ebook img

Unconfounding genetic and nonshared environmental effects PDF

60 Pages·2009·7.49 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Unconfounding genetic and nonshared environmental effects

BEHAVIORAL AND BRAIN SCIENCES (1987) 10, 1-60 Printed in the United States of America Why are children in the same family so different from one another? Robert Plomin Department of Individual and Family Studies, Pennsylvania State University, University Park, Pa. 16802 Denise Daniels Department of Psychiatry, Stanford University, Stanford, Calif. 94305 Abstract: One of the most important findings that has emerged from human behavioral genetics involves the environment rather than heredity, providing the best available evidence for the importance of environmental influences on personality, psychopathology, and cognition. The research also converges on the remarkable conclusion that these environmental influences make two children i nthe same family as different from one another as are pairs of children selected randomly from the population. The theme of the target article is that environmental differences between children in the same family (called "nonshared environment") represent the major source of environmental variance for personality, psychopathology, and cognitive abilities .One example of the evidence that supports this conclusion involves correlations for pairs of adopted children reared in the same family from early in life. Because these children share family environment but not heredity, their correlation directly estimates the importance of shared family environment. For most psychological characteristics, correlations for adoptive "siblings hover near zero, which implies that the relevant environmental influences are not shared by children in the same family. Although it has been thought that cognitive abilities represent an exception to this rule, recent data suggest that environmental variance that affects IQ is also of the nonshared variety after adolescence. The article has three goals: (1) To describe quantitative genetic methods and research that lead to the conclusion that nonshared environment is responsible for most environmental variation relevant to psychological development, (2) to discuss specific nonshared environmental influences that have been studied to date, and (3) to consider relationships between nonshared environmental influences and behavioral differences between children in the same family. The reason for presenting this article in BBS is to draw attention to the far-reaching implications of finding that psychologically relevant environmental influences make children in a family different from, not similar to, one another. Keywords: behavior genetics; development; environment; heredity; individual differences; intelligence; personality: psycho- pathology; schizophrenia; twins The findings of greatest social significance to emerge from Thus, for personality, psychopathology, and cognition, human behavioral-genetic research to date involve nur- behavioral-genetic research converges on the conclusion ture, not nature. Research in this area, consisting pri- that most behavioral variability among individuals is marily of twin and adoption studies, points to significant environmental in origin. For example, for schizophrenia, genetic influence on individual differences for a wide the concordance for first-degree relatives, whose coeffi- range of behaviors, including personality, psychopath- cient of genetic relationship is .50, is less than 10%. ology, and cognition. When we go beyond the statistical Identical twins are less than 50% concordant for schizo- significance of genetic influence to ask about the effect phrenia. Yet schizophrenia is coming to be viewed as a size, it is also apparent that genetic influence is substan- genetic disease. In the rush to find neural causes of tial. Nonetheless, the same data provide evidence - indeed, schizophrenia, who is now studying the major source of we think the best available evidence - for the importance variability - the environment? of environmental variation in each of these domains. Not only does behavioral genetic research document Ten years ago, in order to redress the imbalance of the importance of environmental influence, it also points environmentalism, it was necessary to emphasize the to a possible treasure of environmental variance hidden in possibility that genetic influence could affect behavioral unexplored territory. This research implies that environ- differences that we observe among individuals. Now mental influences that affect psychological development behavioral geneticists find that they must more often operate in a manner quite different from the way most emphasize the importance of environmental variation. psychologists thought they worked. Whatever they may Behavioral-genetic research seldom finds evidence that be, these environmental influences make children in the more than half of the variance for complex behavioral same family as different from one another as are children traits is due to genetic differences among individuals. in different families. One purpose of this article is to ip) 1987 Cambridge University Press 0140-525x187 J5.00+.00 1 Plomin & Daniels: Nonshared environment describe the evidence that leads to this conclusion and its tives such as cousins will be less similar than second- implications. Our main goal, however, is to draw atten- degree relatives such as half-siblings who, in turn, will be tion to this dramatic discovery and to elicit commentary less similar than first-degree relatives such as full siblings. and suggestions from our peers. Despite the far-reaching If heredity does not affect the trait, then differences in implications of the evidence that psychologically relevant genetic similarity should not affect the resemblance of environmental influences make children in a family dif- these pairs of individuals. ferent from rather than similar to each other, we are The problem is that environmental resemblance often aware of no major criticism of these findings. We expect covaries with genetic relatedness: Cousins, half-siblings, that BBS commentary will rock this boat's smooth sailing and full siblings, respectively, are likely to share in- and perhaps even alter its course. creasingly similar environments. Because relatives share family environment as well as heredity, familial re- semblance can be due to environmental influences as well 1. Quantitative genetics as to hereditary influences. In other words, a portion of In order to understand the evidence pointing to the environmental influence could be shared by relatives, importance of nonshared environment it is necessary to making them similar to one another. Nonetheless, family begin with an overview of the theory and methods of studies are useful in estimating limits of genetic and quantitative genetics, which, when applied to behavioral environmental influences. For example, if the correlation phenomena, is referred to as behavioral genetics. After for first-degree relatives is zero for a particular trait, then describing the basic twin and adoption designs, we shall neither shared heredity nor shared family environment examine the implications of twin and adoption data for the affect the trait. separation of shared and nonshared environmental varia- The two major designs of human behavioral genetics - tion in the three domains with the most relevant data: the adoption design and the twin design - were devel- personality, psychopathology, and cognition. oped to circumvent the problem of conflating genetic and Quantitative genetic theory began in the early part of environmental influences in studies of family members this century as a solution to the problem of reconciling who share heredity and family environments. By doing Mendelian genetics with normal distributions. As anyone so, these designs partition environmental variance into two components: one shared by members of a family and who has taken high school biology knows, about a hun- the other consisting of the remainder of the environmen- dred years ago, the monk Gregor Mendel studied di- tal variance, which is referred to as nonshared envi- chotomous, either/or, characteristics such as round ver- ronment. sus wrinkled seeds in the pea plant. When his work was rediscovered 30 years later it provoked controversy among biometricians who felt that the laws of heredity 1.1. Adoption design. The basic problem in family studies described by Mendel could not apply to human charac- is that resemblance among relatives could be due to teristics because, unlike discontinuous pea plant charac- shared heredity or to shared environment. The adoption teristics, human characteristics nearly always involve a design powerfully cleaves these two sources of familial normal, continuous distribution. The resolution to the resemblance. Genetically related individuals adopted controversy came when it was understood that a normal apart and reared in uncorrelated environments will re- distribution would be observed if several genes affected a semble each other only for genetic reasons. Genetically characteristic. In 1918, when Ronald Fisher put the unrelated individuals adopted together in the same fami- finishing touches on this theory and spelled out the ly will resemble each other only for reasons of shared expectations for familial resemblance based on the theo- environment. ry, quantitative genetics was born. The simplest adoption design to understand is the rare, The theory uses the covariance or correlation among but dramatic, situation in which identical twins are relatives on normally distributed traits to estimate the adopted separately at birth and reared apart in uncorre- role of heredity. Although the theory and its methods are lated environments. The resemblance of these pairs of usually presented in a sophisticated algebraic manner, twins, expressed as a correlation, is a direct estimate of the basic idea - which is all that is needed to understand the proportion of phenotypic variance that is clue to the way in which environmental variation is partitioned in genetic variance, a descriptive statistic known as herit- quantitative genetics - is very simple. Details, such as ability. A correlation of .50 for identical twins reared the distinction between additive and nonadditive genetic apart implies that half of the phenotypic variance is variance, can be found in textbooks on the topic (e.g., genetic in origin.1 Falconer 1981; Hay 1985; Plomin, DeFries & McClearn A technical point that has some bearing on the estima- 1980). The fundamental tenet of the theory is that indi- tion of nonshared environment concerns the distinction viduals in a population differ for both genetic and non- between additive and nonadditive genetic variance. genetic reasons. How can we assess the extent to which Identical twins share all sources of genetic variance, no phenotypic (observed) variability is due to genetic varia- matter how complex the interactions among genes. Thus, tion among individuals or to nongenetic differences? In an estimate of heritabilitv derived from the correlation for studies of human beings, for whom selection studies or identical twins reared apart is referred to as broad comparisons among inbred strains cannot be conducted, heritability - it includes all sources of genetic variance. In the only way is to study pairs of individuals who differ in contrast, first-degree relatives primarily share only ad- genetic resemblance. If heredity is important for a partic- ditive genetic variance, genetic effects that add up line- ular characteristic, pairs of individuals who are more arly in their effect on the phenotype; estimates of similar genetically ought to be more similar for the heritability based on first-degree relatives adopted apart measured characteristic. For example, third-degree rela- are thus primarily limited to additive genetic variance BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment and are thus referred to as narrow heritability. This ment. For example, a correlation of.25 for a trait distinction is important to the extent that nonadditive measured in pairs of adoptees reared in the same adoptive genetic variance is important; if nonadditive genetic vari- homes suggests that 25% of the phenotypie variation in ance affects a trait, behavioral genetic designs that assess the trait can be explained by shared environment. A narrow heritability will misread this genetic variance as correlation of zero for pairs of adoptees, on the other nonshared environment. Although most behavioral ge- hand, implies that shared environment contributes noth- neticists discount the importance of nonadditive genetic ing to phenotypie variance, which implies that all of the variance, some recent work suggests that it contributes to environmental variation is nonshared. certain characteristics (Lykken 1982; Plomin 1986). It should be mentioned that the distinction between Phcnotypic variance not explained by.genetic variance shared and nonshared environment is not limited to is ascribed to environmental sources. More properly, this family relationships in which relatives are the same age component of variance is nongenetic; that is, it is broader (such as twins), or relatives who are nearly the same age than the usual way psychologists think about the environ- (such as siblings). We can also consider shared and non- ment in that it includes accidents and illnesses, prenatal shared environmental factors that affect the resemblance influences, cytoplasmic changes, and even DNA changes between parents and their offspring. In this case, shared environment refers to environmental influences that in- that are not transmitted hereditarily. Data for relatives crease resemblance between parents and offspring. It adopted apart, as in the case of separately adopted identi- does not involve all parental influences on offspring, only cal twins, cannot by themselves separate shared and those environmental influences that increase phenotypie nonshared environmental components of nongenetic similarity between parents and their children. variance. Other adoptions designs can assess shared and non- 1.2. Twin design. The twin design compares the re- shared environment. Comparisons between relatives semblance of identical twins with that of same-sex frater- adopted apart and relatives reared together permit an nal twins. Both types of twins are born at the same time, indirect assessment. Relatives adopted apart share he- share the same womb and home, and are of the same sex. redity but not environment, whereas relatives reared One major difference distinguishes the two types; Identi- together are similar for reasons both of shared heredity cal twins are twice as similar genetically (on the average) and shared environment. If relatives reared together are as fraternal twins. If heredity affects a trait, the twofold no more similar than relatives adopted apart we can greater genetic similarity of identical twins will make conclude that growing up in the same family does not add them more similar than fraternal twins with respect to a to relatives' resemblance beyond the similarity induced particular trait. The difference between the correlations by heredity. In other words, environmental influence for identical twins and fraternal twins is an estimate of operates in a nonshared manner. For example, if, for a roughly half of the genetic variance in the population particular trait, identical twins reared together are no because the coefficient of genetic relationship is 1.0 for more similar than identical twins reared in uncorrelated identical twins and .50 for fraternal twins. Thus, for a trait environments, shared environment is unimportant for completely determined by heredity, the expected cor- that trait and all of the environmental variance must be relations are 1.0 for identical twins and .50 for fraternal nonshared. On the other hand, if the correlation for twins. If the pattern of twin correlations were .75 and .50 identical twins reared together is .75 and the correlation for identical and fraternal twins, respectively, heredity for identical twins reared apart is .50, 25% of the phe- would be estimated to explain half of the phenotypie notypic variance could be attributed to shared environ- variance for the trait. If heredity does not affect the trait, ment and the remaining 25% to nonshared environment. the twofold greater genetic similarity of identical twins We have included this concrete example of partitioning will not make them more similar than fraternal twins for only for purposes of clarification. We do not mean to the particular trait. convey that such estimates will be particularly precise. This discussion has oversimplified the twin method for The accuracy of the estimates depends on all of the usual didactic purposes. For example, assortative mating statistical issues such as sample size as well as on the would raise the fraternal twin correlation and nonadditive assumptions of behavioral genetic designs. The estimates genetic variance would lower it. Also, even though twin of nonshared environment described later in our review partners of both types live in the same family, it is come from large samples, are replicated in many studies, possible that identical twins experience more similar and are based on quite different designs such as adoption family environments than do fraternal twins. If this were as well as twin studies. Moreover, our estimate of non- the ease, some of the greater observed similarity of shared environment would have to be very substantially identical twins might be due to greater similarity of their wrong before it would seriously affect our conclusion that experience. This possible confounding effect has been nonshared environment is responsible for most environ- examined and, in research to date, does not appear to mental variation relevant to psychological phenomena. represent a major problem for the twin design (Plomin, A direct test of the importance of shared environment DeFries & McClearn 1980). Finally, genotype-environ- comes from the other side of the adoption design in which ment interaction and correlation can affect these esti- genetically unrelated individuals are adopted into the mates, as discussed later. same family. These adoptive family members share major features of their environment - the same parents, home, If genetic variance accounts for 50% of the phenotypie social class, community, schools, and so forth - but they variance, the rest of the phenotypie variance is attributed do not share heredity. The correlation for pairs of unrelat- to nongenetic variance, which includes shared and non- ed children adopted together directly estimates the pro- shared environment as well as error of measurement. The portion of phenotypie variance due to shared environ- twin method can be used to partition nongenetic variance BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment into its shared and nonshared components. Consider two due to nonshared environment, and error are removed. patterns of identical vs. fraternal twin correlations: .75 Environmental variance not due to shared environment vs. .50 and .50 vs. .25. Doubling the difference between is called nonshared environment; this portion of environ- the twin correlations suggests a heritability of 50% for mental variance makes family members different from both patterns of correlations. Thus, for both patterns, the one another. This variance component is usually esti- proportion of phenotypic variance due to environmental mated as the remainder of phenotypic variance once variance is 50%. In the first case, however, the one with variance due to heredity, shared environment, and error correlations of .75 and .50 for identical and fraternal of measurement is removed. Differences within pairs of twins, respectively, half of the environmental variance is identical twins reared together provides a direct estimate shared by the twins, making them resemble each other, of nonshared environment as experienced by identical and the other half of the environmental variance makes twins. them different. In the case of identical and fraternal twin Because we are developmentalists, we feel compelled correlations of .50 and .25, all of the environmental to make the point that all components of variance can variance contributes to differences within pairs. change during development. Estimates of genetic and The reasoning behind this conclusion is as follows: environmental components of variance depend upon the Differences within pairs of identical twins are due only to age of the subjects sampled. Genetic change during nongenetic factors not shared by twins because members development is the focus of a new subdiscipline, develop- of identical twin pairs do not differ genetically. Thus, mental behavioral genetics (Plomin 1986). Nonshared when identical and fraternal twin correlations are .50 and shared environmental components can also change and .25, respectively, .50 of the phenotypic variance is during development. Research is needed to trace the genetic and .50 is nongenetic. Because identical twins developmental course of shared and nonshared environ- are identical genetically and yet their phenotypic correla- mental variance. For example, there may be a general tion is only .50, all of the nongenetic variance (specifical- trend for nonshared environmental variance to increase ly, nonshared environment and error of measurement) with age as individuals expand their social and environ- leads to differences within pairs. Variance clue to error of mental networks beyond the family. On the other hand, measurement can be assessed as the difference between as this happens, there may be fewer forces contrasting the reliability coefficient (e.g., test-retest correlation) children in the same family. Research throughout the and 1.0. For example, if a test-retest correlation is .90, lifespan - especially research past adolescence - will be error variance is 10%; the 50% nongenetic variance thus needed to resolve such developmental issues. One strik- consists of 40% nonshared environmental variance and ing example of developmental change in the relative 10% error variance. When the identical and fraternal twin influence of shared and nonshared environmental vari- correlations are .75 and .50, half of the phenotypic vari- ance serves to indicate the potential usefulness of a ance is again environmental, but in this case only hal fof lifespan perspective: For 1Q, the shared environment the environmental variance (25% of the total phenotypic component of variance diminishes dramatically after variance: 1.0 — .75 = .25) is due to nonshared environ- childhood, as discussed in the next section. ment and error and the other half is shared. The shared environment component of variance can be estimated as 2. Evidence for the importance of nonshared twice the fraternal twin correlation minus the identical environmental effects on behavior twin correlation. In summary, the twin design provides a direct estimate This section provides a brief summary of behavioral- of nonshared environment - the component of phe- genetic research in personality, psychopathology, and notypic variance that is not shared by members of identi- cognition that leads to the conclusion that the most cal twin pairs. In addition, the twin design provides an important source of environmental variance is nonshared indirect estimate of shared family environment: It is the environment. This material is based on a recent review of component of phenotypic variance that remains after behavioral-genetic research throughout the lifespan accounting for genetic variance and nonshared environ- which can be consulted for additional studies and details mental variance. The generalizability of twin results con- (Plomin 1986). Although readers might take issue with cerning shared family environment to the population of the precise magnitude of one or another of the estimates, nontwin siblings is questionable, however, because it the forest should not be overlooked for'the trees. Our seems likely that twins share family environments to a point, one that to our knowledge has not been disputed, is greater extent than do siblings who are not twins, as will that nonshared environment is responsible for most en- be discussed later. vironmental variation relevant to psychological develop- Thus, adoption and twin studies can separate environ- ment. Thus, our goal in the following section is not to mental variance for behavioral traits into two compo- provide an encyclopedic review of behavioral-genetic nents. One component, called shared environment, in- studies but rather to summarize the results to the extent cludes all environmental influences that make children in needed to understand their message regarding the im- a family similar to one another. This component of vari- portance of nonshared environment. ance can be estimated in three ways: (1) from the correla- tion for genetically unrelated children reared together in 2.1. Personality. The importance of nonshared environ- the same adoptive families, (2) from the difference in ment was first highlighted by Loehlin and Nichols (1976) correlations for relatives reared together and relatives whose twin analyses of personality data led to the follow- adopted apart, and (3) from twin studies, as the remainder ing conclusion: of phenotypic variance when genetic variance, variance Thus, a consistent - though perplexing - pattern is BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment emerging from the data (and it is not purely idiosyn- These twin studies used self-report questionnaires. cratic to our study). Environment carries substantial Perhaps some artifact exists so that identical twins always weight in determining personality - it appears to rate themselves as 50% similar when asked about their account for at least half the variance - but that environ- personality. Other assessment procedures, however, ment is one for which twin pairs are correlated close to yield similar results. For example, in recent years, sever- zero. ... In short, in the personality domain we seem al twin studies using parental ratings of children's person- to see environmental effects that operate almost ran- ality have been reported (reviewed by Buss & Plomin domly with respect to the sorts of variables that psy- 1984). The average identical twin correlation is about .50, chologists (and other people) have traditionally again suggesting that about half of the variance is due to deemed important in personality development. nonshared environment. The few twin studies that have (Loehlin & Nichols 1976, p. 92) used objective observations of personality yield some- Loehlin and Nichols reached this conclusion because what less ubiquitous evidence for nonshared environ- identical and fraternal twin correlations were consistently mental variance than do paper-and-pencil questionnaires about .50 and .30, respectively, within their large study (Plomin & Foch 1980). Nonetheless, estimates of non- of high-school-aged twins that used self-report person- shared environmental influence from these studies are ality questionnaires. This pattern of correlations suggests still substantial - usually greater than estimates of shared 40% genetic variance and 60% environmental variance, environmental variance, even when error variance is and that over 80% of the environmental component of taken into account. variance is due to nonshared environment plus error. Studies of nontwin siblings and other family rela- (Error accounts for about 20% of the variance.) tionships confirm the hypothesis that shared family en- These results are not peculiar to Loehlin and Nichols's vironment accounts for a negligible amount of environ- study of high-school twins. In a review of 10 recent twin mental variance relevant to personality development. studies of personality (Goldsmith 1983), the average twin For example, one of the earliest studies found an average correlations were .47 for identical twins and .23 for fra- sibling correlation of. 12 (Crook 1937); a recent large ternal twins. This pattern of twin correlations suggests family study (Ahem, Johnson, Wilson, McCleam & Van- that heredity accounts for 50% of the phenotypic variance denberg 1982) yielded an average sibling correlation and that nonshared environment and error of measure- of .16 for three widely used personality questionnaires. ment explain the rest. The average parent/offspring correlations in this study It might seem odd to report average correlations across were also low: .12 for father/son, .10 for father/ a domain as diverse as personality. Nonetheless, the twin daughter, . 13 for mother/son, and .14 for mother/ results are generally similar across the dozens of traits daughter. measured by self-report questionnaires. Consider extra- Four recently reported adoption studies of personality version and neuroticism, the two "super-factors" in per- indicate that this modest familial resemblance is not due sonality, which are associated with Eysenck (e.g., 1967) to shared family environment - the average adoptive but also emerge as major second-order factors from other sibling correlation is .04 and the average adoptive par- personality questionnaires such as Cattell's Sixteen Per- ent/adopted child correlation is .05 (Loehlin, Horn & sonality Factor Questionnaire (Cattell, Eber & Tatsuoka Willerman 1981; Loehlin, Willerman & Horn 1985; 1970). [See also Zuckerman: "Sensation Seeking" BBS Scarr, Weber, Weinberg & Wittig 1981; Scan- & Wein- 7(3) 1986.] A study of over 12,000 adult twin pairs in berg 1978a). Adoptive sibling correlations are also low in Sweden (Floderus-Myrhed, Pedersen & Rasmuson 1980) the first report of infant adoptive siblings, involving 61 revealed twin correlations of .51 and .21 for identical and pairs at 12 months and 50 pairs at 24 months tested as part fraternal twins, respectively, for extraversion and correla- of the Colorado Adoption Project (Daniels 1985). Parental tions of .50 and .23 for neuroticism. ratings of temperament yielded average adoptive sibling Similar results emerge for less central dimensions of correlations of .11 at 12 months and .05 at 24 months; personality as well. For example, Loehlin and Nichols's tester ratings on the Infant Behavior Record (Bayley study used the California Psychological Inventory, which 1969) yielded average adoptive sibling correlations of includes diverse scales such as Sense of Weil-Being, -.14 at 12 months and .05 at 24 months. Tolerance, and Good Impression. The identical and fra- ternal twin correlations, respectively, for these scales 2.2. Psychopathology. Behavioral-genetic data on psy- were .50 and .30, .53 and .35, and .48 and .30. Another chopathology are also consistent with the conclusion that example involves twin results for a new personality ques- environmental variation is preponderantly of the non- tionnaire, the Differential Personality Questionnaire, shared variety. Research on schizophrenia is difficult to which assesses nontraditional dimensions of personality. summarize briefly because concordance rates vary widely A twin study of over 200 identical twin pairs and over 100 depending on the following: whether or not age correla- fraternal twin pairs yielded the following sampling of tions are used, the type of diagnostic criteria used, and correlations for identical and fraternal twins, respec- the selection and severity of probands. Nonetheless, tively: .50 and .36 for Danger Seeking; .61 and .37 for relying on a recent book-length review (Gottesman & Authoritarianism; and .58 and .25 for Alienation (Lyk- Shields 1982), familial concordance rates for schizo- ken, Tellegen & DeRubeis 1978). The only personality phrenia in a dozen studies found about 10% concordance trait that appears to show significant shared environmen- rates for schizophrenia for first-degree relatives. The tal influence is masculinity-feminity, which one might concordance rate for fraternal twins is also about 10%. argue falls more in the category of attitudes than person- Concordance rates for identical twins are substantially ality (Loehlin 1982). higher than those for fraternal twins - indeed, higher BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment than would be expected on the basis of a simple additive relation in over 30 studies is .85 for identical twins genetic model in which identical twins would be about and .58 for fraternal twins (Bouchard & McGue 1981), twice as similar as fraternal twins. For example, Got- which suggests again that about 30% of the variance of IQ tesman and Shields review five recent studies that yield scores can be accounted for by shared environment. an average case-wise concordance of 45% for identical Although these data appear to converge on the reason- twins. Regardless of the complications this pattern of twin able conclusion that shared environment accounts for a concordance causes for estimates of genetic influence, the substantial portion of environmental variance relevant to results indicate that most schizophrenic identical twins IQ, doubts have begun to arise. For fraternal twins, who do not have an affected cotwin. Because these are genet- share environment to a greater extent than do nontwin ically identical pairs of individuals, nonshared environ- siblings, the IQ correlation is about .60, whereas the ment must be the reason for these striking differences correlation for nontwin siblings is about .40 - which within pairs of identical twins. means that the twin method overestimates the impor- This conclusion is confirmed in Gottesman and tance of shared environmental in comparison to family Shields's review of recent adoption studies in Denmark studies. in which the same concordance of about 10% is found for The crucial piece of evidence in support of substantial individuals adopted apart from a first-degree schizo- shared environmental variance is the correlation of .30 phrenic relative. Thus, sharing the same family environ- for adoptive siblings reared together. These studies have ment with a schizophrenic relative does not increase included adoptive siblings still living at home, with two familial concordance. exceptions. The first exception is a study of postadoles- Gottesman and Shields (1982) also review attempts to cent adoptee pairs by Scarr and Weinberg (1978a) which isolate environmental sources of variance and conclude: found a correlation of - .03 for IQ. This unsettling finding So far, no specific environmental source of liability is implies that shared environment is important for IQ known; the most likely environmental contributor, during childhood when children are living at home and stress, may come from many sources and, apparently, then fades in importance after adolescence when children may come during any stage of development. Prenatal have left home. or birth complications, early deprivations, broken The hypothesis that shared environmental influences homes, censuring parents, the death of someone close, have no lasting impact on IQ is supported by results o fa failures in school, poor work or social relationships, recent study of adoptive and nonadoptive siblings (Kent childbirth, a bad drug trip, as well as all kinds of good 1985). The study included 52 pairs of adoptive siblings fortune may have effects on a predisposed individual and 54 pairs of nonadoptive siblings ranging from 9 to 15 that are obvious only in retrospect. In prospect, it will years of age, with the average age of 13 years. A battery of be impossible to prophesy the events themselves, let cognitive ability measures was developed for administra- alone their effects. (Gottesman & Shields 1982, pp. tion over the telephone; this battery correlated with face- 241-42) to-face testing near the reliabilities of the tests. An unro- We suggest, however, that until more systematic re- tated first principal component, used as an index of IQ, search on nonshared environmental variance sources is yielded a reasonable correlation of .38 for nonadoptive conducted it is too early to conclude that the large siblings; however, the IQ correlation for adoptive siblings environmental component of variance in schizophrenia is was -.16, not significantly different from zero. A similar brought about by idiosyncratic experiences. pattern of results emerged for specific cognitive abilities. Research on manic-depressive psychosis yields results The adoptive sibling correlations for verbal, spatial, per- similar to those for schizophrenia (Plomin, DeFries & ceptual speed, and memory abilities were —.06, -.07, McClearn 1980). Environmental influences on less se- -.10, and .16, respectively. vere forms of psychopathology, such as neuroses and Thus, this study leads to the conclusion that shared alcoholism, also appear to be predominantly nonshared. environmental influence on IQ and specific cognitive Sibling concordances are generally less than 20% and abilities is of negligible importance by the end of early when twin and adoption studies have been conducted adolescence. Because these estimates of shared environ- most of this familial resemblance has been found to be mental influences were obtained directly from adoptive genetic in origin (Fuller & Thompson 1978; Rosenthal sibling correlations, reasonable confidence can be at- 1970). In other words, the most important influences on tached to this conclusion. For example, the sample of 52 psychopathology lie in the category of nonshared en- pairs of adoptive siblings permits detection of a true vironment. Much more often than not, affected children correlation of .30 with 70% power; the standard error of in families with more than one child will have unaffected the estimates of shared environment were found to be siblings. between . 10 and . 14 when a multiple regression model- fitting approach suggested by DeFries and Fulker (1985) 2.3. Cognition. Until recently, environmental variance was used. that affects individual differences in IQ was thought to fall In summary, nonshared environmental influence is a primarily in the category of shared environment. In 11 major component of variance for personality, psycho- studies, the average IQ correlation for adoptive siblings pathology, and IQ (after childhood). We conclude that is .30, suggesting that 30% of the variance in IQ scores is nonshared environment explains perhaps as much as 40% due to shared environmental influences (Bouchard & to 60% of the total variance for these domains. Although McGue 1981). Adoptive parent/adopted child IQ correla- one can quibble with the magnitude of our estimates, tions are lower, about .20, but still suggest substantial they would have to be substantially in error before they influence of shared environment on parent-offspring would affect our argument that most of the environmental resemblance. Twin studies agree: The average IQ cor- variance is nonshared. 6 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment 3. Shared and nonshared environmental variance specific environmental factors are responsible for the components of variance, it seems to be a reasonable first The purpose of this section is to consider some conceptual step to ask about components of variance - without this details of the distinction between shared and nonshared tack, we would not have discovered that nearly all en- environment before discussing sources of nonshared en- vironmental variance is of the nonshared variety. It is a vironment. These details include other labels for shared major strength of the approach that it can reveal the and nonshared environment, the distinction between presence of genetic and environmental influences even environmental components of behavioral variance and when these are not assessed directly. the relationship between specific environmental mea- Attempts to isolate specific environmental factors will sures and behavior, the impact of nonshared environ- be presented later. A related issue, however, should be mental influence on the development of singletons, gen- mentioned at this time. Traditional environmental re- otype—environment correlation and interaction, and search attempts to relate measures of family environment model-fitting. to measures of behavior of one child per family. The yield from such research has been disappointing, especially if 3.1. Other labels. Shared and nonshared environmental one considers the amount of variance explained (Maccoby influences were named by Rowe and Plomin in 1981, & Martin 1983). Knowing this research, one might ask although the distinction between environmental influ- why such environmental factors as parental affection ences that contribute to the resemblance between rela- should be important within families when they account tives and those that do not has been implicit in quan- for little variance in behavior across families. That is, if it titative genetics since its inception. Many labels have makes little difference that some parents love their chil- been used to refer to these two components of environ- dren more than other parents love their children, why mental variance. Shared environmental influence has should parental love make a difference within families if a been called E2, between-family, and common environ- parent loves one child more than another? The answer is mental variance, labels that have been used to refer to that there is no necessary relationship between the causes nonshared environmental include El, within-family, in- of differences between families and the causes of dif- dividual, unique, and specific environmental variance. ferences within families. That is, environmental factors Rowe and Plomin suggested that the symbols El and E2 that create differences within families can act indepen- (Jinks & Fulker 1970) are probably best in that they carry dently of factors that cause differences between families. no connotations, although they have the distinct disad- For example, a child really knows only his own parents; vantage that they provide no mnemonic to remember the child does not know if his parents love him more or which is which. Within- and between-family environ- less than other parents love their children. A child is ment are the terms most often used. They are useful for likely to be painfully aware, however, that parental affec- those familiar with the terminology of analysis of variance tion toward him is less than toward his sibling. which considers variance within and between groups. Variance within families refers to differences among fami- 3.3. Singletons. Because over 80% of U.S. families have ly members and variance between families describes more than one child, it is important to understand why resemblance among family members. The term "within- children in a family are so different from one another. family environment, however, connotes factors that oc- How does nonshared environment relate to singletons? cur within the confines of the family; whereas nonshared In general, reasons why two children in the same family influences are those that cause family members to differ differ are likely to yield clues as to the environmental regardless of whether the locus of influence is the family source of variance for singletons as well. The easiest (such as differential treatment by parents) or outside the example involves nonsystematic events such as accidents family (such as different experiences at school or with and illnesses which are just as likely to befall singletons. peers). For these reasons, we suggest that the most However, systematic nonshared influences may also be descriptive and straightforward terms to use are shared found to affect singleton variance. For example, if certain and nonshared. characteristics of peer groups differ within pairs of sib- lings and contribute importantly to behavioral differences 3.2. Components of variance versus specific measures. It within sibling pairs, it is likely that these characteristics should be noted that this discussion pertains to environ- also contribute to variance for singletons. mental components of behavioral variance, not to the Obviously, singletons do not have siblings with whom relationship between specific environmental measures they interact; thus, this potential source of nonshared and behavioral measures. In this sense, quantitative environment cannot contribute variance for singletons. genetic analyses describe the "bottom line" of genetic Although it might seem at first that differential parental and environmental influence. That is, the total impact of treatment of two children in the same family is irrelevant genetic variability on phenotypic variability will be de- to singletons, it is possible that, once identified, such tected regardless of the complexity of the genetic effects - factors might contribute to the variance of singletons. for example, whether the genetic effects arise from vari- There is evidence that parents with more than one child ability in structural genes that code for polypeptides or treat the children similarly if we look at the children at the from regulatory genes. Similarly, quantitative genetics same age, which suggests that parental treatment is not estimates the bottom line of environmental influence, an important source of nonshared environment (Dunn, regardless of the specific mechanisms by which environ- Plomin & Nettles 1985). Except for twins, however, mental factors affect behavior. Although this components siblings are not the same age, and when we examine of variance approach may be unsatisfying for those who contemporaneous parental treatment of children of dif- would like to know which specific genes and which ferent ages, we find that parents treat the children differ- BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment ently (Dunn, Plomin & Daniels 1986). Differences in correlation for adoptive siblings is significant. For this parental behavior during development can also affect reason, and because of the relative inaccessibility of most singletons in that parents will treat their singleton chil- models, we have emphasized the basic correlational data dren differently during the course of development. and merely note that model-fitting approaches confirm Thus, studies of differences within pairs of siblings are our conclusions. likely to illuminate factors responsible for singleton vari- ance as well as sibling variance. The important point in the present context is the obvious one: that the study of 4. Categories of nonshared environmental singletons cannot isolate factors that make two children in influence the same family different from one another. Because this is the best clue we have as to the source of environmental What is happening environmentally to make children in variance relevant to psychological development, it makes the same family so different from one another? One sense to focus on environmental sources of differences gloomy prospect is that the salient environment might be between children in the same family. unsystematic, idiosyncratic, or serendipitous events such as accidents, illnesses, and other traumas, as biographies 3.4. Genotype-environment correlation and interaction. often attest. In his autobiography, Darwin noted one Two complicating factors in the estimation of quantitative example: genetic parameters are genotype-environment correla- The voyage of the Beagle has been by far the most tion and genotype-environment interaction (Plomin, important event in my life, and has determined my DeFries & Loehlin 1977). Genotype-environment cor- whole career; yet it depended on so small a circum- relation refers to an increase in phenotypic variance that stance as my uncle offering to drive me thirty miles to occurs when children experience environments corre- Shrewsbury, which few uncles would have done, and lated with their genetic propensities. Phenotypic vari- on such a trifle as the shape of my nose. (Darwin 1892, ance can also be due to genotype-environment interac- p. 28) tion when children respond differently to the same Darwin's comment about his nose refers to the quixotic environment because of genetic differences among them. captain of the Beagle, Captain Fitz-Roy, who nearly What are the effects of genotype-environment correla- rejected Darwin for the trip because the shape of his nose tion and interaction on estimates of shared and nonshared indicated to Fitz-Roy that Darwin would not possess environment? Consider a direct estimate of nonshared sufficient energy and determination for the voyage. (Dar- family environment: the extent to which the correlation win wrote that, during the voyage, Fitz-Roy became for identical twins reared together is less than 1.0. This convinced that "my nose had spoken falsely" [p. 27].) estimate will not include either genotype-environment It is possible that nonshared environmental influences correlation or interaction because identical twins are could be unsystematic in the sense of stochastic events identical genetically; thus, in terms of genetic propen- that, when compounded over time, make children in the sities, identical twins will correlate and interact with the same family different in unpredictable ways. Such ca- environment in a similar manner. Similarly, the direct pricious events, however, are likely to prove a dead end estimate of shared family environment - the correlation for research. More interesting heuristically are possible for unrelated children reared together - will not include systematic sources of differences within families. genotype-environment correlation or interaction be- Table 1 describes categories of environmental factors cause these children are genetically uncorrelated; thus, that could lead to observed differences between children in terms of their genetic propensities, they will correlate in the same family. These include such systematic sources and interact with the environment in ways that do not add of nonshared influence in the family as birth-order and to their resemblance. However, estimates of nonshared gender differences of siblings, interactions between sib- or shared environment derived as the remainder of phe- lings, differential treatment by parents, and extrafamilial notypic variance after other components of variance are influences such as peers. taken into account can be affected by genotype-environ- In one sense, thinking about environmental influences ment correlation and interaction because of their effects that create differences between children in the same on estimates of genetic variance (Plomin e tid. 1977). family represents a dramatic reconceptualization of psy- chological environments. On the other hand, this recon- 3.5. Model-fitting. Fitting models to adoption and twin ceptualization need not involve mysterious elements in data is a powerful way to estimate quantitative genetic the environment: Any environmental factor can be parameters (Loehlin 1978). Although model-fitting tech- viewed in terms of its contribution to nonshared environ- niques differ in their specifics, they all express family mental variance. For example, parental affection can be resemblance in terms of an underlying model consisting easily construed as a source of differences among children of several unobserved genetic and environmental param- in the same family, because parents may be more affec- eters. The approach is powerful because it makes assump- tionate toward one child than another. tions explicit, it tests a specific model, and it can incorpo- In this sense, our conceptualization of nonshared en- rate into a single analysis different types of data, such as vironmental influence is not new and exciting. Although family and adoption data, rather than analyzing each type any traditional environmental factor can be viewed in of data separately. Model-fitting procedures, however, terms of its contribution to nonshared environmental only find significant parameters when they are implicit in variance, it is important to emphasize the point men- the basic data; for example, in a study of adoptive siblings, tioned earlier: There is no necessary relationship be- a reasonable model-fitting analysis will estimate signifi- tween environmental factors that contribute to dif- cant shared family environmental influences onlv if the ferences between families and those that affect differ- BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 Plomin & Daniels: Nonshared environment Table 1. Categories of environmental influences that cause compare their relative effectiveness in predicting sibling children in the same family to differ differences. A first attempt to assess differences in per- ceived environments of siblings is discussed in the follow- Categories Examples ing section. Error of measurement Test-retest unreliability 5. Attempts to identify nonshared environmental Nonshared environment influences Nonsystematic Accidents, illnesses, trauma Systematic: This section explores attempts to assess specific factors Family composition Birth-order; gender differences within these categories that may be responsible for non- Sibling interaction Differential treatment shared environmental variance. Family constellation var- Parental treatment Differential treatment iables, especially birth order, have been studied exten- Extrafamilial Peer groups; teachers; sively. Other categories of possible nonshared environ- networks television mental influence such as differential parental treatment, differential sibling interaction, and differential extra- Source: Adapted from Rowe and Plomin (1981). familial experiences have not yet received much attention. 5.1. Birth-order. The only specific source of nonshared ences between siblings within a family. In some cases, it family environment to receive considerable attention is seems likely that there is no relationship: Socioeconomic birth-order. For example, over 1,000 entries for "birth- status (SES), for example, is an important factor that order" appear in Psychological Abstracts. Birth-order is a operates between families, but even though the SES of prototype of nonshared environmental influence in that it families changes, it is unlikely that SES is an important is different for children in the same family and yet cannot source of differences between siblings. Conversely, an originate in genetic differences among siblings. Paradox- environmental factor that makes only a slight difference ically, however, most studies have analyzed its effect between families may be critical within families. For across families rather than within families and most of the reasons such as these, what is needed more than specula- relationships are weak for IQ (Galbraith 1982) and for tion about the most relevant nonshared environmental personality (Ernst & Angst 1983). influences is research identifying relevant factors. This research can at the same time provide insights into 5.2. Other systematic nonshared environmental influ- theoretical issues such as the relationship between non- ences. Although birth-order has received considerable shared influences and traditional environmental factors attention, studies of differential parental treatment, sib- studied across families. ling interaction, and extrafamilial influences are more The perspective of nonshared environment does, how- promising. In exploring possible nonshared influences, ever, suggest some new ways to study environmental the first step is to ask whether siblings in a family have influences. For example, we must focus on measures of different experiences. If siblings do not differ in their experience specific to each child. That is, one implication experience for a particular aspect of the environment of our conclusion concerning the importance of non- then that environmental factor cannot be a source of shared environment is that environmental factors shared differences between them. For birth-order, this first step by both children in a family are unlikely to be important is unnecessary because siblings obviously differ in birth- sources of environmental influence. Environmental mea- order. Experiential differences, however, cannot be as- sures are needed that capture the major sources of differ- sumed to affect behavioral differences within pairs of ential experience of siblings. Another strategy for re- siblings, therefore demonstrating that nonshared experi- search is exemplified by the emphasis of family therapists ences are related to differences in sibling behavior is the on systems theory in which the child is viewed as part of second step. The third step is to describe the direction of an organized family system, creating and maintaining effects when associations are found between differential patterns of behavior (Minuchin 1985). Another strategic experience and differences in their behaviors. Do sibling suggestion for the study of nonshared environment is to differences in experience affect or merely reflect dif- explore environmental sources of developmental dif- ferences in sibling behavior? ferences within individuals (McCall 1983): An environ- Thus, there are three steps in research on nonshared mental factor that is responsible for change in a child from environmental influences: identifying experiences that early childhood to school age is also likely to make are not shared by family members, relating such non- children in the same family different from one another. shared environmental factors to differences in sibling Finally, another, even more speculative, meth- behavior, and determining the causal direction of such odological lead for research is that subjective, perceived relationships. Because the topic of nonshared environ- experiences may prove to be important (e.g., Jessor ment is so new, only a few relevant studies have been 1981). For example, even if during home observations reported and most of these address the first step. children in the same family appear to receive the "same" environmental treatment, this does not mean that the 5.3. Sibling inventory of differential experience. One sys- children experienced the treatment similarly. We do not tematic approach to the topic is the Sibling Inventory of mean to suggest that objective assessment of the environ- Differential Experience (SIDE; Daniels & Plomin 1985). ment is not also needed - it would be best to use objective The 73-item self-report SIDE asks each sibling to com- and subjective measures in the same study in order to pare his experiences to those of a sibling in the domains of BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1 9 Plomin & Daniels: Nonshared environment sibling interaction, parental treatment, peer charac- college oriented, delinquent, and popular, respectively. teristics, and events specific to the individual. For all Siblings also agree substantially as to which sibling was items, siblings are asked to compare their relative experi- more jealous (r = .56) and which sibling displayed more ences rather than to make absolute judgments about their caretaking (r = .56). Siblings agree to a lesser extent on experience. For example, rather than asking the extent to differences in parental treatment (r = .26 and .28 for which "my sibling and I show understanding for each maternal and paternal affection). The median sibling other," the SIDE asks, "Who has shown more under- agreement correlation over the 11 SIDE scales is .49, standing for the other?" A 5-point scale is used for the which is above typical interrater agreement on person- siblings' ratings: 1 = My sibling has been much more this ality and environmental paper-and-pencil measures. The way than I have; 2 = My sibling has been a bit more this high sibling agreement found for some of the SIDE scales way than I have; 3 = My sibling and I have been the same may be due to the fact that siblings are asked to make a in this way; 4 = 1 have been a bit more this way than my relative and specific comparison to their sibling rather sibling; and 5 = 1 have been much more this way than my than an absolute judgment in comparison to all other sibling. This provides relative scores indicating, for exam- children of that age. Because the SIDE intentionally ple, the extent to which one sibling feels he is understood assesses siblings' perceptions of their differential experi- by the other. Although somewhat unusual, these relative ence, sibling agreement is not an important criterion for judgments have several advantages. First, they should be the usefulness of the measure as long as the measure is easier to make than absolute judgments - for example, on reliable. Other substantive findings from the SIDE are a 5-point scale, how much do you understand your sib- interwoven throughout the following discussion on the ling? (compared to what?). Second, relative judgments do major categories of systematic nonshared environment. not require that a sibling difference be calculated in order to assess nonshared environment. Third, they can be 5.4. Parental treatment. Environmental research has tra- used when data are available from only one member of a ditionally focused on parental treatment because parents sibling pair. The SIDE can also be coded to indicate the appear first and foremost in young children's lives. It has absolute rather than relative amount of differential sibling not been easy, however, to document parental effects on experience by disregarding the direction of the differen- children's development. A recent review of the rela- tial experience (i.e., 0 = no difference in sibling experi- tionship between parental treatment and children's de- ences; 1 = some difference; 2 = much difference). velopment concludes that "in most cases, the rela- The 11 scales of the SIDE (see Table 2) were devised tionships that have appeared are not large, if one thinks using the results of factor analyses of data on a sample of in terms of the amount of variance accounted for" (Mac- 396 12- to 28-year-old siblings from the Denver metro- coby & Martin 1983, p. 82). Indeed, these findings led politan area. The word "differential" precedes the label the authors to argue for the need to examine intrafamilial for each scale to emphasize that all items involve relative variation in the parent-child relationship. It should be (differential) ratings. The 2-week, test-retest reliabilities reiterated that the importance of nonshared environ- are reasonable, with a mean of .84 and a range from .70 ment does not denigrate the importance of environmen- to .94. The scales are virtually independent of siblings' tal influence. Environmental influence is important but age, birth-order, and gender. Also included in the table it operates differently from the way we thought it oper- are sibling agreement correlations which indicate that ated. In the case of parental influences, the effect that siblings agree quite substantially, especially in the areas parents have on their children has little to do with those of differential sibling interaction and peer group charac- aspects of parenting that are experienced similarly by teristics. The sibling agreement correlations are .55, .73, two children in their family. Whatever these parental and .60 concerning which sibling's peer group was more influences might be, they differentiate rather than inte- Table 2. Scales of nonshared environmental influence from the Sibling, Inventory of Differential Experience (SIDE) Test-retest Sibling Category Scale reliability agreement Sibling Differential Sibling Antagonism .83 .39 interaction Differential Sibling Jealousy .93 .56 Differential Sibling Caretaking .89 .56 Differential Sibling Closeness .70 .23 Parental Differential Maternal Affection .82 .26 treatment Differential Maternal Control .77 .25 Differential Paternal Affection .77 .28 Differential Paternal Control .85 .49 Peers Differential Peer College Orientation .88 .55 Differential Peer Delinquency .94 .73 Differential Peer Popularity .84 .60 Source: Adapted from Daniels and Plomin (1985). 10 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:1

Description:
vironmental influences, then the biometric approach needs to be broadly Nonshared environment," like the solution to Aesop's enigma, Scarr, S. (1985) Constructing psychology: Making facts and fables for our times.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.