IN REVIEW

Why Are Children in the Same Family So Different? NonsharedEnvironment a Decade Later

Robert Plomin, PhD', Kathryn Asbury, MA, PG Dip2, Judith Dunn,

Objective: To review recent developments in the study of nonshared environment; that is, the environmental influences that makechildren growing up in the same family different, rather than similar.

Method: We review several recent influential books and papers on the subject of nonshared environment from the decadefoUow-ing the 1987 paper that highlighted its importance in psychological development.

Results: Modest progress has been made toward identifying the specific aspects of the environment responsible for nonshared en-vironment. A Ithough parents treat their multiple children differently, such differential treatment accounts for only a small amountof nonshared environmental influence, once genetic factors are controlled. It has been suggested that some degree of nonsharedenvironment may be due to the fact that siblings react differently to ostensibly shared environmental influences. Peer influenceand other experiences outside the family may be more important sources of systematic nonshared environment.

Conclusions: Despite the difficulties encountered in identifying specific sources of nonshared environment, the fact remains thatmost environmental variance affecting the development of psychological dimensions and psychiatric disorders is not shared bychildren growing up in the same family. More research and theory are needed to explain why such siblings are so different.Chance, in the sense of idiosyncratic experiences, also needs to be considered.

(Can J Psychiatry 2001;46:225-233)

Key Words: nonshared environment, families, development, behavioural genetics, twins, adoption

One of the most important discoveries arising frombehavioural-genetic research concerns nurture rather

than nature. Behavioural-genetic research provides the bestavailable evidence for the importance of environmental influ-ences, but it shows that the environment works in a surprisingway. Although socialization theories assume that the envi-ronment of the family unit influences psychological develop-ment, with the result that shared experiences lead tosimilarities among siblings, behavioural-genetics researchconsistently indicates that children growing up in the same

Manuscript received and accepted December 2000.'Research Professor, Social, Genetic and Developmental Psychiatty Re-search Centre, Institute of Psychiatry, King's College London, London,UK.^Research Assistant, Social, Genetic and Developmental Psyehiatry Re-search Centre, Institute of Psychiatry, King's College London, London,UK,•'Research Professor, Social, Genetic and Developmental Psychiatry Re-search Centre, Institute of Psyehiatry, King's College London, London,UK,Address for correspondence: Dr R Plomin, Research Professor, Social, Ge-netie and Developmental Psychiatry Research Centre, Institute of Psyehia-try, King's College London, DeCrespigny Park, Denmark Hill, LondonSE5 8AF, UK,e-mail: r,plomin@iop.kel,ac,uk

family do not share effective environmental influences. It isnot shared experiences, but shared genetics, that make sib-lings resemble one another. Indeed, with regard to psycho-logical development, environment makes siblings no moresimilar to one another than to children picked at random fromthe general population. This phenomenon—the fact that ef-fective environments are not shared—has been termed "non-shared environment,"

Our goal is to review recent developments in this field andpoint to new directions for research. We begin with a briefoverview of the phenomenon of nonshared environment.

The Phenomenon

Nonshared environment lay hidden within behavioural-genetic studies since such studies began 80 years ago. Twinstudies and adoption studies were designed to disentangle theelements of nature and nurture in familial resemblance. Forexample, schizophrenia runs in families, and environmental-ists had assumed that it does so for environmental reasons:family members share the same schizophrenogenic environ-ment. Twin studies and adoption studies test this assumptionby studying family members who share nature or nurture tovarying degrees. Although most relevant evidence has come

Can J Psychiatry, Voi 46, April 2001 225

226 The Canadian Journal of Psychiatry Voi 46, No 3

from twin studies, it is easiest to see the logic of behavioural-genetic designs by using adoption studies. Studying geneti-cally related individuals reared apart in uncorrelated (al-though not necessarily different) environments directlyestimates the extent to which familial resemblance is medi-ated genetically. These are the sort of data that convinced psy-chiatrists in the late 1960s of the importance of geneticinfluence. For example, Heston's classic adoption studyshowed that children adopted away at birth from biologicalparents with schizophrenia were nonetheless at risk forschizophrenia, suggesting genetic influence on parent-off-spring resemblance for schizophrenia (1), Indeed these chil-dren, adopted away at birth, were just as likely to suffer fromschizophrenia as are children reared by parents with schizo-phrenia, implying that growing up in a schizophrenogenic en-vironment does not add to the genetic risk for schizophrenia,A more direct test of the nurture assumption, however, lies inthe other prong of the 2-pronged adoption design—studyinggenetically unrelated individuals reared together in the sameadoptive family, which directly estimates the extent to whichfamilial resemblance is mediated by shared environment. Forschizophrenia, there is no risk from growing up in a familywith a parent or sibling who suffers from the disorder (2),

Taken together, these adoption data imply that familial re-semblance for schizophrenia is mediated genetically, not en-vironmentally. The environment is important, however. Forexample, identical twins are only 50% concordant for schizo-phrenia—not 100% concordant, which would be the case ifschizophrenia were entirely due to genetic factors. Becausethey are genetically identical, identical twins growing up inthe same family environment can only be 50% discordant forschizophrenia for environmental reasons. Such environ-mental influences have been called nonshared because theyare not shared by children growing up in the same family. Thedistinction between shared and nonshared has many other la-bels, such as "common" versus "unique," "between-family"versus "within-family," and "E2" versus "Ei," The environ-ment is important, but it does not contribute to familial resem-blance. It should be emphasized that such studies onlyaddress the contributions to resemblance among family mem-bers: their goal is to disentangle nature and nurture in familialresemblance. There may be more to schizophrenogenic fam-ily environment than is indexed by schizophrenia in the rear-ing parents. What we know from research on nonsharedenvironment, however, is that the effect of such measures ofthe family environment is to make children growing up in thesame family different from one another.

The importance of nonshared environment for personality, aswell as for psychopathology, was obvious from behavioural-genetic data since such research began a century ago, but itwas the hidden background in a figure-background illusionin which the figure was genetics. Only in 1976 was this illu-sion reversed to focus on the background. In a classic book on

a large twin study, Loehlin and Nichols reported the typicalfinding of genetic infiuence for personality, but in their con-clusions they noted

,,. a consistetit—though perplexing—pattern is emerging fromthe data (and it is not purely idiosyncratic to our stud)'). Envi-ronment carries substantial weight in deterrnining personal-ity— it appears to account for at least half the variance— butthat environment is one for which twin pairs are correlated closeto zero. , , , In short, in the personality domain we seem to seeenvironmental effects that operate almost randomly with re-spect to the sorts of variables that psychologists (and other peo-ple) have traditionally deemed important in personalitydevelopment (3, p 92),

This theme was highlighted and expanded in a 1981 paper (4),The evidence for the importance of nonshared environment inthe development of psychopathology, personality and cogni-tive abilities was brought together in a Behavioural and BrainSciences (BBS) target article called "Why are children in thesame family so different from one another?" (5), which waspublished with 32 commentaries and a response to the com-mentaries (6), Even though they were addressed in the targetBBS article and response to the commentaries, the followingissues resurfaced during the next decade: 1) nonshared envi-ronment needs to be distinguished from error of measurement(yes); 2) genotype-environment interaction and correlationcould account for nonshared environment (no, these cannotexplain why identical twins are different); 3) shared environ-ment may have more effect in extreme situations, such asthose found in abusive families (yes); 4) perceptions of envi-ronment may be an important source of nonshared experience(yes); and 5) nonshared environment may involve chance, inthe sense of idiosyncratic experiences, including prenatalevents (yes). None of the responses then or since have chal-lenged the basic finding that nonshared environment is im-portant. The message of the BBS paper was upbeat: there is anew way, and an empirical tool to study the environment.Namely, it is to study more than one child per family to findout why they are so different. Three steps were identified forthis research program: 1) document differential experiences,which requires the construction of measures of the environ-ment that are specific to each child in the family; 2) documentthe association between such differential experiences and dif-ferential outcomes; and 3) investigate the extent to which as-sociations between differential experiences and differentialoutcomes are causal.

The 1987 paper also included a table listing possible sourcesof nonshared environment (see Table 1), These are not lim-ited to the psychosocial environments that psychologists typi-cally mean when they use the word "environment,"Behavioural genetics defines "genetic" very narrowly tomean inheritance, which is what is typically meant when wesay, for example, that eye colour is genetic. In contrast, envi-ronment is defined extremely broadly to include anything thatis not genetic in this narrow sense. For example, many DNAevents (such as chromosomal anomalies or somatic mutations

April 2001 Why Are Children in the Same Family So Different? 227

that are not transmitted to offspring) are not genetic in thatthey are not inherited. Thus, this broad definition of environ-mental as nongenetic includes all noninherited intra- and in-terorganismic infiuences. While we are discussing basicissues, we should also reiterate the point that genetic researchdescribes what is rather than predicting what could be. For ex-ample, high heritability for height means that height differ-ences among individuals are largely due to geneticdifferences, given the genetic and environmental infiuencesthat exist in a particular population at a particular time (whatis). Even a highly heritable trait like height can be affected byan environmental intervention, such as improving children'sdiet or preventing illness (what could be). Such environ-mental factors are thought to be responsible for the averageincrease in height across generations, even though individualdifferences in height are highly heritable in each generation.

During the past decade, researchers have often emphasized asingle source of nonshared environment, such as family com-position (7), sibling interactions (8), peer infiuence (9), andnonsystematic factors (10), However, the list shown in Ta-ble 1 has remained unchanged, not because of the research-ers' prescience, but simply because it includes all possibletypes of nongenetic infiuence.

In one sense, thinking about environmental influences that cre-ate differences between children in the same family represents adramatic reconceptualization of psychological environments.On the other hand, this reconceptualization need not involvemysterious elements in the environment: Any environmentalfactor can be viewed in terms of its contribution to nonsharedenvirotimental variance. For example, parental affection can beeasily construed as a source of differences among children in thesame family, because parents may be more affectionate towardone child than another (5).

What was new was the focus on measures of experience spe-cific to each child, including subjective perceptions of experi-ence. As discussed later, the same event (for example,parental divorce) may be experienced differently by 2 chil-dren in the same family, and we can assess the children's dif-ferential experience of a shared event. In trying to understandwhy siblings are so different, however, it is reasonable first toinvestigate experiences that differ within families.

In 1990, the BBS paper spawned a nonacademic book calledSeparate Lives: Why Siblings Are So Different, which reinter-preted sibling socialization research along the lines of non-shared environment. It generally showed that siblingsgrowing up in the same families experience environmentsthat differ from those experienced by their parents and theircosiblings as much as the environments experienced amongsiblings growing up in different families (11), A more de-tailed review of the methods and evidence for nonshared en-vironment for the development of psychopathology,personality, and cognitive abilities was published in 1994(12), Several nonshared environment studies of siblings werelaunched in the late 1980s and early 1990s; the early ones

were described in 1994, in an edited book called Separate So-cial Worlds of Siblings: Impact of Nonshared Environmenton Development (]2). A recent metaanalysis, discussed later,identified 43 papers that have addressed the second stage ofthe nonshared environmental program of research—identify-ing associations between nonshared experiences and sib-lings' differential outcomes (10), The largest study is theNonshared Environment in Adolescent Development(NEAD) project which sprang directly from discussionsbased on the BBS paper, NEAD was a decade-long study thatinvolved a collaboration between a psychodynamic familypsychiatrist, a child psychologist specialising in families, anda behavioural geneticist. The NEAD project aimed to addressall 3 steps in the program of research listed above, focusing onmeasures of the family environment and their effect on ado-lescent psychopathology in a genetically sensitive design. Itsresults are described below.

The Reaction

Despite the revolutionary implications of realizing the impor-tance of nonshared environment, there was until recently lit-tle response from the developmental psychology community,perhaps because the responses to the 1987 BBS paper were sothorough. Nonetheless, an early-warning shot was fired in a1991 paper in Psychological Bulletin (7), The critique beganwith a general broadside criticizing behavioural-geneticmethodology, only some of which was relevant to the issue ofnonshared environment. The paper argued that, althoughsome behaviours studied (such as IQ) showed resemblancefor adoptive siblings, which suggests shared environmentalinfiuence, other behaviours (such as self-concept) whichwould show shared environmental infiuence, had not yetbeen studied. Although we agree that behavioural-geneticmethods are only quasi-experimental designs, the conver-gence of results from the different designs of the twin studyand adoption study that point to the importance of nonsharedenvironment is impressive (14). As indicated above, the easi-est way to see the importance of nonshared environ-ment—one that sidesteps most of the complexities—lies inobserving the considerable differences seen within pairs ofidentical twins reared together (even though studies of identi-cal twins probably underestimate nonshared environment be-cause there are special twin effects that contribute to theirsimilarity). Concerning the hypothesis that IQ and self-concept show shared environmental infiuence, the generalrule that effective environments are nonshared does not de-pend on showing that there are no exceptions to it. Nonethe-less, one of the most interesting results in this area involvesIQ: although younger adoptive siblings resemble each other,studies of postadolescent adoptive siblings consistently showno resemblance, suggesting that, in the long run, the environ-ment operates as nonshared environment, even for IQ(15,16), This finding suggests that in childhood we will findshared environmental infiuences that relate to cognitive

228 The Canadian Journal of Psychiatry Vol 46, No 3

development but that what we need to look for in childhoodare nonshared environmental influences that predict cogni-tive abilities in the long run (17). Conceming the issue ofself-concept, the first behavioural-genetic study of self-concept showed no evidence of shared environment (18), afinding confirmed in subsequent studies (19,20). If the cri-tique had considered psychopathology, however, it couldhave pointed to at least 1 exception to the rule that environ-mental influences are nonshared: juvenile delinquency showssome shared environmental influence (14,21), although eventhis evidence for shared environment may be artificial in thattwins may be partners in crime (22).

Despite its general attack on behavioural genetics, the cri-tique in the end agrees that environmental influences for per-sonality are nonshared:

There is basic consensus across behavioural genetics and devel-opmental psycholog)' that both genetics and environment play arole in personality developtnent and that environmental influ-ences do not result in sibling similarity (7, p 190).

The author suggests that developmental psychologists al-ready knew that the way in which the environment workedwas nonshared:

The behavioural geneticists assume that such variables, if oper-ating, would make siblings alike in personality. Environmentallyoriented developmental psychologists, on the other hand, holdthat these influences operate differenfly for different children,and thus the absence of sibling similarity' does not mean thesevariables are not having an effect (7, p 190).

If developmentalists had known about the importance of non-shared environment, why is it that far less than 1% of all de-velopmental studies of socialization studied more than asingle child per family—the only way to investigate non-shared environment?

A presidential address for the Society for Research in ChildDevelopment (23) and a double-barrelled blast of popularbooks by David Rowe (24) and Judith Harris (9) elicited adecade-delayed reaction from developmental psychologists(8,25,26). These publications provocatively spelled out theimplications of nonshared environment for theories of so-cialization; they also tackled larger issues about the nature ofnurture. The title of the Harris book is The Nurture Assump-tion, and one of the subtitles on the dust jacket is: "Parentsmatter less than you think and peers matter more." This bookattracted much media attention and was the subject of a coverstory in Newsweek, entitled "Who needs parents?" (27). Italso received glowing publicity in a lengthy and influentialreview ("Do parents matter?") that was published in The NewYorker (28). (For the extraordinary media coverage of TheNurture Assumption, see http://home.att.net/~xchar/tna/.)The negative reaction to these publications is largely directedtoward nature-and-nurture issues other than nonshared envi-ronment per se. The Harris book, for example, is criticized forarguing that peers are more important than parents (8), an

argument to which Harris has responded (29). Fueling thiscontroversy is the mistaken interpretation that, if parentingdoes not shape children's personality, parents do not matter;however, what makes siblings similar or different is a muchnarrower issue (30).

One issue specific to nonshared environment that continuesto be raised is that estimates of nonshared environment are re-duced when measurement error is taken into account (26,31).The need to separate nonshared environment from measure-ment error was emphasized in the 1987 BBS paper, and thepoint is this: environmental variance (whether a proportion oftotal variance or variance corrected for measurement error) isnot shared—the effective environments are nonshared. An-other criticism is that adoptive families have restricted vari-ance and thus underestimate the effect of shared environmenton IQ (32). However, even though most families studied bydevelopmentalists do not fully represent the population, theintemal logic of the adoption method is unaffected whenadoptive families and comparison nonadoptive families havesimilar variances, as in the Colorado Adoption Project (33).This is also the case for comparisons between identical andfratemal twins. Moreover, the hypothesis of restricted vari-ance ignores crucial adoption data showing that a correlationof 0.26 was found for IQ for more than 200 pairs of adoptivesiblings at the average age of 8 years, but 10 years later thesesame siblings showed a correlation near 0.0 (34). Becauseshared environment is found for IQ in childhood but not afteradolescence in the same families, restriction of range cannotbe an important factor in finding that the long-term influenceof shared environment for IQ is negligible. A more usefulsuggestion is that more shared environment may be found inhigh-risk families, which implies that nonshared environ-ment might be more likely to be found in low-risk families(35,36). Other useful points are that nonshared environmentmight involve perinatal factors, especially for psychopathol-ogy (37) and sibling social-comparison processes, such assibling rivalry and sibling de-identification (38,39). An evo-lutionary perspective on sibling competition for parental in-vestment suggests that sibling interactions in relation toparental attention might be a useful place to look for non-shared environment (40,41).

A more general criticism concems the need to assess the envi-ronment in behavioural-genetic research:

Behaviour-genetic analyses, however, can establish that non-shared environment contributes to individual differences in adomain but cannot document the connections between objec-tively measured nonshared environmental events and develop-ment (25, p 221)

To the contrary, as discussed in the BBS paper, it is a majorstrength of quantitative genetics that its twin and adoptionstudy methods can describe the net effect of genetic and envi-ronmental influences without attempting the impossible taskof assessing specific genes and specific environments. The

April 2001 Why Are Children in the Same Family So Different? 229

next Step in the environmental program of research is to iden-tify specific aspects of the environment responsible for non-shared environment, just as the next step in the geneticprogram of research is to identify specific genes responsiblefor heritability. As difficult as it is to identify specific genesoperating in multiple-gene systems (42), it may be even moredifficult to identify specific environmental infiuences re-sponsible for nonshared environment. With genes, we at leastknow the unit of transmission and how transmission occurs,but for the environment, especially nonshared environment,we know neither,

A constructive suggestion emerging from the past decade'sliterature is that nonshared environmental effects can befound in children's differential responses to ostensibly sharedevents (26), That is, as they interact with children s character-istics, ostensibly shared events can result in nonshared envi-ronmental effects. This distinction can be seen most clearlyfor variables that are usually measured on a family-wide ba-sis, such as parental illness, education, poverty, unemploy-ment, or neighbourhood. Depending on children'scharacteristics, such as age, sex, and personality, the effect ofsuch factors could differ for children in the same family. Anexample that has been used is research showing that the effectof paternal unemployment during the Great Depression de-pended on the age and sex of the child: a shared experiencecould have nonshared effects (43), Although it seems quitelikely that such interactions could be a source of nonsharedexperience, it should be noted that the unemployment exam-ple is drawn from an analysis between families, rather thanwithin families. To test whether the association occurs withinfamilies, and thus accounts for sibling differences in develop-mental outcomes, it would be necessary to show that the ef-fects of paternal unemployment on sibling differentialoutcomes are moderated by sibling age or sex differences.Empirical research showing the importance of such interac-tions in explaining nonshared environment is needed. In try-ing to account for the differences among siblings in a family,however, it seems reasonable to begin by looking for main ef-fects of experiences that differ for siblings (for example, dif-ferential parental treatment), rather than looking forinteractions using experiences that do not differ for siblings.Also, a better strategy might be to investigate main effects ofsiblings' different perceptions of such shared experiences.

The hypothesis that shared events can have nonshared effectssometimes leads to the conclusion that "there is nothing aboutthe findings of these traditional studies that is invalidated bytheir having studied only one child" (26, p, 16), To the extentthat one is interested in finding environmental infiuences thataffect development, this conclusion is unwarranted, becausewe now know that the effective environments are nonshared.Traditional between-family associations between parentingand children's outcomes can only explain nonshared environ-ment to the extent that they can also be shown to operatewithin families, to make siblings different. Other nonshared

Table 1. Categories of environmental influences that cause childrenin the same family to differ (adapted from Plomin and Daniels, |5|)

Categories Examples

Error of measurement

Nonshared environment

Nonsystematie

Systematie

Family eomposition

Sibling interaetion

Parent-ehild relations

Extrafamilial

Test-retest reliability

Aeeidents, differential prenatal effeets,illness, trauma

Birtb order, sex differenees

Differential treatment or perceptions

Differential treatment or perceptions

Differential experienees with peers, friends,teachers, sports, otber activities andinterests, education, occupations, spouses,family life

environmental explanations of associations between parent-ing and child outcome arising from between-family studiesinclude the possibility that parenting is an effect, rather than acause, of differences in children's outcomes or that the asso-ciations are due to a third variable, genetics, as discussed be-low. The goal of nonshared environmental research is toexplain that siblings growing up in the same family are so dif-ferent because this is how the environment works.

New Research

A consensus exists that children growing up in the same fam-ily experience different environments. This is the first step inthe 3-step program of research on nonshared environment.Concerning the second step—relating differential experienceof siblings to differential outcomes—we predict that all thefactors listed in Table 1 make some contribution to nonsharedenvironment, A metaanalysis of 43 papers addressing thissecond step concluded that "measured nonshared environ-mental variables do not account for a substantial portion ofthe nonshared variability" (10, p 78), Looking at the samestudies, however, an optimist could conclude that this re-search is off to a good start. The proportion of total varianceaccounted for in adjustment, personality, and cognitive out-comes was 0.01 for family constellation, 0,02 for differentialparental behaviour, 0,02 for differential sibling interaction,and 0,05 for differential peer or teacher interaction. More-over, these effects are largely independent, because aggregatemeasures of differential environment account for 13% of thetotal variance. These studies have not yet, however, taken thethird step in the research program: the attempt to disentanglecause and effect. Earlier, we noted that it may be more diffi-cult to identify specific environmental infiuences responsiblefor nonshared environment than it is to identify specific genesresponsible for heritability. Nonetheless, one could argue thatat least as much progress has been made in identifying non-shared environment as has been made in identifying genes(14),

230 The Canadian Journal of Psychiatry Vol 46, No 3

In this section, we focus on the results of the decade-longNEAD project because it represents the leading edge of re-search on nonshared environment in attempting to address all3 steps in the research program (44). During 2 visits of 2 hourseach, made at 3-year intervals to 720 families with 2 same-sexsibling offspring ranging in age from 10 to 18 years, extensivequestionnaire and interview measures of the family environ-ment were administered to both parents and offspring, andparent-child interactions were videotaped during a session inwhich problems in family relationships were discussed. Mul-tiple measures across multiple sources of information werecombined to create highly reliable composite measures. Thefirst step was to identify differential experiences of siblings.Sibling correlations for children's reports of their family in-teractions (for example, reports of their parents' negativity)were moderate, as they were for observational ratings ofchild-to-parent, and parent-to-ehild, interactions. Becausethese measures were highly reliable, this finding suggests thatsuch experiences are largely nonshared. In contrast, parent re-ports yielded high sibling correlations (as, for example, whenparents reported on their own negativity toward each of thechildren). Although this may be due to a rater effect in that theparent rates both children, the high sibling correlations forparent reports of children's environments indicate that parentreports are not good candidates for investigating nonsharedenvironmental factors.

Once child-specific nonshared experiences are identified, thesecond step is to ask whether these nonshared experiences re-late to psychological outcomes. For example, to what extentdo differences in parental treatment account for the non-shared environmental variance known to be important forpersonality and psyehopathology? Although research in thisarea has only just begun, some success has been achieved inpredicting differences in adjustment from differences in sib-ling experiences (13). The NEAD project provides severalexamples using different models and methods to investigateassociations between siblings' differential experiences anddifferential outcomes. (A useful, simple, and general method-ology has been described by Rodgers and others [45]; see alsoTurkheimer and Waldron [10]). For instance, negative paren-tal behaviour directed specifically to one adolescent sibling(controlling for parental treatment of the other sibling) relatesstrongly to that child's antisocial behaviour and, to a lesserextent, to that child's depression (44). Most of these associa-tions involve negative aspects of parenting, such as confliet,and negative outcomes, such as antisocial behaviour. Asso-ciations are generally weaker for aspeets of positive parent-ing, such as affection.

At this point it would appear that NEAD supplied a happyending for the story of nonshared environment: parents treatchildren differently and this differential treatment relates todifferential adjustment. The issue suddenly became morecomplex, however, at the third step in this researeh

program—asking whether differential parenting is a cause oran effect of children's adjustment. Longitudinal cross-laggedanalyses were conducted but were not very informative: theparenting measures and adjustment measures were quite sta-ble over the 3-year interval in whieh NEAD families werestudied, whereas the success of cross-lagged analyses de-pends on change. A unique feature of NEAD is that its exten-sive measures of family environment were embedded in agenetieally sensitive design to assess possible genetic media-tion of associations between nonshared environment and ado-lescent psyehopathology. The 720 NEAD families wereselected to include identical and fraternal twins, full siblings,half-siblings, and genetically unrelated siblings. In thesefamilies, the twins and siblings were the adolescent children,not the parents, so that genetic mediation could be seen to theextent that differential parenting is a response to children'sgenetieally driven differential adjustment.

NEAD used multivariate genetic analysis of eovadance to ad-dress this issue (46). Unlike traditional univariate geneticanalysis of the variance of a single variable, multivariate ge-netic analysis decomposes the eovariance between variablesinto genetic and environmental sources of eovarianee (47).The basic idea is that instead of eomparing identical and fra-ternal twin correlations for a single variable in order to de-compose the variance of that variable, multivariate geneticanalysis eompares identical and fraternal twin cross-correlations—that is, the correlation between variable X forone twin and variable Y for the other twin—in order to de-compose the covariance between X and Y. Genetic mediationof the covariance is implied if the identical twin cross-correlation exceeds the fraternal twin cross-correlation (see14, for details). In addition to families with identical and fra-ternal twins, NEAD applied multivariate genetie analysis tothe covariance between family environment and adolescentoutcome for families with full siblings, half siblings and unre-lated siblings. This approach estimates genetie mediation ofthe eovarianee between family environment and adolescentoutcome, and, just as importantly, it estimates shared andnonshared environmental contributions to the eovariance, in-dependent of genetie effects. Moreover, if the standard as-sumption is made that error of measurement does notcorrelate across measures, the nonshared environmental con-tribution to the eovarianee is free of measurement error.

Multivariate genetie analysis of the associations between pa-rental negativity and adolescent adjustment yielded an unex-pected finding: most of these associations were mediated bygenetie factors. Hardly any nonshared environmental mediat-ing effeets remained independent of genetic mediation. Thefinding that nonshared environment fails to mediate these as-soeiations is supported by simpler analyses that found littleassoeiation between identical twin differences in family envi-ronment and identical twin differences in adolescent out-comes—an analytic approach that directly assesses

April 2001 Why Are Chitdren in the Same Family So Different? 231

nonshared environmental mediation of the association (48).The finding of genetic mediation implies that, to a substantialextent, differential parental treatment of siblings reflects ge-netically influenced differences between the siblings. As im-plausible as this finding might seem on first encounter, it ispart of the second great discovery of genetic research at theinterface between nature and nurture—genetics contributessubstantially to experience (49). The NEAD quest for non-shared environment led to genotype-environment correla-tion; that is, children select, modify, construct, andreconstruct their experiences in part on the basis of their ge-netic propensities. Many of the recent reactions mentionedearlier (8,25,26) focus on this larger issue of genotype-envi-ronment interplay, rather than on nonshared environment perse. The NEAD book (44) is called The Relationship Code be-cause genotype-environment correlation restores to the fam-ily some of the infiuence lost to nonshared environmentfactors, in the sense that genetic propensities are expressed inthe family environment. Genotype-environment correlationis neither a solely genetic nor a solely environment issue—itis both. This, however, is another story for another time.

The Future

The relevance ofthe NEAD results in relation to the story ofnonshared environment is that we must begin again to iden-tify nonshared environment. The traditional measures of fam-ily environment used in NEAD did not result in theidentification of nonshared environmental links with differ-ential sibling outcomes. Perhaps other measures ofthe familyenvironment will be more successful. Perhaps some sourcesof nonshared environment can still be found in NEAD, in ageand sex interactions with shared environmental factors, assuggested by recent papers.

In retrospect, however, as Harris (9) has trenchantly pointedout, it seems odd to have looked for differential experiencesof siblings solely in the family because siblings live in thesame family. In defense of NEAD, investigating the familyenvironment as a first step in the quest for nonshared environ-ment allowed NEAD to capitalize on the huge research effortthat has gone into assessing family environment. Indeed, itturns out that siblings experience very different family envi-ronments and that these differential experiences are stronglyassociated with differential outcomes. The problem is that ge-netic mediation largely accounts for these associations.

As suggested by the results of the metaanalysis mentionedearlier (10), extrafamilial factors seem to be good candidatesfor nonshared environments—candidates that have not beenstudied nearly as much as has family environment. Harris an-ticipates that research will confirm the impact of peers: al-though genetic influence is involved in choosing and beingchosen by friends and peers (50-53), a recent report involv-ing both NEAD and an adoption study found that peer groupslargely involve nonshared environment independent of

genetics, as predicted by Harris (54). Nonetheless, nonsharedenvironment appears to be present early in life, long beforechildren experience peer infiuence, which implies that non-shared environmental factors may differ from age to age—atopic to which we shall return.

No matter how difficult it may be to find specific nonsharedenvironmental factors, it should be emphasized that the effec-tive environment operates in terms of nonshared experience.More research is needed to understand these mechanisms.The price of admission to research on nonshared environmentis to study more than a single child per family. This is not dif-ficult because more than 80% of families have more than asingle child: siblings add synergistically to any research de-sign. You can study whatever you were going to study in thetraditional between-family way and then look at the data froma within-family perspective, which can in turn clarify the in-terpretation of any between-family findings (55). Any studyof siblings can be used to investigate the first 2 steps in the re-search program mentioned earlier: documenting differentialexperiences and documenting the association between differ-ential experiences and differential outcomes. Longitudinalstudies can begin to address the third step: interpreting the di-rection of effects for the association—between, for example,parenting (X) and child outcomes (Y). Does X cause Y, ordoes Y cause X, or is a third factor responsible for the associa-tion between X and Y? Behavioural-genetic studies of twinsor adoptees are useful to investigate genetics as a possiblethird factor mediating such associations.

We also need to consider the gloomy prospect that chancecontributes to nonshared environment in terms of randomnoise, idiosyncratic experiences, or the subtle interplay of aconcatenation of events (5,10,11,56-58). Francis Galton, thefounder of behavioural genetics, suggested that nonsharedenvironment is largely due to chance, when he commentedthat "the whimsical effects of chance in producing stable re-sults are common enough." Sounding a bit like a fortunecookie, he also observed that "tangled strings variouslytwitched, soon get themselves into tight knots" (59, p 195).

Two overlooked findings from behavioural-genetic researchalso point to the importance of chance. The first is multivari-ate genetic research showing that nonshared environmentalinfiuences on one trait, such as depression, are independent ofnonshared environmental infiuences on other traits, such asantisocial behaviour (44,60,61). The second finding involvesthe application of multivariate genetic analysis to longitudi-nal data. This assesses the genetic and environmental originsof age-to-age change and continuity (62). Longitudinal ge-netic analyses indicate that nonshared environmental infiu-ences are age-specific for psychopathology (63,64),personality (65-67), and cognitive abilities (68). In NEAD,although most nonshared environment was unstable, somestability was found for depression and antisocial behaviourover a 3-year interval in adolescence (44,69). That is.

232 The Canadian Journal of Psychiatry Vol 46, No 3

nonshared environmental infiuences at one age are largelydifferent from nonshared environmental infiuences at an-other age.

What environmental processes other than chance could ex-plain these 2 findings? Our view, nonetheless, is that chanceis the null hypothesis, although measures of life events can as-sess some of its aspects (70). Systematic sources of nonsharedenvironment need to be thoroughly examined before we dis-miss it as chance. Chance might only be a label for our currentinability to identify the processes by which children—evenpairs of identical twins—growing up in the same family cometo be so different. For example, to some extent children mightmake their own luck.

A larger implication of nonshared environment is that it pro-vides an excellent indication that some ofthe most importantquestions for genetic research involve the environment, andsome of the most important questions for environmental re-search involve genetics (71). Genetic research will profit if itincludes sophisticated measures ofthe environment, environ-mental research will benefit from the use of genetic designs,and understanding of development will be advanced by col-laboration between geneticists and environmentalists. Inthese ways, behavioural scientists are putting the nature-nur-ture controversy behind them and bringing nature and nurturetogether in the study of development to understand the pro-cesses by which genotypes become phenotypes.

References

1. Heston LL. Psychiatric disorders in foster home reared children of schizophrenicmothers. Br J Psychiatry 1966;! 12:819-25.

2. Kety SS, Wender PH, Jacobsen B, Ingraham LJ, Jansson L, Faber B, and others.Mental illness in the biological and adoptive relatives of schizophrenic adoptees:Replication ofthe Copenhagen study in the rest of Denmark. Arch Gen Psychiatry1994;51:442-55.

3. Loehlin JC, Nicholls J. Heredity, environment and personality. Austin: UniversityofTexas; 1976.

4. Rowe DC, Plomin R. The importance of nonshared (E1) environmental influencesin behavioral development. Dev Psychol 1981; 17:517-31.

5. Plomin R, Daniels D. Why are children in the same family so different from eachother? Behav Brain Sci 1987;IO:1-16.

6. Plomin R, Daniels D. Children in the same family are very different, but why? [re-sponse to commentaries]. Behav Brain Sci l987;10(l):44-55.

7. Hoffman LW, The influence ofthe family environment on personality: accountingfor sibling differences. Psychol Bull I991;l 10:187-203.

8. Vandell D. Parents, peers, and others. Dev Psychol 2000;699-710.9. Harris JR. The nurture assumption: why children turn out the way they do. New

York: The Free Press; 1998.10. Turkheimer E, Waldron M. Nonshared environment: a theoretical, methodologi-

cal, and quantitative review. Psychol Bull 2000;126:78-108.11. Dunn JF, Plomin R. Separate lives: why siblings are so different. New York: Basic

Books; 1990.12. Plomin R, Chipuer HM, Nciderhiser JM. Behavioural genetic evidence for the im-

portance of nonshared environment. In: Hetherington, EM, Reiss D, Plomin R, edi-tors. Separate social worlds of siblings: the impact of non-shared environment ondevelopment. Hillsdale (NJ): Lawrence Eribaum Assoc; 1994. p 1—31.

13. Hetherington EM, Reiss D, Plomin R, editors. Separate social worlds of siblings:impact of nonshared environment on development. Hillsdale {NJ): Lawrence Eri-baum Assoc; 1994.

14. Plomin R, DeFries JC, McCleam GE, McGuffin P. Behavioural Genetics. 4th ed.New York: Worth Publishers; 2001.

15. McGue M, Bouchard Jr TJ, Iacono WG, Lykken DT. Behavioural genetics of cog-nitive ability: A life-span perspective. In: Plomin R, McCleam GE, editors. Nature,nurture, and psychology. Washington (DC): American Psychological Association;1993. p 59-76.

Clinical Implications

• In thinking about environmental causes of psychopathology, clini-cians should consider nonshared environmental factors that makesiblings different.

• Nonshared environtnent is likely to involve extrafamilial experi-ences, such as peer influence, and idiosyncratic experiences, such asthose resulting from chance.

• Perceptions of experience may be an important source of nonsharedenvironment.

Limitations

• Research has not yet investigated the relative influence of shared andnonshared environment at the extremes—in abusive families, for ex-ample.

• Few specific nonshared environmental factors have as yet been iden-tified in genetically sensitive ways.

• More measures are needed of fatnily environment specific to eachchild in the family.

16. Plomin R. The nature and nurture of cognitive abilities. In: Stemberg RJ, editor.Advances in the psychology of human intelligence. Volume 4. Hillsdale (NJ): Law-rence Eribaum Assoc; 1988. p 1-33.

17. Chipuer HM, Plomin R. Using siblings to identify shared and non-shared HOMEitems. Br J Dev Psych 1992; 10:165-78.

18. McGuire S, Neiderhiser JM, Reiss D, Hetherington EM, Plomin R. Genetic and en-vironmental influences on perceptions of self-worth and competence in adoles-cence: a study of twins, full siblings, and step-siblings. Child Dev 1994;65:785-99.

19. McGuire S, Manke B, Saudino KJ, Reiss D, Hetherington EM, Plomin R. Per-ceived competence and self-worth during adolescence: a longitudinal behaviouralgenetic study. Child Dev 1999;70:1283-96.

20. Roy MA, Neale MC, Kendler KS. The genetic epidemiology of self-esteem. Br JPsychiatry. I995;166(6):8I3-2O.

21. Gottesman II, Goldsmith HH. Developmental p.sychopathology of antisocial be-havior. In: Nelson CA, editor. Threats to optimal development. Hillsdale (NJ):Lawrence Eribaum Assoc; 1994. p 69-104.

22. Rowe DC. Genetic and environmental components of antisocial pairs: A study of265 twin pairs. Criminology 1986;24:513-32.

23. Scarr S. Developmental theories for the 1990s: development and individual differ-ences. Child Dev 1992;63:1-I9.

24. Rowe DC. The limits of family influence: Genes experience, and behaviour. NewYork: Guilford Press; 1994.

25. Collins WA, Maccoby EE, Steinberg L, Hetherington EM, Bomstein M. Contem-porary research in parenting: The case for nature and nurture. Am Psychol2000;55(2):218-32.

26. Maccoby EE. Parenting and its effects on children: On reading and misreading be-haviour genetics. Annu Rev Psychol 2000;5I:l-27.

27. Begley S. Who needs parents? Newsweek 1998;August 24:53.28. Gladwell M. Do parents matter?. The New Yorker 1998;August 17:54-64.29. Harris JR. Socialization, personality development, and the child's environments.

Dev Psychol 2000;711-23.30. Plomin R. Parents and Personality. APA Review of Books I999;44(4):269-71.31. Rutter M, Silberg J, O'Connor TG, Simonoff E. Genetics and child psychiatry: I.

Advances in quantitative and molecular genetics. J Child Psychol Psychiat1999;40:3-18.

32. Stoolmiller M. Implications ofthe restrictedrangeof family environments for esti-mates of heritability and nonshared environment in behaviour-genetic adoptionstudies. Psychol Bull I999;I25:392-4O9.

33. DeFries JC, Plomin R, Fulker DW. Nature and nurture during middle childhood.Oxford (UK): Blackwell; 1994.

34. Loehlin JC, Horn JM, Willerman L. Modeling IQ change: evidence from the Texasadoption project. Child Dev I989;6O:993-IOO4.

35. O'Connor TG, Dunn JF, Jenkins JM, Pickering K, Rasbash J. Family settings andchildren's adjustment: differential adjustment within and across families. Br J Psy-chiatry [forthcoming].

36. McGuire S, Clifford J, Fink J. Parental differential treatment in different familycontexts: testing multiple risk models [manuscript submitted for publication].

37. McGue M, Bouchard Jr TJ. Genetic and environmental influences on human be-havioural differences. Annu Rev Neurosci 1998;21:1-24.

38. Fienberg ME, Hetherington EM. Sibling differentiation in adolescence: implica-tions for behavioural genetic theory. Child Dev [forthcoming].

39. McGuire S. Nonshared environment research: what is it and where is it going? Mar-riage and Family Review [forthcoming].

April 2001 Why Are Children in the Same Family So Different? 233

40. Buss DM. Evolutionary hypotheses and behavioural genetic methods: hopes for aunion of two disparate disciplines. Behav Brain Sci 1987; 10:20.

41. Lalumiere ML, Quinsey VL, Craig WM. Why children from the same family are sodifferent from one another—a Darwinian note. Human Nature 1996;7(3):281-90.

42. Plomin R, Owen MJ, McGuffm P. The genetic basis of complex human behaviours.Science 1994;264(5l66):1733-9.

43. Elder GH Jr. Children ofthe great depression. Chicago: University of ChicagoPre.ss; 1974.

44. Reiss D, Neiderhiser JM, Hetherington EM, Plomin R. The relationship code: deci-phering genetic and social patterns in adolescent development. Cambridge (MA):Harvard University Press; 2000.

45. Rodgers JL, Rowe DC, Li C. Beyond nature versus nurture: DF analysis of non-shared influences on problem behaviours. Dev Psychol 1994;30:374-84.

46. Pike A, McGuire S, Hetherington EM, Reiss D, Plomin R. Family environment andadolescent depressive symptoms and antisocial behavior: a multivariate geneticanalysis. Dev Psychol 1996;32:590-603.

47. Martin NG, Eaves LJ. The genetical analysis of covariance structure. Heredity1977;38:79-95.

48. Pike A, Rei.ss D, Hetherington EM, Plomin R. Using MZ differences in the searchfor nonshared environmental effects. J Child Psychol Psychiatry 1996;37:695-704.

49. Plomin R. Genetics and experience: the interplay between nature and nurture.Thousand Oaks (CA): Sage Publications Inc; 1994.

50. Baker LA, Daniels D. Nonshared environmental influences and personality differ-ences in adult twins. J PersSoc Psychol 1990;58:103-10.

51. Daniels D, Plomin R. Differential experience of siblings in the same family. DevPsychol 1985;21:747-60.

52. Mankc B, McGuire S, Reiss D, Hetherington EM, Plomin R. Genetic contributionsto adolescents extrafamilial social interactions: teachers, best friends, and peers.Social Development l995;4:238-56.

53. Pike A, Manke B, Reiss D, Plomin R. A genetic analysis of differential experiencesof adolescent siblings across three years. Social Development 2000;9:96-l 14.

54. lervolino AC, Pike A, Manke B, Reiss D, Hetherington EM, Plomin R. Genetic andenvironmental influences in adolescent peer socialization: evidence from two ge-netically sensitive designs Child Dev [forthcoming].

55. Dick DM, Johnson JK, Vikcn RJ, Rose RJ. Testing between-family associations inwithin-family comparisons. Psych.Sci 2000; 11:409-l 3.

56. Molenaar PCM, Boomsma DI, Dolan CV. A third source of developmental differ-ences. Behav Genet l993;6:519-24.

57. Jensen AR. The problem of genotype-environment correlation in the estimation ofheritability from monozygotic and dizygotic twins. Acta Genet Med Gemellol1974;25:86-99.

58. Jenson AR. The puzzle of nongenetic variance. In: Stemberg RJ, Grigorenko EL,editors. Intelligenee, heredity and environment. Cambridge (UK): Cambridge Uni-versity Press; 1998. p 42-88.

59. Galton F. Natural inheritance. London: Macmillan; 1889.60. Boomsma Dl. Absence or underestimation of shared environment. Behav Brain Sci

1987;IO:19.61. Van den Oord JCG, Boomsma DI, Verhulst FC. A study of genetie and environ-

mental effects on the eo-occurrence of problem behaviours in three-year-old twins.J Abnorm P.syehol 2000; 109:360-72.

62. Plomin R. Development, genetics, and psychology. Hillsdale (NJ): Lawrence Eri-baum Assoe; 1986.

63. Kendler KS, Neale MC, Kessler RC, Heath AC, Eaves LJ. A longitudinal twinstudy of personality and major depression in women. Arch Gen Psyehiatry 1993;50:853-62.

64. Van den Oord JCG, Rowe DC. Continuity and change in children's .soeial malad-justment: A developmental behaviour genetic study. Dev Psychol 1997;33:319-32.

65. Loehlin JC, Horn JM, Willerman L. Heredity, environment, and personalitychange: evidence from the Texas adoption study. J Pers I990;58:221—43.

66. McGue M, Bacon S, Lykken DT. Personality stability and change in early adult-hood: a behavioural genetic analysis. Dev Psychol I993;29:96-1O9.

67. Pogue-Geile MF, Rose RJ. Developmental genetie studies of adult personality.Dev Psychol 1985;21:547-57.

68. Cherny SS, Fulker DW, Hewitt JK. Cognitive development from infancy to middlechildhood. In: Stemberg RJ, Grigorenko EL, editors. Intelligenee, heredity and en-vironment. Cambridge (UK): Cambridge University Press; 1997. p 463-82.

69. O'Connor TG, Neiderhiser JM, Reiss D, Hetherington EM, Plomin R. Genetic con-tributions to continuity, change, and co-occurrence of antisocial and depressivesymptoms in adolescence. J Child Psychol Psychiatry l99B;39:323-36.

70. Eley TC, Stevenson J. Specific life events and ehronic experiences differentiallyassociated with depression and anxiety in young twins. J Abnorm Child2000;28(4):383-94.

71. Rutter M, Dunn JF, Plomin R, Simonoff E, Pickles A, Maughan B, and others. Inte-grating nature and nurture: implications of person-environment correlations and in-teractions for developmental psychopathology. Dev Psychopathol 1997;9:335-64.

Resume— Pourquoi les enfants d'une meme famille sont-ils si differents? Le milieu nonpartage, dix ans plus tard

Objectif: Examiner les recents developpements d'une etude du milieu non partage, c 'est-a-dire, les influences du milieu qui fontque les enfants grandlssant dans la meme famille sont differents plutot que semblables.

Methode : Nous passons en revue plusieurs livres et articles recents et influents sur le sujet du milieu non partage, dans les dixannees suivant I 'article de 1987 qui en soulignait I 'importance sur le developpement psychologique.

Resultats: On a realise desprogres modestes dans I 'identification des aspects specifiques de I 'environnement qui sont responsa-bles du milieu non partage. Bien que les parents traitent leurs multiples enfants differemment, cette difference de traitement ne re-presente qu 'une infime partie de I 'influence du milieu non partage, unefois que les facteurs genetiques sont pris en compte. Oncroit qu 'un certain degre du milieu non partage peut etre attrtbuable aufait quefreres et soeurs reagissent differemment aux in-fluences du milieu manifestement partagees. L 'influence des pairs etd'autres experiences a I exterieur de la famille peuvent etredes sources plus importantes de milieu non partage systematique.

Conclusions: Malgre les difftcultes Hies a I 'identification des sources precises du milieu non partage, it n en demeurepas moinsque la majorite de la variance de I 'environnement qui affecte le developpement des dimensions psychologiques et des troublespsychiatriques n 'est pas partagee par les enfants grandissant dans la meme famille. Des recherches et theories additionnellessont necessaires pour expliquer pourquoi freres et soeurs sont si differents. Ilfaut egalement tenir compte du hasard, au sens desexperiences idiosyncrasiques.