You are on page 1of 8

Psychological Bulletin

1978, Vol. 85, No. 1, 194-201

Comment on Munsinger's Review of Adoption Studies


Leon J. Kamin
Princeton University

The critical review of studies of adopted children's IQ published in Psychological


Bulletin by Munsinger concluded that such studies indicate heredity to be "much
more important" than environment in determining individual differences in IQ.
The Munsinger review evaluated 17 different studies on the basis of 9 methodo-
logical criteria. The data cited from the various studies were, however, highly
selective; further, they were in many instances factually incorrect. The very
large-scale study by Freeman, Holzinger, and Mitchell was rated as among the
least adequate by Munsinger. The incorrect citations of their work by Munsinger
are documented in detail, both to illustrate the unsatisfactory nature of
Munsinger's critical review of adoption studies and to call attention to the
wealth of unique data presented by Freeman et al.

The central role of adoption studies in Munsinger study, as it happens, provided


argumentation about the heritability of IQ extraordinarily clear-cut data, suggestive of
has long been evident. The relevance of adop- an overwhelming genetic determination of
tion studies has now been heightened by a IQ, and the study by Freeman et al. con-
recognition that data provided by Burt tained many unique analyses that suggested
(1966) are not usable for the testing of a powerful environmental effect on IQ.
hypotheses (Jensen, 1974; Wade, 1976). The The methodological superiority of Mun-
largest and methodologically most satisfac- singer's study has not been evident to all crit-
tory study of separated identical twins had ics (Kamin, 1977). The purpose of this com-
been reported by Burt, and his data on IQ ment, however, is to indicate that Munsinger's
correlations among various kinship categories review of the adoption literature is in general
were more extensive than those of any other unreliable. Though any review must be selec-
investigator. tive in its presentation and analysis of data,
The critical review of adoption research Munsinger's is excessively so. Further, his
published in this journal by Munsinger descriptions of published studies contain a
(197Sa) concluded that "available data large number of simple factual misstatements.
strongly suggest that under existing circum- These weaknesses characterize much of Mun-
stances, heredity is much more important singer's review, but they will be documented
than environment in producing individual here primarily with reference to the work of
differences in IQ" (p. 623). This conclusion Freeman et al. (1928). That study examined
flowed from Munsinger's examination of 17 the largest sample of adopted children as-
adoption studies, each evaluated on the basis sembled by any investigator, and contained
of 9 methodological criteria. The most satis- a large number of pertinent data analyses.
factory of all the studies, in terms of the The fact that it is not often cited by theorists
methodological criteria, was said to be Mun- who maintain that IQ is highly heritable—
singer's (197Sb). The very large-scale study and the fact that Munsinger found it so
by Freeman, Holzinger, and Mitchell (1928) methodologically deficient—may bear some
was evaluated as one of the very worst. The relation to the results that it reported.
The Munsinger (197Sa) review of the
Requests for reprints should be sent to Leon J. Freeman et al. (1928) study reported accu-
Kamin, Department of Psychology, Princeton Uni- rately that those authors tested 401 adopted
versity, Princeton, New Jersey 08S40. children and 36 biological children of the

Copyright 1978 by the American Psychological Association, Inc. All rights of reproduction in any form reserved.

194
COMMENT ON MUNSINGER 195

adoptive parents. But Munsinger erred when a larger IQ gain in children placed in poorer
he stated that "all the adopted children's in- homes. Precisely the opposite occurred: Chil-
telligence quotients were measured because dren placed in better homes gained 5.3 points,
they were suspected of being mentally re- compared with a gain of only .1 point by
tarded?' (p. 633). This statement applied, in children placed in poorer homes. Practice, of
part, to a subset (134 of the 401 adopted course, was a constant across the two groups.
children), called the "pretest group." Those These highly relevant data evidently had no
children, later re-tested by Freeman et al., impact on Munsinger, who concluded that "it
had received an initial IQ test before being is difficult to know which methodological arti-
placed in foster homes. The initial testing, ac- fact or combination of artifacts actually
cording to Freeman et al. (1928), had been caused the apparent gain" (p. 634). The
carried out for "various reasons, the chief phrasing suggests that any environmentally
being a suspicion of low mentality. . . . In produced gain would be dismissed as an arti-
some cases a prospective foster parent asked fact by Munsinger; it is in any event clear
that a child be given a test in order to be as- that a significant differential gain standing
sured that he had normal mentality" (p. 115). out clearly over and above the "artifacts" he
Though it is obvious that the subjects of any cited was ignored. The differential gain of 5.2
adoption study cannot be representative of IQ points, it may be noted, could conceivably
the general population, the Freeman et al. have been much larger if the children had
sample was not so extraordinarily biased as been adopted at earlier than 8 years of age.
Munsinger's misstatement implied. Further, it should be recognized that the de-
Munsinger next referred to "a small sub- fective standardization of the Stanford-Binet
sample of 74 children" who had been tested presumably masked a significant gain made
twice, at average ages of 8 and 12 years, even by children adopted into "poorer" foster
and whose mean IQ had increased 2.5 points homes.
during 4 years in adoptive homes. These The Freeman et al. study also reported IQ
children were, of course, part of the pre- correlations for a large number of siblings
test group. Munsinger (1975a) indicated who had been separated and placed in differ-
that "Freeman et al. attributed most of ent foster homes. Munsinger accurately re-
this gain to the superior adoptive home en- ported that if the siblings were separated
vironment, but regression to the mean and after each was 5 years old, the correlation was
practice effects are more likely explanations .49 (n — 38). However, if the separation oc-
for the small gain" (p. 633). The likeli- curred before either was 6 years old, the cor-
hood of Munsinger's preferred explanations, relation was .32 (» = 46). Though this ac-
however, appears sharply reduced by a num- curately reported difference was not statisti-
ber of analyses presented by Freeman et al. cally significant, Munsinger (1975a) went to
(1928). First, they indicated that the mag- some length to minimize any impact it
nitude of the IQ gain in their data was might have. The basis for his guess is unclear,
underestimated, since the 1916 Stanford- but he assumed that siblings separated at an
Binet Intelligence Scale was negatively cor- early age were about 8 years old when
related with age. This was true not only tested and that siblings separated at a later
within their own data but also, as they dem- age were about 12 when tested. Further,
onstrated, within the original standardization he indicated that "early tests of intelligence
sample. Further, they subdivided the 74 chil- are not nearly as valid as later IQ tests" (p.
dren into two approximately equal groups (p. 633). The apparently low correlation of sib-
119). The home environments of all children lings separated early was thus attributed to
in the study had been rated numerically, and a lack of test validity. There appear to be no
the two groups consisted of children placed data that indicate the Stanford-Binet to be
in "better" or "poorer" homes. The children significantly more valid at 12 than at 8, but
placed in better homes had a significantly more importantly, Munsinger again ignored
higher IQ at the time of placement. Thus, "re- data directly relevant to his criticism. Free-
gression to the mean" should have produced man et al. (1928), although they noted that
196 LEON J. KAMIN

there were "very few cases in which the foster gitimate children were placed in lower status
homes of a pair differed in any marked de- adoptive homes and at a later age than were
gree" (p. 129), divided the separated sibling illegitimate children. There is adequate evi-
pairs into those reared in "more similar dence, however, of environmental effects over
homes" (n = 62) and those reared in "widely and above selective placement.
different homes" ( « = 6 3 ) . The correlations The second correlation cited by Munsinger
for the two groups were .39 and .28, respec- was one of —.11 between child's IQ and time
tively. This difference, like the one cited by in the adoptive home. That correlation, how-
Munsinger, was not statistically significant, ever, was observed in another subset of 185
but as Freeman et al. indicated, it did suggest children—the "main sibling" group. For the
an environmental effect—particularly in the entire group of 401 children, the correlation
absence of marked differences between foster was in fact -.05 (p. 194). The length of time
homes. The very low correlation between sib- spent in the adoptive home was of course posi-
lings reared in "different" homes, with an n tively correlated with child's age, and age was
of 63, cannot easily be dismissed as reflecting negatively correlated with IQ. For the entire
low test validity. sample, Freeman et al. reported a partial cor-
Munsinger (1975a) proceeded to cite three relation (age constant) of .08 between IQ and
correlations attributed to Freeman et al. that time in foster home. This very modest posi-
"taken together . . . offer no support for an tive association was, in view of the large
environmental effect on the adopted children's sample size, at the border of statistical sig-
IQs" (p. 634). Two of the cited correlations nificance. There was in any event no real rea-
were derived from subsets of the full Free- son to expect a large correlation, even if
man et al. sample, and the third did not ap- large environmental effects occurred. Free-
pear in their article. First, Munsinger re- man et al. cited suggestive evidence from the
ported that "Freeman et al. found a correla- pretest group to speculate that the beneficial
tion of .34 between the adoptive homes' social effect of adoption may occur relatively soon
status and the children's IQs at the time of after placement. There is obviously no reason
placement" (p. 633). This correlation was in to expect it to increase over time with no
fact observed in the pretest group of 74 chil- limit.
dren who, it will be recalled, were tested be- The final coefficient cited by Munsinger
fore placement. Freeman et al. indicated that was described as follows: "Also, the correla-
in this group the child's IQ had presumably tion between the children's IQs and the adop-
been used to accomplish selective placement, tive parents' IQs when the children were 12
but they also reported that within this group, years of age was .32" (p. 634). This particu-
the correlation with home rating increased to lar coefficient was not reported by Freeman et
.52 after 4 years. That significant increase al. The mean age at which children were
was sensibly interpreted as an environmental tested was 11, not 12, and the child's IQ cor-
effect. Their data also indicated that the much related .39 with that of adoptive midparent,
larger remaining group of adopted children .37 with adoptive father, and .28 with adop-
averaged less than 4 years old when placed, tive mother. (The .32 correlation cited by
compared with a mean placement age of 8 in Munsinger may have been the average of the
the pretest group. This large group of younger father's, .37, and mother's, .28, correlations,
children had not been tested prior to place- rounded down). The exact numerical value of
ment. For many of them, information about the parent - adopted child correlation was of
biological family background was minimal or some significance to Munsinger (1975a), who
absent. From these facts, Freeman et al. ar- argued that the
gued that selective placement in the main
group of children must have been less efficient selective placement correlation of .34 is very similar
than in the pretest group. There seems to be to the final correlation between the adoptive parents'
little doubt that selective placement did occur IQs and the children's IQs (r = .32), suggesting that
the adoptive parents' environment has no effect on
in the Freeman et al. study, as in all adoption children who are adopted into new homes at 4 or 5
research. For example, they reported that le- years of age. (p. 634)
COMMENT ON MUNSINGER 197

We have already indicated that the "selective managed both to inflate the alleged IQ cor-
placement correlation of .34" applied only to relation of biological parent-child pairs and
the pretest group and that within that group to suppress the revealing fact that adopted
the correlation between child's IQ and "adop- and biological children's IQs are equally cor-
tive parents' environment" increased signifi- related with the same home environments!
cantly, from .34 to .52! Those children, fur- This last observation is, of course, incon-
ther, were placed at 8, not "4 or 5," years of sistent with the claim that in normal families
age. There is, it may be noted, a logical error the correlation between child's IQ and home
in Munsinger's assumption that identical "se- environment is in large measure mediated by
lective placement" and parent - adopted child the common genes of parent and child. The
correlations would indicate that "adoptive equal correlations with home environment of
parents' environment has no effect." The adopted and biological children in the same
adoptive parents' IQs were not measured until families is one of the most interesting of the
long after the children had been placed, and many observations reported by Freeman et
the correlation between parental IQ and home al., but it is not likely to be discovered by
environment is of course far from perfect (p. readers who assume that Munsinger has pro-
143). vided an accurate or impartial critical review.
The most startling false citation by Mun- Throughout his review, Munsinger (1975a)
singer (197Sa) was his assertion that in the commented on the apparently inadequate data
case of adoptive parents who had their own presentation of Freeman et al. Thus,
biological children, "the correlation of the
adoptive parents' IQs with their own chil- it is difficult to know exactly the distribution and
mean age of separation and placement, because the
dren's IQs is .47" (p. 634). This coefficient, data are reported only in terms of before 7 years and
of course, is very close to that typically re- after 6 years of age, with no notion of how many
ported for biological parent and child, and it children are included in both categories, (p. 634)
is presumably to be contrasted with the .32
coefficient reported by Munsinger for adop- This is false. The Freeman et al. (1928) re-
tive parent - adoptive child. The coefficient port presented (p. 153) complete distribu-
actually reported by Freeman et al. was not tions plus means and standard deviations,
.47 but .35! Further, this was the correlation both of age when committed and of age when
between midparent and biological child. The placed in foster homes. Further, Munsinger
correlation between single parent and own child, reported "statistical problems" because "the
though not reported, was presumably even authors did not report the standard deviations
lower. This low correlation within adoptive of the parents' IQs, the adopted children's
families between parent and biological child IQs, or the adoptive parents' own children's
is to be expected in view of the restricted en- IQ scores" (p. 635). These assertions were
vironmental variance that characterizes adop- also false. The data that Munsinger's critical
tive families (Kamin, 1974). The same re- reading of Freeman et al. failed to detect were
stricted variance, of course, would be ex- presented on pages 177, 154, and 136, respec-
pected to produce low correlations between tively.
adoptive parent and adopted child. To this point I have focused on Munsinger's
The coefficient of .47 cited by Munsinger treatment of the work of Freeman et al., but
did appear in the Freeman et al. report, but the same kind of unreliability is evident in
Munsinger misidentified it. What Freeman Munsinger's descriptions of other adoption
et al. actually wrote about families contain- studies. To document this, below I briefly cite
ing both biological and adopted children was errors in his reports on seven additional
the following: "The correlation between the studies. There is, however, no intended impli-
intelligence of the own child and his home cation that all of Munsinger's errors are speci-
rating was .47 (« = 36), while the corre- fied in this comment.
sponding relation between the foster child The Munsinger review correctly indicated
and the home was also .47 (« = 34)" (p. that Wells and Arthur (1939) compared two
137). Thus, in a single stroke, Munsinger groups of children of "feeble-minded" par-
198 LEON J. KAMIN

ents. The groups, each 100 in number, differed of each group's mean difference score. This
with respect to their rearing conditions. The can be obtained by utilizing the standard
children of one group lived with their biologi- deviations given by Wells and Arthur, to-
cal parents, and children of the other group gether with the wildly implausible assumption
were reared by foster families. The children that, in each group, there was a perfect nega-
were given two IQ tests, spaced approximately tive correlation between scores on the two
5 years apart. The two groups did not differ test administrations. This absurdly conserva-
significantly in mean IQ at the time of the tive procedure still indicates that the differ-
first testing. However, children reared by ential shift in IQ between the two groups is
their own parents showed a decline of 6.68 significant. Finally, it may be noted that if—
IQ points between the two tests; at the same entirely inappropriately—the means and stan-
time, children reared by foster parents showed dard deviations given by Wells and Arthur
an increase of 1.43 points. are assumed to represent four independent
To quote Munsinger (1975a), groups of 100 subjects each, the "interaction
between group and first-second measure-
The authors interpreted this differential shift in IQ ments" would still be regarded as statistically
between the first and second testing to be the result significant!
of foster home placement, but they gave no statistical
significance test of the interaction between group and The Wells and Arthur study provides strong
first-second measurements to support their assump- evidence for the effect of rearing conditions
tion. This interaction could be tested statistically be- on the IQs of children. The design of the
cause the authors gave standard deviations for each study controlled for artifacts frequently cited
group. I computed appropriate analysis of variance
and in fact the interaction is not significant, (p. 643)
by Munsinger, such as regression to the mean
and practice effects. This positive evidence
The computation performed by Munsinger, was disposed of by Munsinger with a false
however, was in no sense "appropriate." The and undocumented assertion that his "ap-
data given by Wells and Arthur were as fol- propriate analysis of variance" had shown the
lows: For children reared by own parents, effect to be "not significant."
mean IQ on Test 1 was 81.06, and on Test 2, The Munsinger review correctly reported
74.38. The standard deviations were 15.39 that Snygg (1938) obtained only a very
and 12.89, respectively. For children reared modest correlation (.13) between the IQs of
by foster parents, the mean IQs were 79.69 312 adopted children and the IQs of their bio-
and 81.12, with standard deviations of 14.40 logical mothers. However, Munsinger (197Sa)
and 14.61. The correlations in IQ between indicated that the
first and second testings were not reported by sample of biological mothers was biased toward low
Wells and Arthur. IQ, so the distribution of IQ scores would be asym-
To "appropriately" test the interaction, of metrical and should have a smaller variance com-
course, knowledge of the within-subject error pared to a representative sample of the total popula-
tion. As you will recall, restriction of range usually
term is required, and this was simply not attenuates the correlation of scores with other vari-
available. However, it is easy to demonstrate ables, (p. 643)
that the differential shift reported by Wells
and Arthur was statistically significant. The The mean IQ of the 312 mothers was given
logical equivalent of the F test for interaction by Snygg as 78.30, with a probable error of
is a simple t test of the signed difference .61. Thus it can be calculated that the stan-
scores (Test 1 — Test 2) of the two groups. dard deviation of mothers' IQ was 1S.97—
The groups differed by 8.11 IQ points in clearly not a smaller variance than in the
mean difference scores. With 198 degrees of total population.
freedom, the difference would be significant Munsinger (197Sa) correctly indicated that
even under the absurd assumption that the Burks (1928) reported a multiple correlation
standard deviation of difference scores, ob- of .35 within a sample of adopted children
tained by testing the same individual twice, between child's IQ and a number of "en-
was as high as 29.08 IQ points! We can in fact vironmental variables." The multiple correla-
assign an upper bound to the standard error tion, Munsinger indicated, was "uncorrected
COMMENT ON MUNSINGER 199

for shrinkage," and this was said to consti- Munsinger's invention; it appeared nowhere
tute "a serious statistical problem in Burks' in the Skeels report.
data" (p. 637). Munsinger presented a for- The Munsinger (1975a) review asserted
mula (p. 637) designed to correct for "capital- that Skodak and Skeels "gathered IQ scores
ization on chance in a multiple correlation." on some of the biological fathers (they re-
The formula includes a term for the number ported having tested 12, but never published
of variables entering into the correlation. The any of the biological fathers' IQ data)" (p.
number of variables, according to Munsinger, 651). This statement was also false; the in-
was nine, so that, corrected for shrinkage, dividual IQs of the 12 tested biological fathers
Burks' "environmental correlation" was re- were presented in the article cited by Mun-
duced to .255. But Burks in fact employed singer (Skodak & Skeels, 1945, p. 51). The
only four variables in her multiple correla- fathers' IQs may be of interest to those who
tion. Thus, if Munsinger's correction formula wish to estimate the children's IQs from those
is appropriate, it yields an adjusted correla- of biological midparents. The biological fa-
tion of .317. thers' mean IQ of 82 was slightly lower than
The review of the several articles by Skeels that of the mothers with whom they had
and Skodak contained a number of errors. mated.
Having first suggested that Skeels "did not The Munsinger (1975a) review of the
really understand the meaning of a correla- Claeys (1973) report was not entirely ac-
tion coefficient" (p. 647), Munsinger (197Sa) curate in maintaining that "he found no rela-
stated that Skeels (1938), in the second re- tion between the primary mental abilities of
port on his cumulative and longitudinal study, the adopted children and the socioeconomic
"did not report the correlations between the class of their adopted parents" (p. 653).
biological fathers' education and the chil- Though the correlations involving various
dren's IQs or between biological midparent mental abilities were very modest in size, they
education and children's IQs, although the were positive in sign and, in one instance, sig-
latter correlation was .29 in the first sample nificant. Peculiarly, Munsinger neglected to
of 73 children" (p. 647). These assertions report that Claeys could not detect any re-
were false; both pieces of data were in fact lation between adopted child's ability scores
given in Skeels' report (p. 39). The correla- and the socioeconomic status of his or her
tion between biological father's education and biological parents. Perhaps this was because,
child's IQ in the enlarged second sample was as Munsinger indicated, information about
an insignificant .05. The correlation between biological parents' socioeconomic status was
biological midparent's education and child's "very vague and uncertain" (p. 653). The in-
IQ was an insignificant .08 in the second formation, however, had been sufficiently pre-
sample. Though Munsinger correctly asserted cise to reveal a significant selective placement
that the latter correlation had been .29 in correlation between the social class of biologi-
the first sample, that coefficient had been de- cal and adoptive parents.
rived from a sample size of 42 (not 73) and When discussing the results of his own
fell just short of statistical significance. study (Munsinger, 1975b), Munsinger re-
With a fine eye for numerical inconsistencies lated data from his samples of Mexican-
on the part of others, Munsinger (197Sa) American and Anglo adoptive families to
wrote that "Skeels' (1938) second report in- data reported earlier by Skodak and Skeels
cluded most of the initial sample of 73 in- (1949). For these three samples, Munsinger
fants plus 75 additional adopted children for described and graphed "a perfect positive as-
a total sample of 147 (see Skeels' report for sociation" between average IQ of the group
discrepant total)" (p. 647). There was no of adopted children and "average social
discrepant total in the Skeels report, which status" (p, 654) of the relevant group of
explicitly and consistently indicated that pre- biological parents. When group average IQs
cisely 74 additional children had been added are related to average social status of the rele-
to the entire initial sample of 73 to form a vant adoptive parents, "the association is
composite sample of 147. The number 75 was exactly reversed." This "pattern of group
200 LEON J. KAMIN

means . . . suggests that nature has a much to cite this low correlation based on what he
stronger effect than nurture on adopted chil- believed to be 13 pairs, Munsinger failed to
dren's IQs" (p. 654). cite an analogous correlation reported by
The "perfect positive association" seems Freeman et al. (1928). Those authors re-
unimpressive in a statistical sense. There is no ported a significant IQ correlation of .34 for
indication of a significant difference in the 40 pairs of adopted-own children. The basic
average IQs of the three groups of children, tenor of Munsinger's review seems evident.
and there are only six possible outcomes of a When a genetic interpretation of IQ calls for
rank ordering of three groups. Thus, granted no correlation, a low coefficient based on a
the social statuses of the biological parents, small sample was reported (but garbled). The
for example, the probability is .17 that, by existence of a higher and significant correla-
chance, the average IQs of the three groups tion based on a larger sample was simply
of children would exhibit "perfect positive as- ignored.
sociation" with parental social status. The point has been made elsewhere
To obtain this unimpressive result, Mun- (Kamin, 1974) that many reviews of the
singer arbitrarily employed the results of the literature on heritability of IQ are both ex-
fourth testing of the Skodak and Skeels chil- tremely selective and uncommonly rife with
dren, with the 1916 Stanford-Binet. The simple factual misstatements. The Munsinger
children had been tested during the same ses- review, in its entirety, is a case in point.
sion with the 1937 version of that test. They Though readers of the review will have ac-
had also, on two earlier occasions, been tested quired misinformation about many adoption
with the 1916 form. By Munsinger's scaling, studies, a particular disservice will have been
the Skodak and Skeels biological parents had done if they are dissuaded by Munsinger
the lowest social status of the three groups. from consulting the large-scale and extra-
Thus, to observe "perfect positive associa- ordinarily interesting Freeman et al. study.
tion" it is necessary for the Skodak and Perhaps the major point to be made is that
Skeels children to have the lowest average IQ readers interested in evaluating the evidence
of the three groups. There was only one of on heritability of IQ ought not to depend on
the several sets of test scores given by Skodak published summaries. Those who wish to
and Skeels that would produce such a result. speak or to teach accurately about what is
This was the testing session employed by and is not known have no realistic alterna-
Munsinger in his analysis, without comment. tive but to read this literature themselves.
There is, further, grave question concerning
the validity of Munsinger's ratings of pa- References
rental social status (Kamin, 1977). For the
two samples in his own study Munsinger Burks, B. S. The relative influence of nature and
(1977) indicated that he "cannot report pre- nurture upon mental development: A comparative
cisely how the ... ratings were generated be- study of foster parent-foster child resemblance and
true parent-true child resemblance. Yearbook of
cause they were not done by me personally" the National Society for the Study of Education
(p. 407), and that rating parental occupa- (Part 1), 1928,27,219-316.
tion "was a more difficult, subjective and Burt, C. The genetic determination of differences in
sometimes ambiguous procedure" (p. 409). intelligence: A study of monozygotic twins reared
The final example simultaneously illus- together and apart. British Journal of Psychology,
1966, 57, 137-153.
trates both the unreliability and the extreme Claeys, W. Primary abilities and field-independence
selectivity of Munsinger's review: He re- of adopted children. Behavior Genetics, 1973, 3,
ported that Leahy (193S) "found a correla- 323-338.
tion of only .06 between the IQ of adopted Freeman, F. N., Holzinger, K. J., & Mitchell, B. C.
children and own children living in the same The influence of environment on the intelligence,
school achievement, and conduct of foster children.
family; however, the statistic was based on Yearbook of the National Society for the Study of
only 13 pairs of foster siblings" (p. 639). The Education (Part 1), 1928, 27, 103-20S.
sample size on which Leahy's correlation was Jensen, A. R. Kinship correlations reported by Sir
based was in fact 25 pairs. Though choosing Cyril Burt. Behavior Genetics, 1974, 4, 1-28.
COMMENT ON MUNSINGER 201

Kamin, L. J. The science and politics of I.Q. Po- dren in adoptive homes. Journal of Genetic Psy-
tomac, Md.: Erlbaum, 1974. chology, 1945, 66, 21-58.
Kamin, L. J. Comment on Munsinger's adoption Skodak, M., & Skeels, H. A final follow-up study of
study. Behavior Genetics, 1977, 7, 403-406. one hundred adopted children. Journal oj Genetic
Leahy, A. Nature-nurture and intelligence. Genetic Psychology, 1949, 75, 85-125.
Psychology Monographs, 1935,17, 241-306. Snygg, D. The relation between the intelligence of
Munsinger, H. The adopted child's IQ: A critical re- mothers and of their children living in foster homes.
view. Psychological Bulletin, 1975, 82, 623-659. (a) Journal of Genetic Psychology, 1938, 52, 401-406.
Munsinger, H. Children's resemblance to their bio- Wade, N. IQ and heredity: Suspicion of fraud be-
logical and adopting parents in two ethnic groups. clouds classic experiment. Science, 1976, 194, 916-
Behavior Genetics, 1975, S, 239-254. (b) 919.
Munsinger, H. A reply to Kamin. Behavior Genetics, Wells, J., & Arthur, G. Effect of foster-home place-
1977, 7, 407-409. ment on the intelligence ratings of children of
Skeels, H. Mental development of children in foster feeble-minded parents. Mental Hygiene, 1939, 23,
homes. Journal of Consulting Psychology, 1938, 2, 277-285.
33-42.
Skodak, M., & Skeels, H. A follow-up study of chil- Received December 21, 1976 •

You might also like