23
JOURNAL OF APPLIED DEVELOPMENTAL PSYCHOLOGY 17, 215-237 (1996) A Social Learning Model of Peer Choice in the Natural Environment JAMES SNYDER LISA WEST VANESSA STOCKEMER SHANNON GIBBONS Wichita State University LAURA ALMQUIST-PARKS Cowley County Mental Health Center The social processes used by preschool-aged children to choose peer affiliates were assessed in a natural classroom setting. The results indicated that children were quite selective in their affiliation with peers; they spent substantial social time with a small number of peers and little time with remaining peers. Children’s initiol affiliotion with each of an array of same- and opposite-gender peers and temporal changes in those affiliations were strongly related to the relative proportion of positive conse- quences the children experienced during social interaction with those peers. Strong mutual affiliations or friendships were established between children who provided each other with the highest levels of positive social consequences available from peers in the classroom. Research in developmental psychology has begun to document the manner in which peer relationships influence individual variation in children’s social, emo- tional, and cognitive development (e.g., Hartup, 1989). A unique property of peer relationships is that they occur in an open field. Children can and do make choices about which peers they prefer and with whom they associate. These choices strongly influence the amount of time spent interacting with peers of various characteristics and, consequently, the nature of the social interactional environment experienced and the developmental effects of that environment. Those children with whom a child associates most will provide the bulk of social contingencies, modeling experiences, support and aid that the child derives from the peer group. From this perspective, assessment of the unique dyadic niches children establish within a larger peer setting, and identification of the processes by which they select and develop those niches are central to understanding between-individual variation in developmental outcomes. The choice of peer associates represents an important vehicle by which children are active agents in Correspondence and requests for reprints should be sent to Jim Snyder, Department of Psychology, Wichita State University, Wichita, KS 67260-0034. 215

A social learning model of peer choice in the natural environment

Embed Size (px)

Citation preview

JOURNAL OF APPLIED DEVELOPMENTAL PSYCHOLOGY 17, 215-237 (1996)

A Social Learning Model of Peer Choice in the Natural Environment

JAMES SNYDER

LISA WEST

VANESSA STOCKEMER

SHANNON GIBBONS

Wichita State University

LAURA ALMQUIST-PARKS

Cowley County Mental Health Center

The social processes used by preschool-aged children to choose peer affiliates were assessed in a natural classroom setting. The results indicated that children were quite selective in their affiliation with peers; they spent substantial social time with a small number of peers and little time with remaining peers. Children’s initiol affiliotion with each of an array of same- and opposite-gender peers and temporal changes in those affiliations were strongly related to the relative proportion of positive conse- quences the children experienced during social interaction with those peers. Strong mutual affiliations or friendships were established between children who provided each other with the highest levels of positive social consequences available from peers in the classroom.

Research in developmental psychology has begun to document the manner in which peer relationships influence individual variation in children’s social, emo- tional, and cognitive development (e.g., Hartup, 1989). A unique property of peer relationships is that they occur in an open field. Children can and do make choices about which peers they prefer and with whom they associate. These choices strongly influence the amount of time spent interacting with peers of various characteristics and, consequently, the nature of the social interactional environment experienced and the developmental effects of that environment. Those children with whom a child associates most will provide the bulk of social contingencies, modeling experiences, support and aid that the child derives from the peer group. From this perspective, assessment of the unique dyadic niches children establish within a larger peer setting, and identification of the processes by which they select and develop those niches are central to understanding between-individual variation in developmental outcomes. The choice of peer associates represents an important vehicle by which children are active agents in

Correspondence and requests for reprints should be sent to Jim Snyder, Department of Psychology, Wichita State University, Wichita, KS 67260-0034.

215

216 SNYDER, WEST, STOCKEMER, AND GIBBONS

their own development (Buss, 1987). The goal of this research was to examine selective peer association and preference among young children, and the social processes by which association and preference are established.

Previous research supports the hypothesis that children are selective in their association with peers. Hinde, Titmus, Easton, and Tamplin (1985) reported that the distribution of children’s observed association time with an array of peers shows a highly positive skewed distribution; children spend most of their time with a few peers and very little time with others. The majority of children appear to have at least one “strong associate” or “friend” as defined by association time with a specific peer or by mutual preference, regardless of their group socio- metric standing, social competence, or deviance (Cairns, 1983; Cairns, Cairns, Neckerman, Gest, & Gariepy, 1988; Dishion, Andrews, & Crosby, 1995; Howes, 1990; Warr & Stafford, 1991). Additionally, there is considerable tempo- ral continuity in these close relationships even in preschool-aged children (Hinde et al., 1985; Howes, 1983; Ladd, Price, & Hart, 1990).

Given that children report differential preference for peers and selectively distribute the time they associate with peers, the next question concerns the basis on which these differential preferences and selective associations are established. Previous research suggests that children seek out and prefer compatible peers. Initially, they choose peers who are similar on external, surface characteristics such as age (e.g., Rodgers, Billy, & Urdry, 1984), gender (e.g., Asher & Hymel, 1981), and ethnicity (e.g., Singleton & Asher, 1984).

After initial contacts, children make more refined choices about peer associ- ates and develop peer preferences on the basis of similarities in behavior such as styles of play (Rubin, Lynch, Coplan, Rose-Krasnor, & Booth, 1994), aggres- sion (Cairns et al., 1988; Dishion et al., 1995), drug use (Kandel, 1978), sexual activity (e.g., Rodgers et al., 1984), and school motivation (Kinderman, 1993). Children also tend to associate with peers who share similar group sociometric status (Ladd, 1983; Ladd, Price, & Hart, 1990). The importance of various behaviors in making peer choice varies with age (Gottman, 1986). For preschool- aged children targeted in this study, peer choice appears to be most closely tied to how “nice” (or reinforcing) peers are toward a child (Youniss, 1986). These descriptive data suggest that a person-environment linkage operates in peer choice such that children selectively seek out peers who are compatible with their own characteristics.

Children show a consistent same-gender bias in preference and association beginning as early as age 3 years. Previous research on peer choice has primarily dealt with this gender bias by focusing on same-gender relationships. Such an approach is inadequate. Although opposite-gender children have less social im- pact in terms of association and interaction time, that time is still substantial (20%-40% depending on setting and age; Daniels-Beimess, 1989). An adequate theory of selective peer choice should account for between-gender (group) varia-

PEER CHOICE 217

tion in preference, and for variation in selective preference within same- and

opposite-gender peer groups. Existing research concerning the processes involved in peer choice has several

limitations. First, most research has been done with older children and adoles- cents; beyond the role of surface characteristics, relatively little is known about the basis of preference and association with specific peers during early child- hood. The processes involved in peer choice may be more clearly delineated as children make the transition into their first systematic peer experiences, before those processes become “contaminated” with cumulative peer exposure and ex- perience. Second, much of the research on peer selection has been based on children’s self-report; there are few behavioral observation data relevant to pref- erence and association for specific peers. This may be particularly important because preschool children’s language abilities are taxed by even simple socio- metric assessment techniques. Third, most research has used a cross-sectional assessment of established dyadic relationships; a clearer picture of the processes involved in children’s selective peer choice at the dyadic level can be obtained by a longitudinal study of those processes in newly forming groups. Fourth, the research with younger children has primarily assessed the processes associated with peer choice in environments that are relatively artificial in terms of the number of children involved, the gender composition of the peer group, and the duration of peer contact. The generalization of these “lab” findings to the natural environment needs to be assessed.

In summary, selective peer preference and association by children in a social environment consisting of peers with varying characteristics play a potentially powerful role in between-child differences in development. Selective peer asso- ciation and preference are made on the basis of behavioral as well as surface characteristics indicating similarity or compatibility. Peer choice is likely to have a meaningful, cumulative developmental impact, given its moderate temporal continuity. However, the complementary, moderate discontinuity suggests a con- tinued adjustment in peer selection based on unfolding behavioral exchanges experienced and engendered in peer relationships. Peer preference and associa- tion are changing products of dynamic and iterative social processes. A model of the processes by which selective peer preference and association are established should account for the formation of initial peer choice in a newly forming group consisting of individuals of varying characteristics. The model should also speci- fy how this process operates over time to account for both temporal continuity and change in peer choice. Finally, it should account for between-gender vari- ance in peer choice as well as for variance in peer choice within gender groups.

This research tested a learning model of the social processes involved in selective peer preference and association. The first premise of the social learning model is that behavior is sensitive to the social environmental contingencies it generates. More specifically, the matching law (McDowell, 1988, 1989) sug-

218 SNYDER, WEST, STOCKEMER, AND GIBBONS

gests that individuals allocate their time between (and preference for) response alternatives in proportion to the relative rate of reinforcement available under those alternatives. Applying this notion to peer choice, a child would be expected to distribute associational time with and indicate verbal preference for each peer in proportion to the probability at which the child received positive social conse- quences during interaction with that peer relative to other peers. Over time, changes in a child’s preference for and association with each peer would follow shifts in the relative probability of positive social consequences received from that peer. This matching formulation should also account for differences in children’s preferences for same- versus opposite-gender children, consistent with Maccoby’s (1990) hypothesis that same- relative to opposite-gender relationships are more reinforcing.

The second premise of the model is that preference for and association with specific peers are complementary aspects of peer choice. Verbal reports of prefer- ences for and observed behavioral associations with peers are different mani- festations resulting from the social contingencies experienced in dyadic relationships with those peers. Consequently, a reliable correlation between peer preference and association time is expected.

This matching law, social learning model of peer choice is compatible with theory and research on young children’s peer relationships. Youniss (1986) sug- gested that young children (targeted in this report) define relationships on the basis of concrete outcomes experienced in those relationships. The basis for choice is literal: The degree to which a child likes and spends time with a peer depends on how the child is treated by that peer. Masters and Furman (1981) found that children’s liking and disliking of specific peers were reliably associ- ated with the rate of positive and negative actions directed toward the children by those peers. Gottman (1983) found that new acquaintances are likely to establish a friendship insofar as they interact in a reciprocally positive manner, exchange information with ease, find common-ground play activities, and are able to quickly resolve conflict. The formation of peer preferences and selection of peer associates appears to be based on the unique consequences engendered during the interaction of an individual with each peer rather than the general behavior of the individual or peer toward others (Kenny & LaVoie, 1984; Masters & Furman, 1981; Ross & Lollis, 1989). What’s important is how nice someone is to me rather than how nice they are to people in general.

The third premise of the social learning model is that reciprocity plays a dual role in peer choice. At a social interactional level, the consequences a child experiences in a relationship with a specific peer are strongly influenced by how the child behaves toward that peer. Individuals program their own social environ- ment by the behavior they direct toward others (Cairns et al., 1988). Interactional reciprocity is a nearly universal finding in research on social relationships (Buss, 1987; Patterson, 1982). This reciprocity increases in close relationships (Dis- hion, Andrews, Patterson, & Poe, 1994; Hinde et al., 1985; Olson, 1992) and

PEER CHOICE 219

changes as a function of the duration of these relationships, suggesting that reciprocity may play a role in the temporal stability and change in peer choice (Ross & Lollis, 1989; Olson, 1992).

At a second level, reciprocity refers to the degree to which there is dyadic mutuality or concordance in children’s preference and affiliation. This type of reciprocity is also important. Only a third of friendship nominations in young children are reciprocated, although most children are involved in at least one reciprocated friendship (Howes, 1988, 1990). Reciprocated relative to non- reciprocated friendships show higher temporal stability (Bukowski & Hoza, 1989; Howes, 1988; Kandel, 1978). Children in reciprocated friendships spend more time with one another than those in nonreciprocated friendships (Hartup, 1992) and are more likely to interact in a reciprocal, cooperative, and positive manner (Bukowski & Hoza, 1989; Hartup, 1992). Finally, a more powerful mutual shaping of behavior occurs in reciprocated than in nonreciprocated friendships (Kandel, 1978).

The following hypotheses derived from the social learning model of peer choice were tested:

1.

2.

3.

4.

5.

6.

The amount of time each child spends associating with each of an array of peers in a classroom environment will be different from chance. There will be a bias toward increased association with same-gender peers, but system- atic variation in association will also be found within same- and opposite- gender groups. There will be a reliable relationship between children’s stated preference for each of an array of peers and the amount of time they spend associating with those peers. There will be stronger preference for same- than for opposite- gender peers. There will be a linear relationship between the relative probability of pos- itive social consequences that children receive from each of an array of peers during interaction and (a) the relative amount of time children asso- ciate with each of those peers, and (b) the stated preference (liking- disliking) of children for each of these peers. Children will receive a higher proportion of positive social consequences from same- than from opposite-gender peers. The relative probability at which children receive positive social conse- quences from each of an array of peers will be linearly related to the relative probability with which those children distribute positive social consequences to each of those peers. Most children will be involved in at least one reciprocated or mutual friend- ship (defined by mutual liking and mutual high levels of association). Most mutual friendships will be among same-gender children. Children will designate as friends those peers from whom they receive the highest nrobabilities of nositive social conseauences.

220 SNYDER, WEST, STOCKEMER, AND GIBBONS

METHOD

Participants In this investigation, 74 children (36 boys and 38 girls) from four preschool classrooms participated. The children ranged in age from 49 to 62 months at the beginning of the preschool year. All but 2 children (whose data were not an- alyzed) were African American. Each classroom contained between 17 to 20 children, and had a child gender ratio no greater than 55%:45%. All of the children were from families with incomes below the poverty line, and 64% came from mother head-of-household families. Only 2 children who were enrolled in the same classroom had more than minimal contact (more than twice per month over the previous year) with each other prior to entry into the preschool.

Setting The children attended classes at a center 4 half-days per week. Space and equip- ment were comparable in the two classrooms (morning and afternoon classes in each) used in this research. The teachers in both classrooms followed a similar educational philosophy, in part as the result of program policies and training. This philosophy was that children learn from natural activities and interaction and need few structured activities to promote learning. Intervention in the inter- action between children occurred primarily when there was excessive aggression or a child was very withdrawn. Children were assessed from early September to late December so that data were collected as the children were becoming initially acquainted and establishing relationships with each other. All observations were made during “free play” time inside the center during which children had free access to each other and to an array of play materials and activities typical of a preschool room, including building materials (e.g., blocks), manipulatives (e.g., puzzles), nonviolent action figures (e.g., cars, trucks, zoo animals), and dramat- ic play (e.g., dress-up clothing, kitchen materials).

Measures

Peer Preference. Preference ratings for each classmate were collected indi- vidually from each child twice during the fall (October and December), using the rating procedure described by Asher, Singleton, Tinsley, and Hymel (1979). In this procedure, children are shown pictures of their classmates and are asked to place each picture on a “happy, ” “neutral,” or “sad” face to indicate whether they liked, neither liked nor disliked, or disliked playing with the pictured classmate. Thus, each child rated each peer on a 1 (disliked) to 3 (liked) preference scale. The only variation from the Asher et al. (1979) procedure entailed having the children make preference ratings using pictures of foods prior to making peer ratings in order to enhance their understanding of the procedure. The test-retest reliability of preferences over a 2-week period in a different, preschool-aged, African American sample was K = .73.

PEER CHOICE 221

Peer Association. The amount of time each child associated with each of the other children during free play periods in the classroom was observed and coded for twenty 15min sessions. These observations were made on a rotational basis such that no child was observed more than twice weekly. Half of these observa- tion sessions occurred in October and half in December.

Association was defined as being in physical proximity and verbally or physi- cally interacting with a peer (either as initiator or recipient), such that both parties actively contributed to interaction and that their behavior was interdependent. The criterion for determining the termination of behavioral association was 10 s without proximity or interaction. These data were collected by two observers (other than those used to collect other measures). The observers were trained to an interobserver agreement criterion of r = .88 in a separate classroom prior to collecting data in this study. Weekly training and recalibration sessions were used to minimize observer drift. One-fourth of the observations of peer associations were made by both observers and were used to assess interobserver consistency. The interobserver consistency for the association data derived for this study was r = .81.

Because of absences, not all peers were equally available as associates during the observation of various target children. This was corrected by calculating the duration of each child’s association with peers as percentages of the total amount of time those peers were present in the classroom during observation of that child. Children also varied in the amount of time they associated with peers versus played alone or interacted with teachers. This was corrected by stan- dardizing corrected association times by individual target child.1 These stan- dardized association times, unless otherwise noted, were used in all subsequent analyses.

Peer Interaction. Twenty 15min samples of the peer interaction of each child were observed and coded during free play periods in the classroom using the Peer Interaction Process code (Snyder, 1986). The observations were made on a rotational basis such that no child was the target of observation more than once every other day. Half of the observation samples were obtained in September and half in December.

The coding system provides a running descriptive account of the interaction of a child with peers along a real-time line. More specifically, for each social behavior of the target child or of peers toward the target child, the following data were coded: (a) the initiator of the action (target child or peer by name), (b) the content of the action in one of 21 social behavior categories (both verbal and nonverbal), (c) the recipient of the action (target child or peer by name), and (d)

‘The correction for peer availability = amount of time target child spends with a peer/(total

observation time of the target - amount of time peer was not in the classroom when the child was observed). Correction for individual differences in total peer association time entailed standardizing the corrected amount of time spent with each peer separately for each target child.

222 SNYDER, WEST, STOCKEMER, AND GIBBONS

the affect with which the action was delivered in one of three categories (posi- tive, neutral, or negative).* In the application of this coding system, observers stationed themselves as unobtrusively as possible (typically, 3 to 4 feet from the children), and softly dictated the coding of the interaction into a tape recorder.

Prior to coding the data for this study, four observers were trained using videotaped and live interaction in a classroom not used in this study. The crite- rion for completing training in terms of interobserver consistency was a Cohen’s Kappa of .95 for child identification and of .70 for content and affect categories for three consecutive 15min observational sessions. Observer training and recal- ibration continued weekly throughout data collection. One-fourth of all peer interactions were coded by two observers and used to estimate interobserver consistency.

For all analyses in this study, the 21 Content X 3 Affect code combinations were collapsed into two summary categories, positive and negative. The negative summary category consisted of the following Behavior X Affect combinations: all content codes delivered with negative affect; and the following content codes regardless of affect: noncompliance, strong commands, disagreement, destroy- ing another’s possessions, physical aggression, and verbal aggression. The fol- lowing content categories, when delivered with positive or neutral affect, were used to define the positive summary category: compliance, agreement, “let’s” requests, physical positive, verbal positive, describing another’s feelings, de- scribing another’s actions, asking about another’s actions, giving information, asking for information, describing one’s own feelings, describing one’s own actions, vocalizations, and getting another’s attention.3

Interobserver consistency as assessed by Kappa on the summary positive- negative categories averaged .8 1 and ranged from .72 to .87. Interobserver consistency as assessed for the 46 Specific content category X Affect combina- tions (that occurred with a frequency > 10 across all subjects) averaged .63 and ranged from a low of .51 (verbal positive with negative affect) to a high (giving information with neutral affect) of .70.

RESULTS

Three variables (preference, association, and positive social consequences re- ceived) were measured for each possible combination of dyadic relationships among children within a classroom. These variables were collected under the following conditions: Classroom (4) X Time (early vs. late fall) X Gender of target child (2) X Gender of relationship partner (2). No main effects were found

ZAdditional information concerning the coding categories and the psychometric properties of the

Peer Interaction Process Code can be received from the first author.

sTo be included in this and any subsequent analysis, the percentage of positive consequences received or given in a dyadic relationship were calculated only for those relationships in which n >

10 consequences were generated.

PEER CHOICE 223

TABLE 1 Gender Bias in Peer Preference, Peer Association Time,

and Positive Consequences Received From Peers

Male preference for Male % association time

with Male % positive conse-

quences from

Female preference for Female % association

time with Female % positive conse-

quences from

Boys Girls

M SD M SD t

2.14 .89 1.78 .86 3.01* 10.8 9.1 4.6 7.8 6.11**

57.5 16.2 49.2 18.3 3.01*

1.74 .84 2.37 .86 4.53** 4.7 8.1 10.1 8.8 4.77**

50.3 17.6 60.5 14.0 4.17”’

*p < .Ol. **p < ,001.

for any of the three variables, and no interaction effects were found involving classroom or time. The mean percent of positive social consequences experi- enced by children was 54, SD = 17. Children spent an average 52% of their free play time in the classroom interacting with peers, SD = 19, 15% interacting with the teacher, SD = 15, and 33% in nonsocial activities, SD = 27. The mean peer preference rating was 2.01 on a 3-point scale, SD = .56.

Gender Bias Significant interaction effects of Target child gender X Gender of relationship partner were found for peer preference, association time with peers, and for positive social consequences received from peers, as shown in Table 1. Both boys and girls showed a distinct, temporally consistent, and comparable positive bias toward same-gender peers on all three variables. They expressed greater liking for, spent more time with, and received a higher proportion of positive consequences from same- relative to opposite-gender peers. These data provide support for the gender biases predicted in subordinate Hypotheses 1, 2, and 3.

Distribution of Association Time with Peers Frequency distributions of the percent of total peer time that each boy and girl spent associating with each of his or her same- and opposite-gender peers (aggre- gated across dyadic relationships and classrooms conditional on gender) during the early fall are shown in Figure 1. A very similar set of distributions were observed in the late fall. The distributions of boys’ and girls’ association with same- and opposite-gender peers were very positively skewed. Children were selective in their peer associations; they spent the majority of their time with only

2 5

oz-ST ‘2 GI-01 g

OI-so a so-00

spr?da 30 huanbaq

224

80

60

40

20 0

Fem

ales

w

ith

Fem

ales

Perc

ent

Tim

e A

ssoc

iatio

n Pe

rcen

t T

ime

Ass

ocia

tion

Figu

re 1.

Dis

trib

utio

n

of

bo

ys’

an

d

girl

s’

ass

oc

iatio

n

time

with

sch

oo

l ye

ar.

-$

80

R

a

0

Fem

ales

w

ith

Mal

es

sam

e-

an

d

op

po

site

-ge

nd

er

pe

ers

in

th

e e

arly

fa

ll o

f th

e

226 SNYDER, WEST, STOCKEMER, AND GIBBONS

a few of their (typically same-gender) classmates, and minimal time with the rest of their classmates.

Assuming a random distribution of peer association, each child would spend approximately 7% of his or her total association time with each peer. The ob- served distributions of association time were compared with an expected distri- bution (defined as equal proportions of the total association time spent with each available classmate conditional on gender) using Kolmogorov-Smimov two- sample tests. Reliable differences between observed and expected distributions were found in same-gender associations of boys and girls during both the early and late fall, all chi-squares (20) > 50, p < .OOl . The distributions of boys’ and girls’ association time with opposite-gender classmates were also reliably differ- ent from chance during the early and late fall, all chi-squares (20) > 38, p < .Ol. As described in the previous section, children spend more association time with same- compared with opposite-gender peers. This same-gender bias is also ap- parent in Figure 1. Boys and girls have nearly three times as many strong associates (defined as spending > 15% of total time with a target peer) of the same (about 30%) compared with the opposite (about 10%) gender.

This provides support for Hypothesis 1, suggesting that children are selective in their behavioral associations with both opposite- and same-gender peers. How- ever, even given this selectivity applies to relationships with peers of both genders, the normative gender-split characteristic of young children is still apparent.

Peer Association, Peer Preference, and Social Reinforcement Aggregating across dyadic relationships (conditional on gender) within a class- room, children’s preference for each of an array of peers was reliably correlated, r = .44 to .64, all ps < .OOl, with the percentage of time that the children spent associating with those peers during free play time in the classroom setting. Despite a same-gender bias in preference and association, there were no differ- ences in the magnitude of these correlations for same- and opposite-gender dyads. The correlations were also comparable in the early and late fall assess- ment periods. This supports Hypothesis 2, which suggests that verbal preference and behavioral association are complementary indicators of the selective affilia- tions that children create with peers.

Children’s preference and association scores were standardized by participant (indicating each child’s relative preference for and relative association with each of an array of peers), and summed to form an affiliation construct for each dyadic relationship in which a child was involved. The relative proportion of positive social consequences received by a child from each of an array of peers was calculated using the binomial z-statistic comparing the proportion of conse- quences that a child received from each specific peer with the marginal value of positive social consequences received from peers in general. The affiliation construct scores for each dyadic relationship of a child and the relative proba- bility with which a child received positive consequences from each dyadic part-

PEER CHOICE 227

ner were aggregated across children and classrooms conditional on gender to test the proposed matching law model of selective peer affiliation.

The matching law hypothesizes that children’s selective affiliations (i.e., the sum of standardized preference and association scores) with each of an array of peers is predicted by the relative probability with which they receive positive social consequences from each of those peers. As shown in Figure 2, significant correlations, .72 to .82, were found between the probability of positive conse- quences derived by a child from a dyad partner relative to peers in general in September and the child’s affiliation with that partner in October. These findings were replicated in the November consequences-December affiliation assess- ment, correlations of .69 to .83, and were consistent for both boys’ and girls’ relationships with same- and opposite-gender peers. Despite a same-gender bias in preference, association, and positive consequences, children’s relative afilia- tion with same- and opposite-gender peers related closely to the relative proba- bility of positive social consequences received in relationships with those peers.4

As also shown in Figure 2, there was a modest l-month stability in the relative probability of positive consequences received by boys (.45) and girls (.26) from same-gender peers, and a lower stability in positive consequences received from opposite-gender peers (.28 and .05, respectively). There was a higher but still modest temporal stability in same- and opposite-gender peer affiliation over a l-month period (.22 to .68).

A stronger test of the matching hypothesis entails ascertaining whether the relative rate of positive consequences received from peers predicts changes in affiliation with those peers. This was assessed by calculating correlations be- tween positive consequences in November and peer affiliation in December after partialling out affiliation in October. As shown in Figure 2, relative probability of positive consequences received from peers in November predicted October-to- December changes in the degree to which a child affiliated with those peers, partial TS from .68 to .79. These partial correlations were reliable for boys’ and girls’ relationships with both same- and opposite-gender peers. A combination of peer affiliation patterns in October and the relative probability of positive conse- quences in November accounted for 45% to 76% of the variance in December peer affiliations. These data are consistent with Hypothesis 3.

In a similar fashion, it was possible to assess the impact of children’s peer affiliations on the relative proportion of positive consequences received from those peers. The correlations between October peer affiliations and the relative probabilities of positive consequences received from peers in November were

4The relationships between peer affiliation and social reinforcement were also separately assessed for each child. The median correlations for the relation of reinforcement and peer affiliation were

smaller than those reported in the aggregated analyses. There was modest variability in these correla- tions across subjects; the correlations in 90% of the cases was greater than .40, suggesting an

association that can be generalized across subjects. Results of individual analyses can be obtained from the first author.

September October November December

Females with Females:

Females with Males:

Males with Males:

Males with Females:

R*~08~

Reinforcement .26 p Reinforcement Received Received

.&

R2=.72b

Reinforcement Received

R2=.02’

Reinforcement * Received

.22

R2= 45b

Reinforcement Received

R2=.30b

) Reinforcement .45 Received

R’~76~

Reinforcement _ Received

.28

R’~14~

+ Reinforcement Received

R2=.b6b

preference

preference

association

preference

association

preference

association

apartial correlation b multiple R2

Figure 2. Relationships between the relative probability of reinforcement received from

peers and affiliation with peers.

228

PEER CHOICE 229

low to moderate (rs from .13 to .55). These correlations diminished after par- tialling out positive consequences received from peers in September (partial rs from .l 1 to .35). September positive consequences and October affiliation ac- counted for a small amount of variance (2% to 8%) in the relative rate of positive consequences received by girls, and a somewhat higher but still modest amount of variance (14% to 30%) in the relative rate of positive consequences received by boys. Changes in positive social consequences experienced during peer inter- action were not strongly influenced by antecedent dyadic affiliations (especially for girls), at least as assessed during the first 3 months of group formation.

It was hypothesized that there would be reciprocity in positive social conse- quences within dyadic relationships. To test this hypothesis, correlations were calculated between the relative probability of positive consequences a child provided for each peer and the relative probability with which the child received positive consequences from each peer, aggregated across children and class- rooms conditional on gender. Reliable correlations, all ps < .OOl , were found for dyadic reciprocity in positive consequences in boys’ relationships with boys, Sept. r = .56, Nov. r = .70, and girls, Sept. r = .51, Nov. r = .67, and in girls’ relationships with girls, Sept. r = .61, Nov. r = .75, and boys, Sept. r = .56, Nov. r = .72. These data support Hypothesis 4.

Mutual and Unilateral Affiliations Strong affiliations were defined in those cases in which a child indicated a preference for a classmate (rating of 3) and spent at least 15% of total peer association time with that classmate (approximately twice that expected by chance). These strong affiliations could be mutual (each child in the dyad was a strong affiliate of the other) or unilateral (only one child in a dyad showed strong affiliation for the other). As shown in the top half of Table 2, 72% to 87% of the children had one or more mutually strong affiliation (i.e., friendships) during October and December, and between 60% and 70% of these friendships were stable over the October-to-December interval. In October 13% of all friendships and in December 9% of all friendships involved opposite-gender dyads. These data support Hypothesis 5.

As shown in the bottom half of Table 2, 63% to 84% of children indicated at least one strong affiliation for a peer that was unilateral. Only one-fourth of these unilateral affiliations were stable over time. Unilaterally strong affiliations with opposite-gender peers occurred more often than mutual strong affiliations, but were still relatively uncommon: 24% of all strong unilateral affiliations in Octo- ber and 29% in December targeted opposite-gender peers. All children indicated a strong affiliation with one peer (either unilateral or bilateral) and 75% of children indicated two or more strong affiliations.5

5However, 10% to 20% of children are not targets of any other child’s strong affiliations; no one

wants to be their friend. These children are, by definition, without strong mutual affiliations or friendships. They want to have friends, but no one wants to be friends with them.

230 SNYDER, WEST, STOCKEMER, AND GIBBONS

TABLE 2 Incidence and Temporal Stability of Strong Peer Affiliations

Boys Girls

Boys Girls

Children with Strong Affiliations 1%)

Oct.-Dec. Continuity October December in Affiliationa (%.)

Mutual affiliations 78 72 61 82 87 68

Unilateral affiliations 67 63 25 84 63 24

aExpressed as a percentage of the total number of strong affilia- tions in October that were still strong affiliations in December.

Table 3 provides data testing the notion that children select as strong affiliates peers who are optimal sources of positive social consequences within the peer group. Mutual same-gender strong affiliates provided the first- or second-best sources of positive social consequences for each other 85% to 90% of the time. The observed distribution of rank orders of the amount of positive consequences from mutually strong affiliates was significantly different than that expected by chance, all chi-squares (7) > 56, ps < .OOl , for boys and girls in the early and late fall.

The bottom half of Table 3 provides comparable data for the positive conse- quences available from same-gender unilateral strong affiliates. These affiliates

TABLE 3 Relationship of Social Reinforcement to Strong Affiliations With Peers

Affiliations of:

Rank Order of Probability of Reinforcement Available in Relationships with Same-Gender Peers

Early Fall Late Fall

1 2 3 >4 1 2 3 >4

Mutually strong affiliations Boys: n 15 7 3 0 13 7 3 0

% 60 28 12 0 57 30 15 0 Girls: n 18 5 4 0 21 6 3 0

% 65 19 15 0 69 20 11 0

Unilaterally strong affiliations Boys: n 5 5 3 5 3 6 8 10

% 21 21 37 21 27 18 24 30 Girls: n 8 4 5 8 7 5 7 5

% 32 16 20 32 23 21 29 21

PEER CHOICE 231

TABLE 4 Comparisons of the Probability of Positive Social Reinforcement

Received From Same-Gender Mutually Strong Affiliates, Unilaterally Strong Affiliates, and Non-Affiliatesa

Tvpe of Relationship

Mutual Affiliate Unilateral Affiliate Nonaffiliate M SD N M SD N M SD N

Early Fall Among boys 63 04 25 56 08 24 49 06 34 Among girls 62 07 27 55 06 25 50 04 38

Late Fall Among boys 66 05 23 58 04 33 47 10 34 Among girls 67 04 30 53 03 24 50 09 38

aAll comparisons between reinforcement received from mutual, unilateral, and nonaffiliates were significantly different, all rs (df = 48 to 56) 1 3.75, ps < .OOl.

ranked among the top three best sources of positive consequences approximately 70% of the time. The observed distribution of rank orders of positive conse- quences from unilaterally strong associates was significantly different than that expected by chance, all chi-squares (7) > 17, ps < .05, for boys in the early fall, and for girls in the early and late fall.

Table 4 compares the mean probability of receiving positive social conse- quences from mutually strong affiliates, unilaterally strong affiliates, and non- affiliates (one randomly selected for each child) of the same gender. In each case, children receive a significantly higher proportion of positive social consequences from mutually than from unilaterally strong affiliates, and a significantly higher proportion of positive social consequences from unilaterally strong affiliates than from nonaffiliates. These data support Hypothesis 6.

DISCUSSION

The data clearly indicate, congruent with previous research (Hinde et al., 1985; Ladd et al., 1990), that children are highly selective in their peer affiliations. In an open field, the young children in this study indicated preferences for specific peers and were observed to allocate a majority of their social time with a small number of available peers. Children’s learning and social experience, beginning as early as the preschool years, are most powerfully influenced by those peers with whom children choose to most closely affiliate. Depending on the charac- teristics of a child’s close peer affiliates, these social experiences may lead to divergent developmental outcomes, including the development of antisocial be- havior and aggression, or of prosocial behavior, cooperation, and sharing (Cairns et al., 1988). Given temporal stability in peer choice, the specific dyadic niches

232 SNYDER, WEST, STOCKEMER, AND GIBBONS

established by children within the larger peer group may critically influence the trajectory of children’s social development.

Given the potentially powerful effects of peer choice on child development, it is important to delineate the processes by which such choices are made. The data suggest that the degree to which children affiliate with various peers is closely related to the social consequences experienced when interacting with those peers. This relation was observed during early group formation, but more importantly, the positive consequences engendered in relationships with peers predicted changes in selective affiliation with those peers from earlier to later stages of group formation. The relative proportion of positive consequences experienced in peer relationships was predictive of peer affiliation later in the school year even after accounting for earlier affiliation.

These data extend previous research on the role of social consequences in peer choice (e.g., Masters & Furman, 1981). The data reflect the sensitivity with which each individual child affiliates with peers on the basis of his or her own social experience rather than a child’s sociometric standing averaged across peer preference. The data were collected in a newly forming peer group so that the social processes involved in the establishment of peer choice were assessed from their inception without contamination by previous social contacts. The use of a short-term, longitudinal design provided the means to examine the processes involved in changes in peer choice using a more rigorous (though still correla- tional) test of the causal role of social consequences in peer affiliation (Gollub & Reichardt, 1987). The relationships between social consequences and selective affiliation did not reflect shared method variance; different methods and/or infor- mants were used to define these constructs.

Gender biases in social consequences received from peers and affiliation with peers were apparent during the early and later stages of group formation. This suggests that the preschool-aged children in this study had already learned gener- alized gender-based rules about the exchange of positive behavior, preference, and behavioral association. More importantly, the relative proportion of positive consequences experienced in peer relationships accounted for substantial vari- ance in children’s affiliative choice among opposite- as well as same-gender classmates. Children are discriminating and selective toward opposite- as well as same-gender peers despite gender bias. The findings are consistent with the social processes hypothesized to account for peer choice, but they also accom- modate well-documented differences in children’s relationships with same- ver- sus opposite-gender peers (Daniels-Beimess, 1989).

More specifically, the findings are compatible with Maccoby’s (1988, 1990) theory concerning gender segregation. Interaction with opposite-gender peers is less rewarding than that with same-gender peers. The data also extend Mac- coby’s theory. Children’s view of opposite-gender children is neither uniform nor blind. Opposite-gender peers are not dismissed in a blanket way. Rather, children make refined affiliative distinctions among (and thus presumably are influenced

PEER CHOICE 233

by) opposite- as well as same-gender peers despite substantial differences in the social payoffs resulting from interaction with those peers.

Theoretically, it is equally plausible that peer affrliative choices would influ- ence the relative proportion of positive consequences received in peer relation- ships. The more you like and hang out with someone, the more rewarding the relationship. Children’s selective affiliation with peers showed low to moderate correlations with the proportion of positive consequences received from those peers and accounted for modest amounts of temporal change in the relative proportion of social consequences received from peers.

The panel design used in this research is inherently correlational. Neverthe- less, this suggests that the positive consequences experienced in relationships with peers is a critical process involved in the development and shaping of peer affiliations. Temporally unstable behavior-consequence exchanges during inter- action between group members contributes to an iterative refinement in affilia- tion patterns within the group. Selective peer affiliations, once initially established, have relatively modest effects on the relative proportion of positive consequences subsequently received from peers.

It is likely that measures of social interaction (e.g., mutual coordination of behavior to achieve more extended cooperative play sequences or enhanced means to quickly resolve conflict) other than positive consequences may be more sensitive indicators of the relational changes engendered by selective peer affilia- tion. This is supported by the observation of high levels of reciprocity in positive social consequences between dyad partners. Children contribute in an active, behavioral manner to the quality of the relationships they experience with their peers (Patterson, 1982).

All children designated at least one other peer as a strong affiliate, and most children were involved in at least one strong mutual affiliation (or friendship). Mutually strong affiliations were relatively stable over the 2-month period of this study, but unilaterally strong affiliations were not. This is consistent with previ- ous research (Howes, 1990; Kandel, 1978). How do children choose their friends? Peers are designated as strong mutual affiliates or friends insofar as they are the source of the highest proportions of positive social consequences avail- able from classmates. Unilaterally strong affiliates also provide higher than aver- age proportions of positive social consequences, but not as high as those from mutual friends. This is consistent with previous research suggesting that the interaction among friends is characterized by high levels of cooperation and support and efficient means of resolving conflicts (e.g., Gottman, 1983, 1986).

These data are compatible with a social learning, reinforcement model of peer choice (Snyder, 1995). More specifically, the substantial linear relationship be- tween the proportion of positive consequences received from each peer relative to peers in general and children’s affiliation with each of those peers is consistent with the matching law formulation of the social processes involved in peer choice (Dishion & Andrews, 1995). Children allocate social time and express a verbal

234 SNYDER, WEST, STOCKEMER, AND GIBBONS

preference for peers consistent with the functional value of those peers in chil- dren’s recent social experience. As this functional value changes as a result of reciprocal interaction within relationships, children adjust their time allocation and preferences to reflect those experiences. The choice of mutually strong associates or friends is compatible with the ecological model of optimal foraging (e.g., Staddon, 1980); individuals spend the most time and energy in those environments providing the best payoffs, and change environments as optimal sources of payoffs shift. Simultaneously, individuals spend a small amount of time surveying other less-reinforcing environments in order to stay in contact with possible shifts in contingencies among available environments. Individuals’ behavior reciprocally influences the contingencies found in those environments (e.g., reduction in foodstuffs or prey as a result of eating).

Social learning and reinforcement theories have been criticized for describing individuals as passive agents, buffeted by environmental forces beyond their control (e.g., Maccoby, 1992). As described here, this is not an inherent char- acteristic of these theories. Children can be active agents in their own develop- ment and demonstrate behavioral sensitivity to environmental contingencies. As shown by reciprocity in the exchange of social consequences in peer relation- ships, children’s behavior influences the social environment they experience as well as being influenced by that environment. As demonstrated by peer choice, children are active agents in that they select the environments to which they are exposed. Theories focusing on affect (e.g., Gottman, 1991) or cognition (e.g., Dodge & Coie, 1987) might also be used to account for these data. However, there are no inherent characteristics of learning or reinforcement theories incom- patible with the notion of persons as active agents contributing to the dynamic processes involved in socialization.

Two decades of research have documented that children’s peer relationships have a substantial influence on their short- and long-term social development and adjustment. However, attempts to foster better peer relations in preventive or clinical interventions have resulted in very modest effects (LaGreca, 1993) and may occasionally be counterproductive (Dishion & Andrews, 1995). In part, this is the result of a failure to carefully delineate the social interactional pro- cesses that mediate peer relational “effects” (Hartup, 1992; Schneider, 1989). Peer affiliative choices strongly influence the types of contingencies and social stimuli experienced by a child. The data in this study suggest that interventions focusing exclusively on changing a target child’s behavior without simul- taneously altering the child’s peer affiliations and the contingencies provided by peer affiliates may be inherently limited. The child, despite behavior changes engendered by intervention, is likely to spend most of his or her interactional time with established strong affiliates (those peers who previously provided optimal reinforcement) upon return to the natural peer group. Insofar as the stimuli and contingencies offered by the child’s strong peer affiliates (as well as

PEER CHOICE 235

other children) are unchanged, behaviors newly acquired and shaped during intervention will be quickly dissipated or extinguished, and there will be no shift in strong peer affiliations. Bringing problematic children together to deliver group or specialized interventions may actually foster increased affiliation with other problematic children. To have more powerful and lasting effects, interven- tions may need to alter peer behavior and to actively structure peer affiliations as well as modify the target child’s behavior.

The increased understanding of the processes involved in peer choice provides only half the picture. The next step entails assessing the nature of the mutually shaping processes occurring during the interaction of strong, mutual peer affili- ates. Once afftliates are selected in a manner that matches or optimizes the relative payoffs available in the natural group, how is children’s behavior changed, and how does that change contribute to the child’s developing reper- toire, subsequent stability or change in peer affiliation in the current group, and peer affiliative choices in other groups or settings?

The simultaneous assessment of all possible dyadic relationships between individuals comprising a natural group provides a rich and ecologically valid approach, but it also presents difficult methodological and data analytic prob- lems. The data collected in this research were ipsative, interdependent, and skewed in distribution, often violating the assumptions required for powerful statistical procedures. Most naturally occurring groups of manageable size (in terms of the intensive and repeated data collection needed to study social pro- cesses) are relatively small and are further reduced given the heterosocial compo- sition of these groups requiring control for gender effects. Aggregation of ipsative data across children to increase sample size may further confound prob- lems of statistical independence. Whereas statistical methodology should not dictate the theoretical models being offered and pursued, the degree to which those models can be unambiguously tested by the data depends on the availability of appropriate statistical methods. This is certainly the case in this report and, as a result, it may be possible to argue that the observed relationships are a reflec- tion of methodological or statistical artifacts rather than actual processes occur- ring in the environment.

REFERENCES

Asher, S., & Hymel, S. (1981). Children’s social competence. In J. Wine & D. Syme (Eds.), Social competence (pp. 1255157). New York: Guilford.

Asher, S.R., Singleton, L.C., Tinsley, B.R., & Hymel, S. (1979). A reliable sociometric measure for preschool children. Developmental Psychology, 15, 443-444.

Bukowski, W.M., & Hoza, B. (1989). Popularity and friendship: Issues in theory, measurement and outcome. In T. Bemdt & G. Ladd (Eds.), Peer relationships in child development (pp. 15- 45). New York: Wiley.

Buss, D.M. (1987). Selection, evocation and manipulation. Journal of Personality and Social Psy- chology, 53, 1214-1221.

236 SNYDER, WEST, STOCKEMER, AND GIBBONS

Cairns, R.B. (1983). Sociometry, psychometry, and social structure: A commentary on six recent studies of popular, rejected and neglected children. Merrill-Palmer Quarterly, 29, 429-438.

Cairns, R.B., Cairns, B.D., Neckerman, H.J., Gest, S.G., & Gariepy, J.L. (1988). Social networks and aggressive behavior: Peer support or peer rejection. Developmental Psychology, 24, 815- 823.

Daniels-Beimess, T. (1989). Measuring peer status in boys and girls: A problem of apples and

oranges. In B.H. Schneider, G. Attili, J. Nadel, & R.P. Weissberg (Eds.), Social competence

in developmental perspective (pp. 107-120). Boston: Kluwer.

Dishion, T.J., Andrews, D.W., Patterson, G.R., & Poe, J. (1994). A process analysis of confluence in antisocial behavior in male adolescent friendships. Unpublished manuscript, Oregon So- cial Learning Center, Eugene, OR.

Dishion, T., & Andrews, D. (1995). Preventing escalation in problem behaviors with high risk

children: Immediate and one year outcomes. Journal of Consulting and Clinical Psychology, 63, 538-548.

Dishion, T.J., Andrews, D.W., & Crosby, L. (1995). Antisocial boys and their friends in early adolescence: Relationship characteristics, quality, and interactional process. Child Develop- ment, 66, 139-151.

Dodge, K.A., & Coie, J.D. (1987). Social information processing factors in reactive and proactive

aggression in children’s peer groups. Journal of Personality and Social Psychology, 53, 1146-

1158.

Gollub, H.F., & Reichardt, C.S. (1987). Taking account of time lags in causal models. Child

Development, 58, 80-92.

Gottman, J.M. (1983). How children become friends. Monographs of the Society for Research in

Child Development, 48 (Serial No. 201).

Gottman, J.M. (1986). Conversations with friends. Cambridge: Cambridge University Press.

Gottman, J.M. (1991). Chaos and regulated change in families: A metaphor for the study of transi-

tions. In P.A. Cowen & E.M. Hetherington (Eds.), Family transitions (pp. 247-272). Hills-

dale, NJ: Erlbaum. Hartup, W. W. (1989). Social relationships and their developmental significance. American Psycholo-

gist, 44, 120-126.

Hartup, W.W. (1992). Peer relations in early and middle childhood. In B.V. Hasselt & M. Hersen (Eds.), Handbook of social development: A lifespan perspective (pp. 257-281). New York:

Plenum.

Hinde, R.A., Titmus, G., Easton, D., & Tamplin, A. (1985). Incidence of “friendship” and behavior

toward strong associates versus nonassociates in preschoolers. Child Development, 56, 234-

245

Howes, C. (1983). Patterns of friendship. Child Development, 54, 1041-1053.

Howes, C. (1988). Peer interaction of young children. Monographs of the Sociery for Research in Child Development, 53 (No. 217).

Howes, C. (1990). Social status and friendship from kindergarten to third grade. Journal of Applied Developmental Psychology, I I ~ 32 I-330.

Kandel, D. B. (1978). Similarity in real life adolescent friendship pairs. Journal of Personality and Social Psychology, 36. 306-312.

Kenny, D.A., & LaVoie, L. (1984). The social relations model. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 18, pp. 141-181). New York: Academic.

Kinderman, T.A. (1993). Natural peer groups as contexts for individual development: The case of

children’s motivation in school. Developmental Psychology, 29, 970-977.

Ladd, G.W. (1983). Social networks of popular, average and rejected children in school settings.

Merrill-Palmer Quarterly, 29. 283-307.

Ladd, G.W., Price, J.M., & Hart, C.H. (1990). Preschoolers behavioral orientations and patterns of

PEER CHOICE 237

peer contact: Predictive of peer status? In S.R. Asher & J.D. Coie (Eds.), Peer rejection in

childhood (pp. 90- 118). Cambridge: Cambridge University Press.

LaGreca, A.M. (1993). Social skills training with children: Where do we go from here? Journal of Clinical Child Psychology, 22, 288-298.

Maccoby, E.E. (1988). Gender as a social category. Developmenrul Psychology, 24, 755-765.

Maccoby, E.E. (1990). Gender and relationships. American fsychologisr, 45, 513-520.

Maccoby, E.E. (1992). The role of parents in the socialization of children: A historical overview. Developmenral Psychology, 28. 1006-1017.

Masters, J.C., & Furman, W. (1981). Popularity, individual friendship selection, and specific peer interaction among children. Developmental Psychology, 17, 344-350.

McDowell, J.J. (1988). Matching theory in natural human environments. Behavior Analyst, I I, 95- 109.

McDowell, J.J. (1989). Two modem developments in matching theory Behavior Analyst, 12, 153- 166.

Olson, S.L. (1992). Development of conduct problems and peer rejection in preschool children: A social systems analysis. Journal of Abnormal Child Psychology, 20, 327-349.

Patterson, G.R. (1982). Coercive family process. Eugene, OR: Castilia.

Rodgers, J.L., Billy, J.O., & Urdry, J.R. (1984). A model of friendship similarity in mildly deviant behaviors. Journal of Applied Social Psychology, 14, 413-425.

Ross, H.S., & Lollis, S.P. (1989). Peer relationships. Child Developmenr, 60, 1082-1091.

Rubin, K.H., Lynch, D., Coplan, R., Rose-Krasnor, L., & Booth, C.L. (1994). Concordances and preferred personal attraction in children. Child Developmenr, 65, 1778-1785.

Schneider, B.H. (1989). Between developmental wisdom and children’s social-skills training. ln B.H. Schneider (Ed.), Social competence in developmental perspective (pp. 339-353). Bos- ton: Kluwer.

Singleton, L., & Asher, S. (1984). Racial integration and children’s peer preferences. ChildDevelop- ment, 50, 936-941.

Snyder, J.J. (1986). Peer interaction process coding system: Procedures, reliabiliry and validity. Unpublished manuscript, Wichita State University, Wichita, KS.

Snyder, J. (1995). Coercion: A two-level theory of antisocial behavior. In W. O’Donohue & L. Krasner (Eds.), Theories in behavior therapy. Washington, DC: American Psychological

Association.

Staddon, J. (1980). Limits to action. New York: Academic.

Warr, M., & Stafford, M. (1991). The influence of delinquent peers: What do they think or do? Criminology, 29, 85 l-866.

Youniss, J. (1986). Development in reciprocity through friendship. In C. Zahn-Waxler, E. Cum- mings, & R. Iannotti @is.), Altruism and aggression: Biological and social origins (pp. 88- 108). Cambridge: Cambridge University Press.