Positive and Negative Peer Interaction in 3- and 4-Year-Olds in Relation to Regulation and Dysregulation

ABSTRACT

Using a sample from the NICHD Study of Early Child Care (N = 435; 219 girls), the authors derived several measures of regulation and dysregulation that predicted, both concurrently and longitudinally, children's positive and negative peer interactions in multiple contexts. Observers rated peer interactions in child care and during dyadic play with a friend, and mothers rated peer behavior. The authors based the derived measures on resistance to temptation (36 months) and delay of gratification (54 months) tasks, as well as observations in child care of children's compliance and defiance with adults at both ages and maternal reports. Preschoolers who had better impulse control and who were more compliant and less defiant with adults engaged more often in friendly, positive, peer play and were less negative in their peer play across contexts. Associations between regulation and dysregulation and peer interaction were broader and more consistent at 54 months than at 36 months. Longitudinally, regulation at 36 months was only modestly associated with more positive and less negative peer play at 54 months. The authors discuss findings in the context of developing self-regulation and its importance for early peer relationships.

Keywords:

• dysregulation
• peer relations
• preschool
• regulation

The toddler and early preschool years represent a watershed in the development of competent peer interaction. During toddlerhood children commence regular, meaningful interactions with peers, and the skills and social knowledge acquired during these early encounters lay the foundation for many aspects of social competence (Brownell & Brown, 1992; Eckerman & Peterman, 2001; Rubin, Bukowski, & Parker, 1998). Peer interactions become increasingly complex over the preschool years, with greater communication, coordination, and turn-taking (Ashley & Tomasello, 1998; Holmes-Lonergan, 2003); more complex and reciprocal fantasy play (Howes, 1992); and more shared positive affect (Denham, McKinley, Couchoud, & Holt, 1990). Early peer play, however, is not entirely cooperative and positive. Preschoolers’ disputes over toys can escalate into verbal and physical aggression (O'Brien, Roy, Jacobs, Macaluso, & Peyton, 1999). Indeed, aggression is at its peak during this period (Campbell, 2002; Tremblay, 2000). Thus, early peer interactions are often highly emotional and arousing both positively and negatively. Young children's ability to regulate their emotions and behavior may therefore be critical for successful, competent interactions with peers during this period and may set the stage for later developing peer social competence. The purpose of this study was to investigate the concurrent and longitudinal associations between preschool children's self-regulation and dysregulation and their positive and negative peer interactions.

Self-Regulation and Preschool Peer Relations

Both negative and positive behavior are central components of young children's peer interactions. In some cases, early aggressive behavior remains stable and becomes associated with peer rejection and other peer problems in childhood (Campbell, Shaw, & Gilliom, 2000; Cummings, Iannotti, & Zahn-Waxler, 1989; Denham, Blair, Schmidt, & DeMulder, 2002; NICHD Early Child Care Research Network [ECCRN], 2004). On the other hand, positive peer behavior, which includes prosocial behavior, voluntary behavior that benefits others, enhances children's peer relationships beginning quite early. Preschool children who are more helpful and cooperative in their play with peers are more popular with their age mates, and their peers rate them as more likable (Denham et al.; Howes, Hamilton, & Phillipsen, 1998; Slaughter, Dennis, & Pritchard, 2002). Hence, beginning in preschool both negative and positive behaviors are associated with children's experiences and success in the peer group.

Among the many factors that contribute to individual differences in children's positive and negative peer interaction, several aspects of self-regulation appear important, including the ability to control emotional reactivity (Calkins, Gill, Johnson, & Smith, 1999; Denham et al., 2003; Eisenberg et al., 1993), moderate impulsive or inhibited social behavior (Calkins, Smith, Gill, & Johnson, 1998; Kochanska, Murray & Harlan, 2000; Rubin, Coplan, Fox, & Calkins, 1995), and attend to others’ emotions and behavior (Denham et al.; Raver, Blackburn, Bancroft, & Torp, 1999). Each of these components of self-regulation appears to be associated with children's peer social competence by early to middle-childhood (Eisenberg et al., 1996; Eisenberg et al., 2000; Fabes et al., 1999).

The preschool period may be particularly important and informative for examining relations between self-regulation and peer interaction. According to Kopp (1982, 1989) self-regulation undergoes significant development during the toddler and preschool years. As young children become more independent and goal-directed, their behavior and emotions become more autonomously regulated, and thus more adaptive, flexible, and socially appropriate across diverse contexts (Kochanska, Coy, & Murray, 2001; Kopp, 1982). During this period, peer interactions also increase dramatically in both frequency and complexity, thus placing new demands on children's growing self-regulatory abilities.

Central components of regulation during this period include the ability to inhibit impulses, regulate the emotional arousal associated with prohibitions on behavior, and activate behavior appropriate to the situation. Researchers conceptualize and measure these aspects of regulation in a number of ways. For example, some use delay of gratification or resistance to temptation tasks to assess children's ability to control emotional arousal, focus attention, and inhibit behavior (Campbell, Pierce, March, Ewing, & Szumowski, 1994; Mischel, Shoda, & Peake, 1988; Raver et al., 1999). Compliance and defiance in response to adults’ requests are also used to measure young children's self-control and self-regulation (Calkins et al., 1998; Crockenberg & Litman, 1990; Kochanska et al., 2001; Vaughn, Kopp, & Krakow, 1984) insofar as these depend on the ability to regulate both behavior and emotion (Eisenberg et al., 1995; Gralinski & Kopp, 1993).

Resisting temptation, delaying gratification, and complying with adults all demonstrate children's ability to manage their attention and arousal, inhibit inappropriate or impulsive responding, and generate appropriate behavior, including behavior that may run counter to the child's own immediate interests or desires. These skills likely contribute to children's growing ability to manage the demands of dynamic social exchange with others, and lay the groundwork for developing standards of behavior that children use in later peer interactions (Calkins, 2007), especially in generating positive, prosocial behavior with peers. In contrast, failure to control emotions or impulses and modify behavior according to others’ requests in this period is associated with later behavior problems (Calkins, 1994; Rubin, Burgess, Dwyer, & Hastings, 2003; Shaw, Bell, & Gilliom, 2000). For example, defiance reflects emotional, usually angry, behavioral opposition to adults’ requests or demands and may occur as a result of breakdowns in children's ability to manage their arousal and behavioral impulses (Crockenberg & Litman, 1990; Kuczynski, Kochanska, Radke-Yarrow, & Girnius-Brown, 1987). Children who lose control of their behavior and emotions when adults block their goals or desires may likewise find it difficult to manage the frustrations of peer encounters (Denham et al., 2002). In the present study, we used children's ability to resist temptation and to delay gratification, and their adherence to or defiance in response to their mothers’ and caregivers’ requests and prohibitions to index regulation and dysregulation.

Regulation Versus Dysregulation

Researchers often conceptualize children's self-regulation as varying from low to high along a continuum (for a review, see Eisenberg, Vaughan, & Hofer, 2009). However, some investigators have distinguished conceptually and empirically between regulation and dysregulation. Dysregulation has been defined as loss of behavioral and emotional control, especially in anger- or frustration-inducing contexts (Bridges, Denham, & Ganiban, 2004; Cole, Michel, & Teti, 1994; Denham et al., 2003; Dodge & Garber, 1991; Rubin et al., 2003; Shields & Cicchetti, 1997; Shipman et al., 2007). Shields and Cicchetti, for example, characterized regulation as “situationally appropriate” behavior and emotion, including “empathic toward others,” or “can say when s/he is feeling sad, angry or mad, fearful or afraid” (p. 910). They conceptualized dysregulation as lack of flexibility in responding, lability in responses, and dysregulated affect, including “wide mood swings,” and “prone to angry outbursts” (p. 910). Accordingly, it is possible for a child to be low on the dimension of regulation without necessarily being dysregulated. For example, a child who is relatively withdrawn and who exhibits little concern for peers would be low in regulation according to Shields and Cicchetti; this child may have difficulty generating situationally appropriate behavior in the context of peer play. But this child would not be considered dysregulated, in contrast to children who exhibited intense anger reactions, perhaps with aggression or tantrums, when peers disrupted or failed to accommodate to their play, even momentarily. Although dysregulated children are, by definition, low in regulation, the converse is not necessarily true—children who are low in regulation (e.g., low empathy) are not necessarily dysregulated (e.g., intensely, impulsively angry). Children high in dysregulation respond to many situations, sometimes even neutral ones, with atypical, disorganizing, volatile, and high-intensity emotional or behavioral responses that they may have difficulty controlling and getting over (Hessler & Katz, 2007; Keenan, 2000; Shields & Cicchetti). Correspondingly, children on the low end of the dysregulation dimension are not necessarily high in regulation; their behavioral or emotional responses are also unmodulated, but they may be weak, ineffective, and overly restricted (Keenan). Dysregulation and regulation of emotions also appear to have distinct etiologies (Shipman et al., 2007). Thus, dysregulated emotion and behavior appears to serve as a unique vulnerability, putting children at risk for maladaptive development (Campbell et al., 1994; Cole et al.; Dodge & Garber; Keenan; Rubin et al.).

Distinguishing between regulation and dysregulation may be important for predicting individual differences in peer social competence during the preschool years. Positive peer interactions likely depend on effective regulation because reactions to momentary interruptions of one's goals during peer play (e.g., during brief conflicts or disruptions) could overwhelm a child's incipient interaction skills and otherwise playful exchanges, turning them negative (Denham et al., 2002; Fabes et al., 1999; Halberstadt, Denham, & Dunsmore, 2001; Miller & Olson, 2000). Children who can regulate their arousal and behavior during emotionally intense interactions with peers should be able to respond more adaptively to their peers in such situations. Because prosocial skills such as empathy, cooperation, and sharing with peers are just emerging in the early preschool years (Farver & Branstetter, 1994; Hay, Castle, Davies, Demetriou, & Stimson, 1999; Zahn-Waxler, Radke-Yarrow, Wagner, & Chapman, 1992) and are likely to be fragile, young children's ability to regulate emotions and behavior appropriately may be especially important in permitting them to attend to others’ emotions and communications during peer interaction and to engage in complementary, reciprocal, and helpful exchanges (Brownell & Hazen, 1999; Eisenberg & Fabes, 1998). Whereas effective regulatory skills may promote positive peer interactions, dysregulation may be independently associated with peer aggression (Calkins et al., 1999; Eisenberg et al., 1995; Rubin et al., 2003; Rubin et al., 1995). Although most toddlers and preschoolers develop and master self-regulation in emotionally arousing contexts, for some young children situations with peers may regularly lead to disorganization and loss of behavioral and emotional control, resulting in negative, aggressive peer behavior. Such response patterns, if stable, could account in part for the emerging stability of peer aggression in the preschool years (NICHD ECCRN, 2004).

The empirical research to date offers suggestive evidence for these hypothesized relations between regulation and dysregulation and children's peer competence, but few studies have compared them in relation to developing peer interaction. Peer acceptance and popularity are associated with self-regulation in both preschool age and older children (Denham et al., 2003; Eisenberg et al., 1993; Hubbard & Coie, 1994; Maszk, Eisenberg, & Guthrie, 1999). Preschool children prefer to play with peers who use more effective attention regulation strategies, such as looking away from a prohibited object during a delay of gratification task (Raver et al., 1999), and the number of preschool children's mutual friendships relates to teachers’ ratings of their emotion regulation (Walden, Lemerise, & Smith, 1999). Elementary school children perceive their peers as more prosocial when parents and teachers rate them as having better regulation or attentional control (Eisenberg et al., 1996). Conversely, toddlers and preschoolers who are dysregulated, as indexed by poor frustration tolerance and anger management, are more likely to initiate conflict and aggression with peers (Calkins et al., 1999; Garner & Estep, 2001; Rubin et al., 1995; Rubin, Hastings, Chen, Stewart, & McNichol, 1998) and are rated by parents as higher in externalizing behavior and lower in social competence (Diener & Kim, 2004). Among slightly older children, impulsivity, poor inhibitory control, and emotion dysregulation such as angry venting are associated with parents’ reports of problematic social behavior in their children (Campbell et al., 1994; Denham et al.; Eisenberg et al., 1995; Eisenberg et al., 2000).

In the present study we investigated associations between both regulation and dysregulation and early peer interactions in the same children, thereby permitting us to both distinguish and compare these processes. We expected that as regulation becomes more autonomous, flexible, stable, and automatic in the preschool period, it would begin to play a more substantial role in children's peer relations, becoming more broadly and consistently associated with peer interaction between 36 and 54 months. Based on the previous conceptualizations and empirical findings, we hypothesized that regulation and dysregulation would be differentially associated with positive and negative peer interaction in preschool children. We also expected that associations between regulation or dysregulation and peer social competence would increase over the preschool years as children's regulatory skills develop. To test these hypotheses we examined concurrent associations between regulation or dysregulation and both positive and negative peer behavior at 36 and at 54 months of age, and we determined whether different patterns of association were evident earlier (36 months) than later (54 months) in the preschool years.

Longitudinal Relations Between Self-Regulation and Peer Competence

The literature suggests that children's self-regulation may play an important role not only in concurrent social competence at a given age, but in predicting later social behavior as well. Toddlers who were more compliant with their mothers were rated as having fewer behavior problems at age 5 years (Kuczynski & Kochanska, 1995), and toddlers who were more emotionally and behaviorally dysregulated exhibited more behavior problems at age four (Rubin et al., 2003) and were less socially competent as kindergarteners (Denham et al., 2002). Preschoolers who performed better in a delay of gratification task, that is, exhibited greater behavior regulation, were more academically and socially competent as adolescents (Mischel et al., 1988). Little research, however, has addressed relations between early self-regulation and children's social behavior with their peers across the preschool years, even though this is the first period of rapid developmental advances in both peer social competence and self-regulation. Therefore, in addition to examining concurrent relations between regulation and dysregulation and young children's peer social behavior, we also tested the hypothesis that early regulation and dysregulation at 36 months of age would predict later peer social competence at 54 months.

We obtained longitudinal data from children and mothers who participated in the NICHD Study of Early Child Care (SECC; NICHD ECCRN, 1996), a study of approximately 1,300 families from 10 locations around the United States. One of the strengths of the NICHD SECC sample is that many of the constructs were assessed in multiple contexts and at multiple ages using different informants and different, converging measures. The measures of children's self-regulation used in the present study were obtained in three different contexts at 36 and 54 months to provide complementary evidence of regulation and dysregulation. Likewise, negative and positive behavior with peers was assessed at 36 and 54 months by maternal report and by observations of play in two settings: child care and dyadic play with a friend.

We examined three specific hypotheses using these data. First, because positive peer behavior requires effective control and management of arousal, behavior, and emotions associated with early peer interaction, and because negative peer behavior may result from loss of regulatory control over emotion and behavior with correspondingly intense and inappropriate responses to frustrating social situations, we expected that regulation and dysregulation, respectively, would be differentially associated with positive and negative peer interaction in the preschool years. That is, we expected regulation to be uniquely associated with positive peer behavior, and dysregulation to be uniquely associated with negative peer behavior. Second, because developmental theory holds that regulatory skills become more autonomous, automatic, and consolidated over the preschool years, we expected that associations between self-regulation and peer social competence would become stronger and more consistent between 36 and 54 months of age. Third, we expected that early regulation and dysregulation at age three would predict later positive and negative behavior with peers at age 4.5 years.

The study extends previous work in several ways. First, the age studied was extended downward to young preschool children (36 months) in whom both self-regulation and peer competence are rapidly developing; they were then followed longitudinally to the more typical preschool age (4.5 years) previously studied with similar questions. Second, a larger and more diverse sample was available than in much previous research. Third, peer interaction was assessed in multiple contexts, as were regulation and dysregulation. Finally, regulation and dysregulation were empirically distinguished to examine whether they contribute in unique ways to the development of positive and negative peer behavior.

Method

Participants

Families participating in the NICHD SECC were recruited in 1991 from 10 urban and rural sites around the United States. When infants were 1 month old, 1,364 families with healthy newborns were enrolled (for complete description of the recruitment and sampling procedure, see NICHD ECCRN, 1996). The present study included 435 families from the larger sample, seen when the children were 36 and 54 months old (216 boys, 219 girls). Families were from all 10 sites and were included in the present sample if the child had regular child care (at least 10 hr per week) and if the child had participated in the dyadic peer play assessment at 36 months. Many families did not have regular child care or participate in the dyadic play interactions because of scheduling issues, and so forth. However, because child care and dyadic play assessment were the two contexts in which observations of peer interaction were made, only families with both measures from both contexts were included in the study. The families in the present study were diverse with respect to ethnic–racial background (18% belonged to an ethnic minority); were relatively educated (maternal education M = 14.95 years, SD = 2.49); were of moderate but variable income (income to needs ratio [i.e., family income divided by the appropriate poverty threshold for each household size], M = 3.27, SD = 2.82); and were predominantly two-parent families (90%).

Overview of Data Collection

Data for this report were collected when the child was 1 month of age through 54 months of age. At 1 month, family demographic information was gathered. When children were 36 months old compliance with their mothers and impulse control (resistance to temptation) were assessed in the laboratory. When children were 54 months old impulse control (delay of gratification) was assessed in the laboratory, and maternal ratings of children's compliant and defiant behavior were obtained. At both 36 and 54 months, behavior with peers was assessed via mothers’ reports, from observations in the children's child-care setting, and during dyadic play with a familiar, same-age playmate in a semistructured play session.

Measures

We used measures from the larger study for the present study. Subsequently we describe assessment procedures and specific measures. Details for all data collection, coding, and reliability procedures are documented in the NICHD SECC Manuals of Operation (http://public.rti.org/secc/).

Self-Regulation: Regulation and Dysregulation

Measures of regulation and dysregulation were obtained at 36 and 54 months of age in three different measurement contexts. As detailed subsequently, measures of compliance, resistance to temptation, and delay of gratification indexed regulation, whereas measures of defiance indexed dysregulation. The measures of compliance and defiance obtained in each context (lab, child care, home–maternal report) are described first, followed by measures of resistance to temptation and delay of gratification, both obtained in the laboratory (see Table 1). Each context makes slightly different self-regulation demands on the children; together these measures provide a broader picture of children's self-regulation competence than any one individually.

TABLE 1 Means and Standard Deviations for Measures of Regulation and Dysregulation and Positive and Negative Peer Behavior at 36 and 54 Months

	36 months		54 months
Variable	M	SD	M	SD
Regulation
Laboratory compliance: Cleanup compliance^a	8.57	2.77
Child care compliance: Child care compliance^b	0.59	0.59	0.81	0.21
Maternal report: Maternal compliance ratings			4.43	1.25
Resistance to temptation^c	125.39	56.38
Delay of Gratification^d			5.04	2.79
Dysregulation
Laboratory defiance: Cleanup defiance^a	2.31	2.31
Child care defiance: Child care defiance	0.08	0.19	0.03	0.06
Maternal report: Maternal defiance ratings			4.18	1.56
Positive peer behavior
Maternal Report: Positive peer sociability^e	24.16	2.50	13.04	2.45
Child care observation: Positive peer interaction	18.85	11.85	14.71	9.68
Dyadic play observation: Positive dyadic behavior	7.64	2.86	8.95	1.53
Negative peer behavior
Maternal Report: Negative peer sociability^e	5.76	1.30	2.12	0.99
Child care observation: Negative peer interaction	4.12	4.40	3.08	3.76
Dyadic play observation: Negative dyadic behavior	4.22	2.54	3.70	0.82
Note. Composite scores before standardization.
^a5-point global rating. ^bProportions. ^cTime (s). ^dTime (min).
^eAdaptive Social Behavior Inventory (A. Hogan, K. Scott, & C. Bauer, 1992) at 36 months; Social Skills Rating System (F. M. Gresham & S. N. Elliot, 1990) at 54 months.

Laboratory compliance and defiance at 36 months

During the 36-month laboratory visit children and mothers were videotaped during a 5-min cleanup period following freeplay. Fourteen children could not be scheduled for the laboratory session at 36 months. There were no significant demographic differences between the children with and without laboratory compliance. Observers blind to information about children and families coded the videotapes. Observers rated five behaviors on 5-point Likert-type global scales developed for the larger study ranging from 1 (not at all characteristic) to 5 (very characteristic): compliance (cooperates with mother's request), defiance (opposes requests), passive noncompliance (ignores a request), assertive noncompliance (says no), and negative affect. Intercoder reliability ratings were based on repeated measures analysis of variance (ANOVA) developed by Winer (1971, p. 287). Reliability was .93 for compliance, .91 for defiance, .86 for passive noncompliance, .86 for assertive noncompliance, and .89 for negative affect.

Two composite scores were created based on the results of a principal components analysis with Varimax rotation that included codes from the laboratory cleanup task and from observations of compliant and defiant behaviors in child care, described below (see Table 2). We created a Cleanup Compliance composite by standardizing and summing the passive noncompliance score (reversed), assertive noncompliance score (reversed), and the compliance score. We created a Cleanup Defiance composite by standardizing and summing negative affect and defiance.

TABLE 2 Factor Loadings for Compliance in Laboratory and Child Care at 36 Months

	Factor	Factor	Factor
	1 (Cleanup	2 (Cleanup	3 (Child care
Variable	compliance)	defiance)	defiance)
Laboratory clean up task
Passive noncompliance	.86	–.07	–.01
Assertive noncompliance	.71	.43	–.04
Compliance	–.86	–.38	.06
Negative affect	.18	.89	.06
Defiance	.13	.90	–.02
Child care observation
Defiance toward caregiver	.07	.01	.79
Autonomy behavior	–.01	–.03	.61
Compliance	.04	.01	–.17
Note. Significant factor loadings, based on J. P. Stevens's (1992) recommendations, are in bold.

Compliance and defiance in child care at 36 and 54 months

Coders observed children's compliant, noncompliant, and defiant behaviors with caregivers in their child-care settings at 36 and 54 months, using an observational instrument designed for the larger study (Observational Record of the Caregiving Environment, or ORCE; see NICHD ECCRN, 1996). Coders observed each child on two days for approximately 2.5 hr (total) at 36 months, and on one day for 2.5 hr at 54 months. Of the 435 children observed in child care at 36 months, 375 children were observed again at 54 months. Children observed in child care at 54 months had higher income to needs ratios than the children not observed (M = 4.41, SD = 4.04 vs. M = 2.65, SD = 1.76, respectively, p < .01).

Observers recorded the frequency of children's compliance (cooperates with adult's request), noncompliance (says no), and defiance (angry or aggressive expressions to adult's requests). Because coding was live and observers also recorded many other behaviors of children and their caregivers, this coding system was less detailed than for compliance observed in the lab. Intercoder reliability was established with a common set of videotapes that had been master coded by a committee of experts. Reliability ratings at 36 and 54 months respectively, based on repeated measures ANOVA developed by Winer (1971), were .84 and .92 for compliance, .86 and .85 for noncompliance, and .85 and .88 for defiance.

We created two composite scores at each age based on the results of a principal components analysis (see Table 2). Because they loaded together, the proportion of defiance toward the caregiver and the proportion of noncompliance were standardized and summed to create a summary score for Child Care Defiance at 36 months. The compliance score did not load on any factor so we analyzed it separately as Child Care Compliance at 36 months. At 54 months, defiance and noncompliance scores were also standardized and summed to create a composite of Child Care Defiance; again, Child Care Compliance was separately analyzed. Thus, there were parallel measures of compliance and defiance with caregivers at each age.

Maternal reports of compliant and defiant behavior at 54 months

Mothers rated children's compliant and defiant behavior at 54 months using the Social Skills Rating System (SSRS; Gresham & Elliot, 1990). We created two subscales. The Maternal Compliance Ratings subscale consisted of three items: follows your instructions, attends to your instructions, and follows household rules (Cronbach's α = .79). Possible scores ranged from 0 to 9 with higher scores indicating greater compliance. The Maternal Defiance Ratings subscale consisted of five items (e.g., controls temper in conflict situations with you [reversed], disobeys rules or requests, ends disagreements with you calmly [reversed]; Cronbach's α = .64). Possible scores ranged from 0 to 15 with higher scores indicating greater defiance. There were no significant demographic differences at 54 months between children with and without maternal ratings of compliance (n = 391, n = 44, respectively) and defiance (n = 386, n = 49, respectively) on the SSRS.

Resistance to temptation at 36 months

A forbidden toy task was used to measure the child's ability to resist temptation during the 36-month laboratory visit. After the child briefly played with a new and attractive toy, the experimenter told the child that she had work to do and that the child could play with the other toys in the room (from an earlier procedure), but could not touch the new toy until told it was okay. The experimenter placed the toy at arm's length from the child for 2.5 min, then did paperwork in a corner of the room while the child's mother filled out questionnaires. Resistance to temptation scores were available for 392 children. The remaining children (n = 43) either could not be scheduled for the laboratory session or did not experience the standard version of the task. They differed significantly from children with resistance to temptation scores on income to needs ratio (M = 4.56, SD = 3.82 vs. M = 3.25, SD = 2.70, respectively, p < .01).

Coders blind to child and family characteristics coded the videotapes of the children's behaviors at a central location. Latency to first active engagement and total active time engaged with the forbidden toy were recorded. Reliability estimates based on the repeated measures reliability formulation presented in Winer (1971) were .98 and .92, respectively. We created a Resistance to Temptation composite by standardizing and summing these two scores with active time reversed (Cronbach's α = .93).

Delay of gratification at 54 months

A waiting procedure (Mischel, 1974, 1981) was used to measure children's ability to delay gratification during the 54-month laboratory visit. An experimenter showed the child two plates of food, one with a smaller pile of food (M&Ms, animal crackers, or pretzels, according to the child's preference) and the other with a larger pile of the same food. The experimenter then told the children they could eat the larger pile if they waited until the experimenter returned, but that if they were unable to wait, they could ring a bell to summon the experimenter sooner. The experimenter then left the room and observed the children from an adjacent observation room for up to 7 min. The experimenter permitted the children who rang the bell before the end of the waiting period to eat the smaller pile of food. The single measure of Delay of Gratification at 54 months was the amount of time waited. There were no significant demographic differences between children with (n = 366) and without (n = 69) delay of gratification scores.

Because we have conceptualized regulation and dysregulation as different, albeit related constructs in this study, it is important to know to what degree the measures of regulation and dysregulation are related. The two constructs shared from 0–22% of variance within each measurement context, depending on the setting and the measure. This nil to moderate degree of association is consistent with the notion that the two constructs are not completely overlapping and do not simply represent the ends of a single continuum from low to high regulation.

Positive and Negative Behavior with Peers

We obtained three measures of negative peer interaction and three measures of positive peer interaction at both 36 and 54 months, corresponding to the three different measurement contexts. First, we describe measures obtained by maternal reports of the child's everyday behavior, followed by measures obtained in observed peer play in child care, and finally measures obtained in observed dyadic play with a friend (see Table 1).

Maternal reports at 36 and 54 months

At 36 months, mothers rated their children's social behavior using the Adaptive Social Behavior Inventory (ASBI; Hogan, Scott, & Bauer, 1992). Mothers rated 30 behaviors on a 3-point Likert type scale ranging from 1 (rarely) to 3 (almost always). We calculated two subscales, for positive and negative peer social behavior. The Positive Peer Sociability subscale consisted of 10 items (e.g., shares toys, cooperates, is helpful to other children; Cronbach's α = .74). Possible scores could range from 10 to 30 with higher scores indicating higher ratings of positive peer behavior. The Negative Peer Sociability subscale consisted of four items (e.g., teases others, bullies others; Cronbach's α = .57). Possible scores could range from 4 to 12 with higher scores indicating higher ratings of negative behavior with peers. There were no significant demographic differences between children with (n = 421) and without (n = 14) scores on the ASBI.

At 54 months, mothers rated their children's social behavior using the Social Skills Rating System (SSRS) using 3-point Likert type scale for frequency of occurrence ranging from 0 (never) to 2 (very often). The Positive Peer Sociability subscale consisted of 9 items (e.g., responds appropriately when hit or pushed by other children, follows rules when playing games with other children, joins group activities without being told; Cronbach's α = .70). Possible scores ranged from 0 to 27 with higher scores indicating more positive peer behavior. The Negative Peer Sociability subscale consisted of three items (aggressive toward people or objects, disturbs ongoing activities, argues with others; Cronbach's α = .49). Possible scores ranged from 0 to 9 with higher scores indicating more negative behavior. Note that alphas for negative peer sociability were low because there were few items and because any given child did not typically exhibit all of the individual items indexing negative peer behavior. We retained this composite because we wished to identify children who exhibited any form of negative peer behavior even if they did not exhibit all forms of negativity and aggression. That is, some children were aggressive whereas others were disruptive and annoying. Thus, a child with a higher score on the negative subscale might evidence this in just one way and different children could be negative with peers in different ways, but both are types of negative behaviors. SSRS Positive Peer Sociability scores were available for 394 children at 54 months and SSRS Negative Peer Sociability scores for 386 children. There were no significant demographic differences between children with and without maternal ratings of peer behavior on the SSRS.

Peer interaction in child care at 36 and 54 months

Children's positive and negative behaviors with peers in child care were observed at 36 and 54 months as part of the ORCE described previously, during successive 1-min time samples for three 10-min periods at each age. Observers also recorded qualitative ratings of the children's social behavior over four 44-min observation periods.

Positive or neutral peer interaction reflected how frequently children engaged other children in any type of nonnegative interaction (number of observation segments in which a peer was available and the child was engaged in positive or neutral interaction). Prosocial actions such as sharing that were initiated by the children were also coded. Positive or neutral interaction and prosocial actions were coded at both ages. At 36 months observers coded mutual pretend play reflecting how frequently children engaged in fantasy play or pretense with other children. At 36 months, observers also rated children for positive sociability, using a 4-point Likert type scale ranging from 1 (not at all characteristic) to 4 (highly characteristic), reflecting the child's interest in and positive engagement with other children. At 54 months observers coded cooperative play, reflecting older children's greater social skill, which indexed how frequently children engaged in organized, nondisruptive, mutual activity with peers and included but was not limited to fantasy play. Interobserver reliability ratings, based on a repeated measures ANOVA developed by Winer (1971), were .96 for positive or neutral interaction, .72 for prosocial actions, and .96 for mutual pretend play. Interobserver reliability at 54 months was .94 for positive or neutral interactions, .75 for prosocial actions, and .82 for cooperative play. We created a composite of Positive Peer Interaction at 36 months by summing standardized scores for positive or neutral interaction, mutual pretend play, prosocial actions, and positive sociability. We created a composite of Positive Peer Interaction at 54 months by summing standardized scores for positive or neutral interaction, prosocial actions, and cooperative play.

At both ages observers coded negative acts such as initiating an unfriendly, nonaggressive act (e.g., taking a toy from another child), verbal aggression such as teasing or name-calling, and physical aggression such as pushing. Negative interaction reflected how frequently children engaged in any other type of negative interaction with peers. Interrater reliability estimates, based on a repeated measures ANOVA developed by Winer (1971), were .94 and .90 for negative interactions, .90 and .83 for negative actions, .70 and .91 for verbal aggression, and .75 and .91 for physical aggression at 36 and 54 month observations, respectively. We summed these behaviors to create a composite of Negative Peer Interaction.

Dyadic play with friend at 36 and 54 months

At 36 and 54 months, children participated in a semistructured play session with a same-sex peer close in age and identified by the child's parent or child-care provider as a close friend or regular playmate. The procedures were identical at 36 and 54 months, except for the toys that were used. Interactions were videotaped and coded at a central location and were independent from the ORCE observations. Dyadic play sessions occurred inside a portable cardboard playroom (91.44 cm high and 152.45 cm in diameter) to provide a standardized play environment; it was set up in the children's regular child-care setting or in one of the children's homes. Experimenters videotaped children's behavior through a curtained opening into the playroom as they presented three different age-appropriate toys to the children in fixed order for 5 min each. The toys were chosen to permit a range of positive and negative peer interaction and play behavior. Children observed in peer play (n = 327 at 54 months) had higher income to needs ratios than the children not observed (n = 108; M = 4.45, SD = 4.21, vs. M = 3.59, SD = 2.52, respectively; p < .01). Coders blind to child care, family history, and other measures of children's functioning coded the videotapes at a central location. Behaviors were averaged across the three toy episodes.

At 36 months amount of positive social interaction and cooperation (sharing, turn-taking and playing together) were coded on 5-point Likert-type scales ranging from 1 (low) to 5 (very high). Concern (expressions of caring and concern for peer) was coded on a 3-point Likert-type scale ranging from 1 (none) to 3 (high). Intercoder reliability ratings based on a repeated measures ANOVA (Winer, 1971) were .95 for positive interaction, .96 for cooperation, and .94 for concern. In addition, Response to provocation: Shares was scored as the number of toy sessions (0–3) in which a child attempted to resolve provocation from the peer by sharing a contested toy; Social problem solving: Property rights was scored as the number of toy sessions (0–3) in which the child tried to gain possession of the peer's toy using an approach other than physical force or verbal demands (e.g., negotiation). We standardized and summed these five scores to form the composite variable Positive dyadic behavior at 36 months.

Coders also rated hostile and instrumental aggression on 5-point Likert type scales ranging from 1 (low) to 5 (high) at 36 months. Intercoder reliability ratings based on a repeated measures ANOVA (Winer, 1971) were .94 for hostile aggression and .97 for instrumental aggression. In addition, Response to provocation: Takes was scored as the number of toy sessions (0–3) in which the child attempted to resolve provocation from the peer by taking away a contested toy; Social problem solving: Negative power assertion was scored as the number of toy sessions (0–3) in which the child tried to gain possession of the peer's toys using physical force or verbal demands. We standardized and summed these four scores to create the composite Negative dyadic behavior at 36 months.

At 54 months the positive peer behaviors during dyadic play included prosocial behavior (e.g., turn-taking and sharing behaviors), positive mood (e.g., happiness and enthusiasm), and contribution to positive interaction (actively contributes to coordinated interaction), all rated on 5-point Likert-type scales ranging from 1 (low) to 5 (very high). Intercoder reliability (Winer, 1971) was .72 for prosocial behavior, .82 for positive mood, and .86 for contribution to positive interactions. We standardized and summed these three ratings to form the composite Positive dyadic behavior at 54 months.

Coders rated aggression, negative mood (e.g., unhappiness or negative affect), and contribution to negative interaction (e.g., expresses negativity in interactions) on 3-point Likert-type scales ranging from 1 (none) to 3 (high) during the 54-month dyadic play session. Intercoder reliability based on a repeated measures ANOVA (Winer, 1971) was .80 for aggression, .89 for negative mood, and .81 for contribution to negative interaction. We standardized and summed these three behaviors to create the composite Negative dyadic behavior at 54 months.

Demographic Characteristics

Preliminary analyses showed that maternal education and income were related to some of the measures so these two demographic variables were combined and included as a covariate in analyses. Maternal education was the number of years of education mothers had completed at the time of the child's birth. Total family income was obtained from mother report when the children were 36 (n = 432) and 54 months (n = 397) old. These were standardized and summed to create a Demographic Composite at 36 and 54 months, with higher scores indicating greater education and family income.

In summary, there were three measures of regulatory abilities at 36 and 54 months corresponding to different contexts in which self-regulation is demanded (compliance with mothers; compliance with caregivers in child care; resistance to temptation at 36 and delay of gratification at 54 months), and two measures of dysregulation at 36 and 54 months (defiance with mothers; defiance with caregivers in child care). There were three outcome measures for positive peer behavior at both ages, also corresponding to different contexts (maternal ratings; child care observations; dyadic play observations) in which the expression of peer interaction skills may differ and three outcome measures of negative peer behavior from the same contexts (maternal ratings; child care observations; dyadic play observations). Table 1 presents the means and standard deviations of each measure. Table 3 presents the zero-order correlations among the measures of positive and negative peer behavior at both ages.

TABLE 3 Correlations Between Positive and Negative Peer Behavior Across Three Measurement Contexts at 36 and 54 months

Variable	1	2	3	4	5	6	7	8	9	10	11	12
36 months
Maternal report
1. Positive peer sociability		.20***	.12*	.01	.12*	–.07	.46***	–.12*	. 03	.01	.12*	.01
2. Negative peer sociability			.10	.01	.05	.08	–.24***	.32***	.05	.02	–.01	.14**
Child care
3. Positive peer interaction				.28***	.08	.01	.04	–.06	.11*	.16**	.08	.01
4. Negative peer interaction					.01	.11*	–.02	.10	.03	.27***	.08	.07
Dyadic play
5. Positive dyadic behavior						–.22***	.13*	–.01	.07	.00	.09	–.01
6. Negative dyadic behavior							–.06	.09	–.08	–.05	.08	–.14*
54 months
Maternal report								–.26**	.09	–.04	.10	–.01
7. Positive peer sociability
8. Negative peer sociability									–.02	–.01	–.02	.07
Child care
9. Positive peer interaction										.27***	.13	–.03
10. Negative peer interaction											–.05	.20***
Dyadic play
11. Positive dyadic behavior												–.20***
12. Negative dyadic behavior
*p < .05. **p < .01. ***p < .001.

Results

We conducted analyses to identify associations between concurrent measures of regulation and dysregulation and peer social behavior at 36 and 54 months, and to test longitudinal relations between regulation and dysregulation at 36 months and peer social behavior at 54 months. Predictive analyses were conducted with missing data (less than 8%) imputed using an expectation–maximization (EM) algorithm to compute maximum likelihood (ML) estimates based on all of the available data in the present data set. ¹

Results are presented in three sections. First, we conducted hierarchical regression analyses to determine whether measures of regulation and dysregulation at 36 months predicted children's concurrent positive behavior and negative behavior with peers. Second, we conducted a parallel set of regressions to examine whether measures of regulation and dysregulation at 54 months predicted children's concurrent behavior with peers. Third, we conducted hierarchical regression analyses to examine longitudinal associations between measures of regulation and dysregulation at 36 months and children's behavior with peers at 54 months.

We entered the demographics composite first as a covariate in all analyses. Because there were several sex differences at both ages (Table 4), gender was entered second as a covariate in all analyses (separate analyses by gender yielded essentially identical patterns of association). When predicting positive behavior with peers, we entered measures of regulation as a block after the control variables, followed by measures of dysregulation. We reversed the order of these two blocks when predicting negative behavior with peers. This strategy reflects the hypothesis that regulation would predict positive peer behavior whereas dysregulation would be more strongly associated with negative peer interaction. However, because we did not want to rule out the possibility that regulation might be negatively associated with negative peer behavior and that dysregulation might be negatively associated with positive peer behavior, so we included both regulation and dysregulation in each analysis. Note that this strategy is relatively conservative insofar as regulation and dysregulation share variance. ²

TABLE 4 Gender Differences in Regulation and Dysregulation and Peer Behavior at 36 and 54 Months

	Girls (n = 219)		Boys (n = 216)
Variable	M	SD	M	SD	t
36 months
Cleanup compliance	0.25	2.48	–0.25	2.59	2.02*
Resistance to temptation	0.31	1.78	–0.32	2.03	3.26**
Positive peer sociability (Maternal report)	24.61	2.51	23.70	2.41	3.79***
Positive dyadic behavior (Dyadic play)	0.42	3.24	–0.42	2.84	2.88**
54 months
Maternal defiance ratings	4.02	1.71	4.37	1.34	2.20*
Positive peer sociability (Maternal report)	13.34	2.48	12.69	2.39	2.64**
Negative peer interaction (Child care)	–.17	2.53	0.92	3.36	3.35***
df = 433. *p < .05. **p < .01. ***p < .001.

Predicting Peer Behavior at 36 Months From Concurrent Regulation and Dysregulation

We conducted six hierarchical regression analyses to predict the three measures of children's concurrent positive behavior and the three measures of negative behavior with peers. Tables 5 and 6 present the findings for the regression analyses at 36 months, including beta weights. Both regulation and dysregulation predicted peer social behavior, especially in the child-care context.

TABLE 5 Predictive Analyses of Positive Peer Behavior at 36 Months From Concurrent Regulation and Dysregulation

	Maternal report: Positive peer sociability		Child care: Positive peer interaction		Dyadic play: Positive dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	.06	.01	–.06	.01	.05	.00
Gender	.16**	.04***	.02	.01	.13*	.02**
Regulation		.05		.03*		.02
Cleanup compliance	.08		.11*		.05
Child care compliance	.03		.10*		.03
Resistance to temptation	.08		.03		.01
Dysregulation		.06		.05*		.03
Cleanup defiance	–.01		.05		.02
Child care defiance	–.03		–.13**		–.02
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 3.54**, 2.96**, and 1.56, respectively. *p < .05. **p < .01. ***p < .001.

TABLE 6 Predictive Analyses of Negative Peer Behavior at 36 Months From Concurrent Regulation and Dysregulation

	Maternal report: Negative peer sociability		Child care: Negative peer interaction		Dyadic play: Negative dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	–.12	.02**	-.13**	.02**	–.09	.01
Gender	.07	.02	.02	.02	.08	.01
Dysregulation		.02		.06***		.02
Cleanup defiance	–.05		.12*		.04
Child Care defiance	–.01		.20***		.03
Regulation		.07***		.09**		.02
Cleanup compliance	.01		.04		.02
Child care compliance	.04		.08		.01
Resistance to temptation	–.21***		–.17***		–.08
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 4.23***, 5.98***, and 1.35, respectively.
*p < .05. **p < .01. ***p < .001.

Positive peer behavior

As hypothesized, regulation predicted positive peer interaction at 36 months. Compliance with mothers during lab cleanup and compliance with caregivers in child care both predicted more positive interaction with peers in child care, β = .11, t(427) = 1.99, p < .05; β = .10, t(427) = 2.08, p < .05, respectively. After controlling for regulation, one measure of dysregulation, child care defiance, was also associated (negatively) with positive peer interaction in child care, β = –.13, t(427) = 2.75, p < .01. Girls also received higher ratings of peer competence from mothers than did boys, β = .16, t(427) = 3.22, p < .01. Thus, 3-year-old children who were more compliant and less defiant with adults were observed to engage in more positive, cooperative exchanges with their peers in child care.

Negative peer behavior

As hypothesized, dysregulation predicted negative behavior with peers at 36 months. Both defiance with mothers during cleanup and defiance with caregivers in child care were associated with negative peer interaction in child care, β = .12, t(427) = 2.04, p < .05; β = .20, t(427) = 4.16, p < .001, respectively. After controlling for dysregulation, one measure of regulation also predicted negative peer behavior in two of the three interaction contexts. Specifically, children's resistance to temptation in the lab was negatively associated with negative peer interaction observed in child care, β = –.17, t(427) = 3.56, p < .001, and with maternal ratings of negative peer sociability, β = –.21, t(427) = 4.24, p < .001. Thus, 36-month-old children who were more defiant with adults and less able to inhibit behavior in the lab were also more negative with peers both in child care and at home.

Predicting Peer Behavior at 54 Months From Concurrent Regulation and Dysregulation

We conducted a parallel set of six regression analyses at 54 months to predict children's concurrent positive behavior and negative behavior with peers. Tables 7 and 8 present the findings, including beta weights. Generally, regulation and dysregulation predicted children's positive and negative peer behavior more systematically across interaction contexts at 54 months than at 36 months.

TABLE 7 Predictive Analyses of Positive Peer Behavior at 54 Months From Concurrent Regulation and Dysregulation

	Maternal report: Positive peer sociability		Child care: Positive peer interaction		Dyadic play: Positive dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	.02	.02**	.04	.00	.05	.00
Gender	.08*	.04*	–.04	.00	.13*	.02
Regulation		.30***		.03**		.04
Cleanup compliance	.41***		.00		.00
Child care compliance	–.04		–.18***		.02
Resistance to temptation	.09*		.05		.15**
Dysregulation		.34***		.07***		.05*
Cleanup defiance	–.22***		.01		–.03
Child care defiance	–.02		–.19**		–.10*
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 31.17**, 4.34***, and 3.16**, respectively.
*p < .05. **p < .01. ***p < .001.

TABLE 8 Predictive Analyses of Negative Peer Behavior at 54 Months From Concurrent Regulation and Dysregulation

	Maternal report: Negative peer sociability		Child care: Negative peer interaction		Dyadic play: Negative dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	.02	.01*	–.04	.01	–.13**	.03*
Gender	–.04	.02	–.15**	.04***	.02	.03
Dysregulation		.20***		.05		.05
Maternal defiance rating	.35***		.07		–.06
Child care defiance	.10*		.04		.12*
Regulation		.22*		.06		.11***
Maternal compliance rating	–.14**		–.01		–.20**
Child care compliance	.07		–.10*		–.14**
Delay of gratification	–.05		–.05		–.04
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 16.82***, 3.97**, and 7.55***, respectively.
*p < .05. **p < .01. ***p < .001.

Positive peer behavior

As hypothesized, regulation at 54 months predicted positive peer behavior in all three peer-interaction contexts. Delay of gratification in the lab predicted positive peer behavior during dyadic play with a friend, β = .15, t(427) = 2.96, p < .01, and maternal ratings of positive peer sociability, β = .09, t(427) = 2.09, p < .05. Compliance with mothers predicted positive peer sociability as rated by mothers, β = .41, t(427) = 8.87, p < .001, whereas compliance with caregivers in child care was negatively associated with positive peer interaction observed in child care, β = –.18, t(427) = 3.70, p < .001.

After controlling for regulation, the two measures of dysregulation at 54 months also predicted children's positive peer behavior in all three contexts. Children who were more defiant with adults in child care were observed to be less positive with peers both in child care and in dyadic play with a friend, β = –.19, t(427) = –3.95, p < .01; β = –.10, t(427) = –2.13, p < .05, respectively. And children who were rated by mothers as more defiant at 54 months were also rated as less positive with peers, β = –.22, t(427) = –4.83, p < .001.

Thus, 54-month-old children with better regulation as indexed by delay of gratification and compliance with mothers were more often involved in cooperative, playful interactions with their friends both in dyadic play and at home. But children who were more compliant with caregivers in child care were less engaged in positive peer play in child care. Above and beyond regulation, children who were more defiant with adults were less positive with peers in dyadic play, in child care, and at home.

Negative peer behavior

As hypothesized, dysregulation predicted negative peer behavior, and this was true for two of the three interaction contexts. Specifically, defiance with caregivers in child care at 54 months predicted negative dyadic behavior during play with a friend, β = .12, t(427) = 2.53, p < .05, as well as negative peer sociability rated by mothers, β = .10, t(427) = 2.32, p < .05. Defiance with mothers also predicted maternal ratings of negative peer sociability, β = .35, t(427) = 7.07, p < .001. Boys were observed to be more frequently involved in negative peer encounters in child care than were girls, β = –.15, t(427) = –3.22, p < .01.

After controlling for dysregulation, measures of regulation at 54 months also predicted negative peer behavior in all three interaction contexts. Children who were more compliant with caregivers in child care were observed to be less negative with peers in child care, β = –.10, t(427) = –1.97, p < .05, and independently observed to be less negative in dyadic play with a friend, β = .14, t(427) = –3.02, p < .01. Children whose mothers rated them as more compliant were less negative with their friends in dyadic play, β = –.20, t(427) = –3.81, p < .01, and were rated by their mothers as less negative with peers, β = –.14, t(427) = –2.77, p < .01.

Thus, 54-month-old children who were more dysregulated as reflected in defiance with adults were also more negative with their friends in both dyadic play and in peer interactions at home as rated by mothers. Above and beyond dysregulation, children who were more compliant with adults engaged in less negative interaction with their friends during dyadic play and at home.

Predicting Peer Behavior at 54 Months From Regulation and Dysregulation at 36 Months

We conducted hierarchical regression analyses to examine longitudinal associations between measures of regulation and dysregulation at 36 months and children's behavior with peers at 54 months, controlling for gender and 54-month demographic characteristics. Analyses were similar to the concurrent analyses at 36 and 54 months. Thirty-six month peer behavior was not controlled in these analyses because we were not predicting change in peer behavior between 36 and 54 months. Tables 9 and 10 present the findings, including beta weights. Individual differences in regulation at 36 months, but not dysregulation, predicted both positive and negative peer behavior at 54 months.

TABLE 9 Predictive Analyses of 54-Month-Year-Old Positive Peer Behavior From Regulation and Dysregulation at 36 Months

	Maternal report: Positive peer sociability		Child care: Positive peer interaction		Dyadic play: Positive dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	.14**	.02	.02	.00	.07	.01
Gender	.10*	.04**	–.04	.00	–.04	.01
Regulation		.05*		.02		.03*
Cleanup compliance	.03		.10		.16**
Child care compliance	.05		.09		.00
Resistance to temptation	.10*		.06		.02
Dysregulation		.05		.03		.04
Cleanup defiance	–.02		.09		.07
Child care defiance	–.02		.05		.07
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 3.40**, 1.62, and 2.25*, respectively.
*p < .05. **p < .01.

TABLE 10 Predictive Analyses of 54-Month-Year-Old Negative Peer Behavior From Regulation and Dysregulation at 36 Months

	Maternal report: Negative peer sociability		Child care: Negative peer interaction		Dyadic play: Negative dyadic behavior
Variable	B	R ^2	B	R ^2	B	R ^2
Demographic	–.08	.02*	–.07	.01	–.16**	.03***
Gender	–.06	.04	–.17***	.04***	–.01	.03
Dysregulation		.05		.04		.04
Cleanup defiance	–.03		.04		.06
Child care defiance	–.01		–.02		–.03
Regulation		.05**		.04		.06
Cleanup compliance	–.03		.03		–.08
Child care compliance	.01		.05		–.11*
Resistance to temptation	–.18***		–.02		–.01
Note. Significance levels for R ^2 are based on the F ratio associated with the R ^2 change. R ^2 statistics refer to the cumulative variance associated with each step. Betas reported are after all variables are entered into the regression and significance levels are based on t tests. For maternal report, child care, and dyadic play, F(7, 434) = 3.48**, 2.69*, and 3.76**, respectively.
*p < .05. **p < .01. ***p < .001.

Positive peer behavior

Resistance to temptation in the laboratory at 36 months predicted more positive peer sociability at 54 months as rated by mothers, β = .10, t(427) = 2.10, p < .05. Greater compliance with mothers during the lab cleanup at 36 months predicted more positive interaction observed during dyadic play with a friend at 54 months, β = .16, t(427) = 3.01, p < .01. After controlling for regulation, no measure of dysregulation at 36 months predicted positive peer behavior at 54 months of age.

Negative peer behavior

No measure of dysregulation at 36 months predicted negative peer behavior at 54 months. However, after controlling for dysregulation, two measures of regulation were associated (negatively) with negative peer behavior. Greater ability to resist temptation at 36 months predicted less negative peer sociability at 54 months as rated by mothers, β = –.18, t(427) = –3.05, p < .001. Compliance with caregivers in child care at 36 months predicted less negative peer interaction observed in child care at 54 months, β = –.11, t(427) = –2.24, p < .05. In addition, 54-month-old boys were observed to be more frequently involved in negative peer encounters than were girls, β = –.17, t(427) = –3.55, p < .001. Although dysregulation predicted peer social competence concurrently at both 36 and 54 months, 36-month dysregulation was not found to be associated with later peer interaction at 54 months.

Discussion

Consistent with the primary hypotheses of this study, we found that regulation and dysregulation in preschool children were associated with their ability to manage the emotional and behavioral challenges of peer interaction. Three- and 4-year-old children who were more regulated in their interactions with their mothers and caregivers and who demonstrated better impulse control on laboratory tasks were more positive and less negative with their peers and friends across multiple interaction contexts. Also as hypothesized, these associations were more consistent at 4.5 years of age than at 3 years. Thus, the results from the present study suggest that young preschool-aged children who are better at managing arousal, affect, and behavior in response to requests, demands, and prohibitions by adults are also better at managing the positive and negative arousal that accompanies peer play both in group settings and in dyadic interaction. In contrast, having poor impulse control and being noncompliant with adults, particularly if coupled with overt dysregulation such as anger and defiance, may carry social costs in the peer group beginning quite early.

Concurrent Associations at 36 and 54 Months

Three-year-old children who were more compliant and less defiant with adults or who were better able to resist temptation were more positive and less negative with peers during play, particularly when observed in child care. In contrast, 4-year-old chilren's compliance and defiance with adults (with one exception) and their ability to delay gratification were associated with peer interaction in all three of the settings in which we collected peer interaction data—in child care, during dyadic play with a friend, and in the home according to mothers’ reports. It is notable that self-regulation observed in one setting was associated with peer interactions observed in different settings. Thus, neither method variance nor context-specific factors account for these associations. Because peer interaction at 54 months was more systematically associated with regulation and dysregulation than it was at 36 months of age, the results are consistent with Kopp's (1982, 1989) position that more active, more autonomous, and more mature strategies for regulating emotion and behavior emerge over the preschool period and confirm our hypothesis that self-regulation comes to play a broader role in children's peer relations over the preschool period.

According to Kopp (1982), during the fourth year of life, self-regulation comes to replace self-control, an earlier precursor marked by limited ability to act flexibly in varying situations. When children are able to regulate their behavior with adults more flexibly in response to varying situational demands, they can also more consistently modulate their behavior in dynamic, emotionally arousing, and sometimes unpredictable peer interactions. Kochanska has also noted that self-regulation in interaction with adults has implications for the young child's ability to manage interactions with other social partners (Kochanska et al., 2000). The one puzzling finding in the present study was that children who were more compliant with adults in child care at 54 months were less often engaged in positive peer play. Although we can only speculate as to the source of this unexpected relation, one possibility is that for some 4-year-old children high rates of compliance with the caregiver may represent dependency, which we know from prior research is associated with shyness and social withdrawal from peers (Coplan & Prakash, 2003; Howes, Hamilton, & Matheson, 1994). Overall, the pattern of results in the present study suggests that emotion and behavior regulation begin to play an important role in children's peer relationships as early as age 3 years, during the period when social interactions with peers become more frequent, complex, and reciprocal. Children's ability to regulate their emotions and behaviors across varying social situations, including the unique demands of peer relationships, probably continues to develop through early childhood (Cole et al., 1994).

Longitudinal Relations: 36 Months to 54 Months

Regulation at 36 months predicted both positive and negative peer behavior at 54 months, although it was not as broad a predictor as were concurrent measures of regulation and dysregulation at each age. Interestingly, although earlier regulation was predictive of later peer interaction, earlier dysregulation was not. Children who were better at delaying gratification and regulating their arousal and behavior in response to adults’ requests or commands at age three were better at managing the arousal that accompanies peer play at age 4 years. Thus, the ability to manage emotions and behavior in early childhood, when both self-regulatory abilities and peer relations are just beginning to blossom, forecasts the quality of developing social relationships. This could be a result of emerging continuity in early personality factors such as agreeableness, or the emerging ability to manage peer conflict more gracefully, or a reciprocal relationship between children's regulatory abilities and peer relations such that more positive peer relations at age 3 years helps children better manage their emotions and behavior which, in turn, feeds forward to more positive peer interaction at age 4 years (Eisenberg et al., 2009).

In contrast, individual differences in 3-year-old children's dysregulation in response to adults did not predict greater negativity (or less positivity) with peers as did that of 4-year-old children. It was somewhat surprising that dysregulation at age 3 years did not predict peer social competence at age 4 years, given the concurrent relations at both ages between dysregulation and peer interaction, as well as previous reports of longitudinal relations between dysregulation at age 2 years and externalizing behavior at age 5 years (Rubin et al., 2003). One possibility is that dysregulation may be expressed differently at different ages or may take different forms, such that it reflects unique, age-specific breakdowns in the ability to manage emotions and behavior; these may not be relevant to peer interaction at every age. For example, temper tantrums may still be prevalent for some children at age 3 years, thus be part of the dysregulation construct at that age. As a reflection of unmodulated emotions and behavior, they would be expected to predict negative peer interaction concurrently. But temper tantrums at age 3 years would not be expected to continue to predict poor peer relations at age 4.5 unless they were continuous with some other index of dysregulation. One potential direction for future researchers is to clarify specific developmental continuities and discontinuities in dysregulation over the toddler and preschool years.

Although each measure of self-regulation assessed in the present study was related to peer social behavior concurrently or longitudinally, impulse control (resistance to temptation at 36 months; delay of gratification at 54 months) was the most stable across age (see also Kochanska, 2002; Kochanska et al., 2000), and was the most widely associated with positive and negative peer interaction. We recognize that these results are correlational in nature, thus directional inferences are necessarily limited. However, taken together with Mischel's (Mischel et al., 1988) impressive longitudinal findings of relations between preschool impulse control and adolescent social competence, the results from the present study suggest that impulse control may be an early-emerging and relatively stable form of regulation, and may be particularly important in managing the demands and challenges of early peer interaction. Thus, impulse control may play a formative role in children's early developing peer competence.

Regulation Versus Dysregulation

Building on prior research (e.g., Keenan, 2000; Shields & Cicchetti, 1997; Shipman et al., 2007), we conceptualized regulation and dysregulation as distinct but related constructs, and the resulting operationalization did yield only modestly related measures. We believe that this is the first study to compare regulation and dysregulation directly in relation to early peer interaction. In predictive models both regulation and dysregulation contributed uniquely to individual differences in peer social competence after controlling for the other. These findings amplify those from a recent report that negative peer relations among disruptive preschool children appear to derive more from problems in behavioral regulation than from immaturities in social understanding (Hughes, White, Sharpen, & Dunn, 2000). Perhaps children's ability to modulate their behavior, attention, and arousal during the preschool years takes precedence in managing their peer relations when their understanding of others’ emotions and mental states is undergoing profound change and may still be fragile and unstable.

Several limitations of the present study should be noted. First, we defined dysregulation in terms of negative behavior only; a child could have dysregulated positive behavior, such as unrestrained, disruptive exuberance, but this was not measured. Second, we did not address dynamic processes of regulation, for example, what strategies and coping procedures children used to regulate their behavior, attention, and emotion. The measures available from the NICHD SECC do not afford these kinds of fine-grained analyses. Instead, as Dodge and Garber (1991) and Cole, Martin, and Dennis (2004) advised, we assessed interdomain associations, that is, how regulation or dysregulation in one or more domains (behavior, attention, emotion) related to the theoretically distinct domain of peer social interaction. Third, because the power available with a sample of this size permits the detection of small effects, some of the significant associations accounted for somewhat modest amounts of variance. Of course, a multitude of other factors also contribute to the development of peer social competence, including child, parent, cultural, and situational factors. It would be interesting for future researchers to consider how and whether regulation and dysregulation add to, amplify, or weaken other important influences on children's early peer relations. Finally, although the sample for the NICHD SECC is relatively diverse, it is possible that the present findings would be either stronger or weaker in a sample characterized by greater educational or economic risk. These limitations are offset in part by the strengths of the NICHD SECC data set, which include a large sample and the availability of multiple, repeated measures of regulation and dysregulation in different contexts, as well as multiple, repeated measures of peer social competence in several settings, using several converging measurement methods.

In sum, together with other recent research the present findings establish that emotion and behavior regulation in the early preschool years, albeit still immature, are nevertheless important components of effective social interaction with peers. Although the individual effects are small, the potential cascade of other possible outcomes as a function of early-appearing differences in peer social competence is worth noting. Perhaps helping children learn to regulate behavior and emotion more effectively during this formative period when self-regulation is differentiating and consolidating could promote more socially competent interactions with peers and reduce the likelihood of later social skills deficits. Intervention research could also be used to better understand the causal and directional relationships between developing self-regulatory abilities and developing social competence.

AUTHOR NOTES

Geetha B. Ramani is an assistant professor of human development at the University of Maryland. Her research focuses on how social interactions with peers and parents influence young children's cognitive development. Specifically, she is interested in children's learning through cooperative play and activities, as well as the development of peer cooperation. Celia A. Brownell is a professor of developmental psychology at the University of Pittsburgh, where she studies the origins, early development, and sequelae of young children's social and social-cognitive competence. Specifically, her research focuses on cooperative peer relations, prosocial behavior, and body self-awareness, all as they emerge and develop in the first three years of life. Susan B. Campbell is a professor of clinical and developmental psychology at the University of Pittsburgh. Her research focuses on the social and emotional development in children with an emphasis on parent-child relationships and family risk factors, as well as the early emergence of behavior problems in young children.

Notes

1. The EM approach is recommended as a more appropriate procedure for handling missing data over traditional approaches, such as list-wise deletion and mean substitution, which can strongly bias estimates and lead to erroneous conclusions (Acock, 2005). The EM approach involves a single imputation in contrast to multiple imputation; therefore, the standard errors of the EM approach may be smaller because multiple imputation can incorporate imputation uncertainty into standard errors. However, the EM approach was used in the present study because it can provide relatively unbiased estimation of missing data in large samples (Schafer & Graham, 2002). Analyses conducted with the valid data and with the missing data imputed yielded similar results. The results are presented from the full sample with missing data imputations.

2. We also conducted the predictive analyses by entering both measures of regulation and dysregulation together as a block in the regression as an alternative approach to assess unique associations of each with peer behavior. These analyses yielded essentially identical results to those reported.