Вы находитесь на странице: 1из 15

Journal of Educational Psychology © 2014 American Psychological Association

2015, Vol. 107, No. 2, 502–516 0022-0663/15/$12.00 http://dx.doi.org/10.1037/a0037457

Teacher–Child Relationship Quality and Academic Achievement in


Elementary School: Does Gender Matter?
Meghan P. McCormick and Erin E. O’Connor
New York University

Using data from the National Institute of Child Health and Human Development Study of Early Child Care
and Youth Development (N ⫽ 1,364) and 2-level hierarchical linear models with site fixed effects, we
examined between- and within-child associations between teacher– child relationship closeness and conflict
and standardized measures of children’s math and reading achievement from 1st through 5th grades. In
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

addition, we tested whether longitudinal effects varied by gender. Results revealed a between-child effect of
This document is copyrighted by the American Psychological Association or one of its allied publishers.

conflict and a within-child effect of closeness on reading achievement for the full sample. In addition, there
were moderated between- and within-child effects of conflict on math achievement: Girls with more
conflictual relationships showed lower overall levels of math achievement and less growth in math achieve-
ment than did boys with similar levels of conflict. Implications are discussed.

Keywords: teacher– child relationships, gender, achievement, middle childhood

A robust body of research has identified associations between girls tend to have higher quality relationships with teachers than boys
high-quality teacher– child relationships in prekindergarten and ele- do (Ewing & Taylor, 2009; Hamre & Pianta, 2001). This trend is
mentary school and children’s academic achievement in middle child- somewhat expected given that elementary-aged boys are more likely
hood (Hamre & Pianta, 2001; O’Connor, 2010; O’Connor & McCart- to engage in disruptive behaviors (Hibel, Farkas, & Morgan, 2010).
ney, 2007; Pianta & Stuhlman, 2004; Roorda, Koomen, Spilt, & Oort, Such findings have led some researchers to theorize that there is a
2011; Rudasill, 2011). This work has found that children who have female advantage in classroom-based education, perhaps related to
early relationships with teachers characterized by high levels of close- girls’ ability to form high-quality relationships with teachers (Legewie
ness and low levels of conflict are more likely to exhibit concurrent & DiPrete, 2012). Indeed, girls tend to score higher than boys on
and prospective academic achievement (e.g., Baker, 2006; Hamre & teacher assessments of academic competence and exhibit higher
Pianta, 2001; Pianta & Nimetz, 1991). Such literature suggests that scores on standardized measures of reading achievement (Robinson &
interventions designed to boost academic achievement during middle Lubienski, 2011). Yet, boys continue to outperform girls on standard-
childhood (i.e., ages 5–10 years) should consider targeting teacher– ized assessments of math across elementary school (Robinson &
child relationship quality. Recent studies have used rigorous longitu- Lubienski, 2011). Thus, an academic risk perspective (e.g., Hamre &
dinal designs to identify positive effects of teacher– child relationship Pianta, 2001; Silver, Measelle, Armstrong, & Essex, 2005) argues that
quality on achievement (Maldonado-Carreño & Votruba-Drzal, 2011; it is important to examine the benefits of high quality teacher– child
Pianta et al., 2008; Spilt, Hughes, Wu, & Kwok, 2012). Yet, empirical relationships differently for boys and girls when considering math and
research has yet to distinguish how effects due to changes in chil- reading achievement separately. Because boys are at higher risk for
dren’s relationships with different teachers over time (i.e., within lower reading achievement, they may have the most to gain from a
child) differ from variations in relationships different children have high-quality teacher– child relationship in terms of reading outcomes
with teachers (i.e., between child). and the most to lose from a low-quality relationship. In contrast, girls’
Furthermore, the extent to which the associations between within- math achievement might have more to gain from a high-quality
and between-child relationships and achievement differ by gender relationship and more to lose from a low-quality relationship (Hamre
have yet to be examined. Multiple studies have noted that, on average, & Pianta, 2001; Silver et al., 2005).
Using a longitudinal framework, this study builds on previous
research and examines between- and within-child effects of
teacher– child closeness and conflict on reading and math achieve-
This article was published Online First August 11, 2014. ment across elementary school. In addition, this study will help
Meghan P. McCormick, Department of Applied Psychology, New York elucidate whether there are gender differences in associations
University; Erin E. O’Connor, Department of Teaching and Learning, New
between dimensions of teacher– child relationship quality and stan-
York University.
The research reported here was supported by the Institute of Education
dardized assessments of math and reading achievement.
Sciences, U.S. Department of Education, through Grant R305A080512 to
New York University. The opinions expressed are those of the authors and Teacher–Child Relationships and Academic
do not represent views of the Institute or the U.S. Department of Education. Achievement in a Longitudinal Framework
Correspondence concerning this article should be addressed to Meghan
P. McCormick, New York University, Department of Applied Psychology, Teacher– child relationships are bidirectional, interpersonal ex-
246 Greene Street, New York, NY 10003. E-mail: meghan.mccormick@ changes that take place in proximal microsystems (Bronfenbrenner
nyu.edu & Morris, 1998; Pianta, 1999). Conceptual studies, based in at-

502
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 503

tachment theory, propose that children who experience positive questions remain about the continuing power of the teacher– child
teacher– child relationships, characterized by low levels of conflict relationship across the course of middle childhood.
and high levels of closeness, are able to rely on teachers as a secure One challenge in using longitudinal study designs, however, is
base and a resource for actively exploring the school environment the importance of decomposing between-child effects from within-
(Howes, Phillipsen, & Peisner-Feinberg, 2000; J. N. Hughes, child effects. Between-child effects describe mean differences
Cavell, & Willson, 2001). Thus, high-quality teacher– child rela- between individuals. For example, an examination of between-
tionships may boost students’ learning by creating a supportive child effects provides information about whether the overall level
environment in which children are motivated to actively and of teacher– child closeness and conflict predicts average levels of
appropriately engage in the classroom (Ladd & Burgess, 1999). academic achievement in middle childhood (Fitzmaurice, Laird, &
Although the correlational research base linking teacher– child Ware, 2004). Within-child effects, in contrast, describe intraindi-
relationships and academic achievement initially seems quite ro- vidual changes and can better test whether changes in one’s own
bust (e.g., Roorda et al., 2011), many of these studies have been teacher– child conflict and closeness predict future changes in
done on samples of children in preschool and early elementary achievement (Singer & Willett, 2003). Such an approach has
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

school (e.g., kindergarten through second grade). Using longitu- methodological benefits as it can help address threats to the inter-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

dinal designs, research conducted in preschool and the first years nal validity of longitudinal study designs posed by omitted vari-
of elementary school has found links between teacher– child rela- able bias (Gelman & Hill, 2007; Shadish, Cook, & Campbell,
tionship quality and contemporaneous development of academic 2002). That is, for between-child effects, the association between
skills (Pianta, Nimetz, & Bennett, 1997; Pianta & Stuhlman, teacher– child conflict and closeness and academic outcomes may
2004), as well as subsequent development on occasion (Burchinal be explained by omitted variables. Examining within-child effects
et al., 2002). Many studies that do focus on children in later middle can reduce the threat posed by omitted variables by eliminating
childhood (e.g., ages 8 –10 years) typically use measures of bias associated with unobserved characteristics that are constant
teacher– child relationship quality from early childhood (e.g., over time (Dearing, McCartney, & Taylor, 2006; Maldonado-
Hamre & Pianta, 2001) or the previous academic year (e.g., J. N. Carreño & Votruba-Drzal, 2011). Although there remains a threat
Hughes, Luo, Kwok, & Loyd, 2008) to estimate prospective of omitted variables bias, testing within-child associations between
teacher– child relationships and achievement can strengthen causal
achievement, which leaves questions regarding concurrent effects
arguments about the importance of teacher– child relationship
of teacher– child closeness and conflict on achievement.
quality for academic achievement.
Although limited, there is some preliminary evidence for the
Previous studies have typically addressed the threat of omitted
causal effect of teacher– child relationship quality on achievement.
variables by adjusting for a host of demographic variables at the
For example, a number of existing studies attempt to increase
child, teacher, and school levels instead of examining within-child
causal inference by using a range of demographic variables to
associations (Ly, Zhou, Chu, & Chen, 2012; Spilt, Hughes, Wu, &
control for between-child differences that may influence selection
Kwok, 2012). A recent study by Maldonado-Carreño and Votruba-
into high-quality teacher– child relationships and subsequent aca-
Drzal (2011), however, did examine within-child associations in
demic achievement (e.g., Hughes, 2011; J. N. Hughes et al., 2008;
teacher– child relationships and achievement, aiming to limit the
Ladd et al., 1999). In addition, a recent study by McCormick,
threat of omitted variable bias. Using data from the National
O’Connor, Cappella, and McClowry (2013) used a propensity Institute of Child Health and Human Development Study of Early
score matching approach to identify a positive effect of teacher– Child Care and Youth Development (NICHD SECCYD) from
child relationship quality on math achievement, in a sample of kindergarten to fifth grade, which is the same data set we used in
low-income racial and ethnic minority kindergarten and first grad- the current study, the authors found positive within-child associ-
ers. Still, when studying teacher– child relationships and achieve- ations between overall teacher– child relationship quality and
ment, it may be important to use longitudinal models that both use teacher reports of math and reading achievement. However, they
covariates to address selection bias and allow one to examine detected no significant within-child effects of teacher– child rela-
change across critical periods of development. tionship quality on standardized achievement scores in elementary
Specifically, it is important to use longitudinal analyses to school (e.g., first, third, and fifth grades).
consider concurrent effects of teacher– child closeness and conflict The authors theorized that observed associations between
because children will likely experience varying levels of relation- teacher– child relationship quality and achievement may be more
ship quality with teachers as they move through elementary school of an artifact of a tendency for teachers to develop higher quality
(Maldonado-Carreño & Votruba-Drzal, 2011). For example, as relationships with students who are more engaged in the classroom
children age and transition to the later elementary school grades or who have better academic skills and less of an effect of teacher–
(e.g., third through fifth grades), they may change teachers for child relationship quality on student achievement. Although this
different subjects, experience larger class sizes, and have fewer conclusion requires further examination by future research, the
opportunities for one-to-one interactions with teachers (Rimm- Maldonado-Carreño and Votruba-Drzal (2011) study is notable
Kaufman & Pianta, 2000). Indeed, empirical evidence suggests because it built on previous work by considering within-child
that teacher– child relationship quality declines over time in ele- effects and controlling for initial levels of achievement when
mentary school (Jerome, Hamre, & Pianta, 2009; O’Connor & predicting later outcomes (Burchinal, Peisner-Feinberg, Pianta, &
McCartney, 2007; Pianta & Stuhlman, 2004). In addition, as Howes, 2002; Pianta & Stuhlman, 2004). It is important to extend
children move through middle childhood, additional factors, such this research using methods that address the problem posed by
as relationships with peers, become increasingly important for omitted variable bias when estimating effects of teacher– child
children’s development (Brechwald & Prinstein, 2011). Thus, relationships on academic achievement (Shadish et al., 2002).
504 MCCORMICK AND O’CONNOR

Measurement of Teacher–Child Relationships and noted that previous studies have shown small to negligible effects
Achievement Outcomes of teacher reports of student–teacher relationship quality on stan-
dardized achievement test results in nonrisk samples (e.g., Baker,
In addition to needing to address concerns related to lack of 2006; Buyse, Verschueren, Verachtert, & Van Damme, 2009).
longitudinal models, additional research should unpack the effects Further studies using standardized or observed achievement mea-
of teacher– child relationships on separate dimensions of teacher– sures are thus needed.
child relationships and academic achievement. First, it is critical to
examine how the closeness and conflict dimensions of teacher–
child relationships uniquely contribute to math and reading Teacher–Child Relationships, Gender,
achievement. In their study discussed earlier, Maldonado-Carreño and Academic Achievement
and Votruba-Drzal (2011) combined closeness and conflict into a In addition, little is known about whether there are gender
single measure of relationship quality. Closeness and conflict, differences in associations between teacher– child relationships
however, are two distinct dimensions of teacher– child relationship and academic achievement. Boys, relative to girls, tend to have
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

quality (e.g., K. Hughes, Bullock, & Coplan, 2014; Rudasill, lower quality relationships with teachers (Ewing & Taylor, 2009;
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Niehaus, Buhs, & White, 2013), and it may be important to Hamre & Pianta, 2001). However, differences in overall levels of
examine them separately to identify nuanced effects of teacher– relationship quality between boys and girls may moderate effects
child relationships on achievement (Birch & Ladd, 1997). Re- on academic achievement over time. Indeed, gender socialization
search examining associations between teacher– child conflict and theory hypothesizes that teachers’ differential treatment of boys
closeness and behaviors has shown that these dimensions do have and girls may reinforce behaviors in children that reflect traditional
differential effects (Hamre & Pianta, 2001; Howes & Hamilton, relational styles (Koch, 2003). More specifically, girls are social-
1993). For instance, Baker et al. (2008) found that although close- ized to be compliant and behave responsibly, whereas compliant
ness in the teacher– child relationship during middle childhood was and responsible boys can be regarded as nonmasculine (Morris,
only associated with concurrent adjustment, conflict in the 2011). Girls, relative to boys, are expected to assume a more
teacher– child relationship during middle childhood was associated traditional role as student, following teachers’ directions, paying
with concurrent and prospective adjustment. attention, and completing schoolwork (Mickelson, 2003).
Moreover, in addressing the dimensionality of constructs, there On the basis of this theory, a teacher– child relationship that is
is growing evidence for the importance of distinguishing between consistent with gender-based relational styles and traditional gen-
math and reading achievement when predicting academic achieve- der role expectations (i.e., one high in closeness and low in
ment in elementary school (e.g., Spilt et al., 2012). A recent study conflict) should be a positive support for girls. However, in cases
by McCormick et al. (2013) using a multilevel propensity score where girls break with traditional gender norms and develop less
matching procedure found that teacher– child relationships had an close and more conflictual relationships with teachers, the conse-
impact on math but not reading achievement for a high-risk sample quences for girls’ achievement may be more negative than they
of kindergarten and first grade children. Using three waves of data would be for boys (Ewing & Taylor, 2009; Roorda et al., 2011).
collected from academically at-risk students, Hughes (2011) found Indeed, even if boys do have low quality teacher– child relation-
that student reports of teacher– child conflict positively predicted ships, it may be less consequential for them given that the tradi-
math achievement. However, there were no unique associations tional school setting does not necessarily expect them to form
between teacher– child closeness or conflict and reading achieve- strong personal bonds with teachers (Ewing & Taylor, 2009).
ment. Additionally, recent research suggests that math achieve- There is some preliminary evidence for the differential effect of
ment in early elementary school is the stronger predictor of sub- teacher– child relationship quality on achievement and related out-
sequent achievement, school completion, and college enrollment comes for boys and girls. Indeed, Baker (2006) found that a
(Duncan, 2011). Such results suggest that it may be particularly positive teacher– child relationship made a stronger contribution to
critical to examine math separately from reading in studies of the academic and behavioral outcomes of girls than of boys. In
effects on academic achievement. another study of Head Start children, Ewing and Taylor (2009)
In addition to considering the multidimensionality of both found stronger effects of teacher– child closeness on school com-
teacher– child relationships and academic achievement, it is im- petence for girls. And in a cross-sectional study of Chinese Amer-
portant to address biases posed by using a teacher report of ican students in first and second grades, Ly et al. (2012) found that
relationship quality to predict a teacher report of a child achieve- teacher ratings of teacher– child conflict were negatively associ-
ment outcome (Shadish et al., 2002). In such an analytic frame- ated with girls’ math achievement. Thus, there is some empirical
work, a significant finding may actually reflect rater effects or research to suggest that academically, girls benefit most from close
monomethod bias (Shadish et al., 2002) For example, because teacher– child relationships and are harmed more by conflictual
teachers are more likely to have high-quality relationships with relationships.
children who are behaviorally regulated, they may perceive those A separate theory, however, labeled the academic risk perspec-
children to have higher levels of academic skills than less behav- tive, argues that boys are most likely to benefit academically from
iorally regulated children do (O’Connor & McCartney, 2007; high-quality teacher– child relationships (e.g., Hamre & Pianta,
Rudasill, Reio, Stipanovic, & Taylor, 2010). Indeed, in their study, 2001; Silver et al., 2005). Boys tend to have higher levels of
Maldonado-Carreño and Votruba-Drzal (2011) found effects of academic and behavior problems at school entry and are at higher
teacher– child relationship quality on teacher reports on academic risk for referrals and placements for special education services
achievement but null effects of teacher– child relationships on compared with girls (Cooper & Farran, 1988; Silver et al., 2005).
standardized measures of achievement. Moreover, it should be As argued by Hamre and Pianta (2001), because boys are at higher
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 505

risk for school failure, they will have more “to gain, or lose, Method
through their ability to adapt to the social environment of the
classroom” (Hamre & Pianta, 2001, p. 627). In their study, Hamre
and Pianta (2001) established some support for this perspective, Participants
finding that boys rated high in conflict in kindergarten had more We conducted this study with data from the first two phases of
problem behaviors later in school, whereas girls’ early teacher the NICHD SECCYD, a prospective study of children from birth
conflict was not related to behavioral outcomes. In addition, through adolescence. Study participants were selected using a
teacher– child conflict was more predictive of disciplinary prob- conditional sampling plan from 10 American cities, referred to as
lems in the upper elementary grades for boys than girls. However, sites throughout this study: Little Rock, Arkansas; Irvine, Califor-
effects of teacher– child conflict translated only to behavioral nia; Lawrence, Kansas; Boston, Massachusetts; Philadelphia,
outcomes in this study and did not extend to academic outcomes. Pennsylvania; Pittsburgh, Pennsylvania; Charlottesville, Virginia;
In addition, differences by type of academic outcome were not Morganton, North Carolina; Seattle, Washington; and Madison,
examined differently by gender. Although boys are thought to be Wisconsin. The conditional random sampling plan was designed to
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

at risk for poorer achievement generally, they actually continue to include families from diverse ethnic groups, economic back-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

outperform girls on standardized assessments of math achieve- grounds, and geographic regions, with varying plans for maternal
ment. Thus, girls are actually at somewhat higher risk for poor employment during the child’s first year of life. Please refer to
math achievement, whereas boys are at higher risk for poor reading NICHD Early Child Care Research Network (ECCRN; NICHD
achievement (Robinson & Lubienski, 2011). ECCRN, 1997) for additional information related to the sampling
Moreover, to date, there are no known studies examining procedure used to recruit study participants. A total of 1,364
between- and within-child effects of teacher– child closeness and mothers and children at ten separate sites enrolled in the study
conflict on academic achievement, differentiated by gender, in a (NICHD ECCRN, 1997).
national sample of elementary school students. However, in a Because we were primarily interested in middle childhood in the
longitudinal study of recently immigrated Latin American youth current study, our analyses relied on outcomes collected in the
(ages 9 –14 years at enrollment), gender was found to moderate first, third, and fifth grades. Data on school type (e.g., elementary,
associations between perceived school support and youth aca- middle) were not available in the current study. However, given
demic engagement. Specifically, youth-perceived staff support research showing that over 95% of elementary schools include
was positively related to girls’ initial engagement levels and pos- fifth grade, we assume that the large majority of students in fifth
itively related to changes in boys’ engagement levels over time grade continued to attend an elementary school (Snyder, Dillow, &
(Green, Rhodes, Hirsch, Suárez-Orozco, & Camic, 2008). Because Hoffman, 2001). We restricted the sample to the 1,118 children
this study was conducted with one ethnic group, generalizability of and families who had data for the outcomes for at least one time
findings is limited. Work using rigorous longitudinal models to point.1 About half of the sample (⬃48%) was female. Most of the
examine heterogeneity of effects of teacher– child closeness and children enrolled in the study were White (⬃80%); 13% of the
conflict on academic achievement in broader samples is needed. study sample was Black, 6% was Hispanic, and 5% was Asian or
Pacific Islander. Levels of maternal education among children
Current Study enrolled in the study were fairly diverse: 10% of mothers had less
In the current study, we use data from the NICHD SECCYD to than a high school diploma, 21% graduated from high school, 33%
examine developmental trajectories of closeness and conflict in attended some college, and 36% graduated from a 4-year college.
teacher– child relationships across the elementary school years and About three quarters of the sample (77%) had mothers who were
within-child and between-children associations between closeness married, and 12% of children had mothers who reported receiving
and conflict and achievement. We also investigate the moderating public assistance. Ninety-six percent of first grade teachers were
role of child gender in predicting achievement trajectories in women; information on teacher gender was not available in third
middle childhood. Specifically, we ask the following research and fifth grades. Across grades, the large majority of teachers were
questions: White (92%).

1. Do higher levels of teacher– child closeness and conflict


relate to higher levels of math and reading achievement Measures
in elementary school? Similarly, are changes in teacher– Teacher– child relationships. We used the 15-item Student
child closeness and conflict associated with changes in Teacher Relationship Scale (STRS; Pianta, 1992) to assess teacher
math and reading achievement in elementary school? perceptions of the quality of the teacher– child relationship in first,
third, and fifth grades. Using a 5-point Likert scale that ranged
2. Do associations between teacher– child closeness and from 1 ⫽ definitely does not apply to 5 ⫽ definitely applies,
conflict and math and reading achievement in elementary
school vary by gender?
1
Data on teacher– child relationship quality were also available in kin-
By engendering understanding of how teacher– child relation- dergarten and fourth grade but not in second grade. Because the achieve-
ships directly relate to achievement across time and gender, this ment outcomes were only available in first, third, and fifth grades, we used
teacher– child relationship quality data from first, third, and fifth grades as
empirical knowledge can inform training for teachers and may the main predictors in the study. However, we do incorporate the kinder-
help develop programming focused on providing emotional sup- garten and fourth grade information when creating lags for a future sensi-
port to students. tivity check.
506 MCCORMICK AND O’CONNOR

teachers rated how applicable statements were to their current Covariates. Given that a number of child, family, and teacher
relationship with a particular child. The items are based on attach- factors are theoretically related to both the quality of the teacher–
ment theory and the Attachment Q-Set (Waters & Deane, 1985). child relationship and children’s achievement, we controlled for a
The STRS has been widely used in studies with preschool and large number of covariates in this study. Each of the covariates we
elementary school children. It is associated with children’s and chose has demonstrated associations with either of the main pre-
teachers’ classroom behaviors and correlates with observational dictors and/or math and reading achievement during middle child-
measures of teacher– child relationship quality (e.g., Birch & Ladd, hood. For example, we controlled for the quality of the home
1997; Howes & Hamilton, 1992; Howes & Ritchie, 1999). Addi- environment when children were 54 months old given research
tionally, STRS scores correlate with Attachment Q-Set ratings of showing that children raised in homes with lower Home Observa-
teachers and students such that higher STRS scores are associated tion for Measurement of the Environment (HOME) scores are
with more secure relationships (Howes & Ritchie, 1999). The more likely to exhibit low academic achievement, relative to
test–retest reliabilities reported by Pianta (2001) were .88 for children raised in contexts with higher HOME scores (McCartney,
closeness and .92 for conflict across a 4-week period. Dearing, Taylor, & Bub, 2007). Including these covariates in
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

This scale contains two subdimensions: closeness and conflict. models allows us to more precisely estimate the effects of teacher-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

The Closeness subscale consists of eight items and is an index of child conflict and closeness on academic achievement.
the amount of warmth and open communication present in the Child characteristics. We included several child-level demo-
relationship (e.g., “I share an affectionate, warm relationship with graphic characteristics as covariates in models. We used child
this child”). The Conflict subscale consists of seven items and grade at the time of assessment as a proxy for age given that age
measures the extent to which the relationship is marked by antag- and grade were highly correlated (r ⫽ .97). We included gender as
onistic, disharmonious interactions (e.g., “This child and I always a dummy variable such that female ⫽ 1, male ⫽ 0. Child race was
seem to be struggling with each other”). Cronbach’s alphas were operationalized with dummy variables indicating whether the child
.94 for conflict at all three time points and ranged from .88 to .91 was White, Black, or other race (referent group). We chose these
for closeness. To operationalize the variables for closeness and racial categories given that there was quite a large proportion of
conflict, we calculated a sum of the individual items for each White students (⬃80%) and Black students made up the second
largest contingent of students (13%). The other category, made up
subscale (for closeness scores, 8 – 40; for conflict scores, 7 – 35).
of Asians, Pacific Islanders, and others was much smaller (⬃6%).
The correlations observed between closeness and conflict in the
In addition, because Hispanic designation is an ethnicity and is not
current study were low to moderate (r ⫽ ⫺.28 for first grade;
mutually exclusive from race, a separate dummy variable was
r ⫽ ⫺.38 for third grade, and r ⫽ ⫺.34 for fifth grade). Notably,
included to indicate whether the child was Hispanic.
teachers completed the teacher– child relationship scale 1–1.5
Maternal characteristics. We also considered a number of
months, on average, prior to children being assessed for achieve-
maternal characteristics in models as covariates. Mother’s educa-
ment.
tion when the child was 1 month old was represented by dummy
Math and reading achievement. Reading and math achieve-
variables indicating less than a high school degree (omitted group),
ment were measured with the Woodcock–Johnson Psycho-
a high school or vocational degree, some college, or a bachelor’s
Educational Battery (WJ–R; Woodcock & Johnson, 1989), admin-
degree or more. Maternal employment was reported in first, third,
istered in first, third, and fifth grades by field interviewers. We and fifth grades and was represented continuously as the number
used W ability scores for two subscales: Letter–Word Identifica- of hours that mothers worked during the week. Finally, maternal
tion (reading achievement) and Applied Problems (math achieve- scores on the Peabody Picture Vocabulary Test—Revised (Dunn
ment). W scores are a special transformation of the Rasch Ability & Dunn, 1981) when children were 36 months old were included
Scale with mathematical properties that facilitate the interpretation as continuous variables in predictive models to account for the
of test performance and make it easier to document change over possible confounding of maternal cognitive competence on chil-
time. W scores provide a common scale of equal-interval mea- dren’s own achievement. Split-half correlations for reliability on
surement that represents both a person’s ability and the task this scale ranged from .80 to .83 in a standardization sample (Dunn
difficulty. The W scale for each test is centered on a value of 500, & Dunn, 1981).
which has been set to approximate the average performance at age Household characteristics. Household income was repre-
10 years 0 months. The W score for any cluster is the average W sented by an income-to-needs ratio when the child was 54 months
score for the tests included in the cluster. The developers of the old, calculated by dividing the family income by the poverty
WJ–R argue that the W score can be considered a growth scale. threshold for the household. Poverty status, measured at first, third,
The Letter–Word Identification subscale consists of 57 items, with and fifth grades, reflected whether mothers reported receiving food
higher values indicating better reading identification skills and stamps or assistance from Aid to Families With Dependent Chil-
word decoding. The Applied Problems subscale includes 60 items dren or the Special Supplemental Nutrition Program for Women,
with higher scores indicating better skills to analyze and solve Infants, and Children (receipt of aid ⫽ 1; no receipt of aid ⫽ 0).
mathematical problems. The WJ–R is a nationally normed and Mother’s caregiving burden was represented by the number of
widely used achievement test with demonstrated internal consis- children living in the household at first, third, and fifth grades.
tency. Average internal consistency from first through fifth grades Finally, a measure of developmental stimulation and support avail-
in the NICHD sample is .90 for Letter–Word Identification and .83 able in the home was provided by the HOME at 54 months
for Applied Problems (NICHD Early Child Care Research Net- (Caldwell & Bradley, 1984). Higher scores indicate a home envi-
work, 2007). Original test–retest reliability ranged from .60 to .90 ronment characterized by higher levels of child stimulation and
across individual subscales. support. There are 55 total items in the scale and raters use a
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 507

yes–no checklist during scoring. Thus, possible HOME scores in In all analyses used to answer the main research questions for
the current study ranged from 18 to 55. Every 4 months, each the study, repeated measures were nested in students, who were
HOME observer coded videotaped visits, and the coding was nested in sites. The intraclass correlation indicating the level of
compared with global standard codes. All observers were required between-child variance was substantial: 65.27% for math achieve-
to maintain a criterion of scoring like the master coder on 90% of ment and 64.63% for reading achievement. In contrast, the
the items. This measure has been validated via correlations with between-site variation was quite small: 0.49% for math achieve-
family social status and maternal IQ (Caldwell & Bradley, 1984). ment and 0.38% for reading achievement, consistent across study
Further, the measure has demonstrated excellent reliability (␣ ⫽ years. We used a likelihood ratio test to compare a two-level
.93; Caldwell & Bradley, 1984). Test–retest reliability, assessed unconditional growth model (Level 1 ⫽ repeated measures; Level
across 18 months, has been demonstrated to be moderate. 2 ⫽ students) with a three-level unconditional growth model
Teacher characteristics. Information about teacher experi- (Level 1 ⫽ repeated measures; Level 2 ⫽ students; Level 3 ⫽ site)
ence was collected once a year in first, third, and fifth grades and and found a significant improvement in model fit (␹2 ⫽ 908.12,
was measured continuously as the total number of years of teach- p ⬍ .01). However, we then compared the three-level model to a
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ing experience. two-level model with site fixed effects (N ⫽ 10 sites) and observed
This document is copyrighted by the American Psychological Association or one of its allied publishers.

an additional statistically significant improvement (␹2 ⫽ 14.64,


p ⬍ .05). Thus, given the complexity of the model, we decided to
Analytic Approach use the more parsimonious two-level model with site fixed effects
Missing data. As discussed earlier, analyses for the current to account for potential site level differences in the outcomes. We
study relied on outcomes collected in the first, third and fifth used restricted maximum-likelihood estimation (REML) in models
grades. As such, we included the 1,099 children and families (of estimating fixed effects parameters and variance components.
the original 1,364 enrolled in the study) who had data for the REML is similar to maximum-likelihood estimation but uses an
outcomes for at least one time point. In the sample consisting of unbiased estimator considered to be more conservative (Fitzmau-
nonattriting participants, missingness was low to moderate for both rice, Laird, & Ware, 2011). In addition, we included a random
predictors and outcomes, ranging from 0 to 16%. Because there slope for time (Grade) and allowed for a correlation between the
were 1,118 participants and we use data from three time points, random intercept and random slope (␳). The unconditional growth
there were a maximum of 3,354 cases (person periods) in the model is as follows:
absence of missing data. However, not all participants answered at
each time point. To recover these lost participants and cases, we Level 1: Y ti ⫽ ␤0i ⫹ ␤1i(Grade ⫺ 1)ti ⫹ ␧ti
used multiple imputation (Rubin, 1987), imputing 10 separate data
sets by chained equations using SAS PROC MI in SAS Version Level 2: ␤0ji ⫽ ␥0 ⫹ ai ⫹ r0i
9.3 (Royston & White, 2011; Van Buuren, Boshuizen, & Knook
␤1ij ⫽ ␥1i ⫹ r1i
1999). All of the predictor and outcome variables used in predic-
tive models were included in the imputation procedure. Proper In the Level 1 model, Yti is the achievement at time t of student
inference for the likelihood-based models relies on the assumption i. ␤0i is the achievement score in first grade for student i. ␤1i is the
that data were missing at random (Little & Rubin, 1987). We growth trajectory for student i and εti is a random error term that
examined the research questions for the current study using both a represents the residual (or unexplained) variation in the outcome.
listwise deletion procedure and the findings aggregated from the In the Level 2 model, the students’ achievement score in first grade
multiply imputed samples for the sample of 1,118 participants.2 is modeled as a function of the mean achievement score in first
We present the findings from the multiple imputation procedure, as grade (␥0), a vector for the dummy-coded site fixed effects (ai) that
they did not differ substantially from the results of the listwise adjust for time-invariant characteristics related to site membership,
deletion analysis. and a random effect (r0i) that allows the intercept to vary randomly
Hierarchical linear modeling. We used hierarchical linear around the student (Level 2) mean. The growth rate in achievement
modeling to examine the main research questions for the current for student i is modeled as a function of the mean growth in
study. Hierarchical linear modeling allows one to model change achievement (␥1i) and a random effect that allows students’ tra-
over time in an outcome with repeated measures (Raudenbush & jectories to vary randomly around the student mean trajectory (r1i).
Bryk, 2002; Singer & Willett, 2003). We fitted all models with These analyses assume that the residuals at Level 1 and Level 2 are
SAS PROC MIXED and used grade as a metric of time (Singer, identically distributed normal random variables with a mean equal
1998), centering the grade variable in first grade so that the to zero and a variance of ␴2. In addition, because tests of model
intercept would represent the math and reading achievement levels specifics provided empirical support for a correlation between the
when children were beginning full-day elementary school.3 All random slope and random intercept, we allowed those random
models adjusted for the level of the outcome variable measured effects to covary in all models. For significant outcomes, we
when children were 54 months old (i.e., pre-elementary school). calculated effect sizes using Feingold’s (2009) framework for
This allowed us to examine how teacher– child closeness and identifying effect sizes in multilevel models.
conflict related to math and reading adjustment in elementary
school, adjusting for the level of academic skills children had when 2
We do not include participants lost to attrition prior to elementary
they began school. For ease of interpretation, we centered the level school in any analyses.
of the outcome at 54 months, which was time invariant, around the 3
Of the children in the study sample, 84% attended half-day kindergar-
sample mean. ten so their full-day elementary school began in first grade.
508 MCCORMICK AND O’CONNOR

Research questions. To examine the between-child effects of and conflict (Level 1) predicting math and reading achievement.
teacher– child closeness and conflict on achievement, we created The multilevel equation for testing Research Question 2, which
time-invariant student means of closeness and conflict. These builds on the model from Research Question 1, is as follows:
variables represented students’ average level of closeness and
conflict across elementary school. We centered the time-invariant Level 1: Y ti ⫽ ␤0i ⫹ ␤1i(Grade ⫺ 1)ti ⫹ ␤2i(TCClosness)ti
student means around the sample mean and included those vari-
⫹ ␤3i(TCConflict)ti ⫹ ␪ti ⫹ ␧ti
ables as predictors at Level 2. Because we found that the correla-
tion between the average levels of closeness and conflict was low Level 2: ␤0ij ⫽ ␥0 ⫹ ␤4i共TCClosness兲i ⫹ ␤5i共TCConflict兲i
to moderate (average r ⫽ ⫺.35), we decided to include both
closeness and conflict as predictors in the same model. Such an ⫹ ␤6i共TCClosnessxFemale兲I
approach allows us to isolate the effect of one dimension of
teacher– child relationship quality, over and above the effect of the ⫹ ␤7i共TCConflictxFemale兲i ⫹ ␦i ⫹ ai ⫹ r0i
remaining dimension. In addition to controlling for the level of
␤1ij ⫽ ␥1i ⫹ r1i
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

math or reading achievement at 54 months, we also adjusted for


This document is copyrighted by the American Psychological Association or one of its allied publishers.

child, parent, and teacher characteristics that have been previously ␤2ij ⫽ ␤8i共Female兲i
linked to teacher– child relationships and/or academic achievement
and listed above. ␤3ij ⫽ ␤9i共Female兲i
To examine the within-child effects of teacher– child closeness
and conflict on achievement (or whether changes in closeness and In this model, ␤6 and ␤7 are the Level 2 interactions testing
conflict predict changes in academic achievement), we included gender differences in average associations between teacher– child
time-varying measures of closeness and conflict as predictors in conflict and closeness. ␤8i and ␤9i represent the cross-level inter-
the Level 1 models. We centered the time-varying predictors for actions testing gender differences in within-child effects of close-
closeness and conflict around the individual students’ mean across ness and conflict on achievement across time.
the full study (i.e., person-centered means). As such, a positive Effect sizes. Given the complexity of interpreting the outcome
effect of this time-varying predictor represents an increase over in the current study, we used procedures originally outlined by the
and above a student’s average level of closeness or conflict across NICHD ECCRN and Duncan (2003) to calculate effect sizes of the
elementary school being associated with higher achievement. Be- predictors’ associations with the math and reading achievement W
cause we found that the correlation between the average levels of score outcomes. Effect sizes were calculated when the teacher–
closeness and conflict was low to moderate (r ⫽ ⫺.28 in first child closeness and conflict were significant predictors of aca-
grade, r ⫽ ⫺.38 in third grade, and r ⫽ ⫺.34 in fifth grade), we demic achievement at either the within- or the between-child level.
decided to include both closeness and conflict as predictors in the Effect sizes were computed to show the anticipated difference in
same model. In addition to controlling for the level of math or standard deviation units of the outcome measure between children
reading achievement at 54 months, models also adjusted for all with differing levels of teacher– child relationship quality that
child, parent, and teacher covariates identified above. The multi- differed by one standard deviation (Belsky et al., 2007). Effect
level equation for testing Research Question 1, which builds on the sizes were calculated as the product of the estimated teacher– child
original specification from the unconditional growth model, is relationship quality coefficient and the average standard deviation
listed below of the teacher– child relationship quality variable divided by the
standard deviation for the outcome measure. The effect sizes can
Level 1: Y ti ⫽ ␤0i ⫹ ␤1i(Grade ⫺ 1)ti ⫹ ␤2i(TCClosness)ti be interpreted like a correlation. For example, the average standard
deviation for teacher– child relationship conflict is 5.65; therefore,
⫹ ␤3i(TCConflict)ti ⫹ ␪ti ⫹ ␧ti the effect size compared predicted achievement scores for children
Level 2: ␤0ij ⫽ ␥0 ⫹ ␤4i共TCClosness兲i ⫹ ␤5i共TCConflict兲i whose teacher– child conflict differed by 5.65 points on the overall
scale (that ranged from 7 to 35). Previous studies that have used
⫹ ␦i ⫹ ai ⫹ r0i this calculation have also examined growth models.

␤1ij ⫽ ␥1i ⫹ r1i


Results
In this model, ␤2 and ␤3 are the within-child effects for teacher–
child closeness and conflict. ␤4 and ␤5 represent the between-child Descriptive Statistics
effects for teacher– child closeness and conflict, ␪ is a vector
representing time-varying covariates, and ␦ is a vector representing Means and standard deviations for study variables for the full
time-invariant covariates (including the dummy variable for fe- sample and by gender are displayed in Table 1. In general, study
male). participants’ reading and math achievement scores increased over
Finally, we tested whether between- and within-child effects of time. Teacher– child closeness, in general, decreased over time. In
teacher– child closeness and conflict on math and reading achieve- contrast, teacher– child conflict increased between first and third
ment differed by gender. We included interactions between female grades and then decreased in fifth grade by a nominal amount.
and teacher– child closeness and conflict in the Level 2 model. Relative to girls, boys had significantly higher levels of conflict
Then, to examine whether the within-child effect of teacher– child with teachers across time: In first grade, t(1124) ⫽ 5.40, p ⬍ .01;
closeness and conflict on achievement varied by gender, we in- in third grade, t(1124) ⫽ 6.21, p ⬍ .01; in fifth grade, t(1124) ⫽
cluded a cross-level interaction of female (Level 2) and closeness 5.40, p ⬍ .01. Girls, on average, had closer relationships with
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 509

Table 1
Descriptive Statistics for Study Variables

Full sample Male Female


Variable M SD M SD M SD

Outcome
Reading achievement at 54 months 369.98 21.98 367.86 21.07 372.10 22.66
Reading achievement in first grade 452.95 24.13 451.45 25.22 454.47 22.90
Reading achievement in third grade 493.84 18.73 493.33 19.65 494.25 17.78
Reading achievement in fifth grade 510.11 17.52 509.92 510.29 510.29 16.78
Math achievement at 54 months 425.04 19.48 422.62 21.60 427.45 16.78
Math achievement in first grade 470.97 15.74 472.12 16.39 469.82 14.99
Math achievement in third grade 497.32 13.19 497.85 13.80 496.79 12.55
Math achievement in fifth grade 509.82 12.85 510.24 13.73 509.40 11.91
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Predictor
This document is copyrighted by the American Psychological Association or one of its allied publishers.

Teacher–child closeness in first grade 33.96 5.04 33.11 5.08 34.81 4.87
Teacher–child closeness in third grade 33.08 5.16 31.94 5.50 34.21 4.53
Teacher–child closeness in fifth grade 31.85 5.37 30.96 5.53 32.73 5.06
Teacher–child conflict in first grade 10.92 5.17 11.79 5.51 10.06 4.66
Teacher–child conflict in third grade 11.63 6.03 12.81 6.54 10.46 5.24
Teacher–child conflict in fifth grade 11.33 5.75 12.44 6.20 10.44 5.06
Maternal characteristic
Mother’s IQ at 36 months 98.98 18.33
Mother’s work hr/week at first grade 26.49 19.01
Mother’s work hr/week at third grade 26.79 19.06
Mother’s work hr/week at fifth grade 27.69 18.98
Household characteristics
Income to needs ratio at 54 months 3.59 3.17
HOME observation at 54 months 46.00 5.46
Number of children in HH in first grade 2.39 0.95
Number of children in HH in third grade 2.41 0.98
Number of children in HH in fifth grade 2.48 1.04
Public assistance status in first grade 0.06
Public assistance status in third grade 0.07
Public assistance status in fifth grade 0.06
Teacher characteristics
Teacher years of experience in first grade 14.53 9.48
Teacher years of experience in third grade 13.97 10.44
Teacher years of experience in fifth grade 14.61 10.68
Note. HOME ⫽ Home Observation for Measurement of the Environment; HH ⫽ household.

teachers than boys did at all three time points: In first grade, the Level 1 (within-child effects) models. Results (see Table 2)
t(1124) ⫽ ⫺5.41, p ⬍ .01; in third grade, t(1124) ⫽ ⫺7.07, p ⬍ suggest a between-child effect of teacher– child conflict on reading
.01; in fifth grade, t(1124) ⫽ ⫺5.07, p ⬍ .01. And although boys achievement (␥ ⫽ ⫺0.25, SE ⫽ 0.12, p ⫽ .03; effect size (ES) ⫽
had higher levels of math and reading achievement than girls .06). This effect can be interpreted such that, on average, a 1-unit
in first grade—math, t(1124) ⫽ ⫺4.05, p ⬍ .01; reading, increase in average teacher– child conflict over and above the
t(1124) ⫽ ⫺2.02, p ⫽ .02—those differences were not statistically sample mean level is associated with a .25-unit decrease in reading
significant in third and fifth grades. achievement, adjusting for child, family, and teacher covariates as
Unconditional growth model. The results of the two-level well as site fixed effects. However, the between-child effect of
unconditional growth models showed a significant random inter- teacher– child conflict was nonsignificant for math achievement
cept in first grade for math achievement (␥ ⫽ 473.04, p ⬍ .01) and (␥ ⫽ ⫺0.07, SE ⫽ 0.08, p ⫽ .31; ES ⫽ .01). In addition, the
reading achievement (␥ ⫽ 456.87, p ⬍ .01). Findings indicated between-child effect of teacher– child closeness was nonsignificant
that scores increased on average 9.81 (p ⬍ .01) points in every for math (␥ ⫽ ⫺0.04, SE ⫽ 0.08, p ⫽ .61; ES ⫽ .02) and reading
grade for math achievement and 14.36 points (p ⬍ .01) for reading achievement (␥ ⫽ ⫺0.21, SE ⫽ 0.13, p ⫽ .22; ES ⫽ .05).
achievement. Furthermore, variance component estimates demon- Results of the within-child effects of teacher– child conflict and
strated significant variation around the Level 2 intercept (for math closeness and math and reading achievement (see Table 2) re-
achievement, ␶ ⫽ 175.13, p ⬍ .01; for reading achievement, ␶ ⫽ vealed a statistically significant within-child effect of teacher–
495.59, p ⬍ .01). There was also variation in the slope for both the child closeness on reading achievement (␥ ⫽ 0.17, SE ⫽ 0.08, p ⫽
math (␶ ⫽ .56, p ⬍ .01) and reading (␶ ⫽ 2.83, p ⬍ .01) .04; ES ⫽ .09). This finding suggests that, on average, a 1-unit
achievement models and significant correlations between the ran- increase in teacher– child closeness over and above one’s own
dom slope and intercept in both models (math, ␶ ⫽ ⫺9.88, p ⬍ average teacher– child closeness is associated with a 0.17 unit
.01; reading, ␶ ⫽ ⫺37.42, p ⬍ .01). higher score in reading achievement. We found that the within-
Research Question 1. Below, we report the results of the child effects of teacher– child closeness were nonsignificant for
Level 2 models (between-child effects), followed by results from math achievement (␥ ⫽ 0.03, SE ⫽ 0.06, p ⫽ .57; ES ⫽ .02). In
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

510

Table 2
Growth Models Predicting Academic Skill Development From Teacher–Child Closeness and Conflict

Unconditional model Main effects model Moderated model


Math Reading Math Reading Math Reading
Fixed effects ␥ SE ␥ SE ␥ SE ␥ SE ␥ SE ␥ SE

Between-child estimates
Intercept 473.04ⴱⴱ 0.47 456.87ⴱⴱ 0.71 439.77ⴱⴱ 9.92 412.26ⴱⴱ 15.53 440.35ⴱⴱ 9.89 412.60 15.66
Grade 9.81ⴱⴱ 0.10 14.36ⴱⴱ 0.15 9.85ⴱⴱ 0.11 14.67ⴱⴱ 0.17 9.84ⴱⴱ 0.11 14.67ⴱⴱ 0.17
Teacher–child closeness ⫺.04 .08 ⫺.21 .13 .01 .11 ⫺.20 .17
Teacher–child conflict ⫺.07 .08 ⫺.25ⴱ .12 .06 .10 ⫺.19 .15
Female ⫺3.72ⴱⴱ 0.62 ⫺0.98 0.97 ⫺3.88ⴱⴱ 0.62 ⫺1.03 0.97
Female ⫻ Teacher–Child Closeness ⫺.07 .16 ⫺.02 .25
Female ⫻ Teacher–Child Conflict ⫺.30ⴱ .14 ⫺.16 .22
Within-child estimates
Teacher–child closeness .03 .06 .17ⴱ .08 .08 .07 .15 .11
Teacher–child conflict .01 .06 ⫺.03 .09 .10 .08 .03 .11
Female ⫻ Teacher–Child Closeness ⫺.10 .11 .05 .16
Female ⫻ Teacher–Child Conflict ⫺.21ⴱ .10 ⫺.13 .17
Covariates
Reading achievement, 54 months .34ⴱⴱ .02 .34ⴱⴱ .02
Math achievement, 54 months .31ⴱⴱ .02 .31ⴱⴱ .02
Child Black ⫺5.04ⴱⴱ 1.69 ⫺1.03 2.63 ⫺4.99ⴱⴱ 1.69 ⫺1.04 2.63
Child White ⫺2.49 1.34 0.62 2.10 ⫺2.51† 1.33 0.59 2.10
Child Hispanic ⫺1.32 1.34 0.96 2.11 ⫺1.28 1.34 0.96 2.11
Maternal ed., high school graduate 3.13ⴱⴱ 1.35 5.80ⴱⴱ 2.12 3.12ⴱ 1.34 5.81ⴱⴱ 2.12
Maternal ed., some college 2.61ⴱⴱ 1.37 5.84ⴱⴱ 2.16 2.71ⴱ 1.36 5.91ⴱⴱ 2.16
MCCORMICK AND O’CONNOR

Maternal ed., college degree 4.42ⴱⴱ 1.52 6.52ⴱⴱ 2.40 4.42ⴱⴱ 1.52 6.56ⴱⴱ 2.40
Maternal employment .01 .02 ⫺.02 .03 .01 .02 ⫺.02 .03
Mother’s IQ .04ⴱ .02 .16ⴱⴱ .04 .04ⴱ .02 .16ⴱⴱ .03
Income to needs ratio .21ⴱ .11 .03 .17 .21ⴱ .11 .03 .17
Poverty status ⫺0.15 1.16 ⫺3.51ⴱⴱ 1.75 ⫺0.15 1.16 ⫺3.51ⴱ 1.75
Mother’s caregiving burden ⫺.34 .34 ⫺.97† .54 ⫺.31 .34 ⫺.94† .54
HOME observation ⫺.17ⴱ .08 .16 .12 .17ⴱ .08 .16 .12
Teacher years experience .07 .04 .05 .07 .07 .04 .05 .07

Random effects and fit statistics


Intercept, r0 175.13ⴱⴱ 14.5 495.59ⴱⴱ 31.2 69.81ⴱⴱ 6.52 227.93ⴱⴱ 18.27 69.27ⴱⴱ 6.49 227.68ⴱⴱ 18.24
Time slope, r4 0.56ⴱⴱ 0.18 2.83ⴱⴱ 0.52 0.52ⴱⴱ 0.19 2.81ⴱⴱ 0.65 0.52ⴱⴱ 0.19 2.81ⴱⴱ 0.65
ⴱⴱ ⴱⴱ
Covariance, slope and intercept ⫺9.88ⴱⴱ 1.87 ⫺37.42 4.46 ⫺6.05 1.33 ⫺25.29ⴱⴱ 3.74 ⫺6.02ⴱⴱ 1.33 ⫺25.28ⴱⴱ 3.74
Level 1, e 70.30ⴱⴱ 2.82 149.2728ⴱⴱ 4.84 69.64ⴱⴱ 2.62 147.69ⴱⴱ 5.61 69.43ⴱⴱ 2.62 147.65ⴱⴱ 5.60
AIC 23,509.78 25,847.26 17,600.65 19,627.75 17,601.04 19,634.35
Note. Coefficients displayed adjust for site fixed effects. HOME ⫽ Home Observation for Measurement of the Environment; ed. ⫽ education; AIC ⫽ Akaike’s information criterion.

p ⬍ .10. ⴱ p ⬍ .05. ⴱⴱ p ⬍ .01.
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 511

addition, the within-child effects of teacher-child conflict were


nonsignificant for both math (␥ ⫽ 0.01, SE ⫽ 0.06, p ⫽ .71; ES ⫽
.01) and reading achievement (␥ ⫽ ⫺0.03, SE ⫽ 0.09, p ⫽ .73;
ES ⫽ .02).
For math achievement, the main effects model showed signifi-
cant between-student variation (␶ ⫽ 69.81, SE ⫽ 6.52, p ⬍ .01).
The within-person random effect remained significant in this
model (␶ ⫽ 69.64, SE ⫽ 2.62, p ⬍ .01), suggesting additional
unexplained variation in within-person changes in math achieve-
ment. An examination of the fit statistics for these models suggests
a statistically significant improvement in the log likelihood for the
main effects model compared with the unconditional growth
model, ␹2(21) ⫽ 5,999.12, p ⬍ .01.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

For reading achievement, the main effects model showed sig-


This document is copyrighted by the American Psychological Association or one of its allied publishers.

nificant between-student variation (␶ ⫽ 22.93, SE ⫽ 18.27, p ⬍


.01). However, the within-person random effect remained signifi-
cant in this model (␶ ⫽ 147.69, SE ⫽ 5.61, p ⬍ .01), suggesting
additional unexplained variation in within-person changes in read-
ing achievement. An examination of the fit statistics for these
models suggests a statistically significant improvement in the log
likelihood for the main effects model compared with the uncon-
ditional growth model, ␹2(21) ⫽ 6,300.24, p ⬍ .01.
Figure 1. Between-child effects of teacher– child conflict predicting math
Research Question 2. Finally, we tested whether between- achievement in elementary school, by gender. The graph adjusts for all
and within-child effects of teacher– child closeness on math and covariates. High teacher– child conflict was operationalized as 1 standard
reading achievement differed by gender. As shown in Table 2, deviation above the mean; low teacher– child conflict was operationalized
there was a significant negative between-child effect of teacher– as 1 standard deviation below the mean.
child conflict on math achievement for girls, relative to boys (␥ ⫽
0.30, SE ⫽ 0.14, p ⫽ .03; ES ⫽ .12). In addition, growth in math
achievement was slower for girls with conflictual teacher– child sistent across models when including the lagged effects, although
relationships, relative to girls with nonconflictual relationships and the effect sizes were a bit smaller.
boys with conflictual relationships (␥ ⫽ 0.21, SE ⫽ 0.10, p ⫽ .04;
ES ⫽ .17). Figure 1 graphs the effects of teacher– child conflict on Discussion
math achievement, by gender, for girls and boys whose teachers
Results of this study advance the understanding of the longitu-
report high (1 standard deviation) and low (⫺1 standard deviation)
dinal connections between teacher– child relationships and aca-
levels of conflict. Both between- and within-child effects of demic development during elementary school. Using a large lon-
teacher– child closeness and conflict on math and reading achieve- gitudinal database, this study is the first to examine both between-
ment did not differ by gender. In addition, there were no gender and within-child associations between teacher– child closeness and
differences in the between- and within-child effects of teacher– conflict and children’s standardized measures of math and reading
child closeness on reading achievement. achievement across elementary school. In addition, by testing
Post hoc analyses. Given the relatively large number of sta- whether these associations varied by gender, we were able to
tistical tests examined, we also ran the models with a Bonferroni identify differential effects that may have important implications
correction to account for the possibility that there was a multiple for practitioners. Results revealed that overall, higher levels of
comparisons problems. Results were consistent although the stan- teacher– child conflict are related to lower levels of reading
dard errors for most variables of interest were a bit inflated after achievement in elementary school, over and above a host of child,
including the Bonferroni correction. A second concern with anal- family, and teacher control variables. In addition, improvements in
yses is that given the relatively small period of time between when teacher– child closeness across elementary school were related to
teachers reported on their relationships with children and when the gains in reading achievement. After testing whether effects dif-
children were assessed for achievement, we cannot necessarily be fered by gender, we also found that girls who had conflictual
sure that the effect of the teacher– child relationship on achieve- relationships with teachers experienced lower average levels of
ment is concurrent. Additionally, because achievement is accumu- math achievement than did boys with conflictual relationships with
lated over time, there may be a lagged effect of teacher– child teachers. In addition, girls with conflictual relationships with
relationship quality. To correct for this possibility, we also tested teachers showed slower growth in math achievement across time
models that included the previous year’s teacher– child relation- than did boys with conflictual teacher– child relationships. Exam-
ship quality as lagged effects in predictive models. Given the ination of the effect sizes by gender suggest that these associations
availability of data, we used the kindergarten teacher– child rela- are small (ranging in effect size from .06 to .17). However, they
tionship as the lag for first grade, the first grade relationship as the are similar in size if not larger to studies that have identified
lag for third grade, and the fourth grade relationship as the lag for significant associations between key predictors of math and read-
fifth grade achievement. Again, we found that results were con- ing achievement such as child care quality (Belsky et al., 2007;
512 MCCORMICK AND O’CONNOR

Vandell, Belsky, Burchinal, Steinberg, & Vandergrift, 2010), par- girls. Although boys had higher overall levels of conflict with
ent involvement (El Nokali, Bachman, & Votruba-Drzal, 2010), teachers, this conflict appeared to matter less for boys than for girls
and academic engagement (Dotterer & Lowe, 2011). in terms of math achievement. This is a notable finding, given that
The results of the main between-child analyses confirm prior girls tend to perform worse than boys do on standardized assess-
research that identified longitudinal associations between teacher– ments of math achievement (Robinson & Lubienski, 2011).
child relationships and reading achievement (Hughes, 2011; Koni- Consistent with a gender socialization perspective (e.g., Baker,
shi, Hymel, Zumbo, & Li, 2010). For example, Maldonado- 2006), having conflictual relationships with teachers may exacer-
Carreño and Votruba-Drzal (2011), using the same sample as in bate girls’ difficulties in learning math. For example, because girls
the current study, found beneficial effects of positive teacher– child are expected to be model students, when they do experience
relationships on teacher reports of reading achievement across conflict with teachers, it is more likely to make them internalize
elementary school. Our findings are somewhat unique, however, feelings of depression and anxiety and perhaps decrease self-
and extend prior work, as we were able to differentiate between- esteem, academic efficacy, and engagement. In contrast, because
and within-child effects specific to teacher– child closeness and boys are not necessarily expected to be model students, experienc-
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

conflict. Although we did find that overall, having a conflictual ing a lower quality relationship with the teacher may not lead to
This document is copyrighted by the American Psychological Association or one of its allied publishers.

teacher– child relationship was associated with poorer reading this series of internalizing behaviors. Thus, girls’ math achieve-
achievement, we did not find that increases in conflict were asso- ment suffers because of a conflictual relationship with a teacher,
ciated with decreases in reading achievement over time. Rather, it whereas boys’ math achievement is less affected. One possible
was closeness that seemed to matter more for reading development implication for this finding concerns teachers’ differential treat-
across elementary school. Within a close relationship, teachers are ment of students (McKown & Weinstein, 2008). Indeed, teachers
able to provide children with positive supports and reinforcement are somewhat inclined to base relationships with students on
that encourage children to develop emotional and behavioral reg- gender stereotypes (e.g., Baker, 2006; Ewing & Taylor, 2009).
ulation (Howes & Hamilton, 1993). This type of closeness may be Using a national sample of children in elementary and middle
particularly important for developing reading competencies school, Lubienski, Robinson, Crane, and Ganley (2013) found that
wherein children must maintain continued engagement in the teachers consistently underrated girls’ math abilities, relative to
academic material and practice those competencies outside of the their standardized math scores. Teachers should be attuned to their
school context. role in using emotionally supportive strategies when teaching girls
The finding relating teacher– child closeness to changes in read- math. This is especially important given rigorous research indicat-
ing achievement is in line with previous research that considered ing that math achievement is the single most powerful predictor of
multiple dimensions of teacher– child relationships. For example, future educational attainment (Duncan et al., 2007).
in their meta-analysis examining studies of teacher– child relation- In addition, although boys are generally thought to be at risk for
ships on overall achievement levels, Roorda et al. (2011) had low overall achievement (Hamre & Pianta, 2005), the academic
initially expected to find stronger associations for negative (e.g., risk perspective may help explain the significant effect of teacher–
conflict) components of teacher– child relationships, relative to child conflict on girls’ math achievement. Research has shown that
positive dimensions of teacher– child relationships (e.g., close- girls, relative to boys, are more likely to perform lower on stan-
ness). Although this expectation was based in prior research (e.g., dardized assessments of math (Robinson & Lubienski, 2011).
Baker, 2006; Hamre & Pianta, 2001), the meta-analysis actually Thus, by experiencing a conflictual teacher– child relationship,
revealed that closeness was just as important as conflict for aca- girls may be at heightened risk for less than optimal math achieve-
demic outcomes. Moreover, as children became older, closeness ment specifically. Originally, we expected that boys would be at
seemed to matter more than conflict for academic outcomes (Ro- higher risk for poor reading achievement, thus standing to benefit
orda et al., 2011). Initially, it seems that closeness would be more most from closer, nonconflictual teacher– child relationships.
important for younger children’s development, given how critical However, in contrast to this expectation, our data suggested that
emotional support is for students transitioning to elementary boys had slightly higher initial reading achievement than girls and
school (Rimm-Kaufman & Pianta, 2000). However, it may be that did not differ in reading scores from girls across later elementary
teacher– child closeness is also vital for older students’ develop- school. Thus, the nonsignificant difference in reading achievement
ment, perhaps because they need to feel supported in the classroom by gender across time may explain the lack of a moderated effect
to take the academic risks necessary to continually develop more for relationship quality by gender. Overall, findings suggest that
advanced reading skills. It may be that because the current study one may need to align the academic risk perspective with the
examined effects throughout middle childhood, rather than focus- specific academic outcome of interest rather than group all aca-
ing on younger children specifically, the effect of closeness was demic outcomes together in future studies.
particularly critical for reading achievement outcomes.
Similarly surprising is the fact that we did not find any overall
Strengths and Limitations
effects of teacher– child closeness or conflict on math achieve-
ment. There is growing research suggesting that teacher relation- The current study has a number of strengths that help to build
ships may matter more for the development of math competencies the research base on teacher– child relationships and achievement.
than reading skills (Crosnoe, Leventhal, Wirth, Pierce, & Pianta, First, this study used a series of rigorous longitudinal models,
2010; McCormick et al., 2013). However, after examining differ- incorporating data from three time points in elementary school to
ential associations between teacher– child relationships and model effects of teacher– child closeness and conflict on achieve-
achievement by gender, we did find both between- and within- ment. As such, we were able to examine both within- and between-
child effects of teacher– child conflict on math achievement for child effects on achievement. In addition, we examined closeness
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 513

and conflict as separate dimensions of teacher– child relationship excluded. The current sample may contain relatively few high-risk
quality and included both dimensions as independent variables in children, thus reducing the magnitude of associations between
predictive models. This was an important consideration, given that teacher– child relationships and academic outcomes. Moreover, we
we did identify differential effects of these dimensions on achieve- must caution readers that the NICHD SECCYD is a low-risk
ment outcomes and are thus able to interpret the effects on sample and may not generalize to populations (e.g., low income,
achievement as being over and above the remaining dimension of immigrant, racial or ethnic minority) that are increasingly coming
teacher– child relationship quality. to represent the broad population of U.S. school children.
Third, we operationalized academic achievement using a stan- Next, we only had three time points of academic achievement
dardized assessment rather than relying on teacher reports of data for this study. Although we were able to examine longitudinal
achievement. This approach allowed us to limit threats to the growth models, we could not test nonlinearity with three time
internal validity of the study posed by rater effects. Notably, we points of data. Future studies should consider how the form of the
examined math and reading achievement separately by gender in a relationship may vary across time. For example, it might be that
large national data set. The academic achievement trends we teacher– child relationships are most important in third grade but
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

identified by gender are important to consider given that the data become less important as children begin to transition into adoles-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

were collected for the purpose of examining developmental tra- cence at the end of elementary school. It may also be important for
jectories and children were randomly selected to participate in the future studies to consider the gender match between the teacher
study before educational differences emerged. Next, we were able and child. We were unable to explicitly test differences by gender
to model trajectories of teacher– child closeness and conflict and match because the overwhelming majority of teachers in the sam-
achievement differentially by gender. These moderated models ple were female. Finally, although the longitudinal models allowed
provided insight into how dimensions of teacher– child relation- us to examine average effects and within-child effects over time,
ship quality differentially predicted a standardized assessment of this study is still correlational in interpretation and we cannot infer
achievement. Finally, we used a large national sample to examine causality from the longitudinal design. Specifically, we still cannot
these effects and controlled for a host of demographic, home-, and be fully certain about the direction of effects in the current study.
school-based covariates. The use of the national sample is critical, Assuming selection bias, it is possible that achievement could
as we now have the ability to generalize results related to the actually be driving the observed levels of teacher– child closeness
gender socialization perspective that had previously been identi- and conflict. Indeed, there was variation in the data collection
fied only in within-group samples (e.g., Green et al., 2008; Ly et process regarding when the teacher reports of relationship quality
al., 2012). were collected relative to when the WJ–R assessments were done.
However, even given these strengths, the current study has a It will be important for researchers conducting future studies to
number of limitations. First, we used a teacher report of teacher– consider methods for examining the causal effects of teacher– child
child relationship quality to operationalize the main predictor for closeness and conflict on reading and math achievement in ele-
the study. Relatedly, the teachers who were reporting changed at mentary school and examine how those effects differ by gender.
each time point, thus potentially introducing additional measure- For example, researchers conducting random assignment studies
ment error. We were only able to account for teacher experience as that aim to improve relational and emotional supports in the
a potential teacher-level confounding covariate. Because of the classroom should consider examining treatment effects on math
data collection, children were not nested within teachers, thus and achievement and test whether those effects differ by gender.
limiting our ability to include teacher fixed effects in models to
account for teacher-level time-invariant differences. Future studies
Implications
in which children are nested in teachers should consider adjusting
for teacher-level differences and should extend this research and This study may have important implications for teacher training
consider child-reported and observed measures of teacher– child and professional development. First, teachers’ role in promoting an
relationship quality to reduce possible biases posed by using emotionally supportive context and creating warm, respectful re-
teacher reports (Hughes, 2011). Second, we did not account for the lationships with all students should not be overlooked because of
possibility that closeness and conflict might have additive or concerns about curriculum and direct instruction. Indeed, we found
interactive effects on achievement. A future study, however, will that although closeness tended to decline across elementary school,
use a person-centered approach to model patterns of closeness and the associations between a close teacher relationship and reading
conflict to determine differential trajectories in achievement. Next, achievement remained constant across time. Results suggest the
we were unable to examine mechanisms linking teacher– child potential need to emphasize formation and maintenance of close
closeness and conflict and achievement and were also unable to relationships with students across the course of elementary school.
test whether mechanisms differed by gender. Although having Given the findings from this study combined with extant literature
conflictual relationships with teachers may affect girls’ internaliz- on social-emotional learning, it may be important to embed rela-
ing behaviors (e.g., depression, anxiety), boys may be more likely tionships into instructional practice to successfully promote aca-
to engage in externalizing behaviors if they experience conflictual demic competencies in reading and math. It may also be beneficial
teacher– child relationships. Thus, future research examining for teachers to be trained on how gender differences and their
mechanisms should consider how effects differ by gender. Third, implicit differential reactions to behaviors engaged in by boys and
there are certain limitations to our findings as a result of recruit- by girls are associated with student achievement. As work on
ment and enrollment methods of the NICHD SECCYD. Specifi- teacher expectations has shown (e.g., McKown & Weinstein,
cally, children with disabilities, children who lived in dangerous 2008), teachers are rarely aware of implicit biases they have
areas, and children whose mothers did not speak English were toward students. As such, it may be critical to share this informa-
514 MCCORMICK AND O’CONNOR

tion with teachers and provide them with training on how best to ulation in early childhood. Child Development, 81, 972–987. doi:
support boys’ and girls’ successful academic development. 10.1111/j.1467-8624.2010.01446.x
This study also has implications for policy development and Dearing, E., McCartney, K., & Taylor, B. A. (2006). Within-child associ-
implementation. Educational policy has put increased emphasis on ations between family income and externalizing and internalizing prob-
teacher accountability systems that use student outcomes to gauge lems. Developmental Psychology, 42, 237–252. doi:10.1037/0012-1649
teacher quality. However, this study provides preliminary evidence .42.2.237
that teachers who create emotionally supportive relationships with Dotterer, A. M., & Lowe, K. (2011). Classroom context, school engage-
ment, and academic achievement in early adolescence. Journal of Youth
their students may also be successful at developing academic
and Adolescence, 40, 1649 –1660. doi:10.1007/s10964-011-9647-5
competencies. Thus, it may be important for policymakers to
Duncan, G. J. (2011, January). The importance of academic skills for
consider teachers’ emotionally supportive practices when creating
preK–3rd [Report]. Retrieved from Foundation for Child Development
and refining teacher evaluation systems. When teachers are pro- website: http://fcd-us.org/resources/importance-academic-skills-prek-
vided with information about their teaching practices and relation- 3rd
ship building, they can reflect on their practice to improve this Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston, A. C.,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

relational component. Studies continue to show that accountability Klebanov, P., . . . Japel, C. (2007). School readiness and later achieve-
This document is copyrighted by the American Psychological Association or one of its allied publishers.

systems alone are unsuccessful in boosting overall levels of aca- ment. Developmental Psychology, 43, 1428 –1446. doi:10.1037/0012-
demic achievement in the United States (e.g., Reardon, 2011). By 1649.43.6.1428
considering the importance of within-school social processes— Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Test—Revised
teacher– child relationships specifically—future policies may be manual for Forms L and M. Circle Pines, MN: American Guidance
able to address a new direction for raising overall achievement and Service.
improving outcomes for American children. El Nokali, N. E., Bachman, H. J., & Votruba-Drzal, E. (2010). Parent
involvement and children’s academic and social development in elemen-
tary school. Child Development, 81, 988 –1005. doi:10.1111/j.1467-
References 8624.2010.01447.x
Ewing, A. R., & Taylor, A. R. (2009). The role of child gender and
Baker, J. A. (2006). Contributions of teacher– child relationships to posi-
ethnicity in teacher– child relationship quality and children’s behavioral
tive school adjustment during elementary school. Journal of School
adjustment in preschool. Early Childhood Research Quarterly, 24, 92–
Psychology, 44, 211–229. doi:10.1016/j.jsp.2006.02.002
105. doi:10.1016/j.ecresq.2008.09.002
Baker, J. A., Grant, S., & Morlock, L. (2008). The teacher–student rela-
Feingold, A. (2009). Effect sizes for growth-modeling analysis for con-
tionship as a developmental context for children with internalizing or
externalizing behavior problems. School Psychology Quarterly, 23(1), trolled clinical trials in the same metric as for classical analysis. Psy-
3–15. chological Methods, 14, 43–53.
Belsky, J., Vandell, D. L., Burchinal, M., Clarke-Stewart, K. A., McCart- Fitzmaurice, L., Laird, N. M., & Ware, J. H. (2004). Applied longitudinal
ney, K., & Owen, M. T. (2007). Are there long-term effects of early analysis. Hoboken, NJ: Wiley.
child care? Child Development, 78, 681–701. doi:10.1111/j.1467-8624 Fitzmaurice, G., Laird, N. M., & Ware, J. (2011). Applied longitudinal
.2007.01021.x analysis (2nd ed.). Hoboken, NJ: Wiley-Interscience.
Birch, S. H., & Ladd, G. W. (1997). The teacher– child relationship and Gelman, A., & Hill, J. (2007). Data analysis using regression and multi-
children’s early school adjustment. Journal of School Psychology, 35, level/hierarchical models. Cambridge, United Kingdom: Cambridge
61–79. doi:10.1016/S0022-4405(96)00029-5 University Press.
Brechwald, W. A., & Prinstein, M. J. (2011). Beyond homophily: A decade Green, G., Rhodes, J., Hirsch, A. H., Suárez-Orozco, C., & Camic, P. M.
of advances in understanding peer influence processes. Journal of Re- (2008). Supportive adult relationships and the academic engagement of
search on Adolescence, 21, 166 –179. doi:10.1111/j.1532-7795.2010 Latin American immigrant youth. Journal of School Psychology, 46,
.00721.x 393– 412. doi:10.1016/j.jsp.2007.07.001
Bronfenbrenner, U., & Morris, P. A. (1998). The ecology of developmental Hamre, B. K., & Pianta, R. C. (2001). Early teacher– child relationships
processes. In W. Damon (Series Ed.) & R. M. Lerner (Vol. Ed.), and the trajectory of children’s school outcomes through eighth grade.
Handbook of child psychology: Vol. 1. Theoretical models of human Child Development, 72, 625– 638. doi:10.1111/1467-8624.00301
development (5th ed., pp. 993–1028). New York, NY: Wiley. Hamre, B. K., & Pianta, R. C. (2005). Can instructional and emotional
Burchinal, M. R., Peisner-Feinberg, E., Pianta, R., & Howes, C. (2002). support in the first-grade classroom make a difference for children at risk
Development of academic skills from preschool through second grade: of school failure? Child Development, 76, 949 –967.
Family and classroom predictors of developmental trajectories. Journal
Hibel, J., Farkas, G., & Morgan, P. L. (2010). Who is placed into special
of School Psychology, 40, 415– 436. doi:10.1016/S0022-
education? Sociology of Education, 83, 312–332. doi:10.1177/
4405(02)00107-3
0038040710383518
Buyse, E., Verschueren, K., Verachtert, P., & Van Damme, J. (2009).
Howes, C., & Hamilton, C. E. (1992). Children’s relationships with child
Predicting school adjustment in early elementary school: Impact of
teacher– child relationship quality and relational classroom climate. The care teachers: Stability and concordance with parental attachments.
Elementary School Journal, 110, 119 –141. doi:10.1086/605768 Child Development, 63, 867– 878. doi:10.2307/1131239
Caldwell, B. M., & Bradley, R. H. (1984). Home observation for measure- Howes, C., & Hamilton, C. E. (1993). The changing experience of child
ment of the environment. Little Rock: University of Arkansas at Little care: Changes in teachers and in teacher– child relationships and chil-
Rock. dren’s social competence with peers. Early Childhood Research Quar-
Cooper, D. H., & Farran, D. C. (1988). Behavioral risk factors in kinder- terly, 8, 15–32. doi:10.1016/S0885-2006(05)80096-1
garten. Early Childhood Research Quarterly, 3, 1–19. doi:10.1016/ Howes, C., Phillipsen, L. C., & Peisner-Feinberg, E. (2000). The consis-
0885-2006(88)90026-9 tency of perceived teacher– child relationships between preschool and
Crosnoe, R., Leventhal, T., Wirth, R. J., Pierce, K. M., & Pianta, R. C. kindergarten. Journal of School Psychology, 38, 113–132. doi:10.1016/
(2010). Family socioeconomic status and consistent environmental stim- S0022-4405(99)00044-8
TEACHER–CHILD RELATIONSHIPS, ACADEMIC ACHIEVEMENT 515

Howes, C., & Ritchie, S. (1999). Attachment organizations in children with Morris, E. W. (2011). Bridging the gap: ‘Doing gender,’ ‘hegemonic
difficult life circumstances. Development and Psychopathology, 11, masculinity,’ and the educational troubles of boys. Sociology Compass,
251–268. doi:10.1017/S0954579499002047 5, 92–103. doi:10.1111/j.1751-9020.2010.00351.x
Hughes, J. N. (2011). Longitudinal effects of teacher and student percep- NICHD Early Child Care Research Network. (1997). The effects of infant
tions of teacher–student relationship qualities on academic adjustment. child care on infant–mother attachment security: Results of the NICHD
The Elementary School Journal, 112, 38 – 60. doi:10.1086/660686 Study of Early Child Care. Child Development, 68, 860 – 879.
Hughes, J. N., Cavell, T. A., & Willson, V. (2001). Further support for the NICHD Early Child Care Research Network. (2007). Age of entry to
developmental significance of the quality of the teacher–student rela- kindergarten and children’s academic and socioemotional development.
tionship. Journal of School Psychology, 39, 289 –301. doi:10.1016/ Early Education & Development, 18, 337–368.
S0022-4405(01)00074-7 NICHD Early Child Care Research Network & Duncan, G. J. (2003).
Hughes, J. N., Luo, W., Kwok, O.-M., & Loyd, L. K. (2008). Teacher– Modeling the impacts of child care quality on children’s preschool
student support, effortful engagement, and achievement: A 3-year lon- cognitive development. Child Development, 74, 1454 –1475.
gitudinal study. Journal of Educational Psychology, 100, 1–14. doi: O’Connor, E. (2010). Teacher– child relationships as dynamic systems.
10.1037/0022-0663.100.1.1 Journal of School Psychology, 48, 187–218. doi:10.1016/j.jsp.2010.01
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Hughes, K., Bullock, A., & Coplan, R. J. (2014). A person-centred analysis .001
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of teacher– child relationships in early childhood. British Journal of O’Connor, E., & McCartney, K. (2007). Examining teacher– child relation-
Educational Psychology, 84, 253–267. doi:10.1111/bjep.12029 ships and achievement as part of an ecological model of development.
Jerome, E. M., Hamre, B. K., & Pianta, R. C. (2009). Teacher– child American Educational Research Journal, 44, 340 –369. doi:10.3102/
relationships from kindergarten to sixth grade: Early childhood predic- 0002831207302172
tors of teacher-perceived conflict and closeness. Social Development, 18, Pianta, R. C. (1992). The Student Teacher Relationship Scale. Charlottes-
915–945. doi:10.1111/j.1467-9507.2008.00508.x ville: University of Virginia.
Koch, J. (2003). Gender issues in the classroom. In W. M. Reynolds & Pianta, R. C. (1999). Enhancing relationships between children and teach-
G. E. Miller (Eds.), Handbook of psychology (pp. 259 –281). Hoboken, ers. Washington, DC: American Psychological Association. doi:
NJ: Wiley. doi:10.1002/0471264385.wei0712 10.1037/10314-000
Konishi, C., Hymel, S., Zumbo, B. D., & Li, Z. (2010). Do school bullying Pianta, R. C. (2001). STRS: Student–Teacher Relationship Scale: Profes-
and student–teacher relationships matter for academic achievement? A
sional Manual. Lutz, FL: Psychological Assessment Resources.
multilevel analysis. Canadian Journal of School Psychology, 25, 19 –39.
Pianta, R. C., Belsky, J., Vandergrift, N., Houts, R., Morrison, F., & the
Ladd, G. W., & Burgess, K. B. (1999). Charting the relationship trajecto-
NICHD Early Child Care Research Network. (2008). Classroom effects
ries of aggressive, withdrawn, and aggressive/withdrawn children during
on children’s achievement trajectories in elementary school. American
early grade school. Child Development, 70, 910 –929. doi:10.1111/1467-
Educational Research Journal, 45, 365–397. doi:10.3102/
8624.00066
0002831207308230
Legewie, J., & DiPrete, T. A. (2012). School context and the gender gap in
Pianta, R. C., & Nimetz, S. L. (1991). Relationships between children and
educational achievement. American Sociological Review, 77, 463– 485.
teachers: Associations with classroom and home behavior. Journal of
doi:10.1177/0003122412440802
Applied Developmental Psychology, 12, 379 –393. doi:10.1016/0193-
Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing
3973(91)90007-Q
data (1st ed.). New York, NY: Wiley.
Pianta, R. C., Nimetz, S. L., & Bennett, E. (1997). Mother– child relation-
Lubienski, S. T., Robinson, J. P., Crane, C. C., & Ganley, C. M. (2013).
ships, teacher– child relationships, and school outcomes in preschool and
Girls’ and boys’ mathematics achievement, affect, and experiences:
kindergarten. Early Childhood Research Quarterly, 12, 263–280. doi:
Findings from ECLS-K. Journal for Research in Mathematics Educa-
tion, 44, 634 – 645. doi:10.5951/jresematheduc.44.4.0634 10.1016/S0885-2006(97)90003-X
Ly, J., Zhou, Q., Chu, K., & Chen, S. H. (2012). Teacher– child relationship Pianta, R. C., & Stuhlman, M. W. (2004). Teacher– child relationships and
quality and academic achievement of Chinese American children in children’s success in the first years of school. School Psychology Re-
immigrant families. Journal of School Psychology, 50, 535–553. doi: view, 33, 444 – 458.
10.1016/j.jsp.2012.03.003 Raudenbush, S. W., & Bryk, A. S. (2002). Hierarchical linear models (2nd
Maldonado-Carreño, C., & Votruba-Drzal, E. (2011). Teacher– child rela- ed.). Thousand Oaks, CA: Sage.
tionships and the development of academic and behavioral skills during Reardon, S. F. (2011). The widening academic achievement gap between
elementary school: A within- and between-child analysis. Child Devel- rich and poor: New evidence and possible explanations. In G. J. Duncan
opment, 82, 601– 616. doi:10.1111/j.1467-8624.2010.01533.x & R. J. Murnane (Eds.), Whither opportunity? Rising inequality,
McCartney, K., Dearing, E., Taylor, B. A., & Bub, K. L. (2007). Quality schools, and children’s life chances (pp. 91–115). New York, NY:
child care supports the achievement of low-income children: Direct and Russell Sage Foundation.
indirect pathways through caregiving and the home environment. Jour- Rimm-Kaufman, S. E., & Pianta, R. C. (2000). An ecological perspective
nal of Applied Developmental Psychology, 28, 411– 426. doi:10.1016/j on the transition to kindergarten: A theoretical framework to guide
.appdev.2007.06.010 empirical research. Journal of Applied Developmental Psychology, 21,
McCormick, M. P., O’Connor, E. E., Cappella, E., & McClowry, S. G. 491–511. doi:10.1016/S0193-3973(00)00051-4
(2013). Teacher– child relationships and academic achievement: A mul- Robinson, J. P., & Lubienski, S. T. (2011). The development of gender
tilevel propensity score model approach. Journal of School Psychology, achievement gaps in mathematics and reading during elementary and
51, 611– 624. doi:10.1016/j.jsp.2013.05.001 middle school examining direct cognitive assessments and teacher rat-
McKown, C., & Weinstein, R. S. (2008). Teacher expectations, classroom ings. American Educational Research Journal, 48, 268 –302. doi:
context, and the achievement gap. Journal of School Psychology, 46, 10.3102/0002831210372249
235–261. doi:10.1016/j.jsp.2007.05.001 Roorda, D. L., Koomen, H. M. Y., Spilt, J. L., & Oort, F. J. (2011). The
Mickelson, R. A. (2003). Gender, Bourdieu, and the anomaly of women’s influence of affective teacher–student relationships on students’ school
achievement redux. Sociology of Education, 76, 373–375. doi:10.2307/ engagement and achievement: A meta-analytic approach. Review of
1519873 Educational Research, 81, 493–529. doi:10.3102/0034654311421793
516 MCCORMICK AND O’CONNOR

Royston, P., & White, I. R. (2011). Multiple imputation by chained Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis:
equations (MICE): Implementation in Stata. The Journal of Statistical Modeling change and event occurrence. Oxford, England: Oxford Uni-
Software, 45(4), 1–20. versity Press. doi:10.1093/acprof:oso/9780195152968.001.0001
Rubin, D. B. (1987). A noniterative sampling/importance resampling al- Snyder, T. D., Dillow, S. A., & Hoffman, C. M. (2001). Digest of
ternative to the data augmentation algorithm for creating a few imputa- education statistics, 2000. Washington, DC: U.S. Department of Edu-
tions when fractions of missing information are modest: The SIR algo- cation, National Center for Education Statistics.
rithm. Journal of the American Statistical Association, 82, 543–546. Spilt, J. L., Hughes, J. N., Wu, J. Y., & Kwok, O. M. (2012). Dynamics of
Rudasill, K. M. (2011). Child temperament, teacher– child interactions, and teacher–student relationships: Stability and change across elementary
teacher– child relationships: A longitudinal investigation from first to school and the influence on children’s academic success. Child Devel-
third grade. Early Childhood Research Quarterly, 26, 147–156. doi: opment, 83, 1180 –1195. doi:10.1111/j.1467-8624.2012.01761.x
10.1016/j.ecresq.2010.07.002 Van Buuren, S., Boshuizen, H. C., & Knook, D. L. (1999). Multiple
Rudasill, K. M., Niehaus, K., Buhs, E., & White, J. M. (2013). Tempera- imputation of missing blood pressure covariates in survival analysis.
ment in early childhood and peer interactions in third grade: The role of
Statistics in Medicine, 18, 681– 694. doi:10.1002/(SICI)1097-
teacher– child relationships in early elementary grades. Journal of
0258(19990330)18:6⬍681::AID-SIM71⬎3.0.CO;2-R
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

School Psychology, 51, 701–716. doi:10.1016/j.jsp.2013.08.002


Vandell, D. L., Belsky, J., Burchinal, M., Steinberg, L., & Vandergrift, N.
Rudasill, K. M., Reio, T. G., Jr., Stipanovic, N., & Taylor, J. E. (2010). A
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(2010). Do effects of early child care extend to age 15 years? Results


longitudinal study of student–teacher relationship quality, difficult tem-
from the NICHD study of early child care and youth development. Child
perament, and risky behavior from childhood to early adolescence.
Development, 81, 737–756. doi:10.1111/j.1467-8624.2010.01431.x
Journal of School Psychology, 48, 389 – 412. doi:10.1016/j.jsp.2010.05
.001 Waters, E., & Deane, K. E. (1985). Defining and assessing individual
Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and differences in attachment relationships: Q-methodology and the organi-
quasi-experimental designs for generalized causal inference. Indepen- zation of behavior in infancy and early childhood. Monographs of the
dence, KY: Cengage Learning. Society for Research in Child Development, 50, 41– 65. doi:10.2307/
Silver, R. B., Measelle, J. R., Armstrong, J. M., & Essex, M. J. (2005). 3333826
Trajectories of classroom externalizing behavior: Contributions of child Woodcock, R. W., & Johnson, M. B. (1989). Woodcock–Johnson Psycho-
characteristics, family characteristics, and the teacher– child relationship Educational Battery—Revised. Itasca, IL: Riverside.
during the school transition. Journal of School Psychology, 43, 39 – 60.
doi:10.1016/j.jsp.2004.11.003
Singer, J. D. (1998). Using SAS PROC MIXED to fit multilevel models, Received November 25, 2013
hierarchical models, and individual growth models. Journal of Educa- Revision received June 18, 2014
tional and Behavioral Statistics, 23, 323–355. doi:10.2307/1165280 Accepted June 23, 2014 䡲

Вам также может понравиться