Growing up and Growing Apart: a Developmental Meta-Analysis of Twin Studies

Growing Up and Growing Apart: A Developmental Meta-Analysis of Twin Studies

Kathleen McCartney Monica J. Harris University of New Hampshire University of Kentucky

Frank Bernieri Oregon State University

Developmental change in twin similarity was examined with age contrasts in a meta-analysis of twin studies from 1967 through 1985. Intraclass re were coded from 103 papers that included data for monozygotic or dizygotic twins, or for both, on personality or intelligence variables. Analyses indi- cated that there was a general tendency for some intraclass rs to decrease with age. In other words, as twins grow up, they grow apart. There were also developmental differences associated with components of variance for heritability, the shared environment, and the nonshared environment. Mecha- nisms through which the nonshared environment may operate are discussed.

The purpose of this article is twofold. First, we offer meta- meta-analysis include (a) obtaining a combined estimate from analysis as a tool for developmental hypothesis testing. The field all available evidence, (b) identification of robust findings, and of developmental psychology is often criticized for being not a (c) the ability to determine whether age interacts with other study of development but rather a study of the child at a given variables of interest (e.g., methodological moderators). How- age (McCall, 1977). Meta-analysis can be used to combine stud- ever, meta-analysis does not permit the same degree of causal ies of subjects at different ages, as in a cross-sectional study. inference found in longitudinal studies (Cronbach. 1984). Spe- Second, we examined whether there are age-related differences cifically, differences among age groups can result from either age in twin similarity. On the basis of recent behavior genetics the- effects or cohort effects, the latter of which is only interesting ory and data (McCall, 1983; Plomin & Daniels, 1987; Rowe & from a historical perspective. Thus, like most methodologies, Plomin, 1981), we predict that as twins grow up, they grow developmental meta-analysis has both strengths and weak- apart; that is, twin similarity should decline with age. The im- nesses. portance of the nonshared environment is presumed to in- Twin studies are based on a philosophy and methodology that crease over time, as children's experiences move outside the make them perfect candidates for meta-analysis. Researchers family. have traditionally studied twins to compute heritability coeffi- Researchers have begun to realize that meta-analysis can be cients. The basic procedure is the same in all twin studies: First, used to summarize a body of research through the computation a sample of twins is obtained. Second, the zygosity of the twin of effect-size estimates. However, most researchers may be un- pairs is determined (i.e., monozygotic, or MZ, twins share aware that specific hypotheses can be tested by performing con- 100% of their genes and dizygotic, or DZ, twins share, on aver- trast analyses on the effect sizes of independent studies (Rosen- age, 50% of their genes). Third, individual difference measures thai, 1984). A developmental meta-analysis is thus denned as a are administered to all twins. Fourth, an intraclass correlation meta-analysis with age contrasts. between twins is computed, once for the MZ twins and once for The advantages of being able to perform age contrasts within the DZ twins. The expectation is that if a given trait or ability has a genetic component, then the intraclass correlation will be higher for the MZ twins than for the DZ twins. This research was supported by grants from the Harvard Graduate Consequently, every twin study contains the data for, and ide- Society and by National Institute of Mental Health Grant MH41807 awarded to Kathleen McCartney, who also acknowledges the support ally should report, the MZ and DZ intraclass correlations. of the Whitcomb family. These correlations can easily be used in a number of meta-ana- We would like to thank Robert Rosenthal for frequent statistical con- lytic procedures, including age contrasts; however, this has not sultations and Maura Kerrigan for her help in coding articles. We would been done. Typically, the twin method has been used to estimate also like to thank the reviewers, David Buss, Elizabeth Jordan, Carolyn components of variance for genetic and environmental influ- Mebert, Edward J. O'Brien, Sandra Scarr, Deborah Vandell, members ence. The assumptions of the method continue to be questioned of Murray Straus's research seminar, and especially Victor Benassi for (Goldsmith, 1983; Hoffman, 1985). The most critical assump- their comments on an earlier draft of this article. This article was pre- tion is that the degree of environmental similarity is the same sented by Kathleen McCartney at the International Conference on Psy- chology in Sydney, Australia, August 1988. for MZ and DZ twins. If MZ twins, because they look alike, Correspondence concerning this article should be addressed to Kath- experience more similar environments than DZ twins, then leen McCartney, Department of Psychology, Conant Hall, University of heritability computations would also reflect this environmental New Hampshire, Durham, New Hampshire 03824. similarity (Plornin, DeFries, & McClearn, 1980).

226 GROWING UP AND GROWING APART 227

However, twin studies are useful for purposes other than the Table 1 estimation of heritability. Twin studies are actually within-fam- Categorization of Identified Studies From Literature ily studies, in that more than one member of a family is studied. Search (Based on Abstracts) Longitudinal, within-family studies are needed to assess the rel- Category n ative importance of shared and nonshared experiences of siblings. Codable 130' 17 Several longitudinal, within-family studies do, in fact, suggest Uncodable that the older siblings become, the less like one another they Anecdotal or case studies 120 16 11 appear on various indexes of intelligence and personality. For Theoretical 81 No relevant data 114 15 example, three adoption studies have shown that adopted sib- Foreign language 83 11 lings, who are genetically unrelated, look moderately similar on Biological 141 18 intelligence tests and personality inventories in early childhood Animal 9 1 but do not resemble one another at all by later adolescence, Predated 9 1 Dissertations 52 7 presumably because the effect of the shared environment wanes Miscellaneous 25 3 (Kent, 1985; Loehlin, Willerman, & Horn, 1987;Scarr&Wein- berg, 1978). The Louisville twin study (Wilson, 1983) showed Total 764 100 that DZ twins became less similar over time on IQ. However, • Of these 130 papers, 103 were actually used in the meta-analysis. The MZ twins became more similar over time, perhaps because they remaining 27 papers did not contain data that contained intraclass rs seek and are exposed to highly similar environments. or that could be converted to intraclass rs. These 27 papers were receded In addition, a preliminary report of the Minnesota study of to the no-relevant-data category after the papers were read. twins reared apart and together showed that the common family environment accounted for little variance on 12 of 14 personal- were excluded because intraclass rs were not included and could not be ity scales (Tellegen, Lykken, Bouchard, Wilcox, Segal, & Rich, calculated; (d) foreign language papers; (e) studies on psychopathology 1988). Because the "environment," broadly calculated, did ac- or biology, which were excluded because the dependent variables were count for substantial variance, one possibility is that nonshared not in the targeted areas of personality and intelligence; (f) animal stud- experiences play a prominent role in development. ies; (g) predated studies, which were excluded because an updated study These findings were at first surprising to developmentalists, with more data was available; (h) dissertations, which were excluded who have traditionally focused on the role of the shared family because many were duplicated in later journal articles; and (i) miscella- environment in psychological development, as reflected in tra- neous studies that included articles we could not locate through interli- ditional socialization theory (see Maccoby & Martin, 1983). brary loan, papers from various conferences and proceedings, and pa- This perspective assumes that influential environmental experi- pere by Sir Cyril Burt, which have been shown to contain fraudulent data (Heamshaw, 1979). It should also be noted that unrefereed books ences are shared by siblings, because they take place within the and book chapters were excluded by the computer literature search pro- family (e.g., parental discipline, attitudes, and values). The pre- cedure. Thus, this search was limited to journal articles on the similar- diction that follows from this perspective is that the longer sib- ity of twins' personality and intelligence. Table 1 displays the number lings live together, the more like one another they should be- and percentage of studies per category. Note that 130 twin studies were come. read for this meta-analysis. Of these 130 papers, 103 actually contained A developmental meta-analysis of twin studies allows the intraclass re or data that could be converted to intraclass re; these 103 comparison of predictions concerning the importance of the papers were used in the meta-analysis (see the appendix for references shared and nonshared environments of siblings over time. for the studies). Methodological moderators, especially those that might be as- Because the validity of all meta-analyses rests on the extent to which sociated with age—for example, method of report (self vs. par- literature searches are both comprehensive and representative, we searched more broadly than was necessary. We believe that the compre- ent)—are also considered in an effort to rule out alternative hensiveness of the search was guaranteed by the broad descriptor terms hypotheses. used in the computer search. Certainly the search identified studies on a wide range of topics, from a wide range of periodicals. Although only Method 13.4% of studies from the literature search were coded, it is important to consider that all the studies excluded, with the exception of dissertations, were those that did not present relevant data on personality or Literature Search intelligence. We conducted a computerized literature search of Psychological Ab- Even if the search were not comprehensive, it would only be a prob- stracts, using twins, heterozygotic twins, and monozygotic twins as de- lem if the identified studies were not representative of studies published, scriptors. The time span covered January 1967 through December and there is no reason to suspect any bias. Some literatures—for exam- 1985. This procedure yielded 749 studies; an additional 15 studies were ple, that on sex differences—are biased with respect to which studies discovered through secondary sources, for a total of 764 studies. are published (see Hall, 1987). This is referred to as the "file drawer" The abstracts of all 764 studies were classified as codable or uncod- problem (Rosenthal, 1984), because unpublished, relevant studies can able. Codable studies were defined as those from journals that included only be located in the file drawers of researchers. However twin studies intraclass re for MZ twins, DZ twins, or both kinds of twins on personal- are unlikely to be found in file drawers because twins have been consid- ity or intelligence variables. Uncodable studies were subdivided into ered to be valuable subjects. In fact, even case studies are likely to be nine categories: (a) anecdotal or case studies, which were excluded be- published (16% of studies identified in this computer search were case cause no r was possible; (b) theoretical papers, which were excluded studies). More important, the file drawer problem is not an issue here, because they presented no new data; (c) no-relevant-data studies, which because this is a study of a moderator, namely age. 228 K.. MCCARTNEY, M. HARRIS, AND F. BERNIERI

Coding the Studies ences between the two selection strategies, so we report the data for all studies in order to simplify our presentation. Three coders each read approximately one third of the studies and Note that there is no agreed-on test of significance to deter- extracted intraclass re separately for MZ and DZ twins for each relevant mine whether these rs are reliably different from zero, because dependent variable. Coders recorded the number of male pairs, the it is not clear whether the degrees of freedom should reflect the number of female pairs, and the number of mixed pairs (for DZ twins). number of studies, which would be conservative, or the number Three moderator variables were recorded: age per group, method of zygosity determination (blood grouping vs. other), and type of report of subjects across studies, which would be liberal given that ob- (self vs. parent). We had planned to use type of DZ twin (same sex servations within studies are not independent. The rs in Table vs. opposite sex) as an additional moderator; however, only 12 studies 2 should, therefore, be treated as effect-size estimates and evalu- included opposite-sex DZ twins (9 of the 12 studies contained data on ated accordingly, where .10 is low, .30 is moderate, and .50 is twin similarity for intelligence). We decided to include data from DZ large (Cohen, 1977). twins that included mixed-sex pairs rather than lose these data; this may Table 2 reveals three important findings. First, the MZ rs are have resulted in a lower DZ intraclass correlation for some traits, al- greater than the DZ rs by .22 for the combined intelligence vari- though sex differences are far less likely for intelligence than for person- ables and by .29 for the combined personality variables. Al- ality variables. Because the mixed-sex DZ data were from studies of though a combined personality variable may not be readily in- seven different dependent variables (see categories described in the fol- terpretable, especially if traits are differentially heritable,2 the lowing section), we did not view this potential bias as a serious problem. Also, there is no reason to expect that mixed-sex DZ data should affect data for a combined personality variable are presented here for the moderator analyses for age. comparison purposes. Consider that these estimates are quite similar to two other reviews that reported median intraclass rs. Nichols (1978) reported a median intraclass r for general intelli- Classification of the Dependent Variables gence of .82 for MZ twins (vs. .81 in this study) and .59 for DZ twins (also .59 in this study). Nichols's review was based on the Dependent variables were classified via group consensus based on the literature through 1971, whereas this meta-analysis covers the knowledge of the researchers and on descriptions by authors. Intelli- gence was subdivided into five categories: verbal intelligence, quantita- years 1967 through 1985. Only nine studies in the present meta- tive intelligence, performance intelligence, perception, and total IQ. analysis came from the 5-year overlap between reviews. This Personality-temperament was subdivided into eight categories: activ- study thus provides a replication of Nichols's review with the ity-impulsivity,1 aggression, anxiety, dominance, emotionality, mascu- report of nearly identical estimates of intellectual concordance linity-femininity, sociability, and task orientation. for MZ and DZ twins. The median intraclass rs for intelligence Several personality constructs were not analyzed because they were reported by Bouchard and McGue (1981), which covered the examined in fewer than five studies. Thus, we lost constructs such as literature through 1980, are comparable: .85 for MZ twins and creativity, altruism, religiosity, hypnotic susceptibility, and humor. We .58 for DZ twins. For personality, this study also reported in- excluded one additional category, attachment, because it was measured traclass rs similar to those of Nichols (1978): .48 for MZ twins very differently across studies (e.g., approach to stranger, approach to (vs. .51 in the present study) and .29 for DZ twins (vs. .22 in mother, crying, and so on). Researchers often assessed a single construct with multiple measures the present study). Although the primary purpose of this study (e.g., self-report and behavioral observation). In such cases, we com- was to conduct a developmental analysis, the value of obtaining puted a combined effect-size estimate using the Fisher's z, transforma- up-to-date and accurate effect-size estimates cannot be over- tion (Rosenthal, 1984, p. 27). This is preferable to including all effect looked. sizes from a given study separately, a procedure that would violate the A second interesting feature of Table 2 is the low variance independence assumption that is essential for many meta-analytic pro- among the rs within the category of intelligence and within the cedures. category of personality. That this is true for the personality scales is especially intriguing. Researchers have recently hy- Results pothesized a genetic role in certain personality variables, most notably sociability (A. Buss & Plomin, 1984) and behavioral Mean Effect-Size Estimates inhibition (Kagan, Reznick, & Snidman, 1988). Variables such as masculinity-femininity and task orientation have typically Table 2 presents mean intraclass rs separately for MZ and been considered as having strong socialization components. DZ twins per dependent variable. Mean rs were calculated using The homogeneity of the intraclass rs for the personality vari- two methods: (a) giving each study unit weight, that is, the sim- ables included here, however, suggests that the various facets of ple mean effect size across all studies, and (b) weighting the in- personality are not differentially affected by genetic or environ- traclass r for each study by the number of twin pairs for that mental influences. study. Because there was little difference between the results obtained by using these two weighting strategies, we limit our- 1 Although these two constructs no doubt seem distinct to some read- selves to a discussion of the data obtained by using unit weight- ers, measures of activity and impulsivity were sometimes combined in ing. In both cases, means were computed following a Fisher's z, research reports and are therefore combined here. transformation. Analyses were computed twice: first for all of 2 Rose (1988) found evidence for differential heritability of personal- the data, which included nonindependent intraclass rs from ity traits on the Minnesota Multiphasic Personality Inventory; however, longitudinal and cross-sectional studies; and second for strictly Loehlin and Nichols (1976) concluded that there is little evidence for independent intraclass rs, excluding those from longitudinal differential heritability of personality on the California Psychological and cross-sectional studies. There were no important differ- Inventory. GROWING UP AND GROWING APART 229

Table 2 have been included, on the basis of some random selection pro- Mean Inlradass rs for Monozygotic (MZ) andDizygotic (DZ) cedure, we decided not to do so because we were concerned Twins From All Studies Coded about practice effects. Alternatively, we could have included the youngest age group, but this procedure would have introduced Weighted Weighted an age bias. This decision to include only independent studies by study by sample Variable type Number does not constitute any bias and is not consequential in that the and variables of studies MZ DZ MZ DZ ratio of nonindependent studies to independent studies was low. Table 3 presents the correlations between age and effect sizes Intelligence for all intelligence and personality variables, computed sepa- Verbal 29a .76 .47 .76 .48 rately for MZ and DZ twins. Note that the age ranges for all Quantitative 8 .71 .51 .74 .57 Performance 29" .71 .47 .70 .47 variables covered at least the period from 7 to 20 years; the me- Perception I) .55 .45 .55 .42 dian of mean ages across studies varied somewhat. The corre- Total IQ 42" .81 .59 .81 .61 lations of effect size and age for the intelligence variables show M» .72 .50 .72 .51 that DZ twins become more dissimilar over time than MZ twins; however, this difference is in large part accounted for by Personality the analysis of quantitative intelligence. The correlations of Activity-impulsivity 31" .58 .22 .51 .20 effect size and age for personality variables are quite similar for Aggression 8 .49 .28 .49 .29 MZ and DZ twins. Anxiety 9 .59 .18 .54 .19 The overall pattern of correlations in Table 3 is consistently Dominance 7 .50 .25 .51 .21 Emotionality 18' .54 .19 .47 .15 negative, which indicates that as twins get older they become Masculinity-femininity 9 .50 .33 .52 .36 less similar. Again, as with Table 2, there is no generally ac- Sociability 33' .50 .16 .51 .19 cepted test of significance for these rs. They should be inter- Task orientation 7 .44 .42 .17 .19 preted as effect-size estimates. M* .51 .22 .50 .22 An alternative way to examine the effect of a moderator is through contrast analysis. Contrast analysis offers the advantage " DZ results are based on I fewer studies. b Arithmetic mean calcu- of providing a significance test. To determine whether age was lated on Fisher's z-transformed rs. a moderator of similarity, age contrasts were computed separately for MZ and DZ twins within the categories of intelligence and personality, following the procedure outlined by Rosenthal Third, the rs for intelligence are consistently greater than the rs for personality. This difference may suggest that intelligence is less susceptible to environmental influences than is personal- Table 3 ity. An alternative explanation is that because personality mea- Correlations Between Intraclass rs and Age of Twins sures are typically more unreliable than intelligence measures, (From Independent Studies) the personality correlations are therefore lower. r Age Age as a Moderator Variable type Number MZ DZ Median range and variables of studies twins twins age in years The most straightforward way to examine the effect of a mod- Intelligence erator within a meta-analysis is to correlate the moderator, in Verbal 22 -.14 .10 10.0 2-84 this case mean age, with the effect size per study. It should be Quantitative 8 -.13 -.82 10.5 6-41 noted that the midpoint of the age range had to be used as the Performance 14 -.25 -.36 10 6^tl best estimate of the mean age when it was not reported. Because Perception 5 -.64 -.79 7.7 6-20 age ranges were sometimes large, this had the effect of lowering Total IQ 16* .15 -.25 7.8 1-59 the power of this analysis, thus making it relatively conservative. Mb -.22 -.50 The analyses were done twice: once using raw mean age in years and once using transformed age in order to minimize points of Personality high leverage, such that 23 years was the maximum score possi- Activity-impulsivity 14' -.48 -.33 7.6 1-50 Aggression 8 -.09 -.06 11.5 7-49 ble (i.e., any age greater than 23, e.g., 40, was coded as 23). Anxiety 5 -.34 -.49 20.4 7-30 Because the results were nearly identical for the two sets of anal- Dominance 5 .67 .07 30.0 7-50 yses, only the correlations between intraclass rs and untrans- Emotionality 8" -.11 .30 6.3 1-50 formed age are reported. Masculinity-femininity 7 -.81* -.74 16 7-50 Only independent studies (i.e., no longitudinal or cross-sec- Sociability 20° -.24 .26 16.5 3-50 Task orientation 5 -.69 -.89 28 1-50 tional studies) were used in the examination of moderator variables. This is also a conservative procedure in that some data M" -.30 -.32 points were excluded. However, this procedure does not violate Note. MZ = monozygotic; DZ = dizygotic. the independence assumption and avoids biasing results that a DZ results are based on 1 fewer study. b Arithmetic mean calculated would occur by heavily weighting a subset of longitudinal stud- on Fisher's z-transformed rs. ies. Also, although data from one age group per study could 230 K. MCCARTNEY, M. HARRIS, AND F. BERNIERI

Table 4 To summarize the results of these two analyses, both demon- Contrasts of the Effects of Twins'Age on Intraclass rs strate a general negative association between twin similarity (From Independent Studies) and age. Both analyses also demonstrate variability in patterns across variables. For example, there is a strong association be- Age contrast Z tween twin similarity and age for activity-impulsivity and for Variable type Number masculinity-femininity, whereas there is little association be- and variables of studies MZ twins DZ twins tween similarity and age for aggression and for emotionality. Intelligence 22 Verbal -0.58 0.29 Methodological Moderators Quantitative 8 -0.42 -2.28" Performance 14 -2.07** -0.90 Meta-analysis has been criticized by some who believe that Perception 5 -0.89 -0.87 all studies should not be treated "equally." This is sometimes Total IQ 16" 1.00 -1.04 referred to as the "apples and oranges" problem, following the -0.09 -0.40 adage that one cannot add apples and oranges. Wachter (1988) described this concern best: "Meta-analysis accommodates it- Personality self to a world in which bad science drives out good by weight Activity-impulsivity 14' -4.79*** -3.53*** Aggression 8 0.05 -0.39 of numbers" (p. 1407). The answer to this criticism is that Anxiety 5 -1.91* 0.38 methodological variables suspected of compromising the sci- Dominance 5 2.44** 0.15 ence, or variables suspected of altering the size of a relation, can Emotionality 8' -0.56 0.84 be examined by contrast analysis or by correlational analysis. Masculinity- femininity 7 -2.95** -1.76* There may be other variables that should be considered here. Sociability 20' -1.42 0.83 Task orientation 5 -1.75* -0.58 For example, method of zygosity determination might compro- mise the results of a twin study. In addition, some variables M* -0.34 -0.39 might moderate the relation between age and twin similarity, Note. MZ = monozygotic; DZ = dizygotic. for example, type of report (self vs. parent) and age differences " DZ results are based on 1 fewer study. " Arithmetic mean calculated in reliability of assessment. These three moderators are consid- on Fisher's z-transformed rs. ered in the following paragraphs for three dependent variables: *p<.W. **p<.05. ***/)<.01. total IQ, sociability, and activity-impulsivity. Total IQ was selected because it is the most robust measure of intelligence. So- and Rubin (1982). To create contrast weights, we simply sub- ciability and activity-impulsivity were selected because they are tracted the mean age of studies from the mean age per study, based on the largest number of studies in the personality do- such that the weights summed to zero. main. In addition, these two variables provide an interesting The results of the age contrasts are displayed in Table 4. We contrast in this context because there are age effects for one, report significant contrasts at p < .10 because each study is activity-impulsivity, but not for the other. based on many observations, which increases confidence in the Type of report. The vast majority of personality constructs estimate. There was a general pattern of negative Zs, indicating were measured by self-report. In 13 studies, personality was as- that re decreased with age. Seventeen of 26 Zs were negative and sessed by parent report. Twelve of these studies concerned tem- 8 of the 9 significant Zs were negative. A negative Z means that perament and 1 concerned sex roles. In 4 of the 12 studies, be- there is decreasing concordance between twins as they get older. havioral observations were collected in addition to parent re- The age contrast was significant for both MZ and DZ twins port. There is an association between age and report method for on two personality variables: activity-impulsivity and mascu- some measures of personality. If parents of MZ twins are prone linity-femininity. In addition, the age contrast on MZ twins was to view their children as more similar (see Jones, 1955; Koch, significant for three other personality variables: anxiety, domi- 1966), then the association between age and report method nance, and task orientation. The contrast on dominance, how- would be a confound. However, several studies have docu- ever, was positive, indicating increasing similarity for MZ twins mented a lack of bias between parents' reports on their twins over time.3 For intelligence, there was one significant age con- and zygosity. Ratings of confusability of twins are not signifi- trast for MZ twins, on performance, and one significant age cantly correlated with difference scores on four temperament contrast for DZ twins, on quantitative intelligence. traits (Plomin, Willerman, & Loehlin, 1976). Moreover, moth- Any differences between the contrast and correlational analy- ers who have misdiagnosed their twins' zygosity report neither ses are probably the result of points of high leverage in the com- greater nor smaller differences in behavioral similarity (Scarr, putation of the Pearson rs; the number of independent studies 1968). per category is relatively small despite the fact that we began with a data base of 103 studies. Although some prefer contrast 3 We plotted the relation between intraclass r and average age of sam- analysis, because it yields a significance test and because studies ple separately for MZ and DZ twins for each dependent variable. In- are weighted by number of subjects, there are advantages to the spection of the data revealed that a single point of high leverage was correlational procedure. The r between a moderator and effect responsible for the positive Z on dominance; the youngest subjects gen- sizes is itself an effect size. There is no straightforward way to erated the lowest intraclass rs on dominance in a unique behavioral rat- compute an effect-size estimate from a contrast analysis within ings task. Inspection of other plots yielded no other points of noticeably meta-analysis, because n is ambiguous. high leverage. GROWING UP AND GROWING APART 231

Nevertheless, method of report remains controversial in twin design. Similarly, sociable behaviors of 2-year-olds and 16-year- studies (Bates, 1980). For this reason, we dummy-coded olds also differ. Differences in the operationalization and reli- method of report (self vs. parent) and correlated it with MZ re ability of assessment tools for subjects of different ages are un- and DZ re for sociability and activity-impulsivity (intelligence solved problems for developmentalists who study constancy was never assessed via parent report). and change in human development (Kagan, 1980). It is for this For sociability, the correlation between method of report and reason that the statistical issues in the assessment of change have twin similarity was large and negative for the MZ twins (r = received great attention (Cronbach & Furby, 1970; Judd & —.55, p < .05), but large and positive for DZ twins (r = .48, p < Kenny, 1981; Richards, 1976). .05). This pattern shows that parent reports are more similar Zygosity determination as a moderator. Blood grouping is than self-reports for MZ twins and are less similar than self- the most reliable method of zygosity determination. When 20 reports for DZ twins. genetic markers are analyzed in blood samples, accuracy ap- For activity-impulsivity, the pattern of results was somewhat proaches 100% (Plomin et al., 1980). Less reliable methods in- different. The correlation between method of report and twin clude ratings of picture similarity, comparison of fingerprints, similarity was large and negative for both MZ twins (r = —.75, and simply asking twins whether they are identical twins. p < .01) and DZ twins (r = -.42, p > .20). For both types of Zygosity was dummy-coded to reflect blood grouping versus twins, intraclass rs were greater for parent reports than for self- other methods. We expected higher intraclass rs for MZ twins reports. It is probable that parent reports are less valid than self- and lower intraclass rs for DZ twins in studies using blood reports, because parents make two sets of ratings. grouping, because this method would tend to purify the MZ For this reason, correlations between intraclass rs and age and DZ samples. Because there was no reason to expect differ- were computed again without data from parent reports. For ac- ences in the zygosity moderator to be associated with specific tivity-impulsivity there was actually a somewhat stronger age traits, we performed contrasts on mean intelligence and mean effect (see Table 3). For the eight self-report studies of MZ twins, personality. the correlation between age and twin similarity was —.51; for For mean personality, as predicted, there was a trend for MZ the seven self-report studies of DZ twins, the corresponding cor- twins to generate higher intraclass rs in studies using blood relation was —.42. Thus, even with self-report data only, there grouping (contrast Z = 1.83, p < .10); for DZ twins, however, is decreasing twin similarity on activity-impulsivity with age. method of zygosity determination did not moderate effect sizes For sociability, there was again little association between twin (Z = .38). Similarly, for intelligence, zygosity determination did similarity and age for the self-report studies. For the 14 self- not moderate the effect sizes of either MZ (Z = -.83) or DZ report studies, the correlation between age and twin similarity (Z = .24) twins. for both MZ and DZ twins was -.02. From the present set of analyses, there is no strong evidence Age and reliability of measures. The reliability of measures to suggest the methodological superiority of blood grouping changes with age. Intelligence tests become more reliable with with respect to the prediction of intraclass rs. The cruder meth- age, with the greatest change occurring between the preschool ods of zygosity determination, which are highly correlated with and early grade-school years (Anastasi, 1982). For personality, blood grouping, perform well. however, this change is less clear. For example, test-retest correlations for the California Psychological Inventory (CPI; Gough, Components of Variance Analyses 1975), the most commonly used scale for adults in this set of studies, are comparable to those of the Emotionality, Activity, As previously discussed, the computation of components of and Sociability Scale (EAS; A. Buss & Plomin, 1984), the most variance from twin data is controversial. Criticism has focused commonly used scale for children in these studies. For sociabil- on heritability (if). In twin studies, h2 is computed by doubling ity, however, reliability is lower on the EAS for 3-year-olds (.58) the difference between MZ and DZ intraclass rs (see Plomin, than on the CPI for young adults (.68 for boys and .71 for girls). 1986, p. 15). A statistical concern with h2 is the relative unreli- For this developmental meta-analysis, increasing measure- ability of difference scores. The standard error of differences ment reliability works against the hypothesis that there is de- between correlations is larger than the standard error of a sim- creasing similarity over time. For this reason, the correlation ple correlation. In addition, there are conceptual problems as- between age and twin similarity was examined in only those sociated with the computation of A2. Scarr and Kidd (1983) studies for which the mean age of subjects was 5 years or more. have outlined three main limitations with the concept of h2: (Of course, this restricts the range of ages considered, which First, r? depends on the range of environmental variation (and may in itself affect the analyses.) The results were quite similar it is difficult to assess whether samples are indicative of popula- to those reported in Table 3. For total IQ, the r for MZ twins tions for many psychological traits); second, A2 requires quanti- was .28; for DZ twins it was -.36. For sociability, the r for MZ tative definition of a trait (this is less of a problem with meta- twins was -.16; for DZ twins it was .00. For activity-impulsiv- analysis, where an estimate would be across measures); third, ity, the r for MZ twins was - .42; for DZ twins it was - .27. Thus, r? estimates assume the genetic and environmental components age differences in assessment reliabilities do not appear to have are additive and not correlated. This final point has been dis- confounded the results of this developmental meta-analysis. puted in recent theoretical work (D. Buss, 1987; Plomin, De- A related problem for developmentalists is that the operation- Fries, & Loehlin, 1977; Scarr & McCartney, 1983)and in recent alization of constructs changes over time. For example, intelli- empirical work (Pederson, Plomin, McClearn, & Friberg, gence for a 2-year-old is tested by tasks such as bead stringing; 1988). for 16-year-olds it is tested by tasks such as mazes and block Environmental components of variance can also be esti- 232 K. MCCARTNEY, M. HARRIS, AND F. BERNIERI

Table 5 of cross-sectional study. The major advantage may be that an Correlations Between Components of Variance entire body of literature can be brought to bear on a develop- and Age of Twins mental issue. For this reason, developmental meta-analyses can be particularly useful in orienting the focus of future longitudi- Nonshared Shared nal research. Interpretation of a developmental meta-analysis is Amount of data Heritability environment environment not straightforward. Essentially, age is a correlate of effect-size All data estimates. Because age cannot be experimentally manipulated, Intelligence .36 -.37 -.15 other variables that might moderate the association between age Sociability -.46 .40 .33 and effect-size estimate must be considered. It is through addi- Activity-impulsivity .03 -.21 .48 tional moderator analyses that alternative hypotheses to those Excluding data where associated with age can be considered. Here, two such modera- mean age is 5 tors were considered: type of report and reliability of assess- years or less ment. As with all meta-analyses, variables associated with the Intelligence .52 -.50 -.28 soundness of the methodology should be considered, so that Sociability -.18 .11 .16 good studies are not outweighed by bad ones. Here, one such Activity-impulsivity -.29 .05 .42 moderator was considered: method of zygosity determination. Excluding parent report data The substantive purpose of this article has been to examine Sociability .08 -.12 .02 changes in twin similarity over time. Recent theory (e.g., Activity-impulsivity .09 -.39 .60 Plomin & Daniels, 1987) and a handful of longitudinal adoption studies suggest that the importance of the nonshared environ- Nate. The monozygotic twin age weight was used in these reported analyses. Corresponding analyses with the dizygotic twin age weight were ment increases with age. We were able to assess the relative im- nearly identical. portance of the shared and nonshared environments by recon- ceptualizing twin studies as within-family studies and by deter- mining that age contrasts in a meta-analysis could be used to mated from twin data, and these components have the same test a developmental hypothesis. The results of this meta-analy- limitations associated with them as h2. In addition, critics have sis demonstrate a developmental trend: To some extent and for questioned the validity of indirect assessments of the environ- some variables, as twins grow up, they grow apart; that is, twin ment (Wachs, 1983). For example, the shared environmental similarity decreases with age. effect (es2) is estimated by MZ, — h2; the nonshared environ- This pattern was more true for some variables than for others. mental effects (ens2) are estimated by 1 — MZ, (Plomin, 1986, In both the correlational analysis and the contrast analysis, cer- p. 69). These two environmental effects and h2 sum to 1. Again, tain domains of personality—for example, activity-impulsivity it is important to remember that an additive model is probably and masculinity-femininity—showed substantial change over simplistic. time. Other domains, such as anxiety and task orientation, To determine the source of decreasing twin similarity, age showed similar changes for MZ twins only. There was less agree- was correlated with these three components of variance for in- ment between the correlational analysis and the contrast analy- telligence, sociability, and activity-impulsivity (see Table 5). sis for intelligence than for personality. The correlations be- These analyses were computed three ways: (a) with all data, (b) tween mean age and effect size of study were comparable to with only those data in which the mean age of subjects was 5 those in the personality realm; however, the contrast Zs were years or more, and (c) excluding parent report data. These anal- small. The correlational analysis showed greater dissimilarity yses must be viewed with caution because of the limitations, over time for DZ twins on overall intelligence. This pattern is outlined earlier, associated with components of variance analy- in contrast to the findings for some indexes of personality. ses. In addition, it is not clear how reliable these analyses are, There were some moderators of effects. For example, parent especially when they are based on so few data points. These reports appear to bias ratings of temperament. Another moder- analyses are presented primarily for heuristic purposes. ator, sex, should be considered in future work, in light of Rose's For intelligence, h2 increased over time, more so when the less (1988) finding of a Gender X Age interaction on 4 of 9 personal- reliable preschool data were excluded. Conversely, es2, and to a ity variables. Other moderators studied here, such as age differ- lesser extent ens2, decreased over time. For sociability, the re- ences in reliability and method of zygosity determination, ap- sults changed when either the preschool data or the parent re- peared to have little, if any, effect. However, power in most anal- port data were excluded. In these cases, there was little evidence yses was low. Although 103 studies were coded in this for change of any sort. For activity-impulsivity, ens2 increased investigation, more than half of the 13 variables were assessed over time and es2 decreased over time, especially for the analy- in fewer than 20 studies. This limitation weakens somewhat the ses of self-report data. robustness of the findings. Nevertheless, it must be remembered that each data point is based on multiple subjects in a study. Discussion This robustness was demonstrated by the consistency between the average effect-size estimates reported here and the The methodological purpose of this article has been to dem- median rs reported by Bouchard and McGue (1981) for intelli- onstrate the utility of developmental meta-analysis, where age gence and by Nichols (1978) for personality and intelligence. is used as a moderator of effect-size estimates. Developmental Nichols's observation about the similarity between the two meta-analysis shares many of the advantages and disadvantages kinds of twins across traits is relevant. His data and our data GROWING UP AND GROWING APART 233 show that "a twin study is likely to find a difference between time for both MZ and DZ twins, although the contrast analysis identical and fraternal correlations of about .20 regardless of shows many more significant effects for MZ twins. the domain or the trait that is being investigated!" (Nichols, An ad hoc explanation for why MZ twins would become 1978, p. 163). For this reason, we believe that in the future, more dissimilar over time than DZ twins is the greater need of singular studies of twins conducted for the sole purpose of ob- MZ twins to appear different. Schachter (1982) introduced the taining heritability estimates on specific psychological indexes term sibling deidentification to describe a process through are unlikely to make any significant contribution. which family members might define siblings as different or con- Certainly, most twin and adoption studies show that the envi- trasting. Schachter and Stone (1985) found evidence for deiden- ronment accounts for a substantial portion of variance (Loeh- tification among same-sex siblings and among first-sibling pairs. lin, Willerman, and Horn, 1988). Our purpose here has not Perhaps this happens because these children are more likely to been to compute components of variance. However, age was be compared with one another. Similarly, MZ twins, who are correlated with components of variance in order to examine the most likely of all sibling pairs to be compared with one an- developmental patterns. Two speculations can be made based other, may be more likely to deidentify than DZ twins, at least on these exploratory analyses. First, heritability seems to in- on some measures of personality. crease over time for IQ. This may reflect an age-related increase Thus, because there is little evidence that DZ twins become in MZ similarity and the age-related decrease in DZ similarity. more dissimilar over time than MZ twins, it appears that non- This pattern was found in the longitudinal Louisville twin study systematic environmental effects might play a more dominant (Wilson, 1983); however, we found it for total IQ only. Second, role than systematic environmental effects. The implications of for IQ and activity-impulsivity the importance of the shared this conclusion for future research are challenging. One ap- environment decreased over time; the nonshared environment proach might be to target certain environmental effects that increased over time for activity-impulsivity. It was difficult to would seem to be unrelated to genetic predispositions or shared analyze age-related change in components of variance with re- family experience, for example, being randomly placed in a spect to sociability, because the pattern changed when the par- classroom with a particularly skilled teacher or being hospital- ent report data were eliminated. ized because of injury. Thus, the results of this developmental meta-analysis of twin A challenge also lies in the nearly certain fact that many de- studies support recent developmental theory and recent family velopmental processes operate simultaneously. For example, research in that they demonstrate the importance of nonshared the shared and nonshared environments each affect develop- experiences generally (e.g., Plomin & Daniels, 1987). This is ment, to some extent. In addition, siblings may both identify especially impressive in this context: Twin studies probably and deidentify, to some extent. Moreover, these processes may overestimate the importance of the shared environment, be- interact differently for different traits and abilities during cause twins no doubt experience more similar environments different developmental periods. As we search for the methods than do other siblings (Plomin, 1986). In other words, this pat- to study multiple developmental processes, developmental tern may be greater for nontwin siblings. meta-analysis will have increasing utility. Patterns associated with nonshared environmental effects may be complicated by the fact that there are probably two dis- References tinct types. The nonshared environment consists of both non- Anastasi, A. (1982). Psychological testing. New York: MacMillan. systematic experiences (stochastic events) and systematic expe- Bates, J. E. (1980). The concept of difficult temperament. Merrill- riences. It may be that nonsystematic experiences lead to Palmer Quarterly, 26, 299-319. differences and systematic experiences increase similarity when Bouchard, T. J., Jr., & McGue, M. (1981). Familial studies of intelli- genetic predispositions are shared. gence: A review. Science, 212, 1055-1059. There are some data on differential parental treatment to Buss, A., & Plomin, R. (1984). Temperament: Early developing person- support this thesis. Lytton's (1977) observations of parent- ality traits. Hillsdale, NJ: Erlbaum. child interaction with both MZ and DZ twins suggest that par- Buss, D. (1987). Selection, evocation, and manipulation. Journal of Per- sonality and Social Psychology, 53, 1214-1221. ems respond to differences between the twins based on actual Cohen, J. (1977). Statistical power analysis for the behavioral sciences. rather than perceived zygosity. Dunn, Plomin, and Nettles New York: Academic Press. (1985) found greater differences in maternal affection between Cronbach, L. J. (1984). Essentials of psychological testing. New York: adopted siblings, who are not genetically related, than between Harper & Row. biological siblings. Cronbach, L. J., & Furby, L. (1970). How we should measure Yet if people make their own environments through probabi- "change"—or should we? Psychological Bulletin, 74, 68-80. listic associations with genetic predispositions, as Scarr and Mc- Dunn, J., Plomin, R., & Nettles, M. (1985). Consistency of mothers' Cartney (1983) have suggested, then one might expect that MZ behavior toward infant siblings. Developmental Psychology, 27,1188- twins would become less dissimilar over time than DZ twins, 1195. because MZ twins would be expected to create similar niches Goldsmith, H. H. (1983). Genetic influences on personality from infancy to adulthood. Child Development, 54, 331-355. in life. The results of this meta-analysis are mixed on this issue. Gough, H. G. (1975). Manual for the California Psychological Inven- Although the correlational analysis shows that DZ twins be- tory. Palo Alto, CA: Consulting Psychologists Press. come more dissimilar over time on intelligence, the contrast Hall, J. A. (1987). Nonverbal sex differences. Communications accuracy analysis shows little evidence for change, in one direction or the and expressive style. Baltimore: Johns Hopkins University Press. other, for either MZ or DZ twins. For personality, the correla- Hearnshaw, L. (1979). Cyril Bwt. psychologist. Ithaca, NY: Cornell tional analysis reveals a similar pattern of dissimilarity over University Press. 234 K. MCCARTNEY, M. HARRIS, AND F. BERNIERI

Hoffman, L. W. (1985). The changing genetics/socialization balance. Plomin, R., DeFries, J. C., & McClearn, G. E. (1980). Behavioral genet- Journal of Social Issues, 41, 127-148. ics: A primer. San Francisco: Freeman. Jones, H. E. (1955). Perceived differences among twins. Eugenics Quar- Plomin, R., Willerman, L., & Loehlin, J. C. (1976). Resemblance in terly, 5, 98-102. appearance and the equal environments assumption in twin studies Judd, C. M., & Kenny, D. A. (1981). Estimating the effects of social of personality. Behavior Genetics, 6, 43-52. interventions. New York: Cambridge University Press. Richards, J. M., Jr. (1976). A simular study comparing procedures for Kagan, 3. (1980). Perspectives on continuity. In D. G. Grin, Jr. & J. assessing individual education growth. Journal of Educational Psy- Kagan (Eds.), Constancy and change in human development (pp. 26- chology. 68, 603-612. 74). Cambridge: Harvard University Press. Rose, R. J. (1988). Genetic and environmental variance in content di- Kagan, J., Reznick, J. S., & Snidman, N. (1988). Biological bases of mensions of the MMPI. Journal of Personality and Social Psychology. childhood shyness. Science, 240, 167-171. 55,302-311. Kent, J. (1985). Genetic and environmental contributions to cognitive Rosenthal, R. (1984). Meta-analytic procedures for social research. Bev- abilities as assessed by a telephone test battery- Unpublished doctoral erly Hills, CA: Sage. dissertation. University of Colorado, Boulder. Rosenthal, R., & Rubin, D. B. (1982). Comparing effect sizes of inde- Koch, H. L. (1966). Twins and twin relations. Chicago: University of pendent studies. Psychological Bulletin, 92, 500-504. Chicago Press. Rowe, D., & Plomin. R. (1981). The importance of nonshared (E,) en- Loehlin, J. C., & Nichols, R. C. (1976). Heredity, environment, and per- vironmental influences in behavioral development. Developmental sonality. Austin: University of Texas Press. Psychology, 77,517-531. Loehlin, J. C., Willerman, L., & Horn, J. M. (1987). Personality resem- Scarr, S. (1968). Environmental bias in twin studies. Eugenics Quar- blance in adoptive families: A 10-year follow-up. Journal of Personal- terly, 15, 34-40. ity and Social Psychology; 53, 961-969. Scarr, S., & Kidd, K. K. (1983). Developmental behavior genetics. In Loehlin, J. C., Willerman, L., & Horn, J. M. (1988). Human behavior M. M. Haith & J. J. Campos (Eds.), Handbook of child psychology, genetics. Annual Review of Psychology, 39, 101-133. Vol. //(pp. 345-434). New York: Wiley. Lytton, H. (1977). Do parents create, or respond to, differences in Scarr, S., & McCartney, K. (1983). How people make their own environ- twins? Developmental Psychology, 13, 456-459. ments: A theory of genotype -*• environment effects. Child Develop- Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of ment, 54, 424-435. the family: Parent-child interaction. In E. M. Heatherington (Ed.), Scarr, S., & Weinberg, R. A. (1978). The influence of "family back- Handbook of child psychology. Vol. IV(pp. 1-102). New York: Wiley. ground" on intellectual attainment. American Sociological Review, McCall, R. B. (1977). Challenges to a science of developmental psychol- 43. 674-692. ogy. Child Development, 48, 333-344. Schachter, F. F. (1982). Sibling deidentification and split-parent identi- McCall. R. (1983). Environmental effects on intellience: The forgotten fication: A family tetrad. In M. E. Lamb & B. Sutton-Smith (Eds.), realm of discontinuous nonshared within-family factors. Child De- Sibling relationships: Their nature and significance across the life- velopment, 54, 408-415. span(pp. 123-151). Hillsdale, NJ: Erlbaum. Nichols, R. C. {1978). Heredity and environment: Major findings from twin studies of ability, personality and interests. Homo, 2V, 158-173. Schachter, F. F, & Stone, T. K. (1985). Difficult sibling, easy sibling: Pederson, N. L., Plomin, R., McClearn, G. E., & Friberg, L. (1988). Temperament and the within-family environment. Child Develop- Ncuroticism, extroversion, and related traits in adult twins reared ment. 56, 1335-1344. apart and reared together. Journal of Personality and Social Psychol- Tellegen, A., Lykken, D. T, Bouchard, T. J., Wilcox, K. J., Segal, N. L., ogy. 55, 950-957. & Rich, S. (1988). Personality similarity in twins reared apart and Plomin, R. (1986). Development, genetics, and psychology. Hillsdale, together. Journal of Personality and Social Psychology, 54, 1031- NJ: Erlbaum. 1039. Plomin, R., & Daniels, D. (1987). Why are children in the same family Wachs, T. D. (1983). The use and abuse of environment in behavior- so different from one another? The Behavioral and Brain Sciences, genetic research. Child Development, 54, 416-423. 10, 1-60. Wachter, K. W. (1988). Disturbed by meta-analysis? Science. 241. Plomin, R., DeFries, J. C., & Loehlin, J. C. (1977). Genotype-environ- 1407-1408. ment interaction and correlation in the analysis of human behavior. Wilson, R. S. (1983). The Louisville twin study: Developmental syn- Psychological Bulletin, 84, 309-322. chronies in behavior. Child Development, 54, 298-316. GROWING UP AND GROWING APART 235

Appendix

Studies Used in the Meta Analysis

Adams, B,, Ghodsian, M., & Richardson, K. (1976). Evidence for a low Halperin, S. L., Rao, D. C, & Morton, N. E. (1975). A twin study of upper limit of heritability of mental test performance in a national intelligence in Russia. Behavior Genetics, 5, 83-86. sample of twins. Nature, 263, 314-316. Hamilton, J., Blewett, D., & Sydiaha, D. (1971). Ink-blot responses of Austin, S. V, & Hemsley, D. R. (1980). Genetic influences on measures identical and fraternal twins. Journal of Genetic Psychology, 119, 37- of selective attention and speed: A twin study. Behavior Genetics, 10, 41. 309-315. Harris, E. L. (1982). Genetic and environmental influences on reading Basil, A. (1972). A Rorschach study of personality development in iden- achievement: A study of first- and second-grade twin children. Acta tical twins. Journal of Personality Assessment, 36, 23-27. Geneticae Medicae et Gemellohgiae: Twin Research, 31((-2), 64- Bettner, L. G., & Blum, J. E. (1972). Kent-Rosanoff Free Association 116. Test in aged twins with and without organic brain syndrome. Proceed- Ho, H., Foch, T. T, & Plomin, R. (1980). Developmental stability of ings of the Annual Convention of the American Psychological Associa- the relative influence of genes and environment on specific cognitive tion, 7.651-652. abilities during childhood. Developmental Psychology, 16, 340-346. Bettner, L. G., Blum, J. E., & Jarvik, L. F. (1982). Kent-Rosanoff Word Horn, J. M., Plomin, R., & Rosenman, R. (1976). Heritability of per- Association Test: Aged twins with and without dementia. Ada Geneti- sonality traits in adult male twins. Behavior Genetics, 6, 17-30. caeMedicae et Gemellohgiae: Twin Research, 31(1-2), 1-7. Jardine, R., & Martin, N. G. (1984). No evidence for sex-linked or sex- Bettner, L. G., Jarvik, L. F., & Blum, J. E. (1971). Stroop Color-Word limited gene expression influencing spatial orientation. Behavior Ge- test, non-psychotic organic brain syndrome, and chromosome loss in netics, 14, 345-354. aged twins. Journal of Gerontology, 26, 458-469. Jarvik, L. F., Blum, J. E., & Varma, A. O. (1972). Genetic components Brooks, J., & Lewis, M. (1974). Attachment behavior in thirteen-month and intellectual functioning during senescence: A 20-year study of old, opposite-sex twins. Child Development, 45, 243-247. aging twins. Behavior Genetics, 2, 159-171. Buss, A. H., Plomin, R., & Willerman, L. (1973). The inheritance of Jensen, A. R. (1970). IQ's of identical twins reared apart. Behavior Ge- temperament. Journal of Personality, 41,513-524. netics, 1, 133-148. Cohen, D. J., Dibble, E., & Grawe, J. M. (1977). Fathers' and mothers' Jensen, A. R., & Marisi, D. Q. (1979). A note on the heritability of perceptions of children's personality. Archives of General Psychiatry, memory span. Behavior Genetics, 9, 379-387. 34, 480-487. Koskenvuo, M., Langinvainio, H., Kaprio, i., & Sarna, S, (1984). Dworkin, R. H. (1979). Genetic and environmental influences on per- Health related psychosocial correlates of neuroticism: A study of son-ituation interactions. Journal of Research in Personality, 13, adult male twins in Finland. Acla Genetica Medicae el Gemetlologiae: 279-293. Twin Research, 33, 307-320. Dworkin, R. H., Burke, B. W., Maher, B. A., & Gottesman, 1.1. (1976). Laginvainio, H., Kaprio, J., Koskenvuo, M., & Lonnqvist, J. (1984). A longitudinal study of the genetics of personality. Journal of Person- Finnish twins reared apart: III: Personality factors. Acta Geneticae ality and Social Psychology, 34, 510-518. Medicae et Gemellologiae: Twin Research, 33, 259-264. Dworkin, R. H., Burke, B. W., Maher, B. A., & Gottesman, 1.1. (1977). Loehlin, J. C. (1982). Are personality traits differentially heritable? Be- Genetic influences on the organization and development of personal- havior Genetics, 12, 417-428. ity. Developmental Psychology, 13, 164-165. Lykken, D. T. (1982). Research with twins: The concept of emergenesis. Elizabeth, P. H., & Green, R. (1984). Childhood sex-role behaviors: Psychophysiology, 19, 361-373. Similarities and differences in twins. Acla Genelicae Medicae et Ge- mellohgiae: Twin Research, 33, 173-179. Lytton, H. (1977). Do parents create, or respond to, differences in twins'? Developmental Psychology, 13, 456-459. Fischbein, S. (1980). IQand social class. Intelligence, 4, 51-63. Fischbein, S. (1984). Self- and teacher-rated school adjustment in Lytton, H., Martin, N. G., & Eaves, L. (1977). Environmental and ge- monozygotic and dizygotic twins. Ada Geneticae Medicae et Gemel- netic causes of variation in ethological aspects of behavior in two- lologiae: Twin Research, 33, 205-212. year-old boys. Social Biology, 24, 200-211. Floderus-Myrhed, B., Pederson, N., & Rasmuson, I. (1980). Assessment Marsh, R. W. (1980). The significance for intelligence of differences in of heritability for personality based on a short form of the Eysenck birth-weight and health within monozygotic twin pairs. British Jour- Personality Inventory: A study of 12,898 twin pairs. Behavior Genet- nal of Psychology, 71, 63-67. ics, 10, 153-162. Matheny, A. P. (1975). Twins: Concordance for Piagetian-equivalent Foch, T. T, & Plomin, R. (1980). Specific cognitive abilities in 5- to 12- items derived from the Bayley Mental Test. Developmental Psychol- year-old twins. Behavior Genetics, 10, 507-520. ogy, 11,224-227. Garfinkle, A. S. (1982). Genetic and environmental influences on the Matheny, A. P. (1980). Bayley's Infant Behavior Record: Behavioral development of Piagetian logico-mathemetical concepts and other components and twin analyses. Child Development, 51,1157-1167. specific cognitive abilities. Ada Geneticae Medicae et Gemellologicae: Matheny, A. P. (1983). A longitudinal twin study of stability of compo- Twin Research, 31, 10-61. nents from Bayley's Infant Behavior Record. Child Development, 54, Goldsmith, H. H., & Gottesman, 1.1. (1977). An extension of construct 356-360. validity for personality scales using twin-based criteria. Journal of Matheny, A. P. (1984). Twin similarity in the development of transfor- Research in Personality, 11, 381-397. mations of infant temperament as measured in a multi-method, lon- Goldsmith, H. H., & Gottesman, 1.1. (1981). Origins of variation in gitudinal study. Acta Geneticae Medicae et Gemellologiae: Twin Re- behavioral style: A longitudinal study of temperament in young search, 33, 181-189. twins. Child Development, 52, 91-103. Matheny, A. P., & Dolan, A. B. (1974). A twin study of genetic influ- Graham, P. J., & Stevenson, J. (1985). A twin study of genetic influences ences in reading achievement. Journal of Learning Disabilities, 7, on behavioral deviance. Journal of the American Academy of Child 99-102. Psychiatry, 24, 33-41. Matheny, A. P., & Dolan, A. B. (1975). Persons, situations, and time: A 236 K. MCCARTNEY, M. HARRIS, AND F. BERNIERI

genetic view of behavioral change in children. Journal of Personality personality in childhood. Journal of Personality and Social Psychol- & Social Psychology, 52,1106-1110. ogy, 39, 680-688. Matheny, A. P., & Dolan, A. B. (1980). A twin study of personality and Plomin, R., Foch, T. T, & Rowe, D. C. (1981). Bobo clown aggression temperament during middle childhood. Journal of Research in Per- in childhood: Environment, not genes. Journal of Research in Per- sonality. 14, 224-234. sonality, 15. 331-342. Matheny, A. P., Dolan, A. B., & Wilson, R. S. (1976). Twins: Within- Plomin, R., & Rowe, D. C. (1977). A twin study of temperament in pair similarity on Bayley's Infant Behavior Record. Journal of Genetic young children. Journal of Psychology, 97, 107-113. Psychology, 128, 263-270. Plomin, R., & Rowe, D. C. (1978). A twin study of temperament in Mather, P. L., & Black, K. N. (1984). Hereditary and environmental young children. Annual Progress in Child Psychiatry and Child De- influences on preschool twins1 language skills. Developmental Psy- velopment, 216-222. chology, 20, 303-308. Plomin, R., & Rowe, D. C. (1979). Genetic and environmental etiology Matthews, K. A., Batson, C. D., Horn, J., & Rosenman, R. H. (1981). of social behavior in infancy. Developmental Psychology, 15, 62-72. "Principles in his nature which interest him in the fortune of others Plomin, R., & Vandenberg, S. G. (1980). An analysis of Koch's (1966) . . ":Theheritabilityofempathicconcernforothers../0W7W/0/№-- Primary Mental Abilities Test data for 5- to 7-year-old twins. Behav- soHallty. 49, 237-247. ior Genetics, 70,409-412. Matthews, K. A., & Krantz, D. S. (1976). Resemblances of twins and Plomin, R., & Willerman, L. (1975). A cotwin control study and a twin their parents in Pattern A behavior. Psychosomatic Medicine, 38, study of reflection-impulsivity in children. Journal of Educational 140-144. Psychology, 67,537-543. McGue, M., Bouchard, T. J., Lykken, D. T, & Fever, D. (1984). Infor- Pogue-Geile, M. F., & Rose, R. J. (1985). Developmental genetic studies mation processing abilities in twins reared apart. Intelligence, 8,239- of adult personality. Developmental Psychology, 21, 547-557. 258. Price, R. A., Vandenberg, S. G., Iyer, H., & Williams, J. S. (1982). Com- Miller, J. Z., & Rose, R. J. (1982). Familial resemblance in locus of ponents of variation in normal personality. Journal of Personality and control: A twin-family study of the internal-external scale. Journal Social Psychology, 43, 328-340. of Personality and Social Psychology, 42, 534-540. Rahe, R. H., Hervig, L., & Rosenman, R. H. (1978). Heritability of Morgan, A. H. (1973). The heritability of hypnotic susceptibility in Type A behavior. Psycosomalic Medicine, 40, 478-486. twins. Journal of Abnormal Psychology, 82, 55-61. Reznikoff, M., Domino, G., Bridges, C., & Honeyman, M. (1973). Cre- Munsinger, H., & Douglass, A. (1976). The syntactic abilities of identi- ative abilities in identical and fraternal twins. Behavior Genetics, 3, cal twins, fraternal twins, and their siblings. Child Development, 47, 365-377. 40-50. Riese, M. L., Wilson, R. S., & Matheny, A. P. (1985). Multimethod Murawski, B. J. (1971). Genetic factors in tests of perception and the assessment of temperament in twins: Birth to six months. Acta Gene- Rorschach. Journal of Genetic Psychology, 119, 43-52. ticae Medicae et Gemellologiae: Twin Research, 34, 15-31. Nathan, M., &Guttman, R. (1984). Similarities in test scores and pro- Rose, R. J., & Ditto, W. B. (1983). A developmental-genetic analysis files of kibbutz twins and singletons. Aaa Geneticae Medicae el Ge- of common fears from early adolescence to early adulthood. Child mellologiae: Twin Research, 3 3, 213-218. Development, 54, 317-330. Nichols, P. L., & Broman, S. H. (1974). Familial resemblance in infant Rose, R. J., Harris, E. L., Christian, J. C., & Nance, W. E. (1979). Ge- mental development. Developmental Psychology. 10, 442-446. netic variance in nonverbal intelligence: Data from the kinship of O'Connor, M., Foch, T, Sherry, T, & Plomin, R. (1980). A twin study identical twins. Science, 205, 1153-1155. of specific behavioral problems of socialization as viewed by parents. Rowe, D. C. (1982a). Monozygotic twin cross-correlations as a valida- Journal of Abnormal Child Psychology, S, 189-199. tion of personality structure: A test of the semantic bias hypothesis. Osborne, R. T, & Gregor, A. J. (1966). Hereditary factors in the perfor- Journal of Personality and Social Psychology, 43, 1072-1079. mance of visual perceptual tasks. Revue Internationale DC Sociolo- Rowe, D. C. (19S2b). Sources of variability in sex-linked personality gie, 2, 274-282. attributes: A twin study. Developmental Psychology, 18,431-434. Osborne, R. T., Gregor, A. J., & Miele, F. (1967). Heritability of numeri- Rowe, D. C. (1983). Biometrical genetic models of self-reported delin- cal facility. Perceptual and Motor Skills. 24. 659-666. quent behavior: A twin study. Behavior Genetics. 13, 473-489. Osborne, R. T, Gregor, A. J., & Miele, F. (1968). Heritability of Factor Rushton, J. P., Fulker, D. W., Neale, M. C, Blizard, R. A., & Eysenck, V: Verbal comprehension. Perceptual and Motor Skills. 26, 191-202. H. J. (1984). Altruism and genetics. Acta Geneticae Medicae et Ge- Osborne, R. X, & Suddick, D. (1973). Stability of IQ differences of mellologiae: Twin Research, 33, 265-271. twins between ages twelve and twenty. Psychological Reports, 32, Salzano, F. M., & Rao, D. C. (1976). Path analysis of aptitude, personal- 1096-1098. ity, and achievement scores in Brazilian twins. Behavior Genetics, 6, Owen, D. R., & Sines, J. Q (1970). Heritability of personality in chil- 461-466. dren. Behavior Genetics, 1, 235-248. Scarr, S. (1982). Effects of birth weight on later intelligence. Social Biol- Pedersen, N. L., McClearn, G. £., Plomin, R., & Friberg, L. (1985). ogy, 29, 230-237. Separated fraternal twins: Resemblance for cognitive abilities. Behav- Stevenson, J., & Fielding, J. (1985). Ratings of temperament in families ior Genetics. 15, 407-419. of young twins. British Journal of Developmental Psychology, 3, 143- Pencavel, J. H. (1976). A note on the IQ of monozygotic twins raised 152. apart and the order of their birth. Behavior Genetics, 6, 455-460. Tambs, K., Sundet, J. M., & Magnus, P. (1984). Heritability analysis of Perry, A. (1973). The effect of heredity on attitudes toward alcohol, ciga- the WAIS subtests: A study of twins. Intelligence, S, 283-293. rettes, and coffee. Archives of General Psychiatry, 37, 1272-1277. Torgersen, A. M. (1982). Genetic factors in temperamental individual- Pezzullo, T. R., Thoisen, E. E.. & Madaus, G. F. (1972). The heritability ity: A longitudinal study of same-sexcd twins from 2 months to 6 of Jensen's Level I and II divergent thinking. American Educational years of age. Annual Progress in Child Psychiatry and Child Develop- Research Journal, 9. 539-546. ment, 219-228. Plomin, R. (1976). A twin and family study of personality in young Torgersen, S. (1979). The nature and origin of common phobic fears. children. Journal of Psychology, 94, 233-235. British Journal of Psychiatry, 134, 343-351. Plomin, R., & Foch, T. T. (1980). A twin study of objectively assessed Torgersen, S. (1980). The oral, obsessive, and hysterical personality syn- GROWING UP AND GROWING APART 237

dromes: A study of hereditary and environmental factors by means Wilson, R. S. (1981). Mental development: Concordance for same-sex of the twin method. Archives of General Psychiatry. 37. 1272-1277. and opposite-sex dizygotic twins. Developmental Psychology, 17, Volger, G. P., & Fulker, D. W. (1983). Familial resemblance for educa- 626-629. tional attainment. Behavior Genetics. 13, 341-354. Wilson, R. S. (1983). The Louisville twin study: Developmental syn- Willerman, L. (1973). Activity level and hyperactivity in twins. Child chronies in behavior. Child Development, 54. 298-316. Development, 44, 288-293. Wilson, R. S. (1984). Twins and chromogenetics: Correlated pathways Willerman, L., & Churchill, J. A. (1967). Intelligence and birth weight of development. Ada Geneticae Medicae el Gemellologiae: Twin Re- in identical twins. Child Development, 38, 623-629. search, 33, 149-157. Wilson, G. D., Rust, J., & Kasriel, J. (1977). Genetic and family origins Wilson, R. S., & Harping, E. B. (1972). Mental and motor development of humor preferences: A twin study. Psychological Reports. 41, 659- in infant twins. Developmental Psychology, 7, 277-287. 660. Young, J. P., Fenton, G. W, & Lader, M. H. (1971). The inheritance of Wilson, R. S. (1974). Twins: Mental development in the preschool years. neurotic traits: A twin study of the Middlesex Hospital Question- Developmental Psychology, 10, 580-588. naire. British Journal of Psychiatry, 119, 393-398. Wilson, R. S. (1975). Twins: Patterns of cognitive development as measured on the Wechsler Preschool and Primary Scale of Intelligence. Developmental Psychology, 11, 126-134. Received July 28, 1988 Wilson, R. S. (1977). Twins and siblings: Concordance for school-age Revision received May 29, 1989 mental development. Child Development, 48, 211 -216. Accepted June 16,1989 •