler, 1997) and may relate to single or multiple events, including social relationships (Baldwin, 1992). Relationship schemas, however, are rather vaguely defined in the literature, including such well-known ones as the "internal working model" (Bowlby, 1969).

In many instances, it is easier to describe these internalized schemas than to specify how they are acquired. Cognitive psychology is still coming to grips with the "explanatory problem." Should we describe learning and knowledge acquisition in terms of stimulus-response connections, perceived similarities, comparisons, evaluations, or equilibrations? No one knows whether or not acquisition is the same for all knowledge structures in all social contexts.

Socialization seems to many investigators to be quite different in reciprocal situations involving two or more children as compared to hierarchical situations such as those involving parents and children (Bugental & Goodnow, 1997; Piaget, 1932; Youniss, 1980). Reciprocal relationships center on the exchange of matched outcomes, leading some investigators (Barker & Wright, 1955; Whiting & Whiting, 1975) to regard sociability (exchanges among equals) as the central norm in child-child relations and power relations as central in hierarchical ones. Certainly, interactive "content" differs between hierarchical and reciprocal relationships, with the former being more likely to involve caretaking (nurturance and instruction) and the latter more likely to involve sociability and play.

Choosing among paradigms to use in studying peer socialization is not so much the task of choosing a social learning perspective, a biological perspective, an information processing perspective, or a cognitive-developmental perspective as choosing constructs and notions that "fit" the socialization domain with which one is concerned. The egalitarian nature of the peer culture, as well as the reciprocities that characterize the interactions children have with one another, suggest that some ideas may be more relevant than others. Piaget (1932) argued that a dialectic between conflict and cooperation is salient in reciprocal social relations but not in hierarchical ones. Similarly, Laursen and colleagues (Laursen, Hartup, & Koplas, 1996) argue that theories of exchange make good sense when applied to reciprocal socialization but may be less effective when applied to hierarchical socialization.

Peer socialization has been regarded by some investigators over the years as a field without a good theory that accounts for behavior change. At one time, I believed that theoretical considerations did not dominate research on peer relations because the grand theories did not account for every instance in which socialization effects can be detected. I now believe that we simply failed to see that we needed to understand the social structure of peer socialization, first, and then it would be easier to choose constructs from the cognitive and emotional armamentarium that could account for change.

Delineating the processes of peer influence is very difficult. Most of the contributors to this issue identify influence sources, determine whether influence flows from the child's selection of associates or mutual socialization, and determine whether exposure to certain associates interacts with child characteristics in ways that will account for significant change over time. Only Sage and Kindermann (this issue) address process issues and their work is a first step in identifying mechanisms of peer influence. Tracing rate changes in classroom task involvement back to contingencies within social interaction is extraordinarily difficult. Nevertheless, an association between these contingencies and some kind of behavior change needs to be established.

The best data connecting contingencies (talk, reward, punishment, and so on) and behavior change come from laboratory experiments or observational studies spanning short time intervals. Although major advances have been made in sequential analysis and its applications to naturalistic events, it is still nearly impossible to trace the contingencies that, over months or years, are associated with changing a child or adolescent from being a non-user to being a user (of drugs) or from being a nondelinquent to being a delinquent. Dishion and his colleagues (see Poulin et al.) identify "deviancy talk" (a quasi-Vygotskian notion) as one relevant construct. Whether deviancy talk has these effects under natural circumstances, however, is not yet known.

Most of the theoretical models tested in this special issue include measures that are relatively distant from social interaction and, consequently, are not very good proxies for "socialization process." The papers in this issue, then, share the shortcomings of the field as a whole. What the field needs to better understand influence processes are: (a) new ideas about mechanisms responsible for behavior change, and (b) methods for documenting them. The kinds of process-oriented studies needed are unclear. Laboratory assessments may once again be needed (Berndt, 1996) but not the kinds of experiments that were popular two decades ago. Rather, field experiments and intervention studies need to be designed and incorporated into multivariate longitudinal studies like those described in this issue.

DOMAIN RELEVANCEContributors to this issue each selected a single criterion to examine in terms of socialization antecedents. Some are narrow banded, including onset of drug use, cigarette smoking, and alcohol use. Others are more broadly banded, including engagement in the classroom, school adjustment, and delinquency. Overall, these studies are representative: Peer influence is not examined in terms of multiple outcomes or generalized measures; almost everyone seems to assume that these influences are contextualized.

Domain differences in the extent to which children influence one another are not easy to explain and probably have different origins from instance to instance. First, greater socialization effects are usually evident when base rates are zero or nearly so (e.g., when children do not smoke or drink) than when they are higher (e.g., when children are already deep into smoking or drinking). Can improvement in school adjustment be demonstrated among children who are already leaders? How much deterioration can be demonstrated among children who already have substantial records of law-breaking? Clearly, one domain-related condition that constrains peer influences is the base rate.

Second, domain specificity is suggested by evidence showing that different friendship features moderate social effects in different normative areas. Friendship stability together with friendship quality, for example, affects changes in sociability and leadership across the transition to adolescence, whereas friendship stability together with the friend's behavior affects changes in behavior problems across this same transition (Berndt et al.). Friends' attitudes toward smoking are linked to smoking onset (Rose et al.) and friends' drinking to alcohol use (Schulenberg et al.) but, on the other hand, friends' substance abuse is more moderately related to drug use (Pilgrim et al.).

Third, child characteristics moderate socialization differently from domain to domain. For example, friends' substance use affects the onset of drug use depending on the subject's sensation seeking and the parenting style employed in the child's home (Pilgrim et al.); susceptibility to social influence moderates the effects of drinking exposure (Schulenberg et al.); friendship quality moderates increases in delinquency depending on delinquency base rates (Poulin et al.). Future investigators may be able to sort out these differences so that a single model will account for socialization outcomes in every arena. For the moment, however, the data suggest that peer influences differ for deviant and nondeviant behavior as well as for different behaviors within these broad classes.

Fourth, it may be that there are some children for whom peers are salient socialization agents and others for whom they are not. "Equifinality" characterizes social development and adaptation, that is, equivalent outcomes may be reached via different developmental routes. Future investigators are certain to be more and more knowledgeable about the range of pathways by which child characteristics combine with parent and/or peer socialization, cycle through time, and bring about behavioral outcome.

CONCLUSIONThe contributors to this special issue of the Merrill-Palmer Quarterly are to be congratulated on their common interest in the processes by which children instigate behavior change in one another. This topic once was at center stage in peer relations research (Hartup, 1970), but interest in it has languished except among researchers dealing with social deviance. Most researchers have chosen recently to study the correlates of sociometric status and their role in social development, the connections between family socialization and peer socialization, and the significance of friendship relations. Longitudinal studies like those included in this issue extend into another realm, however, and leave very little doubt about children's significance in the lives of other children.

These contributions accomplish something more: They demonstrate that a new look must be taken at the "influence construct." We can no longer be content merely to speculate about "what happens" in certain kinds of peer situations. One can guess that negative friendship features, for example, bring about heightened opportunities to learn teasing, arguing, and assertive strategies in conflict resolution (Berndt, 1996), opportunities that may generalize to other social relationships and situations. Deviant talk, for example, may generalize from social interaction between friends to antisocial behavior in other situations (Poulin et al.). Contemporary investigators are not without interesting and reasonable hypotheses to account for behavior change emanating from child-child interaction. Nevertheless, one can rightfully challenge the research community to develop and employ variables that are closer to "process" than the ones used in current work, and to supplement their measurements with experimental assessments and expanded longitudinal designs that will tell us more completely "what happens" in the course of the social interaction that occurs between children or between adolescents.

Added material.

Willard W. Hartup.

University of Minnesota.

Correspondence may be sent to Willard W. Hartup, Institute of Child Development, University of Minnesota, 51 E. River Road, Minneapolis, MN 55455. Electronic mail may be sent via Internet to hartup@tc.umn.edu.

FOOTNOTE1 Except for the statistical modeling across cultures, which is an important contribution to our knowledge of peer influences as well as to what we know about the onset of drug use.

REFERENCESBALDWIN, M. W. (1992). Relational schemas and cognition in close relationships. Psychological Bulletin, 112, 461-484.

BARKER, R. G., & WRIGHT, H. F. (1955). Midwest and its children. New York: Harper & Row.

BERNDT, T. J. (1979). Developmental changes in conformity to peers and parents. Developmental Psychology, 15, 608-616.

BERNDT, T. J. (1996). Exploring the effects of friendship quality on social development. In W. M. Bukowski, A. F. Newcomb, & W. W. Hartup (Eds.), The company they keep: Friendship in childhood and adolescence. Cambridge, UK: Cambridge University Press.

BOWLBY, J. (1969). Attachment and loss. Vol. 1: Attachment. New York: Basic.

BUGENTAL, D. B., & GOODNOW, J. J. (1997). Socialization processes. In W. Damon (Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of child psychology: Vol. 3. Social, emotional, and personality development (5th ed.). New York: Wiley.

CHASSIN, L., PRESSON, C., SHERMAN, S. J., MONTELLO, D., & MCGREW, J. (1986). Changes in peer and parent influence during adolescence: Longitudinal versus cross-sectional perspectives on smoking initiation. Developmental Psychology, 22, 327-334.

FLAVELL, J. H., & MILLER, P. H. (1997). Social cognition. In W. Damon (Series Ed.) & D. Kuhn & R. Siegler (Vol. Eds.), Handbook of child psychology: Vol. 2. Perception, cognition, and language (5th ed.). New York: Wiley.

FRANCE-KAATRUDE, A., & SMITH, W. P. (1985). Social comparison, task motivation, and the development of self-evaluative standards in children. Developmental Psychology, 21, 1080-1089.

GELMAN, R., & WILLIAMS, E. M. (1997). Enabling constraints for cognitive development and learning: Domain specificity and epigenesis. In W. Damon (Series Ed.) & D. Kuhn & R. Siegler (Vol. Eds.), Handbook of child psychology, Vol. 2: Cognition, perception, and language (5th ed.). New York: Wiley.

HARTUP, W. W. (1970). Peer interaction and social organization. In P. H. Mussen (Ed.), Carmichael's manual of child psychology (Vol. 2). New York: Wiley.

HARTUP, W. W. (1983). Peer relations. In P. H. Mussen (Series Ed.) and E. M. Hetherington (Vol. Ed.), Handbook of child psychology: Vol. 4. Socialization, personality, and social development (4th ed.). New York: Wiley.

HUDSON, L. M., FORMAN, E. R., & BRION-MEISELS, S. (1982). Role-taking as a predictor of prosocial behavior in cross-age tutors. Child Development, 53, 1320-1329.

LAURSEN, B., HARTUP, W. W., & KOPLAS, A. (1996). Toward understanding peer conflict. Merrill-Palmer Quarterly, 42, 76-102.

LUDEKE, R. J., & HARTUP, W. W. (1983). Teaching behaviors of 9- and 11-year-old children in mixed-age and same-age dyads. Journal of Educational Psychology, 75, 908-914.

MANDLER, J. M. (1997). Representation. In W. Damon (Series Ed.) & D. Kuhn & R. Siegler (Vol. Eds.), Handbook of child psychology: Vol. 2. Perception, cognition, and language (5th ed.). New York: Wiley.

MOSHMAN, D. (1997). Cognitive development beyond childhood. In W. Damon (Series Ed.) & D. Kuhn & R. Siegler (Vol. Eds.), Handbook of child psychology: Vol. 2. Perception, cognition, and language (5th ed.). New York: Wiley.

PIAGET, J. (1932). The moral judgment of the child. New York: Free Press.

RUBIN, K. H., BUKOWSKI, W., & PARKER, J. G. (1997). Peer interactions, relationships, and groups. In W. Damon (Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of child psychology: Vol. 3. Social, emotional, and personality development (5th ed.). New York: Wiley.

SELMAN, R. L. (1980). The growth of interpersonal understanding. New York: Academic.

URBERG, K. A., DEGiRMENCiOGLU, S. D., & HALLIDAY-SCHERR, K. (1995). The structure of adolescent peer networks. Developmental Psychology, 31, 540-547.

VITARO, F., TREMBLAY, R. E., KERR, M., PAGANI, L., & BUKOWSKI, W. M. (1997). Disruptiveness, friends' characteristics, and delinquency in early adolescence: A test of two competing models of development. Child Development, 68, 676-689.

WHITING, B. B., & WHITING, J. W. M. (1975). Children of six cultures. Cambridge, MA: Harvard University Press.