Polygenic Scores Mediate the Jewish Phenotypic Advantage in Educational Attainment and Cognitive Ability Compared With Catholics and Lutherans

Abstract

A newly released multivariate polygenic score for educational attainment, cognitive ability, and self-rated mathematical ability in the Wisconsin Longitudinal Study was examined as a mediator of the group difference between Jews (n = 53) and 2 Christian denominations, Catholics (n = 2,603) and Lutherans (n = 2,027), with respect to educational attainment, IQ, and performance on a similarities measure. It was found that the Jewish performance advantage over both Catholics and Lutherans with respect to all 3 measures was partially and significantly mediated by group differences in the polygenic score. This result is consistent with the prediction that the high average cognitive ability of Jews may have been shaped, in part, by polygenic selection acting on this population over the course of several millennia. (PsycINFO Database Record (c) 2019 APA, all rights reserved)

Full text

Evolutionary Behavioral Sciences
Polygenic Scores Mediate the Jewish Phenotypic
Advantage in Educational Attainment and Cognitive
Ability Compared With Catholics and Lutherans
Curtis S. Dunkel, Michael A. Woodley of Menie, Jonatan Pallesen, and Emil O. W. Kirkegaard
Online First Publication, January 24, 2019. http://dx.doi.org/10.1037/ebs0000158
CITATION
Dunkel, C. S., Woodley of Menie, M. A., Pallesen, J., & Kirkegaard, E. O. W. (2019, January 24).
Polygenic Scores Mediate the Jewish Phenotypic Advantage in Educational Attainment and
Cognitive Ability Compared With Catholics and Lutherans. Evolutionary Behavioral Sciences.
Advance online publication. http://dx.doi.org/10.1037/ebs0000158

Polygenic Scores Mediate the Jewish Phenotypic Advantage in
Educational Attainment and Cognitive Ability Compared With
Catholics and Lutherans
Curtis S. Dunkel
Western Illinois University
Michael A. Woodley of Menie
Vrije Universiteit Brussel and Unz Foundation,
Palo Alto, California
Jonatan Pallesen
Aarhus, Denmark
Emil O. W. Kirkegaard
New York, New York
A newly released multivariate polygenic score for educational attainment, cognitive
ability, and self-rated mathematical ability in the Wisconsin Longitudinal Study was
examined as a mediator of the group difference between Jews (n⫽53) and 2 Christian
denominations, Catholics (n⫽2,603) and Lutherans (n⫽2,027), with respect to
educational attainment, IQ, and performance on a similarities measure. It was found
that the Jewish performance advantage over both Catholics and Lutherans with respect
to all 3 measures was partially and significantly mediated by group differences in the
polygenic score. This result is consistent with the prediction that the high average
cognitive ability of Jews may have been shaped, in part, by polygenic selection acting
on this population over the course of several millennia.
Public Significance Statement
Ashkenazi Jews exhibit high levels of general intelligence. The hypothesis that
differences in general intelligence between Jews and Catholics and Lutherans is
partially mediated by polygenic scores for educational attainment was tested. The
results support the hypothesized partial mediation.
Keywords: general intelligence, polygenic scores, religious groups
Supplemental materials: http://dx.doi.org/10.1037/ebs0000158.supp
Curtis S. Dunkel, Department of Psychology, Western Il-
linois University; Michael A. Woodley of Menie, Center Leo
Apostel for Interdisciplinary Studies, Vrije Universiteit Brus-
sel, and Unz Foundation, Palo Alto, California; Jonatan Pall-
esen, independent researcher, Aarhus, Denmark; Emil O. W.
Kirkegaard, independent researcher, New York, New York.
This research uses data from the Wisconsin Longitudinal
Study (WLS) of the University of Wisconsin–Madison.
Since 1991, the WLS has been supported principally by the
National Institute on Aging (AG-9775, AG-21079, AG-
033285, and AG-041868), with additional support from the
Vilas Estate Trust, the National Science Foundation, the
Spencer Foundation, and the Graduate School of the Uni-
versity of Wisconsin–Madison. Since 1992, data have been
collected by the University of Wisconsin Survey Center. A
public use file of data from the Wisconsin Longitudinal
Study is available from the Wisconsin Longitudinal Study,
University of Wisconsin–Madison, 1180 Observatory
Drive, Madison, Wisconsin 53706, and at http://www.ssc
.wisc.edu/wlsresearch/data/. The opinions expressed herein
are those of the authors.
Correspondence concerning this article should be ad-
dressed to Curtis S. Dunkel, Department of Psychology,
Western Illinois University, Waggoner Hall, Macomb, IL
61455. E-mail: c-dunkel@wiu.edu
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
Evolutionary Behavioral Sciences
© 2019 American Psychological Association 2019, Vol. 1, No. 999, 000
2330-2925/19/$12.00 http://dx.doi.org/10.1037/ebs0000158
1

Jews, and Ashkenazi Jews in particular, exhibit
possibly the highest group mean for IQ of any
population. Systematic reviews of the Jewish IQ
average indicate that it falls between 109 and 115
(Lynn, 2011), with the difference between Jewish
and non-Jewish populations being greatest on the
more highly g-saturated measures—indicating
that the advantage is primarily on the underlying
general cognitive ability (GCA) factor (Dunkel,
2014; te Nijenhuis, David, Metzen, & Armstrong.,
2014). Jews also exhibit a strong tilt toward verbal
and quantitative reasoning and away from visu-
ospatial ability (Lynn, 2011; Nisbett, 2009). These
psychometric advantages are likely a major factor
associated with their high representation in elite
professions, such as media, academia and among
those winning Nobel Prizes (Cofnas, 2018; Lynn,
2011; Murray, 2007).
Two major theories have been proposed to
account for the Jewish IQ advantage. The first is
based on the observation that their capabilities
and even affinities for various economic niches
may have been shaped by selective pressures
acting on these populations over hundreds of
years, and thus, their advantage might be ge-
netic (MacDonald, 1994). Cochran, Hardy, and
Harpending (2006) proposed that in the Middle
Ages, Jews in Europe were essentially pigeon-
holed into certain social and economic niches
by virtue of religious and social pressure. This
in turn led to culture-gene coevolution shaping
the Jewish ability structure. Cochran et al.
(2006) posited that the primary genetic locus of
this selection might have been rare variants
associated with sphignolipid (lipid storage) dis-
orders, common among Jews, such as Tay-Sachs
disease. This theory has not been tested directly;
however, there are hints in the data that those who
are heterozygous for the Tay-Sachs allele in par-
ticular appear to have higher levels of educational
attainment, when compared with heterozygotes
for other diseases (Kohn, Manowitz, Miller, &
Kling, 1988). This finding is at least in line with
predictions from the theory.
The second major theory is that the Talmudic
tradition among Jews incubates high ability via
the construction of a culture that emphasizes
learning and abstract reasoning and that is trans-
mitted from generation to generation vertically
as an environmental cause (Botticini & Eck-
stein, 2012; Ferguson, 2007). This model pur-
ports to be able to account for the Jewish ad-
vantage in ability and educational achievement
without recourse to genetic selection (Ferguson,
2007). This model should be considered specu-
lative because shared environment and in par-
ticular vertical transfer effects are generally small
or zero for GCA and are not generally found in
adulthood (Bouchard, 2013; Hatemi et al.,
2010; Eaves et al., 1999; Odenstad et al., 2008).
An additional possibility, apparently thus far
not considered at length, is that polygenic se-
lection acting over the course of several thou-
sand years and on multiple genetic variants,
which cumulatively account for variance in
GCA, may also have contributed to the group
difference in ability between Jews and non-
Jews. This could have been engendered by fac-
tors such as cultural group selection favoring
higher group-level GCA as an adaptation to
heightened intergroup competition, as envis-
aged by MacDonald (1994). Culture-gene co-
evolution involving niche provisioning and spe-
cialization of a sort envisaged by Cochran et al.
(2006) may also have been a source of this
polygenic selection. Indeed, the endogenous
cultural forces identified by proponents of cul-
ture-only theories (such as the development and
vertical transmission of scholarship and rule-
based systems of social organization, e.g., Fer-
guson, 2007) might themselves have been
sources of selective pressure acting on these
populations over time, with fitness payoffs hav-
ing accrued to those most capable of learning
and using such innovations. Consistent with
this, MacDonald (1994) has noted that the Tal-
mud contains injunctions against marriage in-
volving those who exhibit signs of low social
status (specifically the ‘am-ha-ares, or the ritu-
ally unclean). The precise nature of the selective
pressures that might have shaped (in particular)
Ashkenazi Jewish GCA are not known with any
certainty at present. A necessary criterion for
invoking these in the first place is the demon-
stration of systematic differences between Jews
and non-Jews with respect to salient genetic
variants.
Selection acting on polygenic scores (PGS)
1
can substantially shift the population means of
traits in relatively short amounts of time. For
1
Polygenic scores are constructed using results from a
GWAS of the trait of interest. Essentially, they are the sum
of the alleles multiplied by their beta on the trait from the
regressions.
2 DUNKEL, WOODLEY OF MENIE, PALLESEN, AND KIRKEGAARD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

example, in the population of Iceland, poly-
genic selection against variants predictive of
educational attainment may have reduced the
IQ of the population by 0.3 points per decade, or
2.1 points over 70 years (Kong et al., 2017).
Indeed, this may even be an excessively con-
servative estimate (for alternative calculations
see Woodley of Menie, Figueredo et al., 2017).
When comparing ancient Bronze and Early Iron
Age genomes, sourced from Eurasia, with those
from ancestrally matched modern European
populations, significant differences in the fre-
quencies of positively predictive alleles for ed-
ucational attainment and GCA have also been
found, favoring the modern populations. This is
consistent with a long-term Holocene selective
sweep in these populations, favoring higher
GCA (Woodley of Menie, Younuskunju, Balan,
& Piffer, 2017). Even among a subsample of the
ancient genomes for which radiocarbon dates
were available, significant associations between
sample age and positive allele frequency were
noted across a span of 3250 years (Woodley of
Menie, Younuskunju et al., 2017).
Given the recent availability of high-quality
PGS on educational attainment and related cog-
nitive phenotypes from large samples (Lee et
al., 2018), some of which contain Jews, it should
be possible to carry out a genetically informed
study on the etiology of the group difference in
GCA and educational attainment between Jews
and non-Jewish Caucasians of other religious
denominations (Catholic and Protestant). The
comparison of these two groups is desirable
because of the following: (a) evolutionary the-
ories of high Jewish ability have emphasized a
role for intergroup competition and niche pro-
visioning between these two groups in particu-
lar (Cochran et al., 2006; MacDonald, 1994,
1998); and (b) population differences studies
using PGS are potentially sensitive to linkage
decay, which results from recombination random-
izing the associations between alleles on chro-
mosomes over time (Bush & Moore, 2012).
This is problematic when the single-nucleotide
polymorphisms are noncausal variants that are
flagged by the genome-wide association study
(GWAS) procedure because they happen to be
in consistent linkage phase with the causal vari-
ants (Zanetti & Weale, 2016). This problem
reduces the utility of PGS when used for pop-
ulations relatively distant to the training sample
(Li & Keating, 2014; cf. Piffer, 2015). Ashke-
nazi Jews and non-Jewish Caucasians have been
found exhibit relatively low levels of genetic
differentiation. Tian et al. (2008) found that
Ashkenazi Jews exhibited F
ST
values ranging
from .0040 when compared with Italians to
.0144 when compared with Basque (across
eight Caucasian populations, the unweighted
F
ST
average is .009). This means that Ashkenazi
Jews exhibit little genetic differentiation, rela-
tive to non-Jewish Caucasians (F
ST
values rang-
ing from 0 to .05 correspond to little genetic
differentiation; Hartl & Clark, 1989). Values
this low also correspond to negligible amounts
of prospective linkage decay because this pa-
rameter has been found to scale quite strongly
with F
ST
(Scutari, Mackay, & Balding, 2016).
To test the polygenic selection theory, a large
sample of predominantly Caucasian individuals
of European descent from the United States,
which also contains Jews will be utilized in a
mediation analysis. We first examine the group
difference between Jews and non-Jews belong-
ing to two large Christian denominations on
PGS and indices of GCA and, second, test ex-
amines the degree to which a PGS capturing
phenotypic variance in educational attainment,
IQ, and self-reported mathematical aptitude
(POLY
MTAG
), mediates the group difference.
Full mediation is not expected for two reasons.
First, the PGS used does not account for all of
the variance in the phenotypes of interest (5–
10%; Lee et al., 2018) and is thus a rather noisy
estimate of the genetic potential. Second, there
may also be contributions stemming from cul-
tural (i.e., environment) causes and additional
nonadditive genetic causes not captured by the
PGS (which captures additive effects only),
such as the heterozygote advantage for certain
carriers of sphingolipid disorders posited by
Cochran et al. (2006). Some indication of poly-
genic mediation is nevertheless what would be
expected if polygenic selection has played a role
in shaping the group differences with respect to
measures of cognitive ability.
Method
Sample and Religious Orientation
Data were sourced from the Wisconsin Lon-
gitudinal Study (WLS). The WLS is a longitu-
dinal study of randomly sampled Wisconsin
high school students beginning in 1957; the last
3POLYGENIC SCORES MEDIATE JEWISH IQ ADVANTAGE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

wave of data collection was in 2011. The 1957
sample included 10,317 Wisconsin high school
seniors. The sample is overwhelmingly of Eu-
ropean descent (Herd, Carr, & Roan, 2014;
Sewell, Hauser, Springer, & Hauser, 2004), re-
flecting mid-20th century state demographics.
In the 1975 wave of data collection, partici-
pants were asked, “What was the main religious
preference of your family in 1957?” A total of
76 options were coded, but for the public re-
lease data set, the codes were collapsed into 17
categories. For the current analyses, three of the
17 categories were used (Catholic, Lutheran,
and Jewish). Catholic and Lutheran were cho-
sen because they are different religious orienta-
tions yet were also most strongly represented in
the original sample, n⫽3690 for Catholic
(29.8% of the WLS sample) and n⫽2619 for
Lutheran (21.2% of the WLS sample). No other
identifiable single orientation or denomination
accounted for more than 5% of the WLS sam-
ple. Beginning in 1977, a subsample of the orig-
inal participants’ siblings was also enrolled in
the study with iterations of sibling enrollment
occurring in the subsequent waves. Among the
original participants and siblings, we first se-
lected only those who had undergone genotyp-
ing. If both an original participant and a sibling
had undergone genotyping, we then randomly
selected from among the pair for inclusion in
the analyses. After these selection criteria, the
sample included 2,603 Catholics (51.2% fe-
male), 2,027 Lutherans (50.5% female), and 53
Jews (58.5% female).
Genotyping
A total of 9,012 WLS study participants were
genotyped on the Illumina HumanOmniExpress
array as part of the recent GWAS for IQ, edu-
cational attainment, and self-reported mathe-
matical ability (Lee et al., 2018). The genetic
samples came from saliva collected first in
2007–2008 and then during the course of home
interviews conducted initially in March 2010.
For full information on sampling and genotyp-
ing procedures, see https://www.ssc.wisc.edu/
wlsresearch/documentation/GWAS/Herd_QC_
report.pdf. In the present study, the educational
attainment polygenic score was used. The edu-
cational attainment phenotype was defined
based on the International Standard Classifica-
tion of Education 1997 United Nations Educa-
tional, Scientific and Cultural Organization
classification, which is associated with seven,
internationally comparable categories of educa-
tional attainment, rescaled as U.S. years-of-
schooling equivalents (Lee et al., 2018). The
polygenic score for educational attainment used
in this analysis (PGS_EA3_MTAG) was com-
puted using multivariate analysis of educational
attainment along with data on cognitive perfor-
mance (evaluated using a single measure IQ test
from U.K. BioBank along with various neuro-
psychological functioning tests and IQ sub-
scales from Cognitive Genomics Consortium)
in addition to self-reported mathematical ability
and highest mathematics class successfully
completed. This multivariate PGS was selected
because it likely captures the largest degree of
shared (i.e., GCA-like) genetic variance com-
mon to these cognitive phenotypes. The PGS
were standardized (transformed to z-scores) to
aid interpretation.
Measures of Cognitive Ability in WLS
Henmon-Nelson Test of Mental Ability.
The Henmon-Nelson Test of Mental Ability is a
30-min test consisting of 90 items of increasing
difficulty in spatial, verbal, and mathematical
ability. Test administration was standardized
across the state of Wisconsin during the first
wave of data collection in 1957. The reliability
of the test is estimated to be high (␣⬇.95; e.g.,
Ganzach, 2016; Hansen, 1968; Harley, 1977)
and scores on the Henmon-Nelson test exhibit a
strong association (r⬇.80 - .85) with full IQ
test scores (e.g., Wechsler Adult Intelligence
Scale [WAIS]) scores (Klett, Watson, & Hoff-
man, 1986; Kling, Davis, & Knost, 1978). The
WLS data file includes a variable labeled as
preferred measure of IQ based on the partici-
pant’s Henmon-Nelson test score, and this vari-
able was the one used in the current analyses. It
was found that siblings had a slightly higher
score than the original participants. Therefore,
the scores for each group were standardized
(transformed into z-scores) prior to merging.
Once the scores were merged, the scores were
transformed again so that the scores represent
IQ values.
Educational level. Education level was
measured in 1975 when participants were in
their mid-30’s. Participants reported their level
of education using a 9-point scale anchored at
4 DUNKEL, WOODLEY OF MENIE, PALLESEN, AND KIRKEGAARD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

high school graduate or less,less than one year
of college, and PhD, MD, other doctorates not
previously included, and post doctorate educa-
tion.
Similarities. During the 1992–1993 wave
of data collection when participants were in
their early 50’s, they were interviewed over the
telephone. The interview included a brief cog-
nitive assessment. Eight items from the WAIS
similarities subtest were also used as the assess-
ment tool (sample item: In what way are air and
water alike?). A total score based on the eight
items was used in the analyses. The total was
standardized by transforming the values into
z-scores.
Results
The correlation matrix for the study variables
for the full study sample can be seen in Table 1.
As seen in Table 1, all the variables were sig-
nificantly and positively correlated; most nota-
bly this includes the correlations between the
PGS and the three measures/proxies of GCA.
The descriptive statistics for the PGS and the
measures of cognitive ability for the three reli-
gious groups can be seen in Table 2. Addition-
ally, four one-way analysis of variance models
were run with religious orientation (Jewish,
Catholic, Lutheran) as the independent variable
with the dependent variables being the PGS and
the three measures of cognitive ability. As seen
in Table 2, all of the analyses of variance were
significant, with Tukey’s-b post hoc tests show-
ing that the Jewish group differed from the
Catholic and Lutheran groups on each variable,
whereas the Catholic and Lutheran groups did
not differ from each other on any variable. Note
that the difference between Jews and the Cath-
olic and Lutheran groups is larger (as measured
by standard deviation units) for educational at-
tainment. This could be due to the measurement
error in measures of GCA or the heightened
effect of differences between groups in genetic
composition and cultural importance placed on
education.
Furthermore, after creating groups of equiv-
alent size, we conducted a random sampling
analysis by taking a subsample of Christians the
same size as the Jewish sample and then ran t
tests looking at the group differences in PGS.
This was done 1,000 times. Each time the p
value of the ttest was recorded. The plot of the
log10 (pvalues) can be seen in the online sup-
plemental material. The mean pvalue for equiv-
alent groups is p⬍.000000001. Thus, it is
reasonably concluded that the effect is reliable.
To illustrate the differences between the Jewish
and two Christian groups, we combined the two
Christian groups and computed Cohen’s dfor
PGS and IQ. For PGS Cohen’s d⫽1.33, which
is a very large effect size. For IQ, Cohen’s d⫽
.57, which is a medium effect size. These group
differences are portrayed in Figure 1.
Next, we tested for the possibility that the
PGS mediates the association between religious
orientation and cognitive ability. The mediation
model and the associated components can be
seen in Figure 2. The PROCESS macro for
SPSS (Hayes, 2012) was utilized for testing for
mediation, and following the recommendations
of Zhao, Lynch, and Chen (2010), the output
from the bootstrap test for the indirect effect
was used as an indicator of mediation. Prior to
analyses, two dummy coded religious orienta-
tion variables were created; one variable (Cath-
olic ⫽1 and Jewish ⫽2) and the other (Lu-
theran ⫽1 and Jewish ⫽2). Thus, for each
GCA index, two analyses were performed: first
with the Jewish-Catholic dummy variable and
second with the Jewish-Lutheran dummy vari-
able. The dummy coded religious orientation
variable was entered as the X variable, the PGS
was entered as the mediator, and the GCA index
was entered as the Y variable. In PROCESS the
number of bootstrap samples was kept at the
default of 5,000, the confidence intervals were
kept at 95%, and the mediation model was set to
the specified model (i.e., Model 4).
Zhao et al. (2010) recommend reporting the
mean value of the indirect path (a⫻b) and
the associated 95% confidence interval from the
bootstrap method. As seen in Table 3, the con-
fidence intervals for the indirect path for each
Table 1
Bivariate Correlations Between Study Variables
Variables PGS IQ
Years of
education Similarities
PGS —
IQ .31 —
Educational level .28 .44 —
Similarities .21 .46 .36 —
Note. PGS ⫽polygenic scores. All correlations are sig-
nificant at p⬍.001, N⫽5513– 6256.
5POLYGENIC SCORES MEDIATE JEWISH IQ ADVANTAGE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

analysis did not include zero, indicating signif-
icant mediation. Zhao et al. (2010) also recom-
mend reporting the unstandardized regression
coefficients to enhance the interpretation of the
results. For example, it was consistently found,
across analyses, that moving from Lutheran or
Catholic to the Jewish religious category (path a
in Table 3) resulted in a .21- or .22-unit increase
in PGS. An additional analysis, included in the
online supplemental material, showed that re-
sults remained when controlling for family so-
cioeconomic status.
Discussion
In the present study, we found that Jews in
a large cohort had higher GCA, educational
attainment, and similarities scores than non-
Jews and that this group difference was par-
tially mediated by a PGS constructed from a
recent GWAS for GCA-related traits. There
are a number of limitations to the present
analysis. First, the number of Jews was rela-
tively small at n⫽53 and may therefore be
unrepresentative, although it appears that
contemporaneous Wisconsin Jews are fairly
representative of the U.S. Jewish population
in terms of socioeconomic characteristics (see
Appendix for analysis). Second, the PGS used
was only a poor estimate of the genetic po-
tential, which would by definition be equal to
the additivity value of IQ in terms of trait-
variance explained. Depending on which part
of the variance of the genetic potential this
proxy captures, it might affect the results in
unknown ways. Third, we relied on religious
denomination as a proxy for Jewish ancestry.
If the ubiquitous negative relationship be-
tween IQ and religiosity that has been noted
in Western populations (e.g., Kanazawa,
2010; Zuckerman, Silberman, & Hall, 2013)
extends to the Jewish population, then it
might be the case that by excluding nonreli-
gious Jews (who will simply not self-identify
as such for the purposes of listing religious
affiliation), we lowered the mean IQ for the
Jewish sample. We believe this to be a minor
problem because relatively few people, Jews
included, were nonreligious in 1975 when the
survey item was asked. Furthermore, the
Christian comparison group has the same
problem, which means both Group IQs are
biased in the same direction and the relative
difference is thus not likely to be strongly
Table 2
Descriptive Statistics and ANOVA Results by Religious Orientation
Variables
Religious Orientation
ANOVAJewish Catholic Lutheran
PGS 1.37 (1.07) .01 (.99) ⫺.04 (.98) F(2, 4680) ⫽52.88, p⬍.001
IQ 109.72 (14.36) 101.48 (14.43) 101.36 (14.66) F(2, 4554) ⫽8.41, p⬍.001
Educational level 5.62 (2.12) 2.45 (2.17) 2.39 (2.13) F(2, 4183) ⫽51.57, p⬍.001
Similarities .42 (.94) .06 (.98) .02 (.96) F(2, 4438) ⫽4.34, p⬍.05
Note. PGS ⫽polygenic scores. Standard deviations are in parentheses.
Figure 1. Distribution of Jewish and Christian IQ and
PGS.
6 DUNKEL, WOODLEY OF MENIE, PALLESEN, AND KIRKEGAARD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

affected (Kanazawa, 2010). Fourth, the PGS
was derived from a GWAS that consisted
mostly of European descent peoples, with
probably only a minor contribution from
Jews. To the degree in which the Jewish pop-
ulation differs genetically from the training
sample, this may reduce the validity of the
derived PGS. However, as was discussed in
the introductory text, Ashkenazi Jews (a re-
cently admixed population) are very closely
related to the training sample used in the
GWAS (Tian et al., 2008), and any reduction
in PGS validity is thus quite minimal, given
that F
st
is strongly and positively associated
with linkage decay (Scutari et al., 2016).
Given the above limitations, we consider the
present results to be tentative and in need of
replication with better PGS data and larger
samples of the Jewish population. Our find-
ings nonetheless yield an initial positive indi-
cation of the polygenic selection model and
critically indicate that in the case of the Jew-
ish versus non-Jewish Caucasian comparison,
the same source of genetic variance that gives
rise to of individual differences in GCA also
contributes substantially to the group differ-
ence. This militates against a substantive role
for Factor Xs (i.e., factors that create differ-
ences between groups but do not influence
individual-level variation) in the etiology of
this particular group difference (for discus-
sion of this, see Jensen, 1998, p. 446).
It finally needs to be stressed that these
findings do not militate against the other mod-
els considered in the introductory text. Rare
variants associated with lipid storage disor-
ders may indeed confer a heterozygote advan-
tage, which may have augmented the Jewish
Group GCA above that which would be pre-
dicted by differences in the level of PGS
alone, perhaps accounting for the relatively
higher frequencies of these disorders in this
population. Direct tests of this model still
need to be carried out, however.
Whereas purely cultural vertical transmis-
sion models involving the passing down
across the generations of the Talmudic Tradi-
tion are unlikely to be causative of the Jewish
advantage in GCA, it is possible that the
Jewish cultural practice of scholarship co-
evolved with, and indeed influenced, via cul-
ture-gene coevolution, Jewish group-level
characteristics, including their high average
GCA (MacDonald, 1994). It is important to
also stress the potential role played by social
epistasis (the moderating effect of a group’s
average PGS on the expressivity of an indi-
vidual’s PGS on a trait of interest, as captured
by the correlation between the PGS and that
trait) in maintaining traits within a group.
Social epistasis effects have been found to
influence educational attainment in human
populations (Domingue et al., 2018); the pat-
terns and rules governing these genetic inter-
actions might therefore constitute a source of
genetic nurture and may potentially be an impor-
tant component of the Jewish cultural inheritance
system that could be profitably researched in fu-
ture work.
Table 3
Mediation Analyses
Variables a⫻b95% CI ab C
IQ
Jewish-Catholic 5.75 [4.31, 7.33] .21 26.90 2.61
Jewish-Lutheran 5.95 [4.52, 7.48] .21 29.00 2.28
Educational level
Jewish-Catholic .74 [.55, .95] .21 3.48 2.24
Jewish-Lutheran .73 [.53, .95] .22 3.29 2.28
Similarities
Jewish-Catholic .27 [.19, .35] .21 1.26 .09
Jewish-Lutheran .26 [.18, .34] .22 1.19 .14
Note.CI⫽confidence interval.
Figure 2. Generalized mediation model.
7POLYGENIC SCORES MEDIATE JEWISH IQ ADVANTAGE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

References
Botticini, M., & Eckstein, Z. (2012). The chosen few:
How education shaped Jewish history (pp. 70–
1492). New Jersey: Princeton University Press.
Bouchard, T. J., Jr. (2013). The Wilson Effect: The
increase in heritability of IQ with age. Twin Re-
search and Human Genetics, 16, 923–930. http://
dx.doi.org/10.1017/thg.2013.54
Bush, W. S., & Moore, J. H. (2012). Chap. 11:
Genome-wide association studies. PLoS Computa-
tional Biology, 8, e1002822. http://dx.doi.org/10
.1371/journal.pcbi.1002822
Cochran, G., Hardy, J., & Harpending, H. (2006).
Natural history of Ashkenazi intelligence. Journal
of Biosocial Science, 38, 659– 693. http://dx.doi
.org/10.1017/S0021932005027069
Cofnas, N. (2018). Judaism as a group evolutionary
strategy: A critical analysis of Kevin MacDonald’s
theory. Human Nature, 29, 134–156. http://dx.doi
.org/10.1007/s12110-018-9310-x
Domingue, B. W., Belsky, D. W., Fletcher, J. M.,
Conley, D., Boardman, J. D., & Mullan Harris, K.
(2018). The social genome of friends and school-
mates in the National Longitudinal Study of Ado-
lescent to Adult Health. Proceedings of the Na-
tional Academy of Sciences of the United States of
America, 15, 702–707.
Dunkel, C. (2014). Reassessment of Jewish cognitive
ability: Within group analyses based on parental
fluency in Hebrew or Yiddish. Open Differential
Psychology. Advance online publication. http://dx
.doi.org/10.26775/ODP.2014.05.13
Eaves, L., Heath, A., Martin, N., Maes, H., Neale,
M., Kendler, K.,...Corey, L. (1999). Comparing
the biological and cultural inheritance of person-
ality and social attitudes in the Virginia 30,000
study of twins and their relatives. Twin Research,
2, 62– 80. http://dx.doi.org/10.1375/twin.2.2.62
Ferguson, R. B. (2007). How Jews become smart: Anti-
natural history of Ashkenazi intelligence. [Working
paper]. Retrieved from https://www.researchgate.net/
profile/R_Brian_Ferguson/publication/273369474_
How_Jews_Became_Smart_Anti-Natural_History_
of_Ashkenazi_Intelligence/links/54ff28410cf2741b6
9f414f9/How-Jews-Became-Smart-Anti-Natural-
History-of-Ashkenazi-Intelligence.pdf
Ganzach, Y. (2016). Cognitive ability and party iden-
tity: No important differences between Democrats
and Republicans. Intelligence, 58, 18–21. http://dx
.doi.org/10.1016/j.intell.2016.05.009
Hansen, E. A. (1968). The relationship between grade
point averages of the Henmon-Nelson Test of mental
ability and the American College Test (Master’s the-
sis). Retrieved from https://digitalcommons.usu.edu/
etd/5649
Harley, D. D. (1977). The Henmon-Nelson: Comput-
erized (Master’s thesis). Retrieved from https://
open.library.ubc.ca/cIRcle/collections/ubctheses/
831/items/1.0094815
Hartl, D. L., & Clark, A. G. (1989). Principles of
population genetics (2nd ed.). Sunderland, MA:
Sinauer Associates.
Hatemi, P. K., Hibbing, J. R., Medland, S. E., Keller,
M. C., Alford, J. R., Smith, K. B.,...Eaves, L. J.
(2010). Not by twins alone: Using the extended
family design to investigate genetic influence on
political beliefs. American Journal of Political Sci-
ence, 54, 798– 814. http://dx.doi.org/10.1111/j
.1540-5907.2010.00461.x
Hayes, A. F. (2012). PROCESS: A versatile compu-
tational tool for observed variable mediation,
moderation, and conditional process modeling
[White paper]. Retrieved from http://www.afhayes
.com/public/process2012.pdf
Herd, P., Carr, D., & Roan, C. (2014). Cohort profile:
Wisconsin longitudinal study (WLS). Interna-
tional Journal of Epidemiology, 43, 34– 41. http://
dx.doi.org/10.1093/ije/dys194
Jensen, A. R. (1998). The g factor: The science of
mental ability. Westport, CT: Praeger.
Kanazawa, S. (2010). Why liberals and atheists are
more intelligent. Social Psychology Quarterly, 73,
33–57. http://dx.doi.org/10.1177/0190272510361602
Klett, W. G., Watson, C. G., & Hoffman, P. T. (1986).
The Henmon-Nelson and Slosson tests as predictors
of the WAIS-R IQ. Journal of Clinical Psychology,
42, 343–347. http://dx.doi.org/10.1002/1097-
4679(198603)42:2⬍343::AID-JCLP2270420221⬎3
.0.CO;2-W
Kling, J. O., Davis, W. E., & Knost, E. K. (1978).
Henmon-Nelson IQ scores as predictors of WAIS
full scale IQ in alcoholics. Journal of Clinical Psy-
chology, 34, 1001–1002. http://dx.doi.org/10.1002/
1097-4679(197810)34:4⬍1001::AID-JCLP2270
340437⬎3.0.CO;2-K
Kohn, H., Manowitz, P., Miller, M., & Kling, A.
(1988). Neuropsychological deficits in obligatory
heterozygotes for metachromatic leukodystrophy.
Human Genetics, 79, 8–12. http://dx.doi.org/10
.1007/BF00291701
Kong, A., Banks, E., Poplin, R., Garimella, K. V.,
Maguire, J. R., & Daly, M. J. (2017). Selection
against variants in the genome associated with edu-
cational attainment. Proceedings of the National
Academy of Sciences of the United States of America,
114, E727–E732.
Lee, J. J., Wedow, R., Okbay, A., Kong, O., Maghzian,
M.,...Cesarini, D. (2018). Gene discovery and
polygenic prediction from a 1.1-million-person
GWAS of educational attainment. Nature Genetics,
50, 1112–1121. http://dx.doi.org/10.1038/s41588-
018-0147-3
Li, Y. R., & Keating, B. J. (2014). Trans-ethnic
genome-wide association studies: Advantages and
challenges of mapping in diverse populations. Ge-
8 DUNKEL, WOODLEY OF MENIE, PALLESEN, AND KIRKEGAARD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

nome Medicine, 6, 91. http://dx.doi.org/10.1186/
s13073-014-0091-5
Lynn, R. (2011). The chosen people: A study of
Jewish intelligence and achievement. Augusta,
GA: Washington Summit Publishers.
MacDonald, K. (1994). A people that shall dwell
alone: Judaism as a group evolutionary strategy.
Westport, CT: Praeger.
MacDonald, K. (1998). Separation and its discon-
tents: Toward an evolutionary theory of anti-
Semitism. Westport, CT: Praeger.
Murray, C. (2007). Jewish genius. Commentary. Re-
trieved from https://www.commentarymagazine
.com/articles/jewish-genius/
Nisbett, R. (2009). Intelligence and how to get it.
New York, NY: W. W. Norton and Co.
Odenstad, A., Hjern, A., Lindblad, F., Rasmussen, F.,
Vinnerljung, B., & Dalen, M. (2008). Does age at
adoption and geographic origin matter? A national
cohort study of cognitive test performance in adult
inter-country adoptees. Psychological Medicine,
38, 1803–1814. http://dx.doi.org/10.1017/S003
3291708002766
Piffer, D. (2015). A review of intelligence GWAS
hits: Their relationship to country IQ and the issue
of spatial autocorrelation. Intelligence, 53, 43–50.
http://dx.doi.org/10.1016/j.intell.2015.08.008
Scutari, M., Mackay, I., & Balding, D. (2016). Using
genetic distance to infer the accuracy of genomic
prediction. PLoS Genetics, 12, e1006288. http://dx
.doi.org/10.1371/journal.pgen.1006288
Sewell, W. H., Hauser, R. M., Springer, K. W., &
Hauser, T. S. (2004). As we age: The Wisconsin
Longitudinal Study, 1957–2001. In K. Leicht
(Ed.), Research in Social Stratification and Mobil-
ity (Vol. 20, pp. 3–111). London, United King-
dom: Elsevier.
te Nijenhuis, J., David, H., Metzen, D., & Armstrong,
E. L. (2014). Spearman’s hypothesis tested on
European Jews vs non-Jewish Whites and vs Ori-
ental Jews: Two meta-analyses. Intelligence, 44,
15–18. http://dx.doi.org/10.1016/j.intell.2014.02
.002
Tian, C., Plenge, R. M., Ransom, M., Lee, A., Vil-
loslada, P., Selmi, C.,...Seldin, M. F. (2008).
Analysis and application of European genetic sub-
structure using 300 K SNP information. PLoS Ge-
netics, 4, e4. http://dx.doi.org/10.1371/journal
.pgen.0040004
Woodley of Menie, M. A., Figueredo, A. J., Sarraf,
M. A., Hertler, S. C., Fernandes, H. B. F., &
Peñaherrera-Aguirre, M. (2017). The rhythm of the
West: A biohistory of the modern era AD 1600 to
the present.Journal of Social Political and Eco-
nomic Studies, Monograph Series, No. 37. Wash-
ington DC: Scott Townsend Press.
Woodley of Menie, M. A., Younuskunju, S., Balan,
B., & Piffer, D. (2017). Holocene selection for
variants associated with general cognitive ability:
Comparing ancient and modern genomes. Twin
Research and Human Genetics, 20, 271–280.
http://dx.doi.org/10.1017/thg.2017.37
Zanetti, D., & Weale, M. E. (2016). True causal
effect size heterogeneity is not required to explain
trans-ethnic differences in GWAS signals. bioRxiv.
Advance online publication. http://dx.doi.org/10
.1101/085092
Zhao, X., Lynch Jr., J. G., & Chen, Q. (2010). Re-
considering Baron and Kenny: Myths and truths
about mediation analysis. Journal of Consumer
Research, 37, 197–206.
Zuckerman, M., Silberman, J., & Hall, J. A. (2013). The
relation between intelligence and religiosity: A meta-
analysis and some proposed explanations. Personal-
ity and Social Psychology Review, 17, 325–354.
http://dx.doi.org/10.1177/1088868313497266
(Appendix follows)
9POLYGENIC SCORES MEDIATE JEWISH IQ ADVANTAGE
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Appendix
Supplemental analysis 1: The representativeness of Wisconsin Jews
It is possible, although unlikely, that the Jew-
ish population in Wisconsin is an outlier in
terms of socioeconomic status among Jewish
populations in the US. This possibility is diffi-
cult to investigate since there are no large stud-
ies of Jewish educational attainment by state.
Instead, to get an approximate estimate, we look
at the income of federal employees in 2017 and
compare the income of Jews to the income of
non-Jews in different states. The assumption is
that if the Jewish population in Wisconsin in
previous generations was an outlier compared
to other states, we would also see a higher
average income among the Jewish population in
Wisconsin in 2017. We acquired 446,603 fed-
eral salaries of people living in the largest cities
in the US from the Federal DataCenter, includ-
ing 14,828 salaries of people with Jewish an-
cestry as determined by surname. For every
person we calculate the relative salary, which is
the salary of that person divided by the mean
salary in the location at which the person works.
Finally, we look at whether the relative salaries
of Jews in Wisconsin cities is higher than the
relative salaries in other US states. We find that
the mean relative salary of Jews compared to
non-Jews is the same in Wisconsin as the US
average. This finding holds when using log
transformed salaries. The boxplots for the rela-
tive log transformed salaries are shown in Fig-
ure A1.
Received August 13, 2018
Revision received November 11, 2018
Accepted November 14, 2018 䡲
Figure A1. The relative (log-transformed) salaries of Jews
compared to non-Jews in Wisconsin vs other US states.
10 DUNKEL, WOODLEY OF MENIE, PALLESEN, AND KIRKEGAARD
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.