INTELLIGENCE 17, 79-104 (1993) Influences of School Environment on the Academic Achievement Scores of Adopted and Nonadopted Children HILARY COON GREGORY CAREY DAVID W. FULKER J.C. DEFRIES University of Utah and University of Colorado, Boulder Associations between academic achievement and characteristics of the school environ- ment can result from direct environmental influences of the school, or from placement of children into particular school environments based on prior ability. To disentangle these potential influences underlying school effects on children, we analyzed data from parents and first-grade adopted and nonadopted children in the Colorado Adoption Project. Mea- sures analyzed included attending a private versus a public school, a variable aggregated at the school level, and several variables aggregated at the classroom level. If such aggregated measures are associated with many intercorrelated individual aspects of school environment, each having a small effect on achievement, the process of aggregation may offer additional power to detect these small individual environment effects. Several of these aggregate variables, in addition to measures of children's attitudes about school, showed direct environmental associations with reading and math achievement indepen- dent of effects of parental IQ. Researchers have found evidence for associations among various aspects of the school environment and children's achievement scores (e.g., Brookover, Beady, Flood, Schweitzer, & Wisenbaker, 1979; Centra & Potter, 1980; Glasman & Biniaminov, 1981; MacPhail-Wilcox & King, 1986; Purkey & Smith, 1983; Rutter, 1983; Spady, 1976; Stevenson, & Lee, 1990; Stockard & Mayberry, This research was supported in part by grants HD-10333, HD-18426, and HD-19802 from the National Institute of Child Health and Human Development (NICHD) and by grant MH-43899 from the National Institute of Mental Health. The article was written while the first author was under the support of training grant HD-07289 from NICHD. Preparation of the article was facilitated by grant RR-07013-20 awarded to the University of Colorado by the Biomedical Research Grant Program, Division of Research Resources, National Institutes of Health. We would like to thank Robert Plomin and three anonymous reviewers for their suggestions and advice on an earlier version of the manu- script. We would also like to thank Rebecca G. Miles for her expert editorial assistance. Correspondence and requests for reprints should be sent to Hilary Coon, Department of Psychia- try, University of Utah Medical School, 50 North Medical Drive, Salt Lake City, UT 84132. 79
26
Embed
Influences of School Environment on the Academic Achievement Scores of Adopted and Nonadopted Children
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
INTELLIGENCE 17, 79-104 (1993)
Influences of School Environment on the Academic Achievement Scores of Adopted and Nonadopted Children
HILARY COON
GREGORY CAREY
DAVID W . FULKER
J . C . DEFRIES
University of Utah and
University of Colorado, Boulder
Associations between academic achievement and characteristics of the school environ-
ment can result from direct environmental influences of the school, or from placement of
children into particular school environments based on prior ability. To disentangle these
potential influences underlying school effects on children, we analyzed data from parents
and first-grade adopted and nonadopted children in the Colorado Adoption Project. Mea-
sures analyzed included attending a private versus a public school, a variable aggregated
at the school level, and several variables aggregated at the classroom level. If such
aggregated measures are associated with many intercorrelated individual aspects of school
environment, each having a small effect on achievement, the process of aggregation may
offer additional power to detect these small individual environment effects. Several of
these aggregate variables, in addition to measures of children's attitudes about school,
showed direct environmental associations with reading and math achievement indepen-
dent of effects of parental IQ.
Resea rche r s have found ev idence for assoc ia t ions a m o n g var ious aspects o f the
schoo l e n v i r o n m e n t and ch i ld r en ' s a c h i e v e m e n t scores (e .g . , Brookover , Beady,
F lood , Schwei tze r , & W i s e n b a k e r , 1979; Cen t r a & Potter , 1980; G l a s m a n &
B in i am in ov , 1981; M a c P h a i l - W i l c o x & King , 1986; Purkey & Smi th , 1983;
Rutter , 1983; Spady, 1976; S t evenson , & Lee, 1990; S tockard & Mayber ry ,
This research was supported in part by grants HD-10333, HD-18426, and HD-19802 from the
National Institute of Child Health and Human Development (NICHD) and by grant MH-43899 from
the National Institute of Mental Health. The article was written while the first author was under the
support of training grant HD-07289 from NICHD. Preparation of the article was facilitated by grant
RR-07013-20 awarded to the University of Colorado by the Biomedical Research Grant Program,
Division of Research Resources, National Institutes of Health. We would like to thank Robert Plomin
and three anonymous reviewers for their suggestions and advice on an earlier version of the manu-
script. We would also like to thank Rebecca G. Miles for her expert editorial assistance.
Correspondence and requests for reprints should be sent to Hilary Coon, Department of Psychia-
try, University of Utah Medical School, 50 North Medical Drive, Salt Lake City, UT 84132.
79
80 COON, CAREY, FULKER, AND DEFRIES
1985). However, controversy remains over whether these associations reflect
true causal relationships between school environment and achievement. Parents
who send their children to schools with characteristics associated with high
achievement may themselves be more supportive or intelligent (Jencks, 1972;
Smith, 1972). Because parents and their children share genes for intelligence, an
observed association between school environment and child achievement may
occur because the school variable is correlated with parental IQ. Plomin,
Loehlin, and DeFries (1985) referred to this type of indirect association as
"genetic mediation" of the environment.
The question of causality has been particularly important in the debate over
the influence of private schools (Coleman & Hoffer, 1987; Hoffer, Greeley, &
Coleman, 1985). Higher achievement scores of Catholic school students may
result from more emphasis on academic subjects, more homework, or smaller
schools, reasons that would reflect a causal effect of the school environment
(Coleman & Hoffer, 1987). However, children in Catholic school also may come
from a selected population of more intelligent parents. Parents willing to spend
the extra money to place their child in a private school may be more supportive
and have higher academic expectations than parents with children in public
schools. Coleman and Hoffer argued for a direct causal effect between Catholic
school characteristics and achievement, citing results showing that children do
not perform at uniformly higher levels across all subjects. They asserted that
these children would have to be differentially more intelligent, or parents would
have to be more supportive only in certain areas, if Catholic schools had no
causal effects on children.
More direct evidence supporting or refuting these findings could be obtained
by investigating the association between school environment and the achieve-
ment scores of adopted children. In adoptive families, an observed school-
achievement relationship cannot be "genetically mediated" in the absence of
selective placement. Because adopted children and their adoptive parents are
genetically unrelated, even if choice of school is correlated with parental geno-
types for IQ (or other genetic factors that might influence achievement), the child
will not share these genotypes.
Although data from adoptive families provide the most powerful test of direct
environmental effects, adoption studies have focused primarily on the home
shows the associations among the variables in nonadoptive families.
SCHOOL ENVIRONMENT AND ACHIEVEMENT 87
11
Figure 2. Path models showing genetic transmission using a traditional parameterization (left) com-
pared to the simple parameterization used in this analysis that does not assume isomorphism between
IQ in the mother and achievement in the child (right). In the figure on the left, the IQ genotype of the
mother (GMo) influences the mother's observed IQ (PMo) by the amount hlo (square root of
heritability of IQ). Similarly, the achievement genotype of the child (Go) influences the child's
observed achievement score (Po) by the amount hgc h (square root of heritability of achievement).
The mother and child genotypes are correlated V2. In the figure on the right, these three parameters
(l/2hlQhAch) have been collapsed into the single parameter gMo.
The genetic contribution to the correlation between sibling pairs (gsib) will
also be zero in adoptive families. For nonadoptive sibling pairs, this correlation
will be the product of the genetic correlation between full siblings (V2) and the
heritability for achievement. Preliminary tests indicated that family resemblance
for parent-offspring and sibling pairs could be modeled using one parameter
(gFamily = gMo,m = gFa,m = gMo,f = gFa,f = gSib:m,m = gSib:m,f = gSib:f,f) ' This result indicates that in these data, heritability of adult IQ and heritability of child
achievement are not significantly different.
The school environment measure influences adopted children through the
direct environmental path w. In nonadoptive families, there is an additional
genetic component (gSch) because the school environment may be correlated
with the child's genotype for achievement. Such a correlation would occur if the
school measure were correlated with the parental genotypes contributing to
achievement, which the parents then transmit to their children (genetic mediation
of the school environment). Thus, the genetic association between the school
measure and child phenotype is the product of this parental genotype-school
correlation, the genetic transmission from parent to child for achievement, and
the heritability for achievement. Results from preliminary analyses indicated that
this parameter could be pooled across gender (gsch = gsch,m = gsch,f)"
The model also allows maternal and paternal IQ phenotypes to be correlated
Ix, and for each to be correlated with the school m e a s u r e (rMo,Sch , rFa,Sch ).
88 COON, CAREY, FULKER, AND DEFRIES
(i . . . . .
/,t
MO Environment Fa
g~
POl I I Po2
z f z
m
S m,f
Figure 3. Path model for the analysis of genetic and environmental transmission among nonadaptive
parents, children, and a measure of the school environment. Arbitrarily, Offspring l is female
(subscript f) and Offspring 2 is male (subscript m). Subscripts Mo, Fa, and Sch represent mother,
father, and school, respectively. The children's phenotypes for achievement (Pol, POE) are influ-
enced by the school environment independent of the direct effects of parental IQ phenotypes (PMo,
PFa)- Genetic associations among variables are now included (see text for further explanation of
variables and parameters).
Residuals (Uot, U o 2 ) may also influence the children's achievement phenotypes
through the z paths; these residuals are allowed to correlate between siblings.
Possible confounding effects using this model include selective placement and
child genotype-environment (GE) correlation. If an association occurs between
a school measure and the IQ of the child's birth parents, selective (nonrandom)
SCHOOL ENVIRONMENT AND ACHIEVEMENT 89
placement of the child has occurred. Correlations between the school environ-
ment of the adopted child and the IQ scores of the 216 matching biological
mothers were low, ranging from - . l l to . 10, indicating negligible selective
placement. In addition, there can be a correlation between a child's genotype and
environment such that a child with a certain genotype may expose himself more
to certain environments (Plomin, DeFries, & Loehlin, 1977). In this analysis,
this effect will only be important to the extent that the GE correlation involves
genetic influences not shared between parent and child (shared genetic effects
that are correlated with the environmental measures will be subsumed in the
genetic mediation pathway). Although such effects are not directly testable using
our model, indirect evidence can be obtained by using child cognitive ability at
age 4 as a predictor of the school environment measures. Such regressions
produced nonsignificant results, with multiple R scores ranging from .01 to .07
for the school questionnaire measures, and from .02 to .12 for the attitude
measures. This indicates minimal effects of the child's prior ability which are
independent of the genetic mediation effects through parental IQ.
Expectations for this model appear in Appendix B; note that the model also
provides estimates of the means and variances of the measures of school environ-
ment, of adult cognitive ability, and of child achievement. This model gives
estimates that are functions of the parameters of interest, rather than the param-
eters themselves. Likelihood-ratio chi-square statistics were used to assess fit.
For a more detailed description of model fitting with adoption data, see Coon,
Carey, and Fulker (1990) and Coon, Fulker, DeFiles, and Plomin (1990).
RESULTS
Parent-Offspring Transmission The influence of school environment on child achievement may be mediated by
influences of parental IQ. our model separates these two potential effects on the
child. Table 3 (p. 90) presents correlations of parent general intelligence and
child achievement scores. Approximate estimates of genetic and 'environmental
influences of parental IQ can be obtained by comparing the correlations for
adoptive families, for whom the association will be entirely environmental, with
those for nonadoptive families, for whom the association has both genetic and
environmental components.
Using this preliminary estimation method, a negative association between
mothers' IQ scores and boys' achievement appears in adoptive families, but not
in nonadoptive families. These contrary findings may suggest a negative en-
vironmental influence mitigated by a positive genetic effect. The positive cor-
relations between mothers and girls in both adoptive and nonadoptive families
provides evidence for a weak positive environmental effect. Fathers' IQ scores
and daughters' reading achievement scores show a strong positive association in
both adoptive and nonadoptive families, again suggesting a positive environmen-
90 COON, CAREY, FULKER, AND DEFRIES
TABLE 3
Correlations Between Parental General Intelligence and Child Achievement
Adoptive Families Nonadoptive Families
Child Measure n Mothers Fathers n Mothers Fathers
Math Achievement
Boys 117 - . 18 a* .05 149 .11 .18"
Girls 108 .17 a .03 123 .17" .16
All 225 - .05 .04 272 .13" .17"*
Reading
Achievement
Boys 114 - . 19 a* - . 00 147 .09 .25**
Gids 105 .16 a .23* 120 .10 .28**
All 219 - .05 .10 267 .09 .26**
aSignificant gender difference.
*p -< .05. **p -< .001.
tal effect. Math achievement in girls and both reading and math achievement in
boys appears to have only a genetic association with paternal IQ.
Gender differences were found in these correlations, and in several of the
correlations with school measures. These gender differences were unexpected
and may be due to chance. However, it is also possible that the gender of the
child may moderate parental and environmental influences on achievement at age
7. Considering both possibilities, we present all subsequent results separately by
gender and pooled across gender.
More rigorous tests of these associations can be obtained by fitting the model
shown in Figure 1 to CAP data. The preliminary results suggested by the correla-
tions are reflected in the maximum likelihood estimates of parameters associated
with parents and offspring. Table 4 presents these estimates from the model fit
with the private-public school factor; the estimates for models testing the other
school factors varied only in the second decimal place from those presented here.
The stability of these results across all school environmental measures indicates
that these parent-offspring parameter estimates are not dependent on which
environmental variable is tested. This stability is to be expected across the
univariate model results for each school environment measure; even if a school
variable is associated with parental IQ or child achievement, this association
should not affect the correlation between spouses or between parent and child.
The pooled estimates were taken from a model where all parameters were equat-
ed across gender; all other estimates were taken from the baseline models used in
the likelihood-ratio chi-square significance tests of parameters.
Model fitting confirmed the negative environmental effect of mothers on
boys' achievement (tin) and the positive influence of fathers on girls' reading
achievement (uf). For reading achievement, the residual correlation between
same-sexed sibling pairs and opposite-sexed pairs were significantly different,
SCHOOL ENVIRONMENT AND ACHIEVEMENT
TABLE 4
Maximum Likelihood Estimates for Parent-Offspring Parameters
91
Parameter Reading Math
Parent-Offspring Parameters
ix (marital correlation) .29 a .28 ~
gFamily (shared genetic effects) .13 a . l 0 a
Environmental transmission:
t m (mother-son) - . 15 a - . 12 t'c
tf (mother-daughter) .04 .13 b¢
tpooled (mother-child) - . 0 9 - . 03
Um (father-son) .10 .08
uf (father-daughter) .19a .04
upootea (father-child) .12 .05
Residuals
z (effect on child phenotype) .80 a .58 a
Ssame (residual r: same-sexed pairs) .62 ad .57 a
Sm. f (residual r: opposite-sexed pairs) - . 1 0 d .44 a
Spooled (residual r: all pairs) .47 a .53 a
Means and Standard Deviations
Xpa r .01 .01
XBoy(adopted )e . ! 5 a --.05
Xgirl(adopted) --. 11 a --. 15 a
Xpooled(adopted ) - - . 12 a - . 10 a
XBoy(nonadopted) . 1 2 a .26 a
XGirl(nonadopted) .10 --.07
Xpooled(nonadopted) .1 I a .10 a
O'pa r .99 a .99 a
OrChil d 1.07 a .99 a
Note. Parameters conform to those depicted in Figure 1, with the following simplifications:
z m = Zf = Z, Sra ,m = Ssame. Pooled estimates were obtained from models equating across gender.
aparameter estimate significantly different from zero.
bSignificant gender difference.
cParameters significantly different from zero only in test with two degrees of freedom.
dSignificant difference between residual correlations for same-sexed versus'opposite-sexed sib-
ling pairs. eAlthough means for boys and girls can be pooled within family types, means for adopted
children were significantly lower than means for nonadopted children.
r e f l e c t i n g d i f f e r e n c e s b e t w e e n a d o p t e d a n d n o n a d o p t e d c h i l d r e n a n d a t r e n d in
t h e d a t a f o r s a m e - s e x e d p a i r s to c o r r e l a t e m o r e s t r o n g l y t h a n o p p o s i t e - s e x e d
pa i r s . ~ T a b l e 4 a l s o p r e s e n t s e s t i m a t e d m e a n s a n d s t a n d a r d d e v i a t i o n s fo r p a r e n t s
a n d c h i l d r e n . F o r b o t h r e a d i n g a n d m a t h a c h i e v e m e n t , n o n a d o p t e d c h i l d r e n
qn nonadoptive families, the gender of the child is determined largely by chance, resulting in an
equal ratio of same-sexed to opposite-sexed sibling pairs. However, adoptive parents more often
choose opposite-sexed adopted sibling pairs. Because of this preference, in the CAP there were too
few same-sexed adoptive pairs (4 boy-boy and 9 girl-girl pairs) to make meaningful comparisons.
92 COON, CAREY, FULKER, AND DEFRIES
performed better than adopted children; however, differences were less than one
quarter of a standard deviation. Moreover, the standard deviations could be
pooled across groups. Because the analyses focus on correlation structures,
differences in variance are more important than mean differences; thus, the mean
differences should not influence the pattern of results.
School Factors
Table 5 presents correlations of school factor scores with parental IQ and child
achievement. Parental IQ and education appear to be largely unrelated to the
school factors. A similar pattern of low correlations appears for the children,
with a few notable exceptions. Because the overall correlation matrix does not
meet a multivariate test of significance (Steiger, 1980), interpretation of the
following univariate effects should be made cautiously. Strong discipline is asso-
ciated with lower reading achievement for both adopted and nonadopted girls,
indicating a negative environmental effect. For boys' math achievement, the
correlations suggest positive effects of private schools due to a combination of
TABLE 5
Correlations of School Questionnaire Factor Scores With Parental Measures and Child
Achievement Scores
School Factors
n Private/Public Traditional Discipline Order/Fair Competitive
Parental Measures
Mothers' IQ 363 .11 *
Mothers' educ. 369 .05
Fathers' IQ 359 .08
Fathers' educ. 370 .08
Reading Achievement
Adopted boys 107 .02
Adopted girls 97 .18
All adopted 204 .09
Nonadopted boys 126 .12
Nonadopted girls 111 .12
All nonadopted 237 .12
Math Achievement
Adopted boys 109 .15
Adopted girls 98 .05
All adopted 207 .06
Nonadopted boys 127 .21 *
Nonadopted girls 114 .06
All nonadopted 241 .14"
.03 .01 - . 0 3 - . 1 3 "
- . 0 5 .04 .05 - . 0 9
- . 0 4 .03 - . 0 3 - . 0 2
- . 0 8 - . 0 0 .03 - . 0 6
- . 0 4 .06 a .04 .02
- . 0 7 - . 2 4 a* .01 - . 0 3
- . 0 5 - . 0 6 .03 - . 01
- . 0 6 .02 a .11 - . 0 5
- . 0 4 - . 2 4 a* .02 - . 0 9
- . 0 5 - . 1 0 .07 - . 0 7
- . 0 9 .00 .14 .00
- . 0 4 --.16 - . 0 6 .03
- . 0 6 - . 0 6 .05 .02
- . 1 5 a - . 0 3 - . 0 8 .22 a*
.17 ~ - . 0 5 .02 - . 0 8 a
.02 - . 0 3 - . 0 4 .11
aSignificant gender difference.
*p -< .05.
SCHOOL ENVIRONMENT AND ACHIEVEMENT 93
genetic and environmental influences, and a genetically mediated effect of com-
petitive environment.
More rigorous model-fitting tests once again revealed much the same results
(see Table 6, pp. 94-95). A significant negative environmental effect of disci-
pline appears for girls' reading achievement (We), and the positive association of
boys' math achievement scores with competitive school environments is genet-
ically mediated (gSch,m)" Positive genetic mediation through the private school
factor appears for boys' reading achievement. The relationship between private
school and boys' math achievement remains inconclusive. Although the genetic
and environmental parameters (gsch,m and Wm) can be dropped individually from
the model, both cannot be dropped together, !2(2) = 9.42, p < .01. Thus,
private schools do have a significant effect on children, but the effect may be due
to a direct environmental association or to genetic mediation.
Estimated correlations between parental IQ and the school factors (rFa.S~h,
rMo,S~h) and school factor means and standard deviations are included for both
reading and math achievement to demonstrate the stability of the model. Al-
though a significant mean difference occurs for the private-public factor, stan-
dard deviations could be pooled across family type.
Results of overall tests of gender differences indicate that gender may be
important for all but the Order/Fair and Discipline factors when math achieve-
ment was the outcome variable. Because this test equates all parameters in the
model, the chi-square values that indicate a worsening in fit are also due to
gender differences in parent-offspring parameters (see Table 4).
School Attitudes
Because of small sample size and because no significant gender differences were
found in the correlations, analyses of the school attitudes measures were carded
out for children pooled across gender groups. As with the school factor scores,
associations between parental measures and ratings from the school events ques-
tionnaire are weak (see Table 7, p. 96). However, several substantial correlations
appear between child achievement and the school events ratings.dndeed, a test of
the overall matrix showed significant associations with achievement at the multi-