teacher practices as predictors of children's classroom social preference

Journal of School Psychology 50 (2012) 95–111

Contents lists available at SciVerse ScienceDirect

Journal of School Psychologyj ourna l homepage: www.e lsev ie r.com/ locate /

j schpsyc

Teacher practices as predictors of children's classroomsocial preference☆

Amori Yee Mikami a,⁎, Marissa Swaim Griggs a,Meg M. Reuland a, Anne Gregory b

a University of Virginia, VA, United Statesb Rutgers University, NJ, United States

a r t i c l e i n f o

☆ This study was supported by the National AcadMikami. Portions of this data have been presented preResearch in Child Development. We thank the teacheand Daphne Keiser for their assistance with recruitmresearch assistants in data collection and data entry.⁎ Corresponding author at: University of British Co

2755.E-mail address: [email protected] (A.Y. MikaACTION EDITOR: Sara Bolt.

0022-4405/$ – see front matter © 2011 Society for tdoi:10.1016/j.jsp.2011.08.002

a b s t r a c t

Article history:Received 21 June 2010Received in revised form 6 August 2011Accepted 9 August 2011

Students who do not get along with their peers are at elevated risk foracademic disengagement and school failure. Research has predomi-nantly focused on factors within such children that contribute to theirpeer problems. This study considers whether teacher practices alsopredict social preference for children in that classroom. Participantswere 26 elementary school teachers and 490 students in theirclassrooms followed for one school year. Results suggested thatteachers who favored the most academically talented students in thefall had classrooms where children had lower average socialpreference in the spring after statistical control of children's fall socialpreference and externalizing behavior problems. Teachers whodemonstrated emotionally supportive relationships with students inthe fall had classrooms where children had greater possibility ofchanging their social preference from fall to spring. Although childrenwith high externalizing behaviors tended to experience decliningsocial preference over the course of the school year, teachers’ learner-centered practices attenuated this progression. However, teachers’favoring of the most academically talented accentuated the negativerelation between externalizing behaviors and social preference.Implications for school psychology practitioners are discussed.

© 2011 Society for the Study of School Psychology. Published byElsevier Ltd. All rights reserved.

Keywords:Social preferenceTeacher practicesExternalizing behaviorsPeer rejectionTeacher-student relationshipsSociometric preference

emy of Education and Spencer Foundation Postdoctoral Fellowship to Amoriviously at the Society for Research on Educational Effectiveness and Society forrs and families who participated in this study, and Mike Coiner, Karen Marcus,ent. We also appreciate the help of many undergraduate and graduate student

lumbia, at 2136 West Mall, Vancouver, B.C., Canada V6T 1Z4. Tel.: +1 604 822

mi).

he Study of School Psychology. Published by Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.jsp.2011.08.002

mailto:[email protected]

Unlabelled image

http://dx.doi.org/10.1016/j.jsp.2011.08.002

Unlabelled image

http://www.sciencedirect.com/science/journal/00224405

96 A.Y. Mikami et al. / Journal of School Psychology 50 (2012) 95–111

1. Introduction

Elementary school students who cannot get along with their peers are relevant to educators becausethese social problems can interfere with the learning environment for the entire classroom (Stormont,2001). Furthermore, children with peer problems are at elevated risk for subsequent academicdisengagement, school failure, and dropout, even after statistical control of the original levels ofachievement (Buhs, Ladd, & Herald, 2006). Collectively, findings underscore the educational importance ofchildren's peer relationships.

Existing research has predominantly focused on the individual characteristics of youth with peerproblems that contribute to their ostracism, such as these children's poor behaviors (Newcomb, Bukowski,& Pattee, 1993). While acknowledging the veracity of this perspective, we propose that social contextualinfluences on peer problems have been understudied (Mikami, Lerner, & Lun, 2010). This article examinesthe possibility that elementary school teachers’ instructional practices and patterns of interaction withstudents may be associated with (a) the overall levels of social preference (a measure of the proportion ofpeers who like the child, minus the proportion of peers who dislike the child) in the classroom; (b) thechange versus stability of children's social preference over the course of the school year; and (c) the extentto which children with externalizing behavior problems show declining social preference.

We note that the traditional method used to assess social preference has limited the investigation ofteacher influences on this construct. Sociometric nominations, in which children name the peers whomthey like and dislike, are the gold standard to assess social preference and are considered superior toreports from parents or teachers (Coie, Dodge, & Coppotelli, 1982; Parker & Asher, 1987). Social preferenceis determined by subtracting the proportion of disliked nominations from the proportion of likednominations a child receives. However, in the traditional model proposed by Coie et al. (1982), childrennominate exactly three peers whom they like and threewhom they dislike, and social preference scores arestandardized within each group of peers providing nominations.

This model has been influential, such that many researchers have subsequently either constrained thenumber of nominations children may provide, standardized nominations within classrooms, or both (e.g.,Cillessen & Bellmore, 1999; DeRosier, Kupersmidt, & Patterson, 1994; Dodge, Coie, Pettit, & Price, 1990;Dodge et al., 2003; Hoza et al., 2005; Masten, Morison, & Pellegrini, 1985). Crucially, these practices restrictthe possibility that some classrooms could have higher average social preference than others, potentiallyowing to differences in teacher behaviors. More recently, researchers interested in social contextualinfluences on peer relationships have begun to depart from this tradition (e.g., Chang, 2004; Donohue,Perry, & Weinstein, 2003), and the current study is aligned with these directions. In sum, the historicalmethodology of sociometric nominations has limited investigation into variability in social preferenceacross classrooms as well as teachers’ influence on this variability. Research about contextual effects isscarce in comparison with the vast literature about child internal characteristics that influence socialpreference.

1.1. Teacher practices and children's social preference

Teachers may affect the overall level of social preference in their classroom because elementary schoolchildren's evaluations of their peers may be based, in part, on their observations of the teacher's reactionsto these students. When teachers demonstrate that they value a child, their behaviors may set an examplefor peers to follow. Using a design where children in kindergarten through second grade viewed avideotaped classroom scene, White and Kistner (1992) and White, Jones, and Sherman (1998)experimentally manipulated teacher responses to a target student in the video whose behavior wasconsistent across conditions. Results revealed that children were influenced by the teacher's responses inthe video, liking the target child more when the teacher expressed positivity relative to when the teachermade derogatory statements toward that child.

A cross-sectional study of third and fourth graders found that peers’ perceptions of a teacher'srelationship quality with a student was correlated with that student's social preference after statisticalcontrol of problem behavior (Hughes, Cavell, & Wilson, 2001). In short-term longitudinal designs, teacherreports of the positive quality of the teacher–student relationship (Hughes & Kwok, 2006) and of personalliking for the student (Taylor, 1989) predicted elementary school children's subsequent gains in social

97A.Y. Mikami et al. / Journal of School Psychology 50 (2012) 95–111

preference over the course of a year. However, although these studies support the hypothesis that teachersaffect students’ evaluations of peers, they still focus on individual differences between children tounderstand social preference. Teachers are assumed to havemore positive interactions with some studentsthan with others, and those children who receive the teacher's favor may grow in social preference relativeto classmates who do not. Not considered is the idea that some teachers may be better than others inestablishing consistently good relationships with all students. Presumably, in those teachers’ classrooms,the average social preference should be higher overall.

The current study tests the hypothesis that teachers who use instructional practices that value allstudents will create a socially accepting peer environment in that classroom that encourages higher socialpreference. One possible practice is the teacher's emotional support of students: the extent to which,during regular instructional time, the teacher is sensitive to student needs, has respect for studentperspectives, and creates warm relationships with students. Such emotionally supportive interactions maydemonstrate that the teacher respects all students, thereby setting an example for peers to follow andraising the average level of social preference in that classroom. High teacher emotional support, asmeasured by the well-validated Classroom Assessment and Scoring System (CLASS; Pianta, La Paro, &Hamre, 2007), has predicted children's greater teacher-rated social competence one year later (Hamre &Pianta, 2005; Mashburn et al., 2008), and has been concurrently associated with children's socialconversation with peers (Rimm-Kaufman, La Paro, Downer, & Pianta, 2005).

A second possible practice involves the teacher's differential interactionwith students at varied levels ofacademic achievement. Teachers may preferentially treat students who have the highest achievement;these teachers are described as promoting an academic status hierarchy whereby some students aresuperior to others based on academic ability (Weinstein, 2002). In addition, teachers vary in their use oflearner-centered (LC) practices, representing teachers’ focus on the process of learning rather thancorrectness, communication that all students are capable of learning, and direction of instruction towardstudents’ interests. By contrast, non-learner-centered (NLC) practices emphasize a single correct answerand child conformity to a uniform standard predetermined by the teacher (APA Learner-CenteredPrinciples Workgroup, 1997). LC practices, because of their sensitivity to students of diverse academicabilities, may be negatively associated with academic status hierarchy (McCombs, Daniels, & Perry, 2008).

We hypothesize that LC practices also convey to peers that the teacher values all students. By contrast,teachers who preferentially treat students based on an academic status hierarchy, as well as NLC practices,may communicate to peers that it is similarly appropriate to have a social status hierarchy where childrenare socially unequal. Although there are other ways in which a teacher could favor some students aboveothers (e.g., based on students’ race or socioeconomic background), we focus on teachers’ differentialtreatment of students based on academic ability because of the high frequency with which this occurs(Weinstein, 2002).

LC practices (McCombs et al., 2008) as well as instruction that equalizes academic status (Cohen & Lotan,1995) have been suggested to improve children's motivation for learning, but less studied are their effects onpeer relationships. However, Donohue et al. (2003) found that first grade teachers’ LC practices predictedchildren's improved social preference over a school year. Cooperative learning interventions (thought todisrupt academic status hierarchies) have led to gains in observations and self-report of children's positiveinteractions with peers (Mikami, Boucher, & Humphreys, 2005; Roseth, Johnson, & Johnson, 2008).

1.2. Change versus stability of social preference

Past research suggests moderate stability over time in elementary school children's social preference.For example, peer liking has been found to be correlated .51 and peer disliking correlated .85 over a 5-weekperiod (Mikami & Hinshaw, 2003), and about half of peer-rejected children are still in that same categoryone year later (Cillessen, Bukowski, & Haselager, 2000). These findings suggest that classrooms often havea set social hierarchy where children's reputations are not malleable. Whether teachers may affect thelikelihood that children's social preference will remain stable versus change is an untested question.

The stability of social preference is typically conceptualized as resulting from children's consistentbehaviors over time, particularly from consistency in problem behaviors of disliked children (Newcombet al., 1993). Yet, the high stability of social preference may also be perpetuated by cognitive biases held bythe peer group—biases that the teacher may be able to influence. For example, peers selectively recall


negative actions of disliked children while forgetting positive actions; yet these cognitive biases arereversed to favor popular children (Hymel, Wagner, & Butler, 1990). Further, peers attribute hostile intentwhen a child with low social preference displays ambiguous behavior, but attribute benign intent whensimilar behavior is performed by a child with high social preference (Peets, Hodges, & Salmivalli, 2008). Insum, peers are disposed to give the benefit of the doubt to their children whom they already like, while thesame actions enacted by a disliked child are negatively viewed, whichmakes it unlikely that peers will everchange their opinions of their classmates (Mikami et al., 2010).

Although we are not aware of any published, empirical studies examining teacher influences on thestability of social preference, we hypothesize that the same teacher practices associated with improvedsocial preference overall may also be associated with lower stability in social preference over time.Teachers’modeling that all students have valuemay draw peers’ attention to children's positive attributes;when this process occurs for disliked children, it may disrupt the negative reputations that perpetuate theirlow social preference. In addition, a teacher's use of LC practices that emphasize the process of learningrather than conformity to a single correct standard and a teacher's dismantling of an academic statushierarchy may communicate to children that flexibility is similarly possible in social status.

Clearly, disliked children potentially benefit most from being in a classroom where social preference ismutable, and they are the group for whom social preference tends to be the most stable. However, aclassroom with weak correlations between fall and spring social preference may be desirable for allchildren, not only those with low social preference. Although speculative, this type of classroommay havestudents who are more flexible in developing their own impressions about peers, as opposed to studentswho follow a tightly-enforced norm about the reputations of each child and who enact a rigid social statushierarchy.

1.3. Externalizing behavior problems and social preference

Evidence suggests that children with high externalizing behaviors (e.g., hyperactivity, aggression andoff-task disruptive behavior) are at great risk for peer problems. Children with clinical levels ofexternalizing behaviors develop low social preference among previously unacquainted peers inmere hours(Erhardt & Hinshaw, 1994). Nonetheless, factors in the environment determined by the classroom teachermay affect this process. A consistent finding is that aggressive, off-task, and hyperactive children havehigher social preference in classrooms where such behavior is common among peers than in classroomswhere such behavior is unusual (Chang, 2004; Stormshak et al., 1999). This pattern of results may occurbecause a high base rate of externalizing behavior in a classroom makes peers view such actions asnormative and socially acceptable. However, these studies do not addresswhether the teacher can increasepeers’ tolerance of diverse behavior in the absence of disruptive peer norms.

Peers may watch the way the teacher responds to students with externalizing problems, and use theteacher as an exemplar when judging those students themselves. Using cross-sectional designs, a teacher'scriticism toward second graders in front of peers (McAuliffe, Hubbard, & Romano, 2009) and a teacher'spersonal disliking of elementary school students (Chang et al., 2007) andmiddle school students (Chang etal., 2004) mediated the negative relation between student externalizing behaviors and social preference.Although not specific to externalizing behavior, students attending LC elementary schools had higher self-reported tolerance for peers who behave differently relative to students in NLC schools (Salinas & Garr,2009). Collectively, these studies suggest that teachers who communicate that a disruptive student hasvalue may break the typical link between externalizing problems and low social preference. In support ofthis hypothesis, a teacher professional development intervention to increase emotional support andsensitivity to students with diverse learning needs yielded improvements in students’ observed and self-reported peer interactions, with results particularly strong for students with externalizing problems(Mikami, Gregory, Allen, Pianta, & Lun, in press).

1.4. Study hypotheses

Our first hypothesis was that elementary school teachers’ high emotional support and LC practiceswould predict increases in children's social preference over the school year while controlling for fall socialpreference and externalizing behavior problems. Conversely, NLC practices and promotion of academic


status hierarchies should predict decreases in social preference. Our second hypothesis was that emotionalsupport and LC practices would predict reduced stability in social preference (and the opposite for NLCpractices and academic status hierarchy). That is, we hypothesized that there would bemore opportunitiesfor children to change in social preference when teachers demonstrated more emotional support and LCpractices, and fewer NLC practices with less academic status hierarchy. Our third hypothesis was thatwhereas youth with externalizing behaviors would show decreases in social preference over the schoolyear, teachers’ high emotional support and LC practices would predict a weaker relation betweenexternalizing behaviors and low social preference (and the opposite for NLC practices and academic statushierarchy).

2. Method

2.1. Participants

Participants were 26 teachers (kindergarten through fourth grade) representing suburban and ruralareas of the southeastern United States. Teachers were drawn from three schools (n=18 at school 1, n=6at school 2, and n=2 at school 3) locatedwithin a 10-mile radius of one another. All teachers werewomen,and most were White, which represents elementary teachers in this region. Their teaching experienceranged from 2 to 38 years (M=16.30; SD=9.27). See Table 1 for a breakdown of teacher demographiccharacteristics by grade level.

During staff meetings, the first author presented the study to all regular education homeroom teacherswho had a core group of children that stayed together in that classroom for at least 80% of the school day(about n=50 potential teachers). Special education and elective teachers were not eligible. Studyparticipation was optional, and 35 teachers agreed to be in the study; however, during the course of theresearch 3 teachers quit their teaching job and 6 additional teachers’ data were not used because fewerthan 40% of their students gave active consent, a cutoff which has been suggested to obtain validsociometric nominations (Hughes & Kwok, 2006; McAuliffe et al., 2009). All teachers provided active,written consent.

Study classrooms ranged in size from 12 to 26 children (M=18.85; SD=3.23). Children in consentedteachers’ classrooms were recruited by presenting the study to parents at the school open house and vialetters. Participation was optional. All children in consented teachers’ homerooms were eligible toparticipate; special education students who were mainstreamed in regular education homerooms wereincluded. There were 296 children nested in these classrooms whose parents provided active, written

Table 1Teacher and child demographics across grade levels.

Kindergarten First grade Second grade Third grade Fourth grade Total sample

TeachersN 8 5 6 5 2 26% Women 100% 100% 100% 100% 100% 100%% White 100% 100% 67% 100% 100% 92%Years experience 17.8 16.2 11.0 17.3 19.5 16.3Active consent rate 57% 59% 65% 63% 62% 61%Passive consent rate 43% 41% 35% 37% 38% 39%

ChildrenzN active consent 91 59 73 50 23 296N passive consent 66 28 49 40 8 194N denied consent 13 7 0 4 1 25% Girls 52% 51% 53% 38% 57% 50%% White 85% 83% 74% 74% 95% 82%

Note. No omnibus test for group differences in demographics by grade was statistically significant (all pN .05). The child gender andethnicity proportions pertain to active consent participants only; we did not obtain demographic data for children with passiveconsent or for children whose parents denied consent. The ethnicity breakdown of children in the district was as follows: 82%White,11% African American, 4% Latino, and 3% Asian American. Free/reduced lunch information was not available for individualparticipants. Twenty-nine percent of the children in the district were identified as eligible for free or reduced-cost lunch.


consent to participation. Themean classroom active consent rate was 61% (range 45–92%). Passive consentwas obtained from an additional 194 students whose parents neither indicated active consent nor repliedthat they did not want their child to participate. The mean classroom passive consent rate was 40% (range8–55%). Therefore, data were obtained on 490 students. Few children had parents who denied consent(n=25, or 5% of eligible participants). Fifty percent of child participants were girls, and 82% were White.Please see Table 1 for a breakdown of child demographics by grade level. All study procedures wereapproved by a university review board.

2.2. Measures

2.2.1. Children's social preference (fall and spring)Children were administered a standard sociometric procedure (Coie et al., 1982) in private, individual

interviews where they were asked to nominate classroom peers whom they liked and disliked. At theschool fromwhich the majority of the participants came (n=18 classrooms), children were provided withthe pictures and names of each of their classmates in order to aid recall. This technique was not possible forthe remaining schools, so children were provided with a printed list of their classmates’ names, andresearchers read the list to the children.

Children were allowed to make unlimited nominations of classroom peers. This nomination processwas accomplished by the research assistant explicitly saying “Is there anyone in your class who you like?It's okay if there is no one.” Then, if the child provided a nomination, the research assistant asked, “Is thereanyone else who you like, or is that all?” This procedure continued until the child indicated there wasnobody else to nominate. Then the research assistant repeated these instructions to solicit dislikednominations. At the fall assessment point, children on average gave 3.75 nominations of peers they likedand 1.94 nominations of peers they disliked; however, the range was large (0–16 for liked and 0–13 fordisliked nominations given). Similarly, at the spring assessment point, children on average gave 3.89 likednominations (range 0–16) and 2.03 disliked nominations (range 0–10). These results suggest that theunlimited nomination procedure yielded variability among children in the number of classmates whomthey liked and disliked.

Proportion scores were calculated for each child by dividing the number of liked and dislikednominations that the child received by the number of children who provided nominations (i.e., thechildren in that classroom who participated in the interview procedure, who were those children withactive consent). Social preference was calculated by subtracting the proportion of disliked nominationsfrom the proportion of liked nominations received (Coie et al., 1982). Sociometric procedures have beenvalidated among children as young as those in our sample (e.g., see Harrist, Zaia, Bates, Dodge, & Pettit,1997).

2.2.2. Observed teacher practices (fall)Teacher practices were measured using the Classroom Assessment and Scoring System (CLASS; Pianta

et al., 2007), on which teachers were observed and their behaviors were coded on 10 dimensions. Previousanalysis of the CLASS suggests these dimensions compose three domains: Emotional Support, InstructionalSupport, and Classroom Organization (Hamre, Pianta, Mashburn, & Downer, 2007). Although studyhypotheses focused on the Emotional Support domain as a predictor of social preference, we coded alldimensions in accordance with standard CLASS procedure. Emotional Support encompasses thedimensions of Positive Climate (which indicates warm teacher–child interactions), Negative Climate(which was reverse scored and indicates teacher irritability), Teacher Sensitivity (which indicates teacherresponsiveness to children's needs), and Regard for Student Perspectives (which indicates teacher respectfor student points of view). Instructional Support includes three dimensions assessing the extent to whichteachers foster students’ higher order thinking skills, foster advanced language, and provide detailedfeedback to students. Finally, Classroom Organization is composed of three dimensions assessing teachers’behavior management, preparedness for lessons, and use of instructional materials.

Research assistants, kept unaware of study hypotheses, received 12 h of training and were required topass a reliability test by scoring within one point of master codes developed by CLASS creators greater than80% of the time on all 10 dimensions. Then, research assistants assessed teachers during in-person visitswhere they completed a 20-minute classroom observation and rated each dimension on a 7-point Likert


scale: 1= low to 7=high. The 20-minute observation plus 10 min of scoring cycle was repeated five timesfor each classroom, which exceeds the recommendation from the CLASS developers that at least four cyclesare needed to obtain a stable estimate of teacher practices (Hamre et al., 2007). During one of the fivecycles for each teacher in the current study, two observers independently scored the same teacher so thatinter-rater reliability could be calculated (ICC=.84 for Emotional Support, .83 for Instructional Support,and .79 for Classroom Organization). The ICC statistic is appropriate because the variable was on a seven-point ordinal scale (Shrout & Fleiss, 1979). We used the average scores on the CLASS domain of EmotionalSupport from the five cycles in the fall.

2.2.3. Self-reported teacher practices (fall)Teachers completed the Learner-Centered Battery (McCombs, Lauer, Bishop, & Peralez, 1997). Using

a 4-point Likert scale (1= strongly disagree, 2=disagree, 3=agree, and 4= strongly agree), teachersrated 13 statements designed to measure LC practices as well as 21 items to measure NLC practices.Items measuring LC practices included “Students have more respect for instructors they see and canrelate to as real people, not just as teachers” and “Seeing things from the students’ point of view is thekey to helping them perform well in school.” Items measuring NLC practices included “I know best whatmy students need to know and what's important so students should take my word that something willbe relevant to them” and “Even with feedback some students just can't learn from their mistakes.”McCombs et al. (2008) reported that the items that load on each scale have been validated using factoranalysis and that moderate correlations exist between teacher self-report and student report of theteacher's LC and NLC practices (rs=.33 to .42, pb .01). Coefficient alpha values in our sample were .80for the LC scale and .66 for the NLC scale. Although these values are lower than the values of .82 for theLC and .72 for the NLC scales reported by McCombs et al. (2008), we left the scales intact so that thisresearch could be compared to existing work.

Teachers also reported their promotion of an academic status hierarchy by using four questionstaken from the Performance Focused Instructional Strategies subscale of the Patterns of AdaptiveLearning Scales (Midgley et al., 2000). Each item is answered on the same 4-point Likert scale as in theLearner-Centered Battery. The items are “I point out students who do well academically as a model forthe other students,” “I give special privileges to students who do the best academically,” “I display thework of the highest achieving students as an example,” and “I encourage students to compete with eachother academically.” Midgley et al. (2000) reported an alpha coefficient of .69, and the alpha coefficientin our sample was .64.

2.2.4. Children's externalizing behaviors (average across year)Using similar procedures as have been established in previous research and suggested to correlate with

adult informant and peer ratings of externalizing behavior (Abikoff, Gittelman, & Klein, 1980; Lee &Hinshaw, 2004), trained observers watched each child for a randomly-chosen 15-second interval duringclass time and rated the child's display of externalizing behavior using a dichotomous scale: 0=behaviornot present and 1=behavior present at any point. Off-task/hyperactive behavior was only recorded duringoccasions that had an academic component. Aggression/noncompliance was recorded any time thechildren were with the teacher, even if they were doing a nonacademic activity. Fifteen percent of theintervals were randomly selected for two observers to code the same child simultaneously in order tocalculate inter-rater reliability. Codes were as follows:

1. Off-task/hyperactive behavior (κ=.62). Observers recorded whether the child made audibleverbalizations that were not permitted or not relevant to the academic task at hand (e.g., talking toself or peers about unrelated topic or when talking not allowed); engaged in motor activity thathindered work on an assigned task (e.g., fidgeting or moving out of seat such that learning iscompromised); or daydreamed and failed to attend to the assigned task for at least 3 consecutiveseconds.

2. Aggression/noncompliance (κ=.80). Observers recorded the presence of verbal or physical aggressionto adults or peers, or noncompliance with adult instructions where the child heard the adult but defiedby talking back or by actively ignoring the command.


On average, each child was observed for a total of 32 15-second periods (range=15–56 periods).Observations were spread across six days during the school year (i.e., across fall, winter, and springassessment points). For each code, proportion scores representing the number of times the childengaged in each behavior divided by the total number of observations conducted on that child werecalculated. The externalizing behavior composite was created by adding the average proportion scorefor off-task/hyperactive behavior to the average proportion score for aggression/noncompliance.

2.3. Procedure

Fall assessments were conducted between September and October. Four hours of observational dataover two days were collected in each classroom during which the teachers’ practices were coded by one setof research assistants (using the CLASS), and different research assistants coded the externalizingbehaviors of children with active and passive consent. In addition, teachers completed questionnairesabout their LC and NLC practices and promotion of academic status hierarchies. Children with activeconsent only were administered a sociometric nomination procedure, but all children with active andpassive consent were included on the roster to be nominated. The entire assessment procedure wasrepeated in winter (December and January) and again in spring (April andMay). The analyses in this articlefocus on the sociometric nominations from fall and spring because this method allowed the largest periodof time to pass between assessment points. Only the observations and self-report questionnaires from theteachers in the fall were used because we wished to determine if baseline indicators of teacher practicespredicted change in child social preference over the school year.

When children with active versus passive consent were compared, those with active consent received asignificantly greater proportion of liked nominations than did those with passive consent in fall, t(485)=3.73, pb .001, and in spring, t(485)=3.32, p=.006. There were no group differences between active versuspassive consent children on disliked nominations received in either fall or spring nor in observations ofexternalizing behavior (all pN .10).

2.4. Data analytic plan

We used Hierarchical Linear Modeling (HLM; Raudenbush & Bryk, 2002) to test hypotheses becauseof the data structure where children were nested in classrooms. HLM accounts for shared variance at theclassroom level in estimation of effects. Unconditional models with spring social preference as theoutcome and fall social preference and externalizing behavior as predictors, but without any teacherpractices, revealed that the intraclass correlation coefficient (ICC; the measure of the variability at theclassroom level) was .11. This ICC suggests that classrooms do differ in levels of social preference,justifying the examination of teacher practices that could explain this variability (Raudenbush & Bryk,2002).

At the child level, social preference in the spring was the criterion and social preference in the fall aswell as externalizing behavior were covariates. At the classroom level, teacher LC practices, NLC practices,promotion of academic status hierarchy, and observed emotional support were entered as predictors. Wedid not include teacher demographics of years of experience, sex, or ethnicity because research suggeststhat these variables are not well-associated with teacher practices (Zumwalt & Craig, 2005) nor do theymoderate the relation between student behaviors and social preference (Chang, 2003). In our sample,there was insufficient variability in teacher sex and ethnicity to assess associations between these variablesand outcomes, and there were no significant correlations between years of experience and any measure ofteacher practices, rs=.02 to .29, all pN .10. In addition, although classrooms were nested in three schools,we did not have hypotheses about school-level effects, nor was there adequate statistical power to testeffects at the school level. Thus, a multilevel model was estimated with the following Level 1 and Level 2equations, respectively:

Level 1:

Yij = β0j + β1j Fall social preferenceð Þ + β2j Externalizing behaviorð Þ + eij


Level 2:

β0j = γ00 + γ01 LC practicesð Þ + γ02 NLC practicesð Þ + γ03 Academic status hierarchyð Þ+ γ04 Emotional supportð Þ + u0j



The outcome variable, Yij, was children's social preference in spring. The statistical significance of the γ01,γ02, γ03 and γ04 coefficients test Hypothesis 1, that fall teacher practices would predict the level of socialpreference in the spring after statistical control of fall social preference and observed externalizing behaviors.If the signs of the coefficients of emotional support and LC practiceswere positive (e.g., associatedwith higherspring social preference) andNLCpractices and academic status hierarchywerenegative (e.g., associatedwithlower spring social preference), these results would be consistent with the hypothesized direction of effects.

The statistical significance of the γ11, γ12, γ13 and γ14 coefficients test Hypothesis 2, that fall teacherpractices would be associated with the stability of social preference between fall to the spring, afterstatistical control of externalizing behaviors. We expected γ10, the coefficient of fall social preference inpredicting spring social preference after controlling for externalizing behaviors, to be positive. If the γ11,γ12, γ13 or γ14 coefficient estimates were negative, this would suggest that in classrooms where teachershave high levels of certain practices, the association between fall and spring social preference was weaker(e.g., the negative γ coefficient operates in the opposite direction of the positive γ10). A positive γ11, γ12, γ13

or γ14 coefficient estimate would suggest that these practices were associated with a stronger relationshipbetween fall and spring social preference. See Chang (2004) for a similar interpretation of cross-levelinteractions. We hypothesized that emotional support and LC practices should have negative coefficientsbut that NLC practices and academic status hierarchy should have positive coefficients.

To test Hypothesis 3, we expected that γ20, the coefficient of externalizing behaviors in predicting springsocial preference after controlling for fall social preference would be negative; that is, children withexternalizing behaviors would decrease in social preference over the school year. A positive γ21, γ22, γ23, orγ24 value would mean that these teacher practices attenuated the typical negative relationship betweenexternalizing behaviors and social preference (e.g., the positive γ coefficient operates in the oppositedirection of the negative γ20). A negative γ21, γ22, γ23, or γ24 value would suggest that these practicesexacerbated this association. We hypothesized that emotional support and LC practices should have positivecoefficients but that NLC practices and academic status hierarchy should have negative coefficients.

All variables were converted to z-scores across the full sample in order to facilitate interpretation ofcoefficients. As suggested by Enders and Tofighi (2007), predictors at Level 1 and Level 2 were grand-meancentered because the primary hypothesis in the current study involves the effects of Level 2 predictors onoutcomes. However, we note that Enders and Tofighi (2007) also recommended group-mean centeringLevel 1 predictors and grand-mean centering Level 2 predictors when testing cross-level interactioneffects. Our study aims involve both main effects of Level 2 predictors and cross-level interactions;however, because we considered the main effects of Level 2 predictors to be the central hypothesis, wecentered Level 1 predictors at the grand mean. We note that when we conducted analyses with Level 1predictors centered at the groupmean, results were nearly identical. Models were interpreted using robuststandard errors, and HLM 6.0 (Raudenbush, Bryk, & Congdon, 2004) was used to conduct data analyses.

A technical note about HLM is that parameter estimates may be quite small in value but still havepractical as well as statistical significance (Chang, 2004). In order to provide estimates of the magnitude ofeffects, descriptions of outcomes for a classroom one standard deviation above and below the mean on thekey teacher practice were included. As an additional metric of effect size, comparisons of the final modelswith teacher practices as predictors relative to the model without these predictors (i.e., the unconditionalmodel) identified the increase in proportion of variance explained in the outcome when teacher practiceswere added (Raudenbush & Bryk, 2002). The number of classrooms (n=26) and the average number ofchildren per classroom (n=19) provided acceptable power; the minimum detectable effect size at Level 2at a statistical power of .80 was d=.25, which is considered to be small (Spybrook, Raudenbush, Congdon,& Martinez, 2009).


3. Results

3.1. Descriptive statistics

All measures of teacher practices were available for the full sample of 26 teachers. Regarding studentmeasures, out of 490 students total, 487 had sociometric data and 473 had observations of theirexternalizing behaviors. Missing data were due to children being frequently absent such that they yieldedfewer than 15 observations of their behavior over at least two different days. Full information maximumlikelihood methods were utilized to address missing data. These procedures have been found to yield lessbiased parameter estimates (versus listwise deletion) when all available data are used for longitudinalanalyses (Enders, 2001).

Table 2 shows the mean, range, and standard deviations of each study variable and the correlationsbetween them. Although teacher LC and NLC practices were significantly correlated, observed emotionalsupport and the self-report scales were not significantly correlated. However, these measures assessdifferent constructs; emotional support captures the affective quality of teacher–student relationships, andthe self-report scales capture academic instructional practices.

Child social preference was normally distributed in both fall and spring. Child externalizing behaviorwas slightly skewed, so that there were more children at the lower end of the distribution and fewerchildren showing high rates of externalizing behaviors. However, because we thought that thesedistributional properties represented meaningful variability on children's externalizing symptoms, andbecause assessment of children's externalizing behavior was an observational measure less likely to beaffected by rater biases, we chose to maintain this variable untransformed. We note that when wetransformed the variable by taking the square root, this transformation resulted in a negatively-skeweddistribution. Substituting this transformed variable resulted in the same overall pattern of results, althoughtwo previously significant predictors remained in the same direction while dropping slightly tononsignificance. All measures of teacher practices could be assumed normally distributed, and there wasadequate range on these measures.

3.2. Teacher practices predict social preference levels

Table 3 displays results showing that when teachers reported more practices that promote an academicstatus hierarchy in fall, there tended to be lower social preference in their classrooms in the spring afterstatistical control of fall social preference and observed externalizing behavior. The magnitude of thisrelation indicated that, after accounting for the other predictors, students in a class that was one standarddeviation below themean in teacher academic status hierarchywould, on average, have a social preferenceraw score of .14, whereas students in a class one standard deviation above the mean in academic status

Table 2Descriptive statistics of child and classroom variables.

Child variable Mean (SD) Min Max 2 3

1. Fall social preference .11 (.23) − .70 .90 .56** − .022. Spring social preference .11 (.23) − .80 .75 – − .073. Externalizing behavior observations .09 (.06) .00 .40 – –

Classroom variable Mean (SD) Min Max 2 3 4 5 6

1. CLASS emotional support 5.12 (0.60) 3.95 6.15 .73** .66** .12 − .15 − .132. CLASS instructional support 3.18 (0.81) 1.73 5.27 – .51** − .16 − .25 − .133. CLASS classroom organization 4.68 (0.63) 3.40 5.73 – .14 .00 − .024. Learner-centered practices 40.05 (4.24) 30 48 – − .44* − .235. Non-learner-centered practices 44.20 (4.52) 35 52 – .306. Promotion of status hierarchy 7.42 (1.60) 4 11 –

Note. CLASS = Classroom Assessment and Scoring System. Columns represent the means with standard deviation values inparentheses, the minimum and maximum values, and correlations. For child variables, n=473, and for classroom variables, n=26.*pb .05; **pb .01.

Table 3Teacher practices as predictors of children's social preference.

Fixed effects Parameter Spring peer social preference

Coefficient (SE) t(21) p

Intercept, β0j γ00 − .016 (.061) −0.26 .801Learner-centered practices γ01 − .041 (.051) −0.81 .425Non-learner-centered practices γ02 .068 (.070) 0.97 .344Academic status hierarchy γ03 − .118 (.055) −2.16 .042*CLASS emotional support γ04 .005 (.052) 0.09 .927

Slope of fall social preference, β1j γ10 .583 (.036) 16.42 b .001**Learner-centered practices γ11 .041 (.045) 0.93 .363Non-learner-centered practices γ12 − .060 (.043) −1.40 .176Academic status hierarchy γ13 .045 (.024) 1.84 .079CLASS emotional support γ14 − .102 (.035) −2.96 .008**

Slope of externalizing behavior, β2j γ20 − .071 (.031) −2.27 .034*Learner-centered practices γ21 .047 (.017) 2.76 .012*Non-learner-centered practices γ22 .038 (.035) 1.08 .292Academic status hierarchy γ23 − .076 (.036) −2.13 .045*CLASS emotional support γ24 .018 (.036) 0.51 .613

Random effects Unconditional model variance component (SE) Final model variance component (SE)

σ2 .62768 (.79226) .61194 (.78227)τ for intercept β0j .07741 (.27823), χ2(25)=58.58 .05327 (.23081), χ2(21)=63.99τ for slope β1j .00689 (.08300), χ2(25)=35.11 .00336 (.05795), χ2(21)=20.27τ for slope β2j .00085 (.02918), χ2(25)=25.28 .00007 (.00820), χ2(21)=23.63

Note. CLASS = Classroom Assessment and Scoring System.*pb .05; **pb .01.


hierarchy would, on average, have a social preference raw score of .08. Given the mean class size of n=19,this finding amounts to a difference in 1.14 more liked nominations per child on average, or conversely,1.14 fewer disliked nominations per child on average. However, teachers’ LC practices, NLC practices, andemotional support did not predict social preference in spring. The addition of all three teacher practices (asa block) accounted for a 31% reduction in the variance of spring social preference.

3.3. Teacher practices predict stability of social preference

As shown in Table 3 and as expected, we also found a significant positive relation between children's falland spring social preference scores, suggesting that social preference was moderately stable over theschool year. However, high observed teacher emotional support in the fall predicted a lower correlation(i.e., reduced stability) between children's fall social preference and spring social preference, afterstatistical control of externalizing behavior problems. The magnitude of this relation indicated that, afteraccounting for other measured variables, a classroom one standard deviation below the mean in emotionalsupport would have an average coefficient of .69 (pb .001) between students’ fall and spring socialpreference scores; whereas a classroom one standard deviation above the mean in emotional supportwould have an average coefficient of .48 (pb .001) between fall and spring social preference. None of theteacher self-reported practices was significant. The addition of the teacher practices (as a block) accountedfor a 51% reduction in the variance associated with the slope of fall social preference predicting springsocial preference.

3.4. Teacher practices affect trajectory between externalizing behavior and social preference

As expected, we found a significant negative relationship between children's observed externalizingbehaviors and spring social preference scores, after statistical control of fall social preference scores. That is,high externalizing behavior problems appear associated with children's declines in social preference over


the course of the school year. Crucially, high teacher LC practices in the fall mitigated this finding,suggesting that for children with externalizing behavior problems in a classroom with high LC practices,the typical strong trajectory between such behavior problems and reductions in social preference wasattenuated. Teachers’ promotion of academic status hierarchies operated in the opposite direction andwasassociated with an accentuated relation between externalizing behavior problems and lowered socialpreference. Emotional support and NLC practices were not significant predictors of the relation betweenexternalizing behaviors and social preference. See Table 3.

After accounting for the other predictors, a classroom one standard deviation below the mean in LCpractices had an average coefficient of −.12 (p=.009) between students’ externalizing behavior andspring social preference scores; whereas a classroom one standard deviation above the mean in LCpractices had an average coefficient of −.02 (p=.659) between externalizing behavior and spring socialpreference. In addition, a classroom one standard deviation below the mean in teachers’ promotion of anacademic status hierarchy demonstrated an average coefficient of .01 (p=.825) between students’externalizing behavior and spring social preference scores, whereas a classroom one standard deviationabove the mean in academic status hierarchy demonstrated an average coefficient of −.15 (p=.001)between externalizing behavior and spring social preference. The addition of the teacher practices (as ablock) explained nearly all (92%) of the variance associated with the slope of externalizing behaviorpredicting spring social preference.

3.5. Additional analyses

We did not have hypotheses that the impact of teacher practices on children's social preference woulddiffer by grade level or child gender, given previous findings of such social contextual effects on peerrelationships from kindergarten through middle school in both sexes (Chang et al., 2004; Donohue et al.,2003; McAuliffe et al., 2009; Stormshak et al., 1999). However, we included exploratory interactionsbetween each teacher practice and grade level (coded as a continuous variable) as predicting β0j, β1j, andβ2j and between each teacher practice and child gender (dummy coded) as predicting β0j. Only 1 of 12possible interactions (8% of analyses) with grade level was significant, a rate that is close to chance, andnone of the interactions with gender was significant, so these findings were not included in this article.

Second, although we conceptualized the CLASS Emotional Support domain to best relate to socialpreference, we re-conducted analyses adding the other CLASS domains Instructional Support andClassroomOrganization tomodels including emotional support (and none of the teacher self-report scales)as predictors. We completed these analyses because there is precedent to consider these factors together inpredicting student outcomes to isolate the independent contribution of one factor over the others (Curbyet al., 2009; Hamre & Pianta, 2001, 2005). The relation between emotional support and lower stability ofsocial preference remained significant. In addition, emotional support also became significantly associatedwith a weaker relation between externalizing behavior and low social preference. In no case was eitherclassroom organization or instructional support statistically significant. Thus, analyses including all CLASSdomains confirmed the particular importance of emotional support for social preference.

4. Discussion

This study provided evidence that teacher practices may relate to children's peer relationships.Teachers who promoted an academic status hierarchy in fall had classrooms in which childrendemonstrated lower social preference in the spring, after statistical control of fall social preference scoresand observed externalizing behavior. Although social preference tended to be stable over the course of theschool year, teachers who displayed high emotional support in the fall had classrooms in which there wereweaker correlations between children's social preference in fall with their social preference in spring aftercontrol of externalizing behavior, suggesting a less rigid peer hierarchy. Finally, children with highexternalizing behaviors tended to decrease in social preference over the course of the school year, butteachers’ use of LC practices mitigated this trajectory. However, in classroomswhere teachers promoted anacademic status hierarchy, the relation between externalizing behavior and decreasing social preferencewas strengthened.


These findings challenge the prevailing model of peer problems that is centered on deficits withindisliked children that cause and maintain their ostracism over time. The current study was the first to useobservational measures to explore the idea that teachers who show positive regard to all children mayraise the average level of social preference in the entire classroom. The findings corroborate previousresearch using teachers’ self-report of their liking of individual students (Hughes et al., 2001; Hughes &Kwok, 2006; Taylor, 1989) and provide additional strength to the claim that teacher practices are linked tostudent social processes (Chang et al., 2004; McAuliffe et al., 2009; Rimm-Kaufman et al., 2005). Effect sizessuggested that children in a classroom one standard deviation below the mean in teacher academic statushierarchy received about 1.14 more liked nominations (or conversely, 1.14 fewer disliked nominations)than did children in a classroom one standard deviation above the mean in academic status hierarchy. Thisdifference may be clinically meaningful because research suggests that the presence of even one friend inthe same classroom can protect otherwise disliked children from victimization (Buhs et al., 2006) andsubsequent psychopathology (Bagwell, Newcomb, & Bukowski, 1998).

To our knowledge, no previous study has investigated influences of teacher preferences on the stabilityof social preference over time. We found substantial relations between fall and spring social preference;however, in classrooms one standard deviation above the mean on teacher emotional support, they weresmaller in magnitude relative to classrooms one standard deviation below the mean in teacher emotionalsupport (B=.48 versus B=.69). These values are in the range reported in other samples (Cillessen et al.,2000; Mikami & Hinshaw, 2003; Newcomb et al., 1993), but importantly, this study suggests that stabilityof social preference may vary somewhat across classrooms.

We hypothesized that a classroom with flexible peer impressions may be positive for all students—notonly for disliked students who potentially have the most to gain from changing their reputations. Forinstance, some initially popular children may be seen in a slightly less positive light by the end of the year,but this change over timemay reflect peers’ valid recognition of changes in these children's behavior ratherthan a tendency to cling to a rigid impression that never shifts. Although speculative, emotionallysupportive teacher–student relationshipsmay highlight to children that it is worthwhile to take the time toget to know their peers and to be flexible in their judgments about classmates. Particularly for dislikedchildren, the teacher may be highlighting their strengths for peers, helping to break down their negativereputations (Mikami et al., 2010).

This study does not imply that child-level factors are irrelevant. In our sample, similar to inprevious work (Erhardt & Hinshaw, 1994), children's externalizing behaviors predicted decreasingsocial preference scores. But crucially, teacher practices exerted partial influence on this relationship,suggesting that the trajectory from externalizing behaviors to poor social preference is not inevitable.In classrooms one standard deviation below the mean in promotion of academic status hierarchy andalso in classrooms one standard deviation above the mean of LC practices, the relation betweenchildren's externalizing behaviors and spring social preference was reduced to nonsignificance. Ourresults are in line with those of Chang (2004), who reported wide variation in correlations betweenchildren's externalizing behavior and social preference in existing research (rs=−.60 to .20) andspeculated that this variability may occur because classroom norms affect the extent to whichchildren with externalizing behaviors face peer problems. Our results support this conclusion, andimportantly, suggest that teacher practices may be a key influence on such classroom norms. Ifteachers show annoyance with children who exhibit hyperactivity, off-task behaviors, and aggression(McAuliffe et al., 2009), our findings raise the possibility that these negative teacher responses maycontribute to the peer problems faced by this population.

The current study advances previous research about teacher–student relationships by consideringhow teachers’ day-to-day, typical academic practices may predict social preference. Observedemotional support on the CLASS indicates how teachers emotionally engage with children duringtheir regular instructional periods; teachers’ LC instruction and academic status hierarchies areintricately pertinent to academic lessons every day. This naturally embedded approach differs fromsome social–emotional learning curricula whereby teachers deliver specific lesson modules toimprove students’ social competence. Although there are well-validated curricula that demonstrateimprovements in children's peer relationships (Durlak, Weissberg, Dymnicki, Taylor, & Schellinger,2011), this study suggests an alternative method that may also hold value. Given pressure to improvechildren's academic test scores, some educators may be unwilling to devote effort or class time


towards social competence lessons to assist children's peer relationships. However, teachers whounderstand the social implications of their regular, day-to-day academic practices may be moreamenable to modifying the way that they teach academic lessons to achieve the additional goal ofimproving peer relationships.

4.1. Implications for school psychologists

Findings have implications for school psychologists who consult with teachers on how to promotestudents’ social inclusion. Typically, teachers refer children with externalizing behavior problems to childstudy teams, whomay conduct a functional analysis of children's behaviors to determine an individualizedintervention that helps children find more socially appropriate ways to meet their needs (Scott et al.,2005). It is assumed that if children improve their behavior, then better peer relationships will necessarilyfollow. Yet practitioners might broaden their focus to include ecological contributors to children's socialfunctioning, such as the teacher's practices.

Our results suggest that minimizing academic status hierarchies may assist children's peer relation-ships, but school psychologists may face challenges implementing this intervention given that manydistricts recommend (and mandate) ability grouping for instruction. In fact, all of the current studyclassrooms used some ability grouping, most typically for reading instruction. However, we speculate thatteachers nonetheless varied in promotion of academic status hierarchies within use of ability grouping;specifically, informal observations and anecdotal reports suggested that some teachers challenged childrenequally across ability groups, picked comparatively engaging activities, seemed equivalently enthusiasticwhen engaging with children of all ability groups, and rearranged children into heterogeneous groups forother activities. Thus, practitioners might consider ways to support teachers in delivering instruction thatreduces academic hierarchies even within ability grouping systems.

4.2. Study strengths and limitations

Strengths of this study include the multi-measure, multi-informant design in which teacher practiceswere assessed by both observations and self-report and in which children's externalizing behaviorssampled by observers. This study also used the gold standard of sociometric methods to determine socialpreference. Indeed, there was nearly complete independence between reporters in analytic models. Inaddition, although the study only covered one school year, it is an advance over work that correlatesteacher practices and student peer relationships at one time point. The longitudinal design allowsstatistical control of social preference in the fall when predicting social preference in spring, a method thathelps to determine whether fall levels of teacher practices may predict growth trajectories in children'ssocial preference.

One weakness of this study pertains to the sample of teachers and students. Our sample of 26 teacherswas demographically homogeneous and also relatively small. According to Mass and Hox (2005),conclusions should be considered tentative when based on fewer than 30 units at Level 2. Additionally,because a significant number of eligible teachers did not choose to take part in the study, it is unknown ifdata reflect an unrepresentative sample of teachers. Further, although there were few children withmissing data due to parents denying consent or because of excessive absenteeism, it is possible that theseare the children most at-risk for social and behavior problems. Although most IRB procedures stipulateexcluding children without active consent from providing sociometric nominations, this stipulation mayintroduce bias in that social preference scores are determined from the perceptions of a better-adjustedgroup of children. Both possibilities are supported by findings that children with active consent receivedmore liked nominations than did children with passive consent (but did not differ on disliked nominationsor externalizing behaviors).

Second, this study failed to consider nesting in schools in data analysis. Future work should considerways in which the schoolmay affect the teacher practices found to influence peer relationships. The level ofhierarchy within the teaching staff or the supportiveness of the principal may influence teachers’ levels ofstatus hierarchies and emotional tone with students in their own classrooms, which are then adopted bypeers in their treatment of one another.


Other limitations pertain to the measurement of variables. Some variables in this study displayed lowreliability (alpha or kappa below .70), and therefore caution should be exercised when interpreting results.An additional limitation pertains to the decision to leave the observational measure of studentexternalizing behavior untransformed. Although the use of untransformed scores may capture meaningfulvariability, there is a risk that extreme scores allot extra influence on findings. When transformed scoreswere used, the overall pattern of results remained the same, but some values dropped slightly tononsignificance. These findings suggest that teachers’ responses to the childrenwith extreme externalizingbehavior may be most influential on peers’ perceptions of those children.

4.3. Implications for future research

Future investigations might examine teacher influences on social preference in other age groups. Mostexisting literature has focused on elementary school students; however, there are suggestions thatteachers may still influence children's peer relationships in middle school, at least in Chinese samples(Chang et al., 2004). Although interactions between teacher practices and grade level were not found inthis study, these tests were statistically underpowered. Theoretically, adolescents may be unlikely tofollow teachers to determine whom they like and dislike, as part of the normative developmental processof seeking autonomy from adults. Alternatively, even though teachers may have little influence over thesocial preferences of adolescents, a teacher's small influence may still carry substantial implications foradjustment because of the importance of peer relationships during this age period (Buhrmester, 1990).

Future work might also suggest direction for interventions that target teachers to improve peerrelationships in their classrooms. A randomized trial of an interventionwould also allow causal conclusionsto be drawn that teacher practices may lead to changes in children's social preference and not vice versa. Inaddition, an intervention training teachers to increase the inclusiveness of the peer group has the potentialfor positive ripple effects extending beyond the children receiving treatment at themoment. Even after theintervention ceases, a lasting increase in peers’ tolerance may discourage disliking of new children theyencounter in subsequent years. A durable change in teacher practices may also prevent peer problems forstudents who are in that teacher's class in the future. Relative to interventions that target the behaviors ofone disliked child, addressing the classroom climate may be a potentially valuable intervention directionthat is cost-effective and reaches more youth because of preventative effects.

References

Abikoff, H., Gittelman, R., & Klein, D. F. (1980). A classroom observation code for hyperactive children: A replication of validity. Journalof Consulting and Clinical Psychology, 48, 555–565.

APA Learner-Centered Principles Workgroup (1997). Learner-centered psychological principles: A framework for school reform andredesign. Washington, DC: American Psychological Association.

Bagwell, C., Newcomb, A. F., & Bukowski, W. M. (1998). Preadolescent friendship and peer rejection as predictors of adult adjustment.Child Development, 69, 140–153.

Buhrmester, D. (1990). Intimacy of friendship, interpersonal competence, and adjustment in preadolescence and adolescence. ChildDevelopment, 61, 1101–1111.

Buhs, E. S., Ladd, G. W., & Herald, S. L. (2006). Peer exclusion and victimization: Processes that mediate the relation between peergroup rejection and children's classroom engagement and achievement? Journal of Educational Psychology, 98, 1–13.

Chang, L. (2003). Variable effects of children's aggression, social withdrawal, and prosocial leadership as functions of teacher beliefsand behaviors. Child Development, 74, 535–548.

Chang, L. (2004). The role of classroom norms in contextualizing the relations of children's social behaviors to peer acceptance.Developmental Psychology, 40, 691–702.

Chang, L., Liu, H., Fung, K. Y., Wang, Y., Wen, Y., Wen, Z., et al. (2007). The mediating and moderating effects of teacher preference onthe relations between students' social behaviors and peer acceptance. Merrill-Palmer Quarterly, 53, 603–630.

Chang, L., Liu, H., Wen, Z., Fung, K. Y., Wang, Y., & Xu, Y. (2004). Mediating teacher liking and moderating authoritative teaching onChinese adolescents' perceptions of antisocial and prosocial behaviors. Journal of Educational Psychology, 96, 369–380.

Cillessen, A. H. N., & Bellmore, A. D. (1999). Accuracy of social self-perceptions and peer competence in middle childhood. Merrill-Palmer Quarterly, 45, 650–676.

Cillessen, A. H. N., Bukowski, W. M., & Haselager, G. J. T. (2000). Stability of sociometric categories. In A. H. N. Cillessen, & W. M.Bukowski (Eds.), Recent advances in the measurement of acceptance and rejection in the peer system, Vol. 88. (pp. 75–93)SanFrancisco, CA: Jossey-Bass.

Cohen, E. G., & Lotan, R. A. (1995). Producing equal-status interaction in the heterogeneous classroom. American Education ResearchJournal, 32, 99–120.

Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimensions and types of social status: A cross-age perspective. DevelopmentalPsychology, 18, 557–570.


Curby, T. W., LoCasale-Crouch, J., Konold, T. R., Pianta, R. C., Howes, C., Burchinal, M., et al. (2009). The relations of observed pre-kclassroom quality profiles to children's achievement and social competence. Early Education and Development, 20, 346–372.

DeRosier, M. E., Kupersmidt, J. B., & Patterson, C. J. (1994). Children's academic and behavioral adjustment as a function of thechronicity and proximity of peer rejection. Child Development, 65, 1799–1813.

Dodge, K. A., Coie, J. D., Pettit, G. S., & Price, J. E. (1990). Peer status and aggression in boys' groups: Developmental and contextualanalyses. Child Development, 61, 1289–1309.

Dodge, K. A., Lansford, J. E., Burks, V. S., Bates, J. E., Pettit, G. S., Fontaine, R., et al. (2003). Peer rejection and social information-processing factors in the development of aggressive behavior problems in children. Child Development, 74, 374–393.

Donohue, K. M., Perry, K. E., & Weinstein, R. S. (2003). Teacher's classroom practices and children's rejection by their peers. Journal ofApplied Developmental Psychology, 24, 91–118.

Durlak, J. A., Weissberg, R. P., Dymnicki, A. B., Taylor, R. D., & Schellinger, K. B. (2011). The impact of enhancing students' social andemotional learning: A meta-analysis of school-based universal interventions. Child Development, 82, 405–432.

Enders, C. K. (2001). The performance of the full information maximum likelihood estimator in multiple regression models withmissing data. Educational and Psychological Measurement, 61, 713–740.

Enders, C. K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multi-level models: A new look at an old issue.Psychological Methods, 12, 121–138.

Erhardt, D., & Hinshaw, S. P. (1994). Initial sociometric impressions of attention-deficit hyperactivity disorder and comparison boys:Predictions from social behaviors and from nonbehavioral variables. Journal of Consulting and Clinical Psychology, 62, 833–842.

Hamre, B. K., & Pianta, R. C. (2001). Early teacher–child relationships and the trajectory of children's school outcomes through theeighth grade. Child Development, 72, 625–638.

Hamre, B. K., & Pianta, R. C. (2005). Can instructional and emotional support in the first-grade classroom make a difference forchildren at risk of school failure? Child Development, 76, 949–967.

Hamre, B. K., Pianta, R. C., Mashburn, A. J., & Downer, J. T. (2007). Building a science of classrooms: Application of the CLASS frameworkin over 4,000 U.S. early childhood and elementary classrooms. Retrieved 4/13/2010, from. http://www.fcd-us.org/resources/resources_show.htm?doc_id=507559.

Harrist, A. W., Zaia, A. F., Bates, J. E., Dodge, K. A., & Pettit, G. S. (1997). Subtypes of social withdrawal in early childhood: Sociometricstatus and social-cognitive differences across four years. Child Development, 68, 278–294.

Hoza, B., Mrug, S., Gerdes, A. C., Bukowski, W. M., Kraemer, H. C., Wigal, T., et al. (2005). What aspects of peer relationships areimpaired in children with Attention-deficit/ Hyperactivity Disorder? Journal of Consulting and Clinical Psychology, 73, 411–423.

Hughes, J. N., Cavell, T. A., & Wilson, V. (2001). Further support for the developmental significance of the quality of the teacher–student relationship. Journal of School Psychology, 39, 289–301.

Hughes, J. N., & Kwok, O. (2006). Classroom engagement mediates the effect of teacher–student support on elementary schoolstudents' peer acceptance: A prospective analysis. Journal of School Psychology, 43, 465–480.

Hymel, S., Wagner, E., & Butler, L. J. (1990). Reputational bias: View from the peer group. In S. R. Asher, & J. D. Coie (Eds.), Peer rejectionin childhood (pp. 156–186). New York: Cambridge University Press.

Lee, S. S., & Hinshaw, S. P. (2004). Severity of adolescent delinquency among boys with and without Attention Deficit HyperactivityDisorder: Predictions from early antisocial behavior and peer status. Journal of Clinical Child and Adolescent Psychology, 33, 705–716.

Mashburn, A. J., Pianta, R. C., Hamre, B. K., Downer, J. T., Barbarin, O. A., Bryant, D., et al. (2008). Measures of classroom quality inprekindergarten and children's development of academic, language, and social skills. Child Development, 79, 732–749.

Mass, C. J. M., & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling methodology. European Journal of Research Methodsfor the Behavioral and Social Sciences, 1, 86–92.

Masten,A. S.,Morison, P.,& Pellegrini,D. S. (1985).A revisedclassplaymethodof peer assessment.Developmental Psychology,21, 523–533.McAuliffe, M. D., Hubbard, J. A., & Romano, L. J. (2009). The role of teacher cognition and behavior in children's peer relations. Journal

of Abnormal Child Psychology, 37, 665–677.McCombs, B. L., Daniels, D. H., & Perry, K. E. (2008). Children's and teacher's perceptions of learner-centered practices, and student

motivation: Implications for early schooling. The Elementary School Journal, 109, 16–35.McCombs, B. L., Lauer, P. A., Bishop, J., & Peralez, A. (1997). Researcher test manual for the Learner-Centered Battery. Aurora, CO: Mid-

continent Regional Educational Library.Midgley, C., Maehr, M. L., Hruda, L. Z., Anderman, E., Anderman, L., Freeman, K. E., et al. (2000). Manual for the Patterns of Adaptive

Learning Scales. Unpublished manuscript, University of Michigan.Mikami, A. Y., Boucher, M. A., & Humphreys, K. (2005). Prevention of peer rejection through a classroom-level intervention in middle

school. Journal of Primary Prevention, 26, 5–23.Mikami, A. Y., Gregory, A., Allen, J. P., Pianta, R. C., & Lun, J. (in press). Effects of a teacher professional development intervention on

peer relationships in secondary classrooms. School Psychology Review.Mikami, A. Y., & Hinshaw, S. P. (2003). Buffers of peer rejection among girls with and without ADHD: The role of popularity with

adults and goal-directed solitary play. Journal of Abnormal Child Psychology, 31, 381–397.Mikami, A. Y., Lerner, M. D., & Lun, J. (2010). Social context influences on children's rejection by their peers. Child Development

Perspectives, 4, 123–130.Newcomb, A. F., Bukowski, W. M., & Pattee, L. (1993). Children's peer relations: A meta-analytic review of popular, rejected,

neglected, controversial, and average sociometric status. Psychological Bulletin, 113, 99–128.Parker, J. G., & Asher, S. R. (1987). Peer relations and later personal adjustment: Are low-accepted children at risk? Psychological

Bulletin, 102, 357–389.Peets, K., Hodges, E. V. E., & Salmivalli, C. (2008). Affect-congruent social–cognitive evaluations and behaviors. Child Development, 79, 170–185.Pianta, R. C., La Paro, K. M., & Hamre, B. K. (2007). Classroom Assessment Scoring System—CLASS. Baltimore, MD: Brooks Publishing Co..Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models: Applications and data analysis methods. Thousand Oaks, CA: Sage

Publications.Raudenbush, S. W., Bryk, A. S., & Congdon, R. (2004). HLM 6 for Windows (computer software). Lincolnwood, IL: Scientific Software

International, Inc.Rimm-Kaufman, S. E., La Paro, K. M., Downer, J. T., & Pianta, R. C. (2005). The contribution of classroom setting and quality of

instruction to children's behavior in kindergarten classrooms. The Elementary School Journal, 105, 377–394.

http://www.fcd-us.org/resources/resources_show.htm?doc_id=507559

http://www.fcd-us.org/resources/resources_show.htm?doc_id=507559


Roseth, C. J., Johnson, D. W., & Johnson, R. T. (2008). Promoting early adolescents' achievement and peer relationships: The effects ofcooperative, competitive, and individualistic goal structures. Psychological Bulletin, 134, 223–246.

Salinas, M. F., & Garr, J. (2009). Effect of learner-centered education on the academic outcomes of minority groups. Journal ofInstructional Psychology, 36, 226–237.

Scott, T. M., McIntyre, J., Liaupsin, C., Nelson, C. M., Conroy, M., Payne, L. D., et al. (2005). An examination of the relation betweenfunctional behavior assessment and selected intervention strategies with school-based teams. Journal of Positive BehaviorInterventions, 7, 205–215.

Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in assessing rater reliability. Psychological Bulletin, 86, 420–428.Spybrook, J., Raudenbush, S. W., Congdon, R., & Martinez, A. (2009). Documentation for the Optimal Design software.Unpublished

manuscript, Western Michigan University.Stormont, M. (2001). Social outcomes of children with AD/HD: Contributing factors and implications for practice. Psychology in the

Schools, 38, 521–531.Stormshak, E. A., Bierman, K. L., Bruschi, C., Dodge, K. A., & Coie, J. D.Conduct Problems Prevention Research Group. (1999). The

relation between behavior problems and peer preference in different classroom contexts. Child Development, 70, 169–182.Taylor, A. R. (1989). Predictors of peer rejection in early elementary grades: Roles of problem behavior, academic achievement, and

teacher preference. Journal of Clinical Child Psychology, 18, 360–365.Weinstein, R. S. (2002). Reaching higher: The power of expectations in schooling. Cambridge, MA: Harvard University Press.White, K. J., Jones, K., & Sherman, M. D. (1998). Reputation information and teacher feedback: Their influences on children's

perceptions of behavior problem peers. Journal of Social and Clinical Psychology, 17, 11–37.White, K. J., & Kistner, J. (1992). The influence of teacher feedback on young children's peer preferences and perceptions.

Developmental Psychology, 28, 933–940.Zumwalt, K., & Craig, E. (2005). Teachers' characteristics: Research on the demographic profile. In M. Cochran-Smith, & K. M. Zeichner

(Eds.), Studying teacher education: The Report of the AERA Panel on Research and Teacher Education (pp. 111–156). Mahwah, NJ:Lawrence Erlbaum Associates, Inc.

teacher practices as predictors of children's classroom social preference

Documents