Peer Effects on Head Start Children’s Preschool Competency

Abstract

The goals of the present study were to investigate whether young children attending Head Start (N=292; M_age=4.3 years) selected peers based on their preschool competency and whether children’s levels of preschool competency were influenced by their peers’ levels of preschool competency. Children’s peer interaction partners were intensively observed several times a week over one academic year. Social network analyses revealed that children selected peer interaction partners with similar levels of preschool competency and were influenced over time by their partners’ levels of preschool competency. These effects held even after controlling for several child (e.g., sex and language) and family factors (e.g., financial strain and parent education). Implications for promoting preschool competency among Head Start children are discussed.

In the US, one of the primary interventions designed to help young economically underprivileged children become better prepared for formal schooling is Head Start (Gilford, 2013). Many of the children attending Head Start programs in the US start school at a considerable disadvantage compared to their more economically advantaged peers. Despite this general pattern, there is considerable variation among Head Start children. For example, in a recent Head Start Family and Child Experiences Survey (FACES) report (Klein, Aikens, West, Lekashanets, & Tarullo, 2013), there was significant variability in nonverbal intelligence both across and within Head Start racial/ethnic groups. Similar variability has been reported on social-emotional characteristics (Aikens, Moiduddin, Xue, Tarullo, & West, 2012) and emergent literacy skills (Cabell, Justice, Konold, & McGinty, 2011). Furthermore, there is evidence that Head Start children’s emergent literacy skills change between the fall and spring of the preschool year (Cabell, Justice, Logan, & Konold, 2013), especially for the most at-risk children (Zill & Resnick, 2006). Thus, in a typical Head Start classroom, each child is surrounded by a classroom of peers who represent a wide and changing continuum of social-emotional and academic skill competencies that may, over time, affect her or his own levels of preschool competency (Reid & Ready, 2013).

The goal of the current study was to better understand how preschoolers’ school-related competencies are associated with the peer environment in Head Start classrooms. Specifically, we assess the efficacy of a “peer social exposure” model in which time spent with peers provides opportunities to become more similar to peers through a variety of peer influence processes. Peer influence may arise from observing peers during play and academic tasks and being concurrently rewarded or punished for imitating these behaviors (Bandura, 1977). Peer influence may also arise, in accordance with social motivation theory, through social experiences that motivate both social (peer) and academic (school) engagement or disengagement in the classroom (Juvonen & Wentzel, 1996). To empirically examine how preschool competencies are related to the peer context, we measured changes in Head Start children’s preschool competency – operationalized as the behaviors that underlie, and are predictive of, academic-related skills -- over the course of the school year. Although changes in preschool competency may be accounted for by many different factors, we focus on peer influence as a mechanism for change. Specifically, we contend that young children actively select certain (similar) peers to interact with and that, in the context of peer interaction, there is an opportunity for peer influence to occur. Thus, interactions with some peers may promote, while interactions with other peers may hinder, the development of preschool competency.

In previous empirical research, social network characteristics of the classroom have been used to predict academic and social adjustment both in the short- (e.g., across the academic year) and long- (e.g., from early childhood into adolescence) term (Ahn, Rodkin, Gest, 2013; Garandeau, Ahn, & Rodkin, 2011; Ialongo, Werthamer, Kellam, Brown, Wang & Lin, 1999; Ialongo, Poduska, Werthamer, & Kellam, 2001; Serdiouk, Rodkin, Madill, Logis, & Gest, 2013). Researchers have also used social network theory to address social differences in the school that may impact academic engagement and performance (Cappella, Kim, Neal, & Jackson, 2013; Neal, Cappella, Wagner, & Atkins, 2011). Nevertheless, when scholars use network features as linear predictors of student outcomes, rather than accounting for the socially embedded characteristics of the classroom, information from various levels of analysis is included in the same linear equation. It follows that, by using network features as linear predictors, there is an important loss of classroom variability that may exist between individuals, within social groups, and ultimately within the social network of the classroom (Cascio & Schanzenbach, 2007; Hedges, Laine, Greenwald, 1994). Linear equations also require decisions to be made on how to transform the data to best represent the social composition of the classroom, a decision that may also bias results (Yudron, Jones, & Raver, 2014). Finally, a failure to analyze complete social network data, or the presence of missing data, can also bias the estimates of peer effects in the classroom (Sojourner, 2013). Taken together, these finding suggest that to adequately test peer influence requires a consideration of a complete social network coupled with all individual behaviors in the classroom to account for the individual and group dynamics that lead to social affinities (selection) and behavioral change (influence). In the current study, complete social network and behavioral data were used to examine peer influence within Head Start classrooms.

Young Children’s Preschool Competency

Few would dispute the fact that children vary in their readiness to learn because some children have behaviors that help them adjust to school environments and others do not. In the present study, we aim to consider predictors of the variability in preschool competency, a critically important, but often understudied, feature of children’s readiness to learn and success in school (Mendez, Fantuzzo, & Cicchetti, 2002; Murphy, Laurie-Rose, Brinkman, & McNamara, 2007). Preschool competency encompasses the social and learning behaviors that are predictive of young children’s concurrent and longer-term academic-related successes (Blair, 2002; Kaiser, Hancock, Cai, Foster & Hester, 2000; Mashburn, Justice, Downer, & Pianta, 2009; McWayne & Cheung, 2009). This includes successful interactions with peers, focused attention, active participation, and classroom engagement (e.g., Kagan, Moore, Bredekamp, 1995; Raver, 2004). Children who exhibit preschool competencies actively participate in class, attend to teachers, and have good peer relationships. These behaviors are interrelated and positively associated with school readiness, a predictor of later academic success (Coolahan, Fantuzzo, Mendez, & McDermott, 2000; Cuhna & Heckman, 2007). Moreover, these non-cognitive competencies are considered to represent important drivers of academic perseverance and academic behaviors that are essential for long-term academic success (Farrington et al., 2012).

However, intervention efforts to promote school readiness typically focus on enhancing children’s cognitive and literacy skills, and little attention is given to how the social ecology of the classroom might be supported in order to build a broader set of school-related competencies. Overlooking the social functioning of peers in a classroom may be a critical oversight (Fabes, Martin, Hanish, Anders, & Madden-Derdich, 2003). We aim to address this concern by directly examining peer influence, or the extent to which children change their own levels of preschool competency to become more similar to the peers with whom they interact most in Head Start classrooms. To be precise, in order to assess the complete social ecology of the classroom; we will analyze peer influence at the level of the classroom peer network rather than at the level of a subset of relationships within the classroom.

Most research, to date, on peer influence processes has focused on peers’ influence on learning-related behaviors among older children and adolescents (Altermatt & Pomerantz, 2003; Fonzi, Schneider, Tani, & Tomada, 1997; Kindermann 2007; Slavin, 1983; Wentzel & Watkins, 2002). The lack of attention to how peers potentially influence younger children’s school success can be credited, in part, to an assumption that older children and adolescents are more susceptible to peer influence than younger children (Hartup, 1989). However, several studies suggest that this presumption is inaccurate. There is evidence that peers impact young children’s externalizing behaviors and language skills (Hanish et al 2005; Martin & Fabes, 2001; Mashburn et al., 2009), suggesting peer influence may occur in younger children. Other early childhood research has examined peer collaborations (Hogan & Tudge, 1999), determining that expert-novice interchanges are mutually beneficial because the less skilled peer learns from the more skilled peer and the more skilled peer’s knowledge is strengthened from teaching the less skilled peer (Azmitia, 1988). Achievement abilities of peers are also positively associated with preschool children’s own achievement level (Mashburn, Justice, Downer, & Pianta, 2009; Justice, Petscher, Schatschneider, & Mashburn, 2011). Martin et al. (2013) provided the only direct test of peer selection and peer influence in early childhood using the complete, longitudinal network data we use in the current study and found evidence of peer influence on gender typed activities, after accounting for peer selection effects. The present study extends this previous work by exploring whether preschool peers can also exert influence on school-related outcomes, namely preschool competency, when interacting in unstructured, freely-chosen, peer settings that are typical of the preschool classroom.

Peer Selection and Influence in Preschool

Within the typical preschool classroom, children have ample opportunities to choose with whom they would like to interact. In these free-play peer settings, we may find that two children who spend time together are similar to one another on certain features (e.g., preschool competencies). This similarity may, in part, result from peer influence in the classroom. However, similarity between frequent interaction partners may also arise from selection effects. If peer selection is left unaccounted for in examinations of peer influence, the estimates of peer influence can be drastically inflated (Kandel, 1978). Therefore, it is incumbent that researchers account for peer selection effects before they can draw conclusions about peer influence.

A child’s selection of a similar peer as an interaction partner may also contribute to the influence process. For example, having freely selected similar peer interaction partners based on preschool competency, there is an opportunity for influence to occur toward increased similarity on the same school-related competency that fostered the initial formation of the peer relationship. Thus, the initial process of shopping for interaction partners may have a significant long-term impact on child outcomes as children continue to influence and change one another’s school-related behaviors. Children may choose peers based on some initial similarities in preschool competency and these chosen peer affiliates may then influence children toward even greater similarity to chosen peers on preschool competency over time.

The Present Study

In the present study, we posed two research questions. The first research question focuses on whether children select their peers based on being similar in preschool competency. Although only a few studies of young children’s peer selection have been conducted (e.g., Farver, 1996; Martin et al., 2013; Santos, Vaughn, Bonnet, 2000; Strayer & Santos, 1996), these studies have shown that even young children select peers based on certain dimensions of similarity (e.g., sex, activity similarity). Thus, there is some evidence to suggest that preschool children are able to detect similarities to peers and use these similarities to select peer interaction partners. It is, however, yet to be determined if children are sensitive to the behavioral cues that might allow preschool children to select peers based on similarity in preschool competency.

The second research question focuses on whether children are influenced by peers such that they become more similar to their peers in preschool competency over time. Several studies suggest that peer influence occurs in preschool (Haun & Tomasello, 2011; Martin et al., 2013), and, so, there is reason to believe that preschool children will be susceptible to peer influence effects. Preschool children have many opportunities to socially engage each other. Thus, children may grow more similar over time as a consequence of the mutual influence that occurs during these social interactions. The classroom context may also make peer behaviors related to preschool competency (e.g. paying attention to the teacher) particularly salient.

In the present study, we used data obtained from a study of children enrolled in Head Start classrooms. The yearlong longitudinal data provides opportunities to explore peer selection and influence effects in young children at a time when peer relationships are beginning to form in the preschool classroom. Understanding the influence of peers in Head Start children’s early preschool competencies, over and above the effects of gender, age, language spoken, financial strain, or cognitive ability, may help educators create classroom environments that enhance the social features of the classroom that will, in turn, promote optimal classroom engagement and student performance in the preschool classroom.

Method

Participants

Participants were preschool-aged children enrolled in 18 Head Start classrooms. Classroom size ranged from 12 to 20 students in 3- to 5-year-old Head Start classrooms. Unique to the current study was sufficient individual variability within these 18 Head Start classrooms to model the effects of important individual differences in the classrooms such as child sex, age, primary language spoken, family financial strain, parent education, and receptive vocabulary.

The children were involved in a cross-sequential longitudinal study in which they were observed intensively for a year during preschool, and then followed for 2 additional years. In the current study, we only investigate peer influence during the preschool year. Children were sampled in three cohorts over the first 3 years of the research project. Children were recruited for participation 2–3 weeks into the start of the academic school year. Consent rates were 99% at recruitment (N=308 of a possible 311); 16 participants left the participating Head Start programs and were not included in these analyses due to a low number of observations (fewer than 20 observations). There were no demographic differences greater than chance in children’s rate of permission or attrition.

Of the 292 children who were part of the present study (boys: n=156, M_age = 51.82 months, SD=5.02, range=38–59; girls: n=136, M_age =51.04 months, SD=5.61, range=37–60), 16 (50% boys) repeated preschool during years 2 and 3 of data collection and were retained in analyses to ensure that we had the complete networks in each classroom. This gave us an effective sample size of N=308. The ratio of girls to boys per classroom ranged from 31% to 62% girls, and 85% of the teachers were women. The majority of children (69%) were Mexican or Mexican American; 60% of the children primarily spoke Spanish. The remaining children were Anglo-American (8%), African American (7%), Asian (2%), Native American (1%), and other or unknown (13%). Consistent with Head Start programs, children were from families of low socioeconomic status (82% earned below $30,000 per year). Almost half of the children (45%) came from two-parent married families.

Procedures

The data for this study were obtained from a larger longitudinal study of Head Start children’s early school adjustment and performance (see Martin et al., 2013). In this study, children were observed across the school year and teacher reports of children’s competencies were obtained at multiple time points.

The data consisted of observations of children’s interactions (used to create the peer interaction network structure), teachers’ reports of children’s preschool competencies (outcome variable), assessment of initial receptive vocabulary (covariate), and demographic information on children and their families (covariates). Data were collected across the preschool academic year (approximately September to May) for the three preschool cohorts. Each year, observational data were divided into 4 waves across the preschool year; wave 1 was the first half of the fall semester, wave 2 was the second half of the fall semester, wave 3 was the first half of the spring semester, and wave 4 was the second half of the spring semester. During Wave 1, all types of data (i.e., observational and teacher data and covariates) were collected. In Waves 2, 3, and 4, only observational and teacher data were collected.

Observational data

Observational data were collected using a brief-observation protocol (e.g., Martin & Fabes, 2001; Martin et al., 2013) in which children were observed indoors and outdoors during free play (e.g., where children freely decide what to play, with whom, and where to do it) in 10-s periods, multiple times a day, 2–3 times a week for several hours a day over the fall and spring semesters. Classroom observers (8–10 per year; 87% female) were intensively trained for the first 3–4 weeks of each semester. Training of observational coders consisted of several meetings to discuss the coding scheme, practice coding sessions, and testing to ensure coders knew the names of the children and all of the codes. For each day of coding, observers began at the top of a randomized list of children (the list was reordered midway through each semester to prevent order effects), completed the entire list, waited 5 minutes, and then returned to the top of the list again (on average about 4 times per day). Coders were instructed to complete a full rotation before ending their day of observing. Prior to recording data, observers noted whether the child was present and available for coding, present but unavailable for coding (e.g., in the bathroom), or absent. If present and available for coding, the observer would then record the child’s identification number and the identification number of any peer interaction partners (up to five). Additional codes (such as the child’s primary activity) that are not relevant to the purpose of the present study were also obtained. To be coded as engaged with a peer, the focal child and the peer had to either engage in a verbal or physical exchange during the 10-s observation or be playing in the same activity alongside another child. Teacher-child interactions were not included in the present analyses.

A total of 38,145 observations were collected for the children in this sample (M=123.55 observations per child, SD=58.95, range = 21–303; 89% with more than 100 observations). As previously noted, 16 children with fewer than 20 observations were dropped. The number of observations recorded for each child varied due to differences in attendance and availability during coding. To control for the varying number of observations and presence in the classroom, we calculated classroom presence: the number of times a child was coded as present divided by the total number of observations attempted for that child.

To determine reliability, two observers independently coded the same children’s behaviors for approximately 1 hour per week. Reliability assessments varied across coders to prevent bias in the time of day or activities for which reliabilities were conducted. For identification of peer partners, percent agreement across semesters ranged from .87 to .97.

Measures

Network measures

For each wave, the observational data on children’s interactions had to be transformed into classroom networks that represent which children “shared a tie” (i.e., were interactional partners). To do this, we first calculated the numbers of times each target child interacted with peers, providing a proxy for young children’s preferences in interaction partners (Baines & Blatchford, 2009; Schaefer, Light, Fabes, Hanish, & Martin, 2010). Nearly all children interacted with every other child in their classroom, so only when children interacted more often than expected by chance (when compared to any other peer interaction partner in the classroom) were their interactions assumed to represent an underlying relationship (i.e., a tie) (see Schaefer et al., 2010 for additional details on tie creation). As such, ties across children can be asymmetrical. However, because these ties represent interactions between two children, our final step was to remove directionality from these peer interactions so that all interactions were coded as mutual (i.e., a symmetrized network) (Santos, Vaughn, & Bost, 2008). Following these procedures, network ties were calculated separately within each of the 4 waves of data. It was this calculation of network ties that allowed for a test of selection (peer partner choice) as a function of preschool competency and influence (peer partners’ effect on behavioral change) on preschool competency.

Preschool competency

Teachers completed questionnaires for each participating child in their classroom at four waves throughout the school year—two in the fall semester and two in the spring (corresponding to the middle of the fall semester, end of the fall semester, middle of the spring semester, and end of the spring semester). Teachers were financially compensated for their participation. From these teacher ratings, a 7-item scale of preschool competency was calculated by averaging items together. Items reflected the early non-cognitive foundations of academic performance identified by Farrington and colleagues (e.g., 2012) and included: “This child enjoys school and actively participates in activities and games,” “This child is skilled, capable, and effective in interactions with other children,” “This child follows classroom rules and complies with teacher requests,” “This child can focus his/her attention when he or she needs to,” “This child adapts well to change in class schedules and activities,” “This child is not disruptive in class,” “This child has strong academic skills”. Items were rated on a 5-point scale (1 = “Not at all true”; 5 = “Very true”). Internal reliability for the preschool competence scale was acceptable (α=.79 to .90 across waves). In addition, all items were significantly correlated (p <.05) with one another. Furthermore, each of the 7-items loaded on a common factor solution at each of the four measurement points. When using an exploratory factor analysis (EFA) there was no evidence that the current scale could be reliably broken down into any subscales. Furthermore, using a confirmatory factor analysis (CFA), all items hung together reliably in a single factor solution with acceptable model fit (CFA=0.99; RMSEA=0.05). The EFA and CFA were assessed in two ways. First, using the total sample (n=308 children) and then using one half of the sample (n=154 children) for the EFA and the other half of the sample (n=154 children) for the CFA. The pattern of results and model fit were similar using both approaches.

Additional support for the use of this measure is provided by evidence of external validity with other measures of academic behaviors and performance. Specifically, preschool competency correlated moderately and positively with the Peabody Picture Vocabulary Test-III (r=.23, p<.001) (PPVT-III; Dunn & Dunn, 1997; TVIP; Dunn, Padilla, Lugo, & Dunn, 1986) and the Woodcock-Johnson Tests of Achievement III (WJ-III) (word identification: r=.22, p<.01; applied problems: r=.31, p<.001) (Woodcock, McGrew, & Mather, 2000); and teacher reports of school-related skills completed at the end of the school year (subscales include social development: r=.50, p<.001; school specific instrumental: r=.54, p<.001; reading and writing: r=.41, p<.001; logic and numbers r=.38, p<.001, p<.001).

Another way to assess the reliability and validity of the preschool competency measure was to explore if the measure predicted change in academic competence upon entry into formal schooling. Hierarchical regression models revealed that the 7-item preschool competency scale predicted change in word identification and applied problem solving skills in kindergarten and 1^st grade as measured by the Woodcock-Johnson Tests of Achievement III (WJ-III; Woodcock, McGrew, & Mather, 2000). Specifically, higher levels of preschool competency were associated with greater increases in applied problem solving skills from preschool to kindergarten (β=.17, p<.01) and from preschool to first grade (β=.14, p<.05); higher levels of preschool competency also were marginally associated with greater increases in word identification from preschool to kindergarten (β=.12, p=.09) and from preschool to first grade (β=.12, p=.10). Together, these findings provide initial support for the reliability and validity of our relatively short teacher-reported measure of children’s preschool competency.

Control Variables

To control for potential confounds, we included a number of variables in the social network analysis to account for individual characteristics that may affect peer selection and influence. The control variables included child factors and family factors. Child factors included: sex, age, primary language spoken, receptive vocabulary, percentage of time present in the classroom, repeating a year in the classroom, and the effect of preschool competency on a child’s number of peer interaction partners. Family factors included: family financial strain and parent education. To assess receptive vocabulary, each participating child was individually administered either the Peabody Picture Vocabulary Test-III (PPVT-III; Dunn & Dunn, 1997) in English (n = 126), or the Test de Vocabulario en Imagenes Peabody (TVIP; Dunn, Padilla, Lugo, & Dunn, 1986) in Spanish (n = 127). The PPVT-III and the TVIP are standardized measures of receptive vocabulary for young children. The language chosen for administration (Spanish or English) was based on teachers’ recommendations. Children were administered the tests by trained bilingual research assistants (RAs). A parent-report measure of family financial strain was completed early in the fall of the preschool year on a 7-item scale of Family Economic Pressure (adapted from Conger & Elder, 1994) (e.g., “How difficult is it to pay for housing? food? clothing? household items? car? medical care? and recreation?”), rated on a 5-point scale (1 = “No trouble paying”; 5 = “A lot of trouble paying”). Internal reliability was good (α=.87) (Conger, Ge, Elder, Lorenz, & Simons, 1994). We included both network and behavioral parameters to control for these effects (described later in more detail).

Plan of Analysis

Longitudinal social network analyses were conducted using RSiena (Ripley, Snijders, & Preciado, 2012) to simultaneously examine young children’s peer relationships and changes in their preschool competency over time. There are many advantages of using the RSiena program and, because of this, it is gaining popularity as a statistical method useful for studying peer relationship dynamics (see Veenstra, Dijkstra, Steglich, & Van Zalk, 2013 for a review).

Several steps had to be taken to analyze the current observational data from Head Start classrooms. First, all 18 classroom networks were estimated simultaneously as a “meta-network” with the constraint that ties could only be between children in the same classroom (i.e., structural zeros prevented cross-classroom ties). By preventing cross-classroom ties we were accounting for the fact that children were nested within 18 classrooms and that the classroom teachers may affect each classroom differently. Thus, although we could not directly estimate the variance accounted for by teacher effects at the classroom level, the nested nature of the data determined that all significance tests were first estimated between classrooms and then combined into a single predictive model to estimate general trends across classrooms (i.e., the estimates reported in the current study). A similar approach was used by Martin et al. (2013).

A critical advantage of RSiena is that it makes use of longitudinal data in order to disentangle peer selection from peer influence. To be precise, RSiena requires behavioral variables that are measured longitudinally and at the same point in time as information about children’s social networks. Thus, in the current study, we used teachers’ assessments of their students’ preschool competency (behavioral variable) measured simultaneously with observations of preschool peer interactions (network variable). In the presence of this type of longitudinal data, RSiena simultaneously considers two functions: one to estimate change in the network (peer selection) and the second to estimate change in behavior (peer influence). Selection effects model the likelihood of a tie, or a peer interaction, to form based on similarity in specific attributes (e.g., selection based on similarity in preschool competency) and influence effects model the likelihood of a behavior to change based on the behaviors of frequent peer interaction partners (e.g., peer influence on preschool competency). The consequent estimates that emerge represent the most probable change at any given moment in time as a function of the current relationship structure of the social network and the behavioral composition of the social network. It is by using longitudinal data to simultaneously account for how relationship change in response to behaviors and how behaviors change in response to relationships that RSiena accounts for the interdependency, or nonindependence, of peer relationship data. As a final step, RSiena provides an omnibus estimate of peer selection and peer influence within the classroom.

A final benefit of the RSiena program is its ability to account for a series of structural, network, and behavioral control parameters in the estimation of selection and influence effects. This methodological innovation allows selection and influence estimates to be modeled without bias from the structural composition (i.e., structure of relationships and behaviors) of the peer network. Below we describe each of these control parameters in more detail. First, we outline the unique network and behavioral controls included in the current model. Next, we outline our effects of primary interest, namely selection based on preschool competency similarity and peer influence on preschool competency.

Structural controls

In keeping with recommendations from other RSiena studies (e.g., Martin et al., 2013), we controlled for the degree parameter, which accounts for the degree to which children are selective in peer interactions. In general, a positive degree parameter would indicate a lack of selectivity. Here we expect the degree parameter to be negative because children are generally selective in their choice of peer interaction partners (Steglich, Snijders, & Pearson, 2010). Second, we controlled for transitive triads, which accounts for the tendency for relationships to form when individuals share a common relationship partner, that is, the idea that friend of friends may be more likely to become friends (Davis, 1970). Because a significant negative degree parameter has been previously documented among preschoolers and over 40% of the ties in preschool networks are determined to be transitive (e.g., Martin et al., 2013; Schaefer et al., 2010), these were important controls to also include in our model. Next we estimated linear tendency and quadratic shape, both of which inform us about the behavioral distribution of preschool competency in these classrooms and prevent an overestimation of influence effects. Specifically, the linear tendency parameter indicates if there is a tendency for children to stay above or below the midpoint on levels of preschool competency. The quadratic shape parameter informs us if there is regression to the mean, or over-dispersion, on levels of preschool competency. The need to account for linear and quadratic tendencies has long been recommended as best practice with applied multiple regression models (Cohen, Cohen, West, & Aiken, 2003) and are considered necessary controls when modeling non-dichotomous outcome variables, such as the preschool competency measure, when using RSiena (Ripley, Snijders, & Preciado, 2012).

Network controls

Both similarity (e.g. sex similarity) and difference (e.g., sex difference) effects were included as network controls for sex, age, language, family financial strain, parent education, receptive vocabulary, percentage of time in the classroom, and repeating a year in the classroom. The similarity parameter describes the tendency for a child to have similar values as their peer interaction partners on that particular attribute (e.g., a positive effect for sex similarity means that children select same-sex interaction partners (e.g., boys are more likely to select boys and girls are more likely to select girls); a negative effect for sex similarity means that children select other-sex interaction partners (e.g., boys are more likely to select girls and girls are more likely to select boys). The difference parameter describes the number of interaction partners for a child with a particular attribute (e.g. sex). Because we coded girls as 2 and boys as 1, a positive sex difference effect means that girls have more interaction partners than boys; a negative sex difference effect means that boys have more interaction partners than girls. Additionally, we included a parameter that indicated the effect of preschool competency on a child’s number of peer interaction partners. Because selection on the basis of preschool competency was of primary interest in the current analysis, only the difference effect for preschool competency was included as a control parameter.

Behavioral controls

Behavioral controls were included for each of the effects of sex, age, language, family financial strain, parent education, receptive vocabulary, percentage of time in the classroom, and repeating a year in the classroom. These behavioral controls were included to account for the effect of each of these individual characteristics (e.g., sex) on changes in preschool competency behaviors across time. As an example, we included a sex effect that was coded so that a positive sex effect would describe the tendency for girls to change their preschool competency behavior more than boys, whereas a negative sex effect would describe the tendency for boys to change their preschool competency behavior more than girls.

After accounting for each of these control parameters in the RSiena model, we were next able to estimate our effects of primary interest, peer selection and influence, free from the bias of each of the estimated structural, network, and behavioral control effects. Below we describe how we estimated selection and influence effects related to preschool competency.

Selection effect

The selection based on preschool competency similarity parameter reflects the extent to which children select interaction partners based on similarity in preschool competency. A positive selection based on preschool competency similarity parameter indicates that children chose their peer interaction partners based on similarities in preschool competency. A negative selection based on preschool competency similarity parameter indicates that children chose their peer interaction partners based on dissimilarities in preschool competency.

Influence effect

The peer influence on preschool competency parameter (i.e. the average alter effect in RSiena) reflects the extent to which interaction partners influence one another on preschool competency. A positive peer influence on preschool competency parameter indicates that children change their level of preschool competency to resemble that of their peer interaction partners. A negative peer influence on preschool competency parameter indicates that children change their level of preschool competency so that they do not resemble that of their peer interaction partners.

To illustrate the magnitude of all reported effects, odds ratios are included for all effects. For these ratios, a higher ratio indicates a stronger effect. For example, odds ratios for peer selection indicate the likelihood of children to select peer interaction partners who are similar to themselves in levels of preschool competency compared to the likelihood of selecting peer interaction partners who are dissimilar in levels of preschool competency; odds ratio for peer influence indicate the likelihood of moving toward the preschool competency of peer interaction partners compared to the likelihood of moving toward the preschool competency levels of children with whom a child does not regularly interact.

As a final step, all effects of primary interest were estimated using an item level analysis to be certain that the same pattern of statistically significant results holds for each individual item on the preschool competency scale.

Results

Descriptive information on the network structure and preschool competency behaviors in these 18 Head Start classrooms is presented first. Next, peer selection for similarities in preschool competency is described. Finally, we describe peer influence on preschool competency. When describing peer selection and influence effects, we first describe the findings for control parameters followed by the findings of primary interest.

Descriptive Data

Table 1 provides descriptive statistics for the network structure and preschool competency behaviors. Degree, which reflects the number of peers children interact with, ranged from 5.16 to 5.62 across the four waves and demonstrates that, on average, children interacted with 5–6 peer interaction partners in their classroom at levels beyond chance. The total number of peer interactions across all 18 classrooms ranged from 1590 to 1732 across the four waves. The mean level of preschool competency increased from 3.98 to 4.07 (1 SD increase) across the four waves of data collection. The Moran’s I network autocorrelation indicates the degree to which frequent peer interaction partners resemble each other on preschool competency behaviors (Veenstra & Steglich, 2011). In the present study, Moran’s I ranged from .14 to .36 and indicates increases across the academic year in the similarity of preschool peer interaction partners in preschool competency. No more than 20 children (6% of the sample) joined or left the 18 classrooms between any data collection period. The Jaccard index provides an indication of the relative stability of the peer relationships, or ties (Snijders, van de Bunt, & Steglich, 2010). The Jaccard index ranged from .37 to .44 meaning that 37 to 44% of the ties observed at one time point also occurred, or were stable, at the next time point. Across four waves, approximately 20% of children increased their levels of preschool competency and approximately 18% of children decreased their levels of preschool competency. Estimates of behavioral stability in preschool competency ranged from a temporal autocorrelation of r=.70 to r=.78 across waves.

Table 1.

Network Descriptive Data

	Wave 1	Wave 2	Wave 3	Wave 4
Degree (per child)	5.47	5.62	5.43	5.16
Number of network ties	1686	1732	1671	1590
Mean levels of preschool competency	3.98	4.00	3.99	4.07
Moran’s I	0.14	0.24	0.36	0.32
		Wave 1 to 2	Wave 2 to 3	Wave 3 to 4
Change in network membership
Number of children leaving		8	20	12
Number of children joining		20	4	0
Change in network ties
Jaccard index		0.37	0.35	0.44
Change in behavior
Percentage of children who increased preschool competency		20.5%	18.2%	22.2%
Percentage of children who decreased preschool competency		22.3%	18.2%	13.0%
Stability: preschool competency (temporal autocorrelation)		r=.70	r=.78	r=.73

Selection on the Basis of Preschool Competency

Table 2 summarizes results of the network analyses and contains the unstandardized RSiena estimates and the standard errors of the estimates. The first major research question addressed in the present study is whether Head Start children select peer interaction partners based on similarities in levels of preschool competency. Before turning to these findings, we first review the findings for the control parameters in the network analyses.

Table 2.

Peer Selection and Socialization on the Basis of Preschool Competency

Network Parameters	Mean Parameter
	Est.	SE	p
Structural control effects
Degree (density of interaction partners in the social network)	−0.37	0.05	.000
Transitive triads (tendency to interact with the interaction partners of your interaction partners)	0.08	0.01	.000
Network control effects
Sex difference (effect of sex on number of interaction partners)	0.01	0.07	.886
Sex similarity (tendency to interact with same-sex interaction partners)	0.79	0.04	.000
Age difference (effect of age on number of interaction partners)	0.01	0.01	.317
Age similarity (tendency to interact with similar-age interaction partners)	0.06	0.14	.669
Language difference (effect of language on number of interaction partners)	−0.10	0.08	.412
Language similarity (tendency to interact with same-language interaction partners)	0.36	0.05	.000
Financial strain difference (effect of financial strain on number of interaction partners)	−0.03	0.03	.317
Financial strain similarity (tendency to interact with similar-financial strain interaction partners)	0.32	0.10	.001
Parent education difference (effect of parent education on number of interaction partners)	0.02	0.03	.505
Parent education similarity (tendency to interact with similar-parent education interaction partners)	0.06	0.10	.549
Receptive vocabulary difference (effect of receptive vocabulary on number of interaction partners)	0.00	0.00	.317
Receptive vocabulary similarity (tendency to interact with similar-receptive vocabulary interaction partners)	0.32	0.13	.014
Proportion of time in classroom difference (effect of the proportion of time in the classroom on number of interaction partners)	0.00	0.04	1.000
Proportion of time in classroom similarity (tendency to interact with similar-proportion of time in the classroom interaction partners)	0.14	2.37	.953
Repeating year in classroom difference (effect of repeating year in classroom on number of interaction partners)	−0.25	0.32	.435
Repeating a year in classroom similarity (tendency to interact with similar-proportion of time in classroom interaction partners)	−0.24	0.18	.184
Preschool competency difference (effect of preschool competency on number of interaction partners)	0.04	0.04	.317
Selection effects
Selection based on preschool competency similarity (preference to interact with others with similar levels of preschool competency)	1.58	0.56	.005
Behavioral Parameters
Structural control effects
Linear tendency (linear tendency for preschool competency)	0.29	0.08	.000
Quadratic shape (deviations from the linear trends of preschool competency)	−0.42	0.08	.000
Behavioral control effects
Effect from sex (change in preschool competency as a function of sex)	0.23	0.17	.176
Effect of age (change in preschool competency as a function of age)	0.01	0.02	.617
Effect of language (change in preschool competency as a function of language)	0.61	0.24	.011
Effect of financial strain (change in preschool competency as a function of financial strain)	−0.14	0.09	.120
Effect of parent education (change in preschool competency as a function of parent education)	−0.01	0.07	.886
Effect of receptive vocabulary (change in preschool competency as a function of receptive vocabulary)	0.01	0.01	.317
Effect of proportion of time spent in classroom (change in preschool competency as a function of proportion of time spent in classroom)	−0.00	0.00	.317
Effect of repeating year in classroom (change in preschool competency as a function of repeating year in classroom)	0.14	0.30	.641
Influence effects
Peer influence on preschool competency (interaction partner influence on preschool competency)	1.40	0.25	.000

Note. N=308 participants from 18 schools. For each parameter, degree estimates, similarity estimates, and behavioral effect estimates were modeled. Statistical significance of the mean parameter estimates is obtained by an approximate t-ratio of the estimate divided by its standard error (SE). Sex: boy=1, girl=2. Language: English=1, Spanish=2. P-values are presented as two-tailed tests of statistical significance.

Structural controls

As can be seen in Table 2, the significant negative effect for degree indicates that children are selective in their peer interaction partners (Odds Ratio=0.69). This indicates that children’s choices about peer interaction partners were not made at random and, instead, were a function of conscious choice regarding with whom they will interact. The significant positive effect of transitive triads (Odds Ratio=1.08) suggests that children tend to be brought into interactions with the peer interaction partners of their peer interaction partners, that is, there is evidence of transitivity in the social relationships found within preschool classrooms.

Network controls

Several network control estimates were significant (see Table 2): Children were more likely to interact with same-sex peer interaction partners (Odds Ratio=2.20), same-language peer interaction partners (Odds Ratio=1.43), peer interaction partners from families with similar levels of financial strain (Odds Ratio=1.08), and with peer interaction partners who had similar receptive vocabulary skills (Odds Ratio=1.05). As an example of the interpretation of these control effects, a significant and positive financial strain similarity effect is interpreted as a tendency to affiliate with peers who were similar to oneself in levels of financial strain. No other control effects were significant.

Selection effect

The first major question addressed in this study was whether preschool children select peers based on similarity in preschool competency. Notice that the selection based on preschool competency similarity effect in Table 2 is positive and significant, which indicates that Head Start children chose interaction partners who were similar to them on preschool competency. The Odds ratio was 1.48 indicating that children were 48% more likely to select a peer interaction partner who was similar in levels of preschool competency than to select a peer interaction partner who was dissimilar in levels of preschool competency.

Influence on Preschool Competency

The second major research question addressed in the present study is whether Head Start children are influenced by their peers on preschool competency. As before, we first review the findings for the control parameters in the behavioral analyses.

Structural controls

As can be seen in Table 2, the significant linear tendency effect (Odds Ratio=1.08) illustrates that children drifted toward levels above the midpoint on the preschool competency scale. The significant quadratic shape effect (odds ratios could not be calculated because this is a nonlinear parameter) suggests that there was some regression to the mean over time. However, a benefit of the RSiena program is that we were able to account for these behavioral tendencies within each classroom, or any evidence of non-normality in the level of preschool competency in each of these 18 Head Start classrooms.

Behavioral controls

The significant effect of language (coded as English=1; Spanish=2) shows that Spanish-speaking children tended to increase preschool competency levels more than English speaking children (Odds Ratio=1.84). None of the other behavioral control effects were statistically significant.

Influence effect

The second major question addressed in this study was whether preschool children influence their peer interaction partners to change preschool competency across the preschool year. Notice that the peer influence on preschool competency effect in Table 2 is positive and significant, which provides evidence that peers influenced behavioral change on children’s preschool competency over time. That is, children were influenced by their frequent peer interaction partners to either increase or decrease their levels of preschool competency in a similar direction to that of their frequent peer interaction partners as compared to the peers with whom they did not frequently interact. The Odds ratio was 1.42, indicating that children were 42% more likely to move toward the preschool competency of their peer interaction partners than to move toward the preschool competency levels of children with whom they did not regularly interact. The particular importance of this finding is that influence effects were found even after controlling for the initial tendency to interact with (or select) peers similar to oneself on preschool competency and after controlling for the structure of peer relationships and behaviors within these 18 Head Start classrooms. Finally, the same pattern of statistically significant results emerged using an item level analysis, rather than the composite score, of the preschool competency scale.

Discussion

Despite the fact that early childhood education programs provide children with important sources of peer socialization, research examining peer effects in early childhood educational settings has been neglected. This is significant because understanding how peers influence early school success can help early childhood educators manage student groupings more effectively to maximize potential benefits and minimize potential problems. Additionally, even when peer effects have been examined, researchers typically only assess a few (4–6 children) of the children in the classroom (e.g., Henry & Rickman, 2007; Mashburn et al., 2009), which contributes to measurement error when extrapolating to classroom level effects. The present study overcomes these shortcomings by utilizing a larger complete network sample of children within Head Start classrooms to determine the role of peer interactions on children’s preschool competency. Furthermore, the present study is unique in two additional ways: (1) Its use of intensive observational data and (2) its use of a rigorous analytical technique that allowed us to disentangle selection from influence effects on preschool competency, allowing for an unbiased estimation of peer influence on preschool competency. Using an intensive data collection design and a rigorous analytic approach, the current study demonstrates that peer influence processes play an important role in young children’s preschool competency.

The goal of the present research was to address two important issues. First, do Head Start preschool children select peers who are similar to themselves in preschool competencies? Second, do preschool peers play a role in influencing preschool competency among Head Start children? To answer the influence question required the use of an innovative analytic technique to examine selection effects simultaneously with influence effects, in order to disentangle one from the other. After modeling selection and influence simultaneously, as well as accounting for normative relationship and behavioral trends within these 18 Head Start classrooms, evidence for significant levels of peer selection and peer influence on children’s preschool competency was detected. The Head Start children in this study selected peers based on similar levels of preschool competency and were influenced to become more like their peer interaction partners on levels of preschool competency across the academic year.

It is important to also note that these effects held after controlling for a number of network, individual, and family covariates, such as sex, age, language, family financial strain, parent education, receptive vocabulary, percentage of time present in the classroom, repeating a year in the classroom, and the effect of preschool competency on a child’s number of peer interaction partners. By including these control parameters we were not only able to control for these characteristics as possible confounding effects in our estimation of peer selection and peer influence, but, by doing so, we also learned that Head Start children have preferences for same-sex, same-language, and similar financial strain peer interaction partners. Although not the primary focus of the current study, these are tendencies for peer affiliation in the preschool classroom that will require further attention and research. For example, we may need to think more carefully about child placement within preschool classrooms. If children select peers based on the severity of economic disadvantage within the Head Start classroom, a classroom peer group characterized by initial economic disadvantage may lead to the kind of social division within the Head Start classroom (i.e., more advantaged children affiliating with other more advantaged children; less advantaged children affiliating with other less advantaged children) that leads to even greater social divisions outside the Head Start classroom. Awareness of such tendencies may help teachers understand the need to create more opportunities for children to engage in diverse peer interactions that promote positive outcomes. For example, children in higher-SES and higher-achieving peer groups and classrooms tend to be viewed as possessing more-advanced skills (Ready & Wright, 2011). Given this, the ways in which children are grouped together may have important consequences for how they are perceived. Thus, further focus on this may create more socially and academically diverse preschool contexts that limit isolation and/or segregation of children perceived to be of lower academic or social competence.

Although it is generally believed that young children’s preschool competency is influenced by the socialization agents to which they are exposed (e.g., Bierman, Domitrovich, Nix, Gest, Welsh, et al., 2008), most research to date has focused on studying the effects of exposure to parents and teachers, but little attention has been given to peers. Peers have been shown to be influential on school competency for older children, and the current findings suggest that younger children are similarly influenced by their peers. To study the influence of peers within classrooms it is important to consider children’s full social networks, and to examine changes over time in both the relationships of peers in the classroom and in children’s behavioral profiles as a consequence of peer interactions in the classroom. Without longitudinally and simultaneously accounting for all social relationships (and their change over time) and all behavioral profiles (and their change over time) in the classroom it is not possible to know if relationships change due to behavioral effects (selection) or if behaviors change due to relationship effects (influence), nor is it possible to account for the relational interdependence that plagues all social relationship data (Laursen, Popp, Burk, Kerr, & Stattin 2007).

After accounting for these tendencies in the present study, peer influence over preschool competency was apparent. Further, the magnitude of peer influence effects found in these preschoolers was similar to the magnitude of peer influence effects on school-related constructs reported in samples of older children and adolescents (e.g. Altermatt & Pomerantz, 2003; Kinderman, 2007; Lomi, Snijders, Steglich, & Torlo, 2011).

Understanding Peer Effects

The present study is not the first to use a social network approach to understand peer processes in the preschool classroom. For example, Martin et al. (2013) used social network analysis to simultaneously model selection and socialization processes and found that young children tended to select peers to interact with on the basis of being the same sex and to influence peers on gender-typed activity involvement. Additionally, Santos et al., (2008) used social network analysis to examine preference for preschool children’s most proximate peer interaction partners, relative to other peer affiliates in the classroom. Children’s use of external cues (e.g., sex, proximity) for peer selection may not be so surprising, but, in the present study, young children selected peers on the basis of being similar on a school-related outcome (e.g., preschool competency), suggesting that similarly engaged preschool children are drawn into interactions with one another through shared interests, similar behavioral profiles (e.g. non-disruptive, socially skilled), and cognitive abilities. Thus, children appear to use a wide array of cues to make distinctions among potential peer interaction partners. The important contribution of the current study is that it extends prior research. Preschool peers are selected, not only on the basis of sex and proximity, and influenced, not only on the basis of gender-typed activities; the present study demonstrates that preschool peers are also selected and influenced on important indicators of school readiness.

In addition to peer selection, the present findings also point to a robust effect of peer influence on preschool competency in Head Start classrooms. Over a period of a few months, children’s levels of preschool competency changed as a function of the preschool competency of their peer interaction partners. Overall, little is known about how peers exert influence at any age (e.g., Brechwald & Prinstein, 2011) but even less is understood about how peers exert influence during early childhood. Thus, interesting questions remain about how this influence occurs.

One obvious candidate is peer modeling (Bandura, 1977). If children model the behaviors of the peers with whom they spend time, they will also begin to behave similarly to those peers. Some limited evidence suggests that children who are able to model positive peer interaction styles have enhanced pre-academic skills (Craig-Unkefer & Keiser, 2002). However, just as older adolescents are influenced by peers who exert negative influence through deviant talk (Dishion, McCord, & Poulin, 1999; Piehler & Dishion, 2007), young children may also be negatively influenced by peers. For example, preschool peers can present children with models for academic disengagement and disruption. If children model these negative behaviors (e.g., academic disengagement, disruption) they may also become disinterested and disruptive, or simply fail to attend to the learning environment.

Evidence shows that peers also provide children with social connection to the classroom context (Wentzel, 2009). Therefore, peer influence may work indirectly, through the classroom as a system, because social experiences can motivate engagement or disengagement in the classroom (Juvonen & Wentzel, 1996). Academic competence involves a system of interactions and transactions between people, settings, and institutions (Mashburn & Pianta, 2006). As such, the effects of peers on preschool competency may operate indirectly via influence on children’s individual skill levels (e.g, Hoxby & Weingarth, 2005), on teachers’ practices (e.g., Howes et al., 2005), on overall classroom climate and quality (e.g., Mashburn et al., 2008), or a variety of other individual and setting level characteristics. Having identified this important form of peer influence on Head Start children, the next important step is to explicate the processes through which peers affect young children’s preschool competency.

Implications

The significance of early peer relationships may be of particular relevance to Head Start children because underprivileged children enter preschool at an increased risk for lower levels of preschool competency (Stipek & Ryan, 1997). A host of factors (e.g., environmental stress, adverse childhood experiences, poor health, genetic risk) have been attributed to low levels of preschool competency among underprivileged children (Bradley & Corwyn, 2002), making it all the more challenging for educators to learn how best to redirect trajectories to support the long-term achievement of underprivileged children. A better understanding of how children form early peer relationships, how these early peer relationships reflect a child’s level of school readiness, and how these early peer relationships may be used to promote long-term academic success upon school entry is critically important knowledge for early childhood educators.

Preschool competency appears to be susceptible to peer influence. This suggests that Head Start teachers may need continued support and resources to understand and enrich the peer interactions in their classrooms. This is especially true because many preschool children are learning, for the first time, how to navigate peer relationships in the classroom. Teachers and parents can be a resource for instructing children how best to communicate and cooperate with peers. This may, in turn, move more children into peer interactions that support, rather than inhibit, learning. Teachers can also break up peer groups in the classroom that undermine learning or work to change the joint behavior exhibited by groups that are uninvolved in learning. Classroom rules and norms may also directly impact the nature of peer interactions. Teachers may be able to encourage children to develop classroom norms that promote collaborative interactions and, in doing so, open the door to positive peer influence among frequent peer interaction partners (e.g., DeLay et al., 2015). By giving careful attention to patterns of peer interactions, and strategically managing peer interaction partners, educators may be able to maximize positive peer influence and prevent, or disrupt, negative influence processes in their classroom. It is also important to note that peer effects on preschool competency and the implications for early childhood educators are not necessarily unique to Head Start classrooms. Other types of preschool classrooms may have even greater variability in levels of preschool competency, giving rise to a similar opportunity for peer selection and peer influence effects to change children’s preschool competency across time.

Limitations and Conclusion

Some caution needs to be taken in generalizing from the present findings. The sample included a large number of Mexican-American children from low-income families, and although we included language spoken as a control variable in our models, we were unable to assess the specific effects of other factors that might be associated with this particular sample (e.g., acculturation). Furthermore, expanding the range of peer effects on academic competencies would be useful to explore whether children are influenced in the same ways for other important aspects of academic readiness because a 7-item construct of preschool competency is limited in its ability to broadly capture all aspects of school readiness observed in a preschool classroom. Thus, research is needed to further explore additional important aspects of preschool students’ preschool competency. Relatedly, the current study was limited to reliance on teacher reports of preschool competency. Although teachers are considered reliable reporters of the characteristics of children in their classroom (Rodkin, Farmer, Pearl, & Van Acker, 2000), it is possible that a single rater may have diminished some of the variability found in the present reports and limited our ability to detect statistically significant peer effects or added some bias to effects as a function of teachers’ perceptions of children who spend time together in their classroom. We also need additional research to determine the classroom context that may lead to peer influence that promotes, rather than inhibits, preschool competency. Previous research demonstrates that collaborative learning will benefit both higher and lower achievers (Webb, 1991), however, a promising direction for future research will be to directly assess the direction of peer influence within preschool classrooms and to determine if there are differences in peer influence effects between classrooms. Classroom factors that may contribute to such variations between classrooms will be important to consider in future research.

Although a longitudinal social network model represents an advance over previous methods used to investigate peer influence, RSiena has not advanced enough to fully take advantage of rich observational data. Thus, information was lost because observations were transformed into discrete affiliation networks. Questions also remain about how best to use observational data to identify ties between children. Currently, a tie was determined to exist when either child in a dyad was observed interacting with the other at a greater than chance level, and, because these interactions were mutual, we symmetrized the network thereby creating symmetrical or reciprocated ties between children. Symmetrical ties were appropriate for our questions of interest concerning influence and selection. However, symmetry may not always be ideal; unreciprocated ties may be more meaningful to study for other questions. For instance, children may aspire to be like people they want to have as friends (unilateral friends – child A wants to be friends with B but B doesn’t want to be friends with A) as could be demonstrated using asymmetrical ties. Future research is also needed to assess individual differences in peer influence trends. Although the current study provides evidence of peer influence within the classroom, we cannot conclude that peer influence holds for all children in the classroom. Finally, we controlled for a wide variety of both structural features of the classroom peer network as well as individual characteristics of students within these classroom peer networks. Nevertheless, we cannot entirely rule out the possibility that a third variable that could not be controlled in a non-experimental design may have added some bias to the present results.

Despite these limitations, our findings provide initial support for the notion that peer relationships in Head Start classrooms can affect children’s preschool competency. This is the first clear evidence that Head Start peers exert influence on Head Start children’s levels of preschool competency. Although early childhood education programs have great potential to enhance children’s school readiness and academic success, most program evaluations do little to explore the mechanisms by which outcomes – and variability in outcomes – are attained (Magnuson & Shager, 2010). Thus, a better understanding of how peers influence outcomes may be critical to understanding the variability between early childhood preschool outcomes and to promoting the social and academic competency of underprivileged youth. A more complete understanding of peer influence during early childhood will inform educators how best to use the peer relationships in their classrooms to foster the academic success of all children.