Adoption and Cognitive Development: A Meta-Analytic Comparison of
Adopted and Nonadopted Children’s IQ and School Performance
Marinus H. van IJzendoorn, Femmie Juffer, and Caroline W. Klein Poelhuis
This meta-analysis of 62 studies (N ? 17,767 adopted children) examined whether the cognitive
development of adopted children differed from that of (a) children who remained in institutional care or
in the birth family and (b) their current (environmental) nonadopted siblings or peers. Adopted children
scored higher on IQ tests than their nonadopted siblings or peers who stayed behind, and their school
performance was better. Adopted children did not differ from their nonadopted environmental peers or
siblings in IQ, but their school performance and language abilities lagged behind, and more adopted
children developed learning problems. Taken together, the meta-analyses document the positive impact
of adoption on the children’s cognitive development and their remarkably normal cognitive competence
but delayed school performance.
Keywords: meta-analysis, adopted children, academic adjustment, learning problems, IQ
Is the cognitive development of adopted children more ad-
vanced compared with that of children who remain in institutional
care or in the birth family? Do adopted children show a less
advanced cognitive development compared with their current,
environmental nonadopted siblings or peers? Several studies indi-
cate that adopted children show externalizing and internalizing
behavior problems to a larger extent than do their environmental
peers (e.g., Brodzinsky, Schechter, Braff, & Singer, 1984). These
socioemotional problems, however, appear to be present in a
relatively small subset of adopted children, whereas the large
majority of them function well (Bimmel, Juffer, Van IJzendoorn,
& Bakermans-Kranenburg, 2003). The central issue for this meta-
analytic review is whether the problems of adopted children are
unique to the socioemotional domain or can also be found in the
domain of cognitive development.
The Theory of Risk and Protective Mechanisms Applied
The theory of risk and protective mechanisms (Rutter, 1990)
assumes that an accumulation of risk factors results in less optimal
development in children, although protective factors may buffer
the negative effects of the risks. The experience of being adopted
involves both protective and risk factors. Adoption means a pos-
itive change, or protective mechanism, for children who are
adopted. They move from a deprived institutional setting or from
an overburdened biological family to an adoptive family. A tra-
jectory that was bound to show cumulative risk factors (because of
maltreatment, neglect, or understimulation) is changed to a posi-
tive direction, one with a greater likelihood of healthy adjustment
However, several risk factors related to adoption or the preadop-
tive background of the child also have been identified to explain
the higher incidence of problems during the life course of adoptees
(Brand & Brinich, 1999; Brodzinsky, 1990; Haugaard, 1998).
Risks may be found in adoptees’ genetic, prenatal, or preadoptive
background. For example, there may have been a psychiatric
disorder in the birth family (Bohman & von Knorring, 1979;
Cadoret, 1990), birth complications, or deprivation in the
adoptee’s home, including malnutrition, neglect, or abuse (Dennis,
1973; McGuinness & Pallansch, 2000; O’Connor et al., 2000;
Rutter et al., 1998). Boys might be more impacted by such adver-
sities than girls (Rutter & Sroufe, 2000).
From a stress and coping perspective, Brodzinsky (1987, 1990)
proposed a psychosocial model of adoption adjustment. He sug-
gested that the experience of adoption exposes parents and chil-
dren to a unique set of psychosocial problems or tasks that may
interact with or complicate normative developmental family tasks.
For example, the child must cope with the loss of the birth parents.
Risks to Cognitive Development
Predictions about a more problematic socioemotional develop-
ment of adopted children have been derived from these models of
risk and protective mechanisms. However, it is less clear whether
adopted children’s cognitive development is at risk in similar
ways. Here, the question is whether adopted children show similar
Marinus H. van IJzendoorn, Femmie Juffer, and Caroline W. Klein
Poelhuis, Centre for Child and Family Studies, Leiden University, Leiden,
The work on this series of meta-analyses was supported by grants from
Stichting VSBfonds, Stichting Fonds 1818, Nationaal Fonds Geestelijke
Volksgezondheid, and Stichting Kinderpostzegels to Femmie Juffer and
Marinus H. van IJzendoorn in cooperation with the Adoptie Driehoek
Onderzoeks Centrum (Adoption Triad Research Center).
We gratefully acknowledge the contributions of Marian J. Bakermans-
Kranenburg to the meta-analyses and the trim and fill computations of Sue
Duval and Elise Dusseldorp.
Correspondence concerning this article should be addressed to Marinus
H. van IJzendoorn, Centre for Child and Family Studies, Leiden Univer-
sity, P.O. Box 9555, Leiden 2300 RB, the Netherlands. E-mail:
2005, Vol. 131, No. 2, 301–316
Copyright 2005 by the American Psychological Association
rates of developmental problems in the cognitive domain as in the
In the current series of meta-analyses, we examine studies that
compared adopted children with their peers and siblings on mea-
sures of cognitive functioning, as defined by IQ and school per-
formance. As described above, many adopted children have expe-
rienced some hardships during their first years of life, including
malnourishment, neglect, and/or maltreatment in their family of
origin or in the institutional setting from which they were adopted
(Johnson, 2000; Verhulst, Althaus, & Versluis-den Bieman, 1992).
Therefore, we might expect their cognitive development to be at
risk (Colombo, de la Parra, & Lopez, 1992; Lien, Meyer, &
Winick, 1977; O’Connor et al., 2000; Rutter et al., 1998; Rutter,
Kreppner, & O’Connor, 2001). Studies have shown that orphan-
ages offer fewer opportunities for young children to acquire or
practice new skills than do home environments, which results in
deficits in social and cognitive functioning (e.g., Kaler & Freeman,
1993). Children are often confined to their crib and lack opportu-
nities to play with toys, interact with adults, or practice locomotion
(Dennis, 1973; Johnson, 2000). Several studies have demonstrated
that poor nutrition or malnourishment during infancy is associated
with cognitive problems or delays (Colombo et al., 1992; Winick,
Meyer, & Harris, 1975), probably resulting from delayed brain
Adoption and Cognitive Development
The association between adoption and cognitive development
has been investigated in behavior genetics studies (Bouchard &
McGue, 1981; Dumaret, 1985; Horn, Loehlin, & Willerman, 1979;
Neiss & Rowe, 2000; Plomin, Fulker, Corley, & DeFries, 1997).
The absence of a genetic relationship between adopted children
and their adoptive parents provides a unique opportunity to study
the influence of family environment without the contamination of
genetic factors. Adoptive families are mostly well above average
socioeconomically (e.g., Maughan, Collishaw, & Pickles, 1998;
Verhulst, Althaus, & Versluis-den Bieman, 1990), and the socio-
economic status of the biological parents is often rather low (e.g.,
Colombo et al., 1992; Schiff et al., 1978). Therefore, adoption
might be conceptualized as a natural quasi-experiment in which
one can look for positive changes in IQ caused by positive changes
in the environment. Of course, this quasi-experiment is certainly
not designed with the inferential power derived from random
assignment to conditions. Adopted children may have been
adopted for reasons or causes related to their preadoption cognitive
status. However, some studies were able to compare the IQ of
adopted children with the IQ of their biological siblings who
stayed behind and with the IQ of the birth parents. Therefore, it
may be possible to estimate the influence of genetic and environ-
mental factors more precisely than can be accomplished with
common between-families studies. More often, however, adopted
children have been compared with their genetically unrelated sib-
lings in the adoptive family.
Neiss and Rowe (2000), for example, compared adopted chil-
dren and matched birth children—unrelated to the adopted chil-
dren—to estimate the genetic and environmental effect of the
mother’s and father’s years of education on children’s verbal
intelligence, as assessed by knowledge of vocabulary words. Ad-
olescent adopted and birth children in the National Longitudinal
Study of Adolescent Health were matched for gender, age, parental
education, and ethnicity. The adolescents all resided with two
parents. The mother–child and father–child correlations in biolog-
ical families were substantially higher than in adoptive families
(.41 and .36, respectively, vs. .16 and .18), indicating both genetic
and shared environmental influences. The authors concluded that
parental education exerts a modest shared environmental effect,
explaining no more than 3%–4% of the variation in adopted
children’s verbal intelligence.
On the basis of earlier studies, Munsinger (1975) already had
suggested that the adoptive parents’ home environment has only a
modest effect on their adopted children’s cognitive development,
whereas the heredity and environment of the birth parents exert a
profound influence. It should be noted, however, that larger vari-
ations in the child-rearing environment might lead to smaller
influences of genetic factors (Plomin et al., 1997).
Drastic Changes of Environment
Comparisons of the IQ of adopted children with that of their
nonadopted birth siblings may provide information on the effects
of adoption even without a complete behavior genetic design.
Unfortunately, we found only a few studies that compared
adoptees’ IQ with the IQ of their nonadopted birth siblings or
nonadopted peers from the same institution. The French Adoption
Study by Schiff et al. (1978) and Colombo et al.‘s (1992) study
from Chile both found a higher mean IQ for the adopted siblings.
In the Schiff et al. (1978) study, school failure rates and the
percentage of IQ scores below 95 were collected for a sample of 32
adopted children placed at approximately 4 months of age. Both
percentages were significantly smaller than those expected from
the social class of birth (13% vs. 55% and 17% vs. 51%, respec-
tively) and the percentages observed in a control group of non-
adopted birth siblings (56% and 49%, respectively). However, the
percentages were close to those expected from the social class of
the adoptive families (both 15%; Schiff, Duyme, Dumaret, &
Colombo et al. (1992) studied 35 school-age children (5–21
years) with a history of early malnutrition. They compared three
groups. The first group consisted of children who were raised after
recovery by adoptive families (n ? 16; placed for adoption before
12 months), the second group of children remained in institutional
care (n ? 8), and the third group consisted of children raised in
their birth family (n ? 11). Adopted children had IQs in the normal
range, and they outperformed the other groups, in particular on the
verbal subscale. The authors concluded that early undernourish-
ment may not cause irreversible damage for children but instead
can be negated by early, drastic, and stable environmental im-
provement. Similar findings have been reported by Tizard and
Hodges (1978), Lien et al. (1977), Winick et al. (1975), and Dennis
The influence of the adoption experience may become larger
when the change of environment becomes more drastic. Scarr and
Weinberg (1976) studied the IQ and school achievement of 130
Black children adopted before the age of 12 months by advantaged
White families. The adoptees from educationally average families
scored above the average level of IQ and school achievement of
the White population. The high IQ scores of the Black adoptees
suggests that IQ is malleable under rearing conditions that are
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
relevant to the tests and the schools and that deviate drastically
from the preadoption social background. Because the preadoption
cognitive status of these adoptees was not assessed, the study
findings, although impressive, remain inconclusive.
Furthermore, the environmental influences of the adoptive fam-
ily may fade as the adopted children grow older. In general,
genetic and environmental factors may not operate on the same
level across the life span. In longitudinal studies, the IQ of adopted
children has been found to become more similar to the IQ of their
birth parents with increasing age (Fulker, DeFries, & Plomin,
1988; Plomin et al., 1997), and in adulthood the correlation be-
tween the IQ of adopted children and that of their adoptive parents
appears to be much lower than the correlation with the IQ of the
biological parents (McGue, Bouchard, Iacono, & Lykken, 1993;
Plomin et al., 1997).
Not only age at assessment but also age at adoption and previous
adverse experiences or deprivation may make a substantial differ-
ence for the influence of the adoption experience. O’Connor et al.
(2000) studied Romanian adopted children in the United Kingdom
who had experienced early malnourishment and circumstances of
severe deprivation, particularly in institutional care. They found
that children adopted at older ages and institutionalized children
had lower IQ scores than younger and noninstitutionalized adopted
children. Extending previous research using data from the same
children, Rutter et al. (1998) assessed the cognitive development
of 6-year-old children adopted from Romania before they were 24
months old or when they were between 24 and 42 months old and
nondeprived, domestic adoptees from the United Kingdom
adopted before they were 6 months old. They found that later
placed Romanian children caught up considerably from entry into
the United Kingdom to age 6, but, as a group, the children
exhibited lower cognitive scores and general developmental im-
pairment compared with earlier adopted Romanian children. This
may indicate that the positive outcomes of adoption may be
hampered or reduced by severe negative experiences before adop-
In the current series of meta-analyses, we first examine whether
adopted children show better cognitive development than their
birth siblings or peers who stayed behind in their original, often
deprived environment and who did not get the chance to grow up
in an enriching adoptive family and school environment. Second,
we test whether the cognitive and language development of
adopted children lags behind compared with that of their non-
adopted peers in the normal population, in their classrooms, or in
their adoptive family. These cognitive and language delays may
exist because of the preadoption hardships these children may have
endured, in particular malnutrition, neglect, and abuse.
On the basis of the theory of cumulative risk and protective
factors, we expect (a) older age at adoption (Howe, 1998), (b)
experiences of serious deprivation before adoption (in particular,
malnutrition, neglect, and abuse), and (c) a higher age at assess-
ment (Plomin et al., 1997) to be associated with larger cognitive
delays for adopted children compared with their nonadopted en-
vironmental siblings or peers. International adoptions may create
more adaptation problems for the adopted children than domestic
adoptions because issues of cultural and ethnic differences may
arise less frequently or intensively in the latter case (Juffer, Stams,
& Van IJzendoorn, 2004; Lanz, Iafrate, Rosnati, & Scabini, 1999).
In the theory of cumulative risk and protective factors, gender of
the developing child has often been found to be relevant: Adopted
boys lag behind adopted girls in various domains of development,
including socioemotional development (Earls, 1987; Eme, 1979;
Rutter, 1990; Stams, Juffer, Rispens, & Hoksbergen, 2000; Ver-
hulst et al., 1990). Thus, in the current meta-analytic review we
test whether adopted boys fare less well than their female coun-
terparts in the cognitive domain.
We used three different search methods for identifying literature for the
meta-analytic review (Cooper & Hedges, 1994; Mullen, 1989). First, we
searched for literature in the electronic databases ERIC, PsycINFO,
MEDLINE, Sociological Abstracts, and Current Contents. We used the
keywords adopt* (where an asterisk indicates that the search contained but
was not limited to that word or word fragment), adopted children, and
adoption in combination with the keywords IQ, school results, academic
adjustment, and related terms, such as academic achievement and cognitive
ability, in searching these databases. Second, we searched the references of
the collected studies for relevant studies on academic achievement, IQ, and
cognitive development of adopted children. Third, we contacted adoption
researchers and asked them to share pertinent studies.
Studies were included in the meta-analysis if they (a) assessed IQ (using,
e.g., the Wechsler Intelligence Scale for Children [WISC], as in Frydman
& Lynn, 1989), presented scores for academic achievement measured by
school results (e.g., school grades, as in Fan et al., 2002) or school
competence (e.g., assessed with the Child Behavior Checklist, as in Pin-
derhughes, 1998), or contained information about school failures (e.g.,
Geerars, Hoksbergen, & Rooda, 1995), special education attendance (e.g.,
Verhulst et al., 1990), or learning problems (e.g., Silver, 1989); (b) used a
nonadopted comparison group (e.g., classmates; e.g., Dalen, 2001) or
included measures with standardized scores (e.g., O’Connor et al., 2000);
and (c) reported sufficient data to permit the calculation of an effect size.
We excluded studies that exclusively sampled adopted children in need of
clinical treatment (e.g., Grotevant, McRoy, & Jenkins, 1988; Jerome,
1993), children exposed to drugs in utero (e.g., Loebstein & Koren, 1997;
Moe, 2002; Thyssen Van Beveren, Spence, & Little, 2000), physically or
mentally handicapped children, and other special needs children (e.g.,
Groze, 1996; J. A. Rosenthal, Groze, & Aguilar, 1991).
The selection procedure yielded 48 studies that used IQ scores as an
outcome and 55 studies that used academic achievement as an outcome.
Data on learning problems, in particular referrals to special education, were
presented in 8 studies. In 14 studies we found results on language abilities
of adopted children (e.g., school results on language, as in Maughan et al.,
1998). In several studies, data on more than one aspect of cognitive
development were presented. For example, scores on both IQ and school
achievement were presented in 8 studies (Castle, Beckett, Groothues, &
ERA Study Team, 2000; Leahy, 1935; Schiff et al., 1978; Stams et al.,
2000; Teasdale & Owen, 1986; Tsitsikas, Coulacoglou, Mitsotakis, &
Driva, 1988; Winick et al., 1975; Witmer, Herzog, Weinstein, & Sullivan,
1963), and we used these outcomes for the separate analyses on these
different domains. We calculated effect sizes separately for subsamples
when data were presented separately for boys and girls or for children at
different ages at the time of adoption or time of the assessments. These
subsamples were considered as independent data points in the meta-
analysis. This is the reason why the number of publications often was
smaller than the number of outcomes used in the meta-analyses. However,
within the domains of IQ, school achievement, and learning problems,
every child was included only once in the meta-analyses. Table 1 provides
ADOPTION AND COGNITIVE DEVELOPMENT
Studies Included in the Meta-Analyses
School results 0.16
Benson et al. (1994)
School results ?0.36
School results 0.03 f/?0.04 m
Language ?0.02 f/?0.05 m
School results 0.09 f/0.07 m
Language 0.02 f/?0.02 m
Learning problems 0.00
Brodzinsky et al.
School competence 0.62 f/0.51 m
Brodzinsky & Steiger
School failure 0.76
Bunjes & de Vries
School results 0.24
Castle et al. (2000)
School results ?0.47, IQ 0.47
Clark & Hanisee
Colombo et al.
Cook et al. (1997)
School competence 0.56 f/0.16 m
School results 0.47 (Colombia), ?0.07
Language 0.43 (Colombia), ?0.05
Learning problems 0.50
IQ ?1.28 (intraracial), ?1.36
De Jong (2001)
School competence 0.65
School results 0.00
Fan et al. (2002)
School grades ?0.02
Education level ?0.03
Fisch et al. (1976)
School results 0.50
Frydman & Lynn
Gardner et al. (1961)
School achievement 0.09
Geerars et al. (1995)
School results 0.19
Hoopes et al. (1970)
1–2 shifts in
Horn et al. (1979)
W. J. Kim et al.
School results 0.74
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
Table 1 (continued)
W. J. Kim et al.
School competence ?0.39
Lansford et al.
School grades 0.46
School grades 0.00
Levy-Shiff et al.
IQ ?1.10 f/?2.00 m
Lien et al. (1977)
Lipman et al. (1992)
School performance ?0.05 f/0.16 m
Long time in
School competence 0.46
IQ ?0.00 f/?1.00 m
Morison & Ellwood
IQ 1.45 (combined)
Neiss & Rowe (2000)
75% United States
O’Connor et al.
IQ ?0.56 (combined)
Palacios & Sanchez
School competence ?0.18
School competence 0.64 (combined)
Plomin & DeFries
Priel et al. (2000)
School competence 0.77 f/1.12 m
School performance ?0.18
Scarr & Weinberg
88% United States
IQ 0.75 (combined)
Schiff et al. (1978)
School results ?0.70
Sharma et al. (1996)
81% United States
School results 0.37 (combined)
Sharma et al. (1998)
School competence ?0.45 f/?0.61 m
Learning problems 1.21
Learning problems 1.38
Skodak & Skeels
Smyer et al. (1998)
Education level ?0.82
Stams et al. (2000)
School results 0.33
IQ ?0.34 f/?0.73 m
Learning problems ?0.05
Teasdale & Owen
Education level 0.32
Tizard & Hodges
IQ ?0.40 (older), ?0.62 (younger)
Tsitsikas et al.
IQ 0.64, school performance 0.29
ADOPTION AND COGNITIVE DEVELOPMENT
Table 1 (continued)
Verhulst et al. (1990)
Perc. special education 0.25 f/0.29 m
& Verhulst (1995)
School competence 0.28 f/0.41 m
Wattier & Frydman
Westhues & Cohen
School performance 0.13
Wickes & Slate
School results 0.09 f/0.07 m
Language 0.07 f/0.03 m
Winick et al. (1975)
School performance 0.00
Witmer et al. (1963)
School performance 0.00
More information about the individual studies is available on request. Country/region of study is the country or region in which the adopted children were living at the time the study was
conducted; country/region of child’s origin is the country or region from which the children were adopted. Preadoption status of the adopted children is given if any information was available. Effect
sizes are given in Cohen’s d separately for female (f) and male (m) adoptees when possible. Norm ? test norm; Perc. ? percentage.
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
an overview of the collected studies and the variables derived from each
The coding system for design and sample characteristics is presented in
Table 2. The design characteristics included sample size, type of publica-
tion outlet, year of publication, and kind of comparison group. Most studies
were published in refereed scientific journals, but the meta-analytic review
also included unpublished reports, books, and book chapters. Therefore, we
examined whether effects found in scientific journals differed from effects
found in nonrefereed reports or books and book chapters. We also con-
trasted recent publications (1990 and later) with older publications (before
We coded for the kind of comparison group. We hypothesized that it
would make a difference whether adopted children were compared with (a)
nonadopted birth siblings who remained with or were restored to the birth
parents (e.g., Colombo et al., 1992) or their institutionalized peers who
remained in the institutional setting where the adopted children once lived
(e.g., Palacios & Sanchez, 1996), (b) classmates (e.g., Tsitsikas et al.,
1988), (c) environmental siblings unrelated to the adopted children but in
the adoptive family (e.g., Westhues & Cohen, 1997), (d) the general
population (e.g., Duyme, 1988), or (e) norms on standardized tests—for
example, in one study adopted children were assessed with the WISC—
Revised (as in Levy-Shiff, Zoran, & Shulman, 1997), and the standardized
IQ mean of 100 (SD ? 15) was used for comparison. We wished to
examine whether different effect sizes for cognitive development of
adopted children could be attributed to the different kinds of comparison
groups. In particular, we were interested in the contrast between adopted
children and children left behind (nonadopted birth siblings or institutional
peers) and in the contrast between adopted children and their environmen-
tal comparisons (classmates, unrelated siblings, the general population, and
standardized test norms).
The number of studies comparing adopted children with their non-
adopted birth siblings or institutional peers was rather small (Colombo et
al., 1992; Dennis, 1973 [with domestic and intercountry adoptions]; Pala-
Coding System for Studies on Adopted Children
Age at adoption1 ? 0–12 months
2 ? 12–24 months
3 ? ? 24 months
4 ? not reported
1 ? 0–4 years
2 ? 4–12 years
3 ? 12–18 years
4 ? ?18 years
5 ? not reported
1 ? male
2 ? female
3 ? mixed
1 ? domestic (inside birth country)
2 ? international (outside birth country)
3 ? not reported
1 ? no samplewide certified abuse, neglect, or malnourishment
2 ? majority of participants in a sample were abused, neglected,
Age at assessment
Gender of comparison groups
Domestic or international adoption
Extreme deprivation or abuse
Sample sizeSample size of adopted children, nonadopted children (control
group), and total sample size for which results are reporteda
1 ? journal article
2 ? other publication (book, book chapter, report, dissertation)
1 ? general population (e.g., Cook et al., 1997)
2 ? classmates (e.g., Andresen, 1992)
3 ? birth siblings (e.g., Colombo et al., 1992) or institutional
peers (e.g., Dennis, 1973)
4 ? environmental siblings (unrelated siblings in the adoptive
family; e.g., Sharma et al., 1998)
5 ? standard scoresb(e.g., Stams et al., 2000) or percentage
comparisonsc(percentage of school failures; e.g., Geerars et
1 ? before 1960
2 ? 1960–1969
3 ? 1970–1979
4 ? 1980–1989
5 ? 1990 or later
Publication outlet (refereed journal
or other publication)
Year of publication
aIf the control group was compared with two or more groups of adopted children, the control group was split
up accordingly (see Bakermans-Kranenburg, Van IJzendoorn, & Juffer, 2003).
the IQ mean of the adopted children in the research group was compared with the standardized scores of an IQ
test (M ? 100, SD ? 15).
group was compared with the percentage of adopted children in the general population.
bStandard scores means that
cPercentage comparisons means the percentage of adopted children in the research
ADOPTION AND COGNITIVE DEVELOPMENT
cios & Sanchez, 1996 [for the meta-analysis on school achievement];
Schiff et al., 1978; Smyer, Gatz, Simi, & Pedersen, 1998 [for the meta-
analysis on school achievement]; Tizard & Hodges, 1978 [with children
adopted at a younger or older age]). It was therefore unwise to conduct
moderator analyses on this set of studies. Because this set of studies was
different from the majority of adoption studies comparing adopted children
with their environmental siblings or peers, we decided to separate the two
sets and to conduct moderator analyses only on the large set of compari-
sons with environmental siblings or peers.
We coded the age of the children when they were adopted, age at
assessment, gender, and whether the children were adopted internationally
or domestically. We examined meta-analytically whether early adoptions
(until 1 year of age) resulted in different effects compared with later
adoptions. In view of the growing genetic influences on adopted children’s
cognitive development in adolescent years and later (e.g., Fulker et al.,
1988; Plomin et al., 1997), we contrasted studies with assessment ages of
younger and older than 12 years. Also, we coded whether the adopted
children experienced deprivation. Deprivation was coded in cases of doc-
umented malnutrition (e.g., Colombo et al., 1992), neglect, and/or abuse
(e.g., De Jong, 2001; Dennis, 1973) or a combination of these circum-
stances (e.g., McGuinness & Pallansch, 2000; O’Connor et al., 2000).
All studies were coded by two coders. The average agreement between
the two coders across the moderator variables was 95%.
We conducted four meta-analyses, one for IQ scores as indexed by
formal IQ tests, one for school achievement as indicated by school grades
and parent or teacher ratings of school competence, one for language
abilities as indicated by specific language tests and parent or teacher ratings
of language development, and one for learning problems as indicated by
school failure and enrollment in special education.
For each study, we calculated an effect size: the standardized difference
between the means of two groups (Cohen’s d). Effect sizes indicating
delays in adopted children’s development got a positive sign (as we
expected that adopted children would be outperformed by their nonadopted
environmental peers or siblings), whereas effect sizes indicating better
cognitive development for adopted children (as we expected in the case of
comparisons with peers or siblings who stayed behind) got a negative sign.
When in an article more than one outcome was reported for the same
domain—for example, two IQ tests—we averaged these outcomes within
the study to have one effect size per study (Cooper & Hedges, 1994;
Mullen, 1989). Because the studies included in the meta-analysis reported
various statistics, we used Mullen’s (1989) Advanced Basic Meta-Analysis
program to transform all results into Cohen’s d. Mullen (1989) and Mullen
and Rosenthal (1985, Chapter 6) provided the formulas for transformation
of t, r, and F statistics into Cohen’s d.
The resulting study effect sizes were analyzed with Borenstein, Roth-
stein, and Cohen’s (2000, Version 1.025) Comprehensive Meta-Analysis
(CMA) program. CMA allows for computation of combined effect sizes
via random effects error models. Significance tests and moderator analysis
in fixed models may be regarded as applying only to the specific set of
studies at hand (Cooper & Hedges, 1994). In random effects models
(Hedges & Olkin, 1985), generalization is to the population of studies from
which the current set of studies was drawn (R. Rosenthal, 1995). Our goal
was to make summary statements about likely differences between adopted
and nonadopted children’s cognitive development even when the sources
of discrepancies between study results were poorly understood (Rauden-
Some extremely large sample sizes were winsorized in the weighting
function to prevent the results from being unduly determined by only one
outlying study (Hampel, Ronchetti, Rousseeuw, & Stahel, 1986). No
outlying effect sizes (z ? ?3.26 or z ? 3.26; Tabachnick & Fidell, 2001)
were detected in any of the meta-analytic data sets after conversion into
Fisher’s z (Mullen, 1989).
We computed the Q statistics to test the homogeneity of the specific set
of effect sizes (Borenstein et al., 2000). We also computed 85% confidence
intervals (CIs) around the point estimate of each set of effect sizes.
Depending on the homogeneity of the set, the CIs were based on either
fixed or random estimates. When we tested moderators, inspection of the
overlap between the CIs provided a test of the differences among the
combined effects of subsets of study effect sizes grouped by moderators. In
the case of combined effect sizes in heterogeneous sets of study outcomes,
this approach of comparing 85% CIs served as the significance test under
a random error model (Goldstein & Healy, 1995). Nonoverlapping 85%
CIs indicated a significantly different effect size in moderator-determined
subsets of study outcomes.
We used the trim and fill method (Duval & Tweedie, 2000a, 2000b) to
calculate the effect of potential data censoring on the outcome of the
meta-analyses. According to this method, a funnel plot is constructed of
each study’s effect size against the sample size or the standard error. These
plots should be shaped like a funnel if no data censoring is present.
However, because smaller or nonsignificant studies are less likely to be
retrieved, studies in the bottom left-hand corner of the plot are often
omitted (Sutton, Duval, Tweedie, Abrams, & Jones, 2000). For the meta-
analyses on IQ, school achievement, and language, the rightmost studies
considered to be symmetrically unmatched were trimmed. The trimmed
studies could then be replaced and their missing counterparts imputed or
filled as mirror images of the trimmed outcomes. This allows for the
computation of an adjusted overall CI (Gilbody, Song, Eastwood, &
Sutton, 2000; Sutton et al., 2000). Analyses were conducted with the
S-Plus trim and fill program (B. Biggerstaff, personal communication,
September 19, 2004; E. Dusseldorp, personal communication, September
20, 2004; S. Duval, personal communication, September 22, 2004).
Comparison of adopted children’s IQ with IQ of siblings or
peers who stayed behind.
The comparison of the adopted chil-
dren with their nonadopted peers who stayed behind in their family
or institution (birth siblings or institutionalized peers) showed a
large and significant effect size (d ? ?1.17) in favor of the
adopted children in a homogeneous set of six studies (n ? 253),
Q(5) ? 3.8, p ? .58. That is, the negative combined effect size
means that the adopted-away children showed higher IQ scores
than their nonadopted siblings and peers who stayed behind. In
contrast, the comparison between the IQ of adopted children and
that of their environmental siblings or peers was not significant
(d ? 0.13, p ? .19, k ? 42, n ? 6,411) in a heterogeneous set of
studies, Q(41) ? 275.6, p ? .01.1
The difference in effect sizes for the two different comparison
groups was significant. The 85% CI around the point estimate
involving siblings or peers who stayed behind (d ? ?1.17, CI ?
?1.36, ?0.99) did not overlap with the CI for the environmental
siblings or peers (d ? 0.13, CI ? ?0.01, 0.26). Thus, the adopted
children (a) outperformed their siblings or peers who were left
behind and (b) performed as well as their environmental siblings,
their classmates, and their peers in the general population of their
1Because Teasdale and Owen’s (1986) study included the outlying
number of 4,450 participants, we decided to winsorize this sample size
before computing the combined effect size (Hampel et al., 1986) and to
divide both the number of adopted and the number of comparison children
in this study by 5. The combined effect sizes with and without winsorizing
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
current environment. Further, the magnitudes of effect revealed by
these comparisons were different.
Comparison of adopted children’s IQ with IQ of environmental
siblings or peers.
Because the set of studies comparing adopted
children’s IQ with the IQ of children who stayed behind was too
small for further moderator analyses (k ? 6), we continued with
the set of studies comparing adopted children’s IQ with that of
environmental siblings or peers. As noted above, the combined
effect size in this set of studies indicated the absence of a differ-
ence in IQ between adopted children and their environmental
siblings or peers. It should be noted, however, that the comparison
with the general population (d ? 0.34, CI ? 0.18, 0.50) and with
environmental peers (d ? 0.79, CI ? 0.48, 1.11) significantly
favored the nonadopted peers, whereas analyses that used average
population IQs showed an advantage for the adopted children of
three quarters of a standard deviation (d ? ?0.75, CI ? ?0.96,
?0.53). However, because of the possible Flynn (1987; see below)
effect, this comparison might be somewhat inflated (O’Connor et
In the set of studies comparing adoptees with environmental
peers, the contrast for publication outlet was not significant. The
85% CIs for the effect sizes were largely overlapping, indicating
that journal articles showed IQ score differences between adopted
and nonadopted children that were similar to those of other pub-
lications (d ? 0.12, CI ? 0.05, 0.29; and d ? 0.08, CI ? 0.00,
0.15, respectively). Contrast for year of publication (before 1990
vs. 1990 and later) was not significant either (d ? 0.14, CI ?
?0.01, 0.28; and d ? 0.06, CI ? ?0.27, 0.39, respectively).
IQ differences between the adopted and comparison children
were not significantly different when we contrasted early arrivals
(before 12 months old) with later arrivals. The CI for the combined
effect size of studies on children adopted before 12 months (d ?
0.14, CI ? ?0.03, 0.31) was largely overlapping with the intervals
for the older ages (combined across adoptions after 12 months, d ?
0.07, CI ? ?0.20, 0.34). Contrasting samples of children who
were younger than 12 years at IQ assessment (d ? 0.12, CI ?
?0.03, 0.28) with the other samples (d ? 0.20, CI ? ?0.16, 0.56)
did not yield a significant difference. The only effect size for a
moderator subgroup that was significant was associated with adult
samples (d ? 0.87; k ? 4; n ? 937, winsorized; p ? .05).
However, the 85% CI for this effect size (CI ? 0.28, 1.45)
overlapped with the CI for the samples assessed at age 12 or
younger (d ? 0.15, CI ? ?0.01, 0.32; and d ? ?0.28, CI ?
?0.90, 0.33, for ages 4–12 years and 0–4 years, respectively).
Type of adoption—domestic versus international—was not re-
lated to the effect sizes of the studies; the 85% CIs for both
estimates were largely overlapping (d ? 0.21, CI ? 0.04, 0.37; and
d ? ?0.19, CI ? ?0.58, 0.20, respectively). Nor did we find
differences in effect sizes between samples with male versus
female adopted children (d ? 0.45, CI ? ?0.06, 0.96; and d ?
0.41, CI ? ?0.05, 0.87, respectively). Eleven samples included
adoptees who had been abused, neglected, and/or malnourished.
These samples showed a somewhat larger combined effect size
(d ? 0.27, CI ? ?0.26, 0.80) than the other studies (d ? 0.05,
CI ? ?0.08, 0.19), but the CIs were large and overlapping.
The set of IQ studies did not show a bias against locating studies
with small effect sizes and a small number of participants (L0? 0;
Comparison of adopted children’s school achievement with
achievement of siblings or peers who stayed behind.
the IQ meta-analyses, the comparison with the nonadopted siblings
or peers who stayed behind showed favorable school achievement
of the adopted children (d ? ?0.55, p ? .05, k ? 3, n ? 523).
However, this set of studies was small, and the CI was rather large.
The 85% CI of the negative combined effect size (d ? ?0.55,
CI ? ?0.88, ?0.21) did not overlap with that of the comparison
with the environmental siblings and peers (d ? 0.19, CI ? 0.14,
0.25). Again, similar to the IQ comparisons, the adopted-away
children outperformed their nonadopted siblings and peers who
were left behind. However, in contrast with IQ, adoptees’ school
achievement did not catch up completely with that of their envi-
ronmental siblings or peers.
Comparison of adopted children’s school achievement with
achievement of environmental siblings or peers.
sion of the studies that compared adopted children with their
siblings or peers who were left behind, the combined effect size for
the remaining set of studies was 0.19 (p ? .01, k ? 52, n ?
78,662). Adopted children did less well in school than nonadopted
peers in their environment. In the comparisons with peers from the
general population (d ? 0.26, p ? .01, k ? 31, n ? 69,085) as well
as with classmates (d ? 0.13, p ? .05, k ? 9, n ? 3,721), this
lagging behind of adopted children’s school achievement was
Age at adoption appeared to be important. For studies with
children who were adopted in the 1st year of their life, the
difference with environmental peers was minimal (d ? 0.09, p ?
.22, k ? 26, n ? 37,991). For studies with children adopted in the
2nd year of life, the combined difference was significant (d ?
0.32, p ? .01, k ? 10, n ? 11,059), as it was for studies with
children adopted after the age of 2 years (d ? 0.42, p ? .01, k ?
10, n ? 5,742). Contrasting samples of children who were adopted
before 12 months of age (d ? 0.09, CI ? 0.00, 0.17) with the other
samples (d ? 0.28, CI ? 0.22, 0.35) showed no overlap between
85% CIs and, thus, a significant larger combined effect in the latter
samples. Only later adoption (after the 1st year of life) appeared to
be associated with a delay in school achievement.
The 85% CIs for the combined effect size of samples with
assessment of school achievement before age 12 (d ? 0.22, CI ?
0.13, 0.31) versus the other samples (d ? 0.18, CI ? 0.10, 0.25)
were largely overlapping. Both studies with domestically adopted
children and studies with internationally adopted children showed
significant combined effect sizes in the same range (d ? 0.22 and
d ? 0.15, respectively), with largely overlapping 85% CIs (CI ?
0.14, 0.30; and CI ? 0.06, 0.24, respectively). We did not find
differences in effect sizes between samples with male (d ? 0.14,
CI ? 0.02, 0.27) and samples with female adopted children (d ?
0.07, CI ? ?0.05, 0.19). Although only three samples with abuse,
neglect, and/or malnourishment were included in this meta-
analysis, this set showed a significantly larger combined effect size
(d ? 0.46, CI ? 0.31, 0.60) than the other studies (d ? 0.19, CI ?
0.13, 0.24), as the 85% confidence boundaries did not show any
Studies published in journals (d ? 0.21, CI ? 0.15, 0.27)
showed a larger effect size than studies reported in other outlets
(d ? 0.03, CI ? ?0.05, 0.11), with no overlap between CIs. Year
With the exclu-
ADOPTION AND COGNITIVE DEVELOPMENT
of publication did not affect the difference, as the 85% CIs for the
studies published before 1990 (d ? 0.20, CI ? 0.11, 0.29) over-
lapped with those of the studies published in 1990 or later (d ?
0.19, CI ? 0.12, 0.26). The set of school achievement studies did
not show a publication bias (L0? 0; R0? 1).
Studies comparing language development of adopted and non-
adopted siblings or institutional peers were missing entirely. In 14
studies we found specific information on language abilities of
adopted children compared with their nonadopted environmental
peers. The combined effect size was small but significant (d ?
0.09, p ? .05, k ? 14, n ? 15,418), and the studies were
homogeneous, Q(13) ? 18.6, p ? .14. Adopted children appeared
to show some delay in language development.
We found no significant moderators of the language effect (see
Table 3). This set of studies contained no abused or malnourished
samples. The set of language studies did not show a publication
bias (L0? 0; R0? 3; but the number of studies was rather small,
and L0therefore was the preferred estimator).
A total of eight study outcomes (N ? 13,291; n ? 3,018 adopted
children) were used for this meta-analysis. All studies compared
adopted children with nonadopted, environmental peers. The com-
bined effect size was significant (d ? 0.55, p ? .01). The 85% CI
ranged from 0.35 to 0.75 in a heterogeneous set of studies, Q(7) ?
119.9, p ? .01. The adopted children showed significantly more
learning problems for which special treatment was needed. To
further specify this elevated risk for learning problems in adopted
children, we computed the percentage of adopted children referred
to special education in four studies (Bohman, 1970; Brodzinsky &
Steiger, 1991; Stams et al., 2000; Verhulst et al., 1990), which
amounted to 11.2% (n ? 2,873 adopted children). In two studies,
the exact numbers of referrals for both the adopted and the non-
adopted pupils were available. Bohman (1970) reported on 125
adopted (7 referrals, 5.6%) and 2,232 nonadopted children (132
referrals, 5.9%) in Sweden. Verhulst et al. (1990) reported on
2,148 adopted children (284 referrals, 13.2%) and 933 nonadopted
children (41 referrals, 4.4%). Combining the two studies, we found
12.8% referrals to special educational services in the adopted
group (n ? 2,273) and 5.5% referrals in the nonadopted group
(n ? 3,165). Thus, in these two studies there was a twofold
increase in special education referrals for adopted children com-
pared with nonadopted comparisons.
Because the data set contained only eight studies, we did not
conduct moderator analyses. The set of studies on learning prob-
lems did not show a publication bias (L0? 0; R0? 1; but the
number of studies was rather small, and L0therefore was the
Our first hypothesis concerned the potential cognitive advan-
tages of adoption over remaining with an overburdened family or
in a deprived institution. With regard to IQ scores, the adopted
children outperformed their siblings or peers who were left behind.
In terms of school achievement, the adopted children also outper-
formed their left-behind siblings and peers. Unfortunately, the
number of comparisons is rather small, as only six studies pre-
sented pertinent data on birth siblings or peers who remained in
their own environment (Colombo et al., 1992; Dennis, 1973;
Palacios & Sanchez, 1996; Schiff et al., 1978; Smyer et al., 1998;
Tizard & Hodges, 1978). Nevertheless, these unique studies are
consistent with the possibility that adopted children are able to
profit from the positive change of environment offered by adoption
and their subsequent upbringing in educationally more stimulating
Our second hypothesis concerned the potential cognitive delays
of adopted children compared with their current, environmental
siblings or peers. Overall, we found that studies reported a negli-
gible difference in the IQ of adopted children and their nonadopted
environmental siblings or peers. Comparing their school achieve-
ment, we documented that the adopted children did somewhat less
well in school, but the effect size was rather small. Their language
abilities also showed a small but significant delay compared with
the abilities of their environmental siblings or peers, but, again, the
effect size was small. The largest delay was found in a set of eight
studies comparing the learning problems of adopted children with
those of their environmental peers. The percentage of adopted
children struggling with learning problems was significantly larger
than that of nonadopted children. We found a twofold increase in
special education referrals in adopted children compared with
nonadopted comparisons in the same countries of study. However,
it should be noted that the percentage of children with learning
problems who needed treatment or referral to special education
was generally rather small, in the adopted group as well as in the
general population, and that this finding is based on a small
number of studies.
Taken together, this series of meta-analyses documents the
potential positive impact of the adoption experience on the adopted
children’s cognitive development compared with that of the chil-
dren left behind. The studies also show the nearly normal cognitive
competence and only somewhat delayed school performance of
adopted children. However, it is not possible to draw a firm
conclusion from the present meta-analysis that adoption has a
positive impact on the cognitive development of adopted children.
Information on the cognitive development of the adopted children
in their preadoption living arrangements is absent, and longitudinal
information on the cognitive abilities of children who were not
adopted and remained in the birth family or institution is also
lacking. Obviously, the natural adoption quasi-experiment is not as
capable of revealing causal effects as are true experimental trials
using pre- and posttests and randomized control groups. Therefore,
it cannot be ruled out that (some) adopted children were selected
for adoption because they seemed brighter or had better social
skills than children not selected for adoption. Without tests before
and after adoption placement, we cannot be certain about the
explanation of the positive effects found in our meta-analyses.
That being said, consistent with concluding a positive effect of
adoption, there is empirical evidence pointing to positive effects of
adoption placement and positive influences of (short) interventions
in institutions (e.g., Spira et al., 2000). For example, Morison,
Ames, and Chisholm (1995) found that, according to the parents,
Romanian children exhibited delays in all areas of development
but that 11 months postadoption many delays had disappeared. A
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
Meta-Analyses of Studies Comparing Cognitive Development of Adopted and Nonadopted Children
Meta-analytic results on IQ
Meta-analytic results on school achievement
Meta-analytic results on language
Siblings/peers left behind
Year of publication
1990 and later
Age at adoption
Age at assessment
Domestic or international adoption
Deprivation or abuse
No abuse or malnourishment
Abuse or malnourishment
CI ? confidence interval.
* p ? .05.
** p ? .01.
ADOPTION AND COGNITIVE DEVELOPMENT
small group of 16 children had been assessed with the Revised
Gesell Developmental Schedules more than once by coders un-
aware of the child’s status, and, on average, children progressed
two developmental quotient points per month in their adoptive
home. Similarly, Rutter et al. (2000) described a dramatic catch-up
for Romanian adoptees from the time of adoption placement (mean
IQ just above 60) to 4 years of age (mean IQ above 100), assessed
with parent report. Interventions within the institutional setting
also have shown positive effects on children’s growth and devel-
opment (Hakimi-Manesh, Mojdehi, & Tashakkori, 1984; T. I.
Kim, Shin, & White-Traut, 2003; Spira et al., 2000).
The overall meta-analytic findings were moderated only by a
few variables, and potential moderators such as the gender of the
children and domestic versus international adoption did not appear
to make a significant difference. Also, moderators such as publi-
cation outlet (journal articles vs. other publications) and year of
publication were not associated with effect sizes, except for school
achievement, for which journal articles documented larger delays
in adopted children than did other publications. Publication bias
and data censoring did not seem to be present in the current data
set of adoption studies.
Age at assessment was not a significant moderator. However, in
adult samples we found a rather large IQ difference between
adopted and nonadopted adults. This is not inconsistent with
findings from longitudinal behavior genetic studies, which have
documented that, with growing age, adopted children show less IQ
similarity to their adoptive parents and more similarity to their
birth parents (McGue et al., 1993; Plomin et al., 1997). However,
school performance and language differences were not associated
with age at assessment. Because preadoption assessments are
lacking, we can only speculate about the existence and size of
age-specific developmental trends and whether they represent true
developmental shifts or are the results of cohort differences at the
time of adoption.
Moderators that did seem to have an important impact on the
cognitive associations with adoption were the age at which the
child was adopted and whether the adopted sample came from
abusive and/or neglecting backgrounds. Age at adoption did not
seem to matter for the IQ of the adopted children, but it did matter
for their school achievement. Children adopted in their 1st year of
life did not show any delays in school achievement, whereas
children adopted after their first birthday lagged behind. Thus,
early adoption may be a protective factor not so much for cognitive
competence (IQ) as for cognitive performance (school achieve-
ment). This may occur because the positive effects of the adoption
environment have greater impact during an important period in the
children’s life. It may also indicate the importance of the first
attachment relationships (with the adoptive parents) developing
around 10–14 months after birth (Main, 1999). These early attach-
ment relationships may function as protective factors, buffering
against the stress of adoption losses.
Another interpretation of the early adoption moderator effect is
the relatively short duration of deprivation before adoption. Briefer
preadoption time may imply shorter exposure to risk factors such
as abuse or neglect. This interpretation converges with the signif-
icant effect we found for the impact of preadoption abusive or
otherwise deprived background on school achievement. Adopted
children who were exposed to severe abuse or neglect prior to
adoption lagged further behind in school achievement than
adopted children without such backgrounds, although their IQ
scores did not show a similar difference. Of course, preadoption
abuse or neglect is a major risk factor that appears to leave its
marks on the children’s school achievement even after adoption
into less deprived social contexts. In fact, it is surprising that we
did not find a similar negative effect on IQ.
The discrepancy between adopted children’s positive attainment
in terms of IQ and their somewhat delayed school achievement (in
children adopted after their first birthday) may indicate an adop-
tion decalage—that is, a gap between adopted children’s compe-
tence and their actual school performance. In our meta-analyses,
the adoption decalage was largest for those children who came
from extremely deprived backgrounds, because their school
achievements lagged further behind than those of adopted children
from less deprived backgrounds, but their IQ was higher than that
of their nonadopted comparisons. Adopted children appear to show
the same cognitive potential as their environmental comparisons,
as these two groups do not differ in IQ. However, they are not able
to catch up completely in school achievement. A larger number of
adopted children develop learning problems requiring referral to
special educational or therapeutic services. This is parallel to the
socioemotional problems present in a minority of adopted children
(e.g., Bimmel et al., 2003).
We speculate that the adoption decalage is intensified by the
socioemotional demands required by the achievement orientation
in a group setting at school. In middle childhood, some adopted
children may begin to struggle with the loss of their birth parents
(Brodzinsky, Schechter, & Henig, 1992; Brodzinsky & Steiger,
1991; Leon, 2002), and the burden of grief may hamper or slow
down their progress through school. Unresolved loss is associated
with intrusions and ruminations that limit the ability to focus on
the tasks at hand (Main, 1999). We suggest that adopted children
are as competent as their nonadopted peers. However, in a small
number of adopted children, school performance does not reach
the expected level because the socioemotional problems related to
their adoption status decrease their ability to concentrate on
An alternative explanation is that, in general, adopted children
are able to profit from the favorable circumstances offered by the
adoptive family (Schaffer, 1998) but that for some children
catch-up is incomplete, in particular for children who experienced
more extreme deprivation (e.g., malnourishment, neglect, and/or
abuse; Rutter & O’Connor, 2004) or later adoption. Genetically
determined problems and enduring effects of deprivation and
institutionalization may result in concentration problems and cog-
nitive delays, hampering successful progress at school. Moreover,
adoptive parents may be more readily inclined to perceive learning
problems, as they are often more aware of available services and
more alert to potential problems (Warren, 1992). They may be
more likely to seek special treatment or special education for their
The current set of meta-analyses suffers from some weaknesses.
The studies included in the current investigation are rather heter-
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
ogeneous. For example, the measures used to assess IQ vary from
the Stanford–Binet and the WISC to the California Mental Matu-
rity tests. The children come from different countries and conti-
nents, and the age at adoption also widely differs. Although we did
not find much systematic influence of these diverse factors, they
nevertheless create more (error) variance than ideally one wants to
see. The heterogeneity of the data sets forced us to use the more
conservative approach of the random effects model instead of
fixed effect models.
Another weakness is the comparison of the adopted children’s
IQ score with the average score (100) of a population (the standard
scores approach). Flynn (1987) noted that intelligence tests tend to
become outdated as people from Western countries increasingly
profit from ongoing education and the information era, thus en-
hancing their intellectual potential (the Flynn effect). Therefore,
older intelligence tests might give inflated profiles of one’s intel-
ligence as the average IQ level of a younger cohort becomes
substantially higher than 100. Because of the Flynn effect (Flynn,
1987), therefore, the average IQ of 100 in our meta-analysis might
underestimate the average IQ level of Western populations, which,
during the past 5 decades, have shown remarkable increases in
mean IQ scores (Flynn, 1987). Thus, similarity of adopted chil-
dren’s IQ with an average IQ of a measure that has not been
updated during the last 10 or even 20 years may underestimate
cognitive delays (O’Connor et al., 2000). This may be the reason
that our comparison with standard mean scores shows a consider-
able advantage for the adopted children, whereas direct compari-
sons of their scores with scores of classmates or other comparison
groups show adopted children to lag behind. Longitudinal studies,
preferably using updated intelligence tests at multiple time points,
are needed to determine whether the Flynn effect plays a role in
explaining our finding. In our meta-analytic database we did not
have a sufficiently large number of studies with multiple time
points to address this issue adequately. Also, in longitudinal stud-
ies, loss of participants could be a potential problem (see, e.g.,
Hodges & Tizard, 1989).
In sum, our meta-analyses lead to three theoretically and prac-
tically important conclusions. First, for many adopted children,
adoption involves a drastic change of environment, and this change
may be an effective intervention that improves their cognitive
development. In a cognitively richer and emotionally safer envi-
ronment, on average, adopted children may recover and nearly get
back on a normal track (see also Morison et al., 1995). Although
it is possible that children with higher cognitive skills are more
likely to be adopted, it is also plausible, and related evidence
suggests, that adoption can be associated with a remarkable recov-
ery from often extremely adverse preadoption circumstances and
should be considered as evidence for children’s resilience. It
should be noted, however, that the number of studies comparing
adopted children with their siblings or peers who stayed behind is
small, and more research is needed to confirm the recovery
Second, adopted children do not completely catch up in school
performance with their nonadopted environmental peers. The dis-
crepancy between the adopted children’s development in terms of
IQ and their school achievement indicates an adoption decalage
similar to the socioemotional problems presented by a minority of
adopted children (e.g., Bimmel et al., 2003). Although the decal-
age is generally small, it is a robust finding.
Third, in a small set of studies we found that the percentage of
adopted children who needed special education for their learning
problems was about twice as large as the percentage of nonadopted
children. This minority of adopted children with learning problems
is clinically important because the children suffer from these
problems and need special treatment. However, their difficulties
should not be confused with those experienced by the average
adopted child. Most adopted children do remarkably well, cer-
tainly much better than their siblings or peers who had to stay
behind in poor institutions or deprived families.
References marked with an asterisk indicate studies included in the
*Andresen, I. L. (1992). Behavioral and school adjustment of 12–13-year-
old internationally adopted children in Norway: A research note. Journal
of Child Psychology and Psychiatry and Allied Disciplines, 33, 427–
Bakermans-Kranenburg, M. J., Van IJzendoorn, M. H., & Juffer, F. (2003).
Less is more: Meta-analysis of sensitivity and attachment interventions
in early childhood. Psychological Bulletin, 129, 195–215.
*Benson, P. L., Sharma, A. R., & Roehlkepartain, E. C. (1994). Growing
up adopted: A portrait of adolescents and their families. Minneapolis,
MN: Search Institute.
*Berg-Kelly, K., & Eriksson, J. (1997). Adaptation of adopted foreign
children at mid-adolescence as indicated by aspects of health and risk
taking: A population study. European Child and Adolescent Psychiatry,
Bimmel, N., Juffer, F., Van IJzendoorn, M. H., & Bakermans-Kranenburg,
M. J. (2003). Problem behavior of internationally adopted adolescents: A
review and meta-analysis. Harvard Review of Psychiatry, 11, 64–77.
*Bohman, M. (1970). Adopted children and their families: A follow-up
study of adopted children, their background, environment and adjust-
ment. Stockholm, Sweden: Propius.
Bohman, M., & von Knorring, A. L. (1979). Psychiatric illness among
adults adopted as infants. Acta Psychiatrica Scandinavica, 60, 106–112.
Borenstein, M., Rothstein, D., & Cohen, J. (2000). Comprehensive Meta-
Analysis. A computer program for research synthesis (Version 1.025)
[Computer software]. Englewood, NJ: Biostat.
Bouchard, T. J., Jr., & McGue, M. (1981, May 29). Familial studies of
intelligence: A review. Science, 212, 1055–1059.
Brand, A. E., & Brinich, P. M. (1999). Behavior problems and mental
health contacts in adopted, foster and non-adopted children. Journal of
Child Psychology and Psychiatry, 40, 1221–1229.
Brodzinsky, D. M. (1987). Adjustment to adoption: A psychosocial per-
spective. Clinical Psychology Review, 7, 25–47.
Brodzinsky, D. M. (1990). A stress and coping model of adoption adjust-
ment. In D. M. Brodzinsky & M. D. Schechter (Eds.), The psychology of
adoption (pp. 3–24). New York: Oxford University Press.
*Brodzinsky, D. M., Schechter, D. E., Braff, A. M., & Singer, L. M.
(1984). Psychological and academic adjustment in adopted children.
Journal of Consulting and Clinical Psychology, 52, 582–590.
Brodzinsky, D. M., Schechter, M. D., & Henig, R. M. (1992). Being
adopted: The lifelong search for self. New York: Anchor Books.
*Brodzinsky, D. M., & Steiger, C. (1991). Prevalence of adoptees among
special education populations. Journal of Learning Disabilities, 24,
*Bunjes, L., & de Vries, A. (1988). Een nieuwe start . . . een nieuwe taal
[A new start . . . a new language]. Kind en Adolescent, 9(3), 199–207.
ADOPTION AND COGNITIVE DEVELOPMENT
Cadoret, R. J. (1990). Biologic perspectives of adoptee adjustment. In
D. M. Brodzinsky & M. D. Schechter (Eds.), The psychology of adoption
(pp. 25–41). New York: Oxford University Press.
*Castle, J., Beckett, C., Groothues, C., & ERA Study Team. (2000). Infant
adoption in England: A longitudinal account of social and cognitive
progress. Adoption and Fostering, 24, 26–35.
*Clark, E. A., & Hanisee, J. (1982). Intellectual and adaptive performance
of Asian children in adoptive American settings. Developmental Psy-
chology, 18, 595–599.
*Colombo, M., de la Parra, A., & Lopez, I. (1992). Intellectual and
physical outcome of children undernourished in early life is influenced
by later environmental conditions. Developmental Medicine and Child
Neurology, 34, 611–622.
*Cook, R., Vatev, I., Michova, Z., & Golombok, S. (1997). The European
study of assisted reproduction families: A comparison of family function
and child development between Eastern and Western Europe. Journal of
Psychosomatic Obstetrics Gynecology, 18, 203–212.
Cooper, H., & Hedges, L. V. (1994). The handbook of research synthesis.
New York: Russell Sage Foundation.
*Dalen, M. (2001). School performance among internationally adopted
children in Norway. Adoption Quarterly, 5, 39–58.
*De Jong, D. K. (2001). The well-being of Russian and Romanian inter-
country adoptees in New Zealand. Unpublished master’s thesis, Massey
University, Palmerston North, New Zealand.
*Dennis, W. (1973). Children of the creche. New York: Appleton-Century
Dumaret, A. (1985). IQ, scholastic performance and behaviour of sibs
raised in contrasting environments. Journal of Child Psychology and
Psychiatry, 26, 553–580.
Duval, S., & Tweedie, R. (2000a). A nonparametric “trim and fill” method
of accounting for publication bias in meta-analysis. Journal of the
American Statistical Association, 95, 89–98.
Duval, S., & Tweedie, R. (2000b). Trim and fill: A simple funnel-plot-
based method of testing and adjusting for publication bias in meta-
analysis. Biometrics, 56, 455–463.
*Duyme, M. (1988). School success and social class: An adoption study.
Developmental Psychology, 24, 203–209.
Earls, F. (1987). Sex differences in psychiatric disorders: Origins and
developmental influences. Psychiatric Developments, 1, 1–23.
Eme, R. F. (1979). Sex differences in childhood psychopathology: A
review. Psychological Bulletin, 86, 574–595.
*Fan, X., Miller, B. C., Christensen, M., Park, K. E., Grotevant, H. D., van
Dulmen, M., et al. (2002). Questionnaire and interview inconsistencies
exaggerated differences between adopted and non-adopted adolescents
in a national sample. Adoption Quarterly, 6, 7–27.
*Feigelman, W. (1997). Adopted adults: Comparisons with persons raised
in conventional families. Marriage and Family Review, 25, 199–223.
*Fisch, R. O., Bilek, M. K., Deinard, A. S., & Chang, P. N. (1976).
Growth, behavioral, and psychologic measurements of adopted children:
The influences of genetic and socioeconomic factors in a prospective
study. Journal of Pediatrics, 89, 494–500.
Flynn, J. R. (1987). Massive IQ gains in 14 nations: What IQ tests really
measure. Psychological Bulletin, 101, 171–191.
*Frydman, M., & Lynn, R. (1989). The intelligence of Korean children
adopted in Belgium. Personality and Individual Differences, 10, 1323–
Fulker, D. W., DeFries, J. C., & Plomin, R. (1988). Genetic influence on
general mental ability increases between infancy and middle childhood.
Nature, 336, 767–769.
*Gardner, D. B., Hawkes, G. R., & Burchinal, L. G. (1961). Noncontinu-
ous mothering in infancy and development in later childhood. Child
Development, 32, 225–234.
*Geerars, H., Hoksbergen, R. A. C., & Rooda, J. (1995). Geadopteerden op
weg naar volwassenheid: De integratie van 68 Thaise jongeren in de
Nederlandse samenleving [Adopted adolescents: The integration of 68
Thai adolescents into Dutch society]. Utrecht, the Netherlands: Utrecht
Gilbody, S. M., Song, F. J., Eastwood, A. J., & Sutton, A. (2000). The
causes, consequences and detection of publication bias in psychiatry.
Acta Psychiatrica Scandinavica, 102, 241–249.
Goldstein, H., & Healy, M. J. R. (1995). The graphical presentation of a
collection of means. Journal of the Royal Statistical Society, Series A
(Statistics in Society), 158, 175–177.
Grotevant, H. D., McRoy, R. G., & Jenkins, V. Y. (1988). Emotionally
disturbed adopted adolescents: Early patterns of family adaptation. Fam-
ily Process, 27, 439–457.
Groze, V. (1996). A 1 and 2 year follow-up study of adoptive families and
special needs children. Children and Youth Services Review, 18, 57–82.
Hakimi-Manesh, Y., Mojdehi, H., & Tashakkori, A. (1984). Effects of
environmental enrichment on the mental and psychomotor development
of orphanage children. Journal of Child Psychology and Psychiatry, 25,
Hampel, F. R., Ronchetti, E. M., Rousseeuw, P. J., & Stahel, W. A. (1986).
Robust statistics: The approach based on influence functions. New
Haugaard, J. J. (1998). Is adoption a risk factor for the development of
adjustment problems? Clinical Psychology Review, 18, 47–69.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
San Diego, CA: Academic Press.
Hodges, J., & Tizard, B. (1989). IQ and behavioural adjustment of ex-
institutional adolescents. Journal of Child Psychology and Psychiatry
and Allied Disciplines, 30, 53–75.
*Hoopes, J. L. (1982). Prediction in child development: A longitudinal
study of adoptive and nonadoptive families. New York: Child Welfare
League of America.
*Hoopes, J. L., Sherman, E. A., Lawder, E. A., Andrews, R. G., & Lower,
K. D. (1970). A follow-up study of adoptions: Vol. 2. Post-placement
functioning of adopted children. New York: Child Welfare League of
*Horn, J. M., Loehlin, J. C., & Willerman, L. (1979). Intellectual resem-
blance among adoptive and biological relatives: The Texas Adoption
Project. Behavior Genetics, 9, 177–207.
Howe, D. (1998). Patterns of adoption. Nature, nurture and psychosocial
development. Oxford, England: Blackwell.
Jerome, L. (1993). A comparison of the demography, clinical profile and
treatment of adopted and nonadopted children at a children’s mental
health centre. Canadian Journal of Psychiatry, 38, 290–294.
Johnson, D. (2000). Medical and developmental sequelae of early child-
hood institutionalization in Eastern European adoptees. In C. A. Nelson
(Ed.), The Minnesota Symposia on Child Psychology: Vol. 31. The
effects of early adversity on neurobehavioral development (pp. 113–
162). Hillsdale, NJ: Erlbaum.
Juffer, F., Stams, G. J. J. M., & Van IJzendoorn, M. H. (2004). Adopted
children’s problem behavior is significantly related to their ego resil-
iency, ego control, and sociometric status. Journal of Child Psychology
and Psychiatry, 45, 697–706.
Kaler, S. R., & Freeman, B. J. (1994). Analysis of environmental depri-
vation: Cognitive and social development in Romanian orphans. Journal
of Child Psychology and Psychiatry, 35, 769–781.
Kim, T. I., Shin, Y. H., & White-Traut, R. C. (2003). Multisensory
intervention improves physical growth and illness rates in Korean or-
phaned newborn infants. Research in Nursing & Health, 26, 424–433.
*Kim, W. J., Shin, Y. J., & Carey, M. P. (1999). Comparison of Korean-
American adoptees and biological children of their adoptive parents: A
pilot study. Child Psychiatry and Human Development, 29, 221–228.
*Kim, W. J., Zrull, J. P., Davenport, C. W., & Weaver, M. (1992).
Characteristics of adopted juvenile delinquents. Journal of the American
Academy of Child & Adolescent Psychiatry, 31, 525–532.
VAN IJZENDOORN, JUFFER, AND KLEIN POELHUIS
*Lansford, J. E., Ceballo, R., Abbey, A., & Stewart, A. J. (2001). Does
family structure matter? A comparison of adoptive, two-parent biolog-
ical, single-mother, stepfather, and stepmother households. Journal of
Marriage and the Family, 63, 840–851.
Lanz, M., Iafrate, R., Rosnati, R., & Scabini, E. (1999). Parent-child
communication and adolescent self-esteem in separated, intercountry
adoptive and intact non-adoptive families. Journal of Adolescence, 22,
*Leahy, A. M. (1935). A study of adopted children as a method of
investigating nature-nurture. Journal of the American Statistical Asso-
ciation, 30, 281–287.
Leon, I. G. (2002). Adoption losses: Naturally occurring or socially con-
structed? Child Development, 73, 652–663.
*Levy-Shiff, R., Zoran, N., & Shulman, S. (1997). International and
domestic adoption: Child, parents, and family adjustment. International
Journal of Behavioral Development, 20, 109–129.
*Lien, N. M., Meyer, K. K., & Winick, M. (1977). Early malnutrition and
“late” adoption: A study of their effects on the development of Korean
orphans adopted into American families. American Journal Clinical
Nutrition, 30, 1734–1739.
*Lipman, E. L., Offord, D. R., Racine, Y. A., & Boyle, M. H. (1992).
Psychiatric disorders in adopted children: A profile from the Ontario
Child Health Study. Canadian Journal of Psychiatry, 37, 627–633.
Loebstein, R., & Koren, G. (1997). Pregnancy outcome and neurodevel-
opment of children exposed in utero to psychoactive drugs: The
Motherisk experience. Journal of Psychiatry and Neuroscience, 22,
Main, M. (1999). Epilogue: Attachment theory: Eighteen points with
suggestions for future studies. In J. Cassidy & P. R. Shaver (Eds.),
Handbook of attachment: Theory, research, and clinical applications
(pp. 845–887). New York: Guilford Press.
*Maughan, B., Collishaw, S., & Pickles, A. (1998). School achievement
and adult qualifications among adoptees: A longitudinal study. Journal
of Child Psychology and Psychiatry and Allied Disciplines, 39, 669–
McGue, M., Bouchard, T. J., Jr., Iacono, W. G., & Lykken, D. T. (1993).
Behavior genetics of cognitive ability: A life-span perspective. In R.
Plomin & G. E. McClearn (Eds.), Nature, nurture, and psychology (pp.
59–76). Washington, DC: American Psychological Association.
*McGuinness, T., & Pallansch, L. (2000). Competence of children adopted
from the former Soviet Union. Family Relations, 49, 457–464.
Moe, V. (2002). Foster-placed and adopted children exposed in utero to
opiates and other substances: Prediction and outcome at four and a half
years. Journal of Developmental and Behavioral Pediatrics, 23, 330–
*Moore, E. G. (1986). Family socialization and the IQ test performance of
traditionally and transracially adopted Black children. Developmental
Psychology, 22, 317–326.
Morison, S. J., Ames, E. W., & Chisholm, K. (1995). The development of
children adopted from Romanian orphanages. Merrill-Palmer Quarterly,
*Morison, S. J., & Ellwood, A. L. (2000). Resiliency in the aftermath of
deprivation: A second look at the development of Romanian orphanage
children. Merrill-Palmer Quarterly, 46, 717–737.
Mullen, B. (1989). Advanced basic meta-analysis. Hillsdale, NJ: Erlbaum.
Mullen, B., & Rosenthal, R. (1985). BASIC meta-analysis: Procedures and
programs. Hillsdale, NJ: Erlbaum.
Munsinger, H. (1975). The adopted child’s IQ: A critical review. Psycho-
logical Bulletin, 82, 623–659.
*Neiss, M., & Rowe, D. C. (2000). Parental education and child’s verbal
IQ in adoptive and biological families in the National Longitudinal
Study of Adolescent Health. Behavior Genetics, 30, 487–495.
*O’Connor, T. G., Rutter, M., Beckett, C., Keaveney, L., Kreppner, J. M.,
& English and Romanian Adoptees Study Team. (2000). The effects of
global severe privation on cognitive competence: Extension and longi-
tudinal follow-up. Child Development, 71, 376–390.
*Palacios, J., & Sanchez, Y. (1996). Nin ˜os adoptados y no adoptados: Un
estudio comparativo [Adopted and nonadopted children: A comparative
study]. Anuario de Psicologia, 71, 63–85.
*Pinderhughes, E. E. (1998). Short term placement outcomes for children
adopted after age five. Children and Youth Services Review, 20, 223–
*Plomin, R., & DeFries, J. C. (1985). Origins of individual differences in
infancy. Orlando, FL: Academic Press.
Plomin, R., Fulker, D. W., Corley, R., & DeFries, J. C. (1997). Nature,
nurture, and cognitive development from 1 to 16 years: A parent-
offspring adoption study. Psychological Science, 8, 442–447.
*Priel, B., Melamed-Hass, S., Besser, A., & Kantor, B. (2000). Adjustment
among adopted children: The role of maternal self-reflectiveness. Family
Relations, 49, 389–396.
Raudenbush, S. W. (1994). Random effects models. In H. Cooper & L. V.
Hedges (Eds.), The handbook of research synthesis (pp. 301–321). New
York: Russell Sage Foundation.
Rosenthal, J. A., Groze, V., & Aguilar, G. D. (1991). Adoption outcomes
for children with handicaps. Child Welfare, 70, 623–636.
Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bul-
letin, 118, 183–192.
*Rosenwald, T. (1995). Intercountry adoptive families in Western Austra-
lia; the well-being of their four to sixteen year-old adoptees. Unpub-
lished manuscript, Edith Cowan University, Joondalup, Western Aus-
Rutter, M. (1990). Psychosocial resilience and protective mechanisms. In J.
Rolf, A. S. Masten, D. Cicchetti, K. H. Nuechterlein, & S. Weintraub
(Eds.), Risk and protective factors in the development of psychopathol-
ogy (pp. 181–214). Cambridge, England: Cambridge University Press.
Rutter, M., Andersen-Wood, L., Becket, C., Bredenkamp, D., Castle, J.,
Dunn, J., et al. (1998). Developmental catch-up, and deficit, following
adoption after severe global early privation. Journal of Child Psychology
and Psychiatry, 39, 465–476.
Rutter, M. L., Kreppner, J. M., & O’Connor, T. G. (2001). Specificity and
heterogeneity in children’s responses to profound institutional privation.
British Journal of Psychiatry, 179, 97–103.
Rutter, M., & O’Connor, T. G. (2004). Are there biological programming
effects for psychological development? Findings from a study of Ro-
manian adoptees. Developmental Psychology, 40, 81–94.
Rutter, M., O’Connor, T. G., Beckett, C., Castle, J., Croft, C., Dunn, J., et
al. (2000). Recovery and deficits following profound early deprivation.
In P. Selman (Ed.), Intercountry adoption: Developments, trends and
perspectives (pp. 107–125). London: British Agencies for Adoption and
Rutter, M., & Sroufe, L. A. (2000). Developmental psychopathology:
Concepts and challenges. Development and Psychopathology, 12, 265–
*Scarr, S., & Weinberg, R. A. (1976). IQ test performance of Black
children adopted by White families. American Psychologist, 31, 726–
Schaffer, H. R. (1998). Making decisions about children (2nd ed.). Oxford,
*Schiff, M., Duyme, M., Dumaret, A., Stewart, J., Tomkiewicz, S., &
Feingold, J. (1978, June 30). Intellectual status of working-class children
adopted early into upper-middle-class families. Science, 200, 1503–
Schiff, M., Duyme, M., Dumaret, A., & Tomkiewicz, S. (1982). How much
could we boost scholastic achievement and IQ scores? A direct answer
from a French adoption study. Cognition, 12, 165–196.
*Segal, N. L. (1997). Same-age unrelated siblings: A unique test of
within-family environmental influences on IQ similarity. Journal of
Educational Psychology, 89, 381–390.
ADOPTION AND COGNITIVE DEVELOPMENT