Separated Twins and the Genetics of Personality Differences: A Critique
Author(s): Jay Joseph
Source: The American Journal of Psychology, Vol. 114, No. 1 (Spring, 2001), pp. 1-30
Published by: University of Illinois Press
Stable URL: http://www.jstor.org/stable/1423378
Accessed: 04/09/2008 17:25
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=illinois.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].
http://www.jstor.org
twinsandthe geneticsof
Separated
A critique
personalitydifferences:
JAYJOSEPH
California School of Professional Psychology, Alameda
This article discussesstudies of separatedtwins,with special emphasison the
Minnesota Study of Twins Reared Apart (MISTRA), to determine whether they
support the existence of an importantgenetic component in behavioraland
personality differences. The methods and conclusions of the
MISTRA
team are
discussedin the context of earlierstudies of separatedidentical twins.I argue
that volunteer-basedstudies are biased towardgreater twin similarity.In addiresearchteam did not publish or share rawdata and case histion, the MISTRA
tory information. Reared-togetherand reared-apartmonozygotic twins share
importantenvironmentalsimilaritiesnot controlled for by comparingpersonality correlations.I propose an alternativecontrol group consisting of biologically unrelated pairs of strangersmatched on all environmentalfactors common to pairsof separatedmonozygotictwins.I conclude that the evidencefrom
studies of twins reared apart does not support the role of genetic factors in
personalityand behavioraldifferences.
This article discusses the findings and conclusions of studies of separated twins, with a special emphasis on the Minnesota Study of Twins
Reared Apart (MISTRA), a project that began in 1979 whose primary
investigators include T. Bouchard, M. McGue, D. Lykken, N. Segal, A.
Tellegen, and L. Heston. The most talked about Minnesota studies have
looked at two aspects of human functioning: cognitive ability and personality. Here, the discussion is limited to articles looking at personality and behavioral differences, although many observations are clearly
relevant to the IQ studies as well.
The conclusions of Bouchard and associates have been widely reported, and their participants have been the subject of stories in the popular press about separated identical twins who discover their supposedly
remarkable similarities upon being reunited-the most well-known case
being the so-called Jim Twins. A Life magazine cover story on genetics
tided "WereYou Born That Way?"(Colt, 1998) answers this question in
the affirmative. L. Wright (1997), the author of a book on twin studying intended mainly for the nonprofessional public, wrote, "The field
of psychology has been shaken by separated-twin studies" (p. 8), and
"The science of behavioral genetics, largely through twin studies, has
AMERICANJOURNAL OF PSYCHOLOGY
Spring 2001, Vol. 114, No. 1, pp. 1-30
? 2001 by the Board of Trustees of the University of Illinois
2
JOSEPH
made a persuasive case that much of our identity is stamped on us from
conception" (p. 143). Bouchard (1994, p. 1701) himself wrote, "The
similarity we see in personality between biological relatives is almost
entirely genetic in origin."
However, the studies of separated twins contain serious flaws, and
their authors' conclusions are questionable. But before examining recent articles published by the MISTRA group, it is necessary to look at
previous studies of separated identical twins and note their shortcomings. The terms monozygotic,or identical, twins rearedapart (MZA) and
dizygotic,orfraternal, twins rearedapart (DZA) are used here for simplicity, although they denote the researchers' classifications and do not
necessarily imply that these twins can be legitimately considered separated. MZTand DZTrefer to monozygotic and dizygotic pairs who were
reared together in the same home.
Previous studies of separated identical twins
The classical twin method compares the correlations or concordance
rates of MZT twins with the same measures of same-sex DZT twins. Because identical twins share 100% of the same genes, whereas fraternals
share only 50% on average, the greater MZ similarity has been claimed
as evidence of the operation of genetic factors for the trait in question.
This claim is based on the assumption that both types of twins experience similar environments. The questionable validity of the MZT-DZT
equal environment assumption has been the most common basis for
criticism of the twin method (seeJackson, 1960;Joseph, 1998, in pressc, 2000; Kamin, 1974; Lewontin, Rose, & Kamin, 1984; Pam, Kemker,
Ross, & Golden, 1996; Smith, 1965). These authors argued that conclusions drawn from the twin method are confounded by the greater environmental similarity of MZTsvs. DZTs.
But what about identical twins separated at birth and raised in disparate socioeconomic and cultural environments? Although such pairs
appear to constitute an ideal test of the nature versus nurture question,
they are quite rare. Most cases of "separated" identical twins in the literature suffer from one or more of the following problems: The twins
were separated only after having been raised together for several years,
they were raised by different members of the same family, they were
placed into families correlated for socioeconomic status, they were
aware of each other's existence and had frequent contact during much
of their lives, they were brought to the attention of researchers on the
basis of their similarity or knowledge of each other's existence, the
material used to evaluate the similarities of twins was collected by the
same researchers, and personality and environmental similarities were
not assessed by blinded raters. In Farber's 1981 review of the 121 cases
SEPARATED IDENTICAL TWINS
3
of separated MZ twins reported to that point, she found only 3 cases
in which the twins were separated during the first year of life, were
reared with no knowledge that they had a twin, and were seen at the
time of their first meeting.
It is important to understand that a study of separated twins is both
a twin study and an adoption study because one or both twins have been
removed from their biological mother and placed into an adoptive or
foster family. A critical assumption of all adoption studies is that placements are random and are not based on the socioeconomic or psychiatric status of the biological family (Joseph, 1999a, 1999b). In the case
of separated twin studies, it must be shown that trait similarities were
not influenced by the possibly correlated environments of the twin pair.
Genetic inferences from studies of twins reared together depend on the
assumption of the rough equality of environments experienced by MZ
and DZ twins. Genetic inferences from studies of separated twins rest
on the assumption that the twins' environments were not systematically more similar than a group of randomly selected unrelated paired
individuals; in contrast to the study of twins reared together, they are
based on an unequal environment assumption. In fact, most studied
MZAs were reared in similar socioeconomic conditions. Farber (1981,
p. 62) observed that about 90% of MZAs were born into poor families
and that "anywhere from 50 to 75 percent of the twins were reared in
clearly deprived homes.... Only two or three individuals were adopted into professional families." It is common for separated twins to be
raised by members of the same family, often with the biological parents
caring for one of the twins.
Most systematic or individual case history studies of separated twin
pairs have recruited MZAs on the basis of media appeal or have located twins who have come to the attention of researchers precisely because
of their similarity. Looking at all studies published through 1980, Farber (1981) noted that about 90% of separated MZ twins were recruited on the basis of similarity. Subsequent claims about the similarities
of these twins were for Farber a form of circular reasoning. OnlyJuelNielsen's (1965/1980) study avoided this bias because his 12 pairs were
identified from a national twin register. Two additional Scandinavian
register-based separated twin studies were reported in the 1980s and are
discussed later in this article.
Through 1981, there were only three systematic studies of separated
identical twins: Newman, Freeman, and Holzinger (1937), who studied
19 pairs; Shields (1962), who studied 44; andJuel-Nielsen (1965/1980),
who investigated 12 pairs. The problems with these studies have been
chronicled in the reviews by Kamin (1974) and Taylor (1980), who were
interested in these reports as they related to IQ and in the exhaustive
4
JOSEPH
study by Farber (1981). The work of British psychologist Cyril Burt, who
claimed to have given IQ tests to 53 separated pairs, is widely believed
to be unreliable and possibly fabricated (Bouchard, 1984; Gould, 1981;
Hearnshaw, 1979; Lewontin, Rose, & Kamin, 1984), and his figures are
not counted by most modern researchers (e.g., Bouchard & McGue,
1981).
Newman et al. (1937). Working in the Chicago area, Newman and
associates (1937) recruited 19 MZA pairs (12 female, 7 male), who were
compared with a control group of 50 MZTs. As Kamin (1974) noted,
MZAs were recruited on the basis of similarity; pairs showing marked
differences in personality were not allowed into the study. Potential
participants were excluded if they answered the following question in
the negative: "Do you yourselves believe that you are far more alike than
any pair of brothers or sisters you know of?" (Newman et al., 1937, p.
135). Newman and associates were attempting to screen out potential
DZA pairs because of the great expense of transporting them to and
lodging them in Chicago. Kamin (1974) also noted that the offer of an
all-expense-paid trip to Chicago in the middle of the Depression may
have induced twins to exaggerate the extent of their separation. According to Newman et al. (1937, p. 134), "Pairafter pair, who had previously been unmoved by our appeals to the effect that they owed it to science and to society to permit us to study them, could not resist the offer
of a free, all-expenses-paid trip to the Chicago Fair."Several pairs were
excluded from the study because they lived far apart from each other
and would have created too much difficulty and expense for the researchers. According to Taylor (1980, p. 82), "Such twins might have
been culturally or environmentally separated as well."
In fact, Newman and associates (1937, p. 31) were quite aware of the
possibility that their MZA group was biased in favor of similarity, writing, "When we come to the separated cases... it seems possible that our
group is more heavily weighted with extremely similar pairs than with
identical twins of less striking similarity." In an attempt to gather new
cases after the study's publication, Gardner and Newman (1940, p. 119)
corresponded with a pair of separated male twins in Pennsylvania. Although one of the twins was "anxious to have their case studied," his
brother "is, we fear, somewhat of a hoodlum and refuses to submit to
examination." Gardner and Newman commented, "This is unfortunate
in view of the fact that the two brothers now seem to be so different in
their personality traits." This case illustrates one aspect of the bias in
studies relying on volunteer participants: These twins were discordant
in their willingness to participate in the study becauseof their personality differences.
SEPARATED IDENTICAL TWINS
5
The age at separation in the Newman et al. study ranged from 2 weeks
to 6 years, and most pairs had much contact before being tested. For
example, Pair I "had lived together for a year before tests were given,"
Pair II "had lived together 7 years when tests were given," Pair IV had
"been acquainted and have visited back and forth at intervals all their
lives,"Pair VI was "living together when examined at the age of 58 years,"
Pair XIII had "visited at long intervals during the last 4 years," and Pair
XIX was separated at age 6 and "spent 6 months studying nursing together when they were 17 years old" (Newman et al., 1937, pp. 144-145).
Newman and associates provided more than 150 pages of case history material and personality test score information for all pairs. Reading
through the cases, one gets the distinct impression that the investigators viewed their twins as having far greater differences than those revealed by the now-outdated and probably invalid personality measures
they used. Some excerpts from case histories demonstrating twin pair
dissimilarities should be noted. Pair III (Paul C. and Paul O.) were MZ
twins who had been separated from infancy to adulthood. The investigators noted that Paul O. had an "affable, free, and unrestrained manner, whereas Paul C. is constrained, taciturn, and might almost be
thought to be sullen.... The tests do not agree very closely with the
observations as to the kind of differences which do exist, and this somewhat minimizes their significance" (Newman et al., 1937, p. 184). Looking at the results for Pair IV (Mabel and Mary), the investigators noted, "The sisters show greater differences in overt manner, as judged by
observation, than in the tests, with the exception of the Downey test"
(p. 195). Fifteen-year-old MZAsMildred and Ruth (Pair VIII) were raised
in homes differing greatly in social status. (Ruth had grown up in a poor
family.) Mildred had "alwaysbeen encouraged to make friends and has
constantly played with other children in groups ... The family of Ruth
has not been socially inclined so that Ruth has led a lonely, rather friendless life" (p. 221). The researchers conceded that Ruth's rearing environment had contributed to her "feeling of inferiority,"whereas Mildred
demonstrated a "feeling of confidence and a more or less expansive and
spontaneous habit of behavior" (p. 225). Pair XI (Gladys and Helen)
revealed further discrepancies between testing and observation: "One
sees at once that [Helen] is by far the more aggressive in her overt acts,
but the Downey test revealed about equally strong aggressiveness in the
two women. Gladys made the impression upon us of a person ill at ease"
(p. 249). In Pair XV (Edwin and Fred), Edwin was characterized as being
"considerably more vivacious than Fred" and as being "less diffident and
more aggressive." Edwin "was in a more cheerful state of mind. One
seemed to feel that Fred was worried" (p. 287). Maxine and Virginia
6
JOSEPH
(Pair XVI) were 11.5 years old when examined. Maxine was described
as being "verylively and active," but Virginia was found to be 'just the
opposite in behavior. She seemed to us be rather sluggish, somewhat
indolent, and difficult to arouse. She seems to be dormant emotionally" (p. 295). A marked contrast was evidenced in Pair XVII (Gene and
James): "Observations of the overt behavior of these boys revealed greater contrasts than did the tests.... Gene was far more confident and
affable toward us than James....James was very silent and glum during the whole period of tests and did not seem happy" (p. 304). Pair
XIX (Augusta and Helen) were found to be quite similar, but the authors also noted that "the sisters lived together during the most impressionable part of their lives" (p. 324).
Newman and associates concluded that it is likely that genetic influences play a role in the development of personality but stated that
their main achievement had been to present a body of data from which
others could draw their own conclusions. Looking back at the study,
Gardner and Newman (1940, p. 126) emphasized their belief in the
importance of heredity but noted, "If one thing is clear in our results
it is that fairly large environmental differences do modify physical,
mental and temperamental traits and produce proportionately large
differences even between hereditarily identical individuals."
Shields (1962).James Shields's (1962) British study reported the cases
of 44 MZA twin pairs, which at the time was the largest MZAsample ever
collected. As noted by Kamin (1974) and Farber (1981), Shields's sample was biased in favor of twin similarity. Twins were recruited through
an appeal made during a 1953 British television program. As Farber
(1981, p. 17) noted, Shields "limit[ed] himself to those sets who knew
of, or who suspected each other's existence." Therefore, as Taylor (1980,
p. 79) pointed out, Shields's sample probably was "biased toward underrepresentation of environmentally dissimilar pairs."
As Shields himself acknowledged, MZA environments "have not as a
rule been extremely different" (1962, p. 20). In fact, 30 of the 44 MZA
pairs were raised in different branches of the same family, and 33 pairs
differed in family socioeconomic status "only to a modest extent" (p.
116). The defining characteristic of Shields's twins' supposed separation
was only that they had been raised in different homes for at least 5 years
during childhood.
Shields established MZT and DZT control groups to compare their
level of similarity to his MZAs. Each twin was administered a Self-Rating Questionnaire (SRQ), specially devised for the study by H. Eysenck.
In addition, Shields ranked each pair for similarity of personality on the
basis of information provided by the twins and from interviews. These
ratings were made by Shields; the consensus among Shields's colleagues
SEPARATED
IDENTICAL
TWINS
7
was that "personality ratings were best made by someone who knew the
twins personally" (p. 71). Shields's twins therefore were rated by a nonblinded judge who also had an interest in the results of the study. MZA
correlations on the SRQ extraversion and neuroticism scales were reported as .61 and .53, respectively (Shields, 1962, p. 139). Interestingly, MZT twins correlated at only .42 and .38, which prompted Shields
(p. 139) to conclude, "There is little difference between the [MZA] and
[MZT] groups, and what there is does not always support the hypothesis that the early family environment has a lasting effect on the traits in
question." Shields failed to explain how separated identical twins could
be more similar than those reared together and did not question the
validity of the methods he had used in obtaining these correlations. In
conclusion, Shields's investigation was affected by most of the previously
discussed problems found in MZA studies, and its findings should therefore be treated with caution.
Juel-Nielsen (1965/1980).Juel-Nielsen's 1965 study (which was updated 15 years later) of 12 MZAs obtained its sample through the use of a
MZA study
twin register and is therefore the only systematic pre-MISTRA
that did not recruit twins on the basis of similarity or their preexisting
knowledge of each other (Farber, 1981). Juel-Nielsen's criteria for selection were that the twins must have been alive, reared apart from early
life, and monozygotic. He did not use a control group in his study.JuelNielsen used both interviews and testing (Rorschach, Word Association
Test) to identify personality similarities and differences between his
MZAs.All interviews were conducted by the investigator, who also made
the final determination of twin personality similarity.Juel-Nielsen concluded that his data indicated that there is a substantial genetic component in human personality variance.
It becomes apparent in readingJuel-Nielsen's detailed case histories
that most of the twins were raised in rural and working-class families,
and many grew up in abject poverty. This sample therefore was drawn
from twins whose rearing environments had been, in the words of Kamin, "highly similar" (Kamin, in Eysenck vs. Kamin, 1981, p. 112).
Although he found MZA similarity sufficient to conclude that the
evidence supported the genetic hypothesis, Juel-Nielsen observed that
his twin pairs generally were quite different from each other:
In all 12 pairs there were markedintra-pairdifferences
in that part of the
personalitygoverning immediate psychologicalinteraction and ordinary
human intercourse.... The twinsbehaved, on the whole, verydifferently, especiallyin their cooperation, and in their form of and need for contact. Corresponding with these observations, the twins gave, as a rule,
expression to very different attitudesto life, and very divergentviewson
general culture,religion and social problems.Their fields of interest,too,
8
JOSEPH
were very different.... Those twins who had children treated, on the
whole, their children differently,and their ideas on upbringingwere, as
often as not, diametricallyopposed. Characterologically,
the twinspresented differences in their ambitionsand in their employmentof an aggressivebehavior.Emotionally,there wasa deep-goingdissimilaritywith regard
to the appearanceof spontaneous emotional reactions or to the control
of affectiveoutbursts.Varioustraitsof personalityfound their expression
in differences in taste, mode of dress, hair style, use of cosmetics, the
wearingof beard or of glasses. (Juel-Nielsen,1965/1980, p. 75)
On the basis of these observations, Juel-Nielsen might have concluded
that separate rearing environments produce large personality differences between genetically identical individuals, but as is often the case in
these types of studies, the researcher was susceptible to "unconscious
investigator bias" (Kamin, 1974, p. 51). This potential bias underscores
the need to have blind raters make evaluations.
Looking at specific cases,Juel-Nielsen was unable to prevent bias from
influencing his final evaluation of twin similarity. For example, Pair V
(Kaj and Robert) was one of the few sets to have grown up under
different socioeconomic conditions. Robert had sought psychiatric
assistance on several occasions and was classified by Juel-Nielsen as a
neurotic; he was considered "introverted" and "more inhibited" (JuelNielsen, 1965/1980, Case Histories, p. 139). His separated twin brother, Kaj, "had shown asocial activity, and had been to prison, convicted
several times, and had been involved in shady business transactions"
(Juel-Nielsen, 1965/1980, p. 81), andJuel-Nielsen found him to be "predominantly extroverted" and "self-assertive"(Case Histories, p. 139). Despite these apparent differences, both twins were characterized as "psychopaths" byJuel-Nielsen because "both lack the check and control of
fundamental, uniform impulses and urges of a similar nature" (p. 81).
Robert was actually very different from Kaj, finding himself "revolted
by the glimpses he got of Kaj'sway of living" after meeting him for the
first time at age 40 (Case Histories, p. 132). Robert considered his brother Kaj to be "the most unpleasant person I have ever come across" (Case
Histories, p. 132). The fact thatJuel-Nielsen considered this pair to have
provided evidence of a "massive, genetic predisposition for a disharmonious development of personality" (p. 81) illustrates how a researcher's
bias can influence his or her interpretation of the data. HadJuel-Nielsen decided that neurosis and psychopathy were vastly different personality types, he might have concluded that these twins were very different, but because he rated them on the wider basis of whether each had
experienced "disharmonious development," he considered them to be
quite similar.
The case of Kaj and Robert illustrates the problems of rater bias in
SEPARATED IDENTICAL TWINS
9
this study. Juel-Nielsen provided a wealth of case-history material, but
despite using a less biased method of twin ascertainment, his study suffered from several important flaws found in most MZA studies.
The Minnesota Study of Twins Reared Apart
Background, method, and conclusions of the study. The MISTRA was
conceived in 1979 by psychologist and behavior geneticist Thomas J.
Bouchard, Jr. The initiative came from a report of the now-celebrated
case ofJim Lewis andJim Springer (commonly called the "JimTwins"),
who had been the subject of stories in the popular press. These identical twins had been separated at birth and were reunited in Ohio at age
39. The Jim Twins were said to have shared an uncanny set of similarities, such as the names of their wives and children, career choices, and
preferences for particular brands of beer and cigarettes. They were invited to the University of Minnesota by Bouchard and thus became the
first pair of MISTRA separated twins.
Twins were recruited on the basis of media appeals and the growing
publicity of the study. It has been claimed that selection on the basis of
similarity was "minimized by vigorously recruiting all reared-apart twins,
regardless of known or presumed zygosity and similarity" (Bouchard,
Lykken, McGue, Segal, & Tellegen, 1990, p. 223). Upon arriving in
Minneapolis, separated twin pairs were given a battery of personality and
mental ability tests, interest and value inventories, psychomotor tests,
information-processing measures, and several life history interviews
(Bouchard, 1984). The researchers had studied 27 MZA pairs by 1984,
and at the time of the most important papers on personality measures
(Bouchard & McGue, 1990; Tellegen et al., 1988), the sample consisted of 44-45 MZA pairs. By 1998, the MZA sample had reached 71 pairs
(Bouchard, McGue, Hur, & Horn, 1998).
According to the MISTRA researchers, their MZAs had "minimal contact" before being brought together for the study (Bouchard & McGue,
1990, p. 267). The average amount of time spent together before separation was 5.1 months, and according to the MISTRA contact time formula, MZAs spent an average of 26.5 months together before being
studied (Bouchard & McGue, 1990).
The results demonstrated that both MZAs and MZTs correlate at
about .50 on personality tests, leading to the conclusion that
The degree of MZtwin resemblanceon self-reportedpersonalitycharacteristics does not appear to depend on whether the twins are reared to-
gether or apart.... The comparabledegree of personalitysimilaritybetween reared-apart and reared-together twins suggests that common
familial environmental factors do not have a substantial influence upon
adult personality. The 50% environmental variance appears to be due
10
JOSEPH
primarilyto idiosyncraticenvironmentalfactorsand measurementerror.
(Bouchard& McGue, 1990, p. 286)
These conclusions are consistent with the current thinking in the
behavior genetics field, which holds that most personality differences
between siblings are caused by genetic differences and nonshared environmental influences, which are personal experiences not shared by
members of the same family (see Plomin & Daniels, 1987). Bouchard
and McGue (1990, p. 289) concluded that "parent child relations and
common family rearing environments have effects of minor magnitude,
if they have any reliable effect at all, on adult personality." In a subsequent paper, McGue and Bouchard (1998, p. 15) wrote that "pre- and
perinatal factors" appear to have a "substantial influence" on several
psychological characteristics.
The authors of the MISTRAarticles concluded that they found evidence that human personality variance is largely a product of human
genetic differences. Our task is to determine whether this is the proper conclusion.
A critique of the MISTRApersonality studies
Bias of the sample. As in the Newman et al. and Shields studies, MISTRAtwins were discovered on the basis of similarity and their knowledge
of each other's existence. As Farber (1981) pointed out, this type of
sample is not even representative of separated twins as a population.
Farber's book appeared 2 years after the beginning of the MISTRA
work,
and it is apparent that Bouchard et al.'s design was similar to the one
that Farber had shown to be seriously flawed. In Bouchard's (1982, p.
190) review of Farber's study, he failed to discuss Farber's devastating
analysis of the similarity bias of volunteer-based MZA studies and
claimed that her work was a "pseudoanalysis."
The Minnesota group's strategy of vigorously recruiting all located
pairs regardless of zygosity or similarity is in direct response to Kamin's
(1974) criticism of the Newman et al. study. How this recruitment strategy was carried out in the actual work remains unknown to the public
because MISTRA articles do not go into detail about how they recruited
their participants, including the Bouchard 1984 article, which is usually cited as the best description of the methods of twin ascertainment.
In any event, the vigorous recruitment of all located pairs does not alter the fact that the MISTRA MZA group constitutes a sample biased in
favor of similarity and is not representative of MZAs as a whole.
The lack of case histories and the failure to share data. The failure
to provide case history or interview material is another serious problem
with the MISTRA
studies. The reports of Newman et al., Shields, andJuel-
SEPARATED IDENTICAL TWINS
11
Nielsen contained hundreds of pages of detailed case histories of twins,
which helped independent observers to draw their own conclusions
from the data. Although most of the earlier researchers were viewing
twins through a hereditarian lens, their descriptions provided a wealth
of information.
Bouchard and associates amassed much data on the life histories of
their twins but have not published the information they collected. According to Bouchard (1984), each twin was given a Life History Interview,
Clinical Interview, Sexual Life History Interview, Life Stress Interview,
Child Rearing/Schooling Interview, and Briggs Life History Questionnaire. In addition, each twin completed the Family Environment Scale
(FES;Moos & Moos, 1986). Unfortunately, the MISTRA investigators have
reported only FES scores and have repeatedly denied critically minded
reviewers access to the interview data (Horgan, 1993; L. Wright, 1997).
MISTRA researchers have argued that it is unnecessary to provide more
information because such requests imply that "the similarities between
the MZA twins might be explained by unreported environmental similarities. We believe this to be highly unlikely" (Bouchard, Lykken,
McGue, Segal, & Tellegen, 1991, p. 192). After citing three studies that
supposedly refuted this idea, the writers continued, "The data we reported in our [Bouchard et al., 1990] article showed that the circumstances of rearing MZA twins could not have been as similar as that experienced by two individuals raised in the same family" (p. 192). Bouchard
and associates therefore refused access to raw data and case histories
because they already "showed" that the similarity of MZA rearing environments would not make a difference anyway.
In a discussion of the problems with Burt's studies, Jensen proposed
that MZA data should be made available to "anyone who wishes,"
Especiallyrare data, such as those of monozygotictwinsrearedapart,siblings from cousin matings, double first cousins, and the offspring of two
mated pairs of monozygotic twins... should be published in full, along
with complete descriptionsof the tests or measurementsand procedures.
Perhapsthis shouldbe a generalrequirementfor the publicationof studies
based on such valuable data, so that quantitativeanalyticaltechniques
other than those used by the original author can be applied to the data
by anyone who wishes. (1974, pp. 26-27)
Unfortunately, Bouchard and colleagues did not followJensen's recommendation. This has made it very difficult for independent researchers to analyze the data and offer alternative interpretations, as they had
with the earlier studies. Apart from an early article discussing the first
15 pairs (Eckert, Heston, & Bouchard, 1981), MISTRA researchers have
not provided a table containing information such as sex, age, age of sep-
12
JOSEPH
aration, and so forth for the individual pairs. Only mean figures for the
various twin types have been provided.
MISTRA researchers now claim that they are forbidden by federal law
(see W. Wright 1998, p. 215) or by the University of Minnesota Human
Subjects Committee from making public specific data on MISTRA participants. The fact that twins had signed an informed consent agreement
protecting their confidentiality was also cited by W. Wright as justification for withholding information. However, these prohibitions do not
explain how the authors of a MISTRA article (Eckert, Bouchard, Bohlen,
& Heston, 1986) were able to discuss the developmental and sexual
histories of six MZAsin a study of homosexuality. The authors discussed
cases while changing details "in order to protect the twins [sic] identities" (p. 422). This article contains no mention of any legal or ethical
restrictions on releasing this information.
Since 1979, Bouchard and associates have discussed individual cases
with journalists who have gone on to publish their stories in leading U.S.
magazines and newspapers. Furthermore, MISTRA researchers have released selected information to friendly reporters and colleagues on twins
who had signed the informed consent agreement and who had specifically
requestedanonymity.Watson (1981) devoted an entire chapter to MISTRAsupplied stories, which included at least two case histories of anonymity-seeking twin pairs:
This completes the number of identical twins raised separatelywhom
Bouchardhas seen and whose names he is willing to release. For a variety
of reasons the other five sets of twinshe has seen do notwantanypersonal
publicity[italics added], and two I now mention are therefore identified
only by a "casenumber."(Watson,1981, p. 61)
As in most other descriptions of MISTRA twin pairs, the readers of Watson's book were presented with the cases of the two most similar pairs;
the other three were not discussed.
Apart from the Jim Twins, the most publicized case has been the MZA
twin pair Oskar St6hr and Jack Yufe. These twins were born in 1933 in
Trinidad. Their parents divorced a few months after their birth, andJack
remained in Trinidad with his father, while Oskar went to live in Germany
with his mother. Oskar was raised a Catholic and Nazi in Germany,whereas Jack was raised as aJew in Trinidad and spent time on an Israeli kibbutz (Holden, 1980). The fact that this could happen seems to run
counter to the behavior genetic argument because the twins grew up to
be what their caregivers had raised them to be. But this was not the way
the case was presented. Although the two men were found to lead "markedly different lives" (Holden, 1980, p. 1324), a supposedly amazing set
of similarities was discovered after they arrived at the airport:
SEPARATED IDENTICAL TWINS
13
Both were wearingwire-rimmedglassesand mustaches,both sported twopocket shirts with epaulets. They shared idiosyncrasiesgalore: they like
spicyfoods and sweet liqueurs, are absentminded,have a habit of falling
asleep in front of the television, think it's funny to sneeze in a crowd of
strangers, flush the toilet before using it, store rubber bands on their
wrists. (Holden, 1980, p. 1324)
What was not mentioned is the fact thatJack and Oskar had met previously and had been in postal contact for more than 25 years (Horgan,
1993; L. Wright, 1997). After their cases were reported in the press, they
sold their life stories to a Los Angeles film producer (Horgan, 1993). Like
the twins recruited by Newman et al., this pair had a personal interest
in exaggerating their similarities and underreporting previous contact.
And it should be remembered that Juel-Nielsen (1965/1980, p. 75)
specifically noted that MZA personality differences express themselves,
among other ways, by the way such people wear beards and glasses.
A major problem with stories about pairs such as the Jim Twins and
Jack and Oskar is that similarities can be found between any two people
if this is what one is looking for. As Bouchard (1997b, p. 53) acknowledged, "Some of those similarities are surely coincidental-complete
strangers at cocktail parties routinely discover 'astonishing' occurrences in their lives; imagine what they might find after fifty hours of filling
out questionnaires." Behavior geneticist Richard Rose (1982, p. 960)
commented, "Such drama makes good show biz, but uncertain science."
Wyatt, Posey, Welker, and Seamonds (1984, p. 64) looked at the similarities of 25 unrelated pairs of college students matched on age and sex
and compared them with a group of MZTs. One pair of unrelated individuals was found to have a lot in common: "Both are Baptist; volleyball
and tennis are their favorite sports; their favorite subjects in school were
English and math (and both listed shorthand as their least favorite); both
are studying nursing; and both prefer vacations at historical places."
Wyatt et al. pointed out that a biased researcher could use the most
convenient level of analysis to make claims about similarity.
The MISTRA contact time formula is open to question. Total twin
contact is calculated on the basis of information obtained in the interview process. Time spent before and after separation is included in the
total and reflects the amount of time (measured in months) that twins
spent together before being studied. Bouchard and McGue (1990, p.
267) wrote that their MZAs had "minimal contact" before entering the
study, thereby implying that a pair of separated twins could influence
each other only when together. However, intimate relationships are
based on an ongoing association between people, which is not necessarily limited to the time they are in physical proximity. According to
the MISTRA contact time formula, "Twinswho met for a week at Christ-
14
JOSEPH
mas and for a week in the summer each year over a 10-year period are
credited with 20 weeks of contact" (Bouchard & McGue, 1990, pp. 266267). Thus, a MISTRA table would show such a pair to have had as much
contact as twins who had spent only the first 20 weeks of life together.
Clearly, these two pairs cannot be considered to have had the same influence on each others' personalities. Rather than regarding twins who
spent 14 days a year together over a 10-year span as having accrued 20
weeks of contact time, it would be far more accurate to note that these
"reared apart" twins had a 10-year relationship. In addition, volunteer
twins seeking inducements or scientific approval, on whose verbal accounts contact time has been calculated, might be tempted to underreport the amount of contact they had (Kamin, 1974).
Farber discussed a psychodynamic explanation of MZA similarity for
twins who "alwaysknew that they had a twin living somewhere else" and
had information about each other:
Psychodynamictheoristswould postulatethat this in itself maybe enough
to allowa child to begin to formulatea fantasy"twinning"with the absent
partner, particularlyif he has scattered bits and pieces of information
abouthim. Such an occurrencewouldbe most likelyto be significantwhen
the environmentin which the child is raised is an unhappyone-which
it was in many of these cases. (1981, pp. 19-20)
According to this theory, all that is needed is knowledge of the existence of a twin and minimal information. Additional contact time, even
if limited, could influence similarity to a far greater extent than it might
appear in a
MISTRA
table. Once again, a MISTRAquantification
is no
substitute for a detailed history of the twins in question.
Whenever large differences occurred between MZA pairs, MISTRA
researchers have searched for or speculated about organic explanations
for the difference. Three examples are provided here. L. Wright (1997,
p. 71) discussed a male MZA pair who had been reared in very different environments. One of the twins had been adopted by illiterate parents, whereas the other had been raised by a better educated family. The
difference in their IQ scores was 29 points, with the better educated twin
scoring higher. According to L. Wright, Lykken believed that "the lower-IQ twin may have suffered some kind of brain damage at birth." One
might ask why such unsupported speculation should be accepted.
The second pair was described by Heston (1988, pp. 211-212), who
discussed a 55-year-old pair of female MZAs who were separated at 4
months and were adopted into different homes. Twin A appeared "overweight and slovenly" and expressed "chronic dissatisfaction with her
life." She had been treated for depression for 8 years before being studied, had experienced anxiety all her life, and had several unexplained
SEPARATED IDENTICAL TWINS
15
medical complaints. She had completed high school at age 17 and
worked as a clerk for the next 30 years. Twin A described her marriage
at age 38 as "miserable." Her MZ sister, Twin B, weighed 60 pounds less
than Twin A and was described as an "engaging, even vivacious woman." Twin B had gone to college and had worked as a head technician
in a large hospital laboratory. She was happily married and described
her life as fulfilling and happy. According to Heston's account (p. 212),
his MISTRA colleagues "searched exhaustively through [the twins'] histories for some explanation of the striking differences between these
women." A clear bias is evident in this description because the researchers did not exhaustively investigate the histories of similartwins for the
purpose of discovering an environmental explanation for their alikeness. It turned out that Twin B had been taking thyroid medication since
age 18, whereas Twin A had not. The MISTRAlaboratories determined
that Twin A's thyroid function was in the "low normal range." Heston
offered Twin B's thyroid medication as a "tenuous explanation" for the
twins' differences.
In a third example, Segal (1999, p. 134) briefly discussed two MZA
pairs showing "marked differences" on the Minnesota Multiphasic Personality Inventory. Segal might have attributed the results to environmental influences but instead concluded that the differences were
"traceable to the development of psychiatric disorder in one case, and
probable brain damage due to head injury in the other case."
Unfortunately, neither Segal, Heston, nor Lykken indicated whether these pairs and their test scores were retained in the study, leading a
reviewer to wonder whether they were removed on the basis of a suspicion of organic problems.
The researchers' assumptions have influenced this study at every step:
The Jim Twins were recruited to the study because they were so similar
and were then offered as evidence of the similarity of MZAs, whereas
differing pairs are assumed to have become that way because of an organic impairment in one of the twins. It is remarkable that the Jim Twins
were included in the study and had their test scores figured into all
subsequent statistical formulations. As in most previous studies, these
twins came to the attention of researchers because of their similarity.
The Jim Twins epitomize the circular reasoning discussed by Farber
(1981).
Instead of publishing case history material, the MISTRA team has provided only scores on the previously mentioned FES. One might ask how
this scale would be necessary in light of the fact that twins underwent
extensive interviews covering most aspects of their life histories or how
it could be considered a substitute for an evaluation of family environment by blinded raters. In fact, the FES does not measure socioeconom-
16
JOSEPH
ic status but only social environment within the family. There is good
reason to believe that selective placement occurred to a large extent in
the MISTRA MZA sample, thereby violating the basic theoretical assumption of all adoption studies (Joseph, in press-a, in press-b). Bouchard
and McGue (1990, p. 278) acknowledged that FES correlations show
that "modest" selective placement occurred in their MZA sample, and
Bouchard (1984, p. 151) wrote earlier that the placement of MZAs in
adoptive homes is "unlikely to be random."
In summary, the absence of case history material and the refusal to
share raw data are reason enough to reject the conclusions of the MISTRA investigators. As one group of critics commented, "It is imperative
that case studies be fully published.... The Minnesota group have not
provided these case histories" (Beckwith, Geller, & Sarkar,1991, p. 191).
MZA personality test score correlations. A common assumption of
researchers studying MZAs has been that from a purely environmental
perspective, MZAswould show a correlation of zero on personality tests
or ratings. Juel-Nielsen (1965/1980, Follow Up, p. 12), for example,
wrote that the theoretical framework of his investigation was "based on
the assumption... that similarities between [MZAs] must be taken as
expressions of their common genotype." Other twin researchers have
thought in similar ways. According to the Swedish/American MZA research team, "The phenotypic resemblance of identical twins reared
apart in uncorrelated environments can be attributed to genetic influences" (Pedersen, McClearn, Plomin, & Nesselroade, 1992, p. 255),
and two leading psychiatric genetics proponents wrote, "Since MZ twins
reared apart do not share a common environment, any phenotypic similarity must be due to genetic factors. We cannot invoke shared environment as a cause of phenotypic concordance" (Faraone & Tsuang, 1995,
p. 91). Even Bouchard (1997b, p. 54) wrote, "When identical twins are
reared apart, their personality correlations must be an effect of genetics" (although Bouchard and associates used an age and sex correction
formula, discussed later in this article). However, there is good reason
to doubt the MZA zero correlation assumption. The failure to recognize that several important environmental factors contribute to MZA
similarity is a major error and is discussed in detail here.
MISTRA researchers have assumed that pen-and-paper inventories are
the best way to assess an individual's personality and have therefore
failed to consider the use of evaluations by blind raters. The authors of
the older separated twin studies cautioned against exclusive reliance on
testing as a way to assess twin similarity. Newman et al. (1937, p. 362),
for example, discussed the importance of "the method of individual
analysis as made in the case studies" as being "a more suitable method"
for estimating environmental influences on personality traits. Shields
was also aware of the pitfalls of total reliance on personality test scores:
SEPARATED IDENTICAL TWINS
17
Personality tests are less reliable and valid than intelligence tests. In any
case, theydo not take theplace of carefulclinical history-takingbytrained observers [italics added]. This aspect has been rather neglected in some of the
pairs reported. (1962, pp. 19-20)
Juel-Nielsen (1965/1980, p. 58) likewise recognized the importance
of obtaining information from interviews as well as from testing, and
after noting the limitations of both methods he decided that "the possibilities of making detailed, statistical comparisons were precluded from
the start." Thus, the authors of the three pre-MISTRA
systematic studies
of separated identical twins found the interview method of determining twin similarity to be an indispensable part of their work.
All systematic studies of MZAshave either failed to use a control group
or have used MZTsfor comparison. But what is often overlooked is the
fact that MZAs and MZTshave a lot more in common than an identical
genetic makeup. The following are the most important environmental
similarities shared by MZAs and MZTs:They are exactly the same age;
they are the same sex; they are almost always the same ethnicity; their
appearance is strikingly similar, which probably will elicit similar treatment; they usually are raised in the same socioeconomic class; they usually are raised in the same culture; they shared the same prenatal environment; and they typically spent a certain amount of time together in
the same family environment and were aware of each other's existence
when studied and often had regular contact over a long period of time.
It is essential that any study of MZAs control for these factors, which
implies that MZTs are unsuitable as controls. In a discussion of possible age and sex confounds in MZA cognitive ability studies, Kamin
(1974, p. 56) concluded that "without an assurance that the test scores
are not confounded with age and sex, there is absolutely no ground on
which to attribute an observed correlation to genetic factors." The confounding nature of these and other factors means that the only acceptable control group would consist of biologically unrelated pairs of individuals matched on the basis of the environmental factors common
to both members of an MZA twin pair. Rose discussed the basic idea
behind the formation of a control group of unrelated pairs:
A colleague suggeststhatwe cannot know [the importanceof MZAresemblance] withoutnecessarycontroldataon similaritiesfound in pairsof agematched strangers.... Were one to capitalize on cohort effects by sampling unrelated but age-matched pairs, born, say, over a half-century
period, the observedsimilaritiesin interests,habits, and attitudesmight,
indeed, be "astonishing."(1982, p. 960)
Let us now examine the most important points of similarity between
MZAs and MZTs. Any two people who are the same sex, were born on
the same day, and were brought up in the same culture are likely to have
18
JOSEPH
far more in common than two randomly selected people (Wyatt et al.,
1984). For example, it is more likely that two unrelated 40-year-olds
would share similar tastes in music than an unrelated pair consisting of
a 20-year-old and a 70-year-old. According to Farber (1981, p. 62), 5075% of MZAs were reared in "clearly deprived homes," and most were
raised in the same national or regional culture. Let us imagine a pair
of American MZAs born in 1920 and raised in working-class families in
different cities. Because they were born at the same time, these twins
would have experienced the same important value-influencing historical events at the same age. Both would have spent their teenage years
during the Great Depression, and their families probably would have
been similarly affected by economic difficulties. Having experienced a
national collapse of the banking system, both would be more likely to
distrust financial institutions than someone born 20 years later. Had they
been male, both would have been eligible for military duty at the same
time, and their personalities might have been influenced by their World
War II experiences. Had they been female, their personalities might
have been shaped by experiences on the home front. A separated pair
of female identical twins born in 1920 probably would both believe that
women can be good riveters; a pair born 30 years earlier might abhor
this idea. Both male and female twins would have been influenced by
the values of their class, ethnic group, and national cultures in similar
ways, and this influence would have continued during all phases of their
lives. We would therefore expect, on purely environmental grounds, that
two biologically unrelated individuals of the same age, sex, and cultural background would have more in common and correlate more similarly on personality and behavioral ratings than would opposite-sex biologically unrelated people of different age groups and cultures.
This phenomenon is known as a cohort effect, which denotes similarity on the basis of common group affiliation. As Farber (1981, p. 77)
noted in her analysis of twins' dental records, MZAs "arenot so muchsimilar to each otheras theyaresimilar to peopleof theireras and SES [socioeconomic statuses]" (italics added). Rose commented further,
[Farber's]argumentis by no means limited to dental health. Cohort effects will be operativein interests,attitudes,and education, and Farber's
box-scoreanalysisof twinpairsborn from the 1890'sto the 1950'smaybe
severelyconfounded with cohort effects. (1982, p. 960)
From the environmental perspective, therefore, MZAs should be far
more similar than two randomly selected members of the population.
MZA researchers' inability to control for the operation of
Pre-MISTRA
cohort effects led them to the questionable conclusions that have already been discussed.
SEPARATED IDENTICAL TWINS
19
Looking at same-sex DZAs, there is reason to believe that they would
be less similar than MZAs for purely environmental reasons. Parents,
families, and adoption agencies probably would attempt to place identical twins into more similar homes because of the greater physical similarity of MZs and societal expectations that they should be treated more
similarly. In the cases of DZ twins separated after infancy, the difference
in their physical appearance would make it more likely that they would
be placed into less correlated environments than MZs, whose similar
appearance might increase the desire to place them in more similar
environments. Because of their less similar physical appearance, DZAs
would experience more dissimilar appearance-related treatment in their
social environment.
It is also more likely that MZAs would be aware of each other's existence and have had more contact than DZAs. Several MZA pairs located each other because people had confused them with their separated
twin sibling who lived in a neighboring community. Upon discovering
the existence of a separated co-twin, an MZA would be more likely to
seek out a twin sibling in the belief that this person would be just like
him or her. This idea is supported by the fact that the MISTRA MZA:DZA
sample ratio includes far more MZAs than chance would expect and by
information demonstrating that MISTRA MZAshad about twice as much
prestudy contact time as DZAs (Bouchard & McGue, 1990, p. 267; Bouchard et al., 1998, p. 309). Therefore, several environmental factors
indicate that separated identical twins would be more similar than separated same-sex fraternals. According to Bouchard et al. (1998, p. 314),
MZA:DZAmean correlations on the 20 California Personality Inventory (CPI) folk scales were .46 and .27, respectively (with several scales
showing no statistically significant differences), and it is likely that environmental factors accounted for the mean scale difference. The 53
DZAs in the sample included 16 opposite-sex pairs (30%), whose correlations were not listed separately and whose scores probably reduced
the total mean DZA correlation. Opposite-sex DZAs should have been
excluded from this comparison because, as acknowledged by McGue
and Bouchard (1984, p. 325), sex (and age) effects on personality variables can be "substantial."
Findings from the MISTRA personality studies have shown that MZAs
correlate at .46 on the CPI (Bouchard et al., 1998) and .50 on the
Multidimensional Personality Questionnaire (MPQ; Bouchard et al.,
1990). To gauge the meaning of these results, one must first assess
whether cohort effects have influenced MZA similarity.
Studies of the relationship between personality test scores and environmental features common to reared-apart twins are rare in the literature. However, a few studies have looked at the effects of age and sex
20
JOSEPH
on scores of unrelated people. Several of these studies placed participants into age groups separated at 5- or 10-year intervals, although a
more meaningful examination would look at pairs matched for the same
age or year of birth because twins are always the same age. Grupp, Ramseyer, and Richardson (1968) found significant age effects on three of
the four CPI scales they examined in a group of 283 participants. Bedeian and Field (1988) found significant age correlations on 8 of 18 CPI
scales despite matching groups on the basis of a 10-yearage range. Cross,
Barclay, and Burger (1978) administered the CPI to 772 volunteer participants and found significant sex and ethnic group effects on several
scales. Another relevant study was performed by Martin, Blair, Dannenmaier,Jones, and Asako (1981). The participants in this study (N= 179,
118 female, 61 male, mean age 28.4 years) completed a CPI, and each
scale was correlated with the age of the respondents. Remarkably, the
mean correlation between age and scale scores was .28 across all 18
scales, with 10 scales showing a correlation of .35 or higher. The Martin et al. sample was drawn from a population of college students who
were probably similar in socioeconomic and educational status and
therefore had more in common than age alone. Nevertheless, it appears
that the effect of common age, which is only one of many environmental similarities shared by MZAs, is enough to account for more than half
the value found in MISTRA MZA personality correlations.
We have seen that, unlike their predecessors, MISTRA researchers
recognized that age and sex effects on psychological variables can be
"substantial."Unfortunately, they have used inadequate measures to deal
with these potentially confounding effects. McGue and Bouchard (1984)
devised a formula that adjusts raw twin correlations for age and sex effects on the basis of "twin-basedinformation." However, their own 1984
data demonstrate that the change produced by this correction is negliarticles have not provided raw test score correlations
gible, and MISTRA
on twin personality measures. Published correlations have been corrected for age and sex, and it is therefore unclear how much impact this
correction had on the data. It is likely that the MISTRA age and sex correction formula greatly underestimated the impact of age and sex effects on twin correlations and failed to correct for other potentially
confounding variables.
The often strikingly similar appearance of even separated MZ twins
can elicit similar treatment and affect behavior. Langlois and Stephan
(1977, p. 1697) studied 120 children (kindergarteners and first-graders)
and concluded that "attractive children... were liked more, were perceived as being smarter, were rated higher on sharing and friendliness
and lower on meanness and hitting other children than were unattractive children." Child participants made these determinations after view-
SEPARATED IDENTICAL TWINS
21
ing photographs of attractive and unattractive children. In a later study
based on the controlled observations of children at play, Langlois and
Downs (1979, p. 409) found that "5-year-old unattractive children aggressed against peers more often than did attractive children. Unattractive children were generally more active than attractive children." After reviewing the literature on the psychosocial aspects of physical
attractiveness, Alley and Hildebrandt (1988, p. 136) concluded, "The
evidence clearly suggests that facial aesthetics has widespread effects on
social development and interpersonal relations." As an extreme example of how similar appearance could affect personality, one would probably expect a pair of MZA beauty contest winners to be more extroverted and self-confident than a separated pair of disfigured MZs, yet no
MZA study has ever made allowances for the effects of the similar appearance of separated identical twins or for the more dissimilar appearance of separated fraternals.
Since the early 1980s, there have been two other studies of MZA personality similarities: The Swedish Adoption/Twin Study of Aging (SATSA; Pedersen, Plomin, McClearn, & Friberg, 1988) and a study carried
out in Finland (Langinvainio, Kaprio, Koskenvuo, & L6nnqvist, 1984).
In contrast to the similarity-biased Minnesota sample, the Scandinavian
MZAs were identified by means of population registers. As Pedersen et
al. correctly noted,
Both the SATSA and the Finnish MZAwere identified from populationbased information, whereas other studies of MZA [e.g., Newman et al.,
1937;Shields, 1962;MISTRA]typicallyrelied on identificationby thirdparties or response to media appeals. Pairsmay have come to the investigator's (and to each other's) attention because of their remarkablesimilarity (Farber,1981). Thus, MZAcorrelationsmaybe inflatedin other studies
with non-population-basedascertainmentprocedures. (1988, p. 955)
As in the other MZA studies, many of the Swedish and Finnish pairs
were reared together in the same home for several years. This was especially true in the Finnish study (Langinvainio et al., 1984), where 12
of 30 MZAswere separated between ages 5 and 10. There is little information in the SATSAor Finnish studies on the amount of postseparation MZA contact, and no case histories were provided. Nevertheless,
these participants were identified in a less biased manner than the MISTRA sample and therefore are a better sample with which to measure
twin personality similarity. The Swedish (N= 95 pairs) and Finnish (N
= 30 pairs) MZA correlations for extraversion were .30 and .38, respectively; for neuroticism, they were .25 and .25. Both studies used a Scandinavian version of the Eysenck Personality Inventory (EPI) short form
(Floderus, 1974). The Swedish correlations were corrected for age and
22
JOSEPH
sex on the basis of the McGue and Bouchard (1984) formula; the Finnish correlations were not corrected for age or sex.
On the basis of MZA and MZT comparisons, both the Swedish and
Finnish researchers concluded that their results replicated the previous
findings and that they supported the genetic hypothesis. Pedersen and
associates wrote that their study provided "further support for the conclusion that shared environmental effects are of negligible importance
for these measures" (1988, p. 954).
However, an alternative hypothesis provides a plausible reinterpretation of the data. Age effects probably have influenced twin correlations
derived from the Scandinavian version of the EPI short form. In discussing the normative data obtained for the original English version,
Eysenck and Eysenck (1964, p. 1106) found that scores differed by age:
"It appears that in nearly all cases older people are less extraverted and
less neurotic." This means that one would expect a person's score on
the EPI short form to be more similar to that of another person his or
her own age than to that of someone several years older or younger.
Looking at other personality tests, we have already seen that with a
sample of 179 unrelated people, Martin et al. (1981) found that the
average correlation between age and CPI scale scores was .28. When we
consider that MZAs are matched on many more factors than sex and
age and that the Finnish correlations were not corrected for age and
sex effects and were based on several pairs who spent a considerable
amount of time together and grew up under very similar socioeconomic
conditions (84% of the MZAs' rearing fathers were "farmers"or "manual workers"; Langinvainio, Koskenvuo, Kaprio, Lonnqvist, & Tarkkonen, 1981, p. 196), a plausible interpretation of the Scandinavian data
would be that the correlations on personality measures they reported
were caused solely by environmental factors. As previously mentioned,
before any conclusions can be drawn from an MZA and MZT comparison, it is necessary to determine the similarity of unrelated people
matched on the environmental variables shared by MZAs. Behavior
geneticists rarely discuss the need to match MZA correlations against
such a control group, which represents an enormous blind spot in their
thinking on the subject. Yet on the basis of MZA versus MZT comparisons, they are willing to come to conclusions that, if accepted, would
radically alter the way we think about human beings.
The MISTRA
articles, based on volunteer participants recruited on the
basis of similarity and a preexisting knowledge of their twinship, reported significantly higher correlations than the population-based Scandinavian work. The MISTRAarticle of Tellegen et al. (1988) reported a
factor-analytic "higher order scale" MZA intraclass correlation of .61 on
the negative emotionality factor, which the authors noted is similar to
SEPARATED IDENTICAL TWINS
23
the EPI neuroticism scale. The .61 negative emotionality correlation is
significantly greater than the .25 neuroticism correlations found in the
Scandinavian studies and is even higher (though not significantly) than
the Tellegen et al. (1988) figure for MZTs (.54). These findings might
have compelled other researchers to examine possible flaws in their
research design. Typically,however, MISTRA researchers argue that their
data demonstrate problems with other MZA studies (see Bouchard,
1997a).
We also can compare MISTRA results with those of other types of studies less vulnerable to environmental confounds. In a recent article coming out of the Colorado Adoption Project (Plomin, Corley, Caspi, Fulker, & DeFries, 1998), the authors found that the mean personality scale
correlation between birth parents and their adopted-away biological
offspring (N = 245)-a relationship that the authors considered "the
most powerful adoption design for estimating genetic influence" (p.
211)-was almost zero (.01, to be precise; p. 214). The results of this
carefully planned longitudinal adoption study by leading behavior geneticists (see Plomin & DeFries, 1985) are strikingly different from the
MISTRA results and cannot be attributed solely to error. One can reasonably conclude that at least one of the studies contains serious methodological flaws.
To illustrate the problems associated with the MISTRA methodology,
we turn to a 1990 article in which the researchers concluded from their
MZA data that 50% of the observed variance of five measures of religiosity was genetically influenced and that "the bulk of environmental
experiences that influence the expression of religiosity, like those that
influence the expression of personality and temperament... are experiences not shared by family members" (Waller, Kojetin, Bouchard,
Lykken, & Tellegen, 1990, p. 140).
It is useful to examine the conclusion that the expression of religiosity is negligibly influenced by shared family environment because it is
based on the same methods used in the other MISTRA studies. As Waller
et al. (1990) concede, their findings differ from those of other behavior genetics investigators. One could rephrase the Waller et al. conclusion to read, "To accept our conclusions about the genetic basis of personality and behavioral differences, one must accept that the expression
of religiosity is determined primarily by genes and environmental influences not shared by family members."
The belief that a person's religiosity is minimally influenced by shared
family environment is, however, absurd. Most parents attempt to instill
religious values in their children beginning in early childhood, and
whether a child attends religious services or schools is decided by the
parents. The continuity of religious ideas is actually one of the major
24
JOSEPH
roles of the traditional family; that it performs this task well is indicated by the fact that the chosen religion of adults is highly correlated with
the religion of their parents. The fact that a religious parent can produce a nonreligious child does not necessarily negate the influence of
shared family influences. As Hoffman (1991) noted, psychodynamic
developmental theory does not hold that parents alwaysproduce clones
of themselves; rather, parental behavior can influence a child in many
ways. Thus, a rebellious daughter of a Presbyterian minister might reject all religious ideas as oppressive, yet her ideas could still be viewed
as a reaction to the (negatively experienced) shared family environment
of her father's household. Attending religious services usually is an event
shared by family members, as are often religious books, television programs, and radio broadcasts. There are few better examples of a shared
family event than the attempt of parents to instill religiosity in their
children. Parents might treat and react to different children in the family in various ways, but it is difficult to imagine that they would have a
specific religious agenda for each child. Waller et al.'s postulated nonshared idiosyncratic environmental influence therefore would have limited opportunity to express itself. Perhaps friends, classmates, and
chance acquaintances are the principal nongenetic component of a
person's level of religiosity. Or perhaps, according to McGue and Bouchard (1998), several personality traits (such as religiosity?) are largely
the product of genetic and prenatal or perinatal factors. Or perhaps,
the entire design of the MISTRA research team is fatally flawed.
In conclusion, significant MZA personality and behavioral correlations can be explained plausibly on the basis of the various environmental similarities shared by separated identical twins and by inflated figures
resulting from bias and error in the various studies.
A proposal for an alternative control group
Although no conclusions about genetic influences on personality
differences can be drawn from the MISTRA data, a description of a valid MZA study seems in order. First, a systematic ascertainment of twins
would be undertaken. In addition toJuel-Nielsen's (1965/1980) criteria that the twins be alive, reared apart from early life, and monozygotic, the twins must not have been aware of each other's existence until
they are contacted by the researchers. As a way of determining whether selective placement had occurred in the sample, each twin pair's
rearing-family socioeconomic status would be ranked and correlated.
Once an experimental group of MZAs is collected in this manner, it
would be compared with a control group of biologically unrelated pairs
of strangers sharing the following characteristics: They should be the
same age, they should be the same sex, they should be the same eth-
SEPARATED IDENTICAL TWINS
25
nicity, the correlation of their rearing environment socioeconomic status should be similar to that of the MZA group, they should be similar
in appearance and attractiveness as determined by blinded raters, and
the degree of similarity of their cultural backgrounds should be equal
to that of the MZA twins. Finally, they should have no contact with each
other until after they are evaluated and tested. These controls will constitute the unrelated group. Of course, it is not possible for unrelated
pairs to share a common prenatal environment, which twins do share.
At this point we would be able to compare the MZA and unrelated
groups for personality and behavioral similarity.The best method would
be to have each individual's personality and behavior rated in a systematic way by blinded raters. After the scoring is completed, the code concealing the individual's group and pair status will be broken, and each
set of scores will be assigned to the proper twin or unrelated pair. The
scores between groups will then be compared. The genetic hypothesis
predicts that the MZA group would be more similar on the basis of their
identical genetic makeup. The environmental hypothesis predicts that
there would be no differences between the two groups other than those
accounted for by prenatal factors.
Summary and conclusions
We have seen that the MISTRA work has suffered from many of the
problems associated with the earlier MZA studies as outlined by Kamin
(1974), Taylor (1980), and Farber (1981). MISTRA researchers have not
demonstrated that MZAs were reared in uncorrelated environments.
The FES used by the research team is an inadequate tool for determining whether the confounding presence of selective placement occurred
in the sample. A far better way to determine similarities in the rearing
environment would have been a blind evaluation of the never-published
information gathered in the extensive life history interviews that were
given to all participants in the study. One could argue that the withholding of this information from publication and from other researchers is
enough reason to refrain from drawing any important conclusions from
the study. This is particularly true for a separated twin study, which is
extremely difficult to replicate.
The authors of MZA studies using a control group have mistakenly
believed that this group should consist of MZTs.I have argued here that
a control group, consisting of biologically unrelated pairs of strangers
matched on environmental factors common to both members of an
MZA twin pair, must be established and compared with the MZA group.
Only in this way, as Rose (1982) pointed out, can cohort effects be ruled
out as a major factor influencing MZA similarity. To the best of my
knowledge, no comparison of this type has ever been attempted. It
26
JOSEPH
might be objected that it would be difficult to obtain a large enough
sample of unrelated pairs on the basis of the requirements outlined in
this article. Even if this proved to be so, it would not permit the acceptance of conclusions about the origins of human psychological trait
variation on the basis of unscientific methods. As a rule, the more important the conclusions of a particular study, the more rigorous and
controlled must be its method. As one group of reviewers put it,
Many of the criticisms of behavior genetic studies that we offer may seem
to place unusual demands for proof upon the researchers. But, given the
potential social misuse of conclusions in this field, we believe that the
utmost care is required. (Billings, Beckwith, & Alper, 1992, p. 236)
In conclusion, the Minnesota separated twin studies, like the studies
that preceded them, are sufficiently flawed that no conclusions about
the role of genetic influences on human behavioral and personality differences can be drawn from them. If the MISTRAresearch philosophy
could be captured in one statement, it would be, "Everything can be
quantified," but this is certainly an illusion. As David Rosenthal (1968,
p. 414) once observed about schizophrenia researchers, "One can divide the investigators in the field into two main types: those who like to
look at numbers and those who like to look at patients. The former tend
to be hereditarians, the latter environmentalists." This observation is
also relevant to investigations looking into human behavioral and personality differences.
Notes
I thank Leon J. Kamin for providing a critical review of an earlier version of
the manuscript. I also thank David Cohen for reading the manuscript and providing valuable support and guidance. Any errors are entirely my responsibility.
Correspondence about this article should be addressed to JayJoseph, 2625
Alcatraz Ave. #328, Berkeley, CA 94705 (e-mail:[email protected]). Received
for publication November 6, 1998; revision received September 10, 1999.
References
Alley, T. R., & Hildebrandt, K. A. (1988). Determinants and consequences of
facial aesthetics. In T. Alley (Ed.), Social and appliedaspectsofperceivingfaces
(pp. 101-140). Hillsdale, NJ: Erlbaum.
Beckwith, J., Geller, L., & Sarkar, S. (1991). IQ and heredity [Letter to the
editor]. Science,252, 191.
Bedeian, A. G., & Field, H. S. (1988). Correlations between age and California
Psychological Inventory scale scores. PsychologicalReports,63, 696-698.
Billings, P. R., Beckwith, J., & Alper, J. S. (1992). The genetic analysis of human behavior: A new era? Social Science& Medicine, 35, 227-238.
SEPARATED IDENTICAL TWINS
27
Bouchard, T.J.,Jr. (1982). Identical twins reared apart: Reanalysis or pseudoanalysis? Contemporary
Psychology,27, 190-191.
Bouchard, T.J.,Jr. (1984). Twins reared together and apart: What they tell us
about human diversity. In S. Fox (Ed.), Individuality and determinism:Chemical and biologicalbases (pp. 147-184). New York: Plenum.
Bouchard, T.J.,Jr. (1994). Genes, environment, and personality. Science,264,
1700-1701.
Bouchard, T. J.,Jr. (1997a). The genetics of personality. In K. Blum & E. Noble (Eds.), Handbookof psychiatricgenetics (pp. 273-296). Boca Raton, FL:
CRC Press.
Bouchard, T.J.,Jr. (1997b, September/October). Whenever twain shall meet.
The Sciences,37, 52-57.
Bouchard, T. J., Jr., Lykken, D. T., McGue, M., Segal, N. L., & Tellegen, A.
(1990). Sources of human psychological differences: The Minnesota study
of twins reared apart. Science,250, 223-228.
Bouchard, T. J., Jr., Lykken, D. T., McGue, M., Segal, N. L., & Tellegen, A.
(1991). Response [Letter to the editor]. Science,252, 191-192.
Bouchard, T. J., Jr., & McGue, M. (1981). Familial studies of intelligence: A
review. Science,212, 1055-1059.
Bouchard, T.J.,Jr., & McGue, M. (1990). Genetic and rearing environmental
influences on adult personality: An analysis of adopted twins reared apart.
Journal of Personality,58, 263-292.
Bouchard, T. J., Jr., McGue, M., Hur, Y, & Horn, J. M. (1998) A genetic and
environmental analysis of the California Psychological Inventory using
adult twins reared together and apart. EuropeanJournal of Personality,12,
307-320.
Colt, G. H. (1998, April). Were you born that way? Life, 39-50.
Cross, D. T., Barclay, A., & Burger, G. K. (1978). Differential effects of ethnic
membership, sex, and occupation on the California Psychological Inventory. Journal of PersonalityAssessment,42, 597-603.
Eckert, E. D., Bouchard, T.J.,Jr., Bohlen,J., & Heston, L. L. (1986). Homosexuality in monozygotic twins reared apart. BritishJournal of Psychiatry,148,
421-425.
Eckert, E. D., Heston, L. L., & Bouchard, T.J.,Jr. (1981). MZ twins reared apart:
Preliminary findings of psychiatric disturbances and traits. Twin research3:
Intelligence,personality,and development(pp. 179-188). NewYork: Alan R. Liss.
NewYork: Wiley.
Eysenck, H.J., vs. Kamin, L.J. (1981). Theintelligencecontroversy.
Eysenck, S. B. G., & Eysenck, H.J. (1964). An improved questionnaire for the
measurement of extraversion and neuroticism. Life Sciences,3, 1103-1109.
Faraone, S. V., & Tsuang, M. T. (1995). Methods in psychiatric genetics. In M.
Tsuang, M. Tohen, & G. Zahner (Eds.), Textbookin psychiatricepidemiology
(pp. 81-134). New York: Wiley-Liss.
Farber, S. L. (1981). Identical twins rearedapart: A reanalysis. New York: Basic
Books.
Floderus, B. (1974). Psycho-social factors in relation to coronary heart disease
and associated risk factors. NordiskHygienisk Tidskrift[Monograph, Suppl.
6], 1-148.
28
JOSEPH
Gardner, I. C., & Newman, H. H. (1940). Mental and physical traits of identical twins reared apart. Journal of Heredity,31, 119-126.
Gould, S.J. (1981). The mismeasureof man. New York: W.W. Norton.
Grupp, S., Ramseyer, G., & Richardson, J. (1968). The effect of age on four
scales of the California Psychological Inventory. Journal of GeneralPsychology, 78, 183-187.
Hearnshaw, L. S. (1979). CyrilBurt: Psychologist.Ithaca, NY:Cornell University
Press.
Heston, L. L. (1988). What about environment? In D. Dunner, E. Gershon, &
J. Barrett (Eds.), Relativesat riskfor mentaldisorder(pp. 205-213). New York:
Raven.
Hoffman, L. W. (1991). The influence of the family environment on personality: Accounting for sibling differences. PsychologicalBulletin, 110, 187-203.
Holden, C. (1980). Identical twins reared apart. Science,207, 1323-1328.
Horgan, J. (1993). Eugenics revisited. ScientificAmerican,268, 122-131.
Jackson, D. D. (1960). A critique of the literature on the genetics of schizophrenia. In D.Jackson (Ed.), Theetiologyof schizophrenia(pp. 37-87). New York:
Basic Books.
Jensen, A. R. (1974). Kinship correlations reported by Sir Cyril Burt. Behavior
Genetics,4, 1-28.
Joseph, J. (1998). The equal environment assumption of the classical twin
method: A critical analysis. Journal of Mind and Behavior,19, 325-358.
Joseph, J. (1999a). A critique of the Finnish Adoptive Family Study of Schizophrenia. Journal of Mind and Behavior,20, 133-154.
Joseph, J. (1999b). The genetic theory of schizophrenia: A critical overview.
Ethical Human Sciencesand Services,1, 119-145.
Joseph, J. (2000). Not in their genes: A critical view of the genetics of attention-deficit hyperactivity disorder. DevelopmentalReview,20, 539-567.
Joseph, J. (in press-a). A critical review of twin and adoption studies of criminality and antisocial behavior. Journal of Mind and Behavior.
Joseph, J. (in press-b). The Danish-American Adoptees' Family studies of Kety
and associates: Do they provide evidence in support of the genetic basis
of schizophrenia? Genetic,Social, and GeneralPsychologyMonographs.
Joseph,J. (in press-c). DonJackson's "Acritique of the literature on the genetics
of schizophrenia": A reappraisal after 40 years. Genetic,Social, and General
PsychologyMonographs.
Juel-Nielsen, N. (1965/1980). Individual and environment:Monozygotictwinsreared
apart (Rev. ed.). New York: International Publishers.
Kamin, L. J. (1974). The scienceand politics of Q. Hillsdale, NJ: Erlbaum.
Langinvainio, H., Kaprio,J., Koskenvuo, M., & Lonnqvist, J. (1984). Finnish
twins reared apart III: Personality factors. Acta GeneticaeMedicaeet Gemellologiae, 33, 259-264.
Langinvainio, H., Koskenvuo, M., Kaprio, J., L6nnqvist, J., & Tarkkonen, L.
(1981). Finnish twins reared apart: Preliminary characterization of rearing environment. Twinresearch3: Intelligence,personality,and development(pp.
189-198). New York: Alan R. Liss.
SEPARATED IDENTICAL TWINS
29
Langlois,J. H., & Downs, A. C. (1979). Peer relations as a function of physical
attractiveness: The eye of the beholder or behavioral reality? Child Development,50, 409-418.
Langlois,J. H., & Stephan, C. (1977). The effects of physical attractiveness and
ethnicity on children's behavioral attributions and peer preferences. Child
Development,48, 1694-1698.
Lewontin, R. C., Rose, S., & Kamin, L. J. (1984). Not in our genes. New York:
Pantheon.
Martin,J. D., Blair, G. E., Dannenmaier, W. D.,Jones, P. C., & Asako, M. (1981).
Relationship of scores on the California Psychological Inventory to age.
PsychologicalReports,49, 151-154.
McGue, M., & Bouchard, T.J., Jr. (1984). Adjustment of twin data for the effects of age and sex. Behavior Genetics,14, 325-343.
McGue, M., & Bouchard, T.J.,Jr. (1998). Genetic and environmental influences
on human behavioral differences. Annual Reviewof Neuroscience,21, 1-24.
Moos, R. H., & Moos, B. S. (1986). FamilyEnvironmentScale manual (2nd ed.).
Palo Alto, CA: Consulting Psychologists Press.
Newman, H. H., Freeman, F. N., & Holzinger, K. J. (1937). Twins: A study of
heredityand environment.Chicago: University of Chicago Press.
Pam, A., Kemker, S. S., Ross, C. A., & Golden, R. (1996). The "equal environment assumption" in MZ-DZ comparisons: An untenable premise of psychiatric genetics? Acta GeneticaeMedicaeet Gemellologiae,45, 349-360.
Pedersen, N. L., McClearn, G. E., Plomin, R., & Nesselroade,J. R. (1992). Effects of early rearing environment on twin similarity in the last half of the
life span. BritishJournal of DevelopmentalPsychology,10, 255-267.
Pedersen, N. L., Plomin, R., McClearn, G. E., & Friberg, L. (1988). Neuroticism, extraversion, and related traits in adult twins reared apart and reared
together. Journal of Personalityand Social Psychology,55, 950-957.
Plomin, R., Corley, R., Caspi, A., Fulker, D. W., & DeFries, J. C. (1998). Adoption results for self-reported personality: Evidence for nonadditive genetic effects? Journal of Personalityand Social Psychology,75, 211-218.
Plomin, R., & Daniels, D. (1987). Why are children in the same family so different from one another? Behavioraland Brain Science, 10, 1-16.
Plomin, R., & DeFries,J. C. (1985). Originsof individual differencesin infancy:The
ColoradoAdoptionProject.Orlando, FL: Academic Press.
Rose, R.J. (1982). Separated twins: Data and their limits. Science,215, 959-960.
Rosenthal, D. (1968). The heredity-environment issue in schizophrenia: Summary of the conference and present status of our knowledge. In D.
Rosenthal & S. Kety (Eds.), The transmissionof schizophrenia(pp. 413-427).
New York: Pergamon.
Segal, N. L. (1999). Entwined lives: Twinsand what theytell us about human behavior. New York: Dutton.
Shields,J. (1962). Monozygotictwins broughtup apart and broughtup together.London: Oxford University Press.
Smith, R. T. (1965). A comparison of socioenvironmental factors in monozygotic and dizygotic twins, testing an assumption. In S. Vandenberg (Ed.),
30
JOSEPH
Methodsand goals in human behaviorgenetics(pp. 45-61). New York:Academic Press.
Taylor, H. F. (1980). The IQ game: A methodologicalinquiry into the heredity-environmentcontroversy.New Brunswick, NJ: Rutgers University Press.
Tellegen, A., Lykken, D. T., Bouchard, T. J., Jr., Wilcox, K. J., Segal, N. L., &
Rich, S. (1988). Personality similarity in twins reared apart and together.
Journal of Personalityand Social Psychology,54, 1031-1039.
Waller, N. G., Kojetin, B. A., Bouchard, T. J., Jr., Lykken, D. T., & Tellegen, A.
(1990). Genetic and environmental influences on religious interests, attitudes, and values: A study of twins reared apart and together. Psychological
Science, 1, 138-142.
Watson, P. (1981). Twins:An uncanny relationship?New York: Viking.
Wright, L. (1997). Twins:And what theytell us about who we are.New York:Wiley.
Wright, W. (1998). Born that way. New York: Knopf.
Wyatt, W.J., Posey, A., Welker, W., & Seamonds, C. (1984). Natural levels of
similarities between identical twins and between unrelated people. Skeptical Inquirer,9, 62-66.