You are on page 1of 24

Predicting School Adjustment in Early Elementary School: Impact of TeacherChild Relationship Quality and Relational Classroom Climate

Evelien Buyse Karine Verschueren Pieter Verachtert Jan Van Damme


Katholieke Universiteit Leuven

Abstract
This longitudinal study evaluated the impact of dyadic and classroom-level teacher-child relationship quality in rst grade on childrens psychosocial and academic adjustment in rst (N 3,784), second (N 3,666), and third (N 3,582) grade, controlling for several child features, namely, child demographics and childrens initial levels of adjustment in kindergarten. Results of multilevel hierarchical regression analyses showed that rst-grade dyadic relationship variables (i.e., teacher-child conict and closeness) as well as classroom relational climate variables (i.e., the average level of teacher-child conict and closeness in the classroom) were associated with childrens psychosocial adjustment in the rst years of primary school. Associations between rst-grade dyadic relationship quality and classroom relational climate, on the one hand, and academic achievement on the other, however, were negligible.

The Elementary School Journal Volume 110, Number 2 2009 by The University of Chicago. All rights reserved. 0013-5984/2009/11002-0001$10.00

Understanding how children adapt to school has been an important objective for researchers interested in the promotion of competence and the prevention of educational and psychological maladjustment. This objective has spurred research on the determinants of early school adaptation (Ladd, Birch, & Buhs, 1999). Historically, researchers have dened school adjustment in terms of childrens academic progress or achievement. The construct of school adjustment has thus been dened rather narrowly, and, as a consequence, the search for its antecedents has also been limited (Birch & Ladd, 1996). Most investigators have focused on internal characteristics of the child as determinants of early school adjustment, such as gender and intelligence. Some researchers tried to elaborate upon these ear-

120

THE ELEMENTARY SCHOOL JOURNAL

lier models of school adjustment by restating the concept of school adjustment itself and rethinking its hypothesized determinants (Birch & Ladd, 1996; Ladd, 1989, 2003). According to them, the concept of school adjustment should, for example, be expanded to include dimensions of psychosocial adjustment as well (e.g., Birch & Ladd, 1997; Ladd, 2003; Ladd et al., 1999; Pianta, Steinberg, & Rollins, 1995). In our study, we followed this broader denition, focusing on reading and mathematics skills as aspects of academic achievement, and including childrens aggressive behavior, popularity with peers, and feelings of wellbeing at school as aspects of psychosocial adjustment at school. In line with the broader denition, new ideas arose considering the possible determinants of school adjustment. One such perspective has been termed a person by environment model, in which successful school adjustment is seen as originating both in the child and in the surrounding interpersonal environment, such as the childs relationships with teachers (Birch & Ladd, 1996; Ladd, 2003; Ladd et al., 1999). Over the past 15 years, increasing attention has been devoted to the role of relationships between children and teachers in inuencing early school adjustment (e.g., Pianta, 1999, 2006; Pianta, Hamre, & Stuhlman, 2003). In these studies, two dimensions are often seen as operationalizations of teacher-child relationship quality, namely, teacher-child closeness and teacher-child conict (Pianta et al., 1995, 2003). Closeness in the teacher-child relationship may function as a support in the school environment (Birch & Ladd, 1996). It is reected in the degree of warmth and open communication between a teacher and a child. It encompasses the extent to which children seem comfortable approaching the teacher, talking about their feelings and experiences, and using the teacher as a source of support and comfort when upset. Having warm and open communication with the teacher may facilitate positive feelings to-

ward school. From close relationships with teachers, children may derive potential resources, such as emotional support and security, which may enhance positive behaviors and exclude more negative behaviors (such as aggression) in social contexts such as classrooms (Birch & Ladd, 1998; Ladd et al., 1999). Children may then use their teachers as resources for other social relationships, including their relationships with peers (Howes, 2000). A supportive teacher-child relationship may also motivate children to become more engaged in the school environment. In this manner, closeness may encourage young childrens learning and performance in school as well (Birch & Ladd, 1996, 1997; Howes, 2000). In line with these hypotheses, ndings from previous studies conducted in kindergarten and/or the early grades of elementary school indeed show that closeness in the teacher-child relationship functions as a support for young children in the school environment. Greater closeness in the teacher-child relationship is associated with better adjustment to school, for instance, with more positive feelings about school (e.g., Birch & Ladd, 1997), fewer behavioral problems, more behavioral competencies and social skills (e.g., Birch & Ladd, 1998; Howes, 2000; Hughes, Cavell, & Jackson, 1999; Pianta et al., 1995; Pianta & Stuhlman, 2004; Silver, Measelle, Armstrong, & Essex, 2005), and higher academic achievement (e.g., Birch & Ladd, 1997; Ladd et al., 1999). Conict in the teacher-child relationship may function as a stressor for children and may impair successful adjustment to school (Birch & Ladd, 1996; Ladd et al., 1999). Conicted relationships are characterized by discordant interactions and a lack of rapport between teachers and children. As a potential stressor in the school environment, teacher-child conict may be emotionally upsetting to young children, yielding negative behaviors. Relationship conict may, for example, exacerbate aggressive behavior (Birch & Ladd, 1998). Friction between teacher and child may also foster negative feelings about school or
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

121

disengagement concerning school matters, causing academic problems as well (Birch & Ladd, 1996). In line with the idea that teacher-child conict may function as a stressor for childrens early school adjustment, empirical research has conrmed that it predicts more negative feelings about school (e.g., Birch & Ladd, 1997), more behavioral problems, fewer behavioral competencies and social skills (e.g., Birch & Ladd, 1997, 1998; Doumen et al., 2008; Howes, 2000; Hughes et al., 1999; Mantzicopoulos, 2005; Pianta et al., 1995; Pianta & Stuhlman, 2004; Silver et al., 2005), and poorer academic performance or readiness (e.g., Hamre & Pianta, 2001; Ladd et al., 1999) in kindergarten and/or the rst years of formal schooling. In sum, teacher-child closeness and conict are found to play a signicant role in regulating the development of the children in these relationships. This regulatory role of both features of teacher-child relationships is thought to be especially relevant in the early years of formal schooling, because those years comprise a sensitive period for development in school (Alexander & Entwisle, 1988). Up until third grade, when school trajectories are generally well established, relationships with teachers may organize and provide direction to development (Pianta et al., 1995). Therefore, in our study we investigated the signicance of rst-grade teacher-child closeness and conict in the prediction of rst-, second-, and third-grade academic and psychosocial adjustment at school. As described above, some research has already been conducted on this topic. In many of these studies, however, principal predictors of the outcomes were not controlled for. Moreover, if control variables were included, this control has rarely been comprehensive, including a set of multiple relevant predictors of outcomes. Therefore, in our study we controlled for demographic factors that have been shown to determine childrens academic and psychosocial adjustment, including childrens gen-

der, socioeconomic status (SES), and ethnic background. In several international studies, girls, children from families with higher SES, and children of native parents adjusted better to school, academically as well as psychosocially (e.g., Hamre & Pianta, 2001; Ladd et al., 1999). In Belgium as well, these are consistent ndings (e.g., Groenez, Van den Brande, & Nicaise, 2003; Hirtt, Nicaise, & De Zutter, 2007). In addition to controlling for these background features, we also controlled for initial levels of childrens academic and psychosocial adjustment, which was measured while the children were in kindergarten. In this manner, we could evaluate whether teacherchild closeness and teacher-child conict provide a unique prediction of child outcomes beyond child demographics and initial levels of adjustment. In addition to child characteristics, it is also worthwhile to control for classroom interactional features when investigating the association between teacher-child relationship quality and child adjustment. Dyadic teacher-child relationships are embedded within the overall interactional environment of the classroom (Birch & Ladd, 1997; Howes, 2000; Pianta et al., 2003). In classroom effectiveness literature, various dimensions of the interactional quality of the classroom, such as instructional and emotional climate, have proved important for childrens school outcomes (e.g., Burchinal et al., 2000; PeisnerFeinberg et al., 2001; Pianta, La Paro, Payne, Cox, & Bradley, 2002; Wilson, Pianta, & Stuhlman, 2007). In our study, the average level of teacher-child closeness and conict within each classroom was used as an indicator of relational classroom climate. As in the studies of Birch and Ladd (1997) and Howes (2000), we controlled for these variables when predicting school adjustment outcomes from dyadic teacher-child closeness and conict. It has been hypothesized that bad classroom composition (e.g., high average levels of conict or low average levels of closeness) places an extra burden or stress on teachers, which can result in teachers behav-

122

THE ELEMENTARY SCHOOL JOURNAL

ing more negatively with children, with associated consequences for childrens adjustment (e.g., Mantzicopoulos, 2005; Pianta et al., 2003; Rimm-Kaufman, Pianta, & Cox, 2000). However, this hypothesis has seldom been tested explicitly. In the aforementioned studies (Birch & Ladd, 1997; Howes, 2000) in which classroom relational composition variables were included, effects were indeed detected of the relational classroom climate on childrens school adjustment in the early years of formal schooling. Furthermore, not only concurrent but also prospective associations have been found between relational variables and young childrens school adjustment in subsequent school years (e.g., Hamre & Pianta, 2001; Howes, 2000; Pianta et al., 1995; Silver et al., 2005). However, associations between relationship variables and childrens outcomes may change over time and tend to be the greatest when the outcomes and relationships were rated in the same school year (Pianta & Stuhlman, 2004). In our study, relationship quality was evaluated in rst grade, but not in second or third grade. We therefore expected the associations between rst-grade relational variables and indicators of childrens school adjustment to be larger in rst grade than in second and third grade. In sum, in the present study we evaluated the predictive value of dyadic teacherchild closeness and conict in rst grade for childrens academic and psychosocial adjustment over the rst 3 years of elementary school, controlling for several child features, namely, child demographics and childrens initial levels of adjustment in kindergarten. Concerning the control variables, we expected childrens levels of adjustment in kindergarten to be positively related to childrens levels of adjustment in early elementary school. Furthermore, we hypothesized that boys, children from families with lower SES, and children from immigrant families would perform worse on academic tests and score lower on psychosocial teacher ratings. When evaluating the

association between teacher-child relationship quality and school adjustment, we also controlled for rst-grade relational classroom features, that is, the average level of teacher-child closeness and conict within each classroom. We expected a more positive relational climate, characterized by high average class levels of closeness or low average class levels of conict, to be associated with better school adjustment. With regard to the key predictors of our study, we expected higher teacher-child closeness and lower teacher-child conict to foster childrens school adjustment over and above relational classroom features and child-level covariates. Finally, we studied the differential effects of dyadic and classroom relational variables on childrens school adjustment in the same and subsequent school years. We hypothesized that we would nd the strongest associations between relationship variables and adjustment in rst grade because relationship quality was evaluated in that school year.

Method
Participants The study described here was part of an ongoing large-scale longitudinal study in Flemish education. It was designed to describe and explain childrens educational careers throughout elementary education.1 To this purpose, a random stratied sample of 122 schools (215 classrooms) was selected. Stratication was based on educational network and school size.2 This sample was found to be representative of the entire Flemish school population in terms of the applied stratication criteria (Verhaeghe, Maes, Gombeir, & Peeters, 2002). Data collection started on September 1, 2002, when children entered their kindergarten year (Verachtert, 2008). A large sample of 3,798 kindergartners was recruited. Of all children, 50.5% were boys. At the beginning of the school year, the mean age of the kindergartners was 5 years and 3 months. To assess the SES of the childrens
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

123

families, a score was compiled based on the educational level of both parents, their professional status, and household income (n 3,461; see Reynders, Nicaise, & Van Damme, 2005, for more detailed information on this variable). Most parents who lled out the questionnaires completed high school (38% of the mothers and 41% of the fathers) or in addition attended higher education (43% of the mothers and 39% of the fathers). Parents nationality at birth was used as an indicator of ethnic background (n 3,296). Children were classied into one of three categories: both parents had Belgian nationality at birth (Belgian, 80%), both parents had a foreign nationality (foreign, 11%), or one parent had Belgian nationality while the other had a foreign nationality (mixed, 9%). When children arrived in rst grade (20032004), second grade (2004 2005), and third grade (20052006), data were available for 3,784 children, 3,666 children, and 3,582 children, respectively, divided over 213 classrooms (117 schools), 202 classrooms (118 schools), and 200 classrooms (116 schools), respectively. Hence, for some children, scores on the outcome variables were missing in rst, second, or third grade because those children did not progress regularly through the successive grades or because they did not stay in a school participating in the study. Preliminary analysis revealed that children for whom outcomes were missing in second and/or third grade differed from children who had complete data throughout the study. In particular, dropouts differed from nondropouts with regard to SES (t(3448) 14.26, p .001), immigrant status ( 2(2) 41.21, p .001), and initial levels of adjustment, concerning aggressive behavior (t(3786) 9.48, p .001), popularity with peers (t(3788) 13.60, p .001), feelings of well-being (t(3789) 9.31, p .001), and language achievement (t(3695) 22.28, p .001) in kindergarten. Higher SES, lower percentage of immigrants, and higher levels of initial adjustment were reported for nondropouts. Effect sizes are rather modest, with eta-squared ranging between 1% and 12%.

Scores on mathematics achievement in kindergarten did not differ between dropouts and nondropouts (t(3685) 1.67, ns). Furthermore, because of additional missing data on some predictor variables, the sample size is sometimes smaller throughout the analyses. The exact number of cases on which results are based is reported for every analysis separately. Instruments Relationship quality. To evaluate the quality of the teacher-child relationship in rst grade, we used a short Dutch version of the Student-Teacher Relationship Scale (STRS; Pianta, 2001). The validity of the STRS has been demonstrated in relation to a range of social and academic outcomes (e.g., Pianta, 2001). The short version (Buyse, Verschueren, Doumen, Van Damme, & Maes, 2008; Cornelissen & Verschueren, 2001) consists of eight items and was used to assess teachers perceptions of two features of their relationships with their pupils. The conict subscale comprises four items that tap the extent to which the teacher-child relationship is characterized by disharmonious interactions (e.g., This child and I always seem to be struggling with each other). The closeness subscale is a four-item index of the degree of warmth and open communication present in the teacher-child relationship (e.g., This child openly shares his/her feelings and experiences with me). In the study of Cornelissen and Verschueren (2001), the correlations between the scores for both shortened subscales and the complete scales were satisfactory: .92 for relational closeness and .91 for relational conict. In the same study, high values of internal consistency were found for both scales of the shorter version of the STRS (conict: .87, closeness: .84), which was conrmed in our study (conict: .82, closeness: .88). Moreover, conrmatory factor analysis showed that closeness and conict are two different constructs (Cornelissen & Verschueren, 2001;

124

THE ELEMENTARY SCHOOL JOURNAL

see also Koomen, Verschueren, & Pianta, 2007). In line with these results, the correlation between both subscales in the present study was signicant, but not substantively, justifying the inclusion of a separate conict and closeness subscale (r(377l) .20, p .001). Relational classroom climate. Firstgrade relational classroom climate variables were obtained by averaging scores on teacher-child closeness and teacher-child conict for all children in each classroom. Again, the correlation between both subscales was signicant, but not substantively, justifying the inclusion of separate subscales for average classroom closeness and conict (r(215) .32). Further (multilevel) analyses showed classroom average closeness to predict dyadic closeness ( .34, p .001), but not conict ( .03, ns), whereas classroom average conict predicted dyadic conict ( .49, p .001), but not closeness ( .05, ns).3 Psychosocial adjustment. Teachers in kindergarten and primary school (i.e., rst, second, and third grade) rated childrens psychosocial adjustment in the classroom on several short subscales. All items on the different subscales were rated on a sixpoint Likert scale with values ranging from 1 (does not apply at all) to 6 (applies completely). Three subscales were selected for this study. For each subscale, scale scores were computed by averaging the item scores. First, the aggressive behavior subscale was derived from the Child Behavior Scale (CBS), developed by Ladd and Prolet (1996). Supportive evidence for the validity of the CBS has been obtained, including signicant correlations in expected directions with observations and peer ratings for behavior (Doumen et al., 2008; Ladd & Prolet, 1996). For use in this study, the subscale was reduced from seven to four items (e.g., Threatens other children). This reduction was based on the results of the original factor analysis, performed and reported by Ladd and Prolet (1996). Ex-

cluded items were the items that had the lowest loadings on the common subscale factor. Internal consistency of this subscale was good, with Cronbachs alpha coefcient equaling .90, .92, .94, and .92 in kindergarten, rst, second, and third grade, respectively. Correlations across grades ranged between .48 and .51 ( p .001). Second, the popularity-with-peers subscale was selected from the PRIMA study (Driessen, van Langen, & Vierke, 2000; Jungbluth, Roede, & Roeleveld, 2001). The items on this subscale resemble the excluded-bypeers subscale from the CBS (Ladd & Prolet, 1996), but because face validity of the popularity scale from the PRIMA study is higher than face validity of the CBS subscale, and because the subscale is shorter, we preferred to work with the PRIMA scale. The scale consists of four items (e.g., Gets along well with peers) with sufcient internal consistency, as Cronbachs alpha coefcients equaled .79, .83, .84, and .88 in kindergarten and rst, second, and third grade, respectively, in our study. This is in accordance with the internal consistency reported in the PRIMA study ( .84). Correlations across grades ranged between .39 and .44 ( p .001). Third, a scale measuring feelings of wellbeing at school was developed for this study, based on another PRIMA scale (Driessen et al., 2000; Jungbluth et al., 2001) and on the school-liking subscale of the Teacher Rating Scale of School Adjustment (TRSSA; Ladd, 1992). The scale consists of four items (e.g., Enjoys most of the classroom activities), with Cronbachs alpha equaling .82, .88, .87, and .88 in kindergarten and rst, second, and third grade, respectively. Correlations across grades ranged between .20 and .33 (p .001). Academic achievement. The wordreading test (Moelands, Kamphuis, & Rymenans, 2003; Moelands & Rymenans, 2003), a Flemish version of a well-validated Dutch test (see Evers et al., 2002, for evidence concerning validity), was administered in rst, second, and third grade. The test contains three reading charts with
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

125

words of increasing difculty. For each chart, the child was asked to read as many words as accurately as he/she could for 1 minute. The number of correctly read words was recorded. Pearson correlation coefcients between the scores for the three reading charts were high in rst grade (.90 .95), second grade (.89 .94), and third grade (.86 .92). Therefore, scores on the three separate reading charts were averaged to yield a composite reading score for each grade. Correlations between the composite reading scores across grades ranged between .57 and .87. At the end of kindergarten, children were administered a shortened form of the language test Taal voor Kleuters voor Vlaanderen (Kindergartners Language Achievement Test for Flanders; Citogroep, 2003), which in turn is an adapted version of a widely used Dutch language prociency test, Taal voor Kleuters (van Kuijk, 1996). The adaptations that were made to the original test are described by Ponjaert-Kristoffersen, Andries, Celestin-Westreich, and Samaey (2000) and Verachtert (2003). The test that was used in this study consisted of ve different subtests (each including eight items): listening comprehension, sound and rhyme, auditory sequencing, literacy knowledge, and sound blending. The internal consistency of the test was good ( .86; Verachtert, 2003). Curriculum-based mathematics achievement tests specically designed for use in this study were administered. Tests for each grade were constructed in such a way that they had a considerable number of items in common with the tests for the preceding and following grades. Hence it was possible to convert the raw mathematics scores to vertically equated Item Response Theory (IRT) based scale scores (Lord, 1980). This conversion enabled the comparison of mathematics scores across time. The mathematics tests administered in kindergarten and rst grade consisted of 40 items, whereas the test for second and third grade included 50 items and 60 items, respectively (Verachtert, 2008). Internal consistencies, as measured by Cron-

bachs alpha coefcient, were .92 in kindergarten, rst, and second grade, and .94 in third grade. The kindergarten mathematics test assessed a number of skills that are often included in denitions of number sense (Berch, 2005; Malofeeva, Day, Saco, Young, & Ciancio, 2004), including comparing magnitudes, counting, and understanding mathematics concepts. The tests in primary school not only contained items on number sense, but also applied items on arithmetic word problems, estimation, number decomposition, number series, geometry, and/or mental arithmetic. Correlations across grades ranged between .81 and .83. Procedure Information on child demographics was initially collected through a kindergarten parent questionnaire distributed among the children in February of 2003. Additional data-collection efforts were performed at the end of kindergarten and during the second and third year of the study in order to retrieve the information for the children whose parents did not return the initial questionnaire. Teachers lled out the questionnaires concerning childrens psychosocial adjustment in February of each school year. In rst grade, teachers additionally rated teacherchild relationship quality items in February. Academic achievement tests were administered by the researchers in May of each school year. The word-reading test was administered for each child individually. The language test in kindergarten and the mathematics tests were all untimed and were administered collectively. The sizes of the testing groups, however, differed between kindergarten and primary school. Because kindergartners tend to have little experience in taking paper-and-pencil tests, the kindergarten mathematics and language tests were administered to small groups of about six to eight children instead of to the entire class. This decision was taken to ensure that test administrators could provide enough guid-

126

THE ELEMENTARY SCHOOL JOURNAL

ance to these young children when they took the tests. From the start of primary school, all mathematics tests were administered to all children in the entire classroom collectively (Verachtert, 2008). Data Analysis All explanatory variables were centered around the grand mean. Among the predictor variables included in this study, variables from at least two different levels can be identied: child-level variables (i.e., child demographics, initial adjustment in kindergarten, and teacher-child relationship quality) and class-level variables (i.e., relational classroom climate variables). Because we wanted to evaluate child and class characteristics as possible predictors of childrens adjustment in primary school, data were analyzed by means of multilevel modeling techniques (Goldstein, 1995), making use of the software program MLwiN (Rasbash, Charlton, Browne, Healy, & Cameron, 2005). These techniques are especially designed to analyze variables from different levels simultaneously (Hox, 2002). Because we had data available for three subsequent grades on all outcome variables, time can be considered an additional predictor at an additional level in the multilevel models (Singer & Willett, 2003). Hence, three levels of information are involved in this study: time (level 1), nested within children (level 2), and nested within classes (level 3).4 An attractive feature of multilevel models in the analysis of longitudinal data is the treatment of missing data (Hedeker & Gibbons, 1997; Singer & Willett, 2003). Specically, subjects who are missing at a given wave are not excluded from the analysis, and the model estimates the subjects trend across time on the basis of whatever data the subject has, augmented by the time trend that is estimated for the sample as a whole and effects of all covariates in the model (Hedeker & Gibbons, 1997, p. 65). Therefore, if subject attrition is related to previous performance,

in addition to other observable subject characteristics, as is the case in our study (see comparison of dropouts and nondropouts in the Participants section above), multilevel models (making use of the maximumlikelihood estimation) provide valid statistical inferences for the model parameters. An overview of the subsequent models that have been tested to predict childrens school adjustment across the early years of elementary school is given in Figure 1. Preliminarily, for each outcome, a threelevel model was tested involving time as the only predictor (this became the baseline model). In this manner, we could evaluate the general trajectories for all children within all classes during the rst three grades of primary school concerning psychosocial and academic adjustment. In the baseline model, and across all further models evaluated, the effect of time was allowed to vary across classes and across children. As can be seen in Appendix A, four additional random parameters (i.e., the covariance term CONS,TIME and the variance term TIME2 at the class and the child level) were therefore estimated, compared to a model in which the effect of time would have been xed (in which case only 2 CONS would have been estimated at each level). We included these extra parameters because we were interested in the possible differential effect of classroom- and childlevel variables across time. To address our research goals, we conducted multiple hierarchical regression analyses, starting from the baseline model for each of the outcomes. In a rst step (Model A), childrens initial level of adjustment in kindergarten, corresponding to the outcome predicted, as well as child demographics (i.e., gender, SES, and ethnic background), were entered as (child-level) predictors. In a second step (Model B), the average level of rst-grade teacher-child conict and closeness within classrooms were added as (class-level) predictors to Model A, already containing childrens initial level of adjustment and child demographics as predictors. In a third step
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

127

FIG. 1.Overview of the subsequent models in the prediction of childrens school adjustment across the early years of elementary school. The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B included relational climate variables as (class-level) predictors, next to the predictors that were included in Model A. In Model C, dyadic teacher-child relationship variables were added as (child-level) predictors, next to the predictor variables that were included in Model B.

(Model C), the individual level of rstgrade teacher-child conict and closeness within teacher-child dyads were added as (child-level) predictors to Model B, already containing childrens initial level of adjustment, child demographics, and class-level relational climate variables as predictors. As mentioned above, we allowed the effect of time to vary across classes and across children, as we were interested in the possible differential effect of classroomand child-level variables across time. More specically, we evaluated the differential effect of classroom relational variables and dyadic teacher-child relationship closeness and conict across subsequent school years. Therefore, interaction terms between these relational variables, on the one hand, and the time variable on the other, were added to Models B and C. If an interaction was signicant, simple slope analyses were performed with the classroom and/or dyadic relational variable(s) predicting the outcome for each grade separately (based

on the procedure described by Aiken & West, 1991). Note that random parameters indicating random slopes for the effect of time across children or classes need not be signicant to justify evaluation of these interaction effects. Evaluation of cross-level interactions can be based on a theoretical argument formulated before looking at the data and can therefore be tested irrespective of whether a random slope has been found (Snijders & Bosker, 1999).

Results
As shown in Table 1, results from the baseline model indicate that childrens feelings of well-being at school, as rated by their teachers, decrease over the rst 3 years of primary school. Childrens teacher-rated aggressive behavior also (slightly) decreases over time. Reading and mathematics competencies, on the other hand, increase over the years. Finally, no change was detected over time concerning chil-

TABLE 1. Predicting Childrens Psychosocial and Academic Adjustment in Grades 13 by Initial Levels of Adjustment in Kindergarten, Child Demographics, Relational Classroom Climate, and Dyadic Teacher-Child Relationship Quality Popularity (n 7,106)a R2 .01 .13/.15/.15 .06/.07/.07 .40*** .70*** .12/.08/.05 R2 R2 .71*** .47/.47/.43 Well-Being (n 7,106)a Reading (n 6,983)b Mathematics (n 6,994)c R2

Aggressive Behavior (n 7,107)a R2

.04*

.23/.26/.26

128 .07/.03/.00 .13*** .09*** .09*** .10*** .08/.03/.00 .26*** .18*** .11*** .04 .43*** .13*** .07*** .02 .09/.08/.07 .24*** .16*** .10** .28*** .24*** .20*** .15*** .20*** .30*** .27*** .14*** .00 .36*** .18*** .08*** .02 .70*** .18*** .09*** .01 .07/.07/.06

.45*** .14*** .09*** .01 .01

.36*** .05*** .16*** .00 .01

.19*** .08*** .13*** .01 .03

.19*** .00 .09*** .02 .06***e

.45*** .10*** .09*** .02 .04**f .00/.00/.00 .02 .03* .01

.09/.03/.00

.18*** .03

.42*** .26*** .15*** .03 .01

.01

.11/.10/.08

.00/.00/.00 .03** .01 .01

Baseline model: Time Model A Initial level kindergarten Genderd SES Ethnicitymixed EthnicityBelgian Model B Class average conict Class average closeness Interactions time: Class average conict grade 1 grade 2 grade 3 Class average closeness grade 1 grade 2 grade 3 Model C Teacher-child conict Teacher-child closeness Interactions time: Teacher-child conict grade 1 grade 2 grade 3

.35*** .01

.39*** .51*** .33*** .16***

Teacher-child closeness grade 1 grade 2 grade 3 R2 total .43/.39/.34 .29/.26/.25 .21/.17/.13 .12/.08/.05 .47/.47/.43

.07

.48*** .29*** .16*** .04***

.52*** .33*** .19*** .05**

.01

NOTE.R and R are reported for each outcome in rst grade/in second grade/in third grade. In the prediction of aggressive behavior, popularity with peers, and feelings of well-being at school, data were available on all predictor variables for 2,657 children, yielding (2,657 3 school years ) 7,971 data points of information. Over the 3 school years, 864, 865, and 865 pieces of information were missing on aggressive behavior, popularity with peers, and feelings of well-being at school, respectively (leading to n 7,107, n 7,106, and n 7,106 for these outcomes, respectively). bIn the prediction of reading, data were available on all predictor variables for 2,672 children, yielding 8,016 data points of information. Over the 3 school years, 1,033 pieces of information were missing on reading (leading to n 6,983). cIn the prediction of mathematics, data were available on all predictor variables for 2,660 children, yielding 7,980 data points of information. Over the 3 school years, 986 pieces of information were missing on reading (leading to n 6,994). d0 boys; 1 girls. eWithout another predictor in the model than ethnicity (except for time from the previous model), equals .03 ( p .05). fWithout another predictor in the model than ethnicity (except for time from the previous model), equals .18 ( p .001). *p .05. **p .01. ***p .001.

129

130

THE ELEMENTARY SCHOOL JOURNAL

drens popularity with peers. Estimation of the random part of the baseline model showed that the variance in intercepts ( CONS2) at the class and child level is significant for every outcome (see Appendix A). When initial levels of childrens psychosocial or academic adjustment and childrens demographics were simultaneously added to the baseline model as predictors (Model A), the deviance of the total model reduced signicantly for every outcome variable (see Appendix A). With regard to the specic predictors (see Table 1), we rst notice that all aspects of childrens adjustment are substantially predicted by previous measures of adjustment. More aggressive behavior, popularity with peers, and feelings of well-being at school as rated by the kindergarten teacher predict more aggressive behavior, popularity with peers, and feelings of well-being at school as rated by the teachers in primary school. Furthermore, more language and mathematics competencies in kindergarten predict better reading and mathematics skills in primary school. Second, teachers in primary school rate girls as more popular with their peers, as feeling better at school, and as less aggressive than boys. Boys perform better at mathematics tests in the rst years of primary education, but no gender differences are detected concerning reading. Third, children from families with a higher SES are rated by their teachers as more popular with their peers and as feeling better at school. Higher SES is also related to less teacher-rated aggressive behavior, better reading skills, and better mathematics performance. Finally, when controlling for other background characteristics and the initial level of adjustment, we see that children whose parents both had Belgian nationality at birth perform worse in reading and mathematics than the children whose parents both had a foreign nationality. Yet, post hoc analyses including ethnic background as the sole child predictor show that Belgians score signicantly higher on both tests. Furthermore, differences be-

tween the three categories referring to ethnic status were signicant with regard to SES (F(2, 3244) 114.81, p .001) and with regard to kindergarten levels of language achievement (F(2, 3233) 129.45, p .001), with eta-squared equaling 8% and 7%, respectively, favoring Belgian children. No signicant differences were detected between ethnic categories concerning mathematics achievement in kindergarten (F(2, 3225) 2.62, ns). Taken together, these post hoc analyses suggest that the negative (univariate) effect of immigrant status on academic development during the rst 3 years in elementary school may mainly be due to its association with low SES and lower kindergarten levels of (language) achievement. Indeed, when controlling for these correlates, the negative effect of immigrant status disappeared and even became positive. After taking all child characteristics into account simultaneously in Model A, a substantial amount of variance is explained in each of the outcomesin rst as well as in second and third grade (see Table 1). Yet a signicant amount of variance is still situated at the class and child level for every outcome variable (see Appendix A). In an attempt to explain the remaining differences between classes, we added the two rst-grade relational classroom climate variables to the models already containing childrens initial levels of adjustment and demographic features (Model B). Overall, the deviance of the total model reduced signicantly in the prediction of every outcome, except in the prediction of reading. Therefore, Model B was not reported for this outcome variable. With regard to the other outcomes, rst, we see that higher average levels of teacher-child conict in rst grade are associated with lower levels of psychosocial adjustment. Specically, higher average levels of conict are associated with more teacherrated aggressive behavior and with less teacher-rated popularity with peers and feelings of well-being at school. Second, higher average levels of teacher-child closeness in rst grade are associated with better psychoDECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

131

FIG. 2.Differential effect of class average conict on aggressive behavior for the subsequent grades.

social adjustment in the classroom (i.e., greater popularity with peers and more feelings of well-being at school). First-grade relational classroom variables explain 7% to 9% of the variance in childrens psychosocial adjustment in rst grade and 3% in second grade, beyond all other predictors already in the model (see Table 1). In third grade, no impact was detected from rst-grade relational climate variables on childrens psychosocial adjustment. Simple slope analyses, which yield a better insight in the differential effects of relational classroom variables across subsequent school years, reveal the same conclusion (see Table 1): Effects of rstgrade relational classroom variables are the largest in rst grade and are absent by the time children are in third grade. As an example, Figure 2 shows how, beyond everything else in the model, the relationship between the average class level of conict and aggressive behavior is strongest in the earlier grades, which are temporally closer to the time at which class-level conict was measured. Considering academic achievement, we detected a small yet signicant effect of the average level of classroom closeness in rst grade on childrens mathematics achievement. The higher the average level of teacher-

child closeness in rst grade, the better the individual children of the classroom achieve mathematically. However, based on the proportion of explained variance (0%), we conclude that classroom relational variables in rst grade do not add substantially to the prediction of childrens academic achievement. Not only in the prediction of academic achievement, but also in the prediction of psychosocial adjustment outcomes, there is still a signicant amount of variance left at the child and class-level to be explained by other predictors, after including several child features and relational climate indicators in Model B (see Appendix A). In a nal step, dyadic teacher-child relationship conict and closeness in rst grade were added to models already containing childrens initial levels of adjustment, demographic features, and rst-grade relational classroom climate variables (Model C). Overall, the deviance of the total model reduced signicantly in the prediction of every outcome, except in the prediction of reading. Therefore, Model C was not reported for this outcome variable. With regard to the other outcomes, rst, we see that more dyadic teacher-child conict in rst grade is associated with lower levels of psychosocial adjustment. Specically, more conict is associated with more teacher-rated aggressive behavior and with less teacher-rated popularity with peers and feelings of well-being at school. Second, more dyadic closeness in rst grade is associated with better psychosocial adjustment in the classroom (i.e., greater popularity with peers and more feelings of well-being at school). First-grade teacher-child relationship quality variables explain 7% to 11% of the variance in childrens psychosocial adjustment in rst grade, 7% to 10% in second grade, and 6% to 8% in third grade, beyond all other predictors already in the model (see Table 1). Simple slope analyses yield the same conclusion: Effects of rst-grade dyadic teacher-child relationship quality are the largest in rst grade and smaller in second and third grade, but still exist by the time children are in third grade. As an example,

132

THE ELEMENTARY SCHOOL JOURNAL

FIG. 3.Differential effect of teacher-child conict on aggressive behavior for the subsequent grades.

Figure 3 shows that beyond everything else in the model, the relationship between teacher-child conict and aggressive behavior is strongest in the earlier grades, which are temporally closer to the time at which teacher-child conict was measured. There is one exception to this trend, which is that more teacher-child conict in rst grade is no longer associated with lower teacher-rated feelings of well-being in third grade (see Table 1). Considering academic achievement, we detected a small yet signicant effect of dyadic teacher-child conict in rst grade on childrens mathematics achievement. The more conict in the dyadic teacher-child relationship, the worse children achieve mathematically. However, based on the proportion of explained variance (0%), we concluded that rst-grade dyadic teacherchild relationship quality variables do not add substantially to the prediction of childrens academic achievement.

Discussion
In this study, we were especially interested in the association between teacher-child relationship quality and childrens school adjustment beyond several child features and

classroom relational climate. We found that more teacher-child closeness, as rated by rst-grade teachers, was associated with better psychosocial adjustment, whereas more teacher-child conict was associated with worse psychosocial adjustment, as rated by teachers. Our ndings are compatible with ndings from other studies in which associations have been detected between teacher-child closeness and conict, on the one hand, and similar aspects of childrens psychosocial functioning on the other, not controlling for any variables or for a selection of the variables we controlled for in our study (e.g., Birch & Ladd, 1997, 1998; Howes, 2000; Hughes et al., 1999; Mantzicopoulos, 2005; Pianta et al., 1995; Pianta & Stuhlman, 2004; Silver et al., 2005). The associations we found between rstgrade teacher-rated relationship quality and childrens psychosocial adjustment were visible until third grade, but were most pronounced for adjustment in rst grade. This is in line with the idea raised by Pianta and Stuhlman (2004) that the teacher-child relationship has primarily concurrent effects, that is, teacher-child relationships may have the greatest effect on childrens competency in that specic classroom or relational context. The stronger within-year compared with across-year associations may, however, also be due to shared informant bias: Teachers who rated childrens psychosocial adjustment in rst grade also rated the quality of the teacher-child relationship. Regarding childrens academic achievement in primary school, a very small additional effect from relationship quality was detected. Specically, more conict in the teacher-child relationship in rst grade was related to worse mathematics achievement over the rst 3 years in primary school. However, the proportion of explained variance by relational conict was negligible for this outcome. This result supports previous studies in which teacher-child relationship quality related more strongly with psychosocial outcomes than with childrens achievement (e.g., Baker, 2006; Pianta &
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

133

Stuhlman, 2004). These differential associations can be explained in several ways. First, according to Hamre and Pianta (2001), psychosocial outcomes may be considered to be more proximal to the predictor, which is itself a measure of psychosocial adjustment. Second, academic achievement may depend more upon the instructional quality of the classroom than upon the quality of interpersonal relationships in the classroom (Hamre & Pianta, 2001). Furthermore, for at least some subgroups of children, academic problems may be rooted in central nervous system dysfunction and may therefore require direct instruction and remediation of learning decits to improve school adjustment (Baker, 2006; Lerner, 2003). Third, compared to psychosocial adjustment, we detected stronger stability for academic achievement across grades, leaving less room for relational variables to explain additional variance. Finally, here too, method bias may play a role (especially in rst grade) because both relationship quality and psychosocial adjustment variables were teacher-rated, while more objective measures were used to evaluate childrens academic adjustment. In contrast to most studies concerning the role of (dyadic) teacher-child relationship quality, our study also evaluated the role of classroom-level closeness and conict. Results showed that classroom relational climate is also associated with childrens school adjustment, beyond several child characteristics. Specically, we found that classroom relational closeness, as rated by rst-grade teachers, was associated with better psychosocial adjustment, whereas classroom relational conict was associated with worse psychosocial adjustment as rated by teachers. Aside from a very small effect, no effects were detected from the classroom relational climate on childrens academic adjustment. Our results were in line with ndings from two other studies that evaluated the effect from classroom relational climate variables on childrens adjustment. First, Birch and Ladd (1997) found that mean conict was (negatively) associated with childrens school

affect and attitudes in kindergarten, but not with childrens academic achievement. Second, Howes (2000) found that a preschool climate of closeness was (positively) associated with childrens social competence with peers in second grade. The effects we found of rst-grade classroom relational climate were visible in rst grade as well as in second grade, but were no longer detectable when children were in third grade. In sum, we can conclude that both dyadic teacher-child relationships and classroom relational climate play signicant roles in the development of child psychosocial competencies that support a broad spectrum of adjustment in the classroom, as has been suggested in previous studies as well (e.g., Howes, 2000). Moreover, ndings from our study not only resonate with these earlier investigations but, in light of the longitudinal method applied, take us beyond a piecemeal understanding of this phenomenon as described by various (important) crosssectional analyses of prior decades. As in previous studies (e.g., Birch & Ladd, 1997, 1998; Howes, 2000; Ladd et al., 1999; Pianta & Stuhlman, 2004), we also found several child features to be associated with childrens adjustment at school. Girls tend to adjust better psychosocially, as do children with a higher SES and higher initial levels of psychosocial adjustment in kindergarten. Boys perform better on mathematics achievement tests, and children with higher kindergarten levels of academic adjustment generally perform better on achievement tests in primary school. When we controlled for other child variables, children whose parents had a foreign nationality performed better academically than Belgian children. This may seem surprising given the myriad of evidence that children from native parents are generally expected to have better academic adjustment (see introduction). However, when ethnic background was the only predictor, native Belgian children did perform better academically. Along with additional analyses, this suggests that the negative effect

134

THE ELEMENTARY SCHOOL JOURNAL

of immigrant status on academic development may mainly be due to its association with other background factors, such as low SES, as well as with initially lower levels of academic achievement. A similar pattern of results has been obtained in other largescale studies in Belgium, suggesting that children with higher SES and native children adjust better to school than children with lower SES and immigrants (e.g., Groenez et al., 2003; Hirtt et al., 2007). Finally, we found childrens academic achievement to increase over the years. This academic progress across the rst years of primary school mirrors a natural process of cognitive maturation. In contrast, childrens aggressive behavior and especially their feelings of well-being at school were found to decrease over time. Several studies with older children have also detected a decline in childrens psychological well-being at school across grades (Harter, 1996; Rhodes, Roffman, Reddy, & Fredriksen, 2004). Although psychosocial adjustment trajectories still need further study, the present ndings suggest that this decline in well-being may already start soon after children enter primary school. Strengths, Limitations, and Future Directions The present study relies on a large sample, from which data were gathered longitudinally in the early years of primary education. Statistically, we took into account the nested structure of the available data, making use of multilevel data-analysis techniques. Furthermore, we controlled for several child and classroom environmental variables in a comprehensive way when evaluating the impact of teacher-child relationship quality on various aspects of childrens school adjustment. The inclusion of a representative Belgian sample in this study has the advantage of extending existing research on teacherchild relationships to countries outside the United States. At the same time, however,

this calls for further research on the generalizability of our specic ndings to U.S. and other samples. Perhaps the most signicant limitation of the present study is its reliance on the teachers perspective in assessing both teacherchild relationships and childrens psychosocial adjustment outcomes. As a result, it can be argued that the associations found stem from teachers response biases rather than reecting genuine relations between relationships and adjustment (Pianta et al., 1995). A halo effect my occur, referring to the possibility that teachers assign the same ratings for different aspects of the childrens performance (Mashburn, Hamre, Downer, & Pianta, 2006). However, it is unlikely that the connections established in this study are solely due to shared response bias, as all of the effects found in rst grade were also found in second grade, using another teacher as the informant of childrens psychosocial adjustment. Furthermore, it is important to note that the quality of the teacher-child relationship and childrens broader psychosocial adjustment, as rated by the teacher, are valuable indicators of school adaptation in their own right (e.g., Birch & Ladd, 1997). Teacher ratings of childrens adjustment are predictive of the future school trajectories and school adjustment of young children. Teachers beliefs and ideas about childrens adjustment in general, and about their relational functioning in particular, affect, for example, whether children are referred for special services or placed in ability groups. They are also used to inform others, such as parents and other teachers (e.g., Birch & Ladd, 1997; Mashburn et al., 2006). It would, however, be worthwhile for future studies to include observations on teacher-child relationship quality and/or psychosocial outcomes because teacher ratings on these variables may also suffer from a social desirability tendency. Yet, although observational information does not suffer from social desirability bias, the fact that it is based on limited episodes of observations, or snapshots, can be considered a
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

135

weakness. Teacher ratings, on the other hand, are based on daily classroom behavior and interactions and hence draw from extensive behavior information, covering an extended period of time. Therefore, teacher ratings also provide an insightful perspective on teacher and child competencies within educational settings (e.g., Kenny & Chekaluk, 1993; Mashburn et al., 2006). Practical Implications The associations we found between classroom and dyadic relationship quality, on the one hand, and childrens adjustment on the other, clearly support the view that teacher-child relationships are an important relational context for the child, and may provide a window for identication of risks and for intervention or prevention (Pianta et al., 1995). However, a focus on teacher-child relationships is currently still missing in several school-based prevention and intervention efforts as these efforts have generally been too centered on the individual child (e.g., Sameroff & Mackenzie, 2003; Zajac & Kobak, 2006). Our study strongly suggests that school practitioners should not only focus on trying to change children, through skill training, for example, to adjust better in class. Instead, they should also pay attention to the inuence of relational aspects of the classroom environment in which children are embedded in order to foster their adjustment (see also DeMulder, Denham, Schmidt, & Mitchell, 2000; Hamre & Pianta, 2005; Sutherland & Oswald, 2005). At this moment, several intervention programs aimed at improving dyadic teacher-child relationship quality and/or more general relational classroom climate (e.g., Driscoll & Pianta, in press; McIntosh, Rizza, & Bliss, 2000; Pianta & Hamre, 2001) are being developed and evaluated. In addition to directly improving the quality of dyadic teacher-child relationships and/or the relational climate, it could also be interesting to investigate the possible determinants of these relational quali-

ties. Studying this could yield more clues for practitioners about how to effectively enhance relationship quality in the classroom and, consequently, about how to foster childrens school adjustment indirectly. Teachers, for example, bring with them a set of beliefs about relationships with children based on their own history of experiences, which inuences the interpretation of the childs behavior and their subsequent responses (Sameroff & MacKenzie, 2003). Furthermore, the quality of the relationships teachers form with children also depends on teachers attitudes, such as authoritarian attitudes (Pianta et al., 2005), and on teachers mental health, such as depressive symptoms (Hamre & Pianta, 2004; Hamre, Pianta, Downer, & Mashburn, 2008). As part of teacher consultation, it may be useful to make teachers more aware of these beliefs, attitudes, and symptoms and their effects on behavior and relationship quality. Concerning the classroom context, it has been shown that some structural features, such as the ratio of children to adults, predict the emotional quality of contacts with students (NICHD ECCRN, 2002; Pianta et al., 2002). Likewise, in more emotionally supportive classrooms, dyadic teacher-child relationships were found to be less conictual (Hamre et al., 2008). Continued work in this direction will provide teachers and other school personnel with a better understanding of how to facilitate positive outcomes for children in school (Pianta & Stuhlman, 2004). As suggested in the current study, fostering childrens academic and psychosocial adjustment in school may each require different measures to be taken. For improving childrens psychosocial adjustment across the early years of formal schooling, this study reveals the impact of strengthening teachers relationships with their pupils as early as rst grade. Early qualitative teacherchild relationships may thus set the stage for further scholastic development (Pianta et al., 1995), at least when it comes to psychosocial aspects of development.

136

THE ELEMENTARY SCHOOL JOURNAL

Appendix A Random Parameters, Deviance, and Signicance of the Reduction in Deviance for Nested Models Fitted in MLwiN

TABLE A1. Prediction of Aggressive Behavior (n Baseline B Random part: Class level:
CONS 2 CONS,TIME 2 TIME

7,107) Model C (SE) B (SE)

Model A (SE) B (SE) B

Model B

.11** .05** .06** .48** .03*

(.02) (.01) (.01) (.02) (.01)

.14** .06** .06** .18** .01

(.02) (.01) (.01) (.02) (.01)

.07** .04** .06** .18** .01

(.01) (.01) (.01) (.02) (.01)

.07** .04** .06** .02 .05**

(.01) (.01) (.01) (.01) (.01)

Child level:
CONS 2 CONS,TIME 2 TIME

Time level Deviance


2 difference deviance

. . .a .45** (.01) 18253.226

. . .a .45** (.01) 16937.046 2(5) 1316**

. . .a .45** (.01) 16875.782 2(2) 61**

. . .a .44** (.01) 16209.537 2(2) 666**

NOTE.The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B included relational climate variables as (class-level) predictors, next to the predictors that were included in Model A. In Model C, dyadic teacher-child relationship variables were added as (child-level) predictors, next to the predictor variables that were included in Model B. aThis parameter is dropped for the model to converge. *p .01. **p .001.

TABLE A2. Prediction of Popularity with Peers (n Baseline B Random part: Class level:
CONS 2 CONS,TIME 2 TIME

7,106) Model C (SE) B (SE)

Model A (SE) B (SE) B

Model B

.07* .04* .05* .26* .02*

(.01) (.01) (.01) (.02) (.01)

.10* .05* .04* .12* .03*

(.01) (.01) (.01) (.01) (.01)

.05* .03* .05* .12* .03*

(.01) (.01) (.01) (.01) (.01)

.05* .03* .05* .03* .06*

(.01) (.01) (.01) (.01) (.01)

Pupil level:
CONS 2 CONS,TIME 2 TIME

Time level Deviance


2 difference deviance

. . .a .39* (.01) 16739.044

. . .a .39* (.01) 15955.449 2(5) 784*

. . .a .39* (.01) 15889.828 2(2) 66*

. . .a .39* (.01) 15379.290 2(2) 511*

NOTE.The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B included relational climate variables as (class-level) predictors, next to the predictors that were included in Model A. In Model C, dyadic teacher-child relationship variables were added as (child-level) predictors, next to the predictor variables that were included in Model B. aThis parameter is dropped for the model to converge. *p .001.
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

137

TABLE A3. Prediction of Feelings of Well-Being at School (n Baseline B Random part: Class level:
CONS 2 CONS,TIME 2 TIME

7,106) Model C (SE) B (SE)

Model A (SE) B (SE) B

Model B

.09* .05* .04* .25* .07*


a

(.01) (.01) (.01) (.01) (.01)

.10* .05* .04* .17* .05*


a

(.01) (.01) (.01) (.01) (.00)

.05* .03* .04* .19* .06*


a

(.01) (.01) (.01) (.01) (.00)

.05* .03* .04* .09* .03*


a

(.01) (.01) (.01) (.01) (.01)

Pupil level:
CONS 2 CONS,TIME 2 TIME

Time level Deviance


2 difference deviance

... .41* (.01) 15558.762

... .40* (.01) 15104.954 2(5) 454*

... .40* (.01) 14917.469 2(2) 187*

... .40* (.01) 14776.263 2(2) 141*

NOTE.The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B included relational climate variables as (class-level) predictors, next to the predictors that were included in Model A. In Model C, dyadic teacher-child relationship variables were added as (child-level) predictors, next to the predictor variables that were included in Model B. aThis parameter is dropped for the model to converge. *p .001.

TABLE A4. Prediction of Reading (n Baseline B Random part: Class level:


CONS 2 CONS,TIME 2 TIME

6,983) Model B B (SE) B Model C (SE)

Model A (SE) B (SE)

24.41* 8.76* 8.75*

(4.32) (1.81) (1.24)

21.14* 7.33* 8.61*

(3.77) (1.65) (1.23)

Pupil level:
CONS 2 CONS,TIME 2 TIME

Time level Deviance


2 difference deviance

176.49* (6.96) 13.85* (2.31) .20 (1.53) 67.50* (2.05) 55238.280

148.23* (6.18) 16.76* (2.19) .25 (1.53) 67.37* (2.05) 54917.170 2(5) 321*

(54993.590)

(54975.880)

NOTE.The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B and Model C were not reported in more detail because the deviance of these models was larger than the deviance of Model A, indicating a worse overall t. *p .001.

138

THE ELEMENTARY SCHOOL JOURNAL

TABLE A5. Prediction of Mathematics Achievement (n Baseline B Random part: Class level:
CONS 2 CONS,TIME 2 TIME

6,994) Model C (SE) B (SE)

Model A (SE) B (SE) B

Model B

11.34** 3.48** 2.79** 44.01** . . .a . . .a

(1.61) (.62) (.37) (1.47)

5.85** 2.32** 2.74** 15.64** . . .a . . .a

(.83) (.44) (.36) (.66)

5.74** 2.32** 2.74** 15.62** . . .a . . .a

(.82) (.44) (.36) (.66)

5.74** 2.32** 2.74** 7.42** . . .a . . .a

(.82) (.44) (.36) (.87)

Pupil level:
CONS 2 CONS,TIME 2 TIME

Time level Deviance


2 difference deviance

16.94** (.37) 45489.06

16.81** (.37) 43332.81 2(5) 2156**

16.81** (.37) 43326.55 2(2) 6.26*

16.81** (.37) 43167.43 2(2) 159**

NOTE.The baseline model only included time as a predictor. Model A included child characteristics as (child-level) predictors (i.e., gender, SES, ethnic background, and initial level of adjustment), next to the time variable. Model B included relational climate variables as (class level) predictors, next to the predictors that were included in Model A. In Model C, dyadic teacher-child relationship variables were added as (child-level) predictors, next to the predictor variables that were included in Model B. aThis parameter is dropped for the model to converge. *p .05. **p .001.

Notes 1. The study referred to here is the SiBO study. SiBO is a Dutch acronym for Schoolloopbanen in het BasisOnderwijs (School Careers in Primary Education). This study was funded by research grant 3H070039 for the Policy Research Centers Study and School Careers, Department of Education, Ministry of the Flemish Community (Belgium). Correspondent: jan .vandamme@ped.kuleuven.be. 2. In Flemish education, three educational networks are usually distinguished. First, community schools are public schools functioning under the authority of the Flemish Community. Second, subsidized public schools provide education organized under the authority of municipalities and provinces. Third, subsidized private (mostly Catholic) schools provide education under the authority of a private person or organization (Verachtert, 2008). 3. A simple Pearson correlation could not be calculated here because average classroom closeness and conict are class-level variables, whereas dyadic closeness and conict are child-level variables. 4. The variance at the school level did not reach signicance for any of the outcomes (equaling 1%2% of the total variance). Therefore, the school level was excluded as a fourth level.

References Aiken, L., & West, S. (1991). Multiple regression: Testing and interpreting interactions. Newbury Park, CA: Sage. Alexander, K. L., & Entwisle, D. R. (1988). Achievement in the rst two years of school: Patterns and processes. Monographs of the Society for Research in Child Development, 53(2, Serial No. 218). Baker, J. A. (2006). Contributions of teacherchild relationships to positive school adjustment during elementary school. Journal of School Psychology, 44, 211229. Berch, D. B. (2005). Making sense of number sense: Implications for children with mathematical disabilities. Journal of Learning Disabilities, 38, 333339. Birch, S. H., & Ladd, G. W. (1996). Interpersonal relationships in the school environment and childrens early school adjustment: The role of teachers and peers. In J. Juvonen & K. R. Wentzel (Eds.), Social motivation: Understanding childrens school adjustment (pp. 199 225). New York: Cambridge University Press. Birch, S. H., & Ladd, G. W. (1997). The teacherchild relationship and childrens early school adjustment. Journal of School Psychology, 35, 6179. Birch, S. H., & Ladd, G. W. (1998). Childrens interpersonal behaviors and the teacherDECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

139

child relationship. Developmental Psychology, 34, 934 946. Burchinal, M. R., Roberts, J. E., Riggins R., Jr., Zeisel, S. A., Neebe, E., & Bryant, D. (2000). Relating quality of center-based child care to early cognitive and language development longitudinally. Child Development, 71, 338 357. Buyse, E., Verschueren, K., Doumen, S., Van Damme, J., & Maes, F. (2008). Classroom problem behavior and teacher-child relationships in kindergarten: The moderating role of classroom climate. Journal of School Psychology, 46, 367391. Citogroep (2003). Taal voor Kleuters voor Vlaanderen [Kindergartners Language Achievement Test for Flanders]. Arnhem, the Netherlands: Citogroep. Cornelissen, G., & Verschueren, K. (2001). Nederlandse vertaling van de Student-Teacher Relationship Scale (STRS) (Pianta, 1996): Toetsing van de factorstructuur en constructie van een verkorte versie [Dutch translation of the Student-Teacher Relationship Scale (Pianta, 1996): Testing of factor structure and construction of a shortened version]. Leuven, Belgium: Center for School Psychology, Katholieke Universiteit Leuven. Retrieved June 16, 2008, from http:// www.steunpuntloopbanen.be/rapporten/ LOA-rapport_03.pdf DeMulder, E. K., Denham, S., Schmidt, M., & Mitchell, J. (2000). Q-sort assessment of attachment security during the preschool years: Links from home to school. Developmental Psychology, 36, 274 282. Doumen, S., Verschueren, K., Buyse, E., Germeijs, V., Luyckx, K., & Soenens, B. (2008). Reciprocal relations between teacher-child conict and aggressive behavior in kindergarten: A three-wave longitudinal study. Journal of Clinical Child and Adolescent Psychology, 37, 588 599. Driessen, G., van Langen, A., & Vierke, H. (2000). Basisonderwijs: Veldwerkverslag, leerlinggegevens en oudervragenlijsten. Basisrapportage PRIMA-cohortonderzoek. Derde meting 1998 1999 [Elementary school: Report on pupil data and parent questionnaires. PRIMA cohort study. Third wave 1998 1999]. Nijmegen, the Netherlands: ITS. Driscoll, K. C., & Pianta, R. C. (in press). Banking Time in Head Start: Early efcacy of an intervention designed to promote supportive teacher-child relationships. Early Education and Development. Evers, A., van Vliet-Mulder, J. C., Resing, W. C. M., Starren, J. C. M. G., van Alphen de Veer, R. J., & van Boxtel, H. (2002). COTAN

Testboek voor het onderwijs [COTAN testbook for education]. Amsterdam: Boom. Goldstein, H. (1995). Multilevel statistical models. London: Arnold. Groenez, S., Van den Brande, I., & Nicaise, I. (2003). Cijferboek sociale ongelijkheid in het Vlaamse onderwijs: Een verkennend onderzoek op de panelstudie van Belgische huishoudens [Report on social inequality in Flemish education: Exploratory research on the panel study on Belgian households]. Leuven, Belgium: HIVA, Steunpunt LOA, Unit Onderwijsloopbanen. Hamre, B. K., & Pianta, R. C. (2001). Early teacher-child relationships and the trajectory of childrens school outcomes through eighth grade. Child Development, 72, 625 638. Hamre, B. K., & Pianta, R. C. (2004). Selfreported depression in nonfamilial caregivers: Prevalence and associations with caregiver behavior in child-care settings. Early Childhood Research Quarterly, 19, 297318. Hamre, K. H., & Pianta, R. C. (2005). Can instructional and emotional support in the rst-grade classroom make a difference for children at risk of school failure? Child Development, 76, 949 967. Hamre, K. H., & Pianta, R. C., Downer, J. T., & Mashburn, A. J. (2008). Teachers perceptions of conict with young students: Looking beyond problem behaviors. Social Development, 17, 115136. Harter, S. (1996). Teacher and classmate inuences on scholastic motivation, self-esteem, and level of voice. In J. Juvonen & K. R. Wentzel (Eds.), Social motivation: Understanding childrens school adjustment (pp. 11 42). New York: Cambridge University Press. Hedeker, D., & Gibbons, R. D. (1997). Application of random-effects pattern-mixture models for missing data in longitudinal studies. Psychological Methods, 2, 64 78. Hirtt, N., Nicaise, I., & De Zutter, D. (2007). De school van de ongelijkheid [The school of inequality]. Berchem, Belgium: EPO. Howes, C. (2000). Social-emotional classroom climate in child care, child-teacher relationships and childrens second grade peer relations. Social Development, 9, 191204. Hox, J. (2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Erlbaum. Hughes, J. N., Cavell, T. A., & Jackson, T. (1999). Inuence of the teacher-child relationship on childhood conduct problems: A prospective study. Journal of Clinical Child Psychology, 28, 173184. Jungbluth, P., Roede, E., & Roeleveld, J. (2001). Validering van het PRIMA-leerlingproel: Se-

140

THE ELEMENTARY SCHOOL JOURNAL

cundaire analyse op het PRIMA-cohort bestand (SCO-rapport nr. 608) [Validation of the PRIMA student prole: Secondary analysis on the PRIMA cohort data (SCO report no. 608)]. Amsterdam: SCO-Kohnstamn lnstituut van de Faculteit der Maatschappij- en Gedragswetenschappen, Universiteit van Amsterdam. Kenny, D. T., & Chekaluk, E. (1993). Early reading performance: A comparison of teacherbased and test-based assessments. Journal of Learning Disabilities, 26, 227236. Koomen, H., Verschueren, K., & Pianta, R. (2007). Leerling Leerkracht Relatie Vragenlijst [Student Teacher Relationship Scale]. Houten, the Netherlands: Bohn Staeu van Loghum. Ladd, G. W. (1989). Childrens social competence and social supports: Precursors of early school adjustment? In B. H. Schneider, G. Attili, J. Naldel, & R. P. Weisberg (Eds.), Social competence in developmental perspective (pp. 277291). Amsterdam: Kluwer. Ladd, G. W. (1992). The Teacher Rating Scale of School Adjustment. Urbana-Champaign: University of Illinois. Ladd, G. W. (2003). Probing the adaptive significance of childrens behavior and relationships in the school context: A child by environment perspective. In R. V. Kail (Ed.), Advances in child development and behavior (Vol. 31, pp. 43104). San Diego: Academic Press. Ladd, G. W., Birch, S. H., & Buhs, E. S. (1999). Childrens social and scholastic lives in kindergarten: Related spheres of inuence? Child Development, 70, 13731400. Ladd, G. W., & Prolet, S. M. (1996). The Child Behavior Scale: A teacher-report measure of young childrens aggressive, withdrawn, and prosocial behaviors. Developmental Psychology, 32, 1008 1024. Lerner, J. W. (2003). Learning disabilities: Theories, diagnosis, and teaching strategies (9th ed.). Boston: Houghton Mifin. Lord, F. M. (1980). Applications of item response theory to practical testing problems. Hillsdale, NJ: Erlbaum. Malofeeva, E., Day, J., Saco, X., Young, L., & Ciancio, D. (2004). Construction and evaluation of a number sense test with Head Start children. Journal of Educational Psychology, 96, 648 659. Mantzicopoulos, P. (2005). Conictual relationships between kindergarten children and their teachers: Associations with child and classroom context variables. Journal of School Psychology, 43, 425 442. Mashburn, A. J., Hamre, B. K., Downer, J. T., & Pianta, R. C. (2006). Teacher and classroom characteristics associated with teachers rat-

ings of prekindergartners relationships and behaviors. Journal of Psychoeducational Assessment, 24, 367380. McIntosh, D. E., Rizza, M. G., & Bliss, L. (2000). Implementing empirically supported interventions: Teacher-child interaction therapy. Psychology in the Schools, 37, 453 462. Moelands, F., Kamphuis, F., & Rymenans, R. (2003). Drie-Minuten-Toets voor Vlaanderen (DMT-V): Wetenschappelijke verantwoording [Three-Minute-Test for Flanders (DMT-V): Scientic justication]. Arnhem, the Netherlands: Citogroep. Moelands, F., & Rymenans, R. (2003). DrieMinuten-Toets voor Vlaanderen: Handleiding [Three-Minute-Test for Flanders: Manual]. Arnhem, the Netherlands: Citogroep. NICHD Early Child Care Research Network. (2002). The relation of global rst-grade classroom environment to structural classroom features, teacher, and student behaviors. Elementary School Journal, 102, 367387. Peisner-Feinberg, E. S., Burchinal, M. R., Clifford, R. M., Culkin, M. L., Howes, C., Kagan, S. L., et al. (2001). The relation of preschool child-care quality to childrens cognitive and social developmental trajectories through second grade. Child Development, 72, 1534 1553. Pianta, R. C. (1999). Enhancing relationships between children and teachers. Washington, DC: American Psychological Association. Pianta, R. C. (2001). Student-Teacher Relationship Scale: Professional manual. Lutz, FL: Psychological Assessment Resources. Pianta, R. C. (2006). Schools, schooling and developmental psychopathology. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopathology (2nd ed., pp. 494 529). Hoboken, NJ: Wiley. Pianta, R. C., & Hamre, B. K. (2001). Student, teachers and relationship support (STARS). Odessa, FL: Psychological Assessment Resources. Pianta, R. C., Hamre, B. K., & Stuhlman, M. (2003). Relationships between teachers and children. In W. M. Reynolds, G. E. Miller, & I. B. Weiner (Eds.), Handbook of psychology: Vol. 7. Educational psychology (pp. 199 234). Hoboken, NJ: Wiley. Pianta, R. C., Howes, C., Burchinal, M., Bryant, D., Clifford, R., Early, C., et al. (2005). Features of pre-kindergarten programs, classrooms, and teachers: Do they predict observed classroom quality and child-teacher interactions? Applied Developmental Science, 9, 144 159. Pianta, R. C., La Paro, K. M., Payne, C., Cox, M. J., & Bradley, R. (2002). The relation of
DECEMBER 2009

PREDICTING SCHOOL ADJUSTMENT

141

kindergarten classroom environment to teacher, family and school characteristics and child outcomes. Elementary School Journal, 102, 225238. Pianta, R. C., Steinberg, M., & Rollins, K. B. (1995). The rst two years of school: Teacher-child relationships and deections in childrens classroom adjustment. Developmental Psychopathology, 7, 295312. Pianta, R. C., & Stuhlman, M. W. (2004). Teacher-child relationship and childrens success in the rst years of school. School Psychology Review, 33, 444 458. Ponjaert-Kristoffersen, I., Andries, C., Celestin Westreich, S., & Samaey, C. (2000). Onderzoek naar het gebruik van het Leerlingvolgsysteem voor jonge kinderen in Vlaanderen (OBPWO-project 96.08): Eindrapport [Investigation of the use of the achievement monitoring system for young children in Flanders (OBPWO project 96.08): Final report]. Brussels: Vrije Universiteit Brussel, Vakgroep Ontwikkelings- en Levenslooppsychologie. Rasbash, J., Charlton, C., Browne, W. J., Healy, M., & Cameron, B. (2005). MLwiNVersion 2.02. Bristol: Centre for Multilevel Modelling, University of Bristol. Reynders, T., Nicaise, I., & Van Damme, J. (2005). Longitudinal onderzoek in het basisonderwijs: De constructie van een SESvariabele voor het SiBO-onderzoek (LOArapport nr. 31). [A longitudinal study in primary education: Construction of an SES variable for use in the SiBO study (LOAreport no. 31)]. Leuven, Belgium: Steunpunt Loopbanen doorheen Onderwijs naar Arbeidsmarkt, Cel Schoolloopbanen in het basisonderwijs (SiBO). Retrieved June 16, 2008, from http://www.steunpuntloopbanen.be/ rapporten/LOA-rapport_31.pdf Rhodes, J., Roffman, J., Reddy, R., & Fredriksen, K. (2004). Changes in self-esteem during the middle school years: A latent growth curve study of individual and contextual inuences. Journal of School Psychology, 42, 243261. Rimm-Kaufman, S. E., Pianta, R. C., & Cox, M. J. (2000). Teachers judgments of problems in the transition to kindergarten. Early Childhood Research Quarterly, 15, 147166. Sameroff, A. J., & Mackenzie, M. J. (2003). Research strategies for capturing transactional models of development: The limits of the possible. Development and Psychopathology, 15, 613 640. Silver, R. B., Measelle, J. R., Armstrong, J. M., & Essex, M. J. (2005). Trajectories of classroom externalizing behavior: Contributions of child characteristics, family characteristics, and the

teacher-child relationship during the school transition. Journal of School Psychology, 43, 3960. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis. New York: Oxford University Press. Snijders, T. A. B., & Bosker, R. J. (1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling. London: Sage. Sutherland, K. S., & Oswald, D. P. (2005). The relationship between teacher and student behavior in classrooms for students with emotional and behavioral disorders: Transactional processes. Journal of Child and Family Studies, 14, 114. van Kuijk, J. J. (1996). Taal voor Kleuters [Kindergartners Language Achievement Test]. Arnhem, the Netherlands: Centraal lnstituut voor Toetsontwikkeling. Verachtert, P. (2003). Longitudinaal onderzoek in het basisonderwijs: Toetsen schooljaar 20022003 (LOA-rapport nr. 14) [A longitudinal study in primary education: Achievement tests administered in school year 20022003 (LOA-report no. 14)]. Leuven, Belgium: Steunpunt Loopbanen doorheen Onderwijs naar Arbeidsmarkt, Cel Schoolloopbanen in het basisonderwijs (SiBO). Retrieved June 16, 2008, from http://www.steunpuntloopbanen.be/ rapporten/LOA-rapport_14.pdf Verachtert, P. (2008). Kindergarten risk factors for early mathematics difculties: Evidence from the longitudinal SiBO study. Unpublished doctoral dissertation, Centrum voor Gezins-en Orthopedagogiek, Centrum voor Methodologie van het Pedagogisch Onderzoek, Katholieke Universiteit Leuven, Belgium. Verhaeghe, J. P., Maes, F., Gombeir, D., & Peeters, E. (2002). Longitudinaal onderzoek in het basisonderwijs: Steekproeftrekking (LOArapport nr. 5) [A longitudinal study in primary education: Sample procedure (LOAreport no. 5)]. Leuven, Belgium: Steunpunt Loopbanen doorheen Onderwijs naar Arbeidsmarkt, Cel Schoolloopbanen in het basisonderwijs (SiBO). Retrieved June 16, 2008, from http://www.steunpuntloopbanen.be/ rapporten/LOA-rapport_05.pdf Wilson, H. K., Pianta, R. C., & Stuhlman, M. (2007). Typical classroom experiences in rst grade: The role of classroom climate and functional risk in the development of social competencies. Elementary School Journal, 108, 8196. Zajac, K., & Kobak, R. (2006). Attachment. In G. G. Bear & K. M. Minke (Eds.), Childrens needs III: Development, prevention and intervention (pp. 379 389). Washington, DC: National Association of School Psychologists.

Copyright of Elementary School Journal is the property of University of Chicago Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

You might also like