British Journal of Health Psychology

See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/6373449
Mind the gap(s)… in theory, method and data: Re-examining Kanazawa
(2006)
Article in British Journal of Health Psychology · June 2007
DOI: 10.1348/135910707X174339 · Source: PubMed
CITATIONS
8
READS
88
3 authors, including:
Some of the authors of this publication are also working on these related projects:
Adolescent Sociality across Cultures: Establishing a Japan-UK Collaboration View project
Evolutionary theory View project
Thomas E. Dickins
Middlesex University, UK
86 PUBLICATIONS 1,230 CITATIONS
SEE PROFILE
Rebecca Sear
London School of Hygiene and Tropical Medicine
105 PUBLICATIONS 3,642 CITATIONS
SEE PROFILE
All content following this page was uploaded by Rebecca Sear on 29 November 2017.
The user has requested enhancement of the downloaded file.
Mind the gap(s) . . . in theory, method and data:
Re-examining Kanazawa (2006)
T. E. Dickins1†, R. Sear2* and A. J. Wells3
1School of Psychology, University of East London & Centre for the Philosophy of the
Natural and Social Sciences, London School of Economics, UK
2Department of Social Policy, London School of Economics, UK
3Institute of Social Psychology, London School of Economics, UK
Kanazawa (2006) has put forward an evolutionarily grounded theory which claims that
individuals in wealthier and more egalitarian societies live longer and stay healthier not because
they are wealthier or more egalitarian but because they are more intelligent (2006: 637). The
claim rests on an argument which asserts that general intelligence is a solution to
evolutionarily novel problems and that most dangers to health in contemporary society
are evolutionarily novel. Kanazawa also claims that this relationship does not hold in
sub-Saharan Africa. These claims are based on a cross-national analysis which finds a
positive correlation between ‘national’ IQ scores and mortality data. The implication is
that intelligence is the principal factor determining longevity in the rest of the world,
regardless of issues such as adequacy of diet and availability of health care. Kanazawa’s
theoretical claims about the evolution of general intelligence as a domain-specific
adaptation are inconsistent with adaptationist analysis: natural selection does not solve
general problems. The assumptions that sub-Saharan Africa is more representative of
the evolutionary past than is the rest of the world, and that most hazards to health in
contemporary society are evolutionarily novel, are implausible. The methods used are
inadequate because Kanazawa argues for causation from correlation and fails to
consider alternative explanations. The IQ data are flawed for reasons to do with sample
size and sampling, extrapolation and inconsistency across measures. Nor are they
temporally compatible with the economic and demographic data.
Over the last decade, evolutionary theory has become more generally applied to the
behavioural sciences and its implications for health have increasingly been recognized
(Eaton et al., 2002). As evolutionists, we applaud this development. Evolutionary theory
improves our understanding of human behaviour and ecology. However, Kanazawa’s
recent paper in this journal falls short of the scientific standards required to further this
synthesis and make a useful contribution to health psychology. In this paper, we address
failings in his theory, his methods and the data he deploys.
* Correspondenceshould be addressed to Rebecca Sear, Department of Social Policy, London School of Economics, Houghton St,
London WC2A 2AE, UK (e-mail: r.sear@lse.ac.uk).
† We have ordered authorship alphabetically; all authors contributed equally to this paper.

167
DOI:10.1348/135910707X174339
Kanazawa attempts to challenge the well-documented argument (Marmot &
Wilkinson, 1999; Wilkinson, 1992, 2000) that differences in health and life expectancy
are explained by social and economic inequalities. He proposes, instead, that
intelligence is the principal determinant of health and longevity. This proposal is linked
to Kanazawa’s theory about the evolution of general intelligence (Kanazawa, 2004). He
sees general intelligence as a domain-specific adaptation for solving evolutionarily novel
problems and argues that most dangers to health in contemporary environments are
evolutionarily novel. As a result, he claims that more intelligent individuals are better
able to cope with these dangers and thus live longer.
As a secondary argument, Kanazawa claims that the environment of sub-Saharan
Africa is evolutionarily more familiar than that of the rest of the world. Consequently,
sub-Saharan Africa provides less evolutionary novelty for its population in all domains,
including health. Kanazawa concludes that intelligence and health outcomes should not
be related in sub-Saharan Africa (SSA).
Kanazawa attempts to support these arguments with statistical analyses combining
IQ data from Lynn and Vanhanen (2002) with economic and demographic data from
United Nations and World Bank databases. He finds a positive relationship between socalled ‘national’ IQ data and mortality rates. When measures of IQ, economic
development and inequality are all correlated with mortality rates, only IQ emerges as a
significant predictor. When sub-Saharan countries are analysed separately, this
relationship is no longer found.
We show that Kanazawa’s claims run counter to evolutionary theory, his methods are
inadequate and the IQ data are seriously flawed.
Problems with theory
Inequality and stress
Kanazawa opens his paper with a brief outline of Wilkinson’s work on socio-economic
inequalities and health. Wilkinson (2000) lays out the thesis that less egalitarian societies
with large differences in relative wealth are less healthy than more egalitarian countries
or those with smaller relative wealth differences. Finding oneself at the bottom of a
social hierarchy with few affiliates and other supports is a highly stressful situation.
Cummins (2005) describes the neuroendocrine system, pointing out that adrenaline is
designed to mobilize energy immediately and increase heart rate, whereas cortisol
produces a slower response that replenishes energy by releasing glucose from stored
fatty acids.
Both Wilkinson and Cummins see neuroendocrine functioning as a conditional
proximate mechanism that has been selected for in order to produce appropriate
energy-releasing responses in problematic situations, and as directly related to social
dominance and subordination. Kanazawa claims that the stress mechanisms discussed
by Wilkinson reduce the fitness of individuals if they are activated because of ‘chronic
low status’ and argues that natural selection would have provided an escape. His
argument rests on a number of questionable assumptions.
First, Kanazawa assumes that there are individuals with permanently low status in
primate groups. He also seems to assume that there is no movement within the lower
ranks. He cites no evidence to support these views and disregards evidence that points
to low-ranking individuals in all primate species, including humans, forming coalitions
to ameliorate their situation and improve their fitness (see Cummins, 2005, for a
comprehensive review).
168 Dickins et al.
A second questionable assumption is that if the position of chronically low status
individuals never changes, they will avoid conflict. In fact, the inherent risks of mate
acquisition, for example, may be far greater for chronically low status individuals than
for those further up the hierarchy, and as a consequence they may face many more shortterm emergencies. In short, Kanazawa provides no reason to reject the hypothesis that
neuroendocrine functioning is an adaptation that is useful to low- and high-ranking
individuals alike. The utility of this adaptation is that it releases energy to deal with
emergencies that would otherwise undermine fitness, perhaps permanently.
The Savanna principle and evolutionary novelty
After discussing inequality and stress, Kanazawa sets out the Savanna principle which
claims that ‘the human brain has difficulty in comprehending and dealing with entities
and situations that did not exist in the EEA [environment of evolutionary adaptedness],
including virtually everything in modern society except for people and many social
relationships’ (Kanazawa, 2006, p. 625). Although Kanazawa lays claim to this idea, the
notion has existed in the literature for sometime under the title of the mismatch
hypothesis (for example, see Dickins, 2006; Eaton et al., 2002; Nesse, 2005). The notion
of a mismatch between ancestral and modern environments was at the heart of early
evolutionary experiments on reasoning (Cosmides, 1989). These experiments made it
clear that the EEA should not be regarded as a specific time and place, but as a statistical
composite of the selection history for a given trait (Tooby & Cosmides, 2005).
Kanazawa identifies the human EEA with a Savanna environment, and is, presumably,
alluding to the Pleistocene era which is discussed in some evolutionary psychology
literature as an important contributory epoch in our evolution (Badcock, 2000), but is
certainly not the whole of it. The notion of mismatch has also attracted criticism (Irons,
1998; Laland & Brown, 2006). There is clear evidence that new physiological
adaptations have emerged since the origin of agriculture (e.g. lactase persistence and
adaptations conferring resistance to malaria; Hill et al., 1991; Holden & Mace, 1997).
Behavioural evolution has also probably occurred. Nevertheless, Kanazawa wants to
argue for a particularly strong form of the mismatch hypothesis claiming that virtually
everything in modern society, except for people and social relationships, is novel. This
cannot possibly be true. The concept of a shelter, for example, is not novel although the
particular forms we have adopted and the materials used are. Is Kanazawa seriously
saying that the human brain has difficulty in comprehending the advantages of running
water, windows and heating mechanisms? Curiously, Kanazawa chooses an
experimental game, the one-shot Prisoner’s Dilemma, to illustrate how poorly humans
are suited to modern environments. Perhaps players of one-shot Prisoner’s Dilemma
games behave ‘irrationally’ because completely anonymous social exchange is unlikely
in any environment, ancient or modern.
The evolution of general intelligence
In the third stage of his theory building, Kanazawa claims that general intelligence is a
domain-specific adaptation to deal with new and non-recurrent problems. Kanazawa is
here tackling what Plotkin (1995) has referred to as the uncertain future problems. The
problem exists for any species whose selection history has resulted in a suite of tailored
adaptations that represent and deal effectively with past problem spaces. The success of
individual members of such a species depends upon the stability of those problem
Mind the gap(s): Re-examining Kanazawa 169
spaces. How can such an organism deal with novelty? Novelty can be thrown up at any
point in the organism’s future and this possibility introduces uncertainty. Behavioural
failure in the face of such uncertainty is at the core of the mismatch hypothesis which
argues, in effect, that in many cases novelty cannot be dealt with. Kanazawa’s Savanna
principle is an extreme version of this view. His proposed solution suggests that general
intelligence affords sufficient flexibility to deal with novelty (see also Hampton, 2004 for
an argument very similar to Kanazawa’s).
Relying on the concept of general intelligence does not solve the novelty problem
(see Dickins, 2005 for a detailed discussion). First, evolution through natural selection
operates, as Kanazawa admits, to select specific solutions to specific recurrent
problems. There is no such thing as a general adaptive problem that could have a general
adapted solution. Second, evolution through natural selection has no foresight and,
therefore, could not select for adaptations to future problems, be they recurrent or nonrecurrent. In other words, natural selection cannot represent future environments
within the genotype of an organism.
Kanazawa discusses novelty in terms of non-recurrent problems that happen
frequently enough to be of consequence. This should strike us as odd. If a problem
occurs with sufficient frequency to create a selection pressure, it is a problem that
evolutionary processes can provide a solution to. The problem will have specific
characteristics and natural selection will find a domain-specific solution. If, on the other
hand, a problem is entirely new and never before encountered the only hope of a
solution is the introduction of trait variation via the usual means, such as mutation, and
subsequent selection. For this to occur, the problem would have to persist which again
calls into question the idea that it can aptly be called novel. Kanazawa’s attempt to build
solutions to novelty into the architecture of the human mind undermines the notion of
novelty. Once a solution is instantiated in the architecture, the problem is solved and
thus no longer novel.
Novelty, health and longevity
Once Kanazawa has introduced his theory of general intelligence, he claims that health
covaries with intelligence. He supports this claim by citing Deary, Whiteman, Starr,
Whalley, and Fox (2004) who looked at Scottish mental surveys from 1932 to 1947.
Deary et al. found a significant relationship between important health outcomes and
childhood intelligence, but stated that ‘caution is needed before inferring any causality
in this relationship. The link between childhood psychometric intelligence and either
specific pathology (with the exception of dementia) or physical frailty is unproven, and
thus the pathway by which childhood psychometric intelligence relates to ill-health in
old age is unclear’ (Deary et al., 2004, p. 143).
Regardless of Deary et al.’s cautions, Kanazawa invokes the Savanna principle
and the concept of general intelligence to support a causal hypothesis. According to
Kanazawa, most health risks faced today are evolutionarily novel. He lists challenges
that he claims are peculiar to modern environments – these include cigarettes,
alcohol, sedentary lifestyles, automobiles and guns and, later, crack cocaine and
vodka. Kanazawa then simply states that ‘high-g individuals can better recognize
such dangers to health, deal with them appropriately and so remain healthier and
live longer’ (p. 626).
This argument rests on a large number of unsupported assumptions. First, the list of
health risks deserves some attention. Smoking, drinking and taking drugs are activities
170 Dickins et al.
with short-term effects, often used to ameliorate stress-related symptoms that are
typically associated with elevated cortisol (and thus consistent with Wilkinson’s story).
Moreover, the assumption that smoking, drinking and risk-taking are evolutionarily
novel behaviours is implausible. Kanazawa buttresses his argument by claiming that
modern dangers are more potent, but this is a pure speculation.
What is more, the health consequences of these risks can equally well be explained
in terms of discounting (Dickins, 2006) where the true costs kick in post-reproductive
age. If one is socially stressed, then one might seek to reduce the immediate negative
effects, as they are experienced as more pressing than possible long-term
negative consequences in the future. Such future discounting effects are well
documented (see, for example Wilson & Daly, 1997) and clearly relate to socioeconomic status, much as Wilkinson would have us believe. Moreover, discounting
effects may displace rational thinking about risks and as far as we know there is no
relationship between intelligence and consideration for future consequences. Kanazawa
presents no hard evidence for a relationship between intelligence and risk-taking
behaviour of the kind his theory requires.
In addition, a link between high intelligence and low rates of risk-taking behaviours
assumes that the costs and benefits of engaging or not engaging in risky behaviour are the
same in all contexts. If extrinsic mortality is high, then the intelligent decision might be to
engage in risky behaviour (with possible immediate benefits), since the risk of dying from
causes unrelated to the risky behaviour is high. Under these circumstances, the degree of
discounting therefore depends on the environment, rather than intelligence.
Sub-Saharan Africa
The final plank of Kanazawa’s argument is the claim that in sub-Saharan Africa ‘the site of
our ancestral environment, where, even today, life in tribal societies is less radically
different from the ancestral environment than in the rest of the world’ (p. 626), there
should be no relationship between intelligence and health and longevity. Even if we take
at face value Kanazawa’s argument for a Pleistocene type of EEA, his suggestion that
sub-Saharan Africans are still living in an evolutionarily familiar environment is
implausible. The epoch that Kanazawa is alluding to was a time when our ancestors
were involved in food production by hunting and gathering, yet only a tiny proportion
of modern sub-Saharan Africans are still hunter-gatherers. Much of sub-Saharan Africa is
involved in agriculture, both subsistence and commercial. There are substantial
differences between agricultural and hunter-gatherer populations, not only in methods
of food production, but also in mating, parenting and demography. It has been argued
that in some respects, the mating and parenting patterns of modern industrialized
countries are more like those of hunter-gatherers (and thus less evolutionarily novel)
than those of agricultural populations. For example, agricultural populations tend to
have higher levels of polygyny and lower levels of paternal investment (Kaplan &
Lancaster, 2003). The fertility, mortality and population growth of agricultural
populations also seem to be higher than those of hunter-gatherers. In any case, a
large and growing proportion of sub-Saharan Africans now live in urban environments
which, like the industrialized world, are evolutionarily novel. In 2003, almost 40% of
Africa’s population lived in urban areas according to UN statistics (this figure includes
North Africa, but in all sub-Saharan African regions, at least a quarter of the population
was urbanized: United Nations, 2004). All of which again calls into question Kanazawa’s
assertions and assumptions about novelty.
Mind the gap(s): Re-examining Kanazawa 171
Problems with methods
Correlation is not causation
Kanazawa finds positive correlations between national IQ and three measures of
mortality (except when only sub-Saharan African countries are included in the models).
He infers a direct causal relationship between IQ and mortality, but correlation is not
causation. If there is an underlying variable which affects both IQ and mortality, then IQ
and mortality rates may not be causally linked. Health is a candidate for such an
underlying variable. Poor health is clearly related to the risk of death and has also been
correlated with cognitive performance (Ezeamama et al., 2005; Fernando et al., 2003;
Liu, Raine, Venables, Dalais, & Mednick, 2003; Pearce, Deary, Young, & Parker, 2005). An
obvious analogy is with height. Height, like IQ, is highly heritable, but variation in height
also has a large environmental component, particularly in resource-stressed
populations. Analysis within societies suggests that height is often negatively related
to mortality rates (Costa, 1993; Marmot, Shipley, & Rose, 1984; Waaler, 1984). The usual
explanation for these findings is that there is an underlying variable which affects both
height and longevity: health. Health and nutritional status influence final adult height
and also affect mortality rates both during childhood and adulthood: height may
therefore be a marker of a healthy phenotype. Kanazawa’s complete failure to consider
alternative explanations for his findings is a serious weakness of the paper. He has simply
not considered the possibility that healthy brains might be part of healthier phenotypes
more generally.
Choice of mortality statistics
Kanazawa correlates three measures of mortality with IQ: life expectancy at birth, infant
mortality rates (,1 year) and mortality rates of young adults (15–19 years). These three
analyses are not independent as life expectancy at birth is calculated from age-specific
mortality rates (including infant mortality and mortality of 15–19-year olds). Quoting
from the technical notes provided by the UN for the data Kanazawa uses on life
expectancy and infant mortality rates, ‘In areas with high infant and child mortality
rates, the indicator (life expectancy at birth) is strongly influenced by trends and
differentials in infant and child mortality’ (United Nations Statistics Division, 2006).
Kanazawa’s choice of infant and young adult mortality rates also raises problems for
his theoretical arguments. He claims that the correlation between the infant mortality
rate and IQ is seen because there is a link between parents’ intelligence and infant
mortality. He also finds that economic development and inequality matter not at all if IQ
is entered into the regression equation. But it is difficult to envisage exactly how the
intelligence of a caretaker could be the only influence on the mortality of very young
children. The biggest killers of babies are congenital abnormalities and infectious
diseases (Murray & Lopez, 1997). Endogenous causes such as congenital abnormalities
are not amenable to changes in the behaviour of caretakers. Exogenous causes such as
infectious disease may conceivably be affected by the propensity of the caretaker to seek
medical attention (which may be linked to intelligence), but this depends on medical
care being available, affordable and of sufficiently high quality. At least when conducting
such a crude, cross-national comparison of infant mortality rates, it seems more
parsimonious to assume that variation in the mortality of very young children is more
strongly affected by variation in the provision of medical care and public health services
than in the intelligence of the child’s caretaker.
172 Dickins et al.
The rationale for choosing to analyse young adult mortality is also unconvincing:
Kanazawa argues that this is the age at which individuals begin to make their own
health-related decisions. But variation in mortality at this age for women may be at least
partly explained by variation in maternal mortality rates (AbouZahr & Wardlaw, 2003).
It is unclear how maternal mortality can be considered a ‘choice’. In addition, as noted
above, many of Kanazawa’s supposedly novel hazards have detrimental effects on health
largely in the long term. Even if individuals were beginning to smoke, drink and take
drugs in their teens, the effects of these behaviours would probably be most noticeable
in later life.
The analysis is not properly controlled
Kanazawa includes only a control for GDP in his analysis, but the factors he really needs
to control for are the availability of education and health services, since these are the
relevant factors which will affect variation in mortality rates. While economic
development and mortality rates are clearly linked (Preston, 1975), countries with
similar levels of economic development can display considerable differences in the
health of their populations because of differences in spending on health and education
(Caldwell, 1986). There are numerous other factors Kanazawa does not consider. For
example, countries with the worst mortality statistics tend to be either those which
have recently engaged in conflicts or those with high levels of HIV/AIDS.
Collinearity
Given that Lynn and Vanhanen (2002) find their national IQ data to be highly correlated
with economic development, entering both IQ and GDP into the same regression model
raises problems of collinearity. This means that the individual regression coefficients
cannot be estimated reliably, so that, at the least, the magnitude of the correlation
between IQ and mortality (stressed at several points by Kanazawa) cannot be taken at
face value.
Problems with data
Problems with IQ data
One of the principal criticisms that must be made of Kanazawa’s paper is its
unquestioning reliance on IQ data obtained from Lynn and Vanhanen (2002). Kanazawa
says of this source, ‘Lynn and Vanhanen compiled a comprehensive list of “national IQs”
of 185 nations in the world, either by calculating the mean scores from a large number of
primary data or carefully estimating them from available sources’ (Kanazawa, 2006,
p. 627). Primary data were available for 81 countries only, so the majority of the so-called
‘national’ figures are estimates. The estimation process was flawed, but this is a
secondary concern. The first consideration is the quality of the primary data.
The primary data are grossly inadequate for two reasons: first, the sampling is
sketchy at best and ludicrously insufficient at worst; second, the calculations of mean
values from multiple samples and the method of adjustment to account for the ‘Flynn
effect’ are both fundamentally inadequate. Consider, first, a few examples of the primary
data. The ‘national’ IQ figure for Barbados is derived from a sample of 108, 9–15-year
olds. The figure for Ethiopia is derived from a sample of 250, 15-year-old immigrants to
Israel. The figure for Nigeria is derived from one sample of 86 adult men and one sample
Mind the gap(s): Re-examining Kanazawa 173
of 375, 6–13-year olds. The figure for Sierra Leone is derived from one sample of 22,
23-year-old skilled workers and one sample of 60 adults. The figure for Russia is derived
from a sample (no sample size reported) of 14–15-year olds drawn from the city of
Briansk. In no case do the data appear to be derived from samples that can plausibly be
regarded as representative of the national populations discussed.
We have looked in detail at the sources from which Lynn and Vanhanen derived their
‘national’ data for Ethiopia, Nigeria and Sierra Leone, choosing these countries because
the figures are exceptionally low and they are sub-Saharan countries which feature
prominently in Kanazawa’s analysis. The figures in question are Ethiopia (63), Nigeria
(67) and Sierra Leone (64). In each case, Lynn and Vanhanen have misused data gathered
for purposes other than the making of national IQ estimates.
The data used by Lynn and Vanhanen for Ethiopia were taken from a paper by Kaniel
and Fisherman (1991) which compared the IQ scores of 250 14–15-year-old Ethiopian
immigrant Jews with a sample of 1740 Israeli Jews aged 9–15 years. Kaniel and
Fisherman attributed the low IQ test performance of the Ethiopian immigrants, relative
to the comparison sample of Israelis of the same age, to cognitive delay rather than to
difference. They also pointed out that there is a consensus that minority and immigrant
groups score lower than dominant groups in IQ tests. Kaniel and Fisherman explicitly
stated that their results do ‘not imply that the subjects will exhibit low cognitive abilities
over the long term’ (op.cit., p. 30) and cited another study which showed that immigrant
Ethiopian Jews could catch up with their Israeli counterparts given a suitable
intervention programme. Given this analysis, it is unacceptable for Lynn and Vanhanen
to have used the immigrant sample’s average score as a stable representation of the IQ of
the Ethiopian population as a whole.
The data for Nigeria and Sierra Leone are equally flawed for purposes of national
comparisons. The data for Sierra Leone are taken from two sources. Berry (1966)
compared a sample of the Temne from Sierra Leone with an Eskimo sample to examine
the effects of cultural and ecological differences on perceptual skills. Binnie-Dawson
(1984) carried out further cross-cultural work matching his Temne sample with one
from central Australia. In neither case was the Temne sample intended to be
representative of the whole population of Sierra Leone.
Berry investigated two groups of Temne in order, in part, ‘to eliminate “race” as a
comprehensive explanation for any perceptual differences found between the societies
(Berry 1966, p. 209). It is not clear which group Lynn and Vanhanen used because the
sample size they report does not match either of the groups in Berry’s study, but it is
quite clear that Berry’s findings explicitly demonstrate intra-national variation as a
result of acculturation. It was therefore inappropriate for Lynn and Vanhanen to take
any data from this study as representative of a fixed national IQ value. Moreover,
Berry explicitly made the point that the tests used were culturally biased and that
culture-free tests are unattainable. For cross-national comparisons of the kind Lynn and
Vanhanen make, this is a crucial point and further invalidates their use of the data.
Binnie-Dawson (1984) was also explicit about the cultural bias of the tests employed.
For this reason his data, too, are fundamentally unsuited for the use to which Lynn and
Vanhanen put them.
The data which Lynn and Vanhanen used for Nigeria were drawn from two sources.
Wober (1969) tested a sample of Nigerian adults on a range of measures. Fahrmeier
(1975) compared a sample of Hausa schoolchildren with a sample of unschooled
children. Wober tested his participants twice with a 6-month gap between initial testing
and retesting. He stated clearly that ‘a strong case can be made that the second testing
174 Dickins et al.
gave a distinctly more valid measure of whatever abilities the Matrices and EFT
[Embedded Figures Test] involve.’ Lynn and Vanhanen ignored this point and reported
the lower figure from the first test as the sample value without comment or justification.
Fahrmeier (1975) explored the question of whether schooling has an effect on
cognitive development. Lynn and Vanhanen simply recorded the average score on
Raven’s Coloured Progressive Matrices for 375 of Fahrmeier’s participants without
regard either to the purposes of the study or to Fahrmeier’s discussion of his results. The
375-test scores used by Lynn and Vanhanen to calculate their average were the aggregate
of eight-group scores reported by Fahrmeier. The results showed that there were
statistically significant differences between different age groups and also between
schooled and unschooled cohorts. It is quite obvious from these results that an average
derived from the whole set of scores should not be used as a fixed IQ value for the
cohort as a whole, never mind for the population of Nigeria as a whole.
It is clear that Lynn and Vanhanen’s ‘national’ data are not accurate, representative or
valid for the purposes to which they have been put. We conclude by commenting on the
sub-Saharan data in general because of Kanazawa’s focus on this area. Lynn and
Vanhanen’s book gives data for 29 sub-Saharan countries but 18 of the figures are
estimates. The primary sub-Saharan African data consist of figures for 11 countries.
These data cannot be representative of the populations of sub-Saharan Africa.
The inadequacy of the sub-Saharan African sampling calls into question one of the
major features of Kanazawa’s results, viz. the finding that mortality data are not
predicted by IQ in that part of the world although they are elsewhere. One explanation
for this is that the finding is a simple statistical artifact. Inspection of the IQ figures for
the 11 countries for which ‘primary’ data are available shows a range of only 14 IQ
points and a standard deviation of 4.34. By contrast, the non-sub-Saharan African
countries have a range of 35 IQ points and a standard deviation of 7.66. There is,
therefore, comparatively little variability in the SSA figures and, consequently, they may
have little predictive power for this reason alone.
Problems with demographic and economic data
Kanazawa uses data on mortality prepared by the UN from national statistical sources.
While this mortality data may be the best available, it is important to note that there is
considerable variation in the quality of national data collected, and that many of the
statistics presented for developing countries are estimates (United Nations Statistics
Division, 2006). The quality of the data on young adult (15–19) mortality is particularly
questionable. This analysis is based on an extremely small number of deaths for some
countries, since mortality rates for young adults are usually relatively low (the 15–19
mortality rates given in the Appendix are expressed per 1000, which is not mentioned
in the Appendix nor the text of the paper). In a few cases, the rates are based on
fewer than 30 deaths, for example both male and female rates in Luxemburg and
female rates in Armenia, Estonia, Slovenia and Macedonia. Age-specific mortality rates
are particularly prone to error because of age misreporting, common in countries
with relatively low levels of literacy (Ewbank, 1981). Again referring to the UN: ‘the
reliability of age data should be of concern to users of these statistics’ (United Nations,
2002, p. 5).
Similar data quality caveats apply to the economic data collected by the World Bank
(2004). It is also worth noting that data on income inequality are not available for a large
number of countries. Lynn and Vanhanen (2002) produced national IQ estimates for 185
Mind the gap(s): Re-examining Kanazawa 175
countries, but only 126 of these have available data on both GDPand income inequality. The
included countries may not be a representative sample of the world’s population (it is also
worth noting that at no point does Kanazawa explain why only two-thirds of the countries
with available IQ data were included in his analysis, indicating a somewhat cavalier attitude
to the usual requirements for rigour when describing scientific methodology).
Comparability of IQ, demographic and economic data
The data used by Kanazawa are not all temporally comparable. Data on GINI coefficients
from the World Bank were collected at various times during the 1990s and early 2000s;
GDPs are presented for 2004; data on life expectancy at birth and infant mortality are
presented for 2000–2005; data on young adult mortality rates were collected variously
between 1993 and 2002. Given that economic development and mortality rates both
change over time, including data in the analysis collected over a 15-year period
introduces both noise and bias. These demographic and economic data are also not
comparable to the IQ data. The raw data used to calculate national IQ figures were
collected between the 1930s and 1990s. Various adjustments were made to the raw data
to account for a potential secular trend over time, but these adjustments involved a
variety of questionable assumptions.
Conclusion
Our examination of Kanazawa’s paper demonstrates clearly the inadequacies of his
theory, methods and data. We can see no reason to believe his claim that intelligence is
the only significant determinant of mortality. To the contrary, the evidence that
inequality, economic development, schooling, health care and nutrition are the
principal determinants of mortality remains unaffected by Kanazawa’s mistaken and
misleading arguments.
Kanazawa’s weak analysis does not negate the value of research exploring the
relations between IQ test scores and health. As Deary et al. (2004) have shown, there is
reliable evidence for just such a relationship in Scotland. Similarly, careful studies in
other countries will also be useful. It remains unclear, however, that such studies will
demonstrate causal relationships between IQ and mortality. If IQ is measuring
something that is sensitive to ecological circumstances and is related to health, then
generalizing the relationship across ecologies will lose the all important detail. What is
needed for all further work is a clear understanding of the relationship between IQ and
health within societies.
References
AbouZahr, C., & Wardlaw, T. (2003). Maternal mortality in 2000: Estimates developed by the
WHO, UNICEF and UNFPA. Geneva: World Health Organisation.
Badcock, C. (2000). Evolutionary psychology: A critical introduction. London: Polity Press.
Berry, J. W. (1966). Temne and Eskimo perceptual skills. International Journal of Psychology,
1(3), 207–229.
Binnie-Dawson, J. L. (1984). Biosocial and endocrine bases of spatial ability. Psychologia, 27,
129–151.
Caldwell, J. C. (1986). Routes to low mortality in poor countries. Population and Development
Review, 12(2), 171–220.
176 Dickins et al.
Cosmides, L. (1989). The logic of social-exchange: Has natural selection shaped how humans
reason? Studies with the Wason selection task. Cognition, 31(3), 187–276.
Costa, D. L. (1993). Height, weight, wartime stress, and older age mortality: Evidence from the
Union Army records. Explorations in Economic History, 30(4), 424–449.
Cummins, D. (2005). Dominance, status and hierarchies. In D. M. Buss (Ed.), The handbook of
evolutionary psychology (pp. 676–697). New Jersey: Wiley.
Deary, I. J., Whiteman, M. C., Starr, J. M., Whalley, L. J., & Fox, H. C. (2004). The impact of
childhood intelligence on later life: Following up the Scottish Mental Surveys of 1932 and
1947. Journal of Personality and Social Psychology, 86(1), 130–147.
Dickins, T. E. (2005). Can there ever be a non-specific adaptation? A response to Simon J. Hampton.
Journal for the Theory of Social Behaviour, 35(3), 329–340.
Dickins, T. E. (2006). Evolutionary health psychology. Health Psychology Update, 15(1), 4–10.
Eaton, S. B., Strassman, B. I., Nesse, R. M., Neel, J. V., Ewald, P. W., Williams, G. C., et al. (2002).
Evolutionary health promotion. Preventive Medicine, 34(2), 109–118.
Ewbank, D. C. (1981). Age misreporting and age-selective underenumeration: Sources, patterns
and consequences for demographic analysis. Washington, DC: National Academy Press.
Ezeamama, A. E., Friedman, J. F., Acosta, L. P., Bellinger, D. C., Langdon, G. C., Manalo, D. L., et al.
(2005). Helminth infection and cognitive impairment among Filipino children. American
Journal of Tropical Medicine and Hygiene, 72(5), 540–548.
Fahrmeier, E. D. (1975). Effect of school attendance on intellectual-development in NorthernNigeria. Child Development, 46(1), 281–285.
Fernando, S. D., Gunawardena, D. M., Bandara, M., De Silva, D., Carter, R., Mendis, K. N., et al.
(2003). The impact of repeated malaria attacks on the school performance of children.
American Journal of Tropical Medicine and Hygiene, 69(6), 582–588.
Hampton, S. J. (2004). Adaptations for nothing in particular. Journal for the Theory of Social
Behaviour, 34(1), 35–53.
Hill, A. V. S., Allsop, C. E. M., Kwiatkowski, D., Anstey, N. M., Twumasi, P., Rowe, P. A., et al. (1991).
Common West African HLA antigens are associated with protection from severe malaria.
Nature, 352, 595–600.
Holden, C., & Mace, R. (1997). Phylogenetic analysis of the evolution of lactose digestion in adults.
Human Biology, 69(5), 605–628.
Irons, W. (1998). Adaptively relevant environments versus the environment of evolutionary
adaptedness. Evolutionary Anthropology, 6(6), 194–204.
Kanazawa, S. (2004). General intelligence as a domain-specific adaptation. Psychological Review,
111(2), 512–523.
Kanazawa, S. (2006). Mind the gap : : : in intelligence: Re-examining the relationship bewteen
inequality and health. British Journal of Health Psychology, 11, 623–642.
Kaniel, S., & Fisherman, S. (1991). Level of performance and distribution of errors in the
progressive matrices test – a comparison of Ethiopian immigrant and Native Israeli
adolescents. International Journal of Psychology, 26(1), 25–33.
Kaplan, H., & Lancaster, J. (2003). An evolutionary and ecological analysis of human fertility,
mating patterns and parental investment. In K. W. Wachter & R. A. Bulatao (Eds.), Offspring:
Human fertility in biodemographic perspective (pp. 170–223). Washington: National
Academies Press.
Laland, K. N., & Brown, G. R. (2006). Niche construction, human behavior and the adaptive-lag
hypothesis. Evolutionary Anthropology, 15(3), 95–104.
Liu, J. H., Raine, A., Venables, P. H., Dalais, C., & Mednick, S. A. (2003). Malnutrition at age 3 years
and lower cognitive ability at age 11 years: Independence from psychosocial adversity.
Archives of Pediatrics and Adolescent Medicine, 157(6), 593–600.
Lynn, R., & Vanhanen, T. (2002). IQ and the wealth of nations. Westport, CT: Praeger.
Marmot, M., Shipley, M. J., & Rose, G. (1984). Inequalities in death–specific explanations of a
general pattern? Lancet, 1(8384), 1003–1006.
Mind the gap(s): Re-examining Kanazawa 177
Marmot, M. & Wilkinson, R. G. (Eds.), (1999). Social determinants of health. Oxford: Oxford
University Press.
Murray, C. J. L., & Lopez, A. D. (1997). Mortality by cause for eight regions of the world: Global
burden of disease study. Lancet, 349(9061), 1269–1276.
Nesse, R. M. (2005). Evolutionary psychology and mental health. In D. M. Buss (Ed.), The
handbook of evolutionary psychology (pp. 903–930). New Jersey: Wiley.
Pearce, M. S., Deary, I. J., Young, A. H., & Parker, L. (2005). Growth in early life and childhood IQ at
age 11 years: The Newcastle thousand families study. International Journal of Epidemiology,
34(3), 673–677.
Plotkin, H. (1995). Darwin machines and the nature of Knowledge. London: Penguin.
Preston, S. H. (1975). The changing relationship between mortality and level of economic
development. Population Studies, 29(2), 231–248.
Tooby, J., & Cosmides, L. (2005). Conceptual foundations of evolutionary psychology. In D. M.
Buss (Ed.), The handbook of evolutionary psychology (pp. 5–67). New Jersey: Wiley.
United Nations. (2002). Demographic Yearbook 2002. New York: Population Division,
Department of Economic and Social Affairs.
United Nations. (2004). World urbanization prospects: The 2003 Revision. New York: Population
Division, Department of Economic and Social Affairs.
United Nations Statistics Division. (2006). Social indicators. [Available at: http://unstats.un.org/
unsd/demographic/products/socind/health.htm].
Waaler, H. T. (1984). Height, weight and mortality: The Norwegian experience. Acta Medica
Scandinavica, (Suppl. 679), 1–51.