Documente Academic
Documente Profesional
Documente Cultură
a r t i c l e
i n f o
Article history:
Received 5 September 2014
Received in revised form 23 July 2015
Accepted 5 August 2015
Available online 18 August 2015
Keywords:
Internalizing problems
Externalizing problems
Teacherchild relationship quality
a b s t r a c t
This study examined three models testing directional associations between childrens adjustment
problems (internalizing and externalizing problems) and teacherchild relationship quality (closeness,
conict, and dependency) over one school term. The relationship-driven model tested the hypothesis
that teacherchild relationship quality contributes to prospective levels of adjustment problems. The
child-driven model tested the hypothesis that childrens adjustment problems contribute to prospective relationship quality. The transactional model tested the hypothesis that adjustment problems and
relationship quality are reciprocally related over time. These models were tested with a sample of lowincome, ethnically diverse children in kindergarten to grade 3 who were assessed at three time points
over one school term (winter, early spring, and late spring). Overall, the child-driven model best explained
the directional associations between adjustment problems and conictual and dependent teacherchild
relationships. When children showed higher levels of externalizing problems, they subsequently experienced more conict in their relationships with their teachers. When children showed more internalizing
problems they experienced more prospective dependency in their relationships with their teachers.
Adjustment problems and teacherchild closeness were modestly related within time but not across
time.
2015 Elsevier Inc. All rights reserved.
Introduction
Teachers are a key source of support for young children in the
school system (Hamre & Pianta, 2001; Pianta, 1999). Many children
share a high-quality, comforting relationship with their teacher
that can enhance their experiences within the school setting. Close
relationships with teachers are characterized by warmth and open
communication where children can use their teacher as a securebase to explore the classroom environment (Birch & Ladd, 1997).
However, some children develop negative relationships with their
teachers that can undermine their experiences in school (Pianta,
1999; Pianta & Stuhlman, 2004). Negative teacherchild relationships can be characterized by conict where there is tension and
anger in the relationship and can also be reected by dependency where children demonstrate a clinginess and overreliance on
the teacher (Birch & Ladd, 1997). Unfortunately, children who are
This research was supported by a grant from the Social Sciences and Humanities
Research Council of Canada (SSHRC) awarded to W.L.G. Hoglund.
Corresponding author at: Department of Psychology, P217 Biological Sciences
Building, University of Alberta, Edmonton, AB, Canada T6G 2E9.
E-mail address: tmejia@ualberta.ca (T.M. Mejia).
http://dx.doi.org/10.1016/j.ecresq.2015.08.003
0885-2006/ 2015 Elsevier Inc. All rights reserved.
14
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
Fig. 1. Hypothesized associations between adjustment problems and teacherchild relationship quality. The faint lines indicate the autoregressive paths and within-time
correlations from the stability models that are also included in the primary models.
internalizing and externalizing problems relate to distinct dimensions of teacherchild relationship quality (closeness, conict, and
dependency) over time, such as over the course of one school term.
The current study tests three alternative theoretical models of
the directional associations between childrens adjustment problems and their relationship quality with teachers over the course
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
15
16
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
17
18
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
applies). Internal consistencies were high across waves for closeness (s = .81.85) and conict (s = .89.91) and moderate for
dependency (s = .68.73).
Baseline covariates
Each cross-lagged path model adjusted for a set of covariates
that included: child age, child gender (0 = boys, 1 = girls), class size
(M =17.80, SD = 4.52), teachers education (0 = Bachelors degree,
1 = 2-year after-degree), and teachers years of teaching experience
(M = 11.78 years, SD = 9.53).
Data analytic strategy
Analyses are presented in three main sections. First, descriptive
statistics of the criterion constructs are examined for the overall
sample and by grade. Bivariate correlations among the constructs
are also presented. Second, a series of auto-regressive, cross-lagged
path analysis models assessing the concurrent and prospective
associations between adjustment problems and teacherchild relationship quality are examined using Mplus 7.2 (Muthn & Muthn,
2011). Third, multiple group models are used to examine agerelated differences in the best tting models (kindergarten and
grade 1 vs. grades 2 and 3). All models are clustered by classroom
to account for the nesting of children within classrooms. In each
model tested, each construct is regressed on child age, gender,
class size, and teachers education and years of teaching experience at each wave.1 Full information maximum likelihood (FIML)
estimation with robust standard errors is used to estimate missing
data. FIML estimation uses data available from each case to produce
unbiased parameter estimates and standard errors. The likelihood
estimate is computed separately for cases with incomplete data and
for cases with complete data, integrating estimates over all possible values to produce parameter estimates that are most likely to
have resulted in the observed data (Allison, 2002).
Model t is assessed using the chi-square statistic (2 ), comparative t index (CFI), root-mean-square error of approximation
(RMSEA), and standardized root-mean-square residual (SRMR:
Kline, 2011). The 2 test evaluates how the hypothesized models differ from the null model but is sensitive to sample size. CFI
values of .95 or greater signify excellent model t and values of
.90.94 signify adequate t. The RMSEA index takes into account
model complexity and parsimony. RMSEA and SRMR values of
.05 or lower signify excellent model t, while values of .06.08
indicate adequate model t (Kline, 2011). Model comparisons are
calculated via the SatorraBentler scaled chi-square (2 ) likelihood
ratio difference test to assess the t statistics of nested models
(e.g., relationship-driven vs. transactional). The Bayesian Information Criteria (BIC) is used to compare non-nested models (e.g.,
relationship-driven vs. child-driven).
Results
Descriptive statistics
Descriptive statistics of the criterion constructs are presented
in Table 1. On average, children showed low levels of internalizing and externalizing problems at each wave. Across waves,
teachers reported high average levels of teacherchild closeness and low average levels of conict and dependency. There
1
As closeness and conict shared a strong positive association at each time point,
we tested additional models where conict was controlled for in the closeness models and closeness was controlled for in the conict models. Findings from these
models were similar to those without these additional controls. For simplicity we
removed these additional control variables.
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
Table 1
Descriptive statistics for child adjustment problems and teacherchild relationship
quality.
Variables
Internalizing problems
.82.85
Wave 1
Wave 2
.83.85
.84.87
Wave 3
Externalizing problems
.91.93
Wave 1
.88.94
Wave 2
Wave 3
.96.94
Teacherchild closeness
.81
Wave 1
Wave 2
.82
.85
Wave 3
Teacherchild conict
Wave 1
.89
.89
Wave 2
.91
Wave 3
Teacherchild dependency
Wave 1
.68
.71
Wave 2
.73
Wave 3
Mean
SD
300
243
203
0.35
0.34
0.35
.36
.37
.36
0.002.18
0.002.29
0.002.00
300
243
203
0.68
0.68
0.64
.56
.54
.56
0.002.44
0.002.41
0.002.59
298
237
201
3.12
3.02
3.21
.71
.71
.66
1.134.00
1.134.00
0.884.00
297
238
201
0.63
0.64
0.61
.91
.85
.89
0.003.86
0.003.71
0.004.00
297
238
200
0.77
0.81
0.66
.74
.73
.71
0.003.40
0.003.20
0.003.20
Range
Note. Internal consistencies for adjustment problems indicate range across subscales.
were modest grade level differences in mean levels of the constructs. On average, children in kindergarten (Ms = 3.263.27,
SDs = 0.720.75) had closer relationships with teachers compared
to children in grade 2 at wave 2 (M = 3.10, SD = 0.69; F[3] = 6.43,
p < .01) and children in grade 3 at wave 2 and 3 (Ms = 2.802.95,
SDs = 0.520.59; F[3] = 3.366.43, p < .05). At wave 2, children in
kindergarten (M = 0.90, SD = 1.08) also experienced more conict in
the teacherchild relationship than children in grade 1 (M = 0.49,
SD = 0.75; F[3] = 2.77, p < .05).
The bivariate correlations between the criterion constructs
at each wave are presented in Table 2. Rank order stability
was high across waves for internalizing and externalizing problems and teacherchild conict and moderate across waves for
teacherchild closeness and dependency. Internalizing and externalizing problems were moderately correlated within and across
waves. Teacherchild conict was weakly and negatively correlated with teacherchild closeness and moderately and positively
correlated with teacherchild dependency within and across
waves. Teacherchild closeness was unrelated to dependency
across waves. Teacherchild closeness was weakly and negatively correlated with externalizing problems within and across
waves but was only correlated with internalizing problems at one
wave. Teacherchild conict and dependency were moderately
and positively correlated with both internalizing and externalizing
problems within and across waves.
19
2
Models with the paths from winter relationship quality to adjustment problems
in late spring were also tested. These additional direct paths were not signicant,
while the initial paths tested were still signicant. Thus, we removed these paths
from winter (wave 1) to late spring (wave 3).
20
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
Table 2
Bivariate correlations between child adjustment problems and teacherchild relationship quality.
Variables
Internalizing problems
1. Wave 1
2. Wave 2
.76**
3. Wave 3
.77**
Externalizing problems
.47**
4. Wave 1
5. Wave 2
.45**
6. Wave 3
.48**
Teacherchild closeness
7. Wave 1
.09
8. Wave 2
.04
9. Wave 3
.06
Teacherchild conict
.48**
10. Wave 1
11. Wave 2
.39**
12. Wave 3
.46**
Teacherchild dependency
.49**
13. Wave 1
14. Wave 2
.54**
15. Wave 3
.54**
10
11
12
13
14
.42**
.40**
.55**
.64**
.61**
.65**
.71**
.40**
.49**
.44**
.10
.10
.22**
.50**
.46**
.54**
.09
.09
.07
.85**
.87**
.13*
.15*
.17*
.90**
.12
.15*
.24*
.24**
.25**
.25**
.59**
.69**
.62**
.45**
.52**
.38**
.49**
.47**
.56**
.73**
.66**
.63**
.61**
.75**
.69**
.66**
.69**
.74**
.29**
.18**
.17*
.20**
.21**
.18*
.27**
.25**
.22**
.71**
.72**
.78**
.44**
.60**
.49**
.47**
.50**
.64**
.49**
.42**
.46**
.42**
.51**
.45**
.40**
.41**
.48**
.04
.00
.03
.07
.01
.01
.04
.05
.04
.55**
.39**
.43**
.44**
.57**
.41**
Baseline covariates
As found in the best tting models described above, younger
children experienced closer relationships with their teachers at
W2 ( = .14, SE = .06, p < .05) and at W3 ( = .10, SE = .05, p < .05)
and experienced higher levels of internalizing problems at W2
( = .10, SE = .05, p < .05) than older children. Older children had
more externalizing problems at W3 than younger children ( = .05,
SE = .02, p < .05). Boys had higher levels of externalizing problems at W1 than girls ( = .21, SE = .06, p < .01). Girls had more
Multiple-group models
Last, a series of multiple-group models testing grade differences
in the best-tting models were examined. First, the within-time,
autoregressive, and cross-time paths were constrained to be equal
Table 3
Cross-lagged path models assessing the directional associations between child adjustment and teacherchild relationship quality.
Model
2 (df)
Teacherchild closeness
1. Stability
2. Relationship-Driven
3. Child-Driven
4. Transactional
24.59 (18), ns
22.87 (14), p < .10
22.04 (14), p < .10
20.38 (10), p < .05
Teacherchild conict
1. Stability
2. Relationship-Driven
3. Child-Driven
4. Transactional
Teacherchild dependency
1. Stability
2. Relationship-Driven
3. Child-Driven
4. Transactional
SRMR
BIC
.994
.992
.992
.990
.032 (.00.06)
.043 (.00.07)
.040 (.00.07)
.055 (.02.09)
.029
.026
.026
.024
2156.21
2176.76
2178.00
2198.70
.971
.980
.998
1.00
.080 (.06.10)
.075 (.05.10)
.025 (.00.06)
.000 (.00.05)
.055
.031
.028
.009
2214.30
2215.06
2184.69
2196.64
.969
.970
.999
.997
.079 (.06.10)
.085 (.06.11)
.018 (.00.06)
.030 (.00.07)
.068
.054
.027
.019
2221.62
2237.31
2193.25
2213.84
CFI
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
21
Fig. 2. Stability model for child adjustment problems and teacherchild closeness. Unstandardized (standard error)/standardized estimates presented. Dashed lines indicate
non-signicant paths. Model Fit: 2 (18) = 24.59, ns; CFI = .994; RMSEA = .032 (.000, .060); SRMR = .029. t p < .10, *p < .05, **p < .01.
across grades (kindergarten and grade 1 vs. grades 2 and 3). The
t of these models was then compared to the t of models where
the estimates for the within-time, autoregressive, and cross-time
paths were sequentially allowed to vary between kindergarten
and grade 1 children versus grades 2 and 3 children. Multiplegroup models indicated signicant grade-level differences in
Fig. 3. Child-driven model for child adjustment problems and teacherchild conict. Unstandardized (standard error)/standardized estimates presented. Dashed lines indicate
non-signicant paths. Model Fit: 2 (14) = 17.11, ns; CFI = .998; RMSEA = .025 (.000, .060); SRMR = .028. t p < .10, *p < .05, **p < .01.
22
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
Fig. 4. Child-driven model for child adjustment problems and teacherchild dependency. Unstandardized (standard error)/standardized estimates presented. Dashed lines
indicate non-signicant paths. Model Fit: 2 (14) = 15.58, ns; CFI = .999; RMSEA = .018 (.000, .060); SRMR = .027. t p < .10, *p < .05, **p < .01.
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
23
24
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326
References
Allison, P. D. (2002). Missing data: Quantitative application in the social sciences.
Thousand Oaks, CA: Sage Publications.
Baker, J. (2006). Contributions of teacherchild relationships to positive school
adjustment during elementary school. Journal of School Psychology, 44,
211229. http://dx.doi.org/10.1016/j.jsp.2006.02.002
Baker, J., Grant, S., & Morlock, L. (2008). The teacherstudent relationship as a
developmental context for children with internalizing or externalizing
behavior problems. School Psychology Quarterly, 23, 315. http://dx.doi.org/10.
1037/1045-3830.23.1.3
Bell, R. Q. (1968). A reinterpretation of the direction of effects in studies of
socialization. Psychological Review, 75, 8195. http://dx.doi.org/10.1037/
h0025583
Bierman, K. L. (2011). The promise and potential of studying the invisible hand of
teacher inuence on peer relations and student outcomes: A commentary.
Journal of Applied Developmental Psychology, 32, 297303. http://dx.doi.org/10.
1016/j.appdev.2011.04.004
Birch, S., & Ladd, G. (1997). The teacherchild relationship and childrens early
school adjustment. Journal of School Psychology, 35(1), 6179. http://dx.doi.org/
10.1016/S0022-4405(96)00029-5
Birch, S., & Ladd, G. (1998). Childrens interpersonal behaviors and the
teacherchild relationship. Developmental Psychology, 34, 934946. http://dx.
doi.org/10.1037/0012-1649.34.5.934
Chang, L., Liu, H., Fung, K. Y., Wang, Y., Wen, Z., Li, H., et al. (2007). The mediating
and moderating effects of teacher preference on the relations between
students social behaviors and peer acceptance. Merrill-Palmer Quarterly, 53,
603630. http://dx.doi.org/10.1353/mpq.2008.0006
Coplan, R., & Prakash, K. (2003). Spending time with teacher: Characteristics of
preschoolers who frequently elicit versus initiate interactions with teachers.
Early Childhood Research Quarterly, 18, 143158. http://dx.doi.org/10.1016/
S0885-2006(03)00009-7
Denham, S. A., Bassett, H. H., & Zinsser, K. (2012). Early childhood teachers as
socializers of young childrens emotional competence. Journal of Early
Childhood Education, 40, 137143. http://dx.doi.org/10.1007/s10643-0120504-2
Doumen, S., Vershueren, K., Buyse, E., Germeijs, V., Luyckx, K., & Soenens, B. (2008).
Reciprocal relations between teacherchild conict and aggressive behavior in
kindergarten: A three-wave longitudinal study. Journal of Clinical and Child and
25
26
T.M. Mejia, W.L.G. Hoglund / Early Childhood Research Quarterly 34 (2016) 1326