Journal of Behavioral Decision Making, J. Behav. Dec.

Making 25: 264–275 (2012)

Published online 24 January 2011 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/bdm.730

Why More Intelligent Individuals Like Classical Music

SATOSHI KANAZAWA1* and KAJA PERINA2
1
Department of Management, London School of Economics and Political Science, UK
2
Psychology Today, NY, USA

ABSTRACT

The origin of values and preferences is an unresolved theoretical problem in social and behavioral sciences. The Savanna‐IQ Interaction
Hypothesis, derived from the Savanna Principle and a theory of the evolution of general intelligence, suggests that more intelligent individuals
are more likely to acquire and espouse evolutionarily novel values and preferences than less intelligent individuals but that general
intelligence has no effect on the acquisition and espousal of evolutionarily familiar values and preferences. Recent work on the evolution of
music suggests that music in its evolutionary origin was always vocal and that purely instrumental music is evolutionarily novel. The
Savanna‐IQ Interaction Hypothesis would then imply that more intelligent individuals are more likely to prefer purely instrumental music than
less intelligent individuals, but general intelligence has no effect on the preference for vocal music. The analyses of American (General Social
Surveys) and British (British Cohort Study) data are consistent with this hypothesis. Additional analyses suggest that the effect of intelligence
on musical preference is not a function of the cognitive complexity of music. Copyright © 2011 John Wiley & Sons, Ltd.

key words origin of values and preferences; evolutionary psychology; the Savanna Principle; the Savanna‐IQ Interaction Hypothesis;
evolutionary origin of music

Where do individuals’ values and preferences come from? arily novel problems is their logical solvability; all evolution-
Why do people like or want what they do? The origin of arily novel problems have logical solutions (Kanazawa, 2010b,
individual values and preferences is one of the remaining pp. 282–283).
theoretical puzzles in social and behavioral sciences The logical conjunction of these two theories, the
(Kanazawa, 2001). In particular, values and preferences Savanna‐IQ Interaction Hypothesis (Kanazawa, 2010a),
occupy a central role in the process of decision making implies that the human brain’s difficulty with evolutionarily
(Ariely & Loewenstein, 2006; Hechter, Ranger‐Moore, novel stimuli may interact with general intelligence, such
Jasso, & Horne, 1999; Yates & Tschirhart, 2006). that more intelligent individuals have less difficulty with
Recent theoretical developments in evolutionary psy- such stimuli than less intelligent individuals. In contrast,
chology may suggest one possible explanation (Kanazawa, general intelligence may not affect individuals’ ability to
2010b). On the one hand, evolutionary psychology comprehend and deal with evolutionarily familiar entities
(Crawford, 1993; Symons, 1990; Tooby & Cosmides, and situations.
1990) posits that the human brain, just like any other organ Evolutionarily novel entities that more intelligent individ-
of any other species, is designed for and adapted to the uals are better able to comprehend and deal with may include
conditions of the ancestral environment (roughly the African ideas and lifestyles, which form the basis of their values and
savanna during the Pleistocene Epoch), not necessarily to preferences; it would be difficult for individuals to prefer or
those of the current environment. It may therefore have value something that they cannot truly comprehend.
difficulty comprehending and dealing with entities and Comprehension does not equal preference. Although not
situations that did not exist in the ancestral environment everyone who comprehends certain entities would thereby
(Kanazawa, 2002, 2004a). acquire preferences for them, we assume some would,
On the other hand, an evolutionary psychological theory of whereas few (if any) who do not comprehend them would
the evolution of general intelligence proposes that general acquire preferences for them.
intelligence may have evolved as a domain‐specific adaptation Hence, applied to the domain of preferences and values,
to solve evolutionarily novel problems, for which there are no the Hypothesis suggests that more intelligent individuals are
predesigned psychological adaptations (Kanazawa, 2004b, more likely to acquire and espouse evolutionarily novel
2008). Such evolutionarily novel, nonrecurrent adaptive preferences and values that did not exist in the ancestral
problems may have included, for example, how to escape a environment than less intelligent individuals, but general
forest fire caused by lightning striking a tree, how to find new intelligence has no effect on the acquisition and espousal of
sources of food in a severe drought that has never been evolutionarily familiar preferences and values that existed in
encountered before, and how to cross a rapid river in the midst the ancestral environment (Kanazawa, 2010a).
of a flash flood. What characterizes the domain of evolution- There has been emerging evidence for the Hypothesis as
an explanation for individual preferences and values. First,
more intelligent children are more likely to grow up to
* Correspondence to: Satoshi Kanazawa, Managerial Economics and espouse left‐wing liberalism (Deary, Batty, & Gale, 2008;
Strategy Group, Department of Management, London School of Economics
and Political Science, Houghton Street, London WC2A 2AE, UK. Kanazawa, 2010a), possibly because genuine concerns with
E‐mail: S.Kanazawa@lse.ac.uk genetically unrelated others and willingness to contribute

Copyright © 2011 John Wiley & Sons, Ltd.

S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 265

private resources for the welfare of such others—liberalism male competition for resources and mates in the ancestral
—may be evolutionarily novel. Even though past studies environment. We may infer this from the fact that behavior
show that women are more liberal than men (Lake & that would be classified as criminal if engaged in by humans
Breglio, 1992; Shapiro & Mahajan, 1986; Wirls, 1986) and are quite common among other species (Ellis, 1998),
blacks are more liberal than whites (Kluegel & Smith, including other primates (de Waal, 1989, 1992; de Waal,
1986; Sundquist, 1983), the effect of childhood intelligence Luttrell, & Canfield, 1993). It also explains the “exception that
on adult liberalism is twice as large as the effect of sex or proves the rule,” why more intelligent individuals are more
race (Kanazawa, 2010a). likely to consume illegal drugs (Kanazawa & Hellberg, 2010).
Second, more intelligent children are more likely to grow Unlike most interpersonal and property crimes, the consump-
up to be atheists (Kanazawa, 2010a), possibly because belief tion of such substances is evolutionarily novel. It is not
in higher powers, as a consequence of over‐inference of legality per se that matters but evolutionary novelty of the
agency behind otherwise natural phenomena, may be part of behavior.
evolved human nature (Atran, 2002; Boyer, 2001; Guthrie, However, there was very little formal third‐party
1993; Haselton & Nettle, 2006; Kirkpatrick, 2005), and enforcement of norms in the ancestral environment, only
atheism may therefore be evolutionarily novel. Even though second‐party enforcement (victims and their kin and allies)
past studies show that women are much more religious than or informal third‐party enforcement (ostracism). It therefore
men (Miller & Hoffmann, 1995; Miller & Stark, 2002), the makes sense from the perspective of the Hypothesis that men
effect of childhood intelligence on adult religiosity is twice with low intelligence may be more likely to resort to
as large as that of sex (Kanazawa, 2010a). evolutionarily familiar means of competition for resources
Third, more intelligent boys (but not more intelligent (theft rather than full‐time employment) and mating
girls) are more likely to grow up to value sexual exclusivity opportunities (rape rather than computer dating) and not to
(Kanazawa, 2010a), possibly because humans were naturally comprehend fully the consequences of criminal behavior
polygynous throughout evolutionary history (Alexander, imposed by evolutionarily novel entities of law enforcement.
Hoogland, Howard, Noonan, & Sherman, 1979; Harvey &
Bennett, 1985; Kanazawa & Novak, 2005; Leutenegger &
Kelly, 1977; Pickford, 1986). Either under monogamy or EVOLUTIONARY ORIGINS OF MUSIC
polygyny, women are expected to be sexually exclusive to
one mate; in sharp contrast, men in polygynous marriage are What is the evolutionary origin of music? Why are humans
not expected to be sexually exclusive to one mate, whereas musical?
men in monogamous marriage are. So, sexual exclusivity In comparison to evolutionary origins and functions of
may be evolutionarily novel for men but not for women. language and art, anthropologists and archeologist have paid
Fourth, more intelligent children are more likely to grow scant attention to the origin of music. In his book The
up to be nocturnal, going to bed and waking up later Singing Neanderthals: The Origins of Music, Language,
(Kanazawa & Perina, 2009), possibly because nocturnal life Mind and Body (2005), the cognitive archeologist Steven
was rare in the ancestral environment where our ancestors Mithen offers a novel theory of the evolution of music.
did not have artificial sources of illumination until the Mithen argues that language and music had a common
domestication of fire. Ethnographies of contemporary precursor, in what Brown (2000) calls “musilanguage,”
hunter‐gatherers suggest that our ancestors may have woken which later developed into two separate systems of music
up shortly before dawn and gone to sleep shortly after dusk. and language.
Night life may therefore be evolutionarily novel. There are two distinct approaches to the evolution of
Fifth, the human consumption of psychoactive sub- language. The compositional approach (Bickerton, 1990;
stances, such as alcohol, tobacco, and drugs, is evolution- Jackendoff, 2000) avers that words came before sentences. A
arily novel, all originating less than 10 000 years ago. Thus, lexicon of words that referred to specific entities, such as
the Hypothesis would predict that more intelligent individ- “meat,” “fire,” and “hunt,” emerged first and were later
uals are more likely to consume alcohol, tobacco, and drugs. combined into phrases and then sentences. Grammar
The analyses of two prospectively longitudinal data sets with emerged at the end to dictate how words could be combined
nationally representative samples in the UK and the USA into sentences.
support the prediction. More intelligent individuals consume In contrast, the holistic approach (Wray, 1998) proposes
more alcohol more frequently, smoke more tobacco (but that sentences came before words. It suggests that the
only in the USA), and use more illegal drugs (Kanazawa & precursor to human language was a communication system
Hellberg, 2010). composed of messages in the form of arbitrary strings of
Finally, criminals on average have lower intelligence than sounds rather than words. Each indivisible utterance or
the general population (Wilson & Herrnstein, 1985; sequence of sounds was associated with a specific meaning.
Herrnstein & Murray, 1994). This is consistent with the These utterances were later segmented into words, which
Hypothesis because, although much of what we call could then be recombined to create further utterances.
interpersonal crime today is evolutionarily familiar, the Mithen favors the latter view. As evidence, he points to
institutions that control, detect, and punish such behavior the fact that all nonhuman primate utterances, such as vervet
are evolutionarily novel (Kanazawa, 2009). Murder, assault, monkey’s alarm calls, rhythmic chatters of geladas, duets of
robbery, and theft were probably routine means of intrasexual pair‐bonded gibbons, and pant‐hoots of chimpanzees, are

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
266 Journal of Behavioral Decision Making

holistic and indivisible (Mithen, 2005, pp. 105–121). In precursor to the modern language and music. Although the
other words, nonhuman primates do not have words, even Pirahã language does have words, it has the fewest number
though their utterances as a whole convey specific meanings of vowels (three) and consonants (seven for women, eight
and emotions. Some primatologists disagree, however. for men) of all known human languages. “The Pirahã people
Zuberbühler (2002, 2003), for example, argues, in support communicate almost as much by singing, whistling, and
of the compositional approach, that Diana and Campbell’s humming as they do using consonants and vowels. Pirahã
monkey calls have both syntactic and semantic rules, which prosody is very rich, with a well‐documented five‐way
can be used to combine elements (“words”) to produce weight distinction between syllable types” (Everett, 2005,
further utterances. The debate on the origin of human p. 622).
language between the compositional and holistic approaches The former professional musician and current academic
is far from closed. linguist (as well as the originator of the holistic approach to
Studies demonstrate that the meanings and the emotions the evolutionary origin of language) Alison Wray (2006)
of primate utterances may be shared by different primate notes that, “To my taste, western classical music (as indeed
species. When macaque vocalizations made in specific most other musical traditions worldwide) is different in kind
social contexts as expressions of contentment, pleading, [from musical expressions in evolutionary history]. Its
dominance, anger, and fear are recorded and then played production is, for a start, subject to a heavy burden of
back, Finnish children and adults are able to interpret learning that few master. There is no naturally facilitated
accurately what the expressed emotions are (Leinonen, access to the comprehension (let alone creation) of the kinds
Linnankoski, Laakso, & Aulanko, 1991; Linnankoski, of melodies, harmonies and rhythms found in the works of
Laakso, Aulanko, & Leinonen, 1994). Another study Bach or Schoenberg: no equivalent—for music of this kind
shows that words spoken by Finnish and English speakers —of first language acquisition.” In other words, classical
in the social context of contentment, pleading, dominance, music of Bach, Schoenberg, and others is evolutionarily
anger, and fear have the same acoustic waveforms as the novel, partly, we contend, because it is largely or entirely
macaque vocalizations made in the corresponding contexts instrumental.
(Leinonen, Laakso, Carlson, & Linnankoski, 2003). It is as Consistent with Wray’s assertion, we observe that a far
though humans and macaques may be able to communicate greater proportion of the general population can (and
with each other through the use of holistic utterances and spontaneously do) sing songs than play musical instruments.
messages. For example, the incidence of tone deafness in the UK is
Mithen contends that human proto‐language was holistic, estimated to be about 4–5% (Kalmus & Fry, 1980). In other
manipulative (it was designed to induce desired emotions words, 95% of people can sing adequately (and some of the
and behavior in other individuals), multimodal (it involved tone‐deaf people nonetheless often do sing). The proportion
not only vocal utterances but also gesture and dance), of the general population who play musical instruments
musical (the utterances had distinct pitches, rhythms, and adequately is nowhere near as high. Further, in many cases
melodies), and mimetic. This proto‐language eventually of playing musical instruments (such as the guitar and the
evolved into two systems of communication: music to piano), it is often accompanied by singing.
express emotions and language to transmit information. To In the context of the Savanna‐IQ Interaction Hypothesis,
demonstrate the common evolutionary origin of music and then, the theory by Mithen suggests that more intelligent
language, Mithen (2005, pp. 28–68) surveys a large number individuals today are more likely to appreciate purely
of clinical cases of individuals with amusia (absence of instrumental music than less intelligent individuals because
musical abilities while retaining some linguistic abilities) and such music is evolutionarily novel, while general intelligence
aphasia (absence of linguistic abilities while retaining some has no effect on the appreciation of vocal music.
musical abilities). These case studies largely show that music The study by Rentfrow and Gosling (2003) shows that
and language are based on discrete modules in the brain; more intelligent individuals prefer “reflective and complex”
some of these are separate and dedicated to one or the other, genre of music (which includes classical, jazz, blues, and
whereas others are shared. folk), but they also prefer “intensive and rebellious” music
If Mithen is right, if music and language share a common (alternative, rock, and heavy metal). Less intelligent
evolutionary origin in holistic, musical utterances designed individuals in their study prefer “upbeat and conventional”
to convey messages, one possible implication is that music, music (country, pop, religious, and sound tracks). In a recent
in its evolutionary origin, was necessarily and invariably study, Chamorro‐Premuzic and Furnham (2007, p. 177)
vocal. If Mithen is right, then all music in its evolutionary identify three distinct “uses” of music: “namely emotional
origin were songs that individuals sang to express their (i.e. music for emotional regulation such as mood manipula-
desires and emotions, in an attempt to induce desired tion), cognitive (i.e. rational musical appreciation or
emotions and behavior in others. In other words, music in its intellectual processing of music), and background (i.e. music
evolutionary origin was never purely instrumental. as background for social events, work or interpersonal
It may be instructive to note in this context that Blackfoot interaction).” Their data show that more intelligent individ-
Indians have a word for “song” but not for “instrumental uals are more likely to use music for “cognitive” purposes,
music” (Nettl, 1983). The language of the Pirahã in the but intelligence is not correlated with the “emotional” use of
Amazon forest in Brazil (Everett, 2005) may be an extant music. If Mithen’s view of the evolution of music is correct,
example of a musilanguage, which Mithen envisions as the then the original function of music was to induce certain

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 267

emotions in self and others, and its “cognitive” use would be more vocal musical genres than instrumental ones in itself
evolutionarily novel. It is therefore possible to interpret the suggests the vocal origin of music.
findings by Chamorro‐Premuzic and Furnham (2007) as For a measure of preference for instrumental music, we
being consistent with the prediction of the Savanna‐IQ compute the mean score for the three instrumental genres;
Interaction Hypothesis. for a measure of preference for vocal music, we compute the
In the following section, we will provide an empirical test mean score for the 15 vocal genres. In constructing these
of our hypothesis that more intelligent individuals are more indices and using them as dependent variables in ordinary
likely to prefer purely instrumental music than less least squares (OLS) regression models, we in effect treat the
intelligent individuals, but general intelligence has no effect original five‐point ordinal scale as interval. However,
on individuals’ preference for vocal music. We will test the treating the dependent variables as ordinal and using the
hypothesis with both American and British data with large, ordinal regression models (McCullagh, 1980) produce
representative samples. identical substantive conclusions.

EMPIRICAL ANALYSES Independent variable: intelligence

The GSS measures the verbal intelligence of its respondents
American sample by asking them to select a synonym for a word out of five
Data candidates. Half the respondents in each GSS sample
The National Opinion Research Center at the University of (including the 1993 sample that we use) answer 10 of these
Chicago has administered the General Social Surveys (GSS) questions, and their total score (the number of correct
either annually or biennially since 1972. Personal interviews responses) varies from 0 to 10. The raw score is then
are conducted with a nationally representative sample of normalized into a standard IQ metric, with a mean of 100
non‐institutionalized adults in the USA. The sample size is and a standard deviation of 15.
about 1500 for each annual survey and about 3000 for each This is strictly a measure of verbal intelligence, not of
biennial one. The exact questions asked in the survey vary general intelligence. However, verbal intelligence is known
by the year. In our analysis, we use the 1993 GSS, which to be highly correlated with (and thus heavily loads on)
includes specific questions about the respondents’ taste in general intelligence. The extensive review of 36 studies by
music. Miner (1957) showed that the median correlation between
vocabulary and general intelligence is 0.83. Wolfle (1980)
reported that the correlation between a full‐scale IQ test
Dependent variable: music preference (Army General Classification Test) and the GSS synonyms
In 1993 only, the GSS asks its respondents about their measure that we use here is 0.71. As a result, the GSS
preference for 18 different types of music. The question is “I’m synonyms measure has been used widely by intelligence
going to read you a list of some types of music. Can you tell me researchers to assess trends in general intelligence (Huang &
which of the statements on this card comes closest to your Hauser, 1998).
feeling about each type of music?” The respondent can answer
on a five‐point scale (reverse coded): 1 = dislike it very much;
2 = dislike it; 3 = have mixed feelings; 4 = like it; and 5 = like it Control variables
very much. The 18 genres of music asked are “big band,” In addition to intelligence, we control for the following
“bluegrass,” “country western,” “blues or R&B,” “Broadway variables in the OLS regression models: age (in years);
musicals,” “classical,” “folk,” “gospel,” “jazz,” “Latin,” “easy race (1 if black); sex (1 if male); education (in years of
listening,” “new age,” “opera,” “rap,” “reggae,” “contemporary formal schooling); annual family income (in 21 equidistant
rock,” “oldie,” and “heavy metal.” ordinal categories, from 1 = less than $1000 to 21 = more
Of the 18 types of music, we classify big band, than $75 000, treated here as interval); religion (in four
classical, and easy listening as entirely or largely dummies for Catholic, Jewish, Protestant, and other, with
“instrumental” and the rest as entirely or largely “vocal.” none as the reference category); whether the respondent is
We admit that the dichotomization is necessarily approx- currently married (1 if yes); whether the respondent has
imate at best; for example, some of both classical and big ever been married (1 if yes); and the total number of
band music is vocal. However, we classify the genres children.
according to whether the majority of the genre is largely
instrumental or vocal.
It is debatable whether or not jazz is largely instrumental Results
or largely vocal. However, all our substantive conclusions The inspection of the variance inflation factors for all
below remain exactly the same if we classify it as independent variables suggests that multicollinearity is not a
instrumental. Because Rentfrow and Gosling (2003) show problem at all in our regression analysis (O’Brien, 2007).
that more intelligent individuals prefer to listen to jazz, our Table 1 presents the analysis of the 1993 GSS data. Column
classification of jazz as largely vocal provides a statistically (1) shows the result of the OLS multiple regression analysis
conservative test of our prediction derived from the Savanna‐ where the GSS respondents’ mean preference for instrumen-
IQ Interaction Hypothesis. The fact that there are so many tal music is regressed on their verbal intelligence and a set of

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
268 Journal of Behavioral Decision Making

Table 1. Effect of intelligence on preference for instrumental versus which is moderate for survey data. The only other significant
vocal music, American sample (U.S. General Social Survey, 1993) predictors of preference for instrumental music are age
(1) (2) (b = 0.0139, p < 0.0001, beta = 0.2880) and education
(b = 0.0549, p < 0.0001, beta = 0.2075). Both age and
Instrumental Vocal
education significantly increase the respondent’s preference
Verbal intelligence 0.0069** 0.0034 for instrumental music, and their effects are stronger than
(0.0022) (0.0018)
0.1245 0.0954 that of intelligence. None of the other sociodemographic
Age 0.0139**** −0.0048** characteristics of the respondent significantly predict prefer-
(0.0019) (0.0016) ence for instrumental music.
0.2880 −0.1462
Race (1 = black) −0.1053 0.0908 Table 1, Column (2), shows the result of the OLS
(0.1002) (0.0782) multiple regression analysis where the GSS respondents’
−0.0369 0.0518 mean preference for vocal music is regressed on their verbal
Sex (1 = male) −0.0606 −0.0789
(0.0548) (0.0438) intelligence and a set of control variables. It shows that, net
−0.0375 −0.0769 of the same control variables, more intelligent Americans are
Education 0.0549**** 0.0199* not more likely to prefer vocal music. Once again, consistent
(0.0111) (0.0092)
0.2075 0.1152 with the prediction derived from the Savanna‐IQ Interaction
Family income 0.0065 0.0041 Hypothesis, intelligence does not significantly affect the
(0.0066) (0.0053) preference for vocal music among the GSS respondents
0.0407 0.0411
Religion (b = 0.0034, not significant, beta = 0.0954). As before, age
Catholic 0.1411 −0.0364 and education significantly predict preference for vocal
(0.1055) (0.0801) music, although, unlike with instrumental music, they have
0.0731 −0.0293
Protestant 0.0031 0.0057 opposite effects. Age significantly decreases preference for
(0.0963) (0.0709) vocal music (b =−0.0048, p < 0.01, beta = −0.1462), whereas
0.0019 0.0054 education significantly increases it (b = 0.0199, p < 0.05,
Jewish −0.2856 0.0724
(0.2251) (0.1628) beta = 0.1152). No other sociodemographic characteristics of
−0.0452 0.0199 the respondent significantly predicts preference for vocal
Other 0.1656 0.0234 music. The analysis of the GSS data presented in Table 1
(0.2034) (0.1642)
0.0299 0.0064 supports the prediction of the Savanna‐IQ Interaction
Currently married (1 = yes) −0.0603 −0.1079 Hypothesis that more intelligent individuals are more likely
(0.0721) (0.0588) to prefer (evolutionarily novel) instrumental music, whereas
−0.0371 −0.1049
Ever married (1 = yes) 0.1344 0.0149 intelligence has no effect on preference for (evolutionarily
(0.0961) (0.0756) familiar) vocal music.
0.0634 0.0121
Number of children −0.0362 −0.0140
(0.0188) (0.0163)
−0.0769 −0.0450 British sample
Constant 1.2875 2.7123 Data
(0.2413) (0.1868)
R2 0.1659 0.0925 The British Cohort Study (BCS), originally developed as
n 786 543 the British Birth Survey and a sequel to the 1958 National
Child Development Study, includes all babies born in
Note: “Instrumental” music includes “big band,” “classic,” and “easy
listening.” Great Britain (England, Wales, and Scotland) during the
“Vocal” music includes “blues,” “bluegrass,” “contemporary rock,” week of 05–11 April 1970. The initial sample contains over
“country,” “folk,” “gospel,” “heavy metal,” “jazz,” “Latin,” “musicals,” 17 000 babies. All surviving members of the cohort, who
“new age,” “oldies,” “opera,” “rap,” and “reggae.”
Main entries are unstandardized regression coefficients. still reside in the UK (Great Britain plus Northern Ireland),
Numbers in parentheses are standard errors. have since been observed in 1975 (at age 5), 1980 (age
Numbers in italics are standardized coefficients (betas). 10), 1986 (age 16), 1996 (age 26), and 2000 (age 30). We
* p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.
use the 1986 follow‐up data because they contain extensive
information on the respondents’ leisure activity, including
musical preferences, as well as a measure of their
control variables. It shows that, net of age, race, sex, intelligence.
education, family income, religion, current and past marital
status, and the number of children, more intelligent
Americans are more likely to prefer instrumental music Dependent variable: music preference
(such as big band, classical, and easy listening) than less The 1986 follow‐up of the BCS asks the respondent’s musical
intelligent Americans. Consistent with the prediction derived preference for 12 different types of music. The question is
from the Savanna‐IQ Interaction Hypothesis, verbal intelli- “What sort of music do you listen to usually?” The respondent
gence significantly increases the preference for instrumental can answer a yes or a no to each of the 12 genres. We catego-
music among the GSS respondents (b = 0.0069, p < 0.01, rize “classical” and “light music” (easy listening or “elevator
beta = 0.1245). The effect size for intelligence, using the music”) as entirely or largely “instrumental” music, and
standardized regression coefficient as a proxy, is 0.1245, “folk music,” “disco,” “reggae,” “soul,” “heavy rock,” “funk,”

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 269

“electronic,”1 “punk,” “other pop music,” and “other” as entirely reference category); religion (with nine dummies for
or largely “vocal” music. For a measure of preference for Roman Catholic, the Church of England, Other Christian,
instrumental music, we compute the mean score across the two Buddhist, Hindu, Islam, Jewish, Sikh, and Other, with
instrumental genres; for a measure of preference for vocal None/Atheist as the reference category; however, there are
music, we compute the mean score for the 10 vocal genres. In no Buddhist respondents in our sample); family income (on
constructing these indices and in using them as dependent an 11‐point equidistant ordinal categories, from 1 = less
variables in OLS regression models, we treat the original binary than £50/week or £2600/year, to 11 = more than £500/week
response as interval. However, treating the dependent variables or £26 000/year, treated here as continuous); and mother’s
as ordinal and using the ordinal regression models (McCullagh, and father’s education (both measured as the age at which
1980) produce identical substantive conclusions. the parent left formal education).
In addition to music preference, BCS86 measures the
respondent’s TV‐viewing habits, by asking about 22 different
types of TV programs. The respondent can answer on a three‐ Results
point scale: 1 = view it as little as I can; 2 = view it some of the The inspection of the variance inflation factors for all
time; and 3 = view it as much as I can. Two of the 22 types of TV independent variables once again suggests that multicol-
programs refer to music: pop/rock music and classical music. linearity is not a problem at all in our regression analysis
We use the respondent’s preference for viewing music programs (O’Brien, 2007). Table 2 presents the analysis of the BCS86
on TV as a secondary measure of music preference. Because the respondents’ preference for music. Column (1) shows the
dependent variables are on three‐point ordinal scale, we use the result of the OLS regression analysis where BCS86
ordinal regression (McCullagh, 1980) to analyze them. respondents’ mean preference for instrumental music is
regressed on their verbal intelligence and a set of control
variables. It shows that, net of teacher‐rated academic
Independent variable: intelligence performance, sex, race, religion, family income, and
The BCS86 measures the verbal intelligence of its respondents mother’s and father’s education, more intelligent British
by asking them to select a synonym for a word out of five 16‐year‐olds are more likely to prefer instrumental music
candidates. These are essentially the same type of questions as (such as classical and light music) than their less intelligent
the GSS vocabulary test, which is known to correlate very highly schoolmates. Consistent with the prediction derived from the
with general intelligence (Huang & Hauser, 1998; Miner, 1957; Savanna‐IQ Interaction Hypothesis, verbal intelligence
Wolfle, 1980). Each BCS86 respondent answers 75 of these significantly increases preference for instrumental music
questions. Their raw score (0–75) is normalized into a standard among the BCS86 respondents (b = 0.0023, p < 0.01,
IQ metric, with a mean of 100 and a standard deviation of 15. beta = 0.0990). The effect size for intelligence, once again
using the standardized regression coefficient as a proxy, is
small (0.0990). Academic performance also has a signifi-
Control variables cantly positive effect on preference for instrumental music
Because all BCS86 respondents are 16 years old and still in (b = 0.0301, p < 0.001, beta = 0.1303). Races do not differ in
school, we cannot control for their educational achievement; their preference for instrumental music, but Hindu and Sikh
all respondents have had the identical number of years of religions both have significant positive effects in this model
formal schooling. In order to separate the effect of (Hindu: b = 0.2435, p < 0.05, beta = 0.0767; Sikh: b = 0.3438,
intelligence from education, we control for the respondent’s p < 0.01, beta = 0.0857). No other sociodemographic vari-
academic performance. The teacher of each BCS86 respon- ables included in this model significantly affect preference
dent rates the student’s academic performance on seven‐ for instrumental music.
point scales (reverse coded): 1 = “bottom” (bottom 5%); Table 2, Column (2), presents the result of the OLS
2 = “well below average” (the next 10%); 3 = “below regression analysis where BCS86 respondents’ mean prefer-
average” (the next 20%); 4 = “average” (the middle 30%); ence for vocal music is regressed on their verbal intelligence
5 = “above average” (the next 10%); 6 = “well above and a set of control variables. It shows that, net of the same
average” (the next 10%); and 7 = “top” (the top 5%). We control variables, more intelligent British teenagers are not
enter this measure of academic performance in our more likely to prefer vocal music than their less intelligent
regression models in lieu of educational achievement. schoolmates. Once again, consistent with the prediction derived
In addition, we control for the following variables: sex from the Savanna‐IQ Interaction Hypothesis, verbal intelli-
(1 = male); race (with five dummies for West Indian, gence does not have a significant effect on preference for vocal
Asian,2 Chinese, Mixed, and Other, with European as the music (b = 0.0003, not significant, beta = 0.0176), nor does any
other sociodemographic variable included in the model.
Table 3 presents the analyses of the BCS86 respondents’
1
The BCS86 code book and data consistently refer to “electric” music. preference for watching musical programs on TV.3 Column
Because, to our knowledge, there is no genre of music called “electric,” we
assume that this is a typographical error for “electronic,” a genre of music
that was particularly popular in the UK in the 1980s.
2
In the UK, the racial category “Asian” refers almost exclusively to South 3
Some of the variables in the two models presented on Table 3, namely West
Asians (chiefly Indian and Pakistani) and usually excludes East Asians Indian, mixed race, and Jewish dummies for the classical music model, and
(Chinese, Korean, and Japanese), hence the separate category “Chinese” West Indian dummy in the popular music model, are excluded in order to
from “Asian.” However, there are no Chinese respondents in our sample. avoid complete separation (singularities in the Fisher information matrix).

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
270 Journal of Behavioral Decision Making

Table 2. Effect of intelligence on preference for instrumental Table 3. Effect of intelligence on watching music programs on TV,
versus vocal music, British Sample (British Cohort Study, 1986 British Sample (British Cohort Study, 1986 Follow‐up)
Follow‐up)
(1) (2)
(1) (2)
Classical Pop
Instrumental Vocal
Verbal intelligence 0.0276** −0.0210**
Verbal intelligence 0.0023** 0.0003 (0.0092) (0.0076)
(0.0008) (0.0005) 0.4133 −0.3146
0.0990 0.0176 Academic performance −0.0096 0.0245
Academic performance 0.0301*** −0.0059 (0.0843) (0.0699)
(0.0082) (0.0050) −0.0121 0.0308
0.1303 −0.0424 Sex (1 = male) −.4317** −.5844****
Sex (1 = male) −0.0234 −0.0082 (0.1569) (0.1277)
(0.0155) (0.0095) −0.2159 −0.2922
−0.0440 −0.0256 Race
Race Asian −0.0219 1.3771
Asian −0.0742 −0.0062 (1.0267) (0.9937)
(0.1074) (0.0660) −0.0026 0.1618
−0.0286 −0.0040 West Indian — —
West Indian −0.0042 0.0559 Mixed — 0.5626
(0.2576) (0.1582) (1.2575)
−0.0005 0.0104 0.0440
Mixed −0.1738 0.0517 Other 0.2789 0.1623
(0.1377) (0.0846) (0.4765) (0.4323)
−0.0388 0.0193 0.0485 0.0282
Other −0.0328 −0.0097 Religion
(0.0491) (0.0302) Roman Catholic 0.5979 0.3191
−0.0195 −0.0097 (0.3880) (0.3050)
Religion 0.2075 0.1108
Roman Catholic 0.0615 0.0080 Church of England 0.3194 0.1661
(0.0365) (0.0224) (0.3458) (0.2502)
0.0762 0.0166 0.1536 0.0799
Church of England 0.0203 0.0163 Other Christian 0.5575 −0.0244
(0.0307) (0.0189) (0.3926) (0.2975)
0.0363 0.0487 0.1754 −0.0077
Other Christian 0.0400 0.0347 Hindu −0.0427 −0.5179
(0.0367) (0.0225) (1.3749) (1.0258)
0.0490 0.0719 −0.0039 −0.0478
Hindu 0.2435* 0.0273 Islam 0.7902 −1.3520
(0.1198) (0.0736) (1.0307) (0.8698)
0.0767 0.0144 0.0909 −0.1555
Islam 0.1168 0.0022 Jewish — −1.6356
(0.1095) (0.0672) (1.3654)
0.0368 0.0012 −0.1130
Jewish 0.0846 −0.0960 Sikh 3.1003** 0.4968
(0.1841) (0.1130) (1.1341) (1.2579)
0.0134 −0.0254 0.2514 0.0403
Sikh 0.3438** 0.0766 Other religion −0.0426 0.0340
(0.1310) (0.0804) (1.1583) (0.9048)
0.0857 0.0320 −0.0035 0.0028
Other religion −0.1030 −0.0065 Family income 0.0095 0.0132
(0.1091) (0.0670) (0.0331) (0.0284)
−0.0281 −0.0030 0.0237 0.0328
Family income −0.0025 0.0011 Mother’s education 0.0034 −0.0235
(0.0034) (0.0021) (0.0401) (0.0343)
−0.0246 0.0182 0.0079 −0.0543
Mother’s education 0.0020 0.0011 Father’s education 0.0080 −0.0132
(0.0043) (0.0026) (0.0305) (0.0266)
0.0172 0.0156 0.0224 −0.0367
Father’s education 0.0020 −0.0018 Threshold [Y = 1] 4.6036 −5.5753
(0.0034) (0.0021) (0.9483) (0.7894)
0.0221 −0.0343 Threshold [Y = 2] 6.8192 −3.3012
Constant −0.3175 0.2722 (0.9667) (0.7748)
(0.0885) (0.0543) χ2 goodness of fit 2170.4309 2125.5258
R2
0.0619 0.0071 −2LogLikelihood 1286.6756* 1719.1149***
n 1160 1160 Cox and Snell pseudo R2 0.0296 0.0404
n 1138 1126
Note: “Instrumental” music includes “classical” and “light.”
“Vocal” music includes “folk,” “disco,” “reggae,” “soul,” “hard rock,” Main entries are unstandardized regression coefficients.
“funk,” “electronic,” “punk,” “other,” and “pop.” Numbers in parentheses are standard errors.
Main entries are unstandardized regression coefficients. Numbers in italics are standardized coefficients (bx sx).
Numbers in parentheses are standard errors. Note: * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.
Numbers in italics are standardized coefficients (betas).
* p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. (1) presents the result of the ordinal regression analysis
where BCS86 respondents’ frequency of watching classical
music TV shows is regressed on their verbal intelligence and

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 271
110.0
a set of control variables. Despite the fact that less intelligent
individuals may enjoy watching TV in general more than 108.0 106.5
more intelligent individuals (Kanazawa, 2006), more
106.0
intelligent British teenagers watch classical music on TV
102.9
more frequently than their less intelligent counterparts. Net 104.0
of the same sociodemographic variables as before, verbal 102.0 100.6

Mean IQ
intelligence has a significantly positive effect on the
frequency of watching classical music TV programs 100.0
(b = 0.0276, p < 0.01, beta = 0.4133); its effect size, measured 98.0
by x‐standardized regression coefficient (0.4133), is larger 94.8
96.0
than that of any other variable included in the model. Sex 93.3
and the Sikh religion are the only other variables included in 94.0
the model that have significant effects; girls watch classical
92.0
music programs on TV significantly more frequently than n=91 n=163 n=240 n=307 n=170
boys (b = −0.4109, p < 0.01, beta = −0.2054), and Sikhs 90.0
Dislike very much Mixed feelings Like very much
watch them significantly more frequently than atheists
Dislike Like
(b = 3.1003, p < 0.01, beta = 0.2514). Like or dislike classical music
Table 3, Column (2), presents the result of the ordinal
regression analysis where BCS86 respondents’ frequency of Figure 1. Mean IQ by preference for classical music (U.S. General
watching pop music TV shows is regressed on their verbal Social Survey, 1993). Note: Error bars represent the standard error
for the mean
intelligence and a set of control variables. It shows that,
consistent with the earlier findings with an American
sample (Kanazawa, 2006), less intelligent British teenagers 110.0
watch popular music on TV more frequently than their
more intelligent counterparts. Net of the same socio- 108.0 106.9
demographic variables as before, verbal intelligence has a
significantly negative effect on the frequency of watching
popular music TV programs (b = −0.0210, p < 0.01, beta = 106.0
−0.3146). Sex is the only other variable included in the
Mean IQ

model that has a significant effect; once again, girls watch 104.0
popular music programs on TV significantly more fre-
quently than boys (b = −0.5904, p < 0.0001, beta = −0.2952).
102.0
It is interesting to note that verbal intelligence has the
opposite effects on the frequency of watching classical and 99.7
popular music programs on TV, whereas sex has the same 100.0
effect on both. The analysis of the BCS86 data presented in
n=5000 n=661
Tables 2 and 3 is consistent with the results from the 98.0
American sample presented in Table 1 and supports the No Yes
prediction of the Savanna‐IQ Interaction Hypothesis that Do you usually listen to classical music?
more intelligent individuals are more likely to prefer
Figure 2. Mean IQ by preference for classical music (British Cohort
(evolutionarily novel) instrumental music, whereas intelli- Study, 1986 Follow‐up). Note: Error bars represent the standard
gence has no effect on preference for (evolutionarily error for the mean
familiar) vocal music.
For illustrative purposes, Figure 1 presents the mean IQ
of GSS respondents by their preference for classical music. It music” (n = 661) is 107, whereas that among those who do
shows that there is a clear monotonic relationship between not usually listen to classical music (n = 5000) is 100. The
IQ and preference for classical music in the American mean difference between the two groups of British teenagers
sample. GSS respondents who like classical music very is highly statistically significant (t(5659) = 11.8270,
much (n = 170) have a mean IQ of 107, those who like it p < 0.00001).
(n = 307) have a mean IQ of 103, those who have mixed
feelings about it (neither like it or dislike it) (n = 240) have a
mean IQ of 101, those who dislike it (n = 163) have a mean COGNITIVE COMPLEXITY OR
IQ of 95, and those who dislike it very much (n = 91) have a EVOLUTIONARY NOVELTY?
mean IQ of 93. The association between IQ and preference
for classical music is highly significant (F(4, 966) = 22.6970, One possible objection to our theory and empirical analyses
p < 0.00001, η2 = 0.0859). above is that our dimension of evolutionary novelty,
Similarly, Figure 2 presents the mean IQ of BCS86 captured by the distinction between instrumental and vocal
respondents by their preference for classical music. It shows music, is confounded with cognitive complexity of music.
that the mean IQ of those who “usually listen to classical For example, classical music, which is largely instrumental,

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
272 Journal of Behavioral Decision Making

is also cognitively complex; it is probably the most Table 4. Partial correlations between preferences for different
cognitively complex form of music in human history. On music genres and intelligence, American Sample (General Social
Survey, 1993)
the other extreme, rap music, which is largely vocal, often to
the exclusion of any discernible melodic structure, is also (1) (2)
cognitively very simple. So critics may argue that the Genre No controls With controls
association between intelligence and preference for instru-
Big band 2.1172** 1.7627**
mental music that we demonstrate above is really an (0.6437) (0.6194)
association between intelligence and cognitively complex 0.1602 0.1366
forms of music. Research on “need for cognition” (Cacioppo Blues or R&B 0.1512 0.2667
(0.6941) (0.6581)
& Petty, 1982) may lead one to expect such an association, 0.0108 0.0191
although it is not clear whether need for cognition is strongly Bluegrass 0.8014 1.3425
correlated with general intelligence (Bors, Vigneau, & (0.7253) (0.6898)
0.0566 0.0975
Lalande, 2006; Cacioppo, Petty, Feinstein, & Jarvis, 1996; Classical 1.8384** 1.0108
Fleischhauer et al., 2010). (0.6458) (0.6204)
In order properly to test this alternative hypothesis, we 0.1540 0.0860
Contemporary rock −0.0953 −0.3574
would ideally need a quantitative “cognitive complexity (0.6695) (0.6331)
score” for each genre of music, in the form (classical = 5, −0.0073 −0.0282
jazz = 4.5, etc.). Further, such “cognitive complexity scores” Country western −1.8759** −1.4840**
(0.5774) (0.5481)
would ideally have been validated and widely in use. We −0.1471 −0.1193
searched the literature extensively, and consulted several Folk 0.6612 −0.1889
experts in music perception, but have not been able to locate (0.6773) (0.6510)
0.0474 −0.0138
such “cognitive complexity scores” for different musical Gospel −2.4617**** −1.3751*
genres. They just do not seem to exist. Yet, most people (0.5766) (0.5936)
seem to understand and agree intuitively that, for example, −0.1855 −0.1052
Heavy metal −1.0502 −.1109
classical music and jazz are far more cognitively complex (0.5478) (0.5433)
than, say, rap music. In the absence of quantitative and −0.0861 −0.0091
validated “cognitive complexity scores,” we must rely on Jazz −0.1418 −0.2620
(0.6657) (0.6493)
such intuitive senses of cognitive complexity of musical −0.0107 −0.0199
genres. Latin 0.1557 0.1534
In our empirical exploration, we have chosen to inspect (0.6505) (0.6236)
0.0113 0.0113
the correlation between intelligence and preference for each Easy listening −1.4360* −1.6884**
genre of music. Because the GSS 1993 data contain a larger (0.6119) (0.5840)
number of musical genres than the BCS86 data (18 vs 12) −0.1060 −0.1280
Broadway musicals 0.8913 0.2417
and measure preference more precisely (five‐point ordinal (0.6444) (0.6448)
scale vs binary), we use the GSS 1993 data to test the 0.0673 0.0185
alternative hypothesis. Using the BCS86 data produces New age 0.6979 0.3228
(0.5711) (0.5381)
substantively identical conclusions, however. 0.0548 0.0258
A potential problem with inspecting the correlation Oldies 1.1830 1.4068*
between intelligence and preference for a specific musical (0.6670) (0.6402)
0.0834 0.1010
genre is that preferences for all musical genres are very highly Opera 0.8213 0.5876
correlated. It appears that there are people who like music, (0.6562) (0.6220)
and there are those who do not, and those who like music like 0.0628 0.0457
Rap −2.1328*** −1.0121
all types of music. For example, preference for classical (0.5839) (0.5602)
music is positively and significantly correlated with prefer- −0.1662 −0.0796
ence for both bluegrass and reggae music; in fact, it is Reggae 0.9920 0.6675
(0.6025) (0.5668)
significantly positively correlated with preference for 12 of 0.0808 0.0548
the 18 genres of music. A factor analysis shows that Constant 96.6085 66.6978
preferences for all 18 genres load on a single latent dimension (3.7340) (5.4398)
R2 0.2581 0.4100
with positive loadings. We therefore regress intelligence on n 564 517
preferences for all 18 musical genres simultaneously to
examine the partial correlation between intelligence and Note: Control variables included in Model (2) are age, race, sex, education,
family income, religion, whether currently married, whether ever married,
preference for a given genre while holding constant and number of children.
preferences for all other genres. Main entries are unstandardized regression coefficients.
Table 4, Column (1), presents the partial correlations Numbers in parentheses are standard errors.
Numbers in italics are standardized coefficients (betas).
between intelligence and preferences for all 18 musical * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001.
genres. As suspected, preference for classical music is
significantly positively correlated with intelligence. Howev- big band music is more cognitively complex than classical
er, preference for big band is even more strongly positively music. On the other extreme, as suspected, preference for rap
correlated with it. It would be difficult to make the case that music is significantly negatively correlated with intelligence.

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 273

However, preference for gospel music is even more strongly study (Kanazawa, 2010a) has shown that more intelligent
negatively correlated with it. It would be difficult to make individuals are more likely to be liberal and atheist, and more
the case that gospel is less cognitively complex than rap. intelligent men (but not women) are more likely to value
(We might point out that, with its close link to religious sexual exclusivity, than their less intelligent counterparts,
rituals, gospel is a particularly evolutionarily familiar form because these values are evolutionarily novel. In this paper,
of music [Mithen, 2005].) At the same time, preference for we have extended the Hypothesis to musical tastes and
opera, another highly cognitively complex form of music, is preferences.
not significantly correlated with intelligence. Its nonsignif- A recent theory of the evolution of music (Mithen, 2005)
icantly positive correlation is smaller than those for oldies, proposes that music and language may have a common
reggae, or Broadway musicals. It would be difficult to make precursor in a holistic, manipulative, multimodal, musical,
the case that oldies, reggae, or Broadway musicals are and mimetic system of communication. If both language and
cognitively more complex than opera. music evolved out of song‐like utterances with emotional
Table 4, Column (2), presents the model that also contents, as Mithen argues, then it follows that music in its
includes the same control variables included in the earlier evolutionary origin was always vocal and purely instrumen-
analysis of the GSS data (age, race, sex, education, family tal music is evolutionarily novel. If Mithen is right, then the
income, religion, whether currently married, whether ever Savanna‐IQ Interaction Hypothesis would suggest that more
married, and number of children). When these additional intelligent individuals are more likely to acquire a taste and
controls are included in the model, the partial positive preference for purely instrumental music, such as big band,
correlation between preference for classical music and classical, and easy listening, but general intelligence has no
intelligence is no longer statistically significant, whereas effect on the acquisition of taste for vocal music.
the partial correlation between preference for big band and Our analyses of an American sample (1993 General
intelligence remains statistically significantly positive. With Social Survey) and a British sample (1986 follow‐up of the
the additional controls, the partial correlation for oldies is British Cohort Study) show that more intelligent individuals
now statistically significantly positive. It would be very are indeed more likely to prefer purely or largely
difficult to make the case that oldies are cognitively more instrumental music than less intelligent individuals, whereas
complex than classical music. Rentfrow and Gosling (2003, intelligence has no effect on preference for purely or largely
p. 1241) categorize blues, jazz, classical, and folk music as vocal music. More intelligent British teenagers watch
“structurally complex,” but the results presented in Table 4, classical music TV programs more frequently than their
Column (2), show that none of these genres are significantly less intelligent classmates. Additional analyses provide little
correlated with intelligence. Two of the genres (folk and support for the alternative hypothesis that more intelligent
jazz) are nonsignificantly negatively correlated with intelli- individuals prefer cognitively more complex forms of music.
gence. All in all, the analysis presented in Table 4 provides It is important to point out that, in every regression model,
little support for the view that more intelligent individuals we control for the respondent’s social class and education (by
necessarily and uniformly prefer cognitively complex genres teacher‐rated academic performance in the case of the BCS86
of music. respondents, all of whom are still in school and thus have the
Given that we do not have quantitative and validated same number of years of formal schooling). Although
“cognitive complexity scores” for different musical genres, education has an independent partial effect on musical
our conclusion in this section must remain tentative. Once preference among both American and British respondents,
such “cognitively complex scores” are constructed and the significant effect of intelligence remains. In no case does
validated, we will need to revisit the issue of the association the respondent’s social class (measured by annual family
between general intelligence and preference for cognitively income and parents’ education) have a significant effect on
complex music to see if such an association indeed exists. musical tastes. The significant effect of intelligence is not
Our preliminary analysis suggests, however, that, unless it confounded by education or social class.
turns out that oldies and big band music are cognitively more There are alternative explanations for the association
complex than classical music and jazz, more intelligent between general intelligence and the preference for
individuals may not necessarily prefer cognitively complex instrumental music. For example, individuals may want to
music. signal their intelligence by publicly expressing their
preferences for such evolutionarily novel, instrumental
music as classical or jazz. However, such an explanation
DISCUSSION based on signaling does not explain why certain genres of
music (such as classical or jazz) have come to be
The Savanna‐IQ Interaction Hypothesis, derived from a associated with higher intelligence. Why do intelligent
logical conjunction of the Savanna Principle and a theory of individuals not signal their intelligence by publicly
the evolution of general intelligence, suggests that more expressing their preference for country western or gospel
intelligent individuals may be more likely to acquire and music? The Savanna‐IQ Interaction Hypothesis can simul-
espouse evolutionarily novel values and preferences than taneously explain the origin of the association between
less intelligent individuals, whereas general intelligence may intelligence and certain genres of music and suggest that
have no effect on the acquisition and espousal of there may be empirical basis for the “stereotype” that
evolutionarily familiar values and preferences. An earlier intelligent individuals listen to classical or jazz music. Once

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
274 Journal of Behavioral Decision Making

the association between intelligence and preference for Chamorro‐Premuzic, T., & Furnham, A. (2007). Personality and
instrumental music is widely known, however, costly music: Can traits explain how people use music in everyday
life? British Journal of Psychology, 98, 175–185.
signaling may explain individuals’ motivation to cultivate Crawford, C. B. (1993). The future of sociobiology: Counting babies
such preference as a signal for their intelligence. or proximate mechanisms? Trends in Ecology & Evolution, 8,
It is important to note that, although the theory of the 183–186.
evolution of general intelligence (Kanazawa, 2004b, 2008) Deary, I. J. G., Batty, G. D., & Gale, C. R. (2008). Bright children
proposes that general intelligence originally evolved to solve become enlightened adults. Psychological Science, 19, 1–6.
Ellis, L. (1998). Neodarwinian theories of violent criminality and
evolutionarily novel and nonrecurrent adaptive problems, antisocial behavior: Photographic evidence from nonhuman
the Savanna‐IQ Interaction Hypothesis (Kanazawa, 2010a) animals and a review of the literature. Aggression and Violent
does not suggest that evolutionarily novel preferences and Behavior, 3, 61–110.
values that more intelligent individuals are more likely to Everett, D. L. (2005). Cultural constraints on grammar and
acquire and espouse are necessarily adaptive and increase cognition in Pirahã: Another look at the design features of
human language. Current Anthropology, 46, 621–646.
their reproductive success in the current environment. It is
Fleischhauer, M., Enge, S., Brocke, B., Ullrich, J., Strobel, A., &
not obvious how being a left‐wing liberal or atheist Strobel, A. 2010. Same or different? Clarifying the relationship
(Kanazawa, 2010a) or being nocturnal (Kanazawa & Perina, of need for cognition to personality and intelligence. Personality
2009) increases reproductive success today. And some of and Social Psychology Bulletin, 36, 82–96.
the evolutionarily novel preferences that more intelligent Guthrie, S. E. (1993). Faces in the clouds: A new theory of religion.
individuals are more likely to acquire and espouse, such as New York: Oxford University Press.
Harvey, P. H., & Bennett, P. M. (1985). Sexual dimorphism and
the consumption of alcohol, tobacco, and psychoactive reproductive strategies. In J. Ghesquiere, R. D. Martin, & F.
drugs (Kanazawa & Hellberg, 2010), are manifestly Newcombe (Eds.), Human sexual dimorphism (pp. 43–59).
detrimental to health and survival. The Savanna‐IQ London: Taylor and Francis.
Interaction Hypothesis does not predict that more intelligent Haselton, M. G., & Nettle, D. (2006). The paranoid optimist: An
individuals are more likely to acquire and espouse healthy integrative evolutionary model of cognitive biases. Personality
and Social Psychology Review, 10, 47–66.
and adaptive preferences and values, only evolutionarily
Hechter, M., Ranger‐Moore, J., Jasso, G., & Horne, C. (1999). Do
novel ones. values matter? An analysis of advanced directives for medical
treatment. European Sociological Review, 15, 405–430.
Herrnstein, R. J., & Murray, C. (1994). The bell curve: Intelligence
ACKNOWLEDGEMENTS and class structure in American life. New York: Free Press.
Huang, M.‐H., & Hauser, R. M. (1998). Trends in black‐white test‐
score differentials: II. The WORDSUM vocabulary test. In U.
We thank Jay Belsky, Tomas Chamorro‐Premuzic, Daniel Neisser (Ed.), The rising curve: Long‐term gains in IQ and
Levitin, Nando Pelusi, Brent T. Simpson, and David Temperley related measure (pp. 303–332). Washington, D.C.: American
for their comments on earlier drafts and other contributions. Psychological Association.
Jackendoff, R. (2000). Foundations of language: Brain, meaning,
grammar, evolution. Oxford: Oxford University Press.
Kalmus, H., & Fry, D. B. (1980). On tune deafness (dysmelodia):
REFERENCES Frequency, development, genetics and musical background.
Annals of Human Genetics, 43, 369–382.
Alexander, R. D., Hoogland, J. L., Howard, R. D., Noonan, K. M., Kanazawa, S. (2001). De gustibus est disputandum. Social Forces,
& Sherman, P. W. (1979). Sexual dimorphisms and breeding 79, 1131–1163.
systems in pinnipeds, ungulates, primates and humans. In N. A. Kanazawa, S. (2002). Bowling with our imaginary friends.
Chagnon, & W. Irons (Eds.), Evolutionary biology and human Evolution and Human Behavior, 23, 167–171.
social behavior: An anthropological perspective (pp. 402–435). Kanazawa, S. (2004a). The Savanna Principle. Managerial and
North Scituate: Duxbury Press. Decision Economics, 25, 41–54.
Ariely, D., & Loewenstein, G. (2006). The heat of the moment: The Kanazawa, S. (2004b). General intelligence as a domain‐specific
effect of sexual arousal on sexual decision making. Journal of adaptation. Psychological Review, 111, 512–523.
Behavioral Decision Making, 19, 87–98. Kanazawa, S. (2006). Why the less intelligent may enjoy television
Atran, S. (2002). In gods we trust: The evolutionary landscape of more than the more intelligent. Journal of Cultural and
religion. Oxford: Oxford University Press. Evolutionary Psychology, 4, 27–36.
Bickerton, D. (1990). Language and species. Chicago: University Kanazawa, S. (2008). Temperature and evolutionary novelty as
of Chicago Press. forces behind the evolution of general intelligence. Intelligence,
Bors, D. A., Vigneau, F., & Lalande, F. (2006). Measuring the need 36, 99–108.
for cognition: Item polarity, dimensionality, and the relation with Kanazawa, S. (2009). Evolutionary psychology and crime. In
ability. Personality and Individual Differences, 40, 819–828. Walsh, A., & Beaver, K. M. (Eds.), Biosocial criminology: New
Boyer, P. (2001). Religion explained: The evolutionary origins of directions in theory and research (pp. 90–110). New York:
religious thought. New York: Basic. Routledge.
Brown, S. (2000). The ‘musilanguage’ model of music evolution. Kanazawa, S. (2010a). Why liberals and atheists are more
In N. L. Wallin, B. Merker, & S. Brown (Eds.), The origins of intelligent. Social Psychology Quarterly, 73, 33–57.
music (pp. 271–300). Cambridge: MIT Press. Kanazawa, S. (2010b). Evolutionary psychology and intelligence
Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. research. The American Psychologist, 65, 279–289.
Journal of Personality and Social Psychology, 42, 116–131. Kanazawa, S., & Hellberg, J. E. E. U. (2010). Intelligence and
Cacioppo, J. T., Petty, R. E., Feinstein, J. A., & Jarvis, W. B. G. substance use. Review of General Psychology, 14, 382–396.
(1996). Dispositional differences in cognitive motivation: The Kanazawa, S., & Novak, D. L. (2005). Human sexual dimorphism
life and times of individuals varying in need for cognition. in size may be triggered by environmental cues. Journal of
Psychological Bulletin, 119, 197–253. Biosocial Science, 37, 657–665.

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm
S. Kanazawa and K. Perina Why Intelligent People Like Classical Music 275

Kanazawa, S., & Perina, K. (2009). Why night owls are more Symons, D. (1990). Adaptiveness and adaptation. Ethology and
intelligent. Personality and Individual Differences, 47, 685–690. Sociobiology, 11, 427–444.
Kirkpatrick, L. A. (2005). Attachment, evolution, and the Tooby, J., & Cosmides, L. (1990). The past explains the present:
psychology of religion. New York: Guilford. Emotional adaptations and the structure of ancestral environ-
Kluegel, J. R., & Smith, E. R. (1986). Beliefs about inequality: ments. Ethology and Sociobiology, 11, 375–424.
Americans’ view of what is and what ought to be. New York: de Waal, F. B. M. (1989). Food sharing and reciprocal obligations
Aldine. among chimpanzees. Journal of Human Evolution, 18, 433–459.
Lake, C. C., & Breglio, V. J. (1992). Different voices, different de Waal, F. B. M. (1992). Appeasement, celebration, and food
views: The politics of gender. In P. Ries, & A. J. Stone (Eds.), sharing in the two Pan species. In T. Nishida, W. C. McGrew,
The American woman, 1992–93: A status report (pp. 178–201). & P. Marler (Eds.), Topics in primatology: Human origins (pp.
New York: Norton. 37–50). Tokyo: University of Tokyo Press.
Leinonen, L., Linnankoski, I., Laakso, M.‐L., & Aulanko, R. de Waal, F. B. M., Luttrell, L. M., & Canfield, M. E. (1993).
(1991). Vocal communication between species: Man and Preliminary data on voluntary food sharing in brown capuchin
macaque. Language & Communication, 11, 241–262. monkeys. American Journal of Primatology, 29, 73–78.
Leinonen, L., Laakso, M.‐L., Carlson, S., & Linnankoski, I. (2003). Wilson, J. Q., & Herrnstein, R. J. (1985). Crime and human nature:
Shared means and meanings in vocal expression of man and The definitive study of the causes of crime. New York:
macaque. Logopedics, Phoniatrics, Vocology, 28, 53–61. Touchstone.
Leutenegger, W., & Kelly, J. T. (1977). Relationship of sexual Wirls, D. (1986). Reinterpreting the gender gap. Public Opinion
dimorphism in canine size and body size to social, behavioral, Quarterly, 50, 316–330.
and ecological correlates in anthropoid primates. Primates, 18, Wolfle, L. M. (1980). The enduring effects of education on verbal
117–136. skills. Sociology of Education, 53, 104–114.
Linnankoski, I., Laakso, M., Aulanko, R., & Leinonen, L. (1994). Wray, A. (1998). Protolanguage as a holistic system of social
Recognition of emotions in macaque vocalizations by children interaction. Language & Communication, 18, 47–67.
and adults. Language & Communication, 14, 183–192. Wray, A. (2006). Joining the dots: The evolutionary picture of
McCullagh, P. (1980). Regression models for ordinal data. Journal language and music. Cambridge Archaeological Journal, 16,
of the Royal Statistical Society, Series B, 42, 109–142. 103–105.
Miner, J. B. (1957). Intelligence in the United States: A survey— Yates, J. F., & Tschirhart, M. D. (2006). Decision making expertise.
with conclusions for manpower utilization in education and In K. A. Ericsson, N. Charness, P. J. Feltovich, & R. R. Hoffman
employment. New York: Springer. (Eds.), Cambridge handbook of expertise and expert perfor-
Miller, A. S., & Hoffmann, J. P. (1995). Risk and religion: An mance (pp. 421–438). New York: Cambridge University Press.
explanation of gender differences in religiosity. Journal for the Zuberbühler, K. (2002). A syntactic rule in forest monkey
Scientific Study of Religion, 34, 63–75. communication. Animal Behaviour, 63, 293–299.
Miller, A. S., & Stark, R. (2002). Gender and religiousness: Can Zuberbühler, K. (2003). Natural semanticity in wild primates. In F.
socialization explanations be saved? The American Journal of B. M. de Waal, & P. L. Tyack (Eds.), Animal social complexity:
Sociology, 107, 1399–1423. Intelligence, culture, and individualized societies (pp. 362–367).
Mithen, S. (2005). The singing Neanderthals: The origins of music, Cambridge: Harvard University Press.
language, mind and body. London: Weidenfeld & Nicholson.
Nettl, B. (1983). The study of ethnomusicology: Twenty‐nine issues
and concepts. Urbana: University of Illinois Press.
O’Brien, R. M. (2007). A caution regarding rules of thumb for Authors’ biographies:
variance inflation factors. Quality & Quantity, 41, 673–690. Satoshi Kanazawa is an evolutionary psychologist and Reader in
Pickford, M. (1986). On the origins of body size dimorphism in Management at the London School of Economics and Political
primates. In M. Pickford, & B. Chiarelli (Eds.), Sexual Science. He is the author of Why Beautiful People Have More
dimorphism in living and fossil primates (pp. 77–91). Florence: Daughters (Penguin, 2007) and The Intelligence Paradox: Why
Il Sedicesimo. Intelligent People Do Unnatural Things (Wiley, 2012).
Rentfrow, P. J., & Gosling, S. D. (2003). The do re mi’s of
everyday life: The structure and personality correlates of music Kaja Perina is the editor in chief of Psychology Today.
preference. Journal of Personality and Social Psychology, 84,
1236–1256.
Shapiro, R. Y., & Mahajan, H. (1986). Gender differences in policy Authors’ addresses:
preferences: A summary of trends from the 1960s to the 1980s. Satoshi Kanazawa, Department of Management, London School
Public Opinion Quarterly, 50, 42–61. of Economics and Political Science, UK.
Sundquist, J. L. (1983). Dynamics of the party system (rev. edn).
Washington, D.C.: Brookings Institution. Kaja Perina, Psychology Today, NY, USA.

Copyright © 2011 John Wiley & Sons, Ltd. J. Behav. Dec. Making, 25: 264–275 (2012)
DOI: 10.1002/bdm