See

discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/16004169

Familial studies of intelligence: A review

ARTICLE in SCIENCE JUNE 1981
Impact Factor: 31.48 DOI: 10.1126/science.7195071 Source: PubMed

CITATIONS

DOWNLOADS

VIEWS

513

364

361

2 AUTHORS:
Thomas J. Bouchard

Matt Mcgue

University of Minnesota Twin Cities

University of Minnesota Twin Cities

143 PUBLICATIONS 9,662 CITATIONS

438 PUBLICATIONS 19,917 CITATIONS

SEE PROFILE

SEE PROFILE

Available from: Thomas J. Bouchard

Retrieved on: 15 July 2015

Reprint Series
29 May 1981, Volume 212, pp. 1055-1059

SCIENCE

Familial Studies of Intelligence: A Review

Thomas J. Bouchard, Jr. and Matthew McGue

Copyright 1981 by the American Association for the Advancement of Science

Familial Studies of Intelligence: A Review

Abstract. A summary of 111 studies identified in a survey of the world literature on
familial resemblances in measured intelligence reveals a profile of average correlations consistent with a polygenic mode of inheritance. There is, however, a marked
degree of heterogeneity of the correlations within familial groupings, which is not
moderated by sex offamilial pairing or by typedf intelligence test used.
In 1963, Erlenmeyer-Kimling and Jarvik (1) published a summary of the world
literature on IQ correlations between relatives. Their. finding that the pattern of
correlations averaged over independent
studies was consistent with the pattern
predicted by a polygenic theory of inheritance has been widely cited as strong
evidence for some genetic determination
ofIQ (2). Although the accumulation of a
great many new data along with the
discrediting of Burt's important study on
monozygotic twins reared apart (3) has
outdated that review, the authors' summary or slightly modified versions of it
(4) continue to be widely reproduced (5).
Recently Plomin and DeFries (6) have
reported a comparison of those summary
data with the results of several large,
recent familial studies of IQ. They conclude that, in general, the recent studies
show less resemblance between relatives
than do the data reported by Erlenmeyer-Kimling and Jarvik. Their summary is
not comprehensive, however, and it
does not identify the factors that distinguish the two bodies of data. Roubertoux
and Carlier (7) have also published a
recent review, but it contains only 37
percent of the studies to be cited here.
The purpose of this report is to provide a comprehensive contemporary
summary of the world literature on the
IQ correlations between relatives. We
have updated the 1963 summary, adding
recent data and deleting several studies
included in the earlier review that do not
meet our methodological criteria for inclusion. Although the pattern of averages reported in this and earlier reviews
is remarkably consistent with polygenic
theory, the individual data points are
quite heterogeneous. Therefore, we have
also assessed the extent to which the
reported correlations are heterogeneous
and have attempted to identify some
factors that contribute to this heterogeneity.
In our survey of the literature we
found 140 studies that reported onfamilial resemblances in broad cognitive ability. These were reduced to 111 by the
application of explicit selection criteria
(8-12). The 111 studies, which include 59
reported in the 17 years subsequent to
the Erlenmeyer-Kimling and Jarvik summary, yielded 526 familial correlations,
based upon 113,942 pairings. Figure 1
SCIENCE, VOL. 212, 29 MAY 1981

displ.ays the correlations between relatives, biological and adoptive, in the 111
studies. The median correlation in each
distribution is indicated by a vertical bar.
The small arrow indicates the correlation
that would be predicted by a genetic
model with no dominance, no assortative
mating, and no environmental effects.
Researchers do not subscribe to such a
simple model, but it pro'fides a noncontroversial pattern against which to compare the results of various familial groupings. Different investigators will undoubtedly fit different models to the
data.
In general, the pattern of average correlations in Fig. 1 is consistent with the
pattern of correlations predicted on the
basis of polygenic 'inheritance. That is,
the higher the proportion of genes two
family members have in common the
higher the average correlation between
their IQ's.
The data set contains considerable
heterogeneity, as indicated by the x2
statistics. In an attempt to identify the
factors contributing to the heterogeneity,
we subdivided the familial groupings into
opposite-sex and same-sex pairings (Fig.
2) and male and female pairings (Fig. 3).
Among dizygotic twins the IQ's of samesex twins are more similar than those of
opposite-sex twins. This may reflect a
social-environmental effect (parents may
treat same-sex twins more similarly than
opposite-sex twins). The difference between non twin same-sex and oppositesex siblings and between same-sex and
opposite-sex parent-offspring pairings is
trivial. The male-female comparison
does not yield consistent trends. For
example, the average correlations are
larger in male twins than in female twins,
but the reverse is true for other siblings.
The absence of any demonstrable sex
effect is consistent with a polygenic theory of inheritance that does not posit the
existence of sex linkage. Environmental
theories that emphasize the importance
of sex-role effects on general cognitive
development are not supported by these
results (13).
Another possible source of heterogeneity is the intelligence test used. There
are many tests that purport to measure
intelligence, and they may not be highly
interrelated. We found great diversity in
test selection. For example, the 34 corre-

0036-8075/8110529-1055$01.00/0

Copyright 1981 AAAS

1055

treme and can be seen to result from a

few rather low correlations (14). The two
most extreme values are the .58 reported
by Blewett (15) and the .62 reported by
Nichols (16). In both cases the sample
sizes are small (26 and 36 pairs, respectively). The observation that 79 percent
of the reported correlations lie above .80
convincingly demonstrates the remarkable similarity of monozygotic twins.
After deleting the Burt data we are left
with test results on but 65 pairs of monozygotic twins reared apart, as reported in
three separate investigations. The
weighted average of. 72 is much less than
that found for the monozygotic twins
reared together, the difference suggesting the importance of between-family
environmental differences. At the same
time, the magnitude of this correlation
would be difficult to explain on the basis
of any strictly environmental hypothesis.
Three studies give midparent-midoffspring correlations, the weighted average of these being .72. In this case the
genetic expectation would depend upon
the number of offspring that define the
midoffspring value, and is thus indeterminate. The correlation between midparent and individual offspring does have a
determinate simple genetic expectation,

lations for monozygotic twins reared together were based upon results from 22
different tests, the 41 dizygotic twin correlations upon results from 25. We do
not have sufficient data to determine
whether the magnitude of the familial
correlation is moderated by the specific
test used. We did investigate whether
individually administered tests and
group-administered tests produced different correlations. For the monozygotic
twins reared together the 24 correlations
calculated on group tests produced a
weighted average of .86, and the 10 calculated on individual tests a weighted
average of .84. For the dizygotic twins
reared together, the weighted average of
32 correlations based on group tests is
.60 and of 9 correlations based on individual tests is .61. In neither case did the
distinction between group and individual
test produce an appreciable effect.
The 34 correlations reported on 4672
monozygotic twin pairs reared together
produce a weighted average correlation
of .86. This value is very close to those
reported in earlier reviews and is approximately the same for male and female
pairs. Although the test of homogeneity
yields a significant X2 value (P < .02),
the degree of heterogeneity is not ex-

0.0

0.10

0.20

0.40

0.30

0.50

0.60

MONOZYGOTIC TWINS
REAREO TOGETHER

0.70

0.80

..

0.90

.. "!!l:'!.:

MONOZYGOTIC TWINS REARED APART

MIDPARENT-MIDOFFSPRING
REAREO TOGETHER

L
)

MIOPARENT-OFFSPRING
REAREO TOGETHER

which is .707. The observed weighted

average of .50 is substantially less than
that, a discrepancy we discuss later.
The weighted' average of the 41 reported correlations in dizygotic twin pairs is
.60, considerably larger than for nontwin
siblings. Same-sex dizygotic pairs show
somewhat greater similarity than opposite-sex dizygotic pairs (.62 versus .57),
males being slightly more similar than
females (.65 versus .61). As with the
monozygotic twins, the test of homogeneity yields a significant value (P < .01),
although 75 percent of the correlations
fall within the narrow range between .50
and .70. The two extreme values were
reported in old studies on rather small
samples, the lowest being the .21 reported by Wingfield (17) in 1928 on 26 pairs,
the highest the .87 reported by Merriman
(18) in 1924 on 51 pairs. The greater
similarity of dizygotic twins than of other
siblings is most often interpreted as a
reflection of greater environmental similarity. It is also likely that bias in the
recruitment of dizygotic twins for study
is in the direction of increasing psychological similarity (19).
The weighted average for siblings
reared together is .47, which although
close to the simple expectation of .50 is

1.00

.
.

.. . 'f
. .

...:l
.1
.l.. 1m . . ..
. :i

ND.
WEIGHT
DF
MEDIAN
EO
ND.
CORREl
OF
CDRREl AVER
ATIONS PAIRINGS ATION
AGE

34
3

4672
65

.85
.67

SI8L1NGS REARED TOGETHER

2.46

.72

0.92
(Z)

0.46

1.33

410

.73

.72

Z.66
(2)

992

.475

.50

8.11
(7)

1.16

2.36
6.31

41

5546

.58

.60

94.5
(40)

69

26,473

.45

.47

403.6
(64)

203

.24

.24

32

8433

.385

.42

211.0
(31)

6.81

.22

9.61
(3)

3.20

1.55

II '1..11

SI8L1NGS REARED APART

.1.
I

SINGLE PARENT-OFFSPRING
REAREO TOGETHER

....

SINGLE PARENT-OFFSPRING
REARED APART

I.

ill'"

.. ..... .. . .

I
I

HAlFSI8LiNGS

COUSINS

NON-BIOLOGICAL SIBLING PAIRS

(ADOPTEO/NATURAl PAIRINGS)

"I' .

..

NON8IDlOGICAl SIBLING PAIRS

(ADOPTED/ADOPTED PAIRINGS)
ADOPTING MIDPARENT -OFFSPRING

II

..

ADOPTING PARENT-OFFSPRING

.
0.0

0.10

I
I

..

0.20

" 'j
0.30

..
0.40

..

..
0.50

0.60

0.70

0.90

.02

200

.35

.31

1,176

.145

.15

1.0Z
(2)

0.51

0.48

345

.29

.29

1.93
(4)

369

.31

.34

10.5
(5)

2.10

758

.19

.24

6.8
(5)

1.36

.19

6.64
(5)

1.33

.33

96.1
(15)

6.41

16

0.80

.oz
(1)

1.55
(1)

6
I

.. :

.22

814

I
I

..

...

.
ASSORTATIVE MATING

d.f.

.86

I.I!. '1

x2 .;.

81.29
(33)

.1

DIZYGOTIC TWINS
REARED TOGETHER

x2 (d.f.)

1397
3817

.18
.365

1.00

Fig. 1. Familial correlations for IQ. The vertical bar in each distribution indicates the median correlation; the arrow, the correlation predicted by a
simple polygenic model.
1056

SCIENCE, VOL. 212

based upon 69 values with a range of

correlations from .13 to .90. Oppositesex and same-sex siblings yield almost
identical weighted averages (.49 versus
.48), as do female and male siblings (.50
versus .47). The sibling correlations are
based on over 25,000 pairs; one large

0.10

0.0

representative study by Record et al.

showed a correlation of .55 on over 5000
pairs (20).
Whereas there is a wealth of information on siblings reared together, there is
a dearth of information on siblings reared
apart. Only 203 such pairs have been

0,20

0.30

SAME SEX DIZYGOTIC TWINS

-'--

.~

SAME SEX PARENT-OFFSPRING

PAIRINGS

SAME SEX ADOPTING

PARENT-OFFSPRING PAIRINGS

.: .. L:

. ..

OPPOSITE SEX PARENT-OFFSPRING

PAIRINGS

0,10

O,D

.:

I
1

0,20

F~milial

0,30

0.50

0.40

0.60

0.70

.57

19

6098

.45

,48

58,6

3,45

(17)

84.65

4,98

(17)

16

5127

.445

,49

70,4
(14)

14

4648

.41

.40

55.4
(13)

4.26

12

4476

,40

.39

70.1
(11)

6,37

460

,18

,IB

461

,12

,12

0.10

0.20

0.3D

0.40

0,6D

0.50

0,70

MALE DZ TWIN PAIRS

..

FEMALE SIBLING PAIRS

..

.: ...

..

FATHER-DFFSPR1NG REARED
TDGETHER

FATHER-SDN REARED
TDGETHER

ADOPTING MOTHER-OFFSPRING

. . . 1. .:

II

869

,835

.86

26,35
(8)

3.29

12

1,013

.86

.B6

24.4
110)

2.44

10

730

.58

.61

14.73
(8)

1.84

.65

20.32
(9)

2,26

3.15

.64

11

1,986

.42

.50

12

2,321

.38

.47

54.69
1101

5.47

,41

119.02
1241

4.96

,43

36.80
191

4.09

3.50

5,660
1,804

.38
.44

12

2,802

,37

.39

38,52
(11)

22

5,497

.43

,41

141.17
(21)

6.72

,46

,39

33.08
(9)

3,68

,38

36.86
(13)

2.B4

,195

,20

4.28
(5)

0.86

10.1
(5)

2.02

10

964

28.34
(9)

25

::

j .

FATHER-DAUGHTER REARED
TDGETHER

X2 "
dJ.

10

10

.1
1

. .. .

WEIGHT
NO.
ED
OF
NO.
MEDIAN
CDRREl
DF
CDRREL AVER
ATIDNS PAIRINGS ATION
AGE
x 2 Id.f.)

11

.. " . . .

.: I

.. ..

MDTHER-SDN REARED
TDGETHER

..

MOTHER-DAUGHTER REARED
TDGETHER

1.00

I
.1

MDTHER-OFFSPRING REARED
TDGETHER

"j,
I

.!
1
:I~

. ..

I
I

...

.1

MALE SIBLING PAIRS

.:

0.90

0.80

FEMALE DZ TWIN PAIRS

14

1,658

2,843

.40

1,393

212

.10

.10

247

.22

.22

1,279

.155

.18

214

,00

.00

248

,25

,25

I
1

. I.

,
j

I
D.l0

0,20

0.30

0.40

0.50

0.60

0.70

0,80

0,9D

1.00

Fig, 3, Familial correlations for IQ organized by male and female pairings,

29 MAY 1981

5,03

1.00

0,90

0.80

...

0,0

,565

2.43

correlations for IQ organized by opposite-sex and same-sex pairings.

MALE MZ TWIN PAIRS

ADDPTlNG FATHER-SDN

1592

FEMALE MZ TWIN PAIRS

18

68.14
(28)

Fig. 2.

ADDPTlNG FATHER-OFFSPRING

,62

X2 _
d.f.

0,0

,61

3,610

x2 IdJ .)

ADDPTlNG MDTHER-SDN

29

.:

"

..

WEIGHT
NO.
EO
OF
NO.
MEOIAN
CORREL
OF
CORREl AVER
ATiONS PAIRINGS AT10N
AGE

OPPOSITE SEX SIBLING PAIRS

ADOPTING FATHER-DAUGHTER

1.00

..:

"

ADDPTlNG MOTHER-DAUGHTER

0.90

0.80

... .:.I.

SAME SEX SIBLINGS PAIRS

OPPOSITE SEX ADOPTING

0,70

..... :.In.i.

OPPOSITE SEX OIZYGOTIC TWINS

PARENT-OFFSPRING PAIRINGS

0,60

0.50

0.40

studied, in two investigations yielding a

weighted average of .24, much less than
the expected value for such pairs and the
average value for siblings reared together.
The weighted average correlation between individual parent and individual

1057

offspring is .42 based upon 32 correlations. There is a marked degree of heterogeneity in the distribution, as evidenced not only by a significant X2 value
(P < .01), but also by the broad range of
the correlations. Their extreme heterogeneity cannot be attributed to a sex
effect, inasmuch as opposite-sex and
same-sex pairings yield equivalent averages, or to a maternal effect, the average
correlation of mother and offspring being
the same as that of father and offspring.
Although the large discrepancies between expected and observed correlations for parent and offspring reared in
the parental home (' 'reared together")
may be easily interpreted as a result of a
generational (social-environmental) effect, one should not hastily discount the
possibility of biological factors. Characteristics that are affected very little by
the social environment, such as height
and total fingerprint ridge count, show
similar generational differences. For example, in one large study of height Wingerd et al. (21) found a midparent-offspring correlation of .51. Two large studies of total fingerprint ridge count (22)
yield a Z-weighted mean midparent-offspring correlation of .63. The single parent-offspring correlations for height and
total fingerprint ridge count are .42 and
.42 (21-23).
As with parent and offspring reared
together, correlations for parent and separated offspring are quite heterogeneous.
The weighted average is .22, much less
than the simple expectation of .50. As
suggested by McAskie and Clarke (24),
one possible explanation could be that
parents and offspring are not given the
same test. In fact, roughly 50 percent of
our intergenerational correlations were
based upon data from cases in which
parents were given different tests from
those given their offspring. The lowerthan-expected morphological correlations found in these studies, however,
suggest that scaling may not be the only
problem.
Two familial pairings that are rarely
studied are half-siblings and cousins.
Two half-sibling correlations, both reported by Nichols (16), produce a
weighted average of .31. The four reported correlations for cousins are quite homogeneous; their average, .15, closely
approximates the simple genetic expectation.
A number of recent adoption studies
have added considerable knowledge.
Enough studies are available to permit
comparison of two sets of nonbiological
sibling
pairs-adopted/natural
and
adopted/adopted. Other things being
1058

equal, the adopted/natural correlation

should be higher than the adopted/adopted correlation, because the former would
contain a component for the covariance
of genotype and environment (25). The
present review finds the reverse (Fig. O.
The weighted average correlation of
adoptive midparent and offspring is .24,
and that of adoptive parent and offspring
is .19. Genetic theory requires the biological midparent-offspring correlation
to exceed the biological single parentoffspring correlation, and it does, although not by much (.50 versus .42).
Some environmental theories predict the
same effect (24); the failure to find any
difference in the adoptive case must be
considered surprising from an environmental point of view.
Unlike the case in natural families,
adopted offspring are somewhat more
similar to the same-sex adoptive parent
than to the opposite-sex adoptive parent
(Fig. 2). This conclusion is based, however, on a single study (26). Overall,
adoptive mothers are no more like their
adopted children than adoptive fathers
are.
The last row in Fig. 1 gives assortative
mating coefficients. There is marked
similarity between mates, but the
weighted mean of .33 is much smaller
than the .50 sometimes reported (27) .
The marked heterogeneity of the distribution indicates the sample-specific nature of these indices.
As in the earlier review, the pattern of
averaged correlations is remarkably consistent with polygenic theory. This is not
to discount the importance of environmental factors; monozygotic twins
reared apart are far from perfectly correlated, dizygotic twins are more similar
than other biological siblings, and adoptive parents' IQ's demonstrate a consistent relation with the IQ's of their adopted offspring. Although the data clearly
suggest the operation of environmental
effects, we found no evidence for two
factors sometimes thought to be important-sex-role effects and maternal effects. That the data support the inference
of partial genetic determination for IQ is
indisputable; that they are informative
about the precise strength of this effect is
dubious. Certainly the large amount of
unexplained variability within degrees of
relationship, while not precluding attempts to model the data, suggests that
such models should be interpreted cautiously.
THOMAS J. BOUCHARD, JR.
MATTHEW MCGUE

Psychology Department, University

of Minnesota, Minneapolis 55455

References and Noles

I. L. Erlenmeyer-Kimling and L. F. Jarvik, Science 142, 1477 (1963).

2. J. D. Matarazzo, Wechsler's Measurement and
Appraisal of Adult Intelligence (Williams & Wilkins, Baltimore, 1972).
3. L. Hernshaw, Cyril Burt, Psychologist (Cornell
Univ. Press, Ithaca, N.Y., 1979); D. D. Dorfman, Science 201, 1177 (1978).
4. L. F. Jarvik and L. Erlenmeyer-Kimling, in
Psychopathology of Mental Development, J.
Zubin and G. A. Jervis, Eds. (Grune & Stratton,
New York, 1967)
.
5. P. E. Vernon, Intelligence, Heredity and Environment (Freeman, San Francisco, 1979).
6. R. Plomin and J. C. DeFries, Intelligence 4, 15
(1980).
7. P. Roubertoux and M. Carlier, Ann. Bioi. Clin.
36, 101 (1978).
8. The reports that met the following criteria were
selected: (i) The research was reported in the
published literature or as a doctoral dissertation.
(ii) Familial resemblance was indicated by a
correlation coefficient, or the information provided permitted the determination of the correlation. In excluded studies results were typically
reported in terms of F ratios or within-pair
variances. (iii) The cognitive measure employed
was a widely accepted, standardized measure of
intelligence, or such a measure could be derived
from the test battery (for example, the first
principal component as a representative of the g
factor). In most studies excluded on the basis of
this criterion either a single measure of special
ability or a SUbjective rating of intellectual level
was used. (iv) In twin studies, the procedure for
zygosity determination was both objective and
valid. Use of a validated questionnaire was
considered an acceptable procedure. Many of
the early twin studies depended on subjective
procedures, for example teacher ratings. In such
cases only the opposite-sex twins (necessarily
dizygotic) were included in our data. (v) The
sample reported on did not overlap with another
sample reported elsewhere. For example, the
sample reported on by Jones (9) was not taken
because it is included in the larger sample reported on later by Conrad and Jones (10). (vi)
Because of the controversy surrounding the
samples reported on by Burt and some of his
colleagues (3), all data based on them were
excluded.
For each selected study the reported correlations were classified into the relevant categories
of familial relationship. If for any single sample
several correlations were recorded within the
same category, a single index of association was
determined as follows: (i) If correlations were
reported for several measures of intelligence one
of which was an individually administered test, .
the correlation for the individually administered
assessment was selected. If all measures were
group intelligence tests, their average correlation was used. Averaging was on the Fisher Z
transformations of the correlations. (ii) If several correlations were reported in a longitudinal
sequence, then the correlation recorded at the
highest age level was selected. (iii) If statistical
corrections were applied (correction for attemiation, correction for range restriction, correction
for age, or the like), only the age-corrected
correlation was used. If no age-corrected correlation was reported, then the uncorrected correlation was used. (iv) If several dependent samples reported on in the same study were relevant
for the category, then the weighted average of
the correlations was used (weighted average of
Z transformations, with sample size used as the
weight). For example, few studies reported a
single correlation between parent and offspring;
instead separate father-offspring and motheroffspring correlations were given. In such cases
the parent-offspring correlation was determined
as the average of the two correlations with the
number of pairings equal to the sum of the
pairings for the two correlations.
Application of the above rules resulted in the
inclusion of only independent correlations within any single classification category. For each
category the total number of pairings reported is
actually the minimum number of pairings, as in
several studies results were reported in terms of
the number offamilies used. The weighted average correlation was determined by taking the
weighted averages of the Z transformations with
the inverse of the variance used as the weight
and is thus the maximum likelihood estimator
(11). The weights employed were N - 3 for
interclass correlations and N - 3/2 for intraclass correlations. The x 2 statistic can be used to
test the hypothesis of homogeneity; that is, all

SCIENCE, VOL. 212

9.
10.

II.
12.
13.

14.
15.
16.

sample correlations were drawn from the same

population (12). The X2 statistic divided by its
degrees of freedom has an expectation equal to
1.0 under the homogeneity hypothesis and can
be used to compare the relative heterogeneity of
different categories.
De.tailed information about the data presented
here and a list of studies included and excluded
are available from the authors.
H. E. Jones, Yearb. Natl. Soc. Study Educ. 27,
61 (1928).
H. S. Conrad and H. E. Jones, ibid. 39, 97
(1940).
C. R. Rao, Linear Statistical Inference and Its
Applications (Wiley, New York, 1973).
M. A. G. Niana, J. Educ. Stat. 5, 83 (1980).
J. A. Sherman, for example [J. A. Sherman,
Sex-Related Cognitive Differences (Thomas,
Springfield, III., 1978), p. 40], argues: "Of the
sex-role characteristics mentioned the most relevant to cognitive development would appear to
be the emphasis on achievement and independence in problem-solving for males rather than
females." Such a pattern of socialization would
be expected to reduce cross-sex correlations
relative to same-sex correlations.
With samples as large as those reported here the
X2 test for heterogeneity is quite sensitive and
could detect even trivial amounts of variability.
D. B. Blewett, J. Ment. Sci. 100, 922 (1954).
P. L. Nichols, thesis, University of Minnesota
(1970); S. H. Broman, P. L. Nichols, W. A.
Kennedy, Preschool IQ, Prenatal and Early

SCIENCE, VOL. 212, 29 MAY 1981

17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.

28.

Developmental Correlates (Eilbaum, Hillsdale,

N.J., 1975).
A. H. Wingfield, Twins and Orphans: The Inheritance of Intelligence (Dent, London, 1928).
C. Merriman, Psychol. Monogr. 33, (No. 152)
(1924).
D. T. Lykken, A. Tellegen, R. DeRubeis, Soc.
BioI. 25, I (1978).
R. G. Record, T. McKeown, J. N. Edwards,
Ann. Hum. Genet. 33, 61 (1969).
J. Wingerd, I. L. Solomon, E. J. Schoen, Pediatrics 52, 555 (1973).
S. B. Holt, The Genetics of Dermal Ridges
(Thomas, Springfield, III., 1968); E. Matsuda,
Jpn. J. Hum. Genet. 17, 293 (1973).
W. E. Nance, Acta Genet. Med. 25, 100 (1976).
M. McAskie and A. M. Clarke, Br. J. Psychol.
67, 243 (1976).
R. Plomin, J. C. DeFries, J. C. Loehlin, Psychol. Bull. 84, 309 (1977).
J. M. Horn, J. C. Loehlin, L. Willerman, Behav.
Genet. 9, 177 (1979).
A. R. Jensen has reviewed this literature and
found a weighted mean of .42 [A. R. Jensen, in
Human Variation, R. T. Osborne, C. E. Noble,
N. Weyl, Eds. (Academic Press, New York,
1978)]. His review is less selective than ours.
This research was supported in part by the
National Institute of Mental Health traineeship
grant 5 T32 MH14647 and a research grant from
the University of Minnesota Computer Center.

22 August 1980; revised 13 January 1981

1059