Psychological Bulletin
2010, Vol. 136, No. 3, 341–343
© 2010 American Psychological Association
0033-2909/10/$12.00 DOI: 10.1037/a0019116
Are There Shared Environmental Influences on
Attention-Deficit/Hyperactivity Disorder? Reply to Wood,
Buitelaar, Rijsdijk, Asherson, and Kunsti (2010)
S. Alexandra Burt
Michigan State University
A recent large-scale meta-analysis of twin and adoption studies indicated that shared environmental
influences make important contributions to most forms of child and adolescent psychopathology (Burt,
2009b). The sole exception to this robust pattern of results was observed for attention-deficit/
hyperactivity disorder (ADHD), which appeared to be largely genetic (and particularly nonadditive
genetic) in origin, with no observable influence of the shared environment. The central thesis of Wood,
Buitelaar, Rijsdijk, Asherson, and Kunsti (2010) is that, contrary to these findings, shared environmental
influences are important for ADHD. As evidence for this thesis, Wood et al. presented a summary of prior
twin studies, followed by a discussion of 4 methodological issues that may account for my findings in
Burt (2009b). I argue that, although the methodological concerns raised by Wood et al. are very
important, they do not undermine my earlier results (Burt, 2009b). I close with a discussion of 2 issues
that may allow for some shared environmental influences on ADHD.
Keywords: attention-deficit/hyperactivity disorder, shared environment, meta-analysis
Wood et al. (2010) summarized 22 child and adolescent twin
studies of ADHD published during the past 10 years. They averaged available genetic and shared environmental parameter estimates across the 100 subpopulations examined within these 22
studies (averages were 61.8% and 22.4%, respectively), and they
used these averages as the basis for their discussion. Although this
straightforward approach to summarizing twin study results is in
some ways intuitive, there are methodological limitations that
reduce confidence in such results. First, and as noted by Wood et
al., 84% of the subpopulations showed no evidence of (significant)
shared environmental influences. Moreover, these investigations
frequently reported results from best fitting reduced models that
omitted C because of lack of significance. This absence of any
estimate of C from such a large proportion of studies makes it
difficult to interpret averages of the available parameter estimates.
Second, analyses were unweighted by sample size (weighting
places more value on larger samples, which should have more
reliable estimates) and treated nonindependent data (e.g., mother
and teacher reports on the same twins) as though they were
independent, both of which are rather strongly advised against
(Lipsey & Wilson, 2001).
I sought to circumvent these difficulties in Burt (2009b) by
analyzing intraclass correlations, which are necessarily weighted
by sample size in the analyses (Neale, Boker, Xie, & Maes, 2003)
and which avoid the issue of missing parameter estimates from
best fitting models. I also omitted nonindependent samples and
used weighted averages to adjust for any nonindependence that
remained in the data after omitting these samples. I was thus able
to include all available (and nonindependent) data in the analysis.
When taking these more rigorous analytical steps, results were
simply not consistent with the presence of shared environmental
influences on ADHD (see Burt, 2009b).
One often-cited conclusion from behavioral genetics research
has been that the crucial environmental influences result in differences between siblings (i.e., referred to as nonshared effects; E),
whereas environmental influences that create similarities between
siblings (i.e., shared effects; C) are indistinguishable from zero
(Plomin & Daniels, 1987; Turkheimer, 2000). Although this finding does appear to hold in adulthood, a recent meta-analysis of
twin and adoption studies (n ! 490) indicated that C makes
important contributions to most forms of psychopathology in
childhood and adolescence (Burt, 2009b). Analyses revealed that
shared environmental influences generally accounted for 10%–
30% of the variance within conduct disorder, oppositional defiant
disorder, anxiety, depression, and broad internalizing and externalizing spectrum disorders. Furthermore, these estimates did not
vary across twin and adoption studies, suggesting that shared
environmental influences reflect actual environmental factors
common to siblings. The only exception to this robust pattern of
results was for attention-deficit/hyperactivity disorder (ADHD),
which instead appeared to be largely genetic (and particularly
nonadditive genetic) in origin (i.e., 26% additive genetic and 44%
dominant genetic), with no significant, observable influence of the
shared environment. The central thesis of Wood, Buitelaar, Rijsdijk, Asherson, and Kunsti (2010) is that, contrary to these findings, shared environmental influences are important for ADHD.
As evidence for this thesis, Wood et al. first presented a summary
of prior twin studies of ADHD, followed by a discussion of four
methodological issues that may account for the findings of Burt
(2009b). I begin with a discussion of the summary.
Correspondence concerning this article should be addressed to S. Alexandra Burt, Department of Psychology, Michigan State University, 107D,
Psychology Building, East Lansing, MI 48823. E-mail: burts@msu.edu
341
342
BURT
Wood et al. (2010) then highlighted four key issues that may
hamper the ability to detect shared environmental influences in the
classical twin study design and thus in Burt (2009b). I review each
of these in turn.
Issue 1
Wood et al. (2010) noted that the power to detect C is low in
traditional twin study designs. This is quite true (see Martin,
Eaves, Kearsey, & Davies, 1978) and, indeed, is a point I made
myself in Burt (2009b) when arguing the merits of the shared
environment. However, the large-scale meta-analyses performed
in Burt (2009b) were not limited by these sorts of power constraints. Even after adjusting for nonindependence, my analysis of
ADHD was based on 25,712 sibling pairs, an extremely large
sample by any standard and one with enough power to detect even
very small estimates of C. Indeed, power analyses suggest that I
had more than 80% power to detect C estimates as low as 5%,
given a 60% heritability estimate.
Issue 2
Wood et al. (2010) correctly stated that shared environmental
influences are confounded with dominant genetic influences in the
classical twin design (Keller & Coventry, 2005; Keller & Medland, 2008). As discussed in Burt (2009b), recent simulation work
by Keller and Medland (2008) revealed that the simultaneous
presence of shared environmental and nonadditive genetic influences (a biologically plausible scenario) led to notable underestimates of shared environmental effects within the classical twin
design (i.e., the true estimate of C was 15%, whereas the observed
estimate was 2%). It is therefore possible that the shared environment does influence ADHD but that these effects are obscured by
the simultaneous presence of nonadditive genetic effects.
Fortunately, the examination of adoptive sibling pairs circumvents this otherwise rather significant potential confound in classical twin studies. In particular, adoptive siblings do not share
any segregating genetic material; thus, any similarity between
adoptive siblings functions as a direct estimate of C. In other
words, should shared environmental influences in adoption studies
be equivalent to those in twin studies, it would argue against this
sort of underestimate of C (as well as other limitations of twin
studies; see Burt, 2009b). There was one adoption study (van den
Oord, Boomsma, & Verhulst, 1994) included in the Burt (2009b)
ADHD analyses (of two available but nonindependent adoption
studies). Because the ADE (i.e., additive genetic, dominant genetic, and nonshared environmental) model emerged as the better
fitting model, I did not originally compare estimates of the shared
environment across twin and adoption studies. However, because
the point raised by Wood et al. (2010) is extremely important, I
now show results of this analysis.
Using the analytic procedures outlined in Burt (2009b), I compared shared environmental influences on ADHD across twin and
adoption studies. Results are presented in Table 1. Genetic effects
emerged as moderate to strong and as significant predictors of
ADHD regardless of study type, as did nonshared environmental
influences. By contrast, shared environmental influences were near
zero (or zero) and nonsignificant in both twin and adoption studies.
Moreover, the shared environmental parameter could be con-
Table 1
ACE Parameter Estimates for ADHD Across Study Type
and Informant
Study type and informant
A
C
E
Adoption study
Twin study
.49! [.04, .90] .04 [.00, .18] .47! [.14, .86]
.70! [.68, .72] .00 [.00, .002] .31! [.30, .33]
Teacher informant report
Child self-report
Peer informant report
.72! [.67, .75] .00 [.00, .04]
.34! [.21, .40] .01 [.00, .12]
.71! [.47, .84] .00 [.00, .21]
.28! [.27, .29]
.64! [.60, .69]
.29! [.24, .36]
Note. 95% confidence intervals for each estimate are reported in brackets. A ! genetic effects; C ! shared environmental effects; E ! nonshared
environmental influences; ADHD ! attention-deficit/hyperactivity disorder.
!
p $ .05, estimate is significantly greater than zero.
strained to be equal across twin and adoption studies without a
significant decrement in fit, "#2(1) ! 0.395, p ! .53. In short,
there was virtually no evidence of shared environmental influence
on ADHD in either twin or adoption studies, arguing against the
notion that methodological limitations of classical twin studies
hampered my ability to identify this effect (Burt, 2009b).
As with any single study, however, it is possible that van den
Oord et al.’s (1994) study yielded peculiar estimates. To evaluate
this possibility, I computed a weighted average of the 44 male–
male, 48 female–female, and 129 male–female adoptive sibling
correlations for each of the six psychopathological conditions
evaluated in van den Oord et al.’s study. For externalizing, internalizing, and anxiety problems, the average adoptive sibling correlations were .18, .34, and .24, respectively, results that are
consistent with moderate shared environmental influences (i.e.,
18%–34%) on those phenotypes. Such results are notably consistent with the overall results of Burt (2009b). For conduct problems,
average correlations were computed separately for the aggressive
and rule-breaking behavioral subtypes. Consistent with findings
from a separate meta-analysis of conduct problems data (Burt,
2009a), shared environmental influences appeared to be more
prominent for rule-breaking than aggressive conduct problems
(average adoptive sibling correlations were .19 and .08, respectively). By contrast, the average adoptive sibling correlation for
attention problems was quite small (weighted r ! .04). Such
findings are consistent with the relative absence of shared environmental influences on ADHD (and aggressive conduct problems) but not on other psychological disorders in this particular
adoption sample.
Issue 3
Wood et al. (2010) next noted that the role of contrast effects
(i.e., when parents overemphasize differences in hyperactivity in
their dizygotic twins) must be considered, since rater contrast can
mimic nonadditive genetic effects by accentuating the difference
in similarity between monozygotic and dizygotic twins (and, in
this way, inhibit researchers’ ability to detect shared environmental
influences). Although theoretically possible as an explanation for
the lack of C observed for parent reports, it is unclear why this
phenomenon would apply only to ADHD and not to other child
behavioral problems, particularly given the high levels of comor-
SHARED ENVIRONMENTAL INFLUENCES ON ADHD: REPLY
bidity between ADHD and other child externalizing disorders.
Regardless, because rater contrast effects are generally not thought
to influence other informant reports (as discussed in
Rietveld, Hudziak, Bartels, van Beijsterveldt, & Boomsma, 2003),
I sought to evaluate the presence of shared environmental effects
in the absence of possible rater contrast by computing estimates of
C for teacher, child, and peer informants. I thus re-examined the
data in Burt (2009b) within an ACE (i.e., additive genetic, shared
environmental, and nonshared environmental) model. As seen in
Table 1, shared environmental influences were uniformly nonsignificant and were estimated to be zero or near zero. In short, rater
contrast effects also do not appear to explain the absence of shared
environmental influences that I observed (Burt, 2009b).
Issue 4
Finally, Wood et al. (2010) argued that the distributional properties of the measures must also be considered, because the presence of significant skew (as well as the use of transformations to
normalize this skew) may give rise to biased parameter estimates.
This issue of distributional properties is an interesting one that
deserves additional research. Indeed, as reported in Burt (2009b),
nonshared environmental estimates were uniformly higher (and
were so at the expense of genetic and/or shared environmental
estimates) when disorders were assessed through a diagnostic
interview as compared with a questionnaire (the former of which
is far more skewed than the latter). In short, there is evidence to
suggest that assessment strategy exerts a profound effect on genetic and environmental parameter estimates. It is thus possible
that, as argued by Wood et al. (2010), shared environmental
influences on ADHD have been artifactually suppressed by these
distributional issues. It is worth noting, however, that shared
environmental influences were estimated to be zero for questionnaire measures of ADHD as well as for diagnostic interviews.
Moreover, it is unclear why shared environmental influences
would be suppressed on ADHD questionnaires but not on questionnaires for other disorders, given that the level of skew is
similar across constructs.
Could Shared Environmental Effects
Influence ADHD?
Although I would argue that the concerns I have discussed here
do not undermine the results of Burt (2009b), there may yet be
some room for shared environmental influences on ADHD. First,
although it would have been statistically preferable to examine
variance– covariance matrices, the broad scope of my analyses
(Burt, 2009b) meant that I was logistically limited to examining
only published data (i.e., intraclass correlations). This difference is
not a trivial one. In their meta-analysis of antisocial behavior, Rhee
and Waldman (2002) found that the exclusion of weight matrices
from the analyses caused them to systematically underestimate
shared environmental influences in particular (i.e., from 28% to
17%). Although this limitation augments the importance of the
shared environmental influences identified in my research (Burt,
2009b), it may also mean that (small) shared environmental influ-
343
ences on ADHD have been obscured. Future research should seek
to use variance– covariance matrices to meta-analyze twin and
adoption studies on ADHD.
Second, Wood et al. (2010) noted that questionnaires designed
to assess the presence of both psychopathological and positive
behaviors were more likely to yield significant estimates of the
shared environment, as were mechanical measures (e.g., actigraphs). Although the former were included in my meta-analysis
(Burt, 2009b), the latter were not (because they measure only
activity level and thus are not clear measures of psychopathology).
However, should these instruments ultimately prove to be more
valid measures of ADHD than are other (more highly skewed)
instruments (e.g., diagnostic interviews, questionnaires), conclusions regarding the absence of shared environmental influences on
ADHD may need to be reconsidered.
References
Burt, S. A. (2009a). Are there meaningful etiological differences within
antisocial behavior? Results of a meta-analysis. Clinical Psychology
Review, 29, 163–178.
Burt, S. A. (2009b). Rethinking environmental contributions to child and
adolescent psychopathology: A meta-analysis of shared environmental
influences. Psychological Bulletin, 135, 608 – 637.
Keller, M. C., & Coventry, W. L. (2005). Quantifying and addressing
parameter indeterminacy in the classical twin design. Twin Research and
Human Genetics, 8, 201–213.
Keller, M. C., & Medland, S. E. (2008, June). Evaluation of the cascade
model—A new extended twin family model. Paper presented at the
conference of the Behavior Genetics Association, Louisville, Kentucky.
Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand
Oaks, CA: Sage.
Martin, N. G., Eaves, L. J., Kearsey, M. J., & Davies, P. (1978). The power
of the classical twin study. Heredity, 40, 97–116.
Neale, M. C., Boker, S. M., Xie, G., & Maes, H. H. (2003). Mx: Statistical
Modeling (6th ed.) [Computer software]. Richmond, VA: Department of
Psychiatry, Virginia Commonwealth University.
Plomin, R., & Daniels, D. (1987). Why are children in the same family so
different from one another? Behavioral and Brain Sciences, 10, 1– 60.
Rhee, S., & Waldman, I. D. (2002). Genetic and environmental influences
on antisocial behavior: A meta-analysis of twin and adoption studies.
Psychological Bulletin, 128, 490 –529.
Rietveld, M. J. H., Hudziak, J. J., Bartels, M., van Beijsterveldt, C. E. M.,
& Boomsma, D. I. (2003). Heritability of attention problems in children:
I. Cross-sectional results from a study of twins, age 3–12 years. American Journal of Medical Genetics, 117, 102–113.
Turkheimer, E. (2000). Three laws of behavior genetics and what they
mean. Current Directions in Psychological Science, 13, 160 –164.
van den Oord, E. J. C. G., Boomsma, D. I., & Verhulst, F. C. (1994). A
study of problem behaviors in 10- to 15-year-old biologically related and
unrelated international adoptees. Behavior Genetics, 24, 193–205.
Wood, A. C., Buitelaar, J., Rijsdijk, F., Asherson, P., & Kunsti, J. (2010).
Rethinking shared environment as a source of variance underlying
attention-deficit/hyperactivity disorder symptoms: Comment on Burt
(2009). Psychological Bulletin, 136, 331–340.
Received January 26, 2010
Revision received January 26, 2010
Accepted January 28, 2010 !