Content uploaded by Jakob Pietschnig
Author content
All content in this area was uploaded by Jakob Pietschnig on Apr 16, 2018
Content may be subject to copyright.
Perspectives on Psychological Science
2015, Vol. 10(3) 282 –306
© The Author(s) 2015
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/1745691615577701
pps.sagepub.com
Early in the 20th century, the first widely used standard-
ized psychometric intelligence measures were devel-
oped (Binet & Simon, 1905, 1908), and the IQ metric as
an estimate of cognitive abilities was introduced (Stern,
1912; Terman, 1921). Although IQ tests were originally
intended as a means to identify children in need of spe-
cial education (Binet & Simon, 1905), further uses for IQ
tests were rapidly discovered. Among other uses, IQ
tests quickly became common in academic contexts
(e.g., as a decision criterion for college acceptance), to
identify leadership personalities for the military, or for
personnel selection (e.g., Brooks, 1922). Nowadays,
even eligibility for subsidized special education place-
ments of children (Ceci & Kanaya, 2010; Kanaya & Ceci,
2007) or potential sentencing to capital punishment in
court (Flynn, 1999, 2009a) may in some countries depend
on IQ test results. Moreover, IQ has been shown to cor-
relate not only with various measures of performance
and job success, but also with phenomena seemingly
unrelated to mental capacity, for example, health and
longevity (Deary, 2009).
From early on, researchers were concerned about the
meaning of expected and actual changes in test scores
within the population (Cattell, 1937). Through the middle
of the century, rising scores were noticed but were mainly
attributed to statistical artifacts or sampling error instead
of being interpreted as genuine changes in population
test scores (Merrill, 1938; Tuddenham, 1948). Schaie and
Strother (1968) were the first researchers to interpret IQ
score changes as cohort effects. However, the first system-
atic description of national and international IQ change
patterns did not appear until the 1980s (Flynn, 1984,
1987). These studies invariably showed increases over
time in test performance on IQ tests and engaged the
attention of many researchers. Since then, generational IQ
test score changes within the general population have
577701PPSXXX10.1177/1745691615577701Pietschnig, VoracekMeta-Analysis of the Flynn Effect
research-article2015
Corresponding Author:
Jakob Pietschnig is now at the Department of Applied Psychology:
Health, Development, Enhancement and Intervention, University of
Vienna, Liebiggasse 5, A-1010 Vienna, Austria
E-mail: jakob.pietschnig@univie.ac.at
One Century of Global IQ Gains: A
Formal Meta-Analysis of the Flynn Effect
(1909–2013)
Jakob Pietschnig1,2 and Martin Voracek2,3
1School of Science and Technology, Middlesex University Dubai, United Arab Emirates; 2Department of
Basic Psychological Research and Research Methods, Faculty of Psychology, University of Vienna, Austria;
and 3Georg Elias Müller Department of Psychology, Georg August University of Göttingen, Germany
Abstract
The Flynn effect (rising intelligence test performance in the general population over time and generations) varies
enigmatically across countries and intelligence domains; its substantive meaning and causes remain elusive. This first
formal meta-analysis on the topic revealed worldwide IQ gains across more than one century (1909–2013), based
on 271 independent samples, totaling almost 4 million participants, from 31 countries. Key findings include that IQ
gains vary according to domain (estimated 0.41, 0.30, 0.28, and 0.21 IQ points annually for fluid, spatial, full-scale,
and crystallized IQ test performance, respectively), are stronger for adults than children, and have decreased in
more recent decades. Altogether, these findings narrow down proposed theories and candidate factors presumably
accounting for the Flynn effect. Factors associated with life history speed seem mainly responsible for the Flynn effect’s
general trajectory, whereas favorable social multiplier effects and effects related to economic prosperity appear to be
responsible for observed differences of the Flynn effect across intelligence domains.
Keywords
Flynn effect, meta-analysis, generational IQ gains, intelligence
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 283
become well known as the Flynn effect, an expression
introduced by Herrnstein and Murray in their widely
disseminated book The Bell Curve (1996, pp. 307–309).
Subsequently, the Flynn effect has been recognized as a
phenomenon of considerable importance, having been
labeled “one of the most striking phenomena in this
field“ and listed under the top research agenda for intel-
ligence research by the Task Force for Intelligence of
the American Psychological Association (Neisser etal.,
1996, p. 96).
Variety of Findings and Explanations
In the present article, we provide the first formal compre-
hensive meta-analysis of the Flynn effect, and we use our
results to assess the proposed theories of it. Past research
on the Flynn effect yielded quite erratic patterns of these
intelligence test score changes in different countries. In
general, these changes seem to be positive and rather
strong in most, but not all, of the investigated countries.
The strongest gains were observed in Austria, France,
Germany, Israel, Japan, Kenya, the Netherlands, and
Spain, whereas gains in Australia, Brazil, Ireland, New
Zealand, the United Kingdom, and the United States of
America were weaker, and in Norway and Sweden gains
seemed to have ceased altogether in recent years (Colom,
Flores-Mendoza, & Abad, 2007; Colom, Lluis-Font, &
Andres-Pueyo, 2005; Daley, Whaley, Sigman, Espinosa, &
Neumann, 2003; Flynn, 1987, 2009b). In addition, more
recent data from Denmark and Finland even suggest a
reversal of gains in the past couple of years (Dutton &
Lynn, 2013; Teasdale & Owen, 2005). Younger ages of
top-ranked chess players, better tournament performance
of younger bridge players, and increasing numbers of
scientific journal articles and patents have been cited as
real-world evidence for rising population intelligence
(Howard, 1999, 2001, 2005).
Different intelligence domains display gains of differ-
ent strengths (for clarity, we provide definitions of these
domains and several further central concepts in Table 1).
Stronger gains have usually been shown for fluid IQ than
for crystallized IQ, especially so for Anglo-American
countries (Jensen, 1998, pp. 319–320; Lynn, 2009a; Nisbett
et al., 2012). However, recent evidence from German-
speaking countries on measures of crystallized IQ showed
gains of about the same size as for fluid IQ in Anglo-
American countries (Pietschnig, Voracek, & Formann,
2010). Accordingly, the role of intelligence domain with
regard to these gains still remains unclear.
Another open question is whether these gains are
expressions of increases in general cognitive ability levels
of individuals (commonly referred to as psychometric g)
or rather whether these gains reflect increases in specific
ability domains. Gains have been reported to be related
to psychometric g (Colom, Juan-Espinosa, & Garcia,
2001), whereas in other cases no such associations were
observed (Te Nijenhuis, 2013; Woodley & Meisenberg,
2013). In contrast, negative associations between psycho-
metric g and IQ gains have been reported in another two
accounts (Must, Must, & Raudik, 2003; Rushton, 1999).
Further investigations revealed age as a moderator vari-
able, indicating differential strength of gains for either
children or adults, depending on specific subtests of
intelligence test batteries (Flynn, 2010).
Irrespective of the evidence for influences of modera-
tor variables, the strength of gains is usually reported in
the literature to amount to 3 IQ points per decade,
although this figure has been established on the basis of
American data only and despite various indications for
country-specific strengths of gains over different periods
(Flynn, 2009b; Williams, 2013; see Figure 1 for the trajec-
tories of gains in the present study). A number of studies
addressed the evidence for differential gains, noting larger
gains between the world wars (e.g., Lynn, 2009a, 2009b)
and decreasing gains since the 1980s (e.g., Flynn, 2007;
Storfer, 1990, p. 89). However, a comprehensive account
of time trends in IQ gains is as of now still unavailable.
The suspected causes of generational IQ gains are
manifold, and theories aiming to explain the Flynn effect
differ considerably. Environmental explanations focus on
the effects of education, family size (which may be seen
as a function of fertility), technology, and changes in test-
taking behaviors. Other proposed explanations relate
environmental to biological causes, including effects of
social multipliers (i.e., societal IQ increases that may act
as IQ-increasing factors in their own right) or health-
related effects, such as nutrition and pathogen stress. We
will provide a detailed discussion of these theories of
causes of IQ gains.
Scope of the Present Meta-Analysis
Here, we provide the first formal meta-analysis of the
available literature of generational IQ test score changes.
We report IQ test score changes as average change in IQ
points per year and assess a set of moderator variables
that are necessary to enable an appraisal of the different
proposed explanations for the Flynn effect.
Using this approach, we can examine the progress and
strength of the Flynn effect since the introduction of psy-
chometric intelligence testing in the early 20th century
and moderating influences of age, economic growth,
sample health status, and sex. In addition, we consider
the influence of test type (high, medium, and low
g-loaded tests, hereafter referred to as g-ness). We
assumed that crystallized IQ measures reflect the highest
g-ness, full-scale IQ a medium g-ness, and fluid IQ the
lowest g-ness (see Johnson, Bouchard, Krueger, McGue,
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
284 Pietschnig, Voracek
& Gottesman, 2004), thus allowing the examination of
associations between psychometric g and IQ gains. IQ
gain trajectories for different continents and the influence
of fertility as a proxy for average family size were exam-
ined in supplemental analyses.
In the Appendix, we describe our literature search and
inclusion criteria. In all, 219 studies met the inclusion cri-
teria, yielding 271 independent samples comprising
3,987,892 participants covering a time span of 105 years
(1909–2013). The earliest included evidence for test score
changes originated from the restandardization of the
Stanford-Binet Scales (period 1909–1932; Merrill, 1938).
The mean age of samples was 17.5 years (range of mean
age = 0.5–74.3 years), comprising 192 children and adoles-
cent samples (defined as mean participant age lower than
17 years) and 79 adult samples. Although mean sample
age varied widely, 90% of samples were younger than 38
years at the time of the testing. In terms of sample health
status, 186 samples comprised healthy participants, 83
comprised patients, and two were mixed samples. Patient
samples comprised individuals with various conditions
that are likely to affect cognitive performance (e.g., learn-
ing disabilities, psychiatric disorders), thus suggesting that
such samples would be expected to score within the lower
tail of the intelligence test performance distribution.
Generational IQ test performance changes were
reported for 31 distinct countries located on six conti-
nents: Africa (Kenya, South Africa, Sudan), Asia (China,
Israel, Japan, Saudi Arabia, South Korea), Europe (Austria,
Belgium, Bulgaria, Denmark, Estonia, Finland, France,
Germany, Ireland, Netherlands, Norway, Spain, Sweden,
Switzerland, Turkey, United Kingdom), North America
(Canada, Dominica, United States), South America
(Argentina, Brazil), and Oceania (Australia, New Zealand).
In two studies, changes were reported for participants
from more than one country but on the same continent
(Pietschnig et al., 2010; Uttl & Van Alstine, 2003). One
study provided changes on Raven’s Progressive Matrices
for participants from 45 different countries (Brouwers,
Van de Vijver, & Van Hemert, 2009). Descriptive charac-
teristics of the included samples are provided in Table S1
in the Supplemental Material available online.
Table 1. Central Concepts
Concept Description
IQ A well-known measurement metric of performance in cognitive tests. By consensus, average population IQ
test performance has been defined to be 100, with a standard deviation of 15 points. Because IQ points are
not directly observable, individual cognitive tests need to be initially standardized to allow a transformation
from raw test scores into IQ scores (or another metric if desired). Thus, the average test raw score of a large
and (ideally) population-representative standardization sample on any given test is assumed to correspond
to 100 IQ points, whereas the standard deviation corresponds to 15 IQ points. This means that the IQ of a
single individual or the average IQ of a sample of individuals reflects the test performance in comparison
to the standardization sample. Consequently, generational IQ test score changes are a result of changes in
population performance compared with the standardization sample.
Full-scale IQ Test/IQ score derived from an IQ test battery consisting of several subtests (e.g., Wechsler Adult Intelligence
Scale). Thus, full-scale IQ can be understood as an average test score of a variety of subtests that assess
several different intelligence domains. In many well-known IQ test batteries, full-scale IQ scores are
calculated as averages of crystallized and fluid IQ scores.
Crystallized IQ Test/IQ score derived from an intelligence test (or several subtests) that consists of knowledge-based questions
that cannot be solved by reasoning (e.g., naming the capital of a certain country). Verbal IQ tests typically
tap this domain (e.g., vocabulary tests).
Fluid IQ Test/IQ score derived from an intelligence test (or several subtests) that consists of reasoning-based tasks that
can be solved with (virtually) no prior knowledge (e.g., providing the next number in a series such as 2, 4,
6, 8, . . . ). Performance IQ tests typically tap this domain (e.g., Raven’s Progressive Matrices).
Spatial IQ Test/IQ score derived from an intelligence test that consists of tasks that require mental rotation of objects in
order to solve the items (e.g., deciding whether a rotated view of a cube is identical with a stimulus cube as,
for instance, in cube rotation tests).
Psychometric gThis term means the general cognitive ability of individuals. The central idea of this concept relies on the well-
known observation that in general test takers’ results on different cognitive tests that tap specific abilities are
positively intercorrelated. This positive manifold of IQ test performance is commonly seen as a manifestation
of psychometric g.
g-ness Different IQ domains have been shown to be more or less related to psychometric g. This means that some
subtests are more strongly related to g than others. In the present article, we use the expression g-ness to
indicate whether an intelligence domain is more or less related to psychometric g. Presently, we categorized
IQ domains according to previous research (Johnson, Bouchard, Krueger, McGue, & Gottesman, 2004) into
either possessing high (crystallized IQ), medium (full-scale IQ), or low g-ness (fluid IQ).
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 285
Summary of Key Findings
Our analyses reveal eight key results that we believe
are central for an appraisal of the specific proposed
explanations for the Flynn effect. These key results are
summarized across the top of Table 2 and detailed below.
First and foremost, we note that this first formal meta-
analysis of the Flynn effect provides strong evidence for
continuous global generational IQ test score gains in the
general population over the past century. Annual changes
across all studies and domains ranged from −0.76 to 1.98
IQ points. Visual distributional inspection of these
changes suggested that annual changes were overwhelm-
ingly positive (90%, 87%, 93%, and 100% of changes were
directionally positive for full-scale, crystallized, fluid, and
spatial IQ, respectively; see Figure 2; IQ test score
changes for all available domains in each investigated
country are summarized in Table S2 in the Supplemental
Material). When taken together, gains of about two stan-
dard deviations from 1909 to 2013 or 0.28 IQ points
annually (or 2.8 points per decade) were observed (Fig.
1). This global gain estimate corresponds closely to pre-
vious estimates of general intelligence test score changes
of about 2.50 (Storfer, 1990, p. 111) and up to 3.00 IQ
points per decade (Flynn, 1987, 2009b).
Changes of gains appeared to be remarkably closely
associated with historical events across all investigated
intelligence domains. Gains were stronger between
World Wars I and II but showed a marked decrease dur-
ing the World War II years (about 0.72 vs. 0.21 IQ points
annually; see upper third of Table S3 in the Supplemental
Material). This observation is consistent with previous
studies reporting larger gains between World War I and
World War II (e.g., Lynn, 2009a, 2009b). Following the
1940s, full-scale IQ test score gains increased and then
remained rather stable until the 1970s but were subse-
quently decreasing again. A virtually identical pattern
was observed for crystallized and fluid IQ gains (Table
S3). These observations may reflect influences of poor
nutrition and marked environmental stress experienced
by the general population in regions that were most
affected by the world wars, although we did not test this
directly. In the below summary of our eight key findings,
we discuss first differences in the strength of gains
between IQ domains, then IQ trajectories over time, and
finally influences of moderator variables (the first five
findings are illustrated by Figure 1 but also have support
in the Supplemental Material).
1. Substantial gains for fluid IQ
Gains amounted to 0.41 IQ points per year from 1924 to
2013 (lower third of Table S3). Joinpoint regression1
yielded five segments (each segment corresponds to a
specific time span), indicating substantial gains in all seg-
ments. The strongest gains were observed in the first seg-
ment, amounting to 0.93 IQ points per year from 1924 to
1935. Subsequently, weaker gains of 0.58 IQ points until
1938 and 0.20 points until 1952 were observed. Yearly
gains increased to 0.43 IQ points until 1985 and showed
gains of 0.22 points until 2013.
2. Substantial gains for crystallized IQ
Annual gains for crystallized IQ were somewhat weaker
than for full-scale IQ, amounting to 0.21 IQ points from
1912 to 2011 (second third of Table S3). Joinpoint regres-
sion analysis again yielded five segments, indicating
rather strong gains in the first segment but leveling off
from 1937 to 1948 (0.26 and 0.04 IQ points annually).
Subsequently, gains proceeded at 0.36 IQ points annually
until 1962 and then resumed at a rate of 0.30 IQ points
until 1987. It is interesting to note that this analysis shows
that gains in crystallized IQ virtually ceased in the last
segment (1987–2011), yielding changes of only 0.04 IQ
points per year.
3. Stronger gains for fluid than
crystallized IQ
Gains appeared to be differentiated with respect to
intelligence domain. Annual gains in fluid IQ were sub-
stantially stronger than those in crystallized IQ (4.1 vs.
2.1 IQ points per decade, respectively). In contrast to
crystallized and full-scale IQ, gains remained substantial
for fluid IQ in each segment, yielding gains of at least
0.20 IQ points annually across the whole study period.
Fig. 1. Domain-specific IQ gain trajectories for 1909–2013. Changes
are based on weighted average annual IQ changes in all available data.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
286
Table 2. Proposed Explanations for IQ Gains
Key Results in Present Evidence
Theory Prior Key Evidence
Substantial
Gains for
Fluid IQ
Substantial
Gains for
Crystallized
IQ
Stronger
Gains for
Fluid Than
Crystallized
IQ
Decreasing
Gains
in More
Recent
Decades
Nonlinear
Gains
Stronger
Gains for
Adults
Than
Children
Positive
Association
With GDP
Change
per Capita
Stronger
Gains on
Low-g
Tests
Environmental Factors
Education Strong IQ gains after controlling for education
(Pietschnig, Tran, & Voracek, 2013); gains
are observable preceding school enrollment
(e.g., Lynn, 2009b)
✓ ✓ ✕ ✓ ✓ ✓ ✓
Exposure to
technology
Influence of technology exposure on IQ task
performance unclear (Boot, Kramer, Simons,
Fabiani, & Gratton, 2008); substantial gains
in countries with comparatively limited
accessibility of technology (Daley, Whaley,
Sigman, Espinosa, & Neumann, 2003)
✓ ✓ ✕ ✕ ✕ ✓ ✓
Family size IQ family size association not unequivocal
(Wichman, Rodgers, & MacCallum, 2006);
gains in countries with little family structure
change (Khaleefa etal., 2008)
✓ ✓ ✕ ✓ ✕
Test-taking
behavior
Direct evidence for gains in both crystallized
and fluid IQ (Must & Must, 2013; Pietschnig
etal., 2013); after controlling for changes in
test-taking behavior, substantial gains remain
(Must & Must, 2013; Pietschnig etal., 2013)
✓ ✓ ✓ ✓ ✕ ✓
Biological Factor
Hybrid vigor Simulation studies show performance-
increasing effects of hybrid vigor (Mingroni,
2007); gains can account for only about 3 IQ
points in 50 years under optimal outbreeding
conditions (Woodley, 2011, 2012b)
✕ ✕ ✓ ✕ ✕ ✕
Hybrid Factors
Blood lead
levels
Coinciding IQ gains and decreasing blood
lead levels in the United States (Kaufman
etal., 2014; Nevin, 2000); strongest gains
should be observed in preindustrial areas
and countries as well as after phasing out
lead gasoline and banning of lead paint
(Steen, 2009)
✓ ✓ ✕ ✕ ✓ ✕
Genomic
imprinting
As of this writing, there is no direct empirical
evidence available
✓ ✓ ✓ ✕ ✕
(continued)
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
287
Key Results in Present Evidence
Theory Prior Key Evidence
Substantial
Gains for
Fluid IQ
Substantial
Gains for
Crystallized
IQ
Stronger
Gains for
Fluid Than
Crystallized
IQ
Decreasing
Gains
in More
Recent
Decades
Nonlinear
Gains
Stronger
Gains for
Adults
Than
Children
Positive
Association
With GDP
Change
per Capita
Stronger
Gains on
Low-g
Tests
Nutrition Decreasing IQ test score variance in general
population (Colom, Lluis-Font, & Andres-
Pueyo, 2005; Pietschnig etal., 2013);
evidence for coinciding IQ gains and
nutritional improvements (Lynn, 2009b)
✓ ✓ ✓ ✓ ✕ ✓ ✕
Pathogen
stress
National IQ test scores associated with
prevalence of infectious disease (Eppig,
Fincher, & Thornhill, 2010)
✓ ✓ ✓ ✕ ✓ ✕
IQ variability Rather a symptom than a cause for IQ gains;
direct evidence for decreasing IQ variability
has been provided in several studies (e.g.,
Colom etal., 2005; Pietschnig etal., 2013)
✓ ✓ ✕ ✕ ✓
Social
multipliers
Social multipliers would be able to account for
increases in IQ heritability with age (Dickens
& Flynn, 2001); direct empirical evidence
for social multipliers unavailable as of this
writing
✓ ✓ ✓ ✓
Life history
speed
Proposes IQ gains as a function of several
different factors that are related to life
history speed, such as education, family size,
nutrition, and pathogen stress; first empirical
evidence supporting this theory has been
provided (Woodley, Figueredo, Brown, &
Ross, 2013); consistent with the present
observation of IQ changes being unrelated
to psychometric g
✓ ✓ ✓ ✓ ✓ ✓
Note: ✓ = theory consistent with observation, ✕ = theory inconsistent with observation; blank cells indicate no specific prediction of the theory; GDP = gross domestic product.
Table 2. (continued)
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
288 Pietschnig, Voracek
4. Decreasing gains in more recent
decades
The decreasing strength of the IQ gains over time was
reflected by meaningful negative effects of time span for
full-scale, fluid, and crystallized IQ (all ηP
2>.18), as well
as year of onset for fluid and spatial IQ (all ηP
2> .04).
Supported by the observed IQ change trajectories, evi-
dence for decreasing gains in recent decades can be con-
sidered to be robust. Regression slopes of joinpoint
regressions significantly decreased in the last segment of
all IQ domains (Table S4 in the Supplemental Material;
decreases of strength of gains were also obvious when
data from different continents were inspected separately;
Fig. S1 in the Supplemental Material). This may conceiv-
ably indicate a development toward an end and perhaps
an ultimate reversal of the IQ gains, as has been recently
reported for Scandinavian countries (Dutton & Lynn,
2013; Flynn, 1987, 2009b; Teasdale & Owen, 2005).
5. Nonlinear gains
Joinpoint regression analyses for all IQ domains showed
a significantly better fit for regression models assuming
changes in the strength of regression slopes over time
than for models without incorporating changes in the
slopes (ps < .001 for comparisons of accepted models vs.
models with no or fewer joinpoints in all segmented line
regressions). This result indicates that IQ test score gains
have not been linear over the past century, but rather
seem to have been alternately accelerating and decelerat-
ing and finally decreasing during more recent years.
Storfer already has proposed such changes in the strength
of gains over time, estimating gains of 3.75 IQ points per
Fig. 2. Domain-specific distributions of observed IQ changes. Frequencies refer to absolute numbers of annual IQ changes
in all observed samples.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 289
decade from 1900 to 1920, of 2.50 from 1920 to 1960, and
slightly smaller gains after the 1960s (Storfer, 1990, p.
439). In contrast, the present evidence indicates that a
decrease in the strength of gains emerged only in the
mid-1970s, yielding moderate gains of 2.30 IQ points per
decade. However, the pattern preceding this period
appears to be considerably more differentiated, indicat-
ing that gains during the early 20th century have been
relatively weak (0.80 IQ points per decade), then showed
a sharp increase in the 1920s (7.20 IQ points per decade),
decreased again from 1935 to 1947 (2.10 IQ points per
decade), but later again recovered until 1976 (3.00 IQ
points per decade; Table S3).
6. Stronger gains for adults than
children
Stronger gains were observed for adults than for children,
showing large effects for fluid and spatial IQ (ηP
2 = .28
and .66, respectively). Past research has attributed increas-
ing gains with age mainly to effects of better education. If
so, then educational effects would be expected to affect
crystallized IQ most (e.g., Flynn, 2010). However, effects
of age on crystallized and full-scale IQ were negligible in
the present study, although the signs of the change coef-
ficients were directionally as expected (Table S4).
7. Positive Associations With Gross
Domestic Product Change per Capita
Gross domestic product (GDP) growth per capita was
substantially positively associated with full-scale, crystal-
lized, and spatial IQ (ηP
2 = .09, .18, and .50, respectively),
but it showed negligible effects for fluid IQ. This finding
is consistent with previous reports of links of economic
prosperity with IQ in several nations (Lynn & Vanhanen,
2002). Associations between IQ gains and GDP have
been linked to educational improvements (Rindermann,
2008), thus conceivably reflecting effects of better educa-
tional infrastructure.
8. Stronger gains on low-g tests
IQ gains appeared not to be taking place on psychomet-
ric g. Findings of meaningful negative effects of medium
and high g-ness of tests on IQ gains (ηP
2 = .12 and .02,
respectively) are supported by the overall lower gains
observed in crystallized IQ (i.e., the domain with highest
g-ness). These findings are consistent with previous evi-
dence showing negative associations between g-ness and
IQ gains (Te Nijenhuis & van der Flier, 2013) and cor-
roborate the importance of environmental influences on
generational IQ test score changes (Rushton, 1999).
Further moderators
Sample type (patients vs. nonpatients) did not play an
important role for the Flynn effect, indicating virtually
identical IQ gains for healthy compared with patient-
based samples across all IQ domains. When average
national fertility rates were added to the regression mod-
els as additional predictors, the signs of all significant
coefficients remained directionally unchanged (except
the sign for year of onset for full-scale IQ, which became
positive but did not reach significance). Average national
fertility rates were positively related to IQ gains in full-
scale and fluid IQ, yielding small to medium effects.
These findings are in contrast to reports linking IQ gains
to decreasing family size (Zajonc & Mullally, 1997). The
positive association should be interpreted with caution,
as the available data allowed investigation of this relation
on only a portion of the total time span (namely, 1960–
2013), and fertility rates may be considered as only a
crude proxy for family size. In any case, the positive sign
emerging in this analysis does not lend credence to the
notion that IQ gains could be due to decreases in family
size (Table S5 in the Supplemental Material).
Causes of Gains
The present account of the accumulated evidence for the
Flynn effect over the past century allows a closer exami-
nation of the proposed causes of this effect. We provide
key findings of the present analysis across the top of
Table 2 and summarize key evidence from previous stud-
ies examining specific proposed theories for the Flynn
effect. Along with the more specific results as detailed
below, Table 2 therefore allows an evaluation of the
potential contributions of the individual theories. In what
follows, we address and evaluate each of these theories,
proposing environmental, biological, and hybrid (i.e.,
interacting biological and environmental) causes for IQ
gains, in turn.
Environmental factors
The environmental factors include education, technol-
ogy, family size, and test-taking behavior.
Education. A number of authors have proposed
increased mean educational years (Williams, 1998) and
improvements in educational systems in industrialized
societies as potential causes for rising IQ scores (Husén
& Tuijnman, 1991; Teasdale & Owen, 2005). Although
both of these reasons seem intuitively quite plausible,
negligible gains or even declining task performance on
achievement tests during periods of observed IQ gains
repeatedly have been cited as evidence against schooling
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
290 Pietschnig, Voracek
as a major cause for the Flynn effect (e.g., Flynn, 1984).
Subsequently, it has been argued that this finding might
be due to the increasing number of individuals taking
achievement tests, specifically the Scholastic Aptitude
Test, in more recent years. In this regard, changes in edu-
cational systems might have allowed individuals from all
educational backgrounds (i.e., lower-performing individ-
uals) to seek entry into higher education, thus explaining
a rise of IQ scores in the presence of a decline of scores
on the Scholastic Aptitude Test (Fuggle, Tokar, Grant, &
Smith, 1992; Teasdale & Owen, 1987). Accordingly,
declining achievement test scores may likely be due to
the changed demographics of the individuals taking the
test rather than actual student ability.
At least three aspects of the present evidence support
the role of education as an important contributing factor
to the explanation of IQ gains: First, we observed sub-
stantial global increases in crystallized IQ. This may
reflect, at least to a certain extent, effects of more and
better schooling. Positive associations between crystal-
lized IQ task performance and highest educational quali-
fication are widely accepted (e.g., McArdle, Ferrer-Caja,
Hamagami, & Woodcock, 2002) and have previously
been shown to be associated with gains on crystallized
IQ measures. Nonetheless, although education has been
shown to account for portions of crystallized IQ gains,
gains have been reported to remain substantial after con-
trolling for education (Pietschnig, Tran, & Voracek, 2013).
Second, larger IQ gains were observed for adults than
for adolescents and children in fluid and spatial IQ
domains. Surprisingly, no meaningful effect of age on
crystallized IQ was found. However, this may be due to
the effect of growing GDP, which could mask age effects
(indeed, preliminary regression analyses of the present
data set without consideration of GDP growth yielded
the expected age effect). Increasing numbers of average
educational years (see, for instance, Ceci, 1991, for such
reports in industrialized countries) may therefore explain
stronger gains for adults than for children.
However, although IQ test performance of children
and adolescents has been observed to increase to a lesser
extent than that of adults, gains for young samples still
were substantial in our data. Although increases in formal
educational years may not play a crucial role among chil-
dren and adolescent IQ gains, increasing exposure to
early childhood education programs, as witnessed in
more recent years, might do so (e.g., Gorey, 2001).
Although most such programs are aimed at more mature
children, some of these programs are aimed at infants
(e.g., the U.S.-based ABCDerian project, which had aver-
age enrollment ages of 4.4 months; for an overview, see
Hunt, 2011, pp. 288–291). Thus, even IQ gains in infants
(Campbell, Siegel, Parr, & Ramey, 1986; Lynn, 2009b;
Thompson, 2012) may be suitably explained by
educational improvements. However, it remains difficult
to decide the explanatory potential of early education
programs for IQ gains in children and adolescents,
because such programs differ considerably in coverage
and availability between (and even within) investigated
countries and time spans.
Third, IQ gains were predicted by average increases in
GDP per capita across all domains, with the exception of
fluid IQ. Positive associations of GDP with IQ gains have
been observed in several studies and countries (for an
overview, see Lynn & Vanhanen, 2002). In particular, the
substantial effect of GDP on crystallized IQ may be linked
to educational effects. It has been proposed that invest-
ments in better education lead to economic growth and
vice versa, thus leading to a positive feedback loop of
economic prosperity, education, and intelligence
(Rindermann, 2008). Of note, it has been shown more
recently that increases in GDP may be better described as
a function of education rather than the other way round
(Rindermann, 2012), which in turn would reverse the
causality assumption of the regression model applied by
us. Regardless of the causality of the observed associa-
tion, the positive sign of the association is consistent with
this theory.
These findings show that there is little doubt that edu-
cation plays a role in explaining the Flynn effect.
Nonetheless, schooling is unlikely to account for the full
extent of the IQ gains, and in particular the large gains
for fluid IQ cannot be attributed to better education.
Effects of technology. Exposure to technology of the
general population, at least in industrial countries, cer-
tainly has increased in more recent decades. It has been
suggested that the reported gains may reflect effects of
increased exposure to modern appliances that inciden-
tally train visual analytical abilities (Neisser, 1997). A
more stimulating visual environment (e.g., owing to more
exposure to computers, television, or video games)
would thus act as a facilitating factor for abilities required
to successfully master intelligence tasks.
This explanation is consistent with the present obser-
vation of the strongest gains having taken place on mea-
sures of fluid IQ. Moreover, consistent with the present
results, effects of technological advancement have been
linked to increases in GDP (Rindermann, 2008). One
main argument for this theory is that the IQ rise has been
observed mainly in Western industrialized countries
where technology has been readily accessible to the
major part of the population (Neisser, 1997). However,
more recent accounts have reported substantial IQ gains
in lesser-developed countries where ubiquitous exposure
to modern visual media would not have been expected.
For instance, IQ gains of about 1.8 IQ points per year
from 1984 to 1998 were demonstrated in rural Kenya
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 291
(Daley etal., 2003), which indicates that substantial IQ
gains may occur without effects of technology playing a
major role. In a similar vein, the strongest IQ gains during
the earliest decades of the present study are inconsistent
with this theory, because during the early 20th century,
the general availability of many now common appliances
to the general population had been very limited.
Finally, there is no conclusive evidence for increased
fluid task performance of individuals that are habitually
more frequently exposed to visual media. Investigations
of video game experts and nonexperts showed no differ-
ences in task performance on the Raven’s Progressive
Matrices between these groups (Boot, Kramer, Simons,
Fabiani, & Gratton, 2008). However, on the basis of the
available evidence, potential effects of incidental expo-
sure to modern technology as root causes for IQ gains
cannot be completely dismissed.
Family size. Decreasing family size has been associated
with increasing cognitive task performance at least in
Western countries (Zajonc & Mullally, 1997). On the basis
of exemplary calculations for the U.S. state of Iowa, it has
been concluded that an increase of about 1.4 IQ points
per decade could be explained by decreasing family size.
Although such an increase would have been suitable to
explain only a portion of the gains found in the same
time span in this state, it has been argued that besides the
birth-order effect, collective potentiation would be able
to contribute further toward explaining the gains (Zajonc
& Mullally, 1997). Nonetheless, family size has not
unequivocally been found to be related to cognitive task
performance (Wichman, Rodgers, & MacCallum, 2006).
In the present study, we were able to directly assess
influences of family size on IQ changes beyond the spe-
cific regional finding, as described above. To do so, we
included average national fertility rates as a predictor of
IQ change from 1960 to 2012 in our regression model. In
contrast to our expectations, fertility rate was significantly
positively related to IQ gains for full-scale and fluid IQ
domains, showing stronger gains in the presence of
higher average fertility rates. The significant positive
association found here should be interpreted with cau-
tion, because fertility rate may be viewed only as a proxy
for family size, and the findings are limited to data from
1960 onward. However, the present findings render it
quite unlikely that effects of decreasing family size are
substantially contributing to IQ gains.
Test-taking behavior. As many modern psychometric
test instruments are based on multiple-choice response
formats, it has been suggested that changed response
behavior (principally, guessing) on such formats may
have led to changes in test scores (Brand, 1987a, 1987b,
1996; Brand, Freshwater, & Dockrell, 1989), although it
has been asserted early on that guessing effects alone
have limited explanatory potential for gains (Flynn, 1990;
for a rejoinder, see Brand, 1990). The highest IQ gains
have typically been shown for measures of fluid IQ,
which are assumed to be largely independent of educa-
tional backgrounds of test takers, rather than for mea-
sures of crystallized IQ, where education should play a
more important role. Because multiple-choice response
formats are more common in fluid measures, it has been
argued that lowered levels of caution, conscientiousness,
and conservatism on social attitudes as well as higher
levels of extraversion across Western countries (see
Twenge, 2001, for evidence of increasing extraversion)
emerging since the advent of such guessing-inviting,
speeded group tests, rather than real-world improve-
ments of cognitive abilities, may be more suitable to
explain IQ test score gains (Brand, 1990). Consequently,
increased risk-taking behavior concomitant with increased
guessing behavior on psychometric test instruments may
be a cause of higher IQ test scores.
Recent direct tests of changes in test-taking behavior
showed strong evidence for guessing effects as important
contributors for fluid IQ gains (Must & Must, 2013) as
well as for crystallized IQ gains (Pietschnig etal., 2013).
Of note, it has been shown that guessing effects are in
fact related to gains on psychometric g, because guessing
is most frequently encountered on difficult rather than
easy test items (i.e., items with higher g-ness). However,
when guessing was controlled for, the Flynn effect
appeared to be independent of g (Woodley, Te Nijenhuis,
Must, & Must, 2014). Accordingly, the present data render
increased guessing behavior as a potentially contributing
but limited source for IQ gains, although, consistent with
guessing-related effects, the strongest gains indeed were
seen for fluid IQ.
Conversely, the independence of IQ gains from psy-
chometric g thus indicates that guessing is not the main
cause of observed IQ gains. This should not come as a
surprise, as we already observed gains for the initial
decades of the 1900s, well preceding the advent and
massive use of most of the guessing-inviting, speeded
group tests, such as Raven’s Progressive Matrices (Raven,
1938), which would be most likely to be affected by
changes in test-taking behavior. Similarly, guessing-
related IQ gains cannot be expected to continue indefi-
nitely, as eventually a ceiling must be reached. Decreasing
IQ gains in more recent years might be an expression of
a beginning saturation of guessing-related gains (see
also Williams, 2013). Moreover, the considerable crystal-
lized IQ gains in our data would be difficult to be
explained by guessing behavior alone, as the majority of
crystallized IQ measures in the present investigation
were knowledge-based tests that leave only little room
for guessing.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
292 Pietschnig, Voracek
Biological factor: Hybrid vigor
In contrast to the theories discussed above, this theory
focuses exclusively on genetic mechanisms as potential
causes for IQ gains. Positive associations between aver-
age allelic heterozygosity within populations and cogni-
tive task performance have been demonstrated in several
studies and now appear to be widely accepted (Jensen,
1998, pp. 189–197). Hybrid vigor refers to the mating of
individuals from genetically dissimilar subpopulations,
thereby increasing allelic heterozygosity and reducing
homozygosity. Consequently, it has been argued that IQ
gains might be due to contemporary increases in the
mobility of individuals and the ensuing lower numbers
of consanguineous (close-relative) and endogamous (in-
group) marriages and offspring during the past decades
(Mingroni, 2004, 2007).
When simulating effects of hybrid vigor on observed
allelic frequencies in isolated tribal villages of native
Indian tribes in Brazil (Mingroni, 2004) and in villages in
the Italian Parma Valley (Mingroni, 2007), it could be
shown that even small trends toward more random (i.e.,
out-group) mating would considerably reduce allelic
homozygosity. However, although this model appears to
offer a plausible mechanism for the explanation of the
Flynn effect, the relevance of this mechanism is limited
when considering the strength and pace of IQ gains. An
evaluation of the proposed model indicated that even
under optimal conditions (i.e., maximal outbreeding),
such a mechanism may be held accountable for only a
portion of observed IQ gains, because hybrid vigor effects
would neither be fast nor strong enough to explain gains
amounting to more than 3 IQ points over 50 years
(Woodley, 2011, 2012b). Keeping the magnitude of the
actually observed global gains in mind (altogether 30 IQ
points over the past century), hybrid vigor appears to play
only a minor role in explaining the Flynn effect. In con-
trast, it has been recently suggested that dysgenic effects
(i.e., disadvantageous genetic effects of higher reproduc-
tion rates of low-ability as opposed to high-ability indi-
viduals within populations) may well be responsible for
the observed reversal of the Flynn effect in Finland
(Dutton & Lynn, 2013) as well as for slower reaction times
on simple intelligence-related reaction tasks of the gen-
eral population in a number of countries over more than
one century (Woodley, Te Nijenhuis, Must, & Must, 2014).
Hybrid factors
This class of proposed explanations includes a number
of basic background factors, such as blood lead levels,
genomic imprinting, nutrition, and reduced pathogen
stress. Moreover, more complex, integrative factors that
may be interpreted as a consequence of these various
basic causes, such as reduced IQ variability, effects of
social multipliers, and decreasing life history speed, have
been proposed in the literature.
Blood lead levels. Detrimental effects of environmental
exposure to lead on cognitive abilities are well docu-
mented. Children have been shown to be particularly sus-
ceptible to lead exposure, because even small increases in
blood lead levels may impair neural development (e.g.,
Steen, 2009). With the advent of industrialization, substan-
tial increases in blood lead levels of the general popula-
tion have been observed (effects of lead paint poisoning
in children were recognized initially in the 1980s), indicat-
ing 100–1,000 times higher blood lead–level concentra-
tions in industrialized compared with preindustrial
societies (Koller, Brown, Spurgeon, & Levy, 2004). More-
over, IQ gains in the United States have been linked to the
banning of lead paint and phasing out of lead gasoline in
the 1970s (Kaufman etal., 2014; Nevin, 2000).
Although consistent with strong gains for fluid IQ and
decreasing gains in more recent years (as one would
expect a beneficial effect of lead reduction to reach a
ceiling), stronger gains in adults than in children cannot
be reasonably explained by this theory. Indeed, the
explanatory potential of this theory appears to be limited
to gains following the 1970s, as only subsequent to this
period did restrictions pertaining to the use of lead paint
and gasoline take effect in most countries (Lovei, 1998).
In fact, increasing lead exposure, preceding lead restric-
tion efforts in the United States and other countries
(Nevin, 2000), might have been responsible for weaker
IQ gains than in environments with lower lead exposure.
Arguably, gains amounting to 4–5 IQ points since the
1970s that have been attributed to lead restrictions in the
United States (Kaufman etal., 2014) may in turn be taken
as a sign for potential IQ gains being depressed prior to
these bans.
It needs to be acknowledged that population expo-
sure to lead is likely to have differed considerably
between countries, thus making it difficult to decide
whether effects of similar magnitude took place in differ-
ent countries. A direct assessment of effects of lead on IQ
gains was not possible in our meta-analysis, because
blood lead–level reports were unavailable or largely
incomplete for several of the included countries. However,
the above evidence indicates that reduced lead exposure
may well account for a portion of IQ gains in more recent
years.
Genomic imprinting. Based on the idea that environ-
mental conditions might be able to evoke fast-emerging
biological effects, genomic imprinting was proposed in
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 293
the 1980s as an epigenetic inheritance mechanism that
works in addition to the well-known Mendelian mecha-
nisms (Surani, Barton, & Norris, 1984). Genomic imprint-
ing means that environmental conditions affect
reproduction information (i.e., the male sperm) and ulti-
mately genetic expressions in children and even in grand-
children. Specifically, for IQ gains it has been argued that
visually stimulating environments may lead to changes of
the father’s sperm, which in turn produces cognitive abil-
ity gains of the offspring (Storfer, 1999).
Although genomic imprinting in principle would be
suitable to explain IQ gains, this hypothesis remains dif-
ficult to test (Mingroni, 2007). Although influences of
genomic imprinting cannot be completely ruled out, in
the light of the observed fluctuations in the slope of IQ
test score gains over time, this mechanism appears to be
insufficient to plausibly explain the observed pattern in
our data.
Nutrition. It is undisputed that nutrition affects the
makeup of the human body. For instance, increases in
average body height in developed countries typically
have been attributed to better nutrition. Such increases
are still ongoing but have been decelerating in the late
twentieth century (e.g., Cole, 2003). Also, nutrition has
been shown to be associated with other biological
characteristics, such as head circumference (Lucas,
Morley, & Cole, 1998), which in turn has been shown
to be a rough correlate of IQ test performance (Rush-
ton, 1997). Although recent meta-analytical findings of
correlations between brain volume, as assessed by
modern brain imaging techniques, and IQ task perfor-
mance suggest that previously calculated effects most
probably were inflated, the association with such phys-
ical features appears to be robust (Pietschnig, Zeiler, &
Voracek, 2011). Moreover, poor nutrition has been
shown to be associated with low IQ test performance
in numerous countries (e.g., Hunt, 2011, p. 260).
Improved nutrition may also plausibly explain stronger
fluid than crystallized IQ gains, as fluid task perfor-
mance appears to be more strongly affected than crys-
tallized task performance in malnourished individuals
(Lynn, 2009b).
Consequently, and based on the observation that IQ
test score gains and improvements in nutrition of the
general population coincided in past decades, better cog-
nitive development due to improved prenatal and post-
natal nutrition has been suggested as a plausible cause of
the Flynn effect (Lynn, 1989, 1990, 2009b; but see Flynn,
2012, pp. 40–52). Trajectories of IQ changes in our data
are consistent with this theory. Decreasing gains in more
recent decades may well be due to the beneficial effects
of nutrition having reached a ceiling (at least in devel-
oped nations), although most likely there remains room
for nutritional improvement in parts of the general popu-
lation, as discussed below.
As the malnourished portions of a respective country’s
population would be expected to benefit more from
improved nutrition than their already well-nourished
compatriots (Nisbett etal., 2012), a narrowing of the dis-
tribution of IQ test performance should be observed.
Although we did not directly test IQ variability, recent
findings show evidence for a decreasing variability of IQ
test scores in the general population of several countries
(see IQ Variability section below).
Conversely, it needs to be acknowledged that not all
members of a specific country’s population are necessar-
ily exposed to the same quality of nutrition. Nutrition
quality has previously been linked to socioeconomic sta-
tus (SES; e.g., Kant & Graubard, 2006), indicating poorer
diets in low-SES individuals and therefore essentially
allowing for further nutritional improvements of these
population segments. As suboptimal nutrition in devel-
oped countries appears to be due to a lack of affordabil-
ity of high-quality nutrition among low-SES individuals, a
lack of dietary awareness, or both, food assistance pro-
grams and dietary guidance may potentially lead to
higher-quality diets and, as a consequence, to further
improvements of IQ test performance among these pop-
ulation segments.
If improved nutrition were the sole cause of IQ gains,
we would expect to observe more or less identical IQ
gains at any age starting from infancy (Lynn, 2013).
Indeed, we observed nontrivial gains for children sam-
ples, thus suggesting an important role of nutrition for the
Flynn effect. However, in our data, adults showed sub-
stantially stronger gains than children and adolescents.
Stronger IQ gains of adults suggest meaningful effects of
further causes that emerge only later in development.
Pathogen stress. Similar to the nutrition hypothesis, the
pathogen stress hypothesis emphasizes the importance
of biological makeup for cognitive performance. Brain
development in children demands a large percentage of
the metabolic turnover (estimated to amount up to 87%
in newborns and 34% at age 10; see Holliday, 1986), an
energy demand that needs to be met to ensure cerebral
development. Unavailability of these resources may
impair optimal cerebral development, consequently
affecting cognitive abilities. Fending off aversive patho-
gens necessitates considerable amounts of energy,
thereby removing important resources from brain devel-
opment in early childhood. In this vein, it has been
shown that average national IQs are negatively related to
the prevalence of infectious diseases around the world
(Eppig, Fincher, & Thornhill, 2010).
Environmental conditions have undoubtedly improved
in the past decades in developed countries, but also in
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
294 Pietschnig, Voracek
the less-developed ones, thereby creating environments
with less pathogen stress to individuals. Increased avail-
ability of health services and better hygienic conditions
lower the prevalence of infectious diseases and other
pathogens, thus providing better conditions for cerebral
development. Following this argument, less risk of expo-
sure to pathogen stress in contemporary populations may
be yet another cause for IQ gains. Thus, modern medical
assistance and health care in developed countries allow
for an allocation of bodily resources to cognitive develop-
ment that in the past would have been needed for the
containment of environmental pathogen stress.
Similar to improved nutrition, reduced pathogen stress
would be suitable to explain stronger fluid than crystal-
lized IQ gains and would be expected to emerge in
infancy. Again, the observed stronger gains for adults
than for children and adolescents would be difficult to
reconcile by this theory, thus suggesting other important
factors contributing to the Flynn effect. However, effects
of nutrition and pathogen stress remain difficult to disen-
tangle in the context of the present research.
IQ variability. In most of the proposed explanations
for the Flynn effect, it has been (implicitly) assumed that
the IQ increases are due to a shift of the overall ability
distribution within the respective populations. However,
systematic shifts of single parts of the distribution could
also result in a mean change of the ability distribution by
either increasing or decreasing the variance (Rodgers,
1999; Rowe & Rodgers, 2002).
There is some evidence for decreasing variability of IQ
task performance and a narrowing of IQ distributions in
several countries, although decreasing variability has not
been found in all accounts (e.g., Dickens & Flynn, 2002;
Pietschnig et al., 2010). Indeed, the available evidence
indicates that decreases of IQ variability may have played
a role in only some, but not all, countries (for an over-
view, see Flynn, 2012, pp. 41–42). However, a consider-
able number of recent investigations reported direct
evidence for decreasing IQ variability over time in Austria,
Denmark, Norway, Spain, and the United States (Colom
et al., 2005; Pietschnig et al., 2013; Rindermann &
Thompson, 2013; Sundet, Barlaug, & Torjussen, 2004;
Teasdale & Owen, 2005).
An upward shift of the lower portion of the IQ distribu-
tion may be plausibly explained by improved nutrition
(Colom et al., 2005; Nisbett et al., 2012). Moreover,
improvements of lower-quality environments or educa-
tional reforms that have been specifically targeted to disad-
vantaged groups may be additional drivers of decreasing
variability (Rindermann & Thompson, 2013). In this regard,
reduced IQ variability in itself may be seen as a conse-
quence of other mechanisms at work rather than an inde-
pendent, genuine cause. Although we did not directly test
for decreasing IQ variability, findings suggest that at least
nutrition and education may play an important role in this
context and thus, consistent with prior research, may well
have led to decreases in IQ variability.
Social multipliers. In his seminal initial publication
about the now eponymous effect, Flynn (1984) ques-
tioned the meaningfulness of the observed IQ gains,
arguing that they were too large to reflect genuine intel-
ligence test performance increases and suggested that
they conceivably might be due to sampling artifacts or
test sophistication. He revised and attenuated this argu-
ment in subsequent articles, arguing that IQ increases
could be due to an actual increase in specific facets of
abstract problem-solving ability but not in intelligence as
such (Flynn, 1994).
Eventually, Dickens and Flynn (2001) proposed an
explanation of their own for an IQ rise that would not be
confined only to such a specific ability facet. They intro-
duced the concept of social multipliers as a potential
mechanism that would be powerful enough to explain
test score gains. The basic idea is that even slight envi-
ronmental advantages typically improve individual per-
formance, which once again will lead to a more
advantageous environment, which in turn will lead to
improved performance, and so on.
This model stands out from other explanatory attempts
because it also leaves room to take genetic factors into
account. Slight advantages in specific abilities are typi-
cally (although not always) found in beneficial environ-
ments (i.e., passive, reactive, and active gene–environment
correlations). These beneficial environmental influences
would then act as multipliers for those abilities, thus pro-
ducing larger increases in performance, which in turn
again positively modify the environment. In this way,
even small initial environmental advantages may produce
large effects over a relatively short time (Dickens & Flynn,
2001). In regard to intelligence task performance, such
social multipliers should lead to a general increase of
cognitive abilities of the general population because of a
more intense societal focus on cognitive task perfor-
mance and subsequent individual exposure to cogni-
tively stimulating environments. This means that social
multipliers should be seen as mechanisms that allow
beneficial societal influences to trigger IQ test score
gains, regardless of the nature of these societal influ-
ences. Whether certain abilities improve in this case
would depend on societal emphasis of a specific ability.
Several commentators have doubted that such envi-
ronmental interactions and feedback loops could be suit-
able to account for IQ gains because of the seemingly
contradictory evidence (Rushton & Jensen, 2005);
because of competing alternative explanations, such as
hybrid vigor (Mingroni, 2007); or because of changes in
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 295
the IQ population variance (Rowe & Rodgers, 2002).
Others have questioned the potential of the model to
explain the strength of the IQ gains, suggesting that the
environmental interaction was not powerful enough to
produce large IQ gains over such a comparatively short
time span (Loehlin, 2002). These criticisms have been
given careful attention by their proponents, and Dickens
and Flynn responded to the raised concerns on a con-
ceptual level (Dickens, 2009; Dickens & Flynn, 2002),
although as of yet there has been no direct test of this
theory (Nisbett etal., 2012).
Most observations from our present analysis fit well to
the social multiplier theory. Specifically, IQ gains inde-
pendent of psychometric g as well as the strongest gains
for fluid IQ are consistent with expected outcomes due
to the effect of social multipliers. Similar to previous
empirical investigations, the present analysis cannot pro-
vide a direct test of social multiplier effects. Accordingly,
social multipliers are likely to explain portions, but not
all, of the observed gains, and evidence from a direct test
of this theory is still needed.
Life history speed. Similar to IQ variability, no novel
cause for IQ gains in itself is postulated in the life history
speed theory, which is more of an attempt to integrate
several proposed factors driving the Flynn effect within a
larger theoretical framework. Specifically, several poten-
tial causes for IQ changes have been linked to life history
speed (Woodley, 2012a). Slow life history individuals are
typically characterized to have fewer lifetime sexual part-
ners, fewer offspring, and later parenthood, as compared
with fast life history individuals. Different environmental
conditions may favor either slower or faster life history
speed.
It has been argued that in environments with high
pathogen stress and adverse conditions, such as insuffi-
cient nutrition, fast life histories are advantageous for a
population’s survival because they facilitate coping with
unpredictable environments. Decreasing life history
speed may thus be seen as a consequence of reduced
perceived mortality threat and would in turn allow for
more energetic investment into cognitive ability matura-
tion and differentiation. This mechanism has recently
been theoretically linked to epigenetic factors, speculat-
ing that cognitive ability increases may be driven by
genome optimization due to decreasing life history speed
(Greiffenstein, 2011).
In other words, when pathogen stress is reduced and
adequate nutrition is ensured, the development of a
slower life history speed is encouraged, thus allowing the
emergence of differentiated cognitive abilities. Ultimately,
these developments of specific abilities may then in turn
be facilitated by improved educational quality and num-
ber of educational years (Woodley, 2012a; Woodley,
Figueredo, Brown, & Ross, 2013; Woodley & Madison,
2013).
A combination of better education, reduced family
size, better nutrition, and lower pathogen stress could
explain IQ test score gains. Accordingly, gains would be
expected to occur in (developed) countries showing
slowing life history speed, whereas no change in IQ test
scores or even decreases should occur in countries that
show accelerating life history speed (Woodley, 2012a).
This theory is suitable to explain the differential gains
on different IQ domains in the present study. Stronger
gains in fluid IQ than in crystallized IQ thus may be
expressions of more individual investments in the devel-
opment of specific abilities in environments that favor
slower life histories (i.e., low-pathogen stress and high-
nutrition environments).
Slower life history has been observed to be associated
with a decline of the strength of g over time and pro-
motes ability differentiation (Woodley, 2012a; Woodley &
Madison, 2013). Consistent with the life history model,
the Flynn effect in the present meta-analysis is apparently
not on g. Indeed, the observed effects of test type suggest
a negative association between IQ gains and psychomet-
ric g.
Predictions of the life history model appear to fit well
to the observed patterns of IQ gains in the present meta-
analysis. As this model does not warrant uniformity of
changes across countries or strength of changes across
time, life history speed would be suitable to explain the
erratic pattern of IQ changes in our data. Different causes
associated with life history speed could thus be either
present or absent in single countries, but they still would
yield overall IQ gains due to compensatory effects of
other factors being present. In other words, not all related
causes need to be present in order to decelerate life his-
tory speed and consequently lead to gains; rather, causes
may be effective one at a time.
A recent first empirical investigation of the life history
model provided evidence for the suitability of this model
to explain the Flynn effect (Woodley et al., 2013).
Furthermore, this model has been shown to be consistent
with reported population increases (Figueredo, de Baca,
& Woodley, 2013) in the personality dimensions of con-
scientiousness, emotional stability, agreeableness (Smits,
Dolan, Vorst, & Wicherts, 2011), and extraversion
(Twenge, 2001). Nonetheless, a comprehensive assess-
ment of life history speed on the country level in combi-
nation with the available IQ trend data seems necessary
to determine the explanatory capabilities of the life his-
tory model for the Flynn effect.
Other causes. A number of further proposed causes
that have gained less attention in the literature include a
general increase in environmental complexity (Schooler,
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
296 Pietschnig, Voracek
1998) and allusions to effects of the collective uncon-
scious (Mahlberg, 1997). Whereas conclusive statements
about the former theory are difficult to make, because a
clear definition of environmental complexity is not avail-
able, the latter assertion should be dismissed as evidently
empirically intractable and thus bearing similarities to
fringe science.
Limitations of the Evidence
Although the present meta-analysis provides a compre-
hensive account of the available data for the Flynn effect,
some considerations when interpreting the current results
should be pointed out. First, the differently designed pri-
mary studies present a potential source of bias for esti-
mating the effect. In two-wave assessments (i.e., tests of
two independent samples on two different time points;
see Appendix), item content might have been more dif-
ficult to comprehend for the latter cohort, particularly so
for crystallized IQ tasks, as linguistic expressions became
outdated. Conversely, in cross-sectional designs, a substi-
tution of outdated items from original tests through newly
constructed test items might have rendered the tests not
perfectly equivalent. In order to deal with these issues,
we took great care during study selection to include only
such studies where potentially biasing influences of com-
prehensibility and content nonequivalence were deemed
to be minimal (see Appendix). However, it is also impor-
tant to note that any such remaining confounders would
actually have made it more difficult to detect IQ gains,
which may be taken as an indicator of the robustness of
the observed Flynn effect.
Second, all gains in the primary studies included in the
present analysis were assumed to be linear. It cannot be
ruled out that the gains, as calculated from the primary
studies, may in fact have followed individual nonlinear
trajectories, thus resulting in a somewhat coarser assess-
ment by the linear models applied here. However, the
assumption of linearity was necessary due to the nature
of the available primary data. Moreover, by means of lin-
ear segmented line regressions on the weighted annual
gains, an analysis of nonlinear trends could be provided
in the present study.
Third, the analyses for full-scale IQ were not exclu-
sively based on studies providing change estimates on
full-scale IQ measures but to a certain amount included
measures of fluid IQ, crystallized IQ, and IQ estimates
from developmental tests. Because we observed substan-
tial differences in the strength of IQ changes between
different IQ domains, change estimates based on fluid
and developmental measures may be expected to show
somewhat stronger gains, whereas crystallized measures
should show somewhat weaker gains than estimates for
full-scale IQ measures. However, we felt it was important
to provide a comprehensive trajectory of intelligence test
score changes over time, and the number of non-full-
scale IQ studies was comparatively small (see Statistical
Analysis in the Appendix).
Of note, we used fixed-effect models for all analyses
rather than random-effects analyses. The latter necessi-
tate estimates of the dispersion of effect sizes. However,
such dispersion measures were unavailable for the major-
ity of data points included in the meta-analysis; for this
reason, application of random-effects calculations was
waived. Moreover, fixed-effect modeling within modera-
tor analyses was deemed appropriate, because this
approach is based on fewer assumptions.
Finally, the accumulated evidence shows a clear lack
of empirical evidence about the Flynn effect in older
individuals. This is due to a vast majority of samples
(90%) having a mean age of 38 years or below. In light of
increasing average population ages, particularly in devel-
oped countries (e.g., Staff, Hogan, & Whalley, 2014),
future Flynn effect research should focus on investigating
older participants.
Conclusion
The present research contributes to the literature in two
ways. First, it provides a comprehensive account of intel-
ligence test norm changes since the introduction of psy-
chometric intelligence tests early in the twentieth century.
Second, by assessing trends of intelligence test perfor-
mance of general population samples worldwide over
more than one century, we were able to evaluate the
plausibility and viability of the different theories pro-
posed to explain these changes. The meta-analytical evi-
dence may be informative for narrowing down the
number of theories already proposed as well as the num-
ber of candidate factors corresponding to these theories
that could account for the Flynn effect.
In summary, the present study clearly demonstrates a
Flynn effect of about 3 IQ points per decade. However,
this estimate reflects global linear gains by assuming uni-
form gains over a period of more than 100 years. The
data suggest that this assumption may well not be justi-
fied, as the strength of gains could be shown to vary
according to country, intelligence domains, and the
investigated time span.
Evaluating mechanisms
Below we evaluate potential contributions of the pro-
posed mechanisms to the Flynn effect (for an overview,
see Table 2). Although the Flynn effect can be demon-
strated across all IQ domains, the gains appear to be
stronger for fluid than for crystallized (or spatial and full-
scale) IQ.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 297
Stronger gains for fluid IQ may be due to social multi-
pliers, because modern societal demands that constitute
the basis for these multipliers may conceivably rely more
heavily on abilities related to fluid intelligence (Dickens
& Flynn, 2001). Increased guessing behavior may most
likely play an additional role in explaining differences in
these gains. Because multiple-choice response formats
are more common for measures assessing fluid than crys-
tallized IQ (e.g., Brand, 1989), stronger guessing effects
on fluid rather than crystallized IQ measures are expected.
Another mechanism that seems suited to explain the
stronger fluid than crystallized IQ gains is the develop-
ment and application of more sophisticated sets of cogni-
tive rules for IQ tasks. Because most measures of fluid IQ
can be solved by applying a rather limited set of basic
mental operations, improvements in cognitive abilities
that aid individuals to develop and successfully apply
such rules (i.e., working memory and implicit learning)
would lead to substantial gains in measures of fluid but
not crystallized IQ (Armstrong & Woodley, 2014). Scores
on crystallized IQ tests should be largely independent
from these effects, because crystallized measures most
commonly do not rely on rule application but rather
assess knowledge. Over time, more frequent or contin-
ued exposure to tasks and tests in modern-day environ-
ments warranting the application of such rules may make
individuals more sensitive to detect and recognize these
rules. Moreover, improvements in specific components of
cognitive abilities should be independent from psycho-
metric g. All of this conforms to the present findings.
Based on our evidence, it seems highly likely that
domain-specific gains are driven by different causes.
Whereas crystallized IQ gains appear to be related mainly
to GDP growth, fluid IQ was observed to be associated
with stronger gains in adults than in children. Although it
may seem surprising that age was strongly related to
gains in fluid but not crystallized IQ, the lack of an asso-
ciation with the latter domain might conceivably be due
to masking effects of GDP growth.
Associations of crystallized IQ gains with GDP growth
support educational, but also nutritional, factors and
decreases in pathogen stress as plausible causes for gains.
Conceivably, such mechanisms may lead to decreasing
IQ variability, which previously has been related to IQ
gains (Colom etal., 2005; Pietschnig etal., 2013; Rodgers,
1999; Rowe & Rodgers, 2002). Thus, decreasing IQ vari-
ability may be viewed as a consequence of such environ-
mental factors at work, although this was not directly
evaluated in the present meta-analysis. Decreases of
gains during times of massive environmental stress (most
notably, World War II) and stronger gains among adults
than among children or adolescents corroborate the
hypotheses that nutritional factors, as proposed by Lynn
(1989, 2009b), and educational factors, as proposed by
Schooler (1998), may play a crucial role in explaining
generational IQ test score changes. Moreover, reduced
pathogen stress may be viewed as another important
contributor (Eppig etal., 2010).
Together with family size, the above three factors have
been integrated into the life history model (Woodley,
2012a), which is consistent with stronger fluid IQ gains as
well as the nonlinearity of these gains. However, we did
not find fertility to be a negative predictor of IQ gains,
nor did we find the expected pattern of decreasing IQ
gains after decades with large birth cohorts (e.g., follow-
ing the years of the baby boom). Consequently, only
three components of decreasing life history speed,
namely, better education, improved nutrition, and
reduced pathogen stress, but not decreasing family size,
appear to be related to the IQ gains. Nonetheless,
decreasing life history speed remains a plausible con-
tributor to the Flynn effect, particularly when bearing in
mind that components within the life history model work
in a compensatory manner.
In more recent years, IQ gains could be observed to
decrease across all intelligence test domains, indicating
that the gains may be coming to an end, as has already
been documented for Scandinavian countries (Dutton &
Lynn, 2013; Teasdale & Owen, 2005). Recent decreases in
IQ gains may be due to a number of different reasons.
On the one hand, decreasing gains may simply be expres-
sions of ceasing effects of beneficial factors, such as satu-
ration or diminishing returns of improved nutrition and
reduced pathogen stress. Similarly, improvements in spe-
cialized cognitive abilities might have reached a ceiling
where further ability differentiation and gains may not be
possible, due to limits of a slowing down of life history
speed (Woodley, 2012a). Likewise, effects of other likely
contributing causes to IQ gains, such as test-taking
behavior or reduced blood lead levels, may be expected
as well to reach a point of saturation. Consequently,
potential further IQ increases due to such factors may be
limited as improvements may eventually yield diminish-
ing returns or cease altogether.
On the other hand, the deceleration of gains may be
due to a picking up of effects that cause IQ decreases
and may ultimately reverse the Flynn effect. Such poten-
tially detrimental effects have been proposed to work
through dysgenic trends in modern populations (Lynn,
2011) or via negative cultural amplifiers (i.e., selective
reproduction patterns due to differences in the use of
contraception in more or less intelligent population seg-
ments; Meisenberg, 2003). Finally, decreases and an ulti-
mate reversal of IQ gains may result from the possibility
that increases in specific cognitive abilities may not be
able to further compensate for decreases in psychomet-
ric g. This notion has received support by a previous
account showing that declines in psychometric g are
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
298 Pietschnig, Voracek
associated with IQ decreases (Woodley & Meisenberg,
2013), and it is consistent with our findings of negative
associations between IQ gains and g-ness of tests.
Although decreasing IQ gains may not be an immediate
cause for alarm, stagnation or an ultimate reversal of IQ
gains might have substantial real-world implications. As
economic prosperity, technological advancement, and
scientific innovation rates have been shown to be posi-
tively related to average national IQs (Lynn & Vanhanen,
2002; Rindermann, 2008; Woodley, 2012b), future IQ
declines may be potentially associated with similar
declines in those areas.
However, the extent to which the above mechanisms
contribute to the strength of gains remains to be further
elucidated. Accordingly, it may be advisable for future
research devoted to this important topic to focus on a
more fine-grained and rigorous assessment of the plausi-
bility and explanatory power of the specific theories pro-
posed for the Flynn effect.
Practical implications
It needs to be acknowledged that IQ should not be
understood as synonymous with intelligence. There is
arguably more to intelligent behavior than IQ scores on a
specific test instrument reflect, and this important differ-
ence needs to be kept in mind when interpreting the
present results. Conceptualizations of intelligence differ,
and so do individual test results on different test instru-
ments. Nonetheless, typically we would expect that indi-
viduals scoring high on psychometric IQ tests also
possess higher mental capacity than low scorers.
However, it would be difficult to argue that the present
IQ increase of about 30 IQ points over the past century
means that the average person born in the early 1900s
had in fact an adjusted IQ of 70 and was therefore accord-
ing to our present classifications learning disabled. It
appears to be equally unlikely that the average person at
present boasts an IQ of 130, which would put about half
of the present population in the gifted range. Although
increases due to guessing behavior are expressions of
either personality traits or increased test sophistication
and therefore can be safely dismissed from constituting
real-world IQ gains, the question remains whether the
portion of IQ gains that cannot be explained by guessing
in fact reflects meaningful gains. Most other potential
explanations that have been put forward are mainly envi-
ronmental in nature, such as factors associated with life
history speed or social multipliers. If the remaining IQ
gains turn out to be driven mainly by such environmental
causes, as it appears to be the case judging from the pres-
ent research, these gains do not reflect changes of the
average mental capacity but rather are expressions of a
facilitation of abilities by the (modern-day) environment.
In other words, adjustments of IQ scores obtained a cen-
tury apart will most certainly not reflect mere differences
in cognitive functioning of two individuals. A person with
an IQ of 100 in the early 1900s may have had quite dif-
ferent capabilities than a person with an “equivalent” IQ
of 70 at the present day (cf. Schooler, 1998).
Final words
In conclusion, in this first formal meta-analysis of the
Flynn effect, we could clearly demonstrate that genera-
tional IQ test score gains have taken place globally over
the past century across all major intelligence domains.
There may be several contributing factors to this ubiqui-
tous but apparently decelerating effect. The totality of
retrievable empirical evidence on this phenomenon, as
quantitatively summarized here, points toward compo-
nents of life history speed, such as improvements of edu-
cation and nutritional factors as well as a reduction of
pathogen-related factors, as the prime candidate causes
of the Flynn effect, whereas differences in the strength of
gains between intelligence domains may be accounted
for by social multipliers and economic prosperity. Future
research will show whether the now observed global
decrease of IQ test score gains will ultimately lead to an
end of these gains or even to a reversal.
Appendix
In this Appendix, we provide details of how we selected
studies, the design of the studies, and how we analyzed
the studies included in our meta-analysis.
Selection
Literature search. Five literature-search strategies
were used to identify relevant studies. First, we per-
formed a cited-reference search in ISI Web of Knowl-
edge for the articles of Flynn (1984), Flynn (1987), and
Schaie and Strother (1968). This strategy was used
because Flynn (1984) and Flynn (1987) are generally
regarded to be the first to systematically describe gen-
erational IQ test score gains and therefore are highly
likely to be cited in subsequent studies investigating
such changes, whereas Schaie and Strother (1968) is
important because it was the first to interpret observed
IQ changes as genuine cohort effects. Second, we
searched the scientific databases PubMed, ISI Web of
Knowledge, and Scopus by using the following search
strings: “Flynn AND effect”; “intelligence AND genera-
tional”; and “IQ AND generational.” Third, we searched
Google Scholar using the string “Flynn AND effect.”
Fourth, we screened reference lists of all relevant arti-
cles obtained through the first three steps of
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 299
our literature search for additional potentially relevant
studies. Finally, we hand searched test manuals of the
most frequently used renormed cognitive measures of
Anglo-American and German origin for reports of vali-
dation studies pertaining to the original and renormed
versions of the same cognitive test instrument. Of note,
it was expected that in this last step of the literature
search only a portion of such validation studies would
be identifiable, because test manuals of cognitive mea-
sures frequently are only available in specific language
areas and may not be obtainable for a variety of reasons
(e.g., discarding and/or stopping production of outdated
manuals, noneligibility of psychometric test manuals for
interlibrary loan). The literature search included all
potentially relevant records up to March 2014.
Subsequently, we screened abstracts of articles from
the database search and titles of records from Google
Scholar for relevance. Following this step, we assessed
the full texts of 1,187 articles, book chapters, theses, and
test manuals (for a flowchart of the literature assessment
procedure, see Figure A1; the comprehensive list of all
included and excluded studies is in Supplementary File
S1). Then, full texts of potentially relevant studies were
coded twice into categories (age group, country, intelli-
gence domain, sample type, study design, and used test)
by the same researcher to ensure reliability of coding,
and the statistical parameters of the investigations (i.e.,
change of test performance on IQ measures, sample size)
were recorded. Of note, categorization of the measures
according to intelligence domains (i.e., full-scale IQ, fluid
IQ, crystallized IQ, and spatial IQ) typically followed the
test descriptions given in the test manuals. If such a
description was unavailable (mostly for tests developed
before the introduction of the concepts of fluid and crys-
tallized IQ; cf. Cattell, 1941), the researcher categorized
test instruments by judging the test content according to
the most closely related intelligence domain (see Table
S1 for these categorizations). In cases of inconsistencies,
Records identified through
database searching
(k = 4,575)
ScreeningIncluded Eligibility Identification
Records after removal of duplicates and
irrelevant results of Google Scholar according
to title screening
(k = 1,188)
Remaining records
screened
(k = 1,188)
Records excluded
(k = 65)
Full-text articles assessed
for eligibility
(k = 1,123)
Full-text articles excluded
because of
Dependent data (k = 112)
Insufficient reporting of
data (k = 24)
Irrelevant design (k = 380)
Comments; Reviews;
Studies reporting no or
insufficient data (k = 387)
Non determinable
outcome measure (k = 1)
Studies included in
quantitative synthesis
(meta-analysis)
(k = 219)
Fig. A1. Flowchart of the literature assessment procedure.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
300 Pietschnig, Voracek
a second researcher coded the respective study indepen-
dently, and discrepancies were resolved by discussion.
Inclusion criteria. To be eligible for inclusion, studies
had to meet four criteria. First, performance was mea-
sured using standardized psychometric test instruments
and developmental tests (i.e., no scholastic achievement
tests, as Flynn effects on such measures would be con-
ceivably masked due to recent increases of individuals
participating in such assessments in the course of col-
lege entrance examinations and the resulting changed
demographic characteristics of such samples). Second,
no correction for assumed gains had to be applied to
arrive at an estimate for the generational IQ test score
changes in the primary studies (for an example, see
Lynn & Hampson, 1986, pp. 30–31). Third, sufficient data
to calculate annual IQ test score changes (i.e., descrip-
tive statistics or change in performance and sample size)
had to be reported. Finally, reported results had to be
independent of results reported in other included stud-
ies. In case of data dependencies, the data sets reporting
the longest time span and the largest sample sizes were
preferred. Our final sample included 219 studies that
met the inclusion criteria, yielding 271 independent sam-
ples comprising 3,987,892 participants covering a time
span of 105 years (1909–2013).
Study designs
Relevant primary studies investigating the Flynn effect
used several distinctly different approaches to assess
generational IQ test score changes, which basically can
be classified into the following five groups: first, assess-
ment of test performance using the same test battery on
two time points using two independent samples display-
ing similar demographic characteristics (two-wave assess-
ment). Thus, differences between the test scores of the
two samples may be interpreted as changes in test per-
formance over the investigated period. Great care was
taken to include only such studies in the analyses where
sample ages were comparable. This was deemed impor-
tant to ensure that differences in test performance
between samples were not confounded with age effects.
Frequently, such investigations were carried out using
samples of military draftees, as this provides in many
cases access to the data of almost entire birth cohorts of
young men in countries where military service is manda-
tory, consequently reducing threats of selection bias in
these samples (e.g., Girod & Allaume, 1976). When test
scores of more than one follow-up sample on the same
test were provided, scores of the most recent of the
reported samples were used to calculate the differences
in order to obtain changes over the longest available time
span in the study (e.g., Flynn, 1998).
Second, in cross-sectional studies, an original and a
revised (restandardized) version of the same test are
administered to the same respondent pool. The difference
between the score on the revised test and the original test
reflects changes in test performance between the time
points of the original and the restandardized test. Typically,
studies using such designs were performed in order to
either explicitly assess the Flynn effect (e.g., Pietschnig,
Voracek, & Formann, 2011) or to examine the validity of
a restandardized measure (e.g., Wechsler, 1981).
In most of these cases, the respective test forms were
administered in a counterbalanced design, thus control-
ling for retest effects. Of note, in a few studies, such
counterbalancing was not performed; rather, for various
reasons, the original test was always administered first.
Nonetheless, inclusion of such studies was deemed
unproblematic, because time intervals between the
administration of test batteries in this subset of studies
typically were large (weighted mean time interval
between two test administrations = 2.9 years), conse-
quently minimizing retest effects.
Another important aspect of administering revised
measures is the change of the test items themselves.
Items that are judged by test authors to be outdated may
be removed, and new items may be introduced in their
place. Such changes of test content range from substitut-
ing single items of a scale up to a subscale’s completely
new construction. Therefore, great care was taken to
include only such studies where original and revised
tests were deemed comparable (i.e., equivalence of test
content was satisfactory).
Third, we used cross-temporal meta-regressions to
estimate IQ changes per year for studies reporting mean
IQ test performance and corresponding sample sizes for
a number of years separately. In cross-temporal regres-
sions, year of data collection is entered as predictor of a
dependent variable (presently, mean IQ of the sample) in
a linear regression model, and data points are weighted
by sample size to account for study precision (i.e., giving
larger weights to studies with larger sample sizes). This
approach allows the interpretation of the slope of these
meta-regression models as the average IQ change per
year over the investigated period (e.g., Pietschnig etal.,
2010). Weighted meta-regression was used in two studies
(Pilliner, Sutherland, & Taylor, 1960; Teasdale & Owen,
2005), unweighted meta-regression in another three
instances (Dutton & Lynn, 2013; Macnamara, 1964;
Schubert & Berlach, 1982). This approach allowed for the
assessment of changes based on a larger number of indi-
viduals than in two-wave assessments and accordingly
more precise estimations.
Fourth, a number of studies compared test results of
samples with characteristics similar to the standardiza-
tion sample of a certain measure with the performance
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 301
of this standardization sample (e.g., Liu & Lynn, 2013).
Consequently, the difference is interpretable as the
change in performance between the test standardization
and the data collection for the respective study.
Fifth, in two studies, the results of a sample of young
participants were compared with the results of a sample
of older participants on the same measure administered
in the same year (Finkel, Reynolds, McArdle, & Pedersen,
2007; Meisenberg, Lawless, Lambert, & Newton, 2005).
Differences between the two cohorts may be interpreted
as the mean change in performance between the two
cohorts, using the mean birth year of the cohorts as onset
and offset points of changes.
Statistical analysis
An approach broadly similar to cross-temporal meta-
analysis (e.g., Pietschnig et al., 2010; Twenge, 2000;
Twenge & Zhang, 2004; see below for a detailed descrip-
tion) and a standard meta-analytical approach (i.e.,
weighted multiple meta-regression) were used. Similar to
the use of standard multiple regressions in primary stud-
ies, meta-regressions allow the examination of influences
of one or more predictor variables on a dependent vari-
able (here, IQ change per year; see below). Different
from standard regressions, in meta-regressions predictors
and dependent variable scores reflect study-level rather
than subject-level observations, and predictors are
weighted according to the precision of the data points.
The use of these two approaches ensured that, on the
one hand, the overall time trend of the generational IQ
test score changes (i.e., by meta-regression) and, on the
other hand, variables moderating the strength of changes
(i.e., by cross-temporal analysis) could be assessed.
Wherever necessary, for both approaches the initially
observed changes in the primary studies were trans-
formed into the IQ metric from other formats reported
(effect sizes, raw scores, or standard deviation units). To
arrive at the unit of analysis, namely, annual IQ test score
changes, absolute changes were divided by the years of
the respective investigated time span per study.
In a number of instances, results of more than one
intelligence test domain were reported (i.e., full-scale,
crystallized, fluid, or spatial IQ test performance). For the
main analyses, full-scale IQ was used as the dependent
variable. In studies where no results for full-scale IQ
were given, changes of other domains were used in the
main analysis in the following order: fluid IQ, crystallized
IQ, and IQ estimates from developmental tests (19.6%,
4.8%, and 13.3% of a total of 271 independent samples,
respectively). In order to provide a comprehensive analy-
sis of all domains, we performed additional separate cal-
culations using crystallized IQ, fluid IQ, and spatial IQ as
the dependent variables.
These domain-specific analyses should make it possi-
ble to provide a more detailed picture of changes. Thus,
the trajectory for full-scale IQ over time should reflect
performance changes on general cognitive ability tasks.
Changes in fluid IQ should reflect changes on tasks asso-
ciated most closely with on-the-spot reasoning ability,
crystallized IQ on tasks requiring knowledge, and spatial
IQ on tasks requiring spatial–temporal abilities.
To assess the overall trend of changes over time, we
calculated annual changes as the mean change per year,
weighted according to the sample size of the respective
primary studies. By using the year preceding the first
available data point of the IQ test score change as a refer-
ence point (i.e., by assuming zero as a reference) and
cumulatively adding annual changes, we could obtain
the overall trend of intelligence changes. This simple lin-
ear transformation was chosen in order to make differ-
ences in strength of gains more easily visible and
intuitively interpretable. The results of inferential statisti-
cal tests are unaffected by this decision.
Subsequently, linear segmented line regression models
(joinpoint regression; see Hudson, 1966; Kim, Fay, Feuer,
& Midthune, 2000; Kim, Yu, & Feuer, 2008), using the grid
search method (Lerman, 1980), were applied on these
data to identify changes in the regression slopes over the
examined period. This method allows modeling of tem-
poral trends by testing significant changes in regression
slopes (i.e., significant increases or decreases) when a
certain number of joinpoints is assumed. This means that
each assumed joinpoint identifies the point on the tempo-
ral axis where a significant change in the regression slope
occurs, thus identifying the point where two ordinary
least squares linear regression segments connect.
To arrive at the best-fitting model, we fit regression
models to the data, starting with the most parsimonious
model (i.e., in the present case assuming linearity, zero
joinpoints). Subsequently, more complex models with
increasing numbers of joinpoints were fitted and com-
pared with the respective simpler models by ratio tests of
permutations of the squared errors of the null and the
alternative model (for details, see Kim etal., 2000).
Models with up to four joinpoints were fitted by means
of Bonferroni-corrected permutation tests using the
Joinpoint Regression Program 4.0.4 (Statistical Research
and Applications Branch, 2011). However, when signifi-
cant numerical changes between two adjacent slopes
were smaller than 0.05 (i.e., absolute difference of half an
IQ point over 10 years), a more parsimonious model was
selected, as practical considerations would suggest that
such small changes in strength would not justify the
selection of more complex models, but rather would lead
to model overfit. For spatial IQ, no segmented line regres-
sions were performed because of the small number of
available observations.
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
302 Pietschnig, Voracek
In our second data-analysis approach, we used a stan-
dard meta-analytical method by combining annual
changes weighted by sample size to yield the overall
observed effect. In a more fine-grained analysis, study
characteristics were entered into multiple meta-regression
models, weighted by sample size to reflect study preci-
sion, in order to identify moderator variables. In addition,
country-specific average changes in gross domestic prod-
uct (GDP) per capita over the respective time spans cov-
ered by individual studies were calculated from the
Maddison Project database (Bolt & van Zanden, 2013) and
included as a predictor.
When interpreting the results of this approach, it is
important to keep in mind that annual IQ test score
changes, as calculated for these analyses, depend solely
on observed changes and on study weights (i.e., the sam-
ple size but not the investigated time span) and therefore
do not reflect linear IQ test score changes over the respec-
tive time span. Rather, these weighted mean changes make
it possible to assess influences of moderator variables by
serving as the dependent variables in regression analyses,
whereas the results of cross-temporal analyses reflect
yearly IQ test score changes. Mean weighted annual
changes turned out to be all positive and are provided in
Table S6. Upon inspection of these gains, it is striking that
most gains were substantially larger when calculated in
this fashion than in the cross-temporal approach. This
observation supports the results from the segmented line
regression models, indicating varying IQ gains for different
time spans (i.e., nonlinearity of IQ gains).
Initially, seven predictors (children vs. adult sample,
GDP change per capita, investigated time span in years,
patient-based vs. healthy sample, proportion of males in
the sample, test g-ness, year of onset of IQ test score
change) were regressed on mean annual IQ test score
changes. As information about participants’ sex frequently
was unavailable in the primary data sources, inclusion of
this variable would have led to a considerable reduction
of samples in the regressions due to missing data (about
a third of otherwise includable samples). Because, consis-
tent with prior reports (Pietschnig et al., 2011), sex of
participants did not emerge as a significant predictor in a
preliminary calculation for full-scale IQ using all seven
predictors, we decided to drop this variable (i.e., propor-
tion of males in the sample) from further analyses.
Inclusion of interaction terms of predictors caused vari-
ance inflation factors (VIFs) to deteriorate drastically (all
VIFs < 4.0 for main effects, but VIFs > 60 when interac-
tions were specified); hence, these interaction terms were
omitted from analysis.
In secondary analyses, subsets of our present data cov-
ering the period from 1960 to 2013 only (i.e., ks = 137
samples for full-scale, 74 for crystallized, 88 for fluid, and
10 for spatial IQ, respectively) were examined to
investigate influences of average fertility rates (i.e., average
fertility within the respective country over the examined
time span) as a proxy for family size. Average fertility rates
per country were obtained from the World Bank databases
for study-specific time spans (World Bank, 2014) and
entered as an additional predictor. Calculations were per-
formed for different intelligence domains separately (full-
scale IQ, crystallized IQ, fluid IQ, and spatial IQ), using
the statistical software R 3.1.1 (R Development Core Team,
2014). We followed Cohen’s classification of effect sizes
into small, medium, and large effects to describe observed
effects (Cohen, 1988).
Acknowledgments
The authors would like to thank James Flynn, Earl Hunt, and
Michael A. Woodley for their helpful comments on an earlier
version of this article.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Supplemental Material
Additional supporting information may be found at http://pps
.sagepub.com/content/by/supplemental-data
Note
1. Joinpoint regression (segmented line regression) is a form of
nonlinear regression that allows the estimation of and compari-
son between changes in the strength of regression slopes as
numerical values of the predictor increase. For a more detailed
description, see the Statistical Analysis section in the Appendix.
References
Armstrong, E. L., & Woodley, M. A. (2014). The rule-depen-
dence model explains the commonalities between the Flynn
effect and IQ gains via retesting. Learning and Individual
Differences, 29, 41–49. doi:10.1016/j.lindif.2013.10.009
Binet, A., & Simon, T. (1905). Méthodes nouvelles pour le
diagnostic du niveau intellectuel des anormaux [New
methods for the diagnosis of intellectual level of subnor-
mals]. L’Année Psychologique, 11, 191–244. doi:10.3406/
psy.1904.3675
Binet, A., & Simon, T. (1908). Le developpement de l’intelligence
chez les enfants [The development of intelligence in chil-
dren]. L’Année Psychologique, 14, 1–90. doi:10.3406/
psy.1907.3737
Bolt, J., & van Zanden, J. L. (2013). The first update of the
Maddison Project: Re-estimating growth before 1820
(Maddison Project Working Paper 4). Groningen, The
Netherlands: University of Groningen.
Boot, W. R., Kramer, A. F., Simons, D. J., Fabiani, M., & Gratton,
G. (2008). The effects of video game playing on attention,
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 303
memory, and executive control. Acta Psychologica, 129,
387–398. doi:10.1016/j.actpsy.2008.09.005
Brand, C. (1987a). British IQ: Keeping up with the times.
Nature, 328, 761. doi:10.1038/328761a0
Brand, C. (1987b). Intelligence testing: Bryter still and bryter?
Nature, 328, 110. doi:10.1038/328110a0
Brand, C. R. (1989). Has there been a “massive” rise of IQ levels
in the West? Evidence from Scottish children. Irish Journal
of Psychology, 10, 383–393.
Brand, C. R. (1990). A ‘gross’ underestimate of a ‘massive’ IQ
rise? A rejoinder to Flynn. Irish Journal of Psychology, 11,
52–56. doi:10.1080/03033910.1990.1010557788
Brand, C. R. (1996). The g factor: General intelligence and its
implications. New York, NY: Wiley.
Brand, C. R., Freshwater, S., & Dockrell, N. (1989). Has there
been a ‘massive’ rise in IQ levels in the West? Evidence
from Scottish children. Irish Journal of Psychology, 10, 388–
393. doi:10.1080/03033910.1989.10557756
Brooks, S. S. (1922). Some uses for intelligence tests. Journal of
Educational Research, 5, 217–238.
Brouwers, S. A., Van de Vijver, F. J. R., & Van Hemert, D.A.
(2009). Variation in Raven’s Progressive Matrices scores
across time and place. Learning and Individual Differences,
19, 330–338. doi:10.1016/j.lindif.2008.10.006
Campbell, S. K., Siegel, E., Parr, C. A., & Ramey, C. T.
(1986). Evidence for the need to renorm the Bayley
Scales of Infant Development based on the performance
of a population-based sample of 12-month-old infants.
Topics in Early Childhood Special Education, 6, 83–96.
doi:10.1177/027112148600600208
Cattell, R. B. (1937). The fight for our national intelligence.
London, England: P. S. King & Son.
Cattell, R. B. (1941). Some theoretical issues in adult intelli-
gence testing. Psychological Bulletin, 38, 592.
Ceci, S. J. (1991). How much does schooling influence general
intelligence and its cognitive components? A reassessment
of the evidence. Developmental Psychology, 27, 703–722.
doi:10.1037/0012-1649.27.5.703
Ceci, S. J., & Kanaya, T. (2010). “Apples and oranges are both
round”: Furthering the discussion on the Flynn effect.
Journal of Psychoeducational Assessment, 28, 441–447.
doi:10.1177/0734282910373339
Cohen, J. (1988). Statistical power analysis for the behavioral
sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
Cole, T. J. (2003). The secular trend in human physical growth:
A biological view. Economics & Human Biology, 1, 161–
168. doi:10.1016/S1570-677X(02)00033-3
Colom, R., Flores-Mendoza, C. E., & Abad, F. J. (2007).
Generational changes on the Draw-a-Man test: A compari-
son of Brazilian urban and rural children tested in 1930,
2002 and 2004. Journal of Biosocial Science, 39, 79–89.
doi:10.1017/S0021932005001173
Colom, R., Juan-Espinosa, M., & Garcia, L. F. (2001). The secu-
lar increase in test scores is a “Jensen effect.” Personality
and Individual Differences, 30, 553–559. doi:10.1016/
S0191-8869(00)00054-4
Colom, R., Lluis-Font, J. M., & Andres-Pueyo, A. (2005). The
generational intelligence gains are caused by decreasing
variance in the lower half of the distribution: Supporting
evidence for the nutrition hypothesis. Intelligence, 33, 83–
91. doi:10.1016/j.intell.2004.07.010
Daley, T. C., Whaley, S. E., Sigman, M. D., Espinosa, M. P., &
Neumann, C. (2003). IQ on the rise: The Flynn effect in
rural Kenyan children. Psychological Science, 14, 215–219.
doi:10.1111/1467-9280.02434
Deary, I. J. (Ed.). (2009). Intelligence, health, and death: The
emerging field of cognitive epidemiology [Special issue].
Intelligence, 37.
Dickens, W. T. (2009). A response to recent critics of Dickens
and Flynn (2001). Retrieved from http://www.brook-
ings.edu/~/media/Files/rc/articles/2001/0401IQ_critics_
dickens/0401IQ_critics_dickens.pdf
Dickens, W. T., & Flynn, J. R. (2001). Heritability estimates ver-
sus large environmental effects: The IQ paradox resolved.
Psychological Review, 108, 346–369. doi:10.1037//0033-
295X.108.2.346
Dickens, W. T., & Flynn, J. R. (2002). The IQ paradox is
still resolved: Reply to Loehlin (2002) and Rowe and
Rodgers (2002). Psychological Review, 109, 764–771.
doi:10.1037//0033-295X.109.4.764
Dutton, E., & Lynn, R. (2013). A negative Flynn effect in Finland,
1997-2009. Intelligence, 41, 817–820. doi:10.1016/j.intell
.2013.05.008
Eppig, C., Fincher, C. L., & Thornhill, R. (2010). Parasite preva-
lence and the worldwide distribution of cognitive ability.
Proceedings of the Royal Society B: Biological Sciences, 277,
3801–3808. doi:10.1098/rspb.2010.0973
Figueredo, A. J., de Baca, T. C., & Woodley, M. A. (2013). The
measurement of human life history strategy. Personality
and Individual Differences, 55, 251–255. doi:10.1016/
j.paid.2012.04.033
Finkel, D., Reynolds, C. A., McArdle, J. J., & Pedersen, N.L.
(2007). Cohort differences in trajectories of cognitive aging.
Journals of Gerontology Series B: Psychological Sciences &
Social Sciences, 62, 286–294. doi:10.1093/geronb/62.5.P286
Flynn, J. R. (1984). The mean IQ of Americans: Massive
gains 1932 to 1978. Psychological Bulletin, 95, 29–51.
doi:10.1037//0033-2909.95.1.29
Flynn, J. R. (1987). Massive IQ gains in 14 nations: What IQ
tests really measure. Psychological Bulletin, 101, 171–191.
doi:10.1037//0033-2909.101.2.171
Flynn, J. R. (1990). Massive IQ gains on the Scottish WISC:
Evidence against Brand etal.’s hypothesis. Irish Journal of
Psychology, 11, 41–51. doi:10.1080/03033910.1990.10557787
Flynn, J. R. (1994). IQ gains over time. In R. J. Sternberg (Ed.),
The encyclopedia of human intelligence (pp. 617–623).
New York, NY: Macmillan.
Flynn, J. R. (1998). Israeli military IQ tests: Gender differences
small; IQ gains large. Journal of Biosocial Science, 30, 541–
553. doi:10.1017/S0021932098005410
Flynn, J. R. (1999). Searching for justice: The discovery of
IQ gains over time. American Psychologist, 54, 5–20.
doi:10.1037//0003-066X.54.1.5
Flynn, J. R. (2007). Capital offenders and the death sentence:
A scandal that must be addressed. Psychology in Mental
Retardation and Developmental Disabilities, 32, 3–7.
Flynn, J. R. (2009a). The WAIS–III and WAIS–IV:
Daubert motions favor the certainly false over the
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
304 Pietschnig, Voracek
approximately true. Applied Neuro psychology, 16, 98–104.
doi:10.1080/09084280902864360
Flynn, J. R. (2009b). What is intelligence? Beyond the Flynn
effect (Expanded paperback ed.). Cambridge, England:
Cambridge University Press.
Flynn, J. R. (2010). Problems with IQ gains: The huge vocabu-
lary gap. Journal of Psychoeducational Assessment, 28,
412–433. doi:10.1177/0734282910373342
Flynn, J. R. (2012). Are we getting smarter? Rising IQ in the twenty-
first century. New York, NY: Cambridge University Press.
Fuggle, P. W., Tokar, S., Grant, D. B., & Smith, I. (1992). Rising
IQ scores in British children: Recent evidence. Journal of
Child Psychology and Psychiatry and Allied Disciplines, 33,
1241–1247. doi:10.1111/j.1469-7610.1992.tb00942.x
Girod, M., & Allaume, G. (1976). L’evolution du niveau intel-
lectuel de la population francaise pendent le dernier quart
de siecle [The evolution of the intellectual level of the
French population during the last quarter of the century].
International Review of Applied Psychology, 25, 121–123.
doi:10.1111/j.1464-0597.1976.tb00863.x
Gorey, K. M. (2001). Early childhood education: A meta-analytic
affirmation of the short- and long-term benefits of educa-
tional opportunities. School Psychology Quarterly, 16, 9–30.
doi:10.1521/scpq.16.1.9.19163
Greiffenstein, M. F. (2011). Secular IQ increases by epigen-
esis? The hypothesis of cognitive genotype optimization.
Psychological Reports, 109, 353–366. doi:10.2466/03.04.10.19
.PR0.109.5.353-366
Herrnstein, R. J., & Murray, C. (1996). The bell curve: Intelligence
and class structure in American society. New York, NY:
Simon & Schuster.
Holliday, M. A. (1986). Body composition and energy needs
during growth. In F. Falkner & J. M. Tanner (Eds.), Human
growth: A comprehensive treatise (pp. 101–117). New York,
NY: Plenum.
Howard, R. W. (1999). Preliminary real-world evidence that
average human intelligence really is rising. Intelligence, 27,
235–250. doi:10.1016/S0160-2896(99)00018-5
Howard, R. W. (2001). Searching the real world for signs
of rising population intelligence. Personality and
Individual Differences, 30, 1039–1058. doi:10.1016/S0191-
8869(00)00095-7
Howard, R. W. (2005). Objective evidence of rising popula-
tion ability: A detailed examination of longitudinal chess
data. Personality and Individual Differences, 38, 347–363.
doi:10.1016/j.paid.2004.04.013
Hudson, D. J. (1966). Fitting segmented curves whose join
points have to be estimated. Journal of the American
Statistical Association, 61, 1097–1129. doi:10.1080/01621459
.1966.10482198
Hunt, E. (2011). Human intelligence. Cambridge, England:
Cambridge University Press.
Husén, T., & Tuijnman, A. (1991). The contribution of formal
schooling to the increase in intellectual capital. Educational
Researcher, 20, 17–25. doi:10.3102/0013189X020007017
Jensen, A. R. (1998). The g factor: The science of mental ability.
Westport, CT: Praeger.
Johnson, W., Bouchard, T. J., Jr., Krueger, R. F., McGue, M., &
Gottesman, I. I. (2004). Just one g: Consistent results from
three test batteries. Intelligence, 32, 95–107. doi:10.1016/
S0160-2896(03)00062-X
Kanaya, T., & Ceci, S. J. (2007). Are all IQ scores created equal?
The differential costs of IQ cutoff scores for at-risk children.
Child Development Perspectives, 1, 52–56. doi:10.1111/
j.1750-8606.2007.00010.x
Kant, A. K., & Graubard, B. I. (2006). Secular trends in the associa-
tion of socio-economic position with self-reported dietary attri-
butes and biomarkers in the US population: National Health
and Nutrition Examination Survey (NHANES) 1971–1975 to
NHANES 1999–2002. Public Health Nutrition, 10, 158–167.
Kaufman, A. S., Zhou, X., Reynolds, M. R., Kaufman, N. L.,
Green, G. P., & Weiss, L. G. (2014). The possible societal
impact of the decrease in U.S. blood lead levels on adult
IQ. Environmental Research, 132, 413–420. doi:10.1016/
j.envres.2014.04.015
Khaleefa, O., Abdelwahid, S. B., Abdulradi, F., & Lynn, R.
(2008). The increase of intelligence in Sudan 1964–2006.
Personality and Individual Differences, 45, 412–413.
Kim, H.-J., Fay, M. P., Feuer, E. J., & Midthune, D. N. (2000).
Permutation tests for joinpoint regression with applica-
tions to cancer rates. Statistics in Medicine, 19, 335–351.
doi:10.1002/sim.811
Kim, H.-J., Yu, B., & Feuer, E. J. (2008). Inference in seg-
mented line regression: A simulation study. Journal of
Statistical Computation and Simulation, 78, 1087–1103.
doi:10.1080/00949650701528461
Koller, K., Brown, T., Spurgeon, A., & Levy, L. (2004). Recent
developments in low-level lead exposure and intellectual
impairment in children. Environmental Health Perspectives,
112, 987–994.
Lerman, P. M. (1980). Fitting segmented regression models by grid
search. Applied Statistics, 29, 77–84. doi:10.2307/2346413
Liu, J., & Lynn, R. (2013). An increase of intelligence in
China 1986-2012. Intelligence, 41, 479–481. doi:10.1016/
j.intell.2013.06.017
Loehlin, J. C. (2002). The IQ paradox: Resolved? Still an
open question. Psychological Review, 109, 754–758.
doi:10.1037/0033-295X.109.4.754
Lovei, M. (1998). Phasing out lead from gasoline: Worldwide
experiences and policy implications (World Bank Technical
Paper No. 397). Washington, DC: World Bank.
Lucas, A., Morley, R., & Cole, T. J. (1998). Randomised trial
of early diet in preterm babies and later intelligence quo-
tient. British Medical Journal, 317, 1481–1487. doi:10.1136/
bmj.317.7171.1481
Lynn, R. (1989). Positive correlation between height, head
size and IQ: A nutrition theory of the secular increases in
intelligence. British Journal of Educational Psychology, 59,
372–377. doi:10.1111/j.2044-8279.1989.tb03112.x
Lynn, R. (1990). The role of nutrition in secular increases in
intelligence. Personality and Individual Differences, 11,
273–285. doi:10.1016/0191-8869(90)90241-I
Lynn, R. (2009a). Fluid intelligence but not vocabulary has
increased in Britain, 1979-2008. Intelligence, 37, 249–255.
doi:10.1016/j.intell.2008.09.007
Lynn, R. (2009b). What has caused the Flynn effect? Secular
increases in the development quotient of infants.
Intelligence, 37, 16–24. doi:10.1016/j.intell.2008.07.008
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
Meta-Analysis of the Flynn Effect 305
Lynn, R. (2011). Dysgenics: Genetic deterioration in modern
populations. Westport, CT: Praeger.
Lynn, R. (2013). Who discovered the Flynn effect? A review
of early studies of the secular increase of intelligence.
Intelligence, 41, 765–769. doi:10.1016/j.intell.2013.03.004
Lynn, R., & Hampson, S. (1986). The rise of national intelligence:
Evidence from Britain, Japan and the U.S.A. Personality
and Individual Differences, 7, 23–32. doi:10.1016/0191-
8869(86)90104-2
Lynn, R., & Vanhanen, T. (2002). IQ and the wealth of nations.
Westport, CT: Praeger.
Macnamara, J. (1964). Zero error and practice effects in Moray
House English quotients. British Journal of Educational
Psychology, 34, 315–320.
Mahlberg, A. (1997). The rise in IQ scores. American
Psychologist, 52, 71. doi:10.1037/0003-066X.52.1.71
McArdle, J. J., Ferrer-Caja, E., Hamagami, F., & Woodcock,
R.W. (2002). Comparative longitudinal structural analyses
of the growth and decline of multiple intellectual abilities
over the life span. Developmental Psychology, 38, 115–142.
doi:10.1037/0012-1649.38.1.115
Meisenberg, G. (2003). IQ population genetics: It’s not as sim-
ple as you think. Mankind Quarterly, 44, 185–210.
Meisenberg, G., Lawless, E., Lambert, E., & Newton, A. (2005). The
Flynn effect in the Caribbean: Generational change of cognitive
test performance in Dominica. Mankind Quarterly, 46, 29–69.
Merrill, M. A. (1938). The significance of IQ’s on the revised
Stanford-Binet scales. Journal of Educational Psychology,
29, 641–651. doi:10.1037/h0057523
Mingroni, M. A. (2004). The secular rise in IQ: Giving heterosis
a closer look. Intelligence, 32, 65–83. doi:10.1016/S0160-
2896(03)00058-8
Mingroni, M. A. (2007). Resolving the IQ paradox: Heterosis as
a cause of the Flynn effects and other trends. Psychological
Review, 114, 806–829. doi:10.1037/0033-295X.114.3.806
Must, O., & Must, A. (2013). Changes in test taking pat-
terns over time. Intelligence, 41, 780–790. doi:10.1016/j.
intell.2013.04.005
Must, O., Must, A., & Raudik, V. (2003). The secular rise in IQs: In
Estonia, the Flynn effect is not a Jensen effect. Intelligence,
31, 461–471. doi:10.1016/S0160-2896(03)00013-8
Neisser, U. (1997). Rising scores on intelligence tests. American
Scientist, 85, 440–447.
Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W.,
Brody, N., Ceci, S. J., . . . Urbina, S. (1996). Intelligence:
Knowns and unknowns. American Psychologist, 5, 77–101.
doi:10.1037//0003-066X.51.2.77
Nevin, R. (2000). How lead exposure relates to temporal changes
in IQ, violent crime, and unwed pregnancy. Environmental
Research Section A, 83, 1–22. doi:10.1006/enrs.1999.4045
Nisbett, R. E., Aronson, J., Blair, C., Dickens, W., Flynn, J.,
Halpern, D. F., & Turkheimer, E. (2012). Intelligence:
New findings and theoretical developments. American
Psychologist, 67, 130–159. doi:10.1037/a0026699
Pietschnig, J., Tran, U. S., & Voracek, M. (2013). Item-response
theory modeling of IQ gains (the Flynn effect) on crys-
tallized intelligence: Rodger’s hypothesis yes, Brands’
hypothesis perhaps. Intelligence, 41, 791–801. doi:10.1016/
j.intell.2013.06.005
Pietschnig, J., Voracek, M., & Formann, A. K. (2010).
Pervasiveness of the IQ rise: A cross-temporal meta-analy-
sis. PLoS ONE, 5, e14406. doi:10.1371/journal.pone.0014406
Pietschnig, J., Voracek, M., & Formann, A. K. (2011). Female
Flynn effects: No sex differences in generational IQ gains.
Personality and Individual Differences, 50, 759–762.
doi:10.1016/j.paid.2010.12.019
Pietschnig, J., Zeiler, M., & Voracek, M. (2011, July). The bigger
they are, the higher they score: Meta-analyzing associations
of in vivo brain volume and IQ. Paper presented at the
International Society for the Study of Individual Differences
2011, London, England.
Pilliner, A. E. G., Sutherland, J., & Taylor, E. G. (1960). Zero
error in Moray House verbal reasoning tests. British Journal
of Educational Psychology, 30, 53–62.
Raven, J. C. (1938). Standardization of progressive matrices,
1938. British Journal of Medical Psychology, 19, 137–150.
R Development Core Team. (2014). R: A language and environ-
ment for statistical computing (Version 3.1.1) [Computer soft-
ware]. Vienna, Austria: R Foundation for Statistical Computing.
Rindermann, H. (2008). Relevance of education and intelligence
at the national level for the economic welfare of people.
Intelligence, 36, 127–142. doi:10.1016/j.intell.2007.02.002
Rindermann, H. (2012). Intellectual classes, technological prog-
ress and economic development: The rise of cognitive capi-
talism. Personality and Individual Differences, 53, 108–113.
doi:10.1016/j.paid.2011.07.001
Rindermann, H., & Thompson, J. (2013). Ability rise in NAEP
and narrowing ethnics gap? Intelligence, 41, 821–831.
doi:10.1016/j.intell.2013.06.016
Rodgers, J. L. (1999). A critique of the Flynn effect: Massive IQ
gains, methodological artifacts, or both? Intelligence, 26,
337–356. doi:10.1016/S0160-2896(99)00004-5
Rowe, D. C., & Rodgers, J. L. (2002). Expanding variance and
the case of historical changes in IQ means: A critique of
Dickens and Flynn (2001). Psychological Review, 109, 759–
763. doi:10.1037/0033-295X.109.4.759
Rushton, J. P. (1997). Cranial size and IQ in Asian Americans
from birth to age seven. Intelligence, 25, 7–20. doi:10.1016/
S0160-2896(97)90004-0
Rushton, J. P. (1999). Secular gains in IQ are not related to
the g factor and inbreeding depression—Unlike Black–
White differences: A reply to Flynn. Personality and
Individual Differences, 26, 381–389. doi:10.1016/S0191-
8869(98)00148-2
Rushton, J. P., & Jensen, A. R. (2005). Thirty years of research
on race differences in cognitive ability. Psychology,
Public Policy, and Law, 11, 235–294. doi:10.1037/1076-
8971.11.2.235
Schaie, K. W., & Strother, C. R. (1968). A cross-sequential
study of age changes in cognitive behavior. Psychological
Bulletin, 70, 671–680. doi:10.1037/h0026811
Schooler, C. (1998). Environmental complexity and the Flynn
effect. In U. Neisser (Ed.), The rising curve: Long-term gains
in IQ and related measures (pp. 67–79). Washington, DC:
American Psychological Association. doi:10.1037/10270-002
Schubert, M. T., & Berlach, G. (1982). Neue Richtlinien zur
Interpretation des Hamburg-Wechsler-Intelligenztests für
Kinder (HAWIK) [New guidelines for the interpretation
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
306 Pietschnig, Voracek
of the Hamburg-Wechsler Intelligence Test for Children
(HAWIK)]. Zeitschrift für Klinische Psychologie, 11, 253–279.
Smits, I. A. M., Dolan, C. V., Vorst, H. C. M., & Wicherts, J. M.
(2011). Cohort differences in Big Five personality factors
over a period of 25 years. Journal of Personality and Social
Psychology, 100, 1124–1138. doi:10.1037/a0022874
Staff, R. T., Hogan, M. J., & Whalley, L. J. (2014). Aging trajec-
tories of fluid intelligence in late life: The influence of age,
practice and childhood IQ on Raven’s Progressive Matrices.
Intelligence, 47, 194–201. doi:10.1016/j.intell.2014.09.013
Statistical Research and Applications Branch. (2011). Joinpoint
Regression Program (Version 3.5.2) [Computer software].
Rockville, MD: National Cancer Institute.
Steen, R. G. (2009). Human intelligence and medical illness:
Assessing the Flynn effect. New York, NY: Springer.
Stern, W. (1912). Die psychologischen Methoden der
Intelligenzprüfung und deren Anwendung an Schulkindern
[The psychological methods of intelligence testing and their
applications in students]. Leipzig, Germany: Barth.
Storfer, M. (1999). Myopia, intelligence, and the expanding
human neocortex: Behavioral influences and evolutionary
implications. International Journal of Neuroscience, 98,
153–276. doi:10.3109/00207459908997465
Storfer, M. D. (1990). Intelligence and giftedness: The contribu-
tions of heredity and early environment. San Francisco, CA:
Jossey Bass.
Sundet, J. M., Barlaug, D. G., & Torjussen, T. M. (2004). The end
of the Flynn effect? A study of secular trends in mean intelli-
gence test scores of Norwegian conscripts during half a cen-
tury. Intelligence, 38, 38–44. doi:10.1016/j.intell.2004.06.004
Surani, M. A., Barton, S. C., & Norris, M. L. (1984). Development
of reconstituted mouse eggs suggests imprinting of the
genome during gametogenesis. Nature, 308, 548–550.
doi:10.1038/308548a0
Teasdale, T. W., & Owen, D. R. (1987). National secular trends
in intelligence and education: A twenty-year cross-sectional
study. Nature, 325, 119–121. doi:10.1038/325119a0
Teasdale, T. W., & Owen, D. R. (2005). A long-term rise and
recent decline in intelligence test performance: The Flynn
effect in reverse. Personality and Individual Differences,
39, 837–843. doi:10.1016/j.paid.2005.01.029
Te Nijenhuis, J. (2013). The Flynn effect, group differences,
and g loadings. Personality and Individual Differences, 55,
224–228. doi:10.1016/j.paid.2011.12.023
Te Nijenhuis, J., & van der Flier, H. (2013). Is the Flynn effect on
g? A meta-analysis. Intelligence, 41, 802–807. doi:10.1016/j.
intell.2013.03.001
Terman, L. M. (1921). Mental growth and the IQ. Journal of
Educational Psychology, 12, 325–341.
Thompson, J. (2012). Richard Lynn’s contribution to personal-
ity and intelligence. Personality and Individual Differences,
53, 157–161. doi:10.1016/j.paid.2011.03.013
Tuddenham, R. D. (1948). Soldier intelligence in World Wars
I and II. American Psychologist, 3, 54–56. doi:10.1037/
h0054962
Twenge, J. M. (2000). The age of anxiety? Birth cohort change in
anxiety and neuroticism, 1952–1993. Journal of Personality
and Social Psychology, 79, 1007–1021. doi:10.1037/0022-
3514.79.6.1007
Twenge, J. M. (2001). Birth cohort changes in extraversion: A
cross-temporal meta-analysis, 1966–1993. Personality and
Individual Differences, 30, 735–748. doi:10.1016/S0191-
8869(00)00066-0
Twenge, J. M., & Zhang, L. Q. (2004). It’s beyond my control: A
cross-temporal meta-analysis of increasing externality in locus
of control, 1960–2002. Personality and Social Psychology
Review, 8, 308–319. doi:10.1207/s15327957pspr0803_5
Uttl, B., & Van Alstine, C. L. (2003). Rising verbal intelligence scores:
Implications for research and clinical practice. Psychology and
Aging, 18, 616–621. doi:10.1037/0882-7974.18.3.616
Wechsler, D. (1981). Wechsler Adult Intelligence Scale—Revised.
New York, NY: The Psychological Corporation.
Wichman, A. L., Rodgers, J. L., & MacCallum, R. C. (2006).
A multilevel approach to the relationship between birth
order and intelligence. Personality and Social Psychology
Bulletin, 32, 117–127. doi:10.1177/0146167205279581
Williams, R. L. (2013). Overview of the Flynn effect. Intelligence,
41, 753–764. doi:10.1016/j.intell.2013.04.010
Williams, W. M. (1998). Are we raising smarter children today?
School- and home-related influences on IQ. In U. Neisser
(Ed.), The rising curve: Long-term gains in IQ and related
measures (pp. 125–154). Washington, DC: American
Psychological Association. doi:10.1037/10270-004
Woodley, M. A. (2011). Heterosis doesn’t cause the Flynn effect:
A critical examination of Mingroni (2007). Psychological
Review, 118, 689–693. doi:10.1037/a0024759
Woodley, M. A. (2012a). A life history model of the Lynn-Flynn
effect. Personality and Individual Differences, 53, 152–156.
doi:10.1016/j.paid.2011.03.028
Woodley, M. A. (2012b). The social and scientific temporal
correlates of genotypic intelligence and the Flynn effect.
Intelligence, 40, 189–204. doi:10.1016/j.intell.2011.12.002
Woodley, M. A., Figueredo, A. J., Brown, S. D., & Ross, K.C.
(2013). Four successful tests of the cognitive differentia-
tion-integration effort hypothesis. Intelligence, 41, 832–842.
doi:10.1016/j.intell.2013.02.002
Woodley, M. A., & Madison, G. (2013). Establishing an asso-
ciation between the Flynn effect and ability differentia-
tion. Personality and Individual Differences, 55, 387–390.
doi:10.1016/j.paid.2013.03.016
Woodley, M. A., & Meisenberg, G. (2013). In the Netherlands
the anti-Flynn effect is a Jensen effect. Personality and
Individual Differences, 54, 871–876. doi:10.1016/j.paid
.2012.12.022
Woodley, M. A., Te Nijenhuis, J., Must, O., & Must, A. (2014).
Controlling for increased guessing enhances the indepen-
dence of the Flynn effect from g: The return of the Brand
effect. Intelligence, 43, 27–34. doi:10.1016/j.intell.2013.12.004
World Bank. (2014). Databank: Fertility rate, total (births per
woman). Washington, DC: Author. Retrieved from http://
data.worldbank.org/indicator/SP.DYN.TFRT.IN
Zajonc, R. B., & Mullally, P. R. (1997). Birth order: Reconciling
conflicting effects. American Psychologist, 52, 685–699.
doi:10.1037/0003-066X.52.7.685
at Vienna University Library on May 18, 2015pps.sagepub.comDownloaded from
