Adapting Minds Evolutionary Psychology and The Persistent Quest For Human Nature
Adapting Minds Evolutionary Psychology and The Persistent Quest For Human Nature
Adapting Minds Evolutionary Psychology and The Persistent Quest For Human Nature
Adapting Minds
David J. Buller
A Bradford Book
The MIT Press
Cambridge, Massachusetts
London, England
© 2005 Massachusetts Institute of Technology
All rights reserved. No part of this book may be reproduced in any form by any elec-
tronic or mechanical means (including photocopying, recording, or information
storage and retrieval) without permission in writing from the publisher.
MIT Press books may be purchased at special quantity discounts for business or sales
promotional use. For information, please e-mail [email protected] or
write to Special Sales Department, The MIT Press, 5 Cambridge Center, Cambridge,
MA 02142.
This book was set in Stone sans and Stone serif by SNP Best-set Typesetter Ltd., Hong
Kong.
Printed and bound in the United States of America.
BF701.B85 2005
155.7–dc22
2004057932
10 9 8 7 6 5 4 3 2 1
To Heide,
for giving me my life,
but mostly for giving me Dabi
Contents
Acknowledgments ix
Introduction 1
1 Evolution 17
The Nature of Evolution 17
The Causes of Evolution 26
Adaptation 31
Phenotypic Variation 37
2 Mind 49
The Evolution of Behavior 50
The Adapted Mind 53
The Environment of Evolutionary Adaptedness 58
Modularity and the Adapted Mind 63
“Human Nature” 70
3 Adaptation 83
Adaptationism? 83
Adaptation Hunting 92
“Our Modern Skulls House a Stone-Age Mind” 107
Psychological Differences and Our “Common Nature” 112
4 Modularity 127
(with Valerie Gray Hardcastle)
Evolutionary Psychology, Meet Developmental Neurobiology 127
The Arguments for Modularity Reconsidered 144
The Empirical Evidence for Modularity Reconsidered 160
On Domain-General Mechanisms 195
viii Contents
5 Mating 201
“The Evolution of Desire” 202
Men Seeking Women 210
Women Seeking Men 228
Concluding Skeptical Remarks 252
6 Marriage 259
What Is Marriage? 259
Keeping Myself Only unto You 277
The Green-Eyed Monster 301
7 Parenthood 347
Discriminative Parental Solicitude 347
What Is “The Truth about Cinderella”? 370
Trying to Understand Parents 410
Epilogue 481
Notes 483
Bibliography 497
Index 527
Acknowledgments
The completion of this book came at the end of a long road that was fre-
quently difficult to navigate, and along the way I received a lot of direc-
tion. I am deeply indebted to all who provided that direction, and to the
extent that I am able in this forum I would like to repay those debts with
some public displays of gratitude.
First, I have general intellectual debts that predate this book project but
that significantly conditioned my approach to it. One of these is to Arthur
Fine, my dissertation advisor of many years ago, well before I became inter-
ested in evolutionary biology and evolutionary psychology. Arthur taught
me a great deal about the relation between science and philosophy, most
of which didn’t truly sink in until years after he ceased being my teacher
and dissertation advisor. But his lessons are evident (at least to me) in all
that follows. When working on my dissertation, I remember expressing
concern to Arthur that I wasn’t really doing philosophy in my dissertation.
He responded with his usual warm-hearted laugh and the gentle comment,
“Don’t worry about that.” Well, Arthur, as you’ll see here, I’ve certainly
stopped worrying about that! The other general intellectual debt is to
David Hull, who succeeded in teaching me a lot about both biology and
the nature of philosophy in the years since graduate school, despite what
must have seemed my stubborn reluctance to learn. As with Arthur’s
lessons, many of David’s often didn’t take until well after the fact, much
to his occasional frustration, I am sure. David also provided detailed and
thoughtful comments on the whole manuscript for this book, and I am
very grateful for his time and thoughtfulness in providing them.
Another general debt, of a very different sort, is to Philip Kitcher for his
1985 book Vaulting Ambition: Sociobiology and the Quest for Human Nature,
which subjected the sociobiology of the day to scathing criticism. For, in
certain respects, Kitcher’s book was a conceptual model for mine—a fact to
which this book’s subtitle gives the nod. I say “in certain respects” for two
x Acknowledgments
Smith, Todd Grantham, Kristen Hawkes, Wade Hazel, Joel Milner, Brad
Sagarin, Terry Sullivan, Denis Walsh, David Sloan Wilson, and anonymous
referees for the National Science Foundation. In addition, I am thankful
to all the students in my seminars over the years, and to the students in
Brad Sagarin’s evolutionary psychology seminar, for providing feedback
on the evolving ideas in this book. One former student, Wendy Parker,
deserves special thanks for particularly incisive feedback in the formative
stages of the project.
As you will see, in order to test some hypotheses in chapter 7, I had to
conduct my own empirical study of patterns of child abuse in the United
States. This wouldn’t have been possible without the efforts of my collab-
orator in this endeavor, Elliott G. Smith, the Associate Director of the
National Data Archive on Child Abuse and Neglect (NDACAN). Elliott was
extremely generous and helpful, over several months, and I learned a great
deal from working with him. On one level it was challenging and fun, but
on another it was deeply depressing. I acquired a lot of respect for those
who work full time, year after year, researching the horrible suffering that
is inflicted upon children by those whom they love and depend on the
most. This research is not easy to stomach, and those who subject them-
selves to it do so in an effort to improve our understanding of, and hope-
fully one day to mitigate, child abuse. They deserve our undying
admiration.
My greatest debt is to my wife, Heide Fehrenbach, who provided me with
loving support and encouragement throughout the long duration of this
project, despite working at the same time on a book of her own. Heide also
provided excellent editorial comments on parts of the manuscript. Finally,
a list of my debts would not be complete without acknowledging our son,
“Dabi,” who brought some perspective to my life and taught me that
writing a book is comparatively unimportant.
Adapting Minds
Introduction
Several years ago, I spent a semester’s leave in London, put up in a flat off
Kensington Gardens on someone else’s dime. During that semester, I was
fortunate enough to spend my days reading evolutionary biology, pursu-
ing applications of evolutionary theory to problems in contemporary phi-
losophy of mind, and my nights in outstanding pubs or at home watching
surprisingly superb British television. (It was always easier to find an inter-
esting program on the four channels we received in our London flat than
on one hundred American channels.)
One night at home, I happened into the end of an immediately grip-
ping program on the BBC. In a scene I was convinced I would not have
found on American television, a naked heterosexual couple was engaged
in foreplay. Within moments, there was a jump cut to another scene, and
I found myself taking a ride that Albert Einstein never imagined. In one
of his famous thought experiments, you may recall, Einstein asked us to
imagine how events would appear to an observer riding a beam of light.
That’s all very interesting, I suppose, but to someone with an abiding and
not-strictly-professional interest in human sexuality, it simply doesn’t
compare to the ride I, as a suddenly devoted BBC viewer, was then taking.
For a camera had been strapped to the male’s erect penis, and I was riding
the male’s penis and viewing the copulatory act from within the female’s
surprisingly well lit vagina.
After a number of penile thrusts and vaginal contractions, the inevitable
ejaculation came. Moments later, I (and no doubt countless other enrap-
tured viewers in Britain) witnessed the female’s cervix take several “dips”
into the semen that had pooled at the back of her vagina. The narrator—
whom I later discovered was Desmond Morris—explained that the cervix
was “sucking” the semen up into the uterus in order to “increase the
chances of fertilization.”1 We were then informed that a female can
increase the chances of being fertilized by a particular partner by “varying
2 Introduction
behavior, which promised to reveal many more truths about human psy-
chology than abstruse philosophy of mind ever could. These applications
of biology to human behavior promised to explain why we desire sex with
some people but not others, why we marry the people we do, why we are
sometimes unfaithful, why the number of sperm in a male’s ejaculate varies
with the amount of time his partner is out of sight, why we care so deeply
for our children, why conflicts arise in all our intimate relationships, and
why “men are from Mars and women are from Venus.” And the best part
was that many of the applications of biology to human behavior concerned
human sexuality. At that moment, my research agenda was redefined. I
immediately abandoned my suddenly boring research into whether evo-
lutionary theory could explain the representational properties of mental
states in order to explore instead what was becoming known as “evolu-
tionary psychology.”
Once I began to focus on evolutionary psychology, I seemed to
encounter it everywhere I turned, especially in the popular media. During
almost every wait in the supermarket checkout line, I would find reference
to the evolutionary psychology of human mating on the covers of women’s
and men’s magazines. On Sundays, I would often find articles in the New
York Times Magazine or the Chicago Tribune Magazine about the evolution-
ary psychology of mating, parent-child relationships, or status seeking.
And it seemed to be all over television, and not just on “high-brow” chan-
nels like PBS and the Discovery Channel. On a Saturday night, ABC aired
an ABC News Special Report by John Stossel examining the evolutionary
psychology of sex differences and their implications for early education. It
appeared that evolutionary psychology was capturing the public con-
sciousness and beginning to condition the human self-conception pre-
sented in popular culture.
Initially, I was completely captivated by evolutionary psychology, and I
was certain that it was providing a deep and accurate understanding of
human mentality and behavior. But after six months’ research, it was
unclear to me how everything that went by the name “evolutionary psy-
chology” fit together, and I began having serious doubts about many of
the confident claims made by evolutionary psychologists (such as Morris’s
claim that “research has proved” that the majority of sperm in an ejaculate
function as sperm warriors). A year’s research later, it was clear to me that
there were distinctly different lines of research being conducted under the
“evolutionary psychology” label, and I became convinced that the line of
research that had garnered the most attention, both within academia and
throughout the popular media, was wrong in almost every detail. This
4 Introduction
with.”7 Too much of the response to Gould then focused on his creden-
tials as an evolutionary biologist instead of on the actual content of his
arguments.
But the most insidious rhetoric employed by evolutionary psychologists
in responding to their critics has involved the claim, often implied, that
there are no biological grounds on which to criticize evolutionary psy-
chology. Evolutionary psychologists have often defended some very spe-
cific hypothesis about human mentality or behavior by arguing that
rejection of that hypothesis requires rejecting the very idea that human
psychology has evolved along with everything else on the planet. If we are
to accept that humans have evolved, the defense often goes, then we have
to accept all of the “evolutionary” arguments offered by evolutionary psy-
chologists and, consequently, all of the specific doctrines that derive from
those arguments. As a result, very specific claims are presented as entailed
by an evolutionary view of humanity. The view that males prefer nubile
females and females prefer high-status males, for example, has been pre-
sented as “the evolutionary view” of human mating. Similarly, the view
that there are evolved sex differences in the intensity of sexual jealousy
has been called “the evolutionary theory of jealousy” by proponents and
critics alike, as though taking an evolutionary view of jealousy entails the
specific theory that there are evolved sex differences in sexual jealousy.
Accordingly, the rhetoric sets up the following dichotomy: Either you
accept biology, in which case you must accept the claims of evolutionary
psychologists, or you don’t. Critics have thus been portrayed as necessar-
ily committed to scientifically empty theories from the social sciences, to
some form of postmodernist relativism, or to creationism. No one who
truly accepts evolution, the rhetoric goes, can seriously question any of the
specific claims of evolutionary psychology.
I have found this lack of civilized, reasoned dialogue to be deeply lam-
entable. On the one side, evolutionary psychologists have not given their
critics their due. In a (perhaps understandable) defensive posture toward
their critics, they have too often simply circled the wagons and become
entrenched in dogmatic views, rather than revising their views in light of
genuine problems. This is not the path of scientific progress. As the
philosopher of science Karl Popper put it, scientific progress consists of iter-
ated rounds of “conjecture and refutation.” To dogmatically reject all “refu-
tations,” by deflecting attention from the content of the criticism and
toward the motives of the critic, is to hinder the growth of scientific knowl-
edge. On the other side, critics have not given evolutionary psychology
its due. Evolutionary psychology is a bold and innovative approach to
Introduction 7
“debate” failed to be clear about how they defined the term “evolutionary
psychology.” In order to forestall this kind of misunderstanding, and in
order to be crystal clear about the target of my critique in this book, let’s
get clear about what evolutionary psychology is.
The term “evolutionary psychology” is sometimes used simply as a
shorthand for “the evolutionary study of mind and behavior” or as a short-
hand for theories “adopting an evolutionary perspective on human behav-
ior and psychology.”8 When used in these ways, “evolutionary psychology”
designates a field of inquiry, which is so broad as to cover work ranging
from studies of foraging and birth spacing in traditional hunter-gatherer
societies to studies of encephalization (the progressive increase in brain
size relative to body size in the human lineage) and the evolution of altru-
ism and language. If one examines all of the work that adopts “an evolu-
tionary perspective on human behavior and psychology,” one will find that
it varies significantly in fundamental theoretical and methodological com-
mitments. Indeed, such work is united only by a commitment to articu-
lating questions about human behavior and mentality, and articulating
hypothetical answers to those questions, with conceptual and theoretical
tools drawn from evolutionary theory. This is why I say that all this work
forms a field of inquiry. For fields of inquiry are defined not by specific sets
of doctrines, but by sets of related questions. Fields of inquiry are defined
not by specific answers to questions, but by the importance they place on
particular kinds of question.
Many researchers in the field of evolutionary psychology often deliber-
ately resist the “evolutionary psychology” label, however, preferring to
classify their work as, for example, human ethology, human behavioral
ecology, or evolutionary anthropology. The reason is that the term “evo-
lutionary psychology” has increasingly come to be used to designate only
work conducted within a specific set of theoretical and methodological
commitments shared and articulated by a prominent and influential group
of researchers, and many researchers outside this group often wish to dis-
tance themselves from it by rejecting the “evolutionary psychology” label.
The most notable members of this influential group of researchers are the
psychologists David Buss, Leda Cosmides, Martin Daly, Steven Pinker, and
Margo Wilson, and the anthropologists Donald Symons and John Tooby.
This group is united in the belief that adoption of an evolutionary per-
spective on human psychology immediately entails a number of very spe-
cific theoretical and methodological doctrines. These theoretical and
methodological doctrines will be elaborated in detail in chapter 2, but a
brief overview of them is useful here.
Introduction 9
that resemble mental organs. But these structures, I will argue, are not bio-
logical adaptations; they are flexible responses to the experiential condi-
tions of an individual’s life.
Chapters 3 and 4 will thus provide most of the meaning of the book’s
title, Adapting Minds, which is a riposte to the title of Evolutionary Psy-
chology’s manifesto, The Adapted Mind. For these chapters will argue that
human populations are characterized by evolved psychological variation,
multiple minds, rather than a single “mind” that is universal within human
populations. And, rather than being adapted to our Pleistocene past, these
chapters will argue that human minds are continually adapting in at least
two senses. First, at the population level, human minds are continuously
adapting to changing environments over evolutionary time; there contin-
ues to be evolution in human psychological characteristics. And, second,
at the individual level, a human mind continually adapts to changing envi-
ronments over the course of an individual’s lifetime; it is not a static struc-
ture preprogrammed with the ways in which it will respond to different
experiences it may encounter.
Some of the ideas I will criticize in these chapters have been criticized
by others. The Norwegian Evolutionary Psychologist Leif Kennair has
responded to these criticisms by claiming that they are rather empty.
Kennair argues that Evolutionary Psychology should be evaluated on the
basis of the empirical results it has produced, rather than criticized on
purely abstract theoretical grounds. In particular, Kennair cites the three
exemplars of Evolutionary Psychology discussed above: Cosmides and
Tooby’s work on “cheater detection,” Buss’s work on mate preferences, and
Daly and Wilson’s work on “discriminative parental solicitude.” The the-
oretical principles of Evolutionary Psychology can’t be wrong, despite
appearances, Kennair argues, if they have led to such impressive results.
Thus, he concludes, “I claim that these [exemplars] are what need to
be addressed critically in any attempt at disqualifying Evolutionary
Psychology.”10
Throughout the central chapters of this book I will take up Kennair’s
challenge, providing detailed critiques of the exemplars he cites as well as
some others. Cosmides and Tooby’s work on “cheater detection” will be
discussed in chapter 4. Chapter 5, Mating, will analyze the research that
is believed to support the claims that males have an evolved mate prefer-
ence for nubile females and that females have an evolved mate preference
for high-status males. Chapter 6, Marriage, will then examine claims that
Evolutionary Psychologists have made about human marriage and the psy-
Introduction 15
For Darwin, and for several generations of biologists after him, evolution
was conceived of as descent with modification. Each component of this def-
inition, descent and modification, requires some comment. Consider first
descent.
18 Chapter 1
eration but that make their appearance in some members of the offspring
generation, who can then transmit that characteristic to their progeny. The
source of these evolutionary novelties was also not successfully explained
by Darwin’s theory.
As it turned out, the development of genetics in the twentieth century
illuminated both of these issues. It was discovered, first of all, that offspring
inherit their parents’ characteristics because parents transmit their genes to
their offspring in the process of reproduction, and genes causally influence
the phenotypes—the anatomical structures, physiological states, or behav-
ioral forms—that organisms exhibit. Thus, offspring resemble their parents
because the genes that causally influenced parental phenotypes are directly
transmitted to offspring, in whom those same genes causally influence the
development of the same phenotypes. Genes were consequently recog-
nized as playing a dual role in evolution: They are the units of heredity,
which get directly transmitted from parents to offspring in reproduction,
and they guide the development of organisms in ways that influence the
phenotypes they possess. It was also discovered, however, that an organ-
ism’s phenotype does not affect the genes it can transmit to its offspring.
As a result, no modifications to an organism’s phenotype during the course
of its life affect the genes its offspring possess. So, genes were seen as the
locus of two causal arrows, one running from the genes of an organism to
its phenotypes and the other running from the genes of an organism to
the genes of its offspring. But there is no causal arrow running from the
phenotypes of an organism to the genes of its offspring. No matter how
much body-building you do in your life, your babies won’t be any stronger
than they would be if you were a couch potato. Similarly, breaking your
arm will not affect the bones of your offspring.
Developments in genetics also led to the discovery that, in the process
of reproduction, genes sometimes mutate into new forms. Consequently,
the evolutionarily novel phenotypes that occasionally appear in a lineage
are the result of mutated genes, which produce novel phenotypes in the
individuals with those genes. Once a mutated gene appears in a popula-
tion, it can be transmitted to the offspring of organisms with that gene,
and the novel phenotype it produces can be transmitted along with it.
Since genes are the key to both inheritance and the appearance of evo-
lutionary novelties, they came to be seen as central to the process of
descent with modification. Indeed, since phenotypes are produced by
genes, and phenotypes have no effect on the genes available to be trans-
mitted across generations, genes came to be seen as the very locus at which
evolution occurs. The discoveries of modern genetics thus gave rise to a
20 Chapter 1
which there are two different alleles, designated A and a, that can occur
at some locus in a population. Then a pair of these alleles can be a pair of
identical alleles (the pair AA or the pair aa) or a pair of different alleles (the
pair Aa). If an organism has the same allele in each opposing slot, if it has
the AA or aa genotype, it is a homozygote; and if it has different alleles in
the opposing slots, if it has the Aa genotype, it is a heterozygote.
In addition to diploid cells human bodies contain some haploid cells, the
nuclei of which contain 23 single, unpaired chromosomes. These cells,
called gametes, are formed by a process called meiosis. In meiosis a diploid
cell first undergoes a process of DNA replication, which generates another
copy of each chromosome contained in the nucleus. This is followed by
two rounds of cell division, in which the chromosomes separate from one
another and divide into four haploid cells. To make this less abstract, con-
sider the process of meiosis with respect to a single locus containing the
Aa genotype. Meiosis is a process whereby that single diploid Aa cell repli-
cates and divides to produce two haploid cells containing A and two
haploid cells containing a. Consequently, the result of meiosis is that an
organism’s DNA gets split in half: Half of the genes in a diploid cell take
up residence in one haploid cell and the other half take up residence in a
different haploid cell. Thus, while all an organism’s diploid cells are genet-
ically identical (with the exception of cells in which there has been a muta-
tion), its gametes are routinely genetically different from one another.
The gametes produced in meiosis are important in the process of repro-
duction, since they form the egg cells in females and the sperm cells in
males. During fertilization an egg cell and sperm cell fuse to form a new
diploid cell, called a zygote, from which a new organism develops. Repro-
duction is thus a process whereby each of two parents contributes a
gamete, which contains half of the parent’s genes, to the formation of a
diploid cell that will develop into an organism of the next generation. Half
of the genes in the diploid cells that form that newly developing organ-
ism’s body will thus have come from its mother’s egg (which contains half
of the mother’s genes) and the other half from its father’s sperm (which
contains half of the father’s genes).
Now consider how zygote genotypes are determined. For simplicity, con-
sider again a single locus at which the three genotypes AA, aa, and Aa occur
in a population. And suppose that mating in this population is random—
that is, there is no overall statistical tendency for like genotypes to mate
with one another. If two AA organisms reproduce, each will contribute only
A gametes, which will fuse to form AA zygotes; so all the offspring of two
AA organisms will also be AA. Similarly, all offspring of two aa organisms
22 Chapter 1
the Aa genotype, there are two alleles, A and a, that occur at the locus that
interests us. Further, since every organism is Aa, half the alleles that occur
at that locus are A and half are a; in other words, the frequency of A is 50
percent and the frequency of a is 50 percent. Now, although unlikely, it is
possible for each couple to produce one AA and one Aa offspring. In that
case, in the offspring generation the frequency of the A allele will increase
to 75 percent (since three out of every four slots in the diploid locus are
occupied by A) and the frequency of the a allele will decrease to 25 percent.
Under the modern genetic definition of evolution, this constitutes signif-
icant evolution.
While this example of a change in gene frequencies also involves a
change in genotype frequencies (since the population evolved from 100
percent Aa to 50 percent AA and 50 percent Aa), it is possible for there to
be a change in genotype frequencies across generations without a corre-
sponding change in gene frequencies. To see how, suppose there is a pop-
ulation consisting of just eight organisms—three AA, three aa, and two Aa
organisms. The alleles A and a each occur with a frequency of 50 percent
in this generation of the population. (There are six copies of A from the
three AA organisms and two copies of A from the two Aa organisms; and
there are six copies of a from the three aa organisms and two copies from
the two Aa organisms. There are thus eight copies each of A and a, out of
a total of sixteen alleles at the locus.) Now suppose that two AA organisms
mate with each other as do two aa organisms, and one Aa organism mates
with the remaining AA organism while the other Aa organism mates with
the remaining aa organism. Suppose further that each of these pairs pro-
duces just two offspring, so the next generation of the population also
contains just eight organisms. We know that the offspring from the
homozygote matings will be homozygotes of the same genotype and that
there is a 50 percent chance that the offspring from the heterozygote-
homozygote matings will also be homozygotes. Suppose that in fact the
heterozygote-homozygote pairs produce only homozygotes. Then the next
generation of the population will consist of four AA organisms (two from
the AA-AA mating and two from the AA-Aa mating) and four aa organisms
(two from the aa-aa mating and two from the Aa-aa mating). Although
the gene frequencies have not changed, both A and a remaining at 50
percent, the genotype frequencies have. For whereas 37.5 percent of
the parental generation was AA, 37.5 percent aa, and 25 percent Aa, the
offspring generation is 50 percent AA and 50 percent aa. Under the
modern genetic definition of evolution, this also constitutes significant
evolution.
24 Chapter 1
The kinds of evolution we have just considered can also produce changes
across generations in the frequencies of phenotypes in a population. The
reason is that genes regulate the synthesis of proteins, the stuff of which
our bodies are made, and differences between bodies or between parts of
the same body are a product of differences in the proteins of which they
are made. By regulating protein synthesis, genes consequently guide the
development of organisms, and this influences the phenotypes that organ-
isms possess. When a gene influences a particular phenotype in this way,
biologists say it is a gene for that phenotype. Thus, a change in gene or
genotype frequencies can produce a change in the frequencies of the phe-
notypes influenced by those genes or genotypes.
But why do I say that genes “influence” phenotypes, rather than saying
that genes “determine” phenotypes? This is because, by themselves, genes
don’t determine anything. One can’t simply put some carefully selected
genes in a petri dish, for example, and grow a cute little button nose. For
how a gene affects the phenotype of an organism depends on precisely
when (or if ) it is switched on and off in the process of development, and
that in turn depends on the properties of the gene’s environment. The
environment of a gene includes not only the environment outside the
organism (which affects the surface of the organism), but also the cells sur-
rounding the one in which the gene resides (which can affect gene action,
sometimes as a result of cascading effects from the environment outside
the organism) and the other genes within the same cell (whose patterns
of activity can affect when a gene is switched on or off ). In short, the devel-
opment of an organism is not simply a matter of gene action, but a matter
of causal interaction between genes and their environment.
For this reason, there is no straightforward relationship of “determina-
tion” between genotypes and phenotypes. Indeed, given the interaction
between genes and environment in development, even if two individuals
possess the same genotype, they can differ in phenotype as a result of
developing under different environmental conditions. For example, if we
plant corn seeds of the same genotype in different soil conditions, and fer-
tilize and water those plants differently, the resulting corn plants can differ
significantly in phenotypes such as height of plant and sweetness of
kernels. So the same genotype can produce a range of different phenotypes
across a range of different developmental environments. Some genotypes
tend to produce the same phenotype across a very wide range of different
environments. But rarely is there a straightforward one-to-one relation
between a genotype and a phenotype. Genotypes typically produce differ-
ent phenotypes if developmental conditions are varied sufficiently.
Evolution 25
But, if this is true, what sense does it make to speak of a “gene for” a
particular phenotype? To say that a gene or genotype G is “for” a pheno-
type P means first of all that, other things being equal, an organism with
G is more likely to have P than is any organism without G (that is, with a
possible rival allele of G). The clause “other things being equal” is impor-
tant here, since it includes the environments in which the organisms
develop. The point is to compare organisms within developmentally
similar environments to see whether having G makes a difference with
respect to having P. For, if we compared organisms with G in one devel-
opmental environment to organisms without G in a different develop-
mental environment, then any difference among them with respect to
their having P could be due to the differences in their developmental envi-
ronments. The clause “other things being equal” enables us to focus on
how a change in having G produces a change in having P, rather than on
how a change in the environment produces a change in having P. But this
first condition is purely correlational, requiring that G be correlated with
P in relevant environments in order to be “for” P. Consequently, this con-
dition alone fails to distinguish the case in which G actually produces P
from the case in which G produces some other phenotype that is corre-
lated with P. Thus, to say that G is “for” P means, second, that G must play
a causal role in the development of P (in those organisms with P). When
these two conditions are met, it is perfectly sensible to speak of genes or
genotypes as being “for” phenotypes.
The fact that phenotypes are produced by the interaction of genes and
environment has a couple of implications with respect to understanding
the connection between evolution at the genetic level and changes across
generations in the phenotypes of organisms in a lineage. First, if a gene
increases in frequency across generations, the phenotype that it is the gene
for can increase in frequency only if the developmentally relevant aspects
of the environment remain relatively constant. This is because there will
only be a particular range of developmental environments in which that
gene will produce the phenotype it is for. So, if the environment changes
so as to fall outside the range in which that gene produces that pheno-
type, then any increase in the gene’s frequency will not be accompanied
by an increase in the frequency of the phenotype it is for. Thus, in order
for patterns of phenotypic change across generations to parallel patterns
of evolution at the genetic level, the developmentally relevant properties
of the environment must remain relatively stable across those generations.
Second, because genotypes can produce different phenotypes in differ-
ent developmental environments, it is possible for there to be phenotypic
26 Chapter 1
So far we have been concerned with what evolution is. But what causes
evolution? As mentioned earlier, evolution can occur only if there is vari-
ation in a population. For, if evolution is change in gene or genotype fre-
quencies, there must be at least two genotypes occurring at a particular
locus in a population, the frequencies of which then get altered across gen-
erations. So, if a population is composed of organisms that are genetically
identical, the only way that evolution can occur is if a new genetic variant
gets introduced into the population. With this in mind, the causes of
evolution can be divided into two very broad types: One type of cause
introduces new variants into a population and the other changes the
frequencies of already existing variants. Consider these in turn.
There are two main processes that cause evolution by introducing new
variants into a population, one of which is mutation. Recall that the first
stage of meiosis involves the replication, or copying, of the genes on each
chromosome in a diploid cell. In the process of gene replication, there are
occasional copying errors, in which one of the nucleic acid bases in a
sequence gets translated incorrectly—for example, an A in a sequence of
bases gets copied as T. The result is a new gene—a new sequence of bases—
that differs from the gene from which it was copied at that one position
in the sequence of bases. A copying error of this kind is called a mutation.
The mutation is then shuttled into one of an organism’s gametes where it
can be transmitted to one of its offspring, in whom the new mutant gene
can then produce some novel phenotype.
The other process that causes evolution by introducing new variants into
a population is recombination. To illustrate, consider a double heterozygote,
Evolution 27
frequencies of the genes with these phenotypic effects, it also changes the
frequencies of the phenotypes they produce. When a phenotypic trait
increases in frequency as a result of natural selection in this way, biologists
say that there has been selection for that trait—that the trait has conferred
a selective advantage, or reproductive advantage, on its bearers.
It is worth noting that some biologists apply the term natural selection
only to selection for traits that affect survival, while applying the term
sexual selection to selection for traits that affect the ability to attract and
mate with members of the opposite sex. In other words, traits that are solu-
tions to adaptive problems posed by members of the opposite sex evolve
under sexual selection, whereas traits that are solutions to adaptive prob-
lems posed by the rest of the environment evolve under natural selection.
Other biologists treat sexual selection as an aspect of natural selection. But
distinguishing the two can be useful when analyzing some traits, since
some traits are detrimental with respect to survival, yet enhance repro-
ductive success by appealing to members of the opposite sex. The classic
example is the peacock’s tail, which is detrimental to survival (since it
attracts predators and impairs the ability to escape), yet appeals to peahens
and, hence, increases the mating ability of well-endowed peacocks. For the
most part I will simply use the term selection, encompassing both natural
and sexual selection. But, when necessary, I will refer specifically to natural
or sexual selection.
Finally, the other process that can cause evolution by altering the fre-
quencies of already existing variants in a population is genetic drift, which
is due to two types of chance event: random survival and random sam-
pling of gametes.
Random survival is due to random events—for example, floods, fires, or
lightning strikes—that kill a much larger number of organisms with one
allele than those with the rival allele. This would have the effect of making
the latter allele more frequent in the next generation, since its bearers
would have survived to reproduce at a higher rate than the bearers of the
unlucky allele. This would constitute evolution, but it would be due to
chance rather than to differences in fitness.
We have already touched on the random sampling of gametes. Recall
that every organism produces many more gametes than will go to form
zygotes. We can thus think of fertilization as a process that randomly
“draws” one gamete from the total pool of gametes created by each parent
organism. When a population contains Aa organisms, which produce
gametes that are 50 percent A and 50 percent a, the random drawing of
their gametes, in each case, has a 50 percent chance of yielding an A and
Evolution 31
Adaptation
These are the nuts and bolts of evolution, but how do they fit together to
build all the complex, functionally integrated organisms that we see in the
world? For organisms are composed of numerous and diverse parts that are
32 Chapter 1
well adapted to one another and to particular features of the world, and
that appear very intricately designed for their functionally specialized
roles. Darwin called such functionally specialized parts of organisms
“organs of extreme perfection and complication” and, in illustration, mar-
veled at the human eye, “with all its inimitable contrivances for adjusting
the focus to different distances, for admitting different amounts of light,
and for the correction of spherical and chromatic aberration.”1 The eye, of
course, is merely one of many examples of such “perfection and compli-
cation.” The wings of birds are very well designed for flight, the echolo-
cation (sonar) system of bats is very well designed for detecting flying
insects at night, and the coloration of many species provides excellent
camouflage from predators.
Such “organs of extreme perfection and complication” appear to be
designed for a purpose. Echolocation, for example, appears to be designed
precisely so that bats can detect the flying insects that make up their diet.
And to say that a part is designed for a purpose is to say that an organism
possesses it because that part solves a particular adaptive problem. So, bats
appear to possess echolocation precisely because possessing echolocation
enables them to eat. But, if all apparent design in nature is the product of
evolution, rather than the product of creation by some intelligent being,
how can there be such design-for-a-purpose in nature? How, in other
words, can the processes discussed so far account for the apparent pur-
poseful design of functionally specialized “organs of extreme perfection
and complication”?
Whether it is a matter of building a trait that appears well designed for
solving some adaptive problem or building an entire organism composed
of numerous such traits that are all functionally integrated, the process is
the same: iterated cycles of modifying a preexisting structure and retain-
ing the modification. This process created all of the world’s diverse organic
forms out of simple replicating molecules. Of course it took a very long
time. But this book is not about the origins of species. For our purposes,
it is sufficient to understand how complex traits that solve adaptive prob-
lems are created by the causes of evolution just discussed—to understand
how traits can develop within a species that make organisms well adapted
to the specific demands of their environment. Consider first how such
traits can evolve under selection, then consider whether they can evolve
under drift.
Suppose there is a population of birds whose beaks vary slightly in size.
The sole food supply for this population is seeds that are digestible only
once they have been extracted from their hulls. To extract the seeds, the
Evolution 33
birds must use their beaks. Suppose that birds with the slightly broader
beaks are the most efficient at hulling the seeds, hence get the most nutri-
tion, and consequently enjoy a slight reproductive advantage over the
other birds in the population. The gene for the broad beak will thus
increase in frequency in the population, as will the broad beak itself.
But suppose also that the broader beak would be even more efficient at
hulling seeds if it had slightly sharper edges. And suppose that there is a
gene in the population that would produce sharper-edged beaks if it
mutated. Of course, since mutation is random, the fact that sharper beaks
would be beneficial doesn’t increase the probability that the desirable
mutation will occur. Also, since mutation is random, the mutation for
sharper-edged beaks is just as likely to occur in a bird without the broad
beak as it is to occur in a bird with the broad beak (in whom it would be
most beneficial). But, as the gene for the broad beak becomes ever more
frequent in the population, there is an increased probability that, if the
mutation for sharper-edged beaks occurs, it will occur in conjunction with
the gene for the broader beak, and thereby provide a beak that is even
better designed for hulling seeds. This increased probability of a better beak
is analogous to rolling dice. Suppose you need a three to turn up on a
rolled die. If you roll just one die, there is a one-sixth probability of getting
a three. And, if you roll twelve dice, each die has a one-sixth probability
of turning up three. But your odds of getting a three are greatly improved
if you can roll twelve dice rather than one die. Similarly, as the gene for a
broad beak spreads in the population, there is an increased probability that
a mutation for sharper edges might occur with it and, hence, further
modify the beak in a way that makes it even better designed for hulling
seeds. If the beak does get further modified in this way, the improved beak
will become more frequent in the population over succeeding generations.
And so on.
This process—a new mutation introduces a beneficial modification that
is retained by selection—can be repeated many times over a very long
period of time. After a very large number of generations, the population
can come to be composed of a large number of birds with beaks that are
extremely well designed for hulling seeds, beaks that have a shape that
conforms to the demands of the seed hulls and are powered by muscles
that exert efficient force in cracking those hulls. And that design will have
been produced by a process of cumulative retention of slight design
improvements introduced by random mutations. (Of course, there may
have been many other mutations that impaired design; but they would
have been selected against and driven to extinction.) In this way, the
34 Chapter 1
for that trait—specifically, the benefit because of which the trait was
selected. The benefit the trait provides is thus the reason why the trait
spreads or persists in the population; that benefit is the purpose of the trait,
since the trait’s providing that benefit is the reason organisms possess the
trait (via inheritance from ancestors in whom the trait was selected for).
A trait that is present in a current population because it performed a
function (solved an adaptive problem) that enhanced fitness in an ances-
tral population, and was thus preserved or proliferated under selection for
it, is called an adaptation. In other words, an adaptation is a trait that has
a history of having been preserved, and possibly modified, by selection for
the beneficial role it plays in an organism. Thus, an adaptation is a trait
that contributed to its own persistence or proliferation; for, by enhancing
the fitness of its bearers, an adaptation contributed to the reproductive
success of its bearers, which contributed to the transmission of the genes
for that adaptation, which in turn contributed to the development of that
trait in other organisms. Adaptations, in short, are self-perpetuating design
features of organisms. Organisms have those traits because they were ben-
eficial to their ancestors.
It is important not to confuse adaptation with adaptiveness. A trait is
adaptive if it enhances fitness, but it is an adaptation if it is possessed by
organisms in a current population because they inherited it from ances-
tors in whom that trait enhanced fitness. As the philosopher of biology
Elliott Sober so nicely puts it: “To say that a trait is an adaptation is to
make a claim about the cause of its presence; to say that it is adaptive is
to comment on its consequences for survival and reproduction.”2
This distinction is important to bear in mind because of the following
two implications. First, just because a trait is adaptive doesn’t mean that
it is an adaptation. A trait could evolve in a population under drift, but
then come to enhance the fitness of its bearers if the environment of the
population changes so as to make the trait useful. In such a case, the trait
would be adaptive, but since it did not evolve under selection it would not
be an adaptation. For adaptation is a historical concept, applying only to
traits with the right sort of evolutionary history. Second, a trait could be
an adaptation yet fail to be adaptive. This, too, could result from a change
in a population’s environment. A trait could evolve under selection, and
even go to fixation in a population, yet the environmental demands to
which that trait was responsive could cease, thereby rendering the trait
useless. In such a case, the adaptation would no longer be adaptive. For
adaptiveness is an ahistorical concept, applying only to traits that currently
enhance fitness. A trait is adaptive, then, if it has current utility; it is an
36 Chapter 1
right, the other must be wrong. Rather, they complement one another by
providing different kinds of information about the same phenomenon.
Indeed, one could see ultimate explanations as explaining why particular
proximate causes are operative. For example, the ultimate explanation of
eggshell removal explains why gulls have neurophysiological mechanisms
that respond to particular stimuli in a way that results in eggshell removal.
But that doesn’t mean that the ultimate explanation can replace a proxi-
mate explanation. Knowing the ultimate causes of eggshell removal
doesn’t give us any information about how eggshell removal gets accom-
plished by any individual gull. Similarly, a proximate explanation doesn’t
exclude an ultimate explanation, since knowing how a particular neuro-
physiological mechanism causes eggshell removal doesn’t inform us about
the causes of the evolution of eggshell removal. Thus, every adaptation can
be explained in terms of both proximate and ultimate causes, where the
former cites the immediate antecedent “mechanistic” causes and the latter
cites the evolutionary causes.
Phenotypic Variation
(Of course, it could be the case that extreme heights or weights would be
selected against, while all of the nonextreme variation in height and weight
would be selectively neutral.)
Third, even if a genotype for a fitness-enhancing phenotype goes to fix-
ation under selection, the fact that the same genotype can produce differ-
ent phenotypes under different developmental conditions means that the
phenotype it’s for won’t necessarily go to fixation also. For a genotype to
be selected, it needn’t always produce the fitness-enhancing phenotype. It
is only necessary that the average fitness of all the phenotypes it produces
(under all its developmental conditions) be higher than the average fitness
of all the phenotypes produced by alternative genotypes (under all their
developmental conditions). So, even a genotype that is increasing in fre-
quency under selection can sometimes produce phenotypes that provide
no selective advantage or are positively maladaptive. Indeed, even if that
genotype goes to fixation, it might still, in certain developmental condi-
tions, produce a phenotype other than the fitness-enhancing phenotype
it was selected for producing. Thus, variation in developmental conditions
can produce phenotypic variation even when a beneficial genotype has
gone to fixation.
These are cases in which phenotypic variation persists in a population
in spite of selection, as it were. More interesting, however, are the ways in
which selection can actively maintain phenotypic variation in a popula-
tion. There are several ways in which selection can maintain phenotypic
variation, but only two of these will be relevant to later discussions. So
here I’ll confine my discussion to those two ways: frequency-dependent
selection and adaptive plasticity.
Consider first frequency-dependent selection. To get a really good handle
on frequency-dependent selection, it is best to take a brief excursion into
cost-benefit analyses of fitness.
Fitness, recall, is a measure of the ability to survive and reproduce in a
particular environment. Many activities in which organisms engage
enhance or diminish that ability. For example, female black-tipped hang-
ingflies mate with males who offer them edible insects. When the male
presents the insect, the female feeds on it while copulation occurs. Con-
sequently, a male hangingfly enhances his ability to reproduce by captur-
ing an insect that will entice a female. Capturing an insect is thus a fitness
benefit for male hangingflies. Conversely, if a female lays eggs to be fertil-
ized by a male who turns out to be sterile, she diminishes her ability to
reproduce as a function of the lost eggs. Losing the eggs is for her a fitness
cost.
Evolution 39
Fitness costs and benefits need not be so drastic or so obvious. Each meal
that we eat contains nutrients that sustain us and thereby enhances our
ability to survive and reproduce relative to the ability we would possess in
the absence of receiving those nutrients. We can thus think of very simple
acts such as eating an apple as having an associated, yet small, fitness
benefit, measured in terms of the nutrients the apple provides and the role
those nutrients play in facilitating survival and reproduction. Similarly, the
very act of engaging in some activity has metabolic costs, diminishing the
energy available for engaging in other activities. Three hours spent in fruit-
less foraging diminishes one’s energy store for fruitful copulations. So every
activity has an associated, though perhaps small, fitness cost, measured in
terms of the depletion of energy available for other activities essential to
survival and reproduction.
The fitness costs and benefits of some activities in which organisms
engage are independent of the behavior of other members of the organ-
ism’s population. The energy gained from eating a particular food item, for
example, is independent of what other individuals in the population are
eating or doing. For male dung flies, there is an optimal amount of time
spent copulating, which maximizes the rate of egg fertilization per unit of
copulation time, and for any given male this optimum is independent of
how long other males spend copulating. And for many animals there is an
optimal amount of time spent foraging for food, which maximizes the
energy intake per unit of foraging time, and this optimum is independent
of how long other population members spend foraging. An activity with
fitness costs and benefits that are independent of how other population
members behave has frequency-independent fitness, since its fitness is inde-
pendent of the frequency of that activity in a population—independent,
that is, of how many population members engage in that activity.
But the fitness costs and benefits of many activities in which humans
and other animals engage are not independent of the behavior of other
population members. For example, in some species males fight with one
another for territory. If most males in such a species only engage in threat-
ening displays and retreat when attacked, a tactic of extreme aggression
might accrue high fitness benefits to any male adopting it. However, if
most males are extremely aggressive in conflicts, then aggression could
exact the fitness costs of injury or death. So the fitness costs and benefits
of any particular form of behavior in a conflict depend on the tactics
adopted by other males in the population. Similarly, when members of one
sex compete to mate with members of the opposite sex, the best tactic to
employ to attract members of the opposite sex can depend on what other
40 Chapter 1
members of your own sex are doing to attract mates. If all other members
of your sex send roses, rather than competing to find the best roses, you
may be better off sending orchids, which are easier to obtain since they
aren’t in demand. It may pay just to be different. In fact, in general, the
fitness costs and benefits associated with any activity that involves
competition with some other population members will be a function of
how one’s competitors behave. Such an activity has frequency-dependent
fitness, since its fitness is dependent on the frequency of that activity in a
population. Consequently, the fitness of an activity with frequency-
dependent fitness changes as the frequency of that activity in a popula-
tion changes.
When activities have frequency-dependent fitnesses, selection often
maintains a particular proportion of alternative variants. To see how this
can occur, consider a simple model known as the “Hawk-Dove game,”
which was first developed in a classic article by the evolutionary biologists
John Maynard Smith and Geoffrey A. Parker. For purposes of illustrating
the game, we’ll represent fitness costs and benefits by whole numbers, or
“fitness points,” where benefits are represented by positive numbers and
costs by negative numbers.
The Hawk-Dove game is a contest for a resource worth +40 points, and
contestants can “play” either Hawk or Dove in competing for the resource.
Hawks always attack and fight aggressively until they win or get seriously
injured. Doves always exhibit a threatening display, but never attack, and
retreat if attacked by their opponent. Since Hawks immediately attack and
Doves retreat when attacked, Doves always immediately lose to Hawks. But
we’ll assume that Hawks have a 50 percent chance of defeating another
Hawk and that Doves have a 50 percent chance of defeating another Dove.
Finally, we’ll assume that the cost of a serious injury is -60 points and that
wasting time and energy in a very prolonged contest costs -10 points.
Given these assumptions, neither Hawk nor Dove can evolve to fixation
and remain there. To see why, consider first a population of Doves. Since
Doves never attack and only retreat when attacked, the absence of attack
in every Dove-Dove contest results in a very prolonged contest of display,
so each Dove accrues -10 points. The eventual winner, however, gets +40
points for acquiring the resource. Since each Dove has a 50 percent chance
of winning, the average payoff for a Dove in a population of Doves is thus
+10 (+40 times 50 percent, plus -10 for wasting time and energy). But
suppose that a mutant Hawk arises in this population. This Hawk will win
every contest, so it will enjoy an average payoff of +40 compared to the
Dove average of +10. Consequently, Hawks will begin to increase in fre-
Evolution 41
produced just because “its number has come up,” not because that phe-
notype is especially suited to the particular environmental conditions in
which the organism happens to find itself. In contrast, the alternative phe-
notypes produced by adaptive plasticity are generated nonrandomly, in
response to the conditions that obtain in the organism’s environment; the
phenotypes are produced to match the environmental conditions. Second,
the alternative phenotypes of a mixed strategy have equal fitness at their
evolutionarily stable ratio. In contrast, the alternative phenotypes pro-
duced by adaptive plasticity need not have equal fitness; in fact, the alter-
native phenotypes can vary significantly in their fitnesses. It need only be
the case that the fitness of each phenotype is greater, in the environment
in which it occurs, than any of the alternative phenotypes would be in
that same environment. This is compatible with one of the alternative phe-
notype’s having lower fitness than the others; it just means that, in those
circumstances, all the other phenotypes would have even lower fitness.
In biology, the concept of adaptive plasticity is applied to a very wide
range of phenomena. But I will discuss just two distinct forms of adaptive
plasticity, developmental plasticity and phenotypic plasticity. These two forms
of adaptive plasticity don’t exhaust the phenomena, but they are the forms
that will be important in later discussions. Before elaborating this distinc-
tion, however, I should issue a caveat. I will not be using the terms adap-
tive plasticity, developmental plasticity, and phenotypic plasticity in a way that
conforms with standard usage in biology. The reason is that there simply
is no standard usage of these terms in biology. Indeed, discussions of plas-
ticity in the biological literature are characterized by widespread termino-
logical inconsistency. Some biologists use all three of the above terms
interchangeably, while other biologists distinguish developmental plastic-
ity from general phenotypic plasticity. In what follows, then, I will be
defining these terms so as to serve my purposes. And, for my purposes, I
will treat developmental plasticity and phenotypic plasticity as distinct
forms of the more general phenomenon of adaptive plasticity.
To illustrate developmental plasticity, consider the caterpillars of the
moth Nemoria arizonaria, the larvae of which develop in oak trees. Cater-
pillars hatched in spring feed on the staminate flowers of the oak and
develop to strongly resemble those flowers. Caterpillars hatched in summer
feed on the leaves of the oak and develop to strongly resemble twigs on
the oak. A difference in diet, due to a difference in chemical composition
of the flowers and the leaves, is responsible for the development of the
very different “flower” and “twig” phenotypes. And each phenotype is
adaptive in its circumstances, since each serves the function of camou-
Evolution 45
nonterritorial males, so they also incur greater fitness costs than nonterri-
torial males. Conspicuous coloration is worth its high costs only if a male
is actively reproducing.
The territories defended by territorial males typically lie within the
recesses of vegetation or behind leaves. In a natural environment, such ter-
ritories prove highly unstable. Leaves move, territories are exposed, and
other areas become visually isolated, hence suitable candidates for defen-
sible territory. When these changes occur, nonterritorial and displaced ter-
ritorial males compete for new territories. If a nonterritorial male captures
a territory, within days it becomes brightly colored and develops mature
testes. If a displaced territorial male fails to secure a new territory, within
days it loses its bright coloration, becoming cryptically colored, and its
testes begin to atrophy. As a displaced territorial male makes the transition
to nonterritorial male, it begins to once again allocate all of its energy to
somatic growth in preparation for later competition for new territories. If
their habitats fluctuate greatly, male H. burtoni can cycle several times in
this way through the territorial and nonterritorial phenotypes.
In order for such phenotypic plasticity to evolve by selection, there must
be variation in some aspect of a population’s environment that is relevant
to fitness, just as with the evolution of developmental plasticity. But, in
order for phenotypic plasticity rather than developmental plasticity to
evolve in response to environmental variation, the environmental varia-
tion must occur relatively rapidly and unpredictably. That is, a population
must face several different environmental conditions within the course of
a single generation, there must be no reliable pattern in the order in which
those different environmental conditions are encountered, and each dif-
ferent environmental condition must be of uncertain duration. In short,
there must be fluctuation in some aspect of a population’s environment
that is relevant to fitness. Finally, the ability to vary phenotype in response
to these fluctuating conditions must have higher average fitness than any
single phenotype would have across all of the conditions. If it weren’t
better to revert to the nonterritorial phenotype when not holding a terri-
tory, for example, H. burtoni males would be brightly colored with mature
testes throughout adulthood.
The following illustration will help to make the distinction between
developmental plasticity and phenotypic plasticity less abstract and more
intuitive. The phenomena to which I am applying the label developmental
plasticity tend to conform to the following model: Organisms can
encounter one of two types of environment, either an environment char-
acterized by round holes or one characterized by square holes. A genotype
48 Chapter 1
When a gene has a phenotypic effect that enhances the fitness of its bearers
(that makes its bearers fitter than population members without the gene),
it tends to increase in frequency in a population, and as a result so does
its fitness-enhancing phenotypic effect. So, a gene that makes a gazelle a
faster runner than others in its population, and hence better able to escape
predators, will increase in frequency in a gazelle population and thereby
increase the average running speed of the population. And a gene that
makes its bearers immune to a common and deadly disease will similarly
increase in frequency. In both cases, the gene codes for a protein that
makes its bearers’ bodies different from those without the gene. In the
former case the effect is on musculature and in the latter on antibody
production.
A gene can also increase in frequency by making its bearers more likely
than nonbearers to perform some fitness-enhancing behavior. For example,
females of many species choose a mate based on the quality of male
courtship displays. If the courtship displays of males differ in quality and
a genetic difference underlies the display difference, the gene for the supe-
rior display will increase in frequency. Of course, courtship behaviors are
not the only behaviors that affect fitness. If parents differ in the quantity
of care they give to their offspring, if the quantity of care affects the via-
bility of offspring, and if a genetic difference underlies this difference in
parental care, then the gene for higher quantity care will increase in fre-
quency. So, as long as a gene makes some fitness-enhancing behavior more
likely, that gene will increase in frequency in a population, and as a result
the behavior may increase in frequency as well. For this reason, biologists
frequently say that, from the standpoint of evolutionary biology, “behav-
ioral traits are like any other class of characters.”1
But treating behavioral traits as just like other traits obscures a funda-
mental difference between the way genes affect morphological traits like
stronger muscles and the way they affect behavioral traits like courtship
displays. For muscles are made of proteins, and genes affect the strength
of muscles by coding for the proteins of which they are made. This is a
stable and relatively long-lasting condition of an animal’s body, which is
not reactive to the rapidly changing conditions in an animal’s environ-
ment. Behavior, on the other hand, isn’t made of proteins; it is what an
animal does in the world with its body. It consists in relatively fleeting
events that are caused by an animal’s brain in reaction to rapidly changing
conditions (although in simple animals behavior is merely reflexive, caused
Mind 51
amino acids (the building blocks of proteins). By coding for the amino
acids and proteins of which these things are made, genes can affect the
neuronal structure and functioning of the proximate mechanisms that
make up the brain (by affecting the goals that the brain entertains or how
the brain processes information relevant to achieving those goals). And
this could make an animal with a particular gene more likely than one
without it to produce a particular type of behavior.
It must not be forgotten, however, that behavior is reactive to current
conditions, some of which are external conditions of the environment,
and genes don’t directly affect those external conditions. For example,
Thomson’s gazelles exhibit a behavior called stotting, which consists in
repeatedly jumping a couple of feet off the ground in plain sight of a
predator, a behavior apparently designed to communicate something like
this to the predator: “I see you watching me and I’m prepared to run if
you charge; but I have very strong leg muscles, as demonstrated by how
high I’m jumping right now, so you would be ill advised to bother trying
to catch me.” A gazelle’s genes don’t affect whether there are predators in
its current environment and, consequently, don’t affect one of the condi-
tions essential for stotting to be exhibited. But a gene can affect a gazelle’s
brain so that it is more likely to stott when it sees a cheetah. Thus, while
genes cannot directly affect behavior, they can affect proximate mecha-
nisms in such a way that certain behavioral responses become more likely
under certain conditions.
Since there can be genes for behaviors in this (indirect) way, the full
range of evolutionary causes discussed in the last chapter can produce
behavioral evolution in a population. But a fitness-enhancing behavior
may not actually increase in frequency in a population. This is because the
external conditions to which the behavior is reactive may decrease in fre-
quency, even if only temporarily. If this happens, behavior that provides
a selective advantage one generation may actually be less frequent in the
next generation. Suppose that there is selection for stotting in a gazelle
population, which increases the frequency of the gene for stotting, but that
the population of gazelle predators is suddenly wiped out by some natural
disaster. Gazelles would then cease stotting, but only because the external
conditions to which stotting is a response would be lacking, not because
the population would no longer be composed of gazelles with a disposi-
tion to stott. Indeed, because of the selection for stotting, more gazelles
would possess the proximate mechanism that underlies the disposition to
stott, so more gazelles would possess the tendency to stott when they see
a predator. They just wouldn’t see predators.
Mind 53
But is there any reason for thinking that human behavior has evolved
under selection? In other words, is there any reason for thinking that the
human brain is an adaptation? Evolutionary Psychologists answer this
question with a strong affirmative (although, as we will see later in this
54 Chapter 2
chapter, they claim that the brain is not a single adaptation, but a network
of many specialized adaptations).
Their argument is simple and compelling. The human brain, which is
capable of producing highly sophisticated and appropriate behavior in
response to an indefinite variety of circumstances, is without doubt the
most complex organ possessed by any species. For reasons we saw in
chapter 1, the probability that something of this complexity would have
evolved purely by random, undirected processes (such as genetic drift) is
so infinitesimally small that such processes can be ruled out as an expla-
nation of the complex design of the human brain. The human brain is
undoubtedly the result of a long period of selection, in which a series of
fitness-enhancing modifications to the brain’s cognitive and motivational
mechanisms were retained and accumulated. But selection retains modifi-
cations only when they provide new or better solutions to adaptive prob-
lems. Thus, the brain must have been designed by selection to produce
behavioral solutions to the adaptive problems faced by human popula-
tions. “The evolutionary function of the human brain,” as Cosmides and
Tooby say, “is to process information in ways that lead to adaptive
behavior.”3 In short, the brain is an adaptation for producing adaptive
behavior. Understanding the evolutionary process that designed the brain
should thus enable us to understand its functional, information-
processing design, to understand what selection designed the brain to do.
The real question, then, is what kind of functional design the brain has
evolved under selection. Since Evolutionary Psychologists follow the over-
whelming majority of cognitive scientists in identifying the mind with the
functional, information-processing design of the brain, this question is
tantamount to asking what kind of mind has evolved in the human lineage.
One possibility is that selection designed the brain to monitor the
current environment and produce whatever behavior maximizes fitness.
According to this view, as the psychologists John and Stamati Crook
endorse it, “humans are endeavouring consciously or unconsciously to
optimize their reproductive success.”4 This leads to the prediction that
humans in current populations should be engaging in behavior that is
designed to maximize reproductive success.
The Evolutionary Psychologists David Buss and Donald Symons argue
that this prediction is everywhere disconfirmed, for contemporary human
behavior too often fails to maximize fitness. For example, donating one’s
sperm or eggs to cryobanks is an obvious way to maximize reproductive
success, but few people pursue this practice. Further, if humans had an
evolved behavioral tendency to maximize reproductive success, the use of
Mind 55
changed relatively little over evolutionary time). For example, the physi-
cal environment of terrestrial animals undergoes climatic changes, which
sometimes devastate a population’s food resources. More significant,
however, are the changes in a population’s biological environment, which
consists of other species that are themselves undergoing evolution—in
particular, parasite, predator, and prey species. For example, if a cheetah
population increases its average running speed, the population of gazelles
on which it preys faces a new adaptive problem (that of escaping faster
cheetahs).
When the environment inhabited by a population changes, what was
previously beneficial, and increasing in frequency under selection, may
cease to be beneficial. When this happens, the population can adapt to the
new environment, but how quickly it does so depends on the extent and
nature of the variation in the population when the environment changes.
If there is significant variation and one of the existing variants provides a
reproductive advantage in the new environment, that variant will begin
to spread in the population. If the existing variants in the population have
equal fitness, on the other hand, so that no one variant provides a repro-
ductive advantage over the others, then the population will only begin to
adapt after a mutation occurs that enhances fitness in the new environ-
ment. Either way, it can take another several hundred to several thousand
generations for the population to adapt to the new environment—depend-
ing on whether a new mutation is required, if so how long it takes before
a beneficial mutation occurs, and how great a reproductive advantage is
provided by the beneficial variant. During these generations, the popula-
tion as a whole will exhibit a lack of fit between its adaptations (the traits
that are present in the population because they were adaptive in ancestral
populations) and its current environment. In short, during this time its
adaptations will not be adaptive.
In the case of the brain, this lack of fit between adaptation and current
environment has a particular character. For, if the brain is an adaptation,
as Evolutionary Psychologists argue, then at some point in human evolu-
tionary history there was selection for brain mechanisms that transformed
information about environmental conditions into behavioral output that
was adaptive to those environmental conditions. Of course, there was
selection for these mechanisms because of how they responded to the envi-
ronment prevailing at the time they were selected. If environmental conditions
are now different, the informational inputs to the brain are different also;
consequently, the brain may respond to this new information by produc-
ing behavior that is not adaptive to the new conditions.
Mind 57
There are at least two ways in which this failure of adaptiveness could
occur. In the simple case, the brain could continue to produce the same
old behavior under the new conditions, but that behavior may fail to have
the beneficial consequences under the new conditions that it had under
the old. On the other hand, the brain could produce some new behavior
in response to the new informational inputs, but the new behavior could
fail to be adaptive. This latter phenomenon is common to a wide range of
mechanisms that are reactive to the conditions around them: They are
designed to respond well to certain circumscribed conditions; but, when
those conditions change, they respond in ways their designers did not
expressly intend. For example, sophisticated computer programs, which
are designed to perform a variety of complex tasks, are programmed to
produce “adaptive” outputs (on the monitor or to a peripheral device) in
response to certain key or key-combination inputs. But a novel key-
combination input can cause the program to give some nonsensical
or undesirable (“nonadaptive”) output—a phenomenon that is all too
familiar to Windows users. The program wasn’t designed by the program-
mers to give this output; but, given how the program was designed, this
output is an incidental response to an unanticipated input. Similarly, a
proximate mechanism that produced adaptive behavior under the condi-
tions in which it was selected may fail to produce adaptive behavior when
it no longer encounters the informational inputs it once responded to
adaptively.
The assumption that humans have an evolved behavioral tendency to
maximize reproductive success, Evolutionary Psychologists argue, conflates
the claim that the human brain was designed by a history of selection
(adaptation) with the claim that it functions to maximize reproductive
success (adaptiveness). The fact that the brain was designed by selection
entails only that it is designed to solve the adaptive problems posed by the
environment in which there was selection for that brain design; it doesn’t
entail that the brain is designed to produce fitness-maximizing behavior
in present environments. For human environments may have changed
since there was selection for the brain’s design, and much contemporary
human behavior could thus be maladaptive because of a “time lag”
between human brain design and the environment.
But is there any reason for thinking that the human mind is lagging
behind modern environments in this way? Evolutionary Psychologists
argue that the answer is yes, and largely because of environmental changes
produced by human action. The invention of agriculture some 10,000
years (four hundred generations) ago; the industrialization of Western
58 Chapter 2
societies some 200 years (eight generations) ago; the attendant rise of the
modern metropolis and megalopolis, which resulted in humans living in
unprecedentedly large groups; the proliferation and increased availability
during the last century of contraceptives hitherto unrivaled in effective-
ness—all of these, Evolutionary Psychologists argue, have changed the
selectively relevant environment for humans in profound and far-
reaching ways. These are changes with which human genetic evolution,
and consequently human psychological adaptation, could not possibly
have kept pace; for, since those changes took place, they argue, there have
been too few generations for significant genetic evolution to have
occurred. For these reasons, among others to be discussed in the next
section, Evolutionary Psychologists think it overwhelmingly likely that the
human mental design built by selection will frequently fail to produce
adaptive behavior in modern environments.
Evolutionary Psychologists conclude that an evolutionary approach to
understanding the design of the human mind should not lead us to look
for fitness-maximizing behavior but should lead us to investigate the adap-
tive history that shaped the mind. Thus, the central premise of Evolution-
ary Psychology is the idea that the brain’s design was produced by a history
of reproductive success, rather than the idea that the brain is designed to
produce reproductive success. This prompts Evolutionary Psychologists to
say that humans are not “fitness maximizers,” but “adaptation executors,”
which means that human behavior is produced not by a brain that is
attempting to maximize fitness, but by a brain functioning in ways that
were selected for because of how they maximized fitness in the past. And,
because our psychological adaptations may fail to produce adaptive
behavior in modern human environments, Evolutionary Psychologists
believe that “studies of the adaptiveness of human behavior are ineffective
in illuminating human psychological adaptations.”5
to function faster than they would if they had to sort through the entire
range of information available in the brain.
Informational isolation entails that much of what one learns can fail to
affect the way that modules process information and go about solving
problems. To borrow an example from the cognitive scientist Jerry Fodor,
optical illusions persist even after you learn that they are merely illusions.
You know that there is no puddle on the desert road ahead, yet you see one
anyway; you know that the lines in the Müller-Lyer illusion are of equal
length, but that doesn’t prevent you from seeing them as having different
lengths. This is because our vision modules are informationally isolated in
the extreme. What we come to know about the world doesn’t affect the
way that our vision modules function in constructing representations of
the spatial layout and orientations of objects in the world. Similarly, due
to their informational isolation, evolved modules process information in
ways that are unaffected by much of what we learn, even when some of
what we learn is relevant to the adaptive problems those modules have
evolved to solve.
Evolutionary Psychologists support this view of the mind with both
general theoretical considerations and some empirical evidence. Let’s begin
by considering three theoretical arguments in support of this view.
First, as we have seen, the adaptive problems faced by our Pleistocene
ancestors varied widely in character, ranging from identifying edible plant
matter and avoiding deadly predators to selecting a reproductively valu-
able mate and cooperating with others in a status hierarchy. Symons argues
that, given the diverse characters of these problems, what constitutes a suc-
cessful solution to one problem is very different from what constitutes a
solution to another. So no single general-purpose problem-solving strategy
can successfully generate solutions to all of the problems in such a diverse
array; instead, each problem requires its own domain-specific problem-
solving strategy. Thus, our Pleistocene ancestors could not have evolved
minds consisting of a single all-purpose problem-solving mechanism, but
must have instead evolved distinct domain-specific mechanisms, each
dedicated to solving a specific adaptive problem posed by the Pleistocene
hunter-gatherer lifestyle. As Symons says: “It is no more probable that
some sort of general-purpose brain/mind mechanism could solve all the
behavioral problems an organism faces (find food, choose a mate, select
a habitat, etc.) than it is that some sort of general-purpose organ could
perform all physiological functions (pump blood, digest food, nourish an
embryo, etc.).”15
Mind 67
“Human Nature”
Tooby and Cosmides offer two arguments to show that human psycho-
logical adaptations must be universal. One argument is what I will call “the
argument from Gray’s Anatomy.” As Tooby and Cosmides put it, “the fact
that any given page out of Gray’s Anatomy describes in precise anatomical
detail individual humans from around the world demonstrates the pro-
nounced monomorphism present in complex human physiological adap-
tations. Although we cannot yet directly ‘see’ psychological adaptations
(except as described neuroanatomically), no less could be true of them.”31
The point is simply that, since selection has designed our minds as well as
our bodies, we should expect selection-designed psychological traits to be
just as universal as selection-designed morphological traits.
This argument is largely a rhetorical appeal to common sense, but Tooby
and Cosmides offer another, more theoretical argument, which I will call
“the argument from sexual recombination.” It first appeared in their 1990
article “On the Universality of Human Nature and the Uniqueness of the
Individual: The Role of Genetics and Adaptation,” and it has been repeated
and refined in several of their subsequent publications. It has become
accepted among Evolutionary Psychologists as definitive proof of a uni-
versal human nature, and it is repeated without modification by Buss,
Pinker, and Symons. The argument is as follows.
Adaptations are complex traits, which possess parts that are functionally
interconnected. (Think of how the cornea, pupil, lens, retina, and so on,
of the eye must be functionally interconnected in order for the eye to
perform its function of transmitting information about objects in the
world to the brain.) Being so complex, such traits “require coordinated
gene expression, involving hundreds or thousands of genes to regulate
their development.”32 Since sexual reproduction is a process in which
random halves of each parent’s genes are “recombined” to form the
genome of a zygote, if parents differed in any of their complex adapta-
tions, the random sexual recombination of the genes for those adaptations
would make it highly improbable that their offspring would receive all the
genes necessary for developing any of the parental adaptations. In other
words, given the randomness inherent in sexual reproduction (in particu-
lar, in crossing over and gamete sampling), “it is improbable that all of the
genes necessary for a complex adaptation would be together in the same
individual if the genes coding for the components of complex adaptations
varied substantially between individuals.”33 As a result, if individuals dif-
fered in the genes underlying adaptations, no adaptation would be reliably
reproduced across generations, and this in turn would mean that no
adaptive trait would persist long enough to be modified by selection in
74 Chapter 2
Thus, just as teeth and breasts are absent at birth and develop later in an
individual’s life history, perceptual organization, domain-specific reason-
ing mechanisms, the language acquisition device, motivational organiza-
tion, and many other intricate psychological adaptations mature and are
elaborated in age-specific fashions that are not simply the product of
the accumulation of ‘experience.’ Consequently, psychological adaptations
may be developmentally timed to appear, disappear, or change operation
to mesh with the changing demands of different age-specific tasks.”48
While individuals of different ages may differ considerably in their
“manifest psychologies or behaviors,” such differences are due to the dif-
ferential activation of common developmental programs as a function of
different maturational cues. In this respect, such differences are very
similar to individual differences that are due to differences in environ-
mental cues during development.
All these adaptive differences among humans—different psychological
phenotypes due to different environmental cues during development,
different adaptive forms due to the genetic switch determining sex, and
different morphological and psychological differences due to differences
in stage of life cycle—are due merely to differences in input to a common
developmental program, a common set of innate adaptations. For “the
genetically universal may be developmentally expressed as different
maturational designs in the infant, the child, the adolescent, and the adult;
in females and males; or in individuals who encounter different cir-
cumstances.”49 If all of us shared the same form of the genetic switch
determining sex, were the same age, and were exposed to the same envi-
ronmental cues during development, our operational psychological
adaptations would be identical. Thus, our “genetically universal” develop-
mental program (innate adaptations) and the psychological modules it is
designed to produce under normal conditions (operational adaptations)
are what constitute our “universal human nature.”
3 Adaptation
As even a casual reading of the previous chapter makes clear, the concept
of adaptation is central to Evolutionary Psychology. For Evolutionary Psy-
chology’s goal is to discover our psychological adaptations and explain
how they function, and the paradigm holds that our network of psycho-
logical adaptations constitutes a “universal human nature” that is adapted
to the lifestyle of Pleistocene hunter-gatherers. This focus on psychologi-
cal adaptation to the Pleistocene EEA led the late paleontologist Stephen
Jay Gould to criticize Evolutionary Psychology for being adaptationist and
unscientific. Indeed, Gould claimed that adaptationism is “the fatal flaw
of Evolutionary Psychology in its current form.”1
Despite its influence in some circles, however, Gould’s critique of Evo-
lutionary Psychology is fallacious and misguided. In the next section, I will
detail some of the problems with Gould’s arguments in order to dispose of
red herrings. As we carry the red herrings toward the garbage, though, we
will be put on the scent of some legitimate problems with Evolutionary
Psychology’s claims about psychological adaptation, and they will be the
focus for the remainder of the chapter.
Adaptationism?
largeness of brain per se. Rather, selection made the human brain large
because of the specific cognitive capacities conferred by larger brains. As
Pinker points out, the mere largeness of the human brain is purely detri-
mental. “If anything is a byproduct, it is the size of the human brain, which
guzzles nutrients [18 percent of the body’s energy intake, although it is
only 2 percent of the body’s weight], makes us vulnerable to blows and
falls, compromises the biomechanical design of the woman’s pelvis, and
makes childbirth dangerous. Bigness of brain is surely a byproduct of selec-
tion for more complex (and hence hardware-demanding) computational
abilities, ones that allowed our ancestors to deal with tools, the natural
world, and one another.”10 Thus, Gould’s claim about what is a psycho-
logical adaptation, which functions to separate adaptation from by-
products, gets things backwards.
This characterization may, however, be uncharitable to Gould. It is pos-
sible that what Gould means by “large brain” is “general-purpose intelli-
gence.” In that case, Gould’s claim that most of our “mental properties and
potentials” are spandrels would actually be a rejection of Evolutionary Psy-
chology’s modular view of the mind and an endorsement of the view that
the mind’s cognitive capacities are domain general. While this interpreta-
tion would charitably avoid attributing to Gould the absurd view that there
was selection for big heads, it nonetheless fails to advance his case against
Evolutionary Psychology, since Gould offers no reasons to doubt Evolu-
tionary Psychology’s modularity thesis or to accept domain generality. He
simply doesn’t address these issues. So, however one reads him, Gould fails
to show that adaptationism is the “fatal flaw” of Evolutionary Psychology.
Of course, escaping Gould’s charge of adaptationism unscathed is mean-
ingless if Gould is correct in claiming that Evolutionary Psychology is
unscientific. But what are Gould’s grounds for that claim?
The claim is grounded in another aspect of Gould and Lewontin’s classic
critique of adaptationism. Gould and Lewontin argue that the assumption
that a trait is an adaptation is often so powerful that it leads evolutionists
to accept an ultimate explanation of a trait even in the absence of evidence
for it. “Often, evolutionists use consistency with natural selection as the sole
criterion [of hypothesis acceptance] and consider their work done when
they concoct a plausible story” about how a trait may have evolved under
selection.11 But, “since the range of adaptive stories is as wide as our minds
are fertile,” Gould and Lewontin argue, it is always easy to concoct a story
about how some trait was adaptive in the long-gone evolutionary past.12
Given the ease with which such stories can be concocted, Gould pejora-
tively dubs them “just-so stories.” What is needed, Gould argues, is not a
Adaptation 87
“just-so story,” which explains how a trait may have evolved under selec-
tion, but some serious evidence of past selection to support the story.
Gould thinks that Evolutionary Psychology, however, has priced itself
out of the market for evidence to support its adaptive hypotheses. His argu-
ment is as follows. First, he rightly points out that Evolutionary Psychol-
ogists don’t assume that “all prominent and universal behaviors must, ipso
facto, be adaptive to modern humans in boosting reproductive success.”13
But, Gould claims, as a result “the task of evolutionary psychology then
turns into a speculative search for reasons why a behavior that may harm
us now must once have originated for adaptive purposes. . . . Much of evo-
lutionary psychology therefore devolves into a search for the so-called EEA,
or ‘environment of evolutionary adaptation,’ that allegedly prevailed in
prehistoric times.”14 However, Gould argues, “claims about an EEA usually
cannot be tested in principle but only subjected to speculation.”15 For “how
can we possibly know in detail what small bands of hunter-gatherers did
in Africa two million years ago? These ancestors left some tools and bones,
and paleoanthropologists can make some ingenious inferences from such
evidence. But how can we possibly obtain the key information that would
be required to show the validity of adaptive tales about an EEA . . .?”16 Since
we have no evidence of how selection acted on human populations in
the EEA, Gould concludes, “the chief strategy proposed by evolutionary
psychologists for identifying adaptation is untestable, and therefore
unscientific.”17
Once again, however, Gould’s salvo misses the mark. For Evolutionary
Psychology is unscientific only if it is untestable in principle, not simply if
it is untestable in practice. A theory or hypothesis is untestable in princi-
ple if there is no possible evidence that could count for or against it. To put
this another way, if a theory or hypothesis is compatible with all possible
evidence, then it is unscientific. For example, young-earth creationism,
which claims that the earth was created by God some six thousand years
ago, is unscientific because there is no possible evidence that can count
against it. If geologists find rock strata that appear to be billions of years
old, according to young-earth creationism that is simply because six thou-
sand years ago God created the rock strata to appear to be billions of years
old. Any possible evidence that appears to count against the theory that
the earth was created six thousand years ago thus gets explained in such
a way as to be compatible with the theory. For that reason, young-earth
creationism is unscientific.
In contrast, a theory or hypothesis is untestable in practice if there is some
possible evidence that would count for or against it, and if all such
88 Chapter 3
evidence is currently unavailable to us. But the mere fact that we are
unable, in practice, to gather the evidence needed to test a theory or
hypothesis does not mean that it is unscientific. Particle physics, for
example, has a history of proposing hypotheses about Very Tiny Things.
Often, at the time these hypotheses are proposed, we have no way to test
them, since we lack the technology to gather the evidence needed to test
them. But this does not make these claims “untestable, and therefore unsci-
entific,” for we typically know precisely what kinds of evidence would test
the claims if only we could obtain it. In short, there is evidence against
which the claims could be tested, so the claims are testable in principle;
we just aren’t able, in practice, to obtain that evidence. Indeed, in particle
physics, knowing what evidence would be needed to test theories often
guides the process of building ever more super particle accelerators and
colliders, which then enable us to create the conditions necessary for
testing hypotheses. When we build the new supercollider, a theory or
hypothesis that was once untestable in practice becomes testable in prac-
tice; but all along it was testable in principle, hence scientific. A claim is
truly unscientific only when there is in principle no evidence against
which it can be tested.
Now, in principle we often know precisely what kinds of evidence would
confirm or disconfirm Evolutionary Psychology’s claims about the EEA. For
example, Evolutionary Psychologists claim that, other things being equal,
Pleistocene males who preferred mating with nubile females would have
produced more offspring than Pleistocene males who preferred mating
with prepubescent or postmenopausal females. As a result, they claim,
human males would have evolved a preference for nubile females, and
consequently this preference is an adaptation in contemporary human
males.
This adaptive hypothesis has clear test implications regarding our EEA.
If it is true, any observer stationed in our EEA should have observed males
with a preference for nubile mates outreproducing males with a preference
for prepubescent or postmenopausal mates (other things being equal). In
principle we know exactly what kinds of evidence would confirm or dis-
confirm the adaptive hypothesis, so the hypothesis is testable in principle,
contrary to Gould’s claim. It’s just that in practice we don’t have access to
the evidence needed to confirm or disconfirm the hypothesis, since we
can’t actually observe our Pleistocene ancestors. But, since such hypothe-
ses are testable in principle, they aren’t unscientific.
Of course, even though Evolutionary Psychology’s adaptive hypotheses
are testable in principle, Gould could still press the point that they are
Adaptation 89
human males would have faced the adaptive problem of choosing a fertile
mate. “Under these conditions, males who happened to mate with females
of ages falling outside the reproductive years would become no one’s ances-
tors. Males who happened to mate with females of peak fertility, in con-
trast, would enjoy relatively high reproductive success. Over thousands of
generations, this selection pressure would, unless constrained, fashion a
psychological mechanism that inclined males to mate with females of high
fertility over those of low fertility.”20 Buss then conducted a large-scale,
cross-cultural survey of male mate preferences in order to determine
whether males possess the predicted psychological mechanism.
Given that the pattern of adaptive reasoning in Evolutionary Psychology
is precisely the reverse of what Gould criticizes, Evolutionary Psychologists
have a ready defense of the scientific character of their enterprise. For Evo-
lutionary Psychologists use adaptive reasoning to predict the presence of
previously undiscovered psychological mechanisms in the human mind;
this is the whole point of evolutionary functional analysis. Evidence that
humans possess the predicted psychological mechanisms is therefore evi-
dence for the adaptive hypotheses that entail the predictions. This is a
standard mode of scientific reasoning. A hypothesis is formulated and a
prediction is derived from it; if the hypothesis correctly predicts the
existence of a previously undiscovered phenomenon, then the hypothesis
is considered confirmed. In other words, the discovery of a previously
undiscovered and unexpected phenomenon confers some degree of
probable truth on the hypothesis that predicted it. Consequently, Evolu-
tionary Psychologists could argue, discovering that humans have a psy-
chological mechanism that is predicted by some adaptive hypothesis about
human evolution in the EEA is evidence for that adaptive hypothesis. Since
Gould mistakes the logic of adaptive reasoning in Evolutionary Psychol-
ogy, he misses the fact that adaptive hypotheses can be supported by
the confirmed predictions they make about human psychological
mechanisms.
While this does mean that Gould’s arguments fail to show that there is
anything wrong with adaptive reasoning in Evolutionary Psychology, it
doesn’t mean that all is well with Evolutionary Psychology. Indeed, adap-
tive reasoning in evolutionary functional analysis involves three steps, and
Evolutionary Psychology can face problems at each of these three steps.
The first step involves the identification of the adaptive problems our
ancestors faced. The second step involves inferring the psychological
mechanisms that must have evolved to solve those adaptive problems.
Since the psychological mechanisms that are predicted in the second step
are not observable, the third step involves conducting experiments
92 Chapter 3
Adaptation Hunting
Evolutionary Psychology claims that it’s well on the way to discovering the
psychological adaptations that form the human mind and that the job will
soon be completed. As Tooby and Cosmides so boldly say, “just as one can
now flip open Gray’s Anatomy to any page and find an intricately detailed
depiction of some part of our evolved species-typical morphology, we
anticipate that in 50 or 100 years one will be able to pick up an equiva-
lent reference work for psychology and find in it detailed information-
processing descriptions of the multitude of evolved species-typical
adaptations of the human mind.”21
This confidence is based on faith in the method of evolutionary func-
tional analysis, which is a kind of reverse engineering. Forward engineering
is a process of designing a mechanism that will be capable of performing
some desired task. Reverse engineering is a process of figuring out the
design of a mechanism on the basis of an analysis of the tasks it performs.
Evolutionary functional analysis is a form of reverse engineering in that it
attempts to reconstruct the mind’s design from an analysis of the prob-
lems the mind must have evolved to solve.
However, because our psychological mechanisms were designed to solve
the adaptive problems faced by our Pleistocene ancestors, according to Evo-
Adaptation 93
With respect to the first problem, we simply don’t know what adaptive
problems our ancestors faced in our EEA. We don’t even know the number
of species in the genus Homo, which species were our direct ancestors, or
in general how Homo species were related to one another, let alone details
about the lifestyles led by those species. This was one of Gould’s complaints
against Evolutionary Psychology. But Gould oversimplified by ignoring the
fact that Evolutionary Psychologists propose three sources of evidence
from which we can obtain information about the adaptive problems faced
by our ancestors. So how, according to Evolutionary Psychologists, can we
obtain information about the adaptive problems faced by our Pleistocene
ancestors?
One source of evidence is the design of adaptations. As the evolution-
ary biologist Randy Thornhill puts it: “The functional design of an adap-
tation is the record of the salient, long-term environmental problem
involved in the creation of the functional design. Thus, we can actually
scientifically go back in time and discover the creative selection pressures
that were effective in human evolutionary history.”25 Following Thornhill’s
lead, David Buss argues that, although “we lack a videotape of the selec-
tive pressures” that affected human psychology, nonetheless “some selec-
tion pressures can be inferred from the . . . analysis of the current design
of our mechanisms.”26
The problem with this suggestion is that we are supposed to discover the
“current design of our mechanisms” via evolutionary functional analysis,
which is supposed to begin with a specification of the adaptive problems
that have shaped human psychology. This suggestion instead requires
antecedent knowledge of our adaptations, which is used as a source of
information about the adaptive problems that have shaped them. But,
while the mind may not be a wholly black box any longer, it still remains
a dark shade of charcoal heather. So this suggestion is a nonstarter if, as in
fact is the case, we haven’t independently identified the psychological
characteristics that are candidate adaptations.
A second potential source of evidence about the adaptive problems faced
by our hunter-gatherer ancestors is the lifestyles of extant hunter-gatherer
populations. Evolutionary psychologists argue that hunter-gatherer popu-
lations that have been unaffected by agriculturalization and industrializa-
tion live much as Pleistocene hunter-gatherers did. In effect, they argue,
there has been a continuity of lifestyle from our ancestral hunter-gatherer
populations to extant hunter-gatherer populations. Consequently, they
conclude, the adaptive problems faced by extant hunter-gatherers should
be the same as those faced by our Pleistocene ancestors.
Adaptation 95
There are, however, two problems with this proposal. First, it is naive to
think that the social lives of extant hunter-gatherer populations have not
changed significantly in the last 10,000 years. As the anthropologist Robert
Kelly argues, “long before anthropologists arrived on the scene, hunter-
gatherers had already been contacted, given diseases, shot at, traded with,
employed and exploited by colonial powers, agriculturalists, and/or pas-
toralists. The result has been dramatic alterations in hunter-gatherers’
livelihoods. . . . There can be little doubt that all ethnographically known
hunter-gatherers are tied into the world economic system in one way or
another; in some cases they have been so connected for hundreds of years.
They are in no way evolutionary relics.”27
Second, as the anthropologist Laura Betzig points out, there is consider-
able variation in the lifestyles of extant hunter-gatherer populations.
Among these populations, the average daily caloric intake from foods gath-
ered by women ranges from 2 percent to 67 percent, average paternal care
ranges from ten minutes a day to 88 percent of the day, and mating systems
vary. This variation is not restricted to differences between sub-Saharan
hunter-gatherers and, for example, South American hunter-gatherers.
There is considerable such variation among African hunter-gatherers in
and around the region Evolutionary Psychologists believe was inhabited
by our ancestors. Which of these populations should we take as our model
of our Pleistocene ancestors? It makes a difference. If we choose a monog-
amous population, we will get a different picture of the adaptive problems
that shaped human sexual psychology than if we choose a polygynous
population. Similarly, a population that gets most of its calories from foods
provided by women will give us a different picture of adaptive problems
than a population that gets most of its calories from foods provided by
men. Since it isn’t at all clear which extant hunter-gatherer population
we should take as our model, we can’t reliably reconstruct the adaptive
problems faced by our Pleistocene ancestors through a study of extant
hunter-gatherers.
The third source of potential evidence about the adaptive problems faced
by our Pleistocene ancestors is comparative analysis, the study of species
related to ours. Evolutionary psychologists claim that we may come to
understand the adaptive problems faced by Pleistocene humans by study-
ing the adaptive problems faced by related nonhuman primates, since they
share with us a common ancestor.
But there are problems with this suggestion too. First, the most recent
ancestor common to us and our closest relative, the chimpanzee, lived 5
to 7 million years ago, a good 3 to 5 million years before the Pleistocene.
96 Chapter 3
The ways in which our lineage diverged from that of our chimpanzee rel-
atives before the Pleistocene, not to mention during the Pleistocene itself,
are profound. By the time early humans emerged in the Pleistocene, the
adaptive problems driving their evolution should have differed profoundly
from the adaptive problems facing nonhuman Pleistocene primates. And,
since nonhuman primates have been evolving since the Pleistocene as well,
the adaptive problems faced by contemporary nonhuman primates should
differ even more from the adaptive problems faced by early humans. As a
result, even our closest relative’s lifestyle holds few clues to the lifestyle of
our Pleistocene ancestors.
Second, nonhuman primate species differ considerably with respect to
foraging, parental care, and mating system. So we face the problem of
which nonhuman primate to take as our model of Pleistocene humans.
The lifestyle of our closest relatives won’t necessarily provide the best clues
to the adaptive problems that drove human psychological evolution.
Studies of 178 mammal and 151 bird species have shown that closely
related species are more similar with respect to morphological and life-
history traits than are distantly related species, but that degree of related-
ness isn’t correlated with similarity in behavioral traits.28 Rather, similarity
of ecological conditions is a more important determinant of similarity of
behavioral traits than is degree of relatedness.
Since human psychology would have evolved in response to selection’s
acting on behavior, the ecological conditions of our ancestors would have
been a primary determinant of the evolution of human psychological
traits. Consequently, rather than simply studying our closest primate rel-
atives, we need to study the primate species whose ecological conditions
most closely resemble those of our Pleistocene ancestors, for that species
will provide the best guide to the adaptive problems faced by our ances-
tors. But, without knowledge of the ecological conditions of our Pleis-
tocene ancestors, which is what we’re trying to obtain, we can’t determine
which nonhuman primates live in ecological conditions similar to those
of Pleistocene humans. Thus, the study of nonhuman primates is unlikely
to shed light on the adaptive problems that helped shape human
psychology.
These are all serious obstacles to the specification of adaptive problems
that constitutes the first step of evolutionary functional analysis. Any spec-
ification of those problems will indeed rank as pure guesswork on the part
of Evolutionary Psychologists, much in the way that Gould complains.
Of course, Evolutionary Psychologists will find this argument unduly
skeptical, much as Pinker finds Gould’s argument unduly skeptical. As
Pinker says, “what makes Gould so certain that our ancestors’ environment
Adaptation 97
lacked written language—the basis for his argument that reading is a span-
drel? Obviously it is the archeological record, which shows that writing is
a recent invention. . . . It is precisely such evidence that leads Evolution-
ary Psychologists to infer that the ancestral environment lacked agricul-
ture, contraception, high-tech medicine, mass media, mass-produced
goods, money, police, armies, communities of strangers, and other modern
features—absences with profound implications for the minds that evolved
in such an environment.”29
But a list of things we know did not affect human psychological evolu-
tion is a far cry from a positive account of the adaptive problems that did
shape human psychology. Of course, Evolutionary Psychologists will still
insist that they can say with some certainty what some of those adaptive
problems were, since many of them seem to follow immediately from
reflection on the demands of survival and reproduction. This is how Tooby
and Cosmides arrived at their laundry list of Pleistocene adaptive problems
discussed in chapter 2. As Tooby and Cosmides would argue, we can be
quite confident that Pleistocene humans would have had to “select mates
of high reproductive value” and to “induce potential mates to choose
them,” for example.
But here we encounter a problem concerning the “grain” at which these
adaptive problems are described. It is true that we can always be certain
that just about all sexually reproducing species face the adaptive problems
of selecting mates of high reproductive value and inducing potential mates
to become actual mates. These descriptions of adaptive problems are so
coarse-grained, however, as to be wholly uninformative about the selec-
tion pressures that act on a species. Consider, for example, what one need
do to attract a mate. Male bowerbirds must build ornately decorated
bowers, male hangingflies must offer captured prey as a nuptial gift, and
male sedge warblers must sing a wider repertoire of songs than other males.
The adaptive problem of attracting a mate thus takes very different forms
depending on the species. For male bowerbirds, the adaptive problem that
is describable at a very coarse grain as “attracting a mate” is, at a finer grain
of description, the problem of building an ornately decorated bower. The
finer-grained description is the one that captures the adaptive problem for
bowerbirds. So, the problem is that, while it’s true that we can be confident
that Pleistocene humans needed to attract mates, that coarse-grained
description doesn’t inform us of the specific form of adaptive problem that
helped shape human sexual psychology.
There is another way to think about this difficulty. Recall from chapter
2 that evolutionary functional analysis begins with a specification of an
adaptive problem and then proceeds to a task analysis of that problem. The
98 Chapter 3
task analysis involves breaking down the adaptive problem into a number
of subproblems, the solutions to which collectively constitute a solution
to the adaptive problem. One purportedly solves the adaptive problem by
solving the subproblems that constitute it, and psychological adaptations
purportedly evolved to solve these subproblems. Seen in this light, simply
knowing that Pleistocene humans needed to attract mates doesn’t inform
us of the subproblems that constituted that adaptive problem for Pleistocene
humans. And it is those more specific subproblems that adaptations would
have evolved to solve. In order to get the more fine-grained and informa-
tive description of the subproblems, however, we would need to have more
detailed knowledge of the lifestyles of our ancestors. And that’s knowledge
we simply don’t have.
There is even reason to think that we couldn’t possibly have any knowl-
edge of the finer-grained adaptive problems faced by our Pleistocene ances-
tors without knowing something about their psychology. The reason is that
the finer-grained adaptive subproblems faced by a species are not inde-
pendent of the morphology and psychology of that species. Indeed, the
morphology and psychology of a species determine which aspects of the
environment are adaptively relevant to the species. As the biologist Richard
Lewontin says: “The bark of trees is part of the environment of a wood-
pecker, but the stones lying at the base of the tree, even though physically
present, are not. On the other hand, thrushes that break snail shells include
the stones, but exclude the tree from their environment. If breaking snail
shells is a ‘problem’ to which the use of a stone anvil is a thrush’s ‘solu-
tion,’ it is because thrushes have evolved into snail-eating birds, whereas
woodpeckers have not. The breaking of snails is a problem created by the
thrushes, not a transcendental problem that existed before the evolution
of the Turdidae.”30 Which features of the environment pose adaptive prob-
lems for a species, in other words, depends on the “equipment” of the
species, since that equipment will antecedently be attuned to certain fea-
tures of the environment but not to others. And features of the environ-
ment to which a species’ equipment is not attuned will generally be
irrelevant to that species’ subsequent evolution by natural or sexual
selection.
This means that, without knowledge of the morphology and psychology
of a species, we can never specify the adaptive problems confronting it
with anything but the most coarse-grained descriptions, and these will not
inform us of the specific selection pressures acting on it. So, in order to
identify the selection pressures that helped shape human psychology, we
would need to know something about ancestral human psychology. For
Adaptation 99
our ancestors’ motivational states and cognitive processes would have been
selectively responsive to certain features of the physical and social envi-
ronments, and only those features would have affected subsequent adap-
tive evolution of early human psychology. At this point we again collide
with our ignorance of our early ancestors. And, given that psychologies
don’t fossilize, this ignorance is likely intractable.
So far I have argued that, without antecedent knowledge of our psy-
chological adaptations, we’re unlikely to ever be in a position to have the
evidence required to make the identification of ancestral human adaptive
problems anything more than pure guesswork. This is a problem for Evo-
lutionary Psychology, which proposes to infer the nature of our psycho-
logical adaptations from the adaptive problems that shaped them. But my
argument so far has merely suggested epistemic obstacles to evolutionary
functional analysis, limitations of our knowledge that prevent us from
identifying ancestral adaptive problems. There are, however, more princi-
pled reasons for being deeply skeptical of evolutionary functional analysis
as a method of discovering our psychological adaptations—reasons for
being skeptical even if all the epistemic obstacles I’ve mentioned could be
overcome. For evolutionary functional analysis presupposes that there
were relatively stable adaptive problems during human evolution to which
selection very slowly tailored psychological solutions. The stability of these
problems is supposedly reflected in the design of our psychological adap-
tations, and consequently we are supposed to be able to get a handle on
the design of those adaptations by identifying the stable problems for
which they were designed. But there are reasons for thinking that there
were no stable adaptive problems driving the majority of human psycho-
logical evolution.
The most widely accepted account of the evolution of human intelli-
gence, which is also accepted by Evolutionary Psychologists, is the social
intelligence hypothesis (also known as the Machiavellian intelligence
hypothesis). According to this hypothesis, most of human psychological
evolution was driven by the human social environment, not by the phys-
ical environment. For the compelling problems whose solutions required
true intelligence were those involved in social life: competing with
members of the same sex for mates, competing with others for resources,
recognizing and responding appropriately to deception and hostility, pro-
tecting and feeding offspring, and so on. In other words, the majority of
adaptive problems that drove human psychological evolution were posed
by other humans, who were themselves responding adaptively to the
demands of human social life. The problems involved in competing for a
100 Chapter 3
mate were posed by the preferences of the opposite sex and by the behav-
ior of competitors of one’s own sex. The problems involved in dealing with
hostilities were posed by the wants, temperaments, and behavioral ten-
dencies of other humans. In short, the need to better one another in social
competition wielded the crop that drove the evolution of our intelligence.
Put crudely, the primary driving force of human psychological evolution
was human psychology.
As the philosophers of biology Kim Sterelny and Paul Griffiths point out,
this state of affairs set up an “arms race” in human psychological evolu-
tion. In an evolutionary arms race, there is progressive modification not
only in the solutions to adaptive problems, but in the adaptive problems
themselves. “Improving” solutions are continually matched by ever more
“difficult” problems, with evolution in either precipitating the evolution
of the other. Many arms races are between predators and prey. As preda-
tors evolve to get better at catching their prey, this creates a selection pres-
sure on the prey to become better at escaping the predator, which then
creates a selection pressure on the predator to catch the more adept prey,
and so on. As cheetahs become faster runners, gazelles are selected to
become faster runners; but their running faster selects for faster cheetahs,
which then selects for even faster gazelles, and so on. Arms races are
common in predator-prey and host-parasite interactions, but they can also
take place within a species or within a sex whenever conspecifics compete
for the same resources.
This latter kind of evolutionary arms race would have characterized all
of human psychological evolution that was responsive to the human social
environment. For any evolution in human psychology would have
changed the psychological composition of the population as a whole,
which would have then created a new adaptive problem for human psy-
chology to adapt to. But this means that, as human psychology evolved,
the adaptive problems driving human psychological evolution would have
evolved in lockstep, so that there would have been no stable adaptive prob-
lems driving human psychological evolution. As Sterelny and Griffiths put
it, “when evolution is driven by features of the social structure of the evolv-
ing species, evolution transforms the environment of the evolving organ-
ism. The evolution of language, of tool use, and of indirect reciprocity are
not solutions to preexisting problems posed to the organism. There are no
stable problems in these domains to which natural selection can grind out
a solution. The ‘adaptive problem’ is always being transformed in an arms
race. As we evolve to detect cheaters, these honesty-mimics evolve better
and better imitations of a trustworthy and honest face.”31
Adaptation 101
Note that the force of this argument is somewhat sensitive to the grain
at which adaptive problems are described. It may be true, for example, that
the adaptive problem of attracting a mate has remained stable throughout
human evolutionary history. But, as I argued earlier, this very coarse-
grained description of the adaptive problem doesn’t inform us of what
one must do to attract a mate. A finer-grained description of the adaptive
problem would specify the precise form the task takes. The adaptive prob-
lems that would have exhibited instability in an evolutionary arms race
are those specified by such finer-grained descriptions. In other words, while
the coarse-grained adaptive problem of attracting a mate may have
remained stable, the finer-grained subproblems involved in the attracting
would have undergone change. Thus, the adaptive subproblems that
shaped human psychology would have been very fluid throughout human
evolution.
But evolutionary arms races are not the only phenomenon that produces
fluidity of adaptive problems. Organisms often actively construct their
niches. When they do, they modify the selection pressures that drive their
evolution. And the more actively they construct their niches, the more
fluid are the selection pressures. An obvious example of niche construc-
tion is dam-building by beavers. But niche construction can take much
simpler forms. Digging a burrow to escape the unusual cold of a severe
winter is an act of constructing a new niche in which the selection pres-
sures differ from those outside the burrow.
Humans have been supreme niche constructors. The development of
agriculture, for example, greatly altered human niches. An acre of farmed
land was able to feed many more mouths than an acre of wild land from
which foods were gathered. As a result, population densities increased in
agricultural areas and birthrates among agriculturists increased as well. At
the same time, however, the transition to a diet composed primarily of
starchy foods brought widespread malnutrition, and the higher population
densities allowed diseases such as cholera to spread quickly to large
numbers of people. The development of industry brought about similar
changes in human niches, and developments in medicine have continu-
ally altered the toll of disease on survival and, as a consequence, oppor-
tunities to reproduce. On the psychological side, techniques of teaching,
whether skill or information based, have altered the cognitive niche in
which humans develop, and the recent development of information tech-
nologies is radically altering the cognitive niche to which future genera-
tions will have to adapt. These are some spectacular ways in which humans
have modified their niches, but there are a multitude of less spectacular
102 Chapter 3
ways that we have done so, ranging from methods of food preparation and
preservation (think of pasteurization, for example) to methods of shelter
construction, from methods of contraception to systems of organized edu-
cation. And we really don’t know precisely how niche construction among
our prehistoric ancestors may have continuously altered the adaptive prob-
lems they faced and helped shape and reshape the direction of human psy-
chological evolution.
The evolutionary biologists Kevin Laland, John Odling-Smee, and
Marcus Feldman offer the following simple nonpsychological example of
how easily, and inadvertently, human niche construction can change selec-
tion pressures. A West African population “increased the frequency of a
gene for sickle-cell anemia in their own population as a result of the indi-
rect effects of yam cultivation. . . . [They] traditionally cut clearings in the
rainforest, creating more standing water and increasing the breeding
grounds for malaria-carrying mosquitoes. This, in turn, intensifies selec-
tion for the sickle-cell allele, because of the protection offered by this allele
against malaria in the heterozygotic condition.”32 There can be little doubt
that similarly subtle niche construction would have affected the direction
of human psychological evolution.
Further, as populations of our direct ancestors diverged around 50,000
years ago (or earlier, depending on whose account you believe) and began
to occupy different parts of the globe, the different habitats they encoun-
tered would have prompted different behavioral responses, at least some
of which would have involved niche construction. Because niche con-
struction would have taken different forms in the different habitats occu-
pied by these early human populations, the adaptive problems faced by
different populations would have diverged, and the problems would have
been very fluid in each different habitat. And diversity in the ways early
human populations formed environmental niches would have com-
pounded the instability of the adaptive problems out ancestors faced.
Because of evolutionary arms races and niche construction, it is doubt-
ful that there were many stable adaptive problems to which humans
evolved psychological solutions. But, if there weren’t stable adaptive prob-
lems driving human psychological evolution, there are no stable adaptive
problems whose identification will give us a handle on the nature of our
psychological adaptations. Once again, the necessary first step in evolu-
tionary functional analysis—the identification of adaptive problems to
which we evolved psychological solutions—appears unachievable.
So far I have argued that there are severe problems with the idea that we
can identify the adaptive problems to which selection slowly tailored
Adaptation 103
If it’s true, however, that the real proof of evolutionary functional analy-
sis is in the pudding of empirical evidence for the psychological mecha-
nisms it predicts, then why did I just argue extensively that evolutionary
functional analysis is an unreliable method for the discovery of our psy-
chological adaptations? Why not just turn directly to an examination of
the empirical evidence of specific predicted psychological mechanisms and
leave it at that?
The reason has to do with the grounds on which we are justified in
accepting claims about new discoveries in science. Here is one typical
pattern of justification for a claim that a new phenomenon has been dis-
covered. We begin with some theoretical principles that we have inde-
pendent reasons for accepting. We discover that these principles, perhaps
together with some independent empirical data, entail the existence of a
previously undiscovered phenomenon. We then reason that, if that phe-
nomenon indeed exists, we should obtain certain results if we perform a
particular kind of experiment. We then conduct the experiment to see if
we obtain those results. If we do obtain those results, then we can be con-
fident that we have discovered a new phenomenon.
Consider how this pattern would be ideally exemplified by evolutionary
functional analysis. We begin with some general theoretical principles
regarding the evolutionary process together with some claims about the
selection pressures operative during human evolutionary history, which we
should have independent reasons for accepting. We deduce from these the
hypothesis that one or both of the sexes should have evolved a particular
psychological mechanism. We then reason that, if people possess that
psychological mechanism, we should obtain certain results if we subject
people to a particular battery of tests or surveys. If we do obtain those
results, we can be confident that people possess the relevant psychologi-
cal mechanism—that is, we conclude that our hypothesis has been
confirmed.
But what if we don’t obtain the right experimental results or the results
aren’t entirely clear? A popular image of science is that a theory makes a
prediction, the scientists check to see whether the prediction is accurate,
and they jettison the theory if it’s not. But this popular image is wrong.
Science does not actually work that way, and it would not be a rational
enterprise if it did. When theoretical principles are independently justified
and have withstood the test of time, we’re often more justified in doubt-
ing the empirical data that contradict them or their entailments than in
doubting the theoretical principles themselves. When your experiments in
high-school chemistry class failed to produce results that exactly matched
106 Chapter 3
those predicted by the textbook, you didn’t reject the relevant aspects of
chemical theory, but concluded that you had not done the experiment
correctly at every stage. And, if your results came out in the ballpark of the
predicted values, you considered the experiment a success and took it as
further confirmation of the theoretical principles in the textbook.
Unfortunately, empirical data in science rarely speak loudly and
unequivocally for or against a hypothesis. This is especially true of empiri-
cal data, like those in Evolutionary Psychology, that are gathered by forced-
choice questionnaires, or other social psychological instruments, and then
analyzed using statistical procedures. Data are often messy as a result of
“noise” in the experimental procedure, factors that interfere with obtain-
ing clean and unequivocal results that clearly confirm or disconfirm a
hypothesis. Consequently, it often requires some work in order to know
what the data are saying about a hypothesis.
Given the frequent messiness of data, how confident can we be in
hypotheses that we derive from independent theoretical principles and
assumptions? That depends on how confident we can be in either the the-
oretical principles that entailed the hypothesis or the empirical evidence
gathered to test it. If we can be very confident in the theoretical principles
from which we derive a hypothesis, then we can be confident in that
hypothesis even if our empirical evidence for it isn’t very good. In this
case, as with high-school chemistry experiments, we can attribute a failure
of exact match between predicted and obtained results to experimental
“noise,” and we can accept results that are merely in the ballpark as fair
confirmation of our hypothesis. On the other hand, if we can’t be confi-
dent in the theoretical principles from which we derive a hypothesis, then
our degree of confidence in the hypothesis is going to derive almost
entirely from the strength of the empirical evidence. In this case, to be
confident in our hypothesis the empirical evidence does need to say very
clearly that the hypothesis is accurate.
The moral of this digression is that confidence in a hypothesis can
derive from two sources—the theoretical principles that entail the hypoth-
esis and the empirical evidence gathered to directly test it. In the ideal case,
we can be confident in both the theoretical principles (when they have a
lot of independent evidence in their favor) and the empirical evidence,
for then we can be maximally confident that our hypothesis is accurate.
But we can nonetheless enjoy a lower degree of confidence in our hypoth-
esis if we can be confident of either the theoretical principles that entail
it or the empirical evidence for it, even though the other is somewhat
dubious.
Adaptation 107
As we have seen, one thing, though not the only thing, that gets evolu-
tionary functional analysis into trouble is its attempt to reverse engineer
the mind from the vantage of our prehistoric past. For it attempts to reverse
engineer the mind by starting from the adaptive problems faced by our
Pleistocene ancestors, and there are severe obstacles to identifying those
adaptive problems. But why do Evolutionary Psychologists think that we
need to begin the process of reverse engineering the mind from our insuf-
ficiently knowable Pleistocene past? The reason is that Evolutionary
Psychologists believe that our psychological adaptations were designed
during the Pleistocene to solve the problems of hunter-gatherer life and
that there has been no significant psychological evolution since the Pleis-
tocene. Consequently, Evolutionary Psychologists believe that the evolved
structure of the mind “reflects completed rather than ongoing selection.”34
The psychologist Henry Plotkin calls this “the thesis of ancient prove-
nance,” since it claims that our minds were fixed by “ancient” selection
processes. In this section, I will argue that the thesis of ancient provenance
is unjustified.
Recall, from chapter 2, why Evolutionary Psychologists think that our
psychological adaptations are adapted to Pleistocene, rather than modern,
environments. The sole reason is a general theoretical argument about how
108 Chapter 3
long it takes for complex adaptations to evolve. The 10,000 years since the
end of the Pleistocene, they argue, “is only a small stretch in evolutionary
terms, less than 1% of the two million years our ancestors spent as Pleis-
tocene hunter-gatherers. For this reason, it is unlikely that new complex
designs—ones requiring the coordinated assembly of many novel, func-
tionally integrated features—could evolve in so few generations.”35 The
four hundred generations since the end of the Pleistocene are too few for
genetic evolution significant enough to “assemble” complex psychological
modules that are adapted to modern environments. Thus, our psycholog-
ical adaptations must have evolved, instead, during the Pleistocene and
hence must be adapted to the Pleistocene conditions under which they
evolved.
There are, however, several problems with this argument. First, the issue
is not whether “new complex designs” that require the “coordinated assem-
bly” of many features could have evolved in the 10,000 years since the
Pleistocene. Without doubt, selection could not build a human mind from
scratch in a mere four hundred generations. But, from the fact that a “new
complex design” could not have evolved since the Pleistocene, it doesn’t
follow that the psychological adaptations of contemporary humans are
identical to those of our Pleistocene ancestors. For the issue is whether old
complex designs, which evolved during the Pleistocene, could have been
modified by selection in the last 10,000 years. Since the argument doesn’t
address this possibility, it fails to show that our psychological adaptations
must have remained adapted to Pleistocene conditions.
Second, as the evolutionary biologist David Sloan Wilson points out, “it
makes no sense to express evolutionary time as a proportion of the species
history (e.g., 1%). If the environment of a species changes, the evolution-
ary response will depend on the heritability of traits [roughly, the propor-
tion of variation in traits that is due to genetic variation], the intensity of
selection, and the number of generations that selection acts. The number
of generations that the species existed in the old environment is irrelevant,
except insofar as it affects the heritability of traits.”36 In other words, it
doesn’t matter whether a lineage spends only 1 percent of its evolution-
ary history in a new environment, what matters is what kinds of change
occur during that 1 percent of its evolutionary history. Thus, Wilson con-
cludes, “rather than marvelling at the antiquity of our species, we should
be asking what kinds of evolutionary change can be expected in 10, 100,
or 1000 generations.”
Finally, the thesis of ancient provenance does indeed underestimate the
kinds of evolutionary change that may have occurred since the end of the
Adaptation 109
change. So, if the subtasks involved in acquiring food, for example, have
changed, there has been selection pressure for evolution in the Darwinian
algorithms involved in acquiring food.
To see how such subtasks have changed, consider that in a hunter-
gatherer population a man must find and kill an animal in order to acquire
food for a (potential) mate and offspring. But what a man must do to
acquire food in an agricultural population differs considerably. And what
a man must do to acquire food within a barter system, or as a wage laborer
in a capitalist system, differs even more. Wage-Laborer Man needs to obtain
a job (often in a large population characterized by intense competition for
jobs), perform the tasks required of that job, manage his wages, locate a
purveyor of food, and negotiate the rules required to obtain the food from
the purveyor (then navigate the traffic back home to the wife and kids).
So, even if Hunter Man and Wage-Laborer Man faced the same coarse-
grained adaptive problem of acquiring food for a mate and offspring, the
subtasks required of each to solve this problem differ significantly. So any
Darwinian algorithm executed by Hunter Man would not be effective in
acquiring food in the world of Wage-Laborer Man. Consequently, changes
in human social structures would have created intense selection pressure
favoring changes in any psychological mechanisms that Pleistocene
males may have evolved to solve the adaptive problem posed by female
preference.
This is just one example, among many possible, of how changes in
human social structures since the Pleistocene would have changed the
selection pressures on psychological mechanisms—even if we assume that
the very coarse-grained adaptive problems faced by humans have not
changed. Such changes in social structures are forms of niche construction,
and they rapidly change the adaptive problems a population faces. As these
changes in social structures occurred, however, humans would have
needed to compete with one another within these changing environments
for the resources essential to survival and reproduction, including members
of the opposite sex. These forms of intraspecific and intrasexual competi-
tion within rapidly changing environments would have in turn spawned
evolutionary arms races with respect to psychological solutions to the
problems faced. Thus, it is safe to conclude that radically changed envi-
ronments since the Pleistocene have created strong selection pressures
favoring psychological evolution.
But has there been sufficient time since the Pleistocene for an evolu-
tionary response to these environmental changes? Let’s be clear about the
question, since Evolutionary Psychologists typically are not. The question
Adaptation 111
is not whether there has been enough time since the Pleistocene for human
populations to evolve minds that are adapted to twenty-first-century envi-
ronments. The question, instead, is whether there has been sufficient time
for modification to whatever psychological adaptations Pleistocene humans
possessed.
Consider first that there are clear cases of post-Pleistocene adaptive
evolution in physiological and morphological traits. Laland, Odling-Smee,
and Feldman point out that “the persistent domestication of cattle, and
the associated dairying activities, did alter the selective environments of
some human populations for sufficient generations to select for genes that
today confer greater adult lactose tolerance.”37 Indeed, they argue, niche
construction typically accelerates the pace of evolution as successive gen-
erations continually modify the sources of selection acting on themselves
and subsequent generations. There is little doubt that human niche con-
struction has been radical in this regard. And there is no reason to think
that such evolution must have been confined to physiological and mor-
phological traits. In fact, the fundamental assumption of Evolutionary Psy-
chology is that selection has shaped both mind and body. So, any evidence
of post-Pleistocene evolution in physiological and morphological traits
should create a presumption that there has been evolution in psychologi-
cal traits as well.
Moreover, the idea that human psychological adaptations cannot have
evolved since the end of the Pleistocene depends on a false assumption
about the rate at which natural selection can alter traits in a population.
Recent work by the biologist David Reznick and his colleagues has shown
that evolution by natural selection can occur very rapidly. Reznick and his
colleagues split populations of guppies living in high-predation environ-
ments, leaving a part of each population in its high-predation environ-
ment and moving the other part to a low-predation environment. They
found that life-history traits of the transplanted guppies evolved signifi-
cantly within a mere eighteen generations. The descendants of the trans-
planted guppies matured to a larger size and achieved reproductive
viability at a later age than the nontransplanted guppies. More signifi-
cantly, they produced fewer litters, with fewer and larger offspring in each
litter, and they allocated less of their total resources to reproduction during
their early reproductive lives. And Reznick and his colleagues were able to
identify both the genetic basis of this change and the mechanism by which
selection drove it (namely, differential mortality by predation). If this
much evolution can occur in eighteen generations, the nearly four
hundred human generations since the end of the Pleistocene has certainly
112 Chapter 3
So far we have followed a trail of issues that arose out of problems with
Gould’s critique of Evolutionary Psychology. Now I want to consider one
other issue before leaving the discussion of Evolutionary Psychology’s
focus on psychological adaptations and its theoretical treatment of them.
For perhaps the strongest claim that Evolutionary Psychologists make
about human psychological adaptations is that they are, of necessity, uni-
versal in the species. As Tooby and Cosmides say, “the psychic unity of
humankind—that is, a universal and uniform human nature—is necessar-
ily imposed to the extent and along those dimensions that our psy-
chologies are collections of complex adaptations.”38 This is allegedly
demonstrated by the argument from sexual recombination discussed in
chapter 2, which purports to show that the genes underlying complex
adaptations must be universal in our species and, hence, that the adapta-
tions they underlie must be universal as well. If the argument is correct,
genetic differences among humans never produce differences in psycho-
logical adaptations—that is, there are no stable psychological polymor-
phisms in human populations.
There is, nonetheless, obvious variation in psychological phenotypes in
human populations, and a significant portion of that variation shows signs
of being adaptive. For example, the “opportunistic” and “investing” female
Adaptation 113
two individuals had different adaptations, they would also have to differ
with respect to hundreds or thousands of genes, which, the argument
purports to show, could not possibly be transmitted collectively intact to
offspring. But the genetics of sex determination in humans shows the
problem with this assumption. The reproductive organs of each sex clearly
constitute a complex suite of functionally coordinated adaptations: in
males, the testicles, vas deferens, seminal vesicles, prostate, penis, and so
on; in females, the ovaries, fallopian tubes, uterus, vagina, and so on. Yet
the difference between these coordinated suites of adaptations is the result
of a single-gene difference, the SRY gene on the Y chromosome. Males and
females don’t have to differ in hundreds or thousands of genes to differ
profoundly in their reproductive anatomy.
Of course, male reproductive anatomy is not built by the SRY gene alone.
Both male and female reproductive organs do, indeed, require the coordi-
nated interaction of hundreds or thousands of genes in order to develop,
and the SRY gene produces male reproductive anatomy only in interaction
with hundreds or thousands of genes that males and females share. But,
against the genetic background shared by males and females, a single
genetic switch initiates a cascade of developmental events that results in
a coordinated suite of male reproductive adaptations, which differs pro-
foundly from the coordinated suite of female reproductive adaptations.
If such differences in whole suites of adaptations can result from a single
genetic switch, then many less profound differences in adaptations could
result from alternative forms of a genetic switch. In such cases, alternative
genotypes at a single locus would causally interact with other, shared genes
to shunt development down alternative pathways leading to different
adaptive phenotypes. Since the argument from sexual recombination fails
to show that there cannot be stable polymorphisms produced by alterna-
tive forms of a genetic switch, the argument fails to show that adaptations
must be species universals. Consequently, the argument fails to show that
no psychological variation in human populations results from genetic
variation maintained by selection. In fact, single-locus polymorphisms
could underlie a significant amount of adaptive psychological variation in
human populations.
Tooby and Cosmides actually acknowledge that the differences in adap-
tations between males and females result from a single genetic switch, and
they grant that single-gene adaptive differences are an exception to the
argument from sexual recombination. But they misinterpret the way that
single-gene differences can function. As Tooby and Cosmides say: “An
adaptation wholly coded for by a single gene can survive this filter [of
116 Chapter 3
of sex determination was most effective had a lot to do with the role
of sex determination within the development and life history of early
mammals. In general, then, whether a genetic switch or adaptive plastic-
ity is the more effective system for determining alternative phenotypes
depends on which phenotypes we’re talking about, the role they play in
the life history of the organism, and the role their development plays in
the overall development of the organism. Whether adaptive plasticity or
a genetic switch is more effective can’t be determined for all phenotypes
in general by a simple a priori argument.
Considerations of development and life history show that the effective-
ness of adaptive plasticity or a genetic switch in producing particular alter-
native phenotypes is relative to a number of other factors “internal” to the
organism. Sometimes adaptive plasticity is more effective, and can be
favored by selection, and at other times a genetic switch is more effective.
But which system is more effective is also relative to factors “external” to
the organism. Tooby and Cosmides’ argument presupposes that a poly-
morphic genetic switch will always be less adaptive to the “local situation”
than a mechanism of adaptive plasticity that relies on cues from the envi-
ronment. But this ignores the question of precisely which aspects of the
“local situation” are being adapted to.
Some “local situations” can favor a genetic switch. For example, when
the fitness of a phenotype depends on its frequency in a population, then
the environment to which that phenotype is adapted is one containing an
evolutionarily stable ratio of alternative phenotypes. Recall the Hawk-Dove
game. If the evolutionarily stable ratio is 75 percent Hawks and 25 percent
Doves, then Dove and Hawk are each well adapted to an environment con-
taining 75 percent Hawks and 25 percent Doves. In such circumstances,
a genetic polymorphism maintained by frequency-dependent selection
is just as effective as a system of adaptive plasticity for maintaining an
evolutionarily stable ratio of alternative phenotypes. There would be no
improving on a genetic polymorphism by “deciding” which phenotype to
possess based on information gathered from the environment, since those
“decisions” would still have to result in the same stable ratio that is pro-
duced by the genetic polymorphism. So adaptive plasticity will entail no
greater benefit in such cases.
In addition, a system for “choosing” the phenotype based on conditions
of the environment requires some method of ascertaining the condition
of the environment. This is particularly apparent in cases like sex “choice”
in fishes. If sex is “chosen” based on the sex ratio in the local population,
there must be some mechanism that monitors the local sex ratio and then
118 Chapter 3
The first thing to note is that it’s not a psychology at all. This follows rel-
atively straightforwardly from facts that Evolutionary Psychologists accept.
To see why, suppose that one psychological adaptation is a mechanism
dedicated to detecting cheaters in social exchanges, as Evolutionary Psy-
chologists claim. First, the cheater-detection mechanism, like all psycho-
logical adaptations, will develop only in individuals exposed to the right
developmental environments—in particular, those developmental envi-
ronments encountered by most humans throughout the period in which
the cheater-detection mechanism evolved. Individuals exposed to devel-
opmentally impoverished environments, which lack the environmental
cues or metabolic resources necessary for the proper development of the
mechanism, will not develop the cheater-detection mechanism. Second,
not all individuals are in fact exposed to the right developmental envi-
ronments; some individuals develop in impoverished environments.
Consequently, some individuals’ psychologies include a cheater-detection
mechanism, whereas other individuals’ psychologies don’t. Since this will
be the case for every psychological adaptation, there simply is no
psychology common to all humans.
This argument relies on differences in developmental environment that
result in some individuals’ lacking an adaptation that other individuals
possess. But psychological differences can also be produced by differences
in environmental cues, which shunt psychological development down
alternative pathways, and by genetic switches, which produce differences
in psychological adaptations. Since psychological characteristics are
phenotypic, individuals who have different psychological phenotypes, for
any of these reasons, ipso facto have different psychologies. Consequently,
individual differences that are due to different environmental cues during
development or to different forms of a genetic switch are psychological
differences, which are irreducible to some underlying psychological
commonality. So there simply is no universal human mind.
To a certain extent this is acknowledged by Evolutionary Psychologists.
The acknowledgment is implicit in Charles Crawford’s distinction between
innate and operational adaptations. Operational adaptations, according to
Crawford, are “the anatomical structures, physiological processes, and psy-
chological processes that develop because of interactions with the envi-
ronment and that actually do the work of helping the organism survive
and reproduce.”50 Since differences in developmental environment can
produce differences in operational adaptations, Crawford claims that what
we really share are innate adaptations, where an “innate adaptation is the
Adaptation 123
only after the species had evolved through that environment. Thus, the
“developmental program” can only contain information about how to
develop in an environment the species has already encountered in its
evolutionary history.
But surely there have been some radical changes in developmental envi-
ronments at particular points throughout the evolutionary history of our
species. That is, there must have been at least some occasions on which a
population encountered an evolutionarily novel developmental environ-
ment. Since humans did not go extinct, some individuals must have devel-
oped adaptively in that evolutionarily novel environment. Further, it is
implausible to suppose that, on such occasions, only a random mutant or
two survived the radical environmental change. Instead, most individuals
were able to develop in the novel environment. But that means that most
individuals were able to develop adaptive phenotypes in an environment
for which their developmental programs could not possibly have con-
tained a rule specifying an adaptive response, since developmental pro-
grams cannot contain rules for evolutionarily novel environments.
If the foregoing is correct, then there have been some occasions in our
evolutionary history on which individuals have developed adaptive phe-
notypes without the need for a rule specifying “how each possible state of
the developmental environment is to be responded to, if encountered.”58
But, if adaptive phenotypes can develop in some environments without
the need for a rule, they can develop in all environments without the need
for a rule. This means that the notion of a set of rules embodied in a devel-
opmental program guiding development is idle; it plays no genuine role
in explaining how development occurs. When we develop differently, it is
not because something that is the same in us simply responds differently
in a programmed way to differences outside us.
Not only is the notion of a “program” guiding development idle in
explaining how development occurs, the goal of discovering such a
program has been altogether abandoned by researchers in genetics. The
geneticist Sydney Brenner has spent decades studying the genetics and
development of a simple nematode, Caenorhabditis elegans, which has only
959 cells in its body. Given the simplicity of this organism and the regular
way in which it develops, Brenner believed it to be an ideal research subject
and expected that study of its development would quickly reveal the
“program” guiding it. Even though Brenner and his team discovered the
complete developmental history of every cell in the nematode’s body, and
every connection in its nervous system, nowhere did they discover an
orderly pattern of developmental events that gave the appearance of
126 Chapter 3
reasoning instincts. As Tooby and Cosmides say, “what is special about the
human mind is not that it gave up ‘instinct’ in order to become flexible,
but that it proliferated ‘instincts’—that is, content-specific problem-solving
specializations.”9
It is important to consider precisely how, according to this picture, these
instincts-as-modules “proliferated” in the human mind. First, according to
Evolutionary Psychologists, each module (or “brain circuit”) was “added”
to the mind at some point in human evolutionary history and then sub-
sequently “shaped” by selection to be highly effective at solving adaptive
problems in its proprietary domain. This “shaping” consisted in selection’s
retaining successive modifications to the existing design of the module,
each of which made the module more effective in solving its adaptive prob-
lems. Second, since Evolutionary Psychologists claim that each module was
shaped by selection pressures specific to the adaptive-problem domain
in which it specializes, each module must have evolved independently of
every other. This does not mean that no two modules evolved during the
same period of human evolution, but only that separate modules evolved
in response to independent selective forces and were functionally modifi-
able independently of other modules.
These two claims entail that each module evolved as a result of numer-
ous mutations over human evolutionary history, each of which added or
modified a specialized “brain circuit,” and all of which were preserved
by selection as the gene complex that regulates the development of the
module. In short, each module must be the developmental product of its
own gene complex, which, as Pinker puts it, is that part of the “genetic
program” that “specifies” the “basic design” of that “mental organ.”10
This is not to say that each module’s gene complex has developmental
effects on that module alone; genes often have multiple developmental
effects. But, if modules were added and modified by selection independ-
ently of one another, it is highly unlikely that any single mutation would
have added more than one module to the brain or that any single
mutation would have produced a beneficial modification to more than one
module. Think of this in terms of Evolutionary Psychologists’ favored com-
parison of modules to bodily organs: It is highly unlikely that any single
mutation would “add” both a kidney and a liver to human anatomy or
enhance the functional effectiveness of both the stomach and the heart.
Thus, if the picture proposed by Evolutionary Psychologists is accurate, as
modules proliferated in the human brain and became shaped by selection,
the number of genes for features of the human brain would have had to
130 Chapter 4
tional DNA in mice. And given the discussion in chapter 3 of how single-
gene differences can produce significant differences in adaptations, it is
possible for significant phenotypic differences to be produced by relatively
small genetic differences. But these are merely reasons for thinking
that we shouldn’t expect to find a perfect cross-species correlation between
genome size and brain complexity. The remarkable thing is that there is
no correlation whatsoever. Species with very complex brains simply don’t
have more “genetic information” available for building those brains than
do species with relatively simple brains.
But, if the complex structure of the human brain isn’t “genetically
specified,” how does it develop? If each special-purpose brain circuit in an
adult brain isn’t constructed in accordance with a “genetic program” that
“specifies its design,” how do such complex and functionally effective
circuits develop?
From the time the human brain begins to develop in utero, at about
twenty-five days after conception, it increases by a remarkable 250,000 cells
per minute, and this rate of cell production continues until birth. The pro-
duction of these cells takes place in two different “zones,” the ventricular
and the subventricular. The cells that make up the evolutionarily oldest
parts of the brain are produced in the ventricular zone. These are the cells
that make up the midbrain and the limbic system, which are regions of
the brain controlling motor coordination, sexual response, and emotion
(such as the fear response). In contrast, the cells produced in the subven-
tricular zone make up the evolutionarily most recent addition to the brain,
the neocortex, which carries out the “higher” cognitive functions.
These two cell-production zones also have different modes of operation.
As the ventricular zone produces cells, newly formed cells simply push
older cells outward from the zone of production. So cell production in the
ventricular zone builds brain structures by adding cells to already existing
layers of cells, in much the way that other parts or organs of the body are
formed. All the building of these structures is done by cells working at very
close quarters, and it appears to be under rather rigid genetic control. While
this is the method for building the evolutionarily older parts of the brain,
the development of the evolutionarily more recent cortex is a very differ-
ent matter. Since Evolutionary Psychologists claim that their postulated
modules are complex information-processing mechanisms, which execute
sophisticated “Darwinian algorithms” in solving adaptive problems, these
modules would most likely be found in the cortex. So cortical develop-
ment will be the principal focus in what follows, although there will be a
brief discussion later of subcortical structures.
132 Chapter 4
Unlike the cells produced in the ventricular zone, the cells produced
in the subventricular zone, which make up the cortex, must actively
“migrate” to their final destinations in the brain, wending their way
through a thicket of other cells. Once they reach their final destinations,
they grow branching axons that form connections with other cells. Much
of this development is also undoubtedly under rather rigid genetic control,
since it is uniform across individuals who develop in very different envi-
ronments. But this is true only of the development of the major structures
of the brain and their primary subdivisions; it is not generally true of the
development of the more fine-grained brain structures that perform spe-
cialized cognitive functions. Indeed, whereas the major cortical structures
and their primary subdivisions are in place at birth or shortly thereafter,
the functionally specialized circuits that characterize an adult brain are not.
But the transition from infant to adult brain is not a matter of adding fine-
grained specialized circuits to the developing brain in the way that adap-
tations such as teeth and breasts are added to the developing organism in
time to meet new age-specific adaptive demands. In fact, since the adult
brain contains fewer cells and connections than the infant brain, the tran-
sition is actually a matter of subtracting connections and cells. And the
primary mechanism of subtraction is cell competition and cell death.
For a simple illustration of this mechanism, consider the development
of motor neurons, which transmit information from the central nervous
system to muscle fibers. In an adult, each motor neuron is connected to
at least one, though usually more than one, muscle fiber, yet each muscle
fiber is innervated by only one motor neuron. But this one-many con-
nection pattern is not present in the developing embryo. Instead, many
motor neurons are connected to many muscle fibers in a vast proliferation
of connections that is not present in the adult. Over the course of devel-
opment, the ends of most of the motor neurons retract until only one
neuron controls each muscle fiber. The best explanation to date of the
mechanism for this sort of neuronal pruning is that motor neurons compete
with each other for sole activation rights of muscle fibers. The neuron with
the strongest activation wins, and the cells that lose the competition lose
their connections to other cells and eventually die.
This process of pruning the overabundance of connections and cells
forms the brain circuits that carry out our specialized cognitive functions.
Central to this process is cell competition and death, in which the cells
with the strongest patterns of innervation retain their connections and the
other cells die. And two types of innervation are relevant in the competi-
tion for cellular survival: spontaneous neural firings internal to the brain
Modularity 133
itself and the brain activity produced by sensory inputs. Indeed, without
both sorts of activity, functional brain circuits simply do not develop. If
we chemically block spontaneous firing of the cells that protrude from the
retina, the structures that relay information to the visual cortex do not
develop in a way that enables normal vision. Similarly, if you keep one eye
closed as your visual system is developing, so that your brain receives little
input from one retina, you will end up functionally blind in that eye. As
a result, even though the cells projecting from the retina produce normal
outputs after you reopen the eye, the areas in the cortex to which they
feed will no longer respond appropriately to visual inputs. Other brain
regions are similarly dependent on environmental stimuli. If infants
are deprived of auditory inputs, they are subsequently unable to process
speech or understand language without special intervention (teaching
them a signed language, for example, or artificially stimulating their
central auditory cortex).
Thus, the precise patterns of environmental stimuli to which the devel-
oping cortex is exposed play an essential role in shaping the brain circuits,
and the functional properties of those circuits, that are connected most
directly to our sensory receptors. But, as we have seen, patterns of activity
internal to the brain are essential in shaping the circuits and functional
properties of less peripheral cortical mechanisms. Indeed, the shaping of
the less peripheral circuits in the cortex is almost entirely dependent on
innervation from other brain cells, rather than from sensory receptors. The
structures in which those other cells reside will have themselves been
shaped either by patterns of innervation from other brain cells or by
patterns of innervation from sensory receptors. And so on. Eventually,
however, we get to structures that were shaped primarily by sensory
innervations. Since patterns of innervation from sensory receptors are
propagated inward in this way, even circuits that are shaped primarily
by patterns of innervation from other brain cells are ultimately, though
indirectly, shaped by the patterns of environmental stimuli encountered
during development. In short, environmental inputs to the brain shape
the more fine-grained cortical structures by determining the outcome of
cell competition.
The cognitive scientist Jeffrey Elman and his colleagues draw a distinc-
tion between additive and subtractive events that is very helpful in
thinking about the process of brain development. Additive events are the
formation of neurons, their migrations to their final locations in the brain,
and their axonal branching to establish connections with other neurons.
Subtractive events are axonal retraction (whereby neurons lose their
134 Chapter 4
the source that overproduces the raw materials that explains the final form,
but the particular environmental demands that prune those materials that
explain it. In short, it is simply not the case that “our mental organs
owe their basic design to our genetic program,” which evolved during the
evolutionary history of the species.13 They owe their basic design to
environment-guided brain activity, which occurs during the lifetime of the
individual organism.
This developmental flexibility of the cortex is known as neural plasticity.
Although the concept of neural plasticity is related to the concept of adap-
tive plasticity discussed in chapter 1, in that both denote forms of flexi-
bility, the two concepts have different meanings, and it is worth being clear
about this difference. Adaptive plasticity, recall, is the ability of a single
genotype to produce more than one adaptive phenotype, producing the
phenotype that is the fittest in its environment. This is a sort of flexibility
of the genotype. The concept of neural plasticity, in contrast, refers to the
ability of brain regions to perform different functions, so that a given brain
region has the capacity take on the function of any other region. This kind
of flexibility entails not only the possibility of multiple developmental out-
comes, which are contingent on the environment, but also the possibility
of change or reorganization of structure in response to changes in the
environment. In other words, the concept of neural plasticity refers not to
a genotype’s ability to produce different adaptive phenotypes, but to the
brain’s ability to remake itself in response to changing environmental
demands. This makes neural plasticity an instance of phenotypic plasticity.
The concept of phenotypic plasticity, in contrast with the concept of
developmental plasticity, refers to cases in which a genotype builds a
mechanism or process that is capable of producing phenotypic change or
reorganization in response to changing conditions in the organism’s
environment. Thus, although the concept of neural plasticity differs in
meaning from the concept of adaptive plasticity in general, and from the
concept of phenotypic plasticity in particular, neural plasticity is actually
a form of phenotypic plasticity.
In fact, neural plasticity isn’t confined to what we mark off as the process
of development from infancy to adulthood. Our brains are changing all
the time, quite rapidly and profoundly. We now know that the number of
neurons in several areas of the cortex increases throughout life, continu-
ally providing neurons and connections to be further pruned by exposure
to environmental demands. In addition, not only do our brains continue
to grow (at least in some places), but they continually reorganize
themselves in response to environmental demands. If a finger is lost, the
138 Chapter 4
cortical region that used to respond to its input will decrease in size and
the neighboring regions will expand until nothing is left of the functional
brain area at all. The converse is true as well. If a digit is overstimulated
for a while, its corresponding area in the cortex increases in size. These
changes occur outside of any “critical period” of development and can
occur within a matter of days or even hours. Indeed, it appears that our
brains change enough over our lifetimes that, by the time we are old, we
use regions in our brains that are different from the ones we used as young
adults to accomplish the same tasks.
How can our brains maintain a plasticity that allows for changes of func-
tion within hours? The evidence now indicates that regions once thought
to be dedicated to a single information-processing task actually receive
inputs from more than one source. When the median nerve of the hand
is severed in adult owl or squirrel monkeys, areas of the cortex that nor-
mally respond to medial nerve stimulation begin almost immediately to
respond to inputs from other nerves in the hand. Currently, the best expla-
nation of the rapidity of this response is that silencing the inputs to the
medial nerve “unmasks” secondary inputs from other nerves. Other animal
studies, human behavioral studies of phantom-limb patients, and func-
tional magnetic resonance imaging (fMRI) studies of the human cortex all
indicate that brain mechanisms process overlapping inputs. In each case,
areas of the cortex allegedly dedicated to processing one sort of informa-
tion are revealed to process very different sorts of information as well.
Perhaps the most striking research that supports this blurring of
information-processing streams concerns how our brains compensate for
vestibular disturbances. If we remove the semicircular canals in our ears
(which help us maintain balance), so that our vestibular system no longer
receives any orientation information, we recover our sense of balance very
quickly, much faster than we could form new neural connections. Single-
cell recordings show that vestibular processing is not rerouted elsewhere
in the brain; the same neurons in the brain stem that respond to normal
vestibular inputs are also used in recovery. Obviously, they must be getting
orientation information from somewhere other than the missing semicir-
cular canals. Some other sensory system must already be feeding into the
vestibular system. One hypothesis is that brains use a form of sensory sub-
stitution to compensate for such information loss. In this case, the brain
would use internally generated signals from the visual system to compen-
sate for the loss of inputs from the semicircular canals. It would substitute
computations from a visual pursuit system (involved in tracking objects as
they move across the visual field), which probably reconstructs informa-
Modularity 139
tion about head movement, for inputs from the semicircular canals.
Perhaps as animals try to orient toward targets, error signals from the retina
help the vestibular system compute head location. Ongoing work is explor-
ing this possibility.
Research in neural plasticity also dovetails with recent data that
highlight cross-modal connections—that is, connections between different
sensory modalities, such as sight and smell. For example, if you touch
someone’s body on the same side and at the same time as you present them
with a visual stimulus, activation in the visual cortex is significantly greater
than when the visual stimulus is presented alone. Imaging studies of this
phenomenon indicate that the somatosensory cortex (which processes
information about touch, pressure, and joint position) projects back to
the visual cortex, telling the visual cortex about tactile stimuli received.
The extent of cross-modal communication among our alleged sensory
“modules” is still a matter of investigation, though we do know that
auditory and visual areas of the cortex exchange a lot of information
regarding speech perception.
So what do all these facts tell us about Evolutionary Psychology’s massive
modularity thesis? First, if information processing about different sensory
domains overlaps in our various brain circuits, and if even “modality-
specific” processing receives inputs from other sensory modalities, there is
little sense in which even our most basic cognitive processes are informa-
tionally isolated in the way that the massive modularity thesis implies.
And, if this is true of our most basic cognitive processes, which function
primarily to convey information about the ambient environment, it is
undoubtedly true in spades for the “higher” cognitive processes involved
in adaptive-problem solving of the sort that interests Evolutionary
Psychologists. The degree of informational overlap in our brains shows
that brain circuits are not “domain specific,” but that they are domain
dominant. One sort of processing in a brain circuit may be more prominent
than others, but other processing is still occurring. Our brain circuits are
not so specialized that they deal only with restricted domains. Instead, they
deal mostly with particular domains, and they do so only contingently; the
dedication of a brain circuit to a particular task is subject to change as the
inputs to that circuit change.
Second, even if our species was faced with recurrent adaptive problems
throughout a significant portion of its evolutionary history, distinct
“genetically specified” brain circuits were not required to solve those prob-
lems. Our brains hit on a different, domain-general solution: a plasticity
that allows particular environmental demands to participate heavily in
140 Chapter 4
having that trait, had higher fitness, on average, than those without it.
And, third, the difference between having and not having the trait was
due (at least in part) to a genetic difference between individuals with and
without the trait. This is why evolutionary biologists say that “adaptive
evolution is caused by natural selection acting on genetic variation.”14
Thus, adaptations are traits that are present in current individuals because
of an evolutionary history in which selection acted on genetic differences
between individuals, preserving the genes for those traits and weeding out
the genes for alternative traits.
Consequently, differences between individuals that are due to environ-
mental differences are not grist for selection’s mill and, hence, do not form
the basis of biological adaptation. Indeed, (micro)evolution, you’ll recall,
just is change in gene or genotype frequencies across generations in a
population. So transgenerational phenotypic changes that are due to
environmental changes—such as the increase in average height in the
developed world during the twentieth century, which was due to improved
nutrition—are not even biological evolution, let alone adaptive evolution.
Only phenotypic changes across generations that are driven by underly-
ing genetic changes are adaptive evolution.
But why does this matter? It matters because, for the reasons discussed,
our functionally specialized cortical circuits are the product of a plastic
system’s response to its local environment. In other words, differences
between individuals with respect to their functionally specialized cortical
circuits are due to environmental differences between them, not to genetic
differences (with the exception of mutations that disrupt development so
severely that “normal” cortical plasticity is lost). And, as it is for us in this
respect, so it was for our ancestors. Consequently, such functionally spe-
cialized brain circuits cannot have formed the basis of adaptive evolution.
They are not present in current individuals because of an evolutionary
history in which selection acted on genetic differences between individu-
als with certain functionally specialized brain circuits and those without.
They are present in individuals because of a plastic brain’s response to its
local environment. In short, even though an adult human brain has many
functionally specialized cortical circuits, which resemble in certain respects
what Evolutionary Psychologists call modules, those circuits are not adap-
tations, and they do not owe their basic design to our “genetic program.”
To put this the other way around, whatever our psychological adaptations
are, they are not modules.
However, the brain’s plasticity—the process by which it forms function-
ally specialized circuits in response to environmental demands—is an
adaptation. Similarly, the immune system’s antibody-assembly process is
142 Chapter 4
emotional responses may be more like the liver or the kidneys than like
antibodies, the result of selection’s having acted on genetic differences
between individuals, preserving the genes for those sexual and emotional
responses while discarding the genes for their alternatives. Consequently,
Evolutionary Psychologists may be right about some of our more basic
emotional adaptations, but nonetheless wrong in its claims that we possess
a lot of cognitive adaptations devoted to very specific forms of problem
solving.
So cortical plasticity is no doubt not our only psychological adaptation.
Some of our affective responses, subserved by circuits in the midbrain and
limbic system, may be adaptations as well. At the same time, it may be
that the emotional adaptations we find in the limbic system are not the
sort of precisely honed psychological adaptations that interest Evolution-
ary Psychologists. Our limbic system produces fear in response to danger-
ous phenomena in the environment, such as coiled snakes. The limbic
system alone, however, is not very sophisticated in the way it produces
these fear responses, since it reacts to all coiled shapes (hoses, ropes, a cir-
cular pile of brush) with the same fear response. Only when the limbic
system is conjoined with our higher cortical structures do we get more spe-
cific responses, such as “Danger! Rattlesnake!” But such specific message
contents derive largely from the functioning of the cortex. The fear
response itself derives from the limbic system, which is simply too crude
to do anything more than react in a rough-and-ready fashion.
There are additional things that we should count among our psycho-
logical adaptations, since brain function is not determined solely by the
brain’s neural circuitry, by how the brain is “wired.” Neurotransmitters and
hormones also affect how the brain’s circuits function and thus contribute
to the regulation and control of behavior. We know that the brain regions
regulating sexual desire and response are affected by levels of the sex hor-
mones in the brain and that these levels sometimes vary cyclically. For
example, although human females are sexually receptive throughout their
menstrual cycles, there is a marked increase in sexual fantasy and female-
initiated sexual activity around ovulation, and this increase is produced by
peaking levels of estrogen around ovulation. This is undoubtedly an adap-
tation, selected for its effects on sexual behavior. Consequently, neuro-
transmitters and hormones, or specific levels of these, also number among
our psychological adaptations (although specific patterns of their timed
release could be adaptations as well).
So we do possess psychological adaptations in addition to brain plastic-
ity. But the question so far has been what accounts for the functionally
144 Chapter 4
mind could possess innate knowledge. In short, the very first premise of
the argument—that all knowledge possessed by a strictly domain-general
mind would have to be derived from experience—is false, so the argument
fails to support massive modularity.
The crucial point here is the distinction between information and the
mechanisms in the mind that process that information. The mind could
come equipped with innate knowledge about particular adaptive prob-
lems—for example, that snakes are dangerous or that you should sacrifice
your own resources to benefit others only in proportion to the degree to
which others are related to you. Fodor and Samuels think that Evolution-
ary Psychology’s argument establishes this much. But Evolutionary Psy-
chologists also claim that the mind possesses a specialized mechanism, or
“minicomputer,” for each adaptively important body of domain-specific
information. Each of these mechanisms, they claim, specializes in pro-
cessing a particular body of domain-specific information, and it processes
that information with a domain-specific body of rules, which applies only
to that information. These are the claims that Fodor and Samuels contend
don’t follow from the argument. For the mind could consist primarily of
a single general-purpose mechanism (supplemented by modules for sen-
sory inputs), which employs rules of reasoning that apply to information
independently of the problem domain that the information is about. Para-
digms of such domain-independent rules of information processing are the
rules of deductive logic and the probabilistic rules of hypothesis testing.
Thus, according to Fodor and Samuels, the picture that emerges from the
argument is one of a mind that is equipped with some innate domain-
specific information that is processed by a domain-general mechanism employ-
ing domain-general rules, which are also used in acquiring additional
information about the world.
While Fodor and Samuels are right about the fallacy in the argument,
they’re too quick to accept that the argument actually demonstrates the
necessity of innate knowledge. All the argument really shows is that a
strictly domain-general mechanism, which employs only domain-general
learning rules, must be given a head start on learning in certain adaptively
crucial domains. Without any head starts, domain-general learning would
face the insoluble problem of learning which of the world’s features are
worth learning about before they set about learning about them. For there
are an overwhelmingly vast number of things in the world that a general-
purpose mechanism could learn about. On entering the world, should it
pay more attention to the walls of the nursery or to mother’s face? Should
it spend hours watching for and documenting subtle changes on the backs
150 Chapter 4
of its hands or should it focus on the noises coming from mother’s mouth?
Since there are far too many features of the world that could be foci of
domain-general learning, a wholly unconstrained general-purpose learn-
ing mechanism would be overwhelmed by useless data unless it had some
built-in “data filters.” Thus, if a general-purpose mechanism as plastic as
the brain is to develop the right problem-solving specializations, there
must be some constraints on its possible courses of development. In short,
its development must be “channeled” in the right directions.
Innate knowledge of adaptively important features of the world is cer-
tainly one way to ensure that a general-purpose mechanism comes to know
the right stuff, since it’s “preequipped” with the knowledge it will need.
But innate knowledge is not the only way to ensure coming to know the
right stuff. Initial biases in the mechanism, which channel attention to par-
ticular environmental inputs more than others, can be all that is required
to get it learning about the right stuff. These initial biases can be far short
of actual knowledge about a problem domain. They can be mere predis-
positions to apply domain-general learning to certain special classes of
environmental inputs.
To illustrate the role of initial biases in learning, consider the phenom-
enon of face recognition. In an adult human, a significant portion of the
brain is involved in processing information about human faces, and the
ability to accurately distinguish our conspecifics from one another and
detect subtle facial cues of emotional state has undoubted adaptive value.
(Think of the problems you’d have if you couldn’t distinguish people
who have previously helped you from those who have hurt you, or if you
couldn’t tell whether someone was angry or happy with you.) Because of
this, some cognitive scientists, including Evolutionary Psychologists, have
proposed that there is a genetically specified module for face recognition,
which embodies a lot of innate knowledge about human faces.
But Jeffrey Elman and his colleagues propose an alternative, according
to which face recognition becomes gradually and progressively “modular-
ized.” This process, however, has a head start in development. An experi-
ment with human newborns has shown that they preferentially attend to
stimuli consisting of three high-contrast blobs in the triangular configura-
tion of two eyes and a mouth. Interestingly, newborns don’t prefer actual
faces over pictures of three high-contrast blobs until a little later in devel-
opment, and even at six months infants show no discrimination of human
faces from monkey faces. Only later still do infants begin to pay attention
to and discriminate among movements internal to the face. Thus, infants
appear to gradually learn about human faces, but learning is aided by an
Modularity 151
massively modular brain it accomplishes the trick with very minimal initial
structure built into the brain. For, rather than needing to build in a
complex specialized circuit for a particular task, or a developmental mech-
anism to “add” that circuit at a particular point in development, it merely
needs to build in a bias to preferentially attend to a restricted class of
inputs.
The initial biases built into the brain, of course, may themselves have
been shaped by selection (indeed, they could even be the product of
frequency-dependent selection). So our psychological adaptations would
consist not only in brain plasticity, neurotransmitters, hormones, and
some affective responses, but in the brain’s specific initial biases as well. It
is worth noting, however, that the biases needn’t be as domain specific
as Evolutionary Psychology’s modules, so they needn’t be as precisely
attuned to adaptive-problem domains as Evolutionary Psychologists
believe modules to be. For, again, face recognition can be accomplished by
starting with learning about three high-contrast blobs. This can solve the
adaptive problem of needing a mechanism that recognizes faces, but the
initial bias pertains to a far wider domain than that of human faces; in
fact, the initial bias isn’t even about faces per se. Thus, while initial biases
number among our psychological adaptations, they cannot be easily
inferred from the highly specialized cognitive abilities exhibited by adult
humans. All that initial biases need do is start the functioning of a general-
purpose mechanism down a path that will lead, in standard developmen-
tal environments, to a brain circuit that specializes in an adaptive-problem
domain.
These points are important enough to belabor a little, so consider them
from another angle. Many cognitive scientists have claimed that there is
a strong similarity between learning and the growth of scientific knowl-
edge. Like the growth of scientific knowledge, learning appears to be a
process in which hypotheses about a particular domain of phenomena are
tested against evidence derived from experience and rejected or revised
when that evidence conflicts with them. And, in both cases, the goal is to
arrive at hypotheses about the world that enable us to accurately antici-
pate the course of experience (to make accurate predictions).
In an argument that long predated Evolutionary Psychology’s argument
for innateness, the philosopher of science Karl Popper argued that some
kind of hypotheses always had to precede the gathering of data in science,
since data gathering that is not guided by hypotheses would be over-
whelmed by the literal infinity of “facts” that could be recorded. Imagine
recording everything that you can observe right now. As you become more
Modularity 153
In this passage, once again, Tooby and Cosmides conflate a few distinct
properties. As we have seen, a general-purpose mechanism need not
necessarily be “content free” or “content independent.” A general-purpose
mechanism can have initial biases that are “content dependent,” about par-
ticular things in the environment that are important targets of learning,
and these can facilitate the acquisition of adaptively crucial information.
Even granting this, however, Tooby and Cosmides would remain uncon-
vinced that learning is driven by domain-general rules. For the real worry
underlying their claim is that a domain-general learning mechanism would
be unable to acquire certain crucial domain-specific rules.
Everyone agrees that human problem solving routinely employs rules
that are domain specific. Figuring out the best move in chess involves rules
that are specific to chess; not only must one have learned the rules of chess,
but one must have rules for reasoning about which moves are most effec-
tive in which circumstances, and these rules will apply only in chess games.
Similarly, figuring out the best strategic maneuver at a critical point in a
baseball game involves use of rules that are highly baseball specific in their
content, and figuring out how to solve an equation in algebra involves use
of rules of inference that are specific to algebra. Thus, learning isn’t simply
a process of acquiring information about the world; it is also a process of
acquiring many domain-specific rules, which are used in solving problems
in their proprietary domains. And the conviction of Evolutionary Psy-
chologists is that a mind equipped with only domain-general learning rules
would be unable to acquire the repertoire of domain-specific rules that
characterizes adult human problem-solving abilities. So, they believe, the
human mind can’t consist solely of domain-general mechanisms, but must
contain some domain-specific mechanisms that are preequipped with rules
of reasoning specific to their proprietary domains.
Obviously, however, humans do acquire chess-specific and baseball-
specific rules of problem solving. Since no one would argue that humans
possess innate rules that are specific to these domains, they must be
acquired through some domain-general mechanism. So what is the source
of Evolutionary Psychologists’ conviction that a domain-general learning
mechanism couldn’t learn to perform the problem-solving tasks that
humans routinely perform? The conviction is based on a widely accepted
view concerning language acquisition and an assumption that what is true
of language acquisition is true of the acquisition of most complex human
competences. So consider language acquisition.
Adult speakers of a language are capable of understanding and generat-
ing a potentially infinite number of grammatical sentences. According to
the received view in psycholinguistics, this capacity can be explained only
156 Chapter 4
by supposing that adult speakers know the grammatical rules that gener-
ate all of the infinitely many grammatical sentences of their languages.
The problem of explaining language acquisition, then, is the problem of
explaining how individuals acquire knowledge of the grammatical rules of
their languages.
Now, suppose that individuals are equipped only with domain-general
learning rules. We can then suppose that learning involves formulating
hypotheses about the grammatical rules of one’s language and testing
those hypotheses against the linguistic data encountered in one’s experi-
ence (the sentences uttered by others, corrections of one’s own uttered
sentences, and any explicit instruction in the rules of grammar). And
suppose further that hypothesis formation and testing employs only
domain-general rules of reasoning—for example, those of probability
theory and deductive logic. Could an individual equipped only with such
domain-general learning rules acquire knowledge of the grammatical rules
of their language?
According to the received view, the answer is no, for the following
reasons. During the course of linguistic development, children are exposed
primarily to speech containing incomplete and ungrammatical sentences
and to comparatively few utterances that are grammatically correct. In
addition, they are exposed to virtually no “positive evidence” of gram-
matical rules (being told what the rules are) and to very little “negative
evidence” of the rules (being told that their utterances are grammatically
incorrect and having them corrected). In short, during the course of lin-
guistic development, the average language learner is exposed to precious
little evidence about the grammatical rules of their language. If language
acquisition involved the employment of domain-general rules in forming
and testing hypotheses about grammatical rules, hypotheses would have
to be tested against this very small body of evidence regarding grammati-
cal rules. Since the evidence is always necessarily sparse compared to the
number of grammatical sentences in a language, however, it would always
be compatible with several competing hypotheses about the rules of
grammar. As a result, a language learner guided solely by domain-general
learning rules would be unable to learn which among several competing
hypotheses about the rules of grammar is the correct hypothesis. Yet native
language speakers do arrive at knowledge of the grammatical rules of their
languages, so they can’t be acquiring this knowledge through domain-
general learning rules. Consequently, language acquisition must have a
head start in the form of domain-specific learning rules, which are designed
specifically to facilitate acquisition of the grammatical rules of one’s native
Modularity 157
sexual attraction to young females. Chess is far more complex than sexual
attraction to young women, yet we learn chess without the benefit of an
innate chess module. Since the majority of the adaptive “tasks” that
Evolutionary Psychologists claim we perform are similarly much less
complex than speaking a language, it is a mistake to assume that what is
true of language acquisition must be true of learning in each of these other
domains as well.
Third, even if learning in certain nonlinguistic domains requires a head
start in the form of domain-specific learning rules, given the lack of com-
plexity in these domains, the necessary domain-specific learning rules
could take the form of minimal initial biases. The initial bias described in
greatest detail earlier was a bias to attend to certain stimuli, but initial
biases could also take the form of dispositions that physically embody a
domain-specific rule. Thermostats, for example, are physically constituted
in such a way that, when set at seventy degrees, they behave in accordance
with the rule, “Turn the furnace on when the temperature falls below
seventy degrees, and turn it off when the temperature rises above seventy
degrees.” Thus, while it remains to be demonstrated that learning in any
area other than language requires a head start in the form of innate
domain-specific learning rules, if such head starts are required, they could
be provided by initial biases.
So far we’ve been examining Evolutionary Psychology’s arguments for
massive modularity, and we’ve found that none of them demonstrates that
the brain is massively modular in the way Evolutionary Psychology claims.
But there is also a good theoretical reason to think that the brain could not
possibly be massively modular in the way Evolutionary Psychologists claim.
Indeed, there is reason to think that most of the modules postulated by
Evolutionary Psychologists couldn’t function without the assistance of a
domain-general mechanism.
To see why, consider Evolutionary Psychology’s claim, to be examined
in detail in chapter 5, that women have an evolved mate-choice module
that implements rules “designed to detect and prefer high-status men.”24
According to Evolutionary Psychologists, modules are activated by the
environmental cues that are specific to the domains in which they spe-
cialize. So a woman’s mate-preference module is activated when exposed
to high-status men, and it functions to produce attraction to those men.
But status is not directly observable in the way that baldness is; it has
to be inferred from other properties that are observable. Thus, in order
to produce attraction to high-status males, the female mate-preference
Modularity 159
module must be activated by signs that are positively correlated with high
status. That is, its input must be a sign of high status in males.
Signs of status, however, vary considerably across cultures. What is
accorded high status in one culture may be denigrated in another. As
Robert Wright, Evolutionary Psychology’s most effective popularizer,
vividly puts it: “The range of things that can bring status in different
cultures and subcultures is astonishing. Making beads, making music,
delivering sermons, delivering babies, inventing drugs, inventing tales, col-
lecting coins, collecting scalps. . . . [T]he Zuni [Native Americans of New
Mexico] confer status on those who don’t seek status too fiercely, and deny
status to those who do. . . . In a monastery, serenity and asceticism can be
sources of status. In some strata of Victorian England, a nearly ludicrous
amount of gentility and humility could help earn status.”25 In spite of this,
Symons says: “The particular correlates or indexes of male status do, of
course, vary; what is invariant is the psychological adaptation that speci-
fies the rule ‘prefer signs of high status.’”26
Since signs of high status can be so varied, however, information from
any number of domains could be relevant to determining which among
all the many observable properties are the signs of high status in one’s local
environment. If delivering sermons or delivering babies could be signs of
high status, then information from “the religion domain” (to determine
who’s the best preacher) or “the obstetric domain” (to determine who’s the
best obstetrician) would be relevant to detecting high-status males. But,
since the mate-preference module employs only narrow domain-specific
rules that are specialized to process information about signs of high status,
it will be unable to wander across an unprincipled variety of domains with
those rules in order to figure out which are the local signs of high status.
For, in order to function, it must be given a sign of high status. Thus, some
domain-general mechanism must be involved in ascertaining the local
signs of high status. But, if a domain-general mechanism has to mediate
between environmental inputs and the inputs to modules, and if it can
effectively perform the tasks necessary in this capacity, it’s very unclear
why we would need highly specialized modules to do the easy jobs after a
domain-general mechanism has done the hard jobs. A domain-general
mechanism that can effectively mediate between the environment and
modules should also be able to perform as well as modules on the spe-
cialized tasks that Evolutionary Psychologists attribute to them.
Fodor calls this the input problem. The input problem arises when the
properties to which a module is allegedly responsive aren’t things that can
160 Chapter 4
be directly detected by the senses but must be inferred from things that can
be detected by the senses—properties such as having high status and being
a social exchange. For, as Fodor puts it, there is no “Lurking Benevolence”
that paints high-status males or social exchanges readily perceptible, pro-
prietary colors. And most of the modules postulated by Evolutionary
Psychologists deal with properties that can’t be detected by the senses, but
must be inferred. Since such properties must be inferred, and since infor-
mation from any number of domains could be relevant in inferring them,
they can’t simply be detected by modules. So domain-specific mechanisms
cannot perform their putative functions without the prior assistance of a
more powerful domain-general mechanism feeding them their inputs. But
the necessity of this more powerful domain-general mechanism under-
mines claims that domain-specific mechanisms are necessary for perform-
ing the tasks attributed to them.
The last section examined strictly theoretical arguments for and against
massive modularity, finding none that supports it and one that poses
serious problems for it. But these are just arguments. Nothing convinces
quite like good empirical evidence; the proof is in the experiments. And
Evolutionary Psychologists believe that there is strong empirical evidence
for some of the modules they postulate. For example, Evolutionary Psy-
chologists believe that there is currently very strong empirical evidence for
the existence of language modules. As argued earlier, however, language is
atypically complex, and acquiring it may require a language acquisition
device. Since nothing would follow from this about the massive modular-
ity of the mind, we need to examine the empirical evidence for other
modules.
Apart from language modules, Evolutionary Psychologists widely tout
two other postulated modules as having very strong empirical evidence in
their favor. One is Cosmides’ “cheater-detection module,” which was dis-
cussed briefly in chapter 2, and the other is the so-called “theory-of-mind
module.” Let’s examine each of these, beginning with the cheater-
detection module.
As we saw in chapter 1, many activities in which organisms engage have
associated fitness costs and benefits. Foraging for food expends energy,
which exacts a fitness cost from the forager, but finding food provides the
forager with a fitness benefit. From a cost-benefit standpoint, then, we
should expect organisms to forage for food in ways that minimize fitness
Modularity 161
performing some act that benefits A at some cost to B. Third, the fitness
benefit to A of B’s act must be greater than the fitness cost to A of A’s act.
Fourth, the fitness benefit to B of A’s act must be greater than the fitness
cost to B of B’s act. When these four conditions are met, A and B both
come out ahead by performing acts that benefit one another, and selec-
tion then favors their propensities to perform altruistic acts. Any interac-
tions that have the above cost-benefit structure are an instance of what
Trivers calls reciprocal altruism and what Cosmides calls a social exchange.
Once individuals evolve propensities to perform altruistic acts, however,
selection can favor what Trivers calls “cheating,” the failure or refusal to
reciprocate an altruistic act. As Cosmides puts it, a cheater is thus someone
who accepts the benefit of another’s altruistic act without paying the cost of
reciprocation. In a population of altruists, a mutant cheater would enjoy
the highest fitness, since the cheater would accrue the benefits of the
altruistic acts of others without ever paying the costs of reciprocation.
So cheaters would begin to increase in frequency in such a population. As
cheaters increased in frequency, however, the number of individuals
performing altruistic acts would decrease until there were no more altru-
istic acts from which cheaters, or anyone else, could benefit. Thus, recip-
rocal altruism appears to be evolutionarily unstable. “Given this unstable
character of the system, where a degree of cheating is adaptive,” Trivers
argues, “natural selection will rapidly favor a complex psychological
system in each individual regulating both his own altruistic and cheating
tendencies and his responses to these tendencies in others. As selection
favors subtler forms of cheating, it will favor more acute abilities to detect
cheating.”27
In keeping with Evolutionary Psychology’s claim that distinct adaptive
problems select for distinct solutions, Cosmides argues that these selection
pressures would have selected for a module dedicated to detecting cheaters
in social exchanges. This module, she argues, should be designed to detect
and process information about social exchanges using its own proprietary
body of principles of reasoning that apply specifically to social exchanges.
These principles of reasoning would include “algorithms that produce and
operate on cost-benefit representations of exchange interactions” and
“inferential procedures that make one very good at detecting cheating.”28
The latter inferential procedures would be designed to detect when
someone has failed “to pay a cost to which one has obligated oneself by
accepting a benefit, and without which the other person would not have
agreed to provide the benefit.”29 The module that implements the infer-
ential procedures to look for those who have accepted a benefit without
Modularity 163
subjects give the logically correct response to the Wason selection task
appears to vary as a function of what the conditional rules are about. For
example, in response to the Wason selection task described above, typi-
cally only 10 percent of subjects choose the E and 7 cards. Most subjects
choose the E card alone or the E and 4 cards, despite the fact that no matter
what is on the other side of the 4 card it cannot falsify the conditional rule
“If a card has a vowel on one side, then it has an even number on the
other.” In contrast, most subjects choose the logically correct cards
when presented with the “drinking-age problem,” for which subjects are
instructed to imagine being a “bouncer” in a bar, whose job is to identify
and eject violators of the conditional rule “If a person is drinking beer,
then they must be at least twenty-one years old.” Subjects are then given
the following four cards, which contain partial information about four cus-
tomers: drinking beer, drinking Coke, twenty-five years old, and sixteen years
old. Typically, 75 percent of subjects choose the drinking beer and sixteen
years old cards, the cards that represent the P and not-Q conditions.
Many psychologists, Evolutionary Psychologists among them, conclude
that these results indicate a “content effect” in the Wason selection task.
For it appears that the conditional rules in both problems have the same
logical form and that they differ only in their content, in what the rules are
about. The fact that the drinking-age problem elicits a high rate of logically
correct responses, while the letter-number problem elicits a very low rate
of logically correct responses, appears to indicate that the content of the
problems accounts for the difference in performance. There must be some-
thing about the drinking-age problem that facilitates subjects’ reasoning
in a way that the letter-number problem does not.
Cosmides and Tooby argue that the “content effect” is due to the fact
that the drinking-age problem exemplifies a social-contract rule, whereas the
letter-number problem does not. Cosmides and Tooby define a social con-
tract as “a situation in which an individual is obligated to satisfy a require-
ment of some kind, usually at some cost to him- or herself, in order to be
entitled to receive a benefit from another individual (or group). The
requirement is imposed because its satisfaction creates a situation that ben-
efits the party that imposed it. Thus, a well-formed social contract expresses
an intercontingent situation of mutual benefit: To receive a benefit, an
individual (or group) is required to provide a benefit. Usually (but not
always) one incurs a cost by satisfying the requirement.”30 In the drinking-
age problem, Cosmides and Tooby argue, drinking beer represents a
“benefit” that is supposed to be available only to those who have satisfied
the “requirement” of being over twenty-one years old. That requirement,
Modularity 165
powerful aphrodisiac that makes the men who eat it irresistible to women.
Given the prohibition on sex between unmarried people, the island’s elders
have thus enacted the following rule: “If a man eats cassava root, then he
must have a tattoo on his face.” Of course, because cassava root is such a
powerful aphrodisiac, many bachelors in this culture “are tempted to cheat
on this law whenever the elders are not looking.”31 Cosmides then pre-
sented subjects with four cards that read eats cassava root, eats molo nuts,
tattoo, and no tattoo. Subjects were instructed that these cards described
four men and that they were to turn over the necessary cards in order to
determine whether any of these four men were in violation of the rule.
Cosmides paired this unfamiliar social-contract problem with the fol-
lowing unfamiliar non-social-contract problem. Subjects were asked to
imagine being anthropologists studying an African hunter-gatherer band
whose members like to eat duiker meat, the meat of a small antelope, and
who like to use ostrich eggshells, which are very light, as canteens to carry
water. Subjects were asked to imagine overhearing the natives frequently
say, “If you eat duiker meat, then you have found an ostrich eggshell.” The
natives’ rationale for this saying is that duikers frequently feed on ostrich
eggs, so duikers and ostrich eggshells are typically found in close proxim-
ity. Cosmides then presented subjects with four cards that provided infor-
mation about four locations with caches of eggs. One side of the cards
described the type of eggshell found at that location, and the other side
described mammal tracks found at that location. The cards read, duiker,
weasel, ostrich eggshell, and quail eggshell. Subjects were instructed to turn
over the necessary cards in order to determine whether the natives’ con-
ditional saying is true.
Cosmides found that 75 percent of subjects turned over the logically
correct P and not-Q cards (eats cassava root and no tattoo) in response to the
unfamiliar social-contract problem, but that only 21 percent of subjects
turned over the logically correct cards (duiker and quail eggshell) in response
to the unfamiliar non-social-contract problem. In a second, similar exper-
iment, Cosmides obtained similar results, finding that 71 percent of sub-
jects selected the logically correct P and not-Q cards in response to the
unfamiliar social-contract problem, while only 25 percent of subjects
selected the logically correct cards in response to the unfamiliar non-social-
contract problem. Thus, she concluded, the “content effect” is robust; sub-
jects reason more effectively about social-contract rules, even when they
are unfamiliar, than about logically identical non-social-contract rules. In
short, social-contract rules “facilitate” performance on the Wason selection
task.
Modularity 167
But why is this finding significant? Cosmides argues that, if the human
mind does not consist of numerous domain-specific psychological mech-
anisms, then it must apply the same set of logical principles to solve all
reasoning problems, regardless of what those problems happen to be about.
In that case, she claims, the mind would apply precisely the same logical
principles to all problems that have the same logical structure. In particu-
lar, if subjects solved Wason selection tasks by applying logical principles,
they would apply the same logical principles to all logically identical tasks,
in which case subjects should perform with the same degree of proficiency
on all problems that are logically identical. The fact that subjects
don’t perform with the same degree of proficiency on all logically
identical problems, Cosmides argues, indicates that they are not simply
applying logical principles in solving those problems, which in turn entails
that the human mind does not employ a single set of logical principles
to solve all reasoning problems. In contrast, Cosmides argues, if humans
have an evolved cheater-detection module, any context that presents a
social contract should activate that module to look for cheaters, but con-
texts that do not present social contracts should not activate the module.
Thus, Cosmides concludes, the robust social-contract “content effect” in
performance on Wason selection tasks is evidence of a cheater-detection
module.
The second finding that supports the cheater-detection hypothesis is that
subjects appear to systematically select logically incorrect cards in selec-
tion tasks when those cards represent cheating, taking a benefit without
paying a cost. This result was obtained in “switched-conditional” experi-
ments conducted by Cosmides and “perspective-switching” experiments
conducted by the evolutionary psychologists Gerd Gigerenzer and Klaus
Hug.
In one of Cosmides’ “switched-conditional” experiments, subjects were
given precisely the same background information and instructions that
they were given in the experiment described above. The only difference
was that the conditionals that subjects were asked to evaluate were
switched around. Thus, in the “switched” cassava-root problem, subjects
were asked to select those cards necessary to discover any violations of the
rule “If a man has a tattoo on his face, then he eats cassava root.” And, in
the “switched” duiker-meat problem, subjects were asked to select those
cards necessary to determine whether “If you have found an ostrich
eggshell, then you eat duiker meat” is true.
Cosmides argued that the logically correct responses to these problems
are to select the P and not-Q cards. But, she claimed, if the “switched”
168 Chapter 4
person gets a day off during the week.” This rule was presented with two
different background stories, one that cued the subject into the perspec-
tive of the employee, and one that cued the subject into the perspective
of the employer. “The employee version stated that working on the
weekend is a benefit for the employer, because the firm can make use of
its machines and be more flexible. Working on the weekend, on the other
hand, is a cost for the employee. The context story was about an employee
who had never worked on the weekend before, but who is considering
working on Saturdays from time to time, since having a day off during the
week is a benefit that outweighs the costs of working on Saturday. There
are rumours that the rule has been violated before. The subjects’ task was
to check information about four colleagues to see whether the rule has
been violated before.”32 In the employer version, subjects were given the
same rationale, but were cued into the perspective of the employer by
being told to check whether any of four employees had violated the rule
before. The four cards giving information about the four employees in both
tasks displayed worked on the weekend, did not work on the weekend, did get
a day off, and did not get a day off.
Gigerenzer and Hug argued that the application of domain-general
logical principles should lead to selecting the P and not-Q cards (worked on
the weekend and did not get a day off) regardless of a subject’s perspective
on the conditional rule. On the other hand, a cheater-detection module
should lead subjects to select whichever pair of cards represents being
cheated from the perspective they occupy. Thus, subjects cued into the per-
spective of the employee should select the P and not-Q cards (worked on the
weekend and did not get a day off), since they represent the conditions under
which an employee was cheated by the employer. In contrast, subjects cued
into the perspective of the employer should select the not-P and Q cards
(did not work on the weekend and did get a day off), since they represent the
conditions under which the employer was cheated by an employee. And,
indeed, Gigerenzer and Hug found that 75 percent of the subjects cued
into the perspective of the employee chose the P and not-Q cards and that
61 percent of the subjects cued into the perspective of the employer chose
the not-P and Q cards. Gigerenzer and Hug conclude that these results
clearly provide evidence of a cheater-detection mechanism.
The third finding that is taken as evidence of a cheater-detection module
was obtained in an experiment conducted by the Evolutionary Psycholo-
gist Lawrence Fiddick along with Cosmides and Tooby. In this experiment,
subjects were all presented with the following scenario: “You are a South
American farmer. At the end of the harvest you find you have more
170 Chapter 4
potatoes than you need so you pack up some of them and travel to the
neighboring village. When you get to the village four different people
approach you, and though you don’t speak the same dialect, you recog-
nize that each of them is telling you. . . .”33 At this point, half the subjects
were presented with the following “conditional version” of the problem:
“If you give me some potatoes, then I will give you some corn.” The other
half of the subjects were presented with the following “want version” of
the problem: “ ‘I want some potatoes.’ You, in turn, know a little bit of
their dialect, and tell them ‘I want some corn.’ ” Then all subjects were
given the same four cards, each providing information about one of the
four people, and were instructed to turn over only those cards necessary
in order to determine whether any of the four people had cheated “you.”
The four cards read you gave this person potatoes, you gave this person nothing,
this person gave you corn, and this person gave you nothing.
Fiddick, Cosmides, and Tooby argued that, if subjects apply domain-
general logical principles in solving these problems, there should be a sig-
nificant difference in the frequency with which subjects choose the P and
not-Q cards in the two versions of the problem. Of course, in the “condi-
tional version,” subjects should select the P and not-Q cards (you gave this
person potatoes and this person gave you nothing), since they represent the
logically correct solution to the problem. But, they argued, since there is
no logical connective connecting the statements “I want some potatoes”
and “I want some corn” in the “want version” of the problem, there is
nothing in the “want version” to which logical principles can apply. As a
result, card selection in the “want version” should be random. Indeed,
since there are sixteen possible ways of selecting cards in response to the
problem, the selection of both you gave this person potatoes and this person
gave you nothing shouldn’t be any more likely than any of the other fifteen
possible responses. Thus, they concluded, if subjects apply domain-general
logical principles in the selection task, vastly more subjects should choose
the P and not-Q cards (you gave this person potatoes and this person gave you
nothing) in the “conditional version” than in the “want version.”
Interestingly, Fiddick, Cosmides, and Tooby found that 67 percent of
subjects chose the P and not-Q cards in response to the “conditional
version” of the problem and that 50 percent chose the same cards in
response to the “want version” of the problem. Although this can’t be
accounted for by the hypothesis that people employ domain-general
logical principles in reasoning about social contracts, they claimed, these
results are precisely what we should expect if humans have an evolved
cheater-detection module. For a cheater-detection module would represent
Modularity 171
selection tasks involving what Cosmides and Tooby call social contracts,
which is a much broader class of phenomena than the class of social
exchanges. For a social contract, as Cosmides and Tooby define it, is “a sit-
uation in which an individual is obligated to satisfy a requirement of some
kind, usually at some cost to him- or herself, in order to be entitled to
receive a benefit from another individual (or group).”34 Social contracts thus
include situations in which an individual must satisfy a requirement in order
to receive a benefit from society. Indeed, this is the kind of situation embod-
ied in the drinking-age problem and the cassava-root problem. In the
drinking-age problem, individuals must satisfy the requirement of being at
least twenty-one years of age before society bestows the benefit of being
able to drink beer, while in the cassava-root problem males must be
married (as signified by a facial tattoo) before society bestows the benefit
of being able to eat cassava root. Social contracts are thus relations between
individuals and society, in which society makes the satisfaction of a par-
ticular requirement a precondition for the receipt of a certain benefit.
Cheating on a social contract thus involves taking a benefit from society
without satisfying the requirement that is a precondition for receipt of that
benefit.
There is, therefore, a disconnect between the theoretical support for the
cheater-detection module and the experimental results that purportedly
provide evidence of its existence. The theory behind the cheater-detection
module should lead us to expect a mechanism that is specialized in detect-
ing cheaters in the domain of social exchanges. But the experimental results
that purportedly support the existence of a cheater-detection module
involve detecting cheaters in the domain of social contracts. This poses the
following dilemma for Evolutionary Psychology’s claim that there is strong
empirical evidence of an evolved cheater-detection module. On the one
hand, even if the experimental results provide evidence of a module for
detecting cheaters in social contracts, there is no theoretical support for
predicting that humans have an evolved module that specializes in social
contracts. The reason is that, although we have a well-developed theoret-
ical understanding of how social exchanges evolved, we have no compa-
rable theoretical understanding of how social contracts evolved (and
Evolutionary Psychologists offer no theory about the evolution of social
contracts). We consequently have no well-developed theoretical basis for
any hypotheses regarding the psychological mechanisms, if any, that have
evolved to deal with social contracts. On the other hand, even if we have
good theoretical reasons to expect that humans have an evolved psycho-
Modularity 173
sentences have the same logical form, any difference in the way that sub-
jects respond to them gets interpreted as due to a difference in what the
sentences are about.
But this presupposition is false. Not all conditionals are alike. Indeed, it
is a commonplace among most philosophers of logic that there are several
different kinds of conditional. Two of these are important for present pur-
poses, indicative conditionals and deontic conditionals. Indicative condition-
als are formed by making the truth of one fact-stating sentence, Q,
conditional upon the truth of another fact-stating sentence, P. The indica-
tive conditional if P, then Q then makes a conditional assertion, another
statement that purports to be true. In short, indicative conditionals are
used to make assertions of fact. The conditional sentence “If a card has a
vowel on one side, then it has an even number on the other” is an example
of an indicative conditional. In contrast, deontic conditionals are used
to impose obligations. Rather than making the truth of one statement
conditional upon the truth of another, they make an obligation condi-
tional upon the truth of a fact-stating sentence. The conditional sentence
“If you are under the age of twenty-one, then you must not drink alcohol”
is an example of a deontic conditional. In the Wason selection tasks
we’ve discussed, all of the conditionals that elicited a low frequency of
P and not-Q selections are indicative conditionals, whereas the condition-
als that elicited a high frequency of P and not-Q selections are deontic.
Indeed, all of the so-called social-contract rules are deontic conditionals,
although the class of deontic conditionals is broader than the class of so-
called social-contract rules that are used in Evolutionary Psychologists’
experiments.
Indicative conditionals and deontic conditionals actually have different
logical properties, and the difference in their logical properties affects the
logic of the Wason selection tasks in which the two types of conditional
are embedded. This can be seen by comparing the problems in Cosmides’
first experiment. In the duiker-meat problem, subjects are asked to evalu-
ate the truth of the conditional “If you eat duiker meat, then you have
found an ostrich eggshell.” The so-called not-Q card reads quail eggshell,
which is elliptical for “you have found a quail eggshell,” which (in the
context of the background story) entails “it is not the case that you have
found an ostrich eggshell.” Thus, the “quail eggshell” card is elliptical for
the negation of the consequent of the conditional whose truth the subjects
are to evaluate. In contrast, in the cassava-root problem, subjects are asked
to determine whether any men are violating the rule “If a man eats cassava root,
then he must have a tattoo on his face.” The so-called not-Q card in this
Modularity 175
problem reads no tattoo. This card, however, does not represent the nega-
tion of the consequent clause of the conditional rule. There are two ways
in which a negation can be inserted into the consequent clause of the rule.
We can either negate it by saying, “it is not the case that he must have a
tattoo on his face,” which removes the obligation to have a facial tattoo,
or we can insert a negation by saying, “he must not have a tattoo on his
face,” which imposes an obligation to not have a facial tattoo. The no tattoo
card, however, corresponds to neither of these. Rather, the no tattoo card
is simply elliptical for “this man has no facial tattoo.” Consequently, the
logic of the cassava-root problem, which contains a deontic conditional,
differs from the logic of the duiker-meat problem, which contains an
indicative conditional.
Other commentators have made roughly the same point in a different
way. For example, the evolutionary anthropologists Dan Sperber, Francesco
Cara, and Vittorio Girotto point out that, in the indicative-conditional
versions of the Wason selection task, “subjects are asked to reason about
the rule,” to determine whether it is true, whereas “in true deontic ver-
sions they are asked to reason from a rule given as axiomatic,” to deter-
mine whether someone is in conformance with the rule.37 What Sperber,
Cara, and Girotto fail to drive home, however, is that the reason that the
logic of the tasks differ in this way is that the conditionals embedded in
the tasks differ in their logical properties.
But, if deontic conditionals have a different logical form than indicative
conditionals, what is it? For it certainly looks like the logical form of “If
you are under the age of twenty-one, then you must not drink alcohol” is
if P, then Q. Fodor has an interesting suggestion, which I believe is correct.
Fodor argues that, in an important sense, deontic conditionals aren’t true
conditionals at all. Rather, Fodor argues, deontic conditionals categorically
impose obligations (in their Q parts), while also indicating on whom the
obligations fall (in their P parts). For example, on Fodor’s analysis, the
above conditional actually has the form “Thou shalt not drink alcohol,
and this prohibition falls on those under the age of twenty-one.” An equiv-
alent way of viewing this is to see the P part of deontic conditionals as
negatively indicating a class of individuals who are exempted from the obli-
gation imposed in the Q part. In this equivalent formulation, the above
deontic conditional actually asserts “Thou shalt not drink alcohol—unless
you happen to be at least twenty-one.” On Fodor’s analysis, then, all
deontic conditionals have the logical form of Old Testament command-
ments. For, in the Old Testament, God’s commandment “Thou shalt not
kill” actually had the form “Thou shalt not kill—unless, of course, I
176 Chapter 4
command you to make a sacrifice of your son or to smite all the inhabi-
tants of Jericho.”
But how does this different logic account for the fact that subjects
perform better on deontic versions of the selection task than on indicative
versions? The reason, Fodor argues, is that, since deontic conditionals actu-
ally require Q (under the condition that P), attention is immediately drawn
to the not-Q case. For example, in the drinking-age problem, the deontic
conditional actually prohibits drinking alcohol (under the condition that
individuals are under twenty-one), so subjects immediately begin looking
for those who are drinking alcohol, since they violate the prohibition. Sub-
jects then examine the violators in order to determine whether they are
among those on whom the prohibition against drinking alcohol actually
falls. Thus, the increase in the frequency with which so-called not-Q cards
are selected in the deontic versions of Wason selection tasks is due to the
fact that the not-Q cards represent direct violations of the obligations that
are categorically imposed by the deontic rules.
In an effort to refute Fodor, the psychologist Philip Beaman conducted
Wason selection tasks in which subjects were presented with “social-
contract rules” formulated in the usual way and also reworded in accor-
dance with Fodor’s analysis. If Fodor’s analysis is right, Beaman argued,
making the logical form of the deontic conditionals even more perspicu-
ous in accordance with Fodor’s analysis should facilitate performance on
“social-contract” selection tasks even more. Rather than refuting Fodor’s
analysis, however, Beaman found that rewording deontic conditionals in
accordance with Fodor’s analysis did, indeed, facilitate performance.
Whereas approximately 67 percent of subjects selected the P and not-Q
cards in response to the usually formulated “social-contract rule,” a full 90
percent of subjects selected those cards in response to the Fodorized rule.
This indicates that Fodor’s analysis has tapped the logic that subjects are
responding to in “social-contract” versions of the Wason selection task and
that, when that logic is made fully explicit, they perform even better.
Thus, the results from the Wason selection tasks don’t demonstrate a
“content effect” according to which subjects are able to reason more effec-
tively about social contracts than about non-social-contract facts. Rather,
the results demonstrate a logic effect, according to which subjects grasp the
different logical properties of indicative and deontic conditionals, and then
select the so-called not-Q cards with greater frequency in response to the
latter because the appropriateness of that response is made more perspicu-
ous by the logical form of deontic conditionals.
Modularity 177
Fodor suggests, subjects are able to accurately focus in on the P and not-Q
cards in those tasks.
If this hypothesis is correct, then if subjects were presented with more
natural indicative conditionals, embedded within background stories that
made very perspicuous the connection between antecedent and conse-
quent conditions, they would select the P and not-Q cards with as high a
frequency as they’re selected in deontic selection tasks. Interestingly,
Sperber, Cara, and Girotto have conducted some experiments that confirm
this prediction.
Sperber, Cara, and Girotto conducted three experiments in which some
subjects were presented with the standard Wason selection tasks involving
conditionals with arbitrary connections between antecedent and conse-
quent conditions, and other subjects were presented with more natural
indicative conditionals. The more natural conditionals were presented in
what they called “relevance conditions,” background stories that served to
make perspicuous the logic of the conditionals being presented. In each of
the experiments, the “relevance conditions” conformed to the following
pattern. In the background story, either a character in the scenario or the
experimenter asserted that there are cases of P and not-Q. Another charac-
ter in the scenario then denied this assertion by claiming, “If P, then Q.”
Subjects were then asked to turn over the necessary cards in order to
evaluate the truth of if P, then Q. The interesting feature of all of the “rel-
evance conditions” is that they present if P, then Q as being the contradic-
tory of P and not-Q. This serves to make it very clear that the logic of the
conditional is that of an indicative, which is falsified by conjunctions of
P with not-Q.
In their first experiment in this vein, they presented some subjects with
the standard letter-number problem, which represented the “nonrelevance
condition.” The other subjects were presented a problem in a “relevance
condition.” This problem involved a scenario in which a religious leader
is suspected of trying to create an elite group of virgin mothers. The reli-
gious leader dismisses the suspicion by asserting, “If a woman has a child,
she has had sex.” Subjects were then given four cards that read children:
yes, children: no, sex: yes, and sex: no. They were instructed that these cards
described four women in the religious group and that they were to deter-
mine whether the religious leader’s claim is true. In this particular experi-
ment, Sperber, Cara, and Girotto found that 78 percent of subjects chose
the P and not-Q cards (children: yes and sex: no), which is as high as the fre-
quency normally obtained in deontic selection tasks. In contrast, only 26
percent of subjects chose the P and not-Q cards in the “nonrelevance”
182 Chapter 4
having logical forms that make sense in the contexts in which those state-
ments are made. And often that process of interpretation involves assign-
ing the statement a logical form that does not accord with the surface
structure of the natural-language statement. Indeed, this is precisely how
logicians apply logic to natural languages and how they teach their stu-
dents to apply logic to natural languages. As a result, simply “switching”
a conditional or making minor changes to the wording of selection tasks
is insufficient to change the way that subjects represent the logical forms
of the conditionals in those tasks.
In fact, the results that Cosmides obtained in the “switched” selection
tasks support the hypothesis that subjects “normalize” the switched con-
ditionals and select cards appropriate to the normalized conditionals. In
the “unswitched” cassava-root problem, 75 percent of subjects selected the
eats cassava root and no tattoo cards, which were the logically correct cards
in that problem. In the “switched” version of the problem, however, only
67 percent of subjects selected those cards. If subjects were ignoring the
logic of the conditionals, and focusing only on whether a man accepted a
benefit without meeting a requirement, what accounts for the drop in per-
formance? If subjects don’t really care about logic in these problems, and
are simply looking to select the eats cassava root and no tattoo cards, then
why didn’t the full 75 percent of subjects choose those cards in the
“switched” version of the problem? Cosmides offers no answer to this
question. But I think that the answer is obvious. The answer is that sub-
jects are focusing on logic, but that the logical form of the “switched”
conditional doesn’t correspond to its surface grammar, so it’s harder to
interpret. While 67 percent of subjects “normalize” the “switched” condi-
tional, and then select the cards that are logically correct relative to the
normalized conditional, some subjects are simply stymied by the task of
evaluating compliance with a conditional rule that doesn’t really make
sense in the context of the problem. Thus, even though subjects are
responding to the same mental representation of logical form in both the
“switched” and “unswitched” versions of the problem, there is a decline
in performance on the “switched” version of the problem because that
logical form is harder to represent in the context of the problem.
A similar phenomenon accounts for the results obtained by Gigerenzer
and Hug. Rather than “switching” the conditional to which subjects were
instructed to determine compliance, Gigerenzer and Hug made minor
modifications to the instructions in order to “switch” the perspective from
which subjects were asked to evaluate the same conditional. In the
“employee version” of the problem, subjects were asked to imagine being
Modularity 187
an employee who wants to know whether the employer has violated the
rule “If an employee works on the weekend, then that person gets a day
off during the week.” In the “employer version,” subjects were asked to
imagine being the employer who wants to know whether any employees
have violated the identically worded rule. But instructing subjects that
they are to adopt the perspective of an employer looking for cheating
employees will affect more than simply the “perspective” that subjects
adopt on the conditional rule. Indeed, it should induce subjects to assign
the conditional a logical form that makes sense of the task they are given.
For “employees,” that form will correspond to the surface grammar of the
conditional given. But, for “employers,” the given conditional imposes
no obligation on employees, if that conditional is simply taken literally.
“Employers,” however, are instructed to find cases of cheating by employ-
ees. If they are to execute that task meaningfully, they need to find cases
in which employees have violated a deontic conditional that actually
imposes an obligation on employees. Working with the wording in the
problem, the only deontic conditional that would actually impose a sen-
sible obligation on employees is “If an employee gets a day off during the
week, then that employee must work on the weekend,” which would be
the “normalized” form of the conditional given (relative to the perspec-
tive of an employer). Relative to the normalized conditional, then, the
logically correct response from an “employer” is to select the did not work
on the weekend and did get a day off cards, which is what 61 percent of the
“employers” did.
So “employers” aren’t actually violating the logic of the conditional by
simply looking for whatever represents being cheated from their perspec-
tive; they are reasoning logically about the only sensible representation of
the logical form of the conditional in the context of the task they’re
assigned. And, again, evidence of this is the fact that “employees” perform
better on the task than “employers.” For, if subjects simply look for an
instance of being cheated in a social contract, what accounts for the fact
that 75 percent of “employees” identify the cards that represent their being
cheated by the employer, whereas only 61 percent of “employers” identify
the cards that represent their being cheated by employees? If subjects
ignore the logic of the problem, and look only for cases in which someone
has accepted a benefit without paying a cost, “employers” should perform
as well as “employees” in the task. On the other hand, if subjects apply
logical principles to solve the problems, then the version of the problem
that doesn’t literally make sense should elicit a lower performance from
subjects, because they have to first assign the conditional rule a logical
188 Chapter 4
form that makes sense given the task demands. Thus, a significant number
of “employers” simply get stymied by a nonsensical task, although 61
percent of them “normalize” the conditional to a deontic conditional that
makes sense given their task and choose the logically correct cards relative
to that normalized conditional.
In sum, the results from Cosmides’ “switched-conditional” experiments
and Gigerenzer and Hug’s “switched-perspective” experiments simply do
not show that subjects are looking for cheaters rather than applying logical
principles in solving the selection tasks in those experiments. Cosmides’
interpretation of the results rests on the mistaken supposition that, if sub-
jects apply logical principles, they do so without interpreting the linguistic
input in the problems. But verbal comprehension tasks routinely involve
interpreting statements as having logical forms that differ from the
forms of their surface grammar, particularly when such interpretation is
necessary to make sense of a statement or a task. Cosmides’ “switched-
conditional” problems and Gigerenzer and Hug’s “switched-perspective”
problems are precisely instances of tasks that don’t make logical sense.
In order to make sense of the problems, subjects interpret the condi-
tionals as having logical forms that differ from their surface grammar, and
then they select the logically correct cards relative to their mental repre-
sentations of logical form. Thus, the findings fail to demonstrate that sub-
jects do not apply logical principles in solving selection tasks involving
social contracts.
The third finding that purportedly provides evidence of a cheater-detec-
tion module is Fiddick, Cosmides, and Tooby’s finding that subjects select
cards that represent cheating even in a version of the problem in which
no conditional rule is provided, but in which two individuals merely state
their wants. Fiddick, Cosmides, and Tooby argue that, if subjects solved
the problem by applying logical principles, they would be unable to solve
this “want version” of the problem, because it contains no logical con-
nectives to which logical principles can apply. But, again, this argument
wrongly presupposes that, if subjects are not given statements whose
surface grammar contains apparent logical form, they will not interpret the
statements as having a logical form. Again, for the purposes of drawing
logical inferences in Wason selection tasks, however, it’s not the surface
grammar of the conditionals (or want statements) that matters, but the
subject’s mental representation of the logical form of the conditionals (or want
statements). And, in the “want version” of the selection task that Fiddick,
Cosmides, and Tooby conducted, there is a clear, albeit implicit, logical
Modularity 189
form, and that logical form is the same as the logical form of the condi-
tional in the “conditional version” of the task with which it was paired.
In the “conditional version,” recall, subjects were given the conditional
“If you give me some potatoes, then I will give you some corn,” and they
were then instructed to turn over the necessary cards to determine whether
“you” have been cheated. In the “want version,” the instructions are identi-
cal, but subjects were to imagine hearing another person say, “I want some
potatoes,” and replying, “I want some corn.” In the “conditional version,”
67 percent of subjects selected the you gave this person potatoes and this
person gave you nothing cards, while 50 percent of subjects selected those
cards in the “want version.” Of course, against the background of being
instructed to determine whether “you” have been cheated, subjects will nat-
urally represent the “want version” of the problem as embodying a deontic
logic. In particular, subjects will represent the want statements as estab-
lishing an agreement that “If you give me some potatoes, then I will give
you some corn,” which imposes an obligation on “me.” They will proba-
bly also represent those statements as establishing an agreement that “If I
give you some corn, then you will give me some potatoes,” which imposes
an obligation on “you.” But the instructions are clear that subjects are to
determine whether “you” have been cheated, so it is the former deontic
conditional that will be operative in their thinking about the problem,
since that is the only sensible representation of the situation in accordance
with which it is possible that “I” can cheat “you” by not giving “you” corn
when “you” give “me” potatoes. Again, Fiddick, Cosmides, and Tooby mis-
takenly assume that, if subjects apply logical principles to selection tasks,
they apply those principles to the surface grammar of the English, rather
than to the mental representations of the logical form of the English that
are the output of verbal comprehension.
And, again, if subjects are sensitive only to accepting benefits without
paying costs, and not to the logical properties of the selection tasks they’re
presented, it’s unclear why the 67 percent response rate in the “conditional
version” of the problem should drop to 50 percent in the “want version.”
In both versions, the acceptance of the benefit without paying the cost is
the same. What is not the same in the two versions is the explicitness of
the logic of the problem. In the “conditional version,” the logic of the
problem is explicit in the surface grammar of the conditional. But in the
“want version,” the logic of the problem is only implicit, and subjects must
perform complex cognitive processing in order to comprehend the logic
of the situation. Given this fact, if subjects apply logical principles to solve
190 Chapter 4
mind given their avoidance of interaction with other people and their
inability to attend to complex and changing environmental stimuli. For
acquiring a theory of mind would surely require a great deal of extended
interaction with others, since learning mechanisms would require a lot of
data about the behavior of others and how others explain their behavior,
and this could only be acquired through extended interaction. But that
interaction in turn would require an attention to environmental com-
plexity that autistic children can’t give.
The philosopher Philip Gerrans points out that autistic children are not
alone in this respect. “Deaf children of hearing parents as well as congen-
itally blind children suffer autistic-like deficits in social, communicative,
and imaginative abilities, as well as selective incapacity to pass reasoning
tasks with a mentalistic component.”40 Indeed, the psychologists Candida
Peterson and Michael Siegal report that deaf children of hearing parents
perform comparably to autistic children on the false-belief test. The reason,
Peterson and Siegal found, is that hearing parents do not attempt to com-
municate with their deaf children about abstractions such as mental states.
Virtually all communication between hearing parents and deaf children,
instead, concerns objects that can be easily pointed to in the visual envi-
ronment. In other words, deaf children of hearing parents, like autistic
children, do not have ready access to information about the mental states
of others, which is a prerequisite for the kinds of learning necessary to
acquire a theory of mind. Consequently, it’s really not clear what we can
infer about the theory-of-mind module hypothesis from the fact that autis-
tic children can’t pass false-belief tests.
Second, Baron-Cohen’s tests of the theory-of-mind module hypothesis
take the false-belief test as a criterion for possessing a theory of mind. But,
as the cognitive scientists Paul Bloom and Tim German argue, it’s not at
all clear that the false-belief test in fact tests specifically for the possession
of a theory of mind. On the one hand, they argue, the ability to pass the
false-belief test requires more than just a theory of mind; and, on the other
hand, possessing a theory of mind involves more than the ability to pass
the false-belief test.
Consider first why passing the false-belief test involves more than a
theory of mind. As noted earlier, three-year-olds typically fail the false-
belief test, while four-year-olds typically pass. The same is true, however,
of the “false-photograph” test. In the false-photograph test, children are
taught how to use a Polaroid camera. They are then instructed to take a
picture of a scene in which a stuffed cat is sitting on a chair next to a bed.
After the picture is taken, and the snapshot is removed from the camera,
194 Chapter 4
the cat is moved to the bed. The child is then asked two questions: “Where
was the cat when you took the photograph?” and “In the photograph,
where is the cat?” Three-year-olds typically answer that, in the photograph,
the cat is on the bed, despite saying that the cat was on the chair when
the photograph was taken, while four-year-olds typically answer both ques-
tions correctly.
Passing the false-photograph test clearly doesn’t involve reasoning with
a theory of mind, yet children typically either fail both it and the false-
belief test or pass both. If three-year-olds passed the false-photograph test
but failed the false-belief test, then their failure on the false-belief test could
reasonably be attributed to their not having developed a theory of mind.
Conversely, if four-year-olds passed the false-belief test but failed the false-
photograph test, then their success with the false-belief test could reason-
ably be attributed to their having developed a theory of mind but not more
general principles of sophisticated reasoning. Since children pass or fail
both tests at the same stages of development, however, it appears that
success in both tests is the result of having acquired sophisticated princi-
ples of counterfactual reasoning that apply to more than just beliefs.
Indeed, one of the first things that a child beginning to understand
mental states would learn is that beliefs are a way of tracking the way the
world is. Consequently, one of the first useful heuristics a child would
acquire is that people’s beliefs tend to be true. That is, a child will typi-
cally reason that Daddy says that he thinks there’s milk in the refrigerator
because there is milk in the refrigerator. Only subsequently, and as a result
of ample experience, will a child learn that people’s beliefs can deviate,
sometimes wildly, from the way the world is. But this step involves more
than simply reasoning about beliefs. It involves reasoning about two
states of affairs—one’s beliefs, and the aspects of the world those beliefs
are about—and sophisticated counterfactual reasoning about what would
happen if the world changed. These principles of counterfactual reasoning
go beyond mere reasoning about beliefs, and they appear to be operative
in the false-photograph test in addition to the false-belief test.
Interestingly, older autistic children typically pass the false-photograph
test, while still failing the false-belief test. This has been taken, by Baron-
Cohen and others, as evidence that autistic children specifically lack a
theory of mind—that they are like “normal” three-year-olds with respect
to reasoning about human behavior, but are like “normal” peers in their
abilities to reason about inanimate objects. But this isn’t necessarily evi-
dence that autistic children have an impaired theory-of-mind module.
Perhaps those autistic children who are able to pass the false-photograph
Modularity 195
On Domain-General Mechanisms
This has been a wide-ranging chapter, so it might pay to take stock of where
we’ve arrived. We’ve seen that none of the arguments for the massive
196 Chapter 4
None of this is to say that the human brain doesn’t learn. It surely does.
It is merely to say that the traditional concept of learning doesn’t ade-
quately capture the process by which the brain forms the specialized cir-
cuits that perform cognitive functions. It is also not to say that the brain’s
circuits don’t process information. They surely do. Some of them process
information using domain-specific rules, and some process information
using domain-general rules. It is merely to say that we don’t come by our
specialized competences solely through a process of applying rules to infor-
mation in order to generate and acquire more rules and information.
Essential to the process of acquiring specialized competences is the process
of forming the brain circuits that subserve those competences, and essen-
tial to that process in turn is the process of proliferate-and-prune by which
the brain adapts to its local environment. This process gives rise to domain-
dominant information-processing circuits in the brain, which in turn give
us the highly specialized cognitive structures that are similar to Evolu-
tionary Psychology’s modules.
At this point Evolutionary Psychologists may be tempted to respond that
brain plasticity isn’t actually inconsistent with the massive modularity
thesis. What matters, they could argue, is not how the brain succeeds in
developing and organizing itself into functionally specialized brain cir-
cuits, but only that it succeeds in doing so, even if the method it employs
to do so is proliferate-and-prune. As long as functionally specialized
mechanisms, which function to solve adaptive problems, somehow emerge
during the course of development, and somehow have regularly emerged
throughout human evolutionary history, these mechanisms can still be
adaptations, having been tailored by natural selection to solve their respec-
tive problems. For, as Tooby and Cosmides argue, “it is primarily the
information-processing structure of the human psychological architecture
that has been functionally organized by natural selection, and the neuro-
physiology has been organized insofar as it physically realized this cogni-
tive organization.”42 Thus, Evolutionary Psychologists might conclude,
details about brain plasticity don’t refute the massive modularity thesis,
since that thesis concerns the information-processing structure of the mind;
and as long as the right information-processing structure emerges, the
physical process by which it emerges is irrelevant.
But there are two problems with this line of argument. First, as explained
in detail earlier in this chapter, characteristics that emerge from the inter-
action between a plastic system and its environment—such as antibodies
or functionally specialized brain circuits—are not biological adaptations.
So the details about brain plasticity do undercut Evolutionary Psychology’s
200 Chapter 4
claim that our functionally specialized brain circuits are adaptations. Those
circuits simply don’t have the right kind of causal history to count as bio-
logical adaptations. They weren’t shaped by selection over our species’ evo-
lutionary history; they are shaped by the local environment during the
course of an individual’s lifetime.
Second, the response presupposes that selection can build adaptations
at an abstractly functional level, without bothering with the details about
how to build the physical structures that implement the functions. But, as
explained earlier, adaptations are traits that have been preserved and
modified by selection. Selection, however, acts only on genetic differences
between individuals. Thus, an adaptation is a trait that has been shaped
by selection through a process in which selection retained the genes that
beneficially modified the trait and discarded genes that detrimentally
modified it.
Genes, however, affect only the protein structures in an organism’s body.
They can affect the functions that are performed by some body part only
by affecting the physical structure of the part that performs the function.
To build an organism that digests, genes must build a physical structure
that performs the function of digestion; they can’t build digestion in some
ethereal, abstract functional space in the organism. Thus, the genes that
were preserved by selection because of their effects on human psychology
had their effects only by altering the neurophysiology of the brain in such
a way that it performed beneficial psychological functions. Genes simply
can’t affect information processing without affecting the neurophysiology
of the brain. Consequently, selection shaped human psychology only by
altering the neurophysiology of the brain over evolutionary time. In con-
trast to Tooby and Cosmides’ claim, it is the neurophysiology of the brain
that has been organized by selection, and it has been organized in the way
it has because of the functional benefits of that neurophysiology.
Thus, rather than being an irrelevant detail along the way to the devel-
opment of functionally specialized brain circuits, cortical plasticity has
been a primary focus of selection’s creative energy throughout human
psychological evolution. Consequently, it is a mistake to assume that the
products of brain development—the functionally specialized brain circuits
that emerge during the course of brain development—are cognitive adap-
tations. Our primary cognitive adaptation is, instead, the process that con-
tinually generates and modifies these specialized brain circuits. However
“modularized” a human brain becomes in the course of development, it
simply doesn’t contain “hundreds or thousands” of modules that are bio-
logical adaptations.
5 Mating
As we saw in chapter 1, life (in the biological, not existential, sense of the
term) is all about reproductive success—how many copies of one’s genes
one contributes to future generations via the bodies of one’s offspring. We
also saw in chapter 1 that many activities have fitness costs and benefits,
which respectively diminish and enhance fitness. Producing offspring, the
very sine qua non of fitness, is no exception. Indeed, producing offspring
is a costly endeavor.
First of all, barring very recently invented reproductive technologies
(which are too new to have affected evolved motives and preferences), in
sexually reproducing species such as ours, you’ve got to have sex with a
member of the opposite sex in order to produce offspring. But, unfortu-
nately, members of the opposite sex don’t have sex with you just because
you want them to. They’ve got to be enticed into it, one way or another,
and the cost of enticement can range from the metabolic costs of produc-
ing a come-hither wink to the costs of building a bower or obtaining and
presenting gifts over an extended period. Once a partner has been enticed,
the sex act costs the energy involved in doing it (plus the contents of an
ejaculate if you’re male). Then, if sex results in conception and you’re a
female, you’ve just begun to pay. If you’re a human female, you pay the
costs of a nine-month gestation, which exacts an enormous physiological
toll on your body. And, throughout most of our evolutionary history,
ancestral women paid the additional metabolic costs involved in breast-
feeding for several years.
So here is one of Nature’s great inequities. If you’re a woman, the
absolute minimum cost for producing a single offspring is quite high. Not
only do you pay the costs of gestation and lactation, but you also pay the
cost of forgoing any other possible reproductive opportunities with males
other than the father of your offspring—possibly better males than the
father of your offspring—during the period of pregnancy and lactation. If
Mating 203
you’re a man, on the other hand, the absolute minimum cost for produc-
ing a single offspring is the energy expended in copulation and the
contents of a single ejaculate (an inexpensive 300 million sperm and three
milliliters of semen). After a fruitful copulation, a man can get up and
pursue reproductive opportunities with other women, whereas a woman
is committed to the costly act of childbearing. This is a radical asymmetry
in the minimum costs the sexes must pay in order to produce a single
offspring.
Although the costs are real, it’s not like flushing money down the toilet,
since you do get an offspring out of the deal. So these expenditures are
really an investment—what is called parental investment. Parental invest-
ment is standardly defined as any characteristics or behaviors of a parent
that enhance the ability of an offspring to survive and reproduce at a cost
to the parent’s fitness, including diminishment in the parent’s future abil-
ities to mate or care for other offspring. Thus, one way of describing the
above asymmetry between the sexes is that the minimum obligatory parental
investment for women is vastly higher than that for men.
Evolutionary Psychologists derive their hypotheses about evolved mate
preferences from this fact about minimum obligatory parental investment,
and the derivation begins in the work of the evolutionary anthropologist
Robert Trivers. In a classic article, Trivers argued that, when there is a sex
asymmetry in parental investment, selection will tend to make the higher-
investing sex choosier in the mating market, because that sex stands to
lose more by making a poor choice of mate. This greater choosiness on
the part of the higher-investing sex will force members of the other sex
to compete among one another to be chosen. As a result, the higher-
investing sex will appear more cautious in the mating market, while the
lower-investing sex will appear more eager and more intensely competi-
tive in its attempts to attract mates. For example, if males invest nothing
beyond the act of copulation and an ejaculate, leaving females to cover all
costs of parental care, females will be very selective in choosing a mate.
Under these circumstances, males are little more than sperm transport, so
a male’s quality is solely a function of the genes he can provide. Females
will then hold out for males who show signs of having “good genes”—
signs such as good health and bodily symmetry (a purported sign of devel-
opmental stability). And males will compete among themselves to be
chosen by females, attempting to present the best advertisements of “good
genes.”
Trivers’s theory is supported by observations of the mating habits of
many species. Some of the strongest support for the theory comes from
204 Chapter 5
our species. With internal fertilization, a female can always be 100 percent
certain that the offspring she births are hers. But no male can be 100
percent certain that the offspring birthed by his mate are his. For we are a
species in which internal fertilization is coupled with concealed ovulation.
This contrasts with other primates, such as chimpanzees. When a chim-
panzee female is ovulating, her genitals swell and become red, a clear sign
to chimpanzee males that it is time to take action. If a chimpanzee
male wants to sire an offspring, he merely needs to ensure that he sexu-
ally monopolizes a female during her fertile period. Ancestral human
males, in contrast, had no idea when females were ovulating, so they could
never be sure whether they were inseminating a fertile female or not. So,
in order to sire an offspring, they had to copulate with ancestral females
round-the-month. But a lot can happen in a month. The demands of sur-
vival would have required frequent periods during which mates were out
of one another’s sight foraging, for example. A female who had been out
of sight for a mere twenty minutes could have been carrying internally the
inseminate of another male. Even if her mate copulated with her imme-
diately upon their reunion, there was never any sure way to know exactly
what was going on in there. As a result, any issue from her womb was of
uncertain provenance from a male’s perspective. This is known as the
problem of paternity uncertainty.
Given the possibility that a male’s putative offspring are not truly his
own, there is always the chance that the male is investing in another male’s
offspring, thus squandering resources that could be better spent in a
competition to fertilize other females. A female, in contrast, never faces
the potential problem of squandering her parental investment on offspring
she mistakenly believes to be hers. Thus, Evolutionary Psychologists argue,
since human male parental investment can be misspent in a way that
human female parental investment cannot, selection should have designed
males to deliver a lower level of parental investment than females as a
hedge against the possibility of misspending it. In fact, Evolutionary Psy-
chologists further predict, the degree of male parental investment should
be a function of the degree to which a male feels confident in his pater-
nity of offspring.
Nonetheless, because human males provided a fairly high level of
parental investment throughout our evolutionary history, they, like
human females, have evolved to be very selective in choosing a female
with whom they will jointly invest in offspring. However, because the
two sexes provided different forms of parental investment throughout
human evolutionary history, Evolutionary Psychologists argue, each sex
has evolved to prefer as mates those members of the opposite sex who
206 Chapter 5
have an evolved preference for young women (that is, women with
physical features that are correlated with peak reproductive potential).
These hypotheses are derived from general theoretical considerations
regarding the nature of parental investment in our species.
It is worth noting, however, that these preferences are for the qualities
of long-term mates. According to Evolutionary Psychologists, when people
are in the market for short-term mates (one-night stands, for example),
their preferences shift. Men still like fertile women as short-term mates,
Evolutionary Psychologists claim, but men’s standards for short-term
mating typically drop so low that they’re willing to copulate with pretty
much anything that is self-moving (since, after all, sperm is cheap).
Women, on the other hand, are less interested in status and more inter-
ested in intelligence and good looks when seeking a short-term mate. In
what follows, I will not examine these claims about short-term mate pref-
erences, but will focus exclusively on the two core hypotheses regarding
long-term mate preferences.
Each of the hypotheses about long-term mate preferences is separable
into two independent claims. One is a claim about what people prefer, and
the other is a claim about why they prefer it. Each hypothesis, that is, con-
tains a claim that a particular universal preference has evolved in each sex
and a claim that that universal preference evolved because of selection for
it in our evolutionary past (that it is an adaptation). These claims are typ-
ically not separated, because empirical studies in Evolutionary Psychology
are presumed to test both claims simultaneously.
To illustrate, consider the male preference for youth. As Buss says,
“because male reproductive success in humans depends heavily on mating
with reproductively capable females, selection over thousands of genera-
tions should favor those males who prefer to mate with reproductively
capable females.”7 Here a hypothesis about what males prefer (reproduc-
tive capability) is derived from a hypothesis about how selection has acted
during human evolutionary history, which would explain why males have
that preference (it is an adaptation). If we get confirmation of the derived
(preference) hypothesis, it seems to be simultaneous confirmation of the
hypothesis (about past selection) from which it was derived. So, if the evi-
dence shows that males indeed prefer youth, that appears to confirm the
hypothesis that the preference is an adaptation.
Universality enters the picture because, for the reasons discussed in
chapters 2 and 3, Evolutionary Psychologists believe that adaptations are,
of necessity, species universals. This is why, in attempting to confirm
hypotheses about evolved mate preferences, Buss conducted a massive
Mating 209
each and every human male prefers young women and that each and every
female prefers high-status men. It is always possible that certain individu-
als have unusual developmental experiences and end up not possessing
the predicted preferences. But, if there is a truly universal developmental
program that has been designed by selection to reliably produce a prefer-
ence for young females in men and a preference for high-status males in
women, that developmental program should produce those preferences
across a very wide range of conditions. Thus, Evolutionary Psychologists
would maintain, to say that those preferences are “universal” means that
they are observable in all cultures, all historical periods, all economic or polit-
ical systems, all social classes, all religious groups, all “races” or ethnicities, and
all relevant ages of the life cycle. It is this more restricted sense of “univer-
sal” that is operative when Buss claims that female preference for high-
status, resource-holding mates is universal. As Buss says, “women across all
continents, all political systems (including socialism and communism), all
racial groups, all religious groups, and all systems of mating (from intense
polygyny to presumptive monogamy) place more value than men on good
financial prospects.”9
I will argue that, even in this more restricted sense of “universal,” the
data on human mate preferences fail to provide convincing support for
claims of a universal male preference for youth and a universal female pref-
erence for high status. Indeed, I will argue, the mate preferences in which
Evolutionary Psychologists are interested tend to vary with age and social
class, among other things. If this is right, then something is wrong
with the hypotheses about human evolution from which Evolutionary Psy-
chology derives its claims about mate preferences. Let’s turn now to the
evidence for Evolutionary Psychology’s core mate-preference hypotheses,
focusing on the studies that are standardly cited in support of those
hypotheses.
marry at 27.49 years of age. He also asked subjects to give the preferred
age of their mates relative to their own ages. So males were asked to state
how much younger or older than themselves their ideal mate would be.
He found that in every one of the thirty-seven samples males indicated a
preference for younger mates, with average preferences ranging from 0.38
to 7.38 years younger. Pooling all the samples, Buss found that, on average,
males preferred a mate who was 2.66 years younger. “By subtracting the
mean age difference preferred between males and their mates (2.66 years)
from the age at which males prefer to marry (27.49 years), it can be inferred
that males in these samples prefer to marry females who are approximately
24.83 years old. This age preference is closer to peak female fertility than
to peak reproductive value.”10
Without splitting hairs about peak fertility versus peak reproductive
value (or peak reproductive potential, which incorporates both), an average
preferred age of 24.83 years is clearly near the height of female reproduc-
tive potential. Given the large cross-cultural scale of Buss’s study, this
appears to show that male preference for females with high reproductive
potential is universal. And this, in turn, appears to confirm the hypothe-
sis that selection has designed male mate preferences to be highly sensi-
tive to female reproductive potential.
As Buss recognizes, however, males may indicate preferences on a survey
questionnaire that don’t accord with the actual decisions they make in
choosing a mate. In addition, offspring are produced not by preferences
for mates with certain qualities, but by actual matings. Consequently,
selection cannot have acted on male preferences unless males throughout
human evolutionary history actually mated in accordance with their
preferences. In particular, a preference for fertile young women could not
increase in frequency in a population unless there was a strong correlation
between that preference and actually mating with fertile young women.
Thus, there can’t have been past selection for a male preference for young
women unless males with that preference actually produced more off-
spring, by actually mating with fertile young women, than did males with
alternative preferences.
To confirm that the preferences for young women that males reported
on his questionnaire are (and presumably were in our evolutionary history)
reflected in actual mating behavior, Buss compared the age-preference data
with the actual ages at marriage of men and women in thirty of his thirty-
seven samples. He found that the average age at marriage was 28.2 years
for males and 25.3 years for females, only slightly higher than males’ pre-
ferred ages of 27.49 years and 24.83 years respectively.
212 Chapter 5
Of course, it takes two to mate. So, while males may prefer mates who
are 24.83 years old, females have their own preferences, and females
expressed a preference to marry at 25.4 years to a man of 28.8 years. Thus,
the discrepancy between males’ preferred ages of self and spouse at mar-
riage and the actual ages at marriage appears to be a product of compro-
mise with female preference. Indeed, Evolutionary Psychologists argue, we
should expect all actual mating decisions and behaviors to differ from the
preferences of both sexes, since the preferences of the sexes will typically
differ; and, when preferences of the mating parties differ, actual mating
decisions and behaviors will reflect a compromise between the preferences.
Despite the expected compromise, however, the actual average age of
women at marriage is very close to the male preference, so male prefer-
ences for young women do indeed appear to be reflected in actual mating
behavior. Therefore, Buss concludes, the preference data together with the
marriage data provide strong “support for the evolution-based hypothesis
that males both prefer and choose females displaying cues to high repro-
ductive capacity.”11
But Buss’s analyzed data do not clearly confirm this hypothesis. Buss’s
analysis of his preference data consists in subtracting the average preferred
age difference between male respondents and their female mates (2.66
years) from the average age at which his male respondents said they pre-
ferred to marry (27.49 years). Since the average age of his male respondents
was 23.49 years, this shows only that young men say that they prefer to
marry relatively younger women and to do so at a fairly young age. As Buss
recognizes, what males say they want in a mate stands in need of a valid-
ity check, which his analysis of his marriage data purportedly provides. But
Buss’s analysis of his marriage data consists only of comparing the average
age of males at marriage (28.2 years) with the average age of females at mar-
riage (25.3 years). While this does show that on average males marry fairly
young women, it also shows that on average the males marrying them are
themselves young—only 2.9 years older than their brides.
If males both prefer and choose young women as mates, however, this
preference should be present across the male life cycle. Older males should
exhibit a preference for young women just as young males do. Since Buss’s
analysis employs only the averages from his samples, it doesn’t show that
older males prefer and choose young women as mates. The mate prefer-
ences of older males disappear into the averages, and the averages present
a profile of the mate preferences of relatively young males. But, to confirm
that males have an evolved preference for young women, it is not enough
to show that young men prefer young women.
Mating 213
potential in mate choice, with the result that actual choices of mate strike
a balance between the two potentially competing considerations.
This has some interesting implications. For a male in his twenties, like
Buss’s average respondent, similarly aged females are also those near their
peak reproductive potential, so males in their twenties should prefer
females in their early twenties. But, as males age, similarly aged females
are increasingly further from their peak reproductive potential, so older
males must trade off the increasingly competing considerations of age
similarity and reproductive potential. Thus, Kenrick and Keefe predicted,
“whereas aging males should prefer progressively older women (because of
similarity), they should also prefer women progressively younger than
themselves (to maximize reproductive opportunities).”13 That is, as males
get older, the average age difference at marriage between self and spouse
should gradually increase. While the age difference at marriage should be
relatively small for males in their twenties, it should be fairly large for older
males, who must choose females no older than their forties in order to
have mates with some remaining, albeit small, reproductive potential.
Kenrick and Keefe examined all the marriages that took place in Seattle
in January 1986 and a sample of those in Phoenix in January and May
1986. To ensure that their results would not simply be an artifact of 1980s
America, they also examined a sample of one hundred marriages in
Phoenix in 1923. And to further ensure that these combined results would
not simply be an artifact of American culture, they examined all marriages
on the Philippine island of Poro between 1913 and 1939.
Kenrick and Keefe found the same pattern in all their samples. The 1986
samples were virtually identical. In these samples, on average, males who
married in their twenties married females a year or so younger; males in
their thirties married females a few years younger; males in their forties
married females about six years younger; males in their fifties married
females about nine years younger; and males in their sixties married
females about ten years younger. The sample of Phoenix marriages in 1923
showed the same pattern for males in their twenties and thirties, but there
were even greater age differences between older males and their spouses.
In 1923 Phoenix, males in their forties married females about thirteen
years younger, and the age difference between spouses increased a year for
each decade of male age after that. Finally, in Poro, on average, males in
their twenties married females three years younger; males in their thirties
married females about nine years younger; males in their forties married
females about twelve years younger; males in their fifties married females
Mating 217
fifteen years younger; and males in their sixties married females a full
twenty years younger.
Although these data appear to provide straightforward confirmation of
Kenrick and Keefe’s hypothesis that males weigh both age similarity and
reproductive potential in selecting a mate, thus striking a balance between
the two considerations, things are not quite that simple. First, as the
psychologist Kim Wallen points out, the principal period of fecundity for
women is between the ages of twenty and forty, and the average age of
menopause is fifty. But the data show older males, on average, marrying
women who are past the period of principal fecundity and much older
males marrying women who are in their postreproductive years or very
nearly so. If reproductive potential is a significant factor in male mate
choice at all, regardless of the male’s age we should not find males marry-
ing women who are at or very near the end of their reproductive careers.
Of course, males aren’t the only ones doing the choosing. It may be that
males in their late fifties and older are unable, for the most part, to attract
and marry women with significant remaining reproductive potential. So
the fact that older males marry women with little or no reproductive
potential could simply be a result of compromise in the mating market.
Perhaps older men would rather marry significantly younger women, but
they can’t, so they settle for women who are postmenopausal or very
nearly so.
But Kenrick and Keefe also gathered data from personal ads, in which
advertisers indicated a preferred age or age range for their respondents,
and the pattern of average preferred age differences from the ads closely
matched the pattern of average age differences in the marriage data. As the
age of male advertisers increased, the average age difference between them
and their desired respondents also increased. However, although men in
their fifties and sixties did express a preference for much younger women,
on average, the ages they preferred still fell near the end of or beyond female
reproductive potential. So, on average, older males not only marry women
who are postreproductive or nearly so, but seek them as well.
This is not what we should expect given Kenrick and Keefe’s hypothe-
sis. Even if males choose mates by weighing both age similarity and repro-
ductive potential, when a potential mate has little or no reproductive
potential, age similarity should count for little or nothing in mate choice.
For, by Kenrick and Keefe’s account, age similarity factors into male
mate choice only because it facilitates extended cooperation in providing
parental care. But, if a postreproductive mate is chosen, there will be no
218 Chapter 5
offspring for whom to provide parental care. So a preference for age simi-
larity can facilitate parental cooperation only if it plays second fiddle to
the preference for reproductive potential. However, Kenrick and Keefe’s
marriage data and preference data appear to show that a preference for age
similarity among older males virtually trumps any preference for repro-
ductive potential.
A second problem is that the samples of marriages of males in their fifties
and sixties consist almost entirely of males who are remarrying, as Kenrick
and Keefe acknowledge. Evolutionary Psychologists argue that it is enlight-
ening to examine the choices that males of those age groups make when
they are seeking new mates. But their mating decisions present only a
partial picture of the mating decisions of males in those age groups.
Consider the fact that the National Marriage Project of 2000 found that
40 to 50 percent of all marriages end in divorce. Suppose we adopt the
extreme estimate that 50 percent of marriages end in divorce. Some of
these divorces are attributable to serial marriers (or serial divorcers, depend-
ing on whether you’re a romantic or not), for whom two out of three, three
out of four, or even seven out of eight marriages end in divorce. So, even
if as many as 50 percent of all marriages end in divorce, it is not the case
that 50 percent of all those who marry end up getting divorced. At least
50 percent of all men who marry do not get divorced, hence never remarry,
so about half of all men in their fifties and sixties have decided to remain
married. Since this half will have married much earlier in life, by Kenrick
and Keefe’s own data, their wives will be relatively close to their own ages.
The divorce data, and independent data about the frequency of infidelity,
however, shows that married people frequently have the option of taking
up with a new mate. So, these males are making genuine choices to remain
married, since they always have the option of divorcing and looking for a
new mate. Remaining married is actually a continual choice of one’s spouse
over others. Thus, half the older male population is choosing to remain in
mateships with women who are no longer capable of bearing children. A
hypothesis about male mate preferences can’t be tested exclusively on the
males who choose to remarry after fifty. The choices of males who remain
in mateships with no reproductive potential have to be considered as well.
Third, Kenrick and Keefe’s analysis of the data suffers from a problem
that plagues Buss’s analysis as well. They base their analysis entirely on the
averages in their samples and ignore the variation. But, as Kenrick and Keefe
admit: “Individual subjects showed wide variation in their preferences,
however, and in their choice of marriage partners. There were older men
who sought, and others who married, women their own age.”14
Mating 219
Kenrick and Keefe don’t report the variation in their data, but the
evolutionary psychologist Karl Grammer reconstructed the variation from
their personal-ad data. Grammer found the variation to be much higher
than one would expect if males prefer mates with high reproductive poten-
tial. Consider a couple of examples. Among 53-year-old males, preferences
for mate age ranged from 35 to 57, and among 56-year-old males they
ranged from 46 to 52. The marriage data undoubtedly exhibit similar
ranges, although they aren’t reported by Kenrick and Keefe. This means,
however, that a significant number of males in their fifties and sixties both
prefer and choose postreproductive women as mates. And this doesn’t
conform to Kenrick and Keefe’s predictions.
Apart from these specific problems with Kenrick and Keefe’s hypothesis,
there is a general reason why it’s problematic to test hypotheses about mate
preferences against sample averages alone, as both Buss and Kenrick and
Keefe do. As we saw in chapter 1, variation is not only the fuel on which
selection burns, but is itself often produced and maintained by selection.
As a consequence, patterns of variation can be highly significant, because
they can indicate that different, possibly frequency-dependent, strategies
are being pursued. To put this another way, Evolutionary Psychologists
assume that each hypothesis about past selection entails a prediction about
a single adaptation that evolved in response to it. As a result, Evolution-
ary Psychologists tend to focus only on the sample average to see whether
it conforms to the phenotypic value they derive from their hypothesis
about past selection. But hypotheses about past selection can entail the
coexistence of multiple adaptive phenotypes in a population. In such cases,
phenotypic values in a population may be bimodally (or trimodally) dis-
tributed. Such distributions, however, are concealed when only sample
averages are calculated. So, rather than collapsing variation inside sample
averages, we should always ask whether there is a potential explanation of
the variation itself.
There is not sufficient data at this point to strongly confirm any hypothe-
ses about the precise nature and source of variation in male preferences
regarding age differences between themselves and their mates. But there is
sufficient data to suggest a possibility. Let’s review a few of the relevant
facts.
First, even if we focus only on the average age differences between males
and their mates, we find that older males, on average, both prefer and mate
with females who are very near or at the end of their reproductive careers.
So, if males are weighing both age similarity and reproductive potential
in choosing their mates, they are placing too great a weight on age
220 Chapter 5
Keefe and their colleagues conducted a study to test whether males actu-
ally prefer females who are near peak reproductive potential or whether
they simply prefer slightly younger females. They argued that the prefer-
ences of adolescent males provide a crucial test of these alternative
hypotheses. For, if adolescent males prefer older females, they claimed, the
preference for females of peak reproductive potential must be deeply
ingrained in male sexual psychology.
Kenrick and Keefe and their colleagues interviewed 103 males between
the ages of twelve and eighteen, asking them to “think of the most attrac-
tive person you could possibly imagine” going on a date with and to indi-
cate how much older or younger than the respondent that person is.15 They
also asked their male subjects to indicate the minimum and maximum
acceptable ages of a potential date, again in terms of a difference from the
respondent’s age. They found that males from twelve to sixteen indicated
that females who were 3 to 4 years older were most attractive, and that
seventeen-year-old males indicated that females 6 years older were most
attractive. They also found that, averaged across ages, these young males
were willing to date females who were between 1.57 years younger and 6.0
years older than themselves.
Both results seem to indicate an adolescent male preference for some-
what older females, especially females near peak reproductive potential.
“The most interesting feature of these data,” the authors say, “is that
adolescent males expressed an interest in females substantially older than
themselves, despite the fact that older females’ age preferences showed
no evidence of a reciprocal interest in younger males.”16 Thus, since these
preferences couldn’t possibly be experientially induced by reciprocal
female interest, they concluded that the results had to indicate an evolved
preference for peak reproductive potential.
But, according to the hypothesis of shifting reproductive effort, these
results are instead a by-product of the fact that adolescent males, who have
yet to reproduce, represent pure, unadulterated mating effort. And, when
reproductive effort is devoted exclusively to mating, we can expect males
to prefer fertile females. Thus, rather than indicating that male preference
is focused laser-like on the female years of peak reproductive potential,
these results only reinforce that, when preferences are a function of mating
effort alone, males will prefer fertile young females.
Further, these results are not as surprising as Kenrick and Keefe and their
colleagues make out. They claim that their results show the robustness of
male preference for fertile young females because adolescent males cannot
possibly have been encouraged in these preferences by having experienced
226 Chapter 5
reciprocal interest from females in their early twenties. But, given the con-
ditions of the interview, there is nothing at all surprising about the fact
that the adolescent males responded as they did. For the interviewers
primed the subjects with the following instructions: “I’d like you to think
for a second about what type of person you would find attractive. Imagine
you were going to go on a date with someone. Assume that the person would
be interested in you. . . .”17 Here the interviewers are explicitly instructing
the respondents to imagine a situation of reciprocal interest, not the real
world in which older females aren’t interested in adolescent males. If the
adolescent subjects hadn’t been explicitly instructed to assume reciprocal
interest on the part of any female they could imagine, their responses may
have been different. Indeed, if they were responding on the basis of their
real-world preferences, they may have indicated a preference for females
closer to their own ages.
One other study deserves comment before we turn our attention to
female mate preferences. In The Evolution of Human Sexuality, Donald
Symons argues that the preferences of homosexuals provide an “acid test”
for evolutionary hypotheses about mate preferences. For, according to
Symons, the two sexes bring very different sets of desires to the act of
mating, and these differences in desire engender conflict. As a result, het-
erosexual mating, Symons argues, always involves both parties’ compro-
mising their true desires in order to form a less-than-perfect union. But
homosexuals need never compromise, since both parties bring the same
sets of desires to homosexual unions. Thus, Symons claims, the preferences
of homosexual men and women should reveal the pure, uncompromised
evolved desires of each sex.
Playing off this idea, the anthropologist William Jankowiak and his
colleagues had both heterosexual and homosexual men and women rank
photographs of members of their desired sex on the criterion of physical
attractiveness. They found that both heterosexual and homosexual
men consistently ranked the photographs of younger individuals as most
attractive, while both heterosexual and homosexual women placed no
emphasis on youth in their attractiveness rankings. Consequently, in The
Evolution of Desire, Buss takes the fact that homosexual men placed such a
clear emphasis on youth in their attractiveness rankings as clear evidence
of an evolved male preference for youth in a mate.
But it’s not at all clear what to make of this result. For recall that Buss
says, “according to evolutionary psychologists, the evolutionary model
predicts that what men desire is not youth per se, but rather features of
women that are associated with reproductive value or fertility.”18 One of the
Mating 227
John is 5¢10≤ tall, 165 lbs. He has been playing tennis for one year and is currently
enrolled in an intermediate tennis class. Despite his limited amount of training he
is a very coordinated tennis player, who has won 60% of his matches. His serve is
very strong and his returns are extremely powerful. In addition to his physical
abilities, he has the mental qualities that lead to success in tennis. He is extremely
competitive, refusing to yield against opponents who have been playing much
longer. All of his movements tend to communicate dominance and authority. He
Mating 231
tends to psychologically dominate his opponents, forcing them off their games and
into mental mistakes.23
John is 5¢10≤ tall, 165 lbs. He has been playing tennis for one year and is currently
enrolled in an intermediate tennis class. Despite his limited amount of training he
is a very coordinated tennis player, who has won 60% of his matches. His serve and
his returns are consistent and well placed. Although he plays well, he prefers to play
for fun rather than to win. He is not particularly competitive and tends to yield to
opponents who have been playing tennis much longer. He is easily thrown off his
game by opponents who play with great authority. Strong opponents are able to
psychologically dominate him, sometimes forcing him off his game. He enjoys the
game of tennis but avoids highly competitive situations.24
Sadalla, Kenrick, and Vershure then asked their female subjects to rate
the sexual attractiveness and dating desirability of both males. Subjects
used a seven-point scale with 7 being highest. On the dimension of sexual
attractiveness, John the Dominant received an average rating of 5.37, while
John the Nondominant received an average rating of 4.05. On the dimen-
sion of dating desirability, John the Dominant received an average rating
of 4.56, while John the Nondominant received an average rating of 3.49.
Sadalla, Kenrick, and Vershure concluded that females are most sexually
attracted to dominant males and find them most desirable as mates.
You’ll note, however, that the experiment fails to include a male who is
described as neither dominant nor nondominant. One description depicts
John as dominating the other males with whom he competes, while the
other depicts him as easily dominated by his competitors. But there is no
description of a male who neither easily dominates nor is easily dominated
by others. So the psychologists Jerry Burger and Mica Cosby decided to run
the experiment again and include the “missing control condition.”
Burger and Cosby had 118 female undergraduates read the same two
descriptions that Sadalla, Kenrick, and Vershure used, but they added the
following “control” description, which says nothing about whether John
is dominant or nondominant:
John is 5¢10≤ tall, 165 lbs. He has been playing tennis for one year and is currently
enrolled in an intermediate tennis class. Despite his limited amount of training he
is a very coordinated tennis player, who has won 60% of his matches.
Burger and Cosby then had their subjects rate the sexual attractiveness and
dating desirability of all three males. Subjects used a seven-point scale, but
this time 1, rather than 7, was highest (so, in this study, lower scores are
better).
232 Chapter 5
Like the previous study, Burger and Cosby found that, on the dimension
of sexual attractiveness, John the Dominant received an average rating of
3.63, although John the Nondominant received an average rating of 4.11.
Also like the previous study, they found that, on the dimension of dating
desirability, John the Dominant received an average rating of 3.72, while
John the Nondominant received an average rating of 3.97. However, John
the Control, who was described as neither dominant nor nondominant,
received an average rating of 3.19 on the dimension of sexual attractive-
ness and 3.11 on the dimension of dating desirability. Not only are these
scores higher than John the Dominant’s scores, but they are significantly
higher. These results indicate that, although women may prefer dominant
males to males who are easily dominated, they may prefer males who are
neither to both of the others.
The concept of dominance, however, may carry negative connotations,
which may have influenced how female subjects responded to the descrip-
tions they were given. Given only the descriptions provided, female
subjects may have (unconsciously) inferred that John the Dominant’s dom-
inance and competitiveness would be directed at his mate as well. And this
may have made John the Dominant appear less desirable than a male who
is not depicted as so aggressively dominant, especially when that male is
also not depicted as a pushover. It may be that females actually do desire
mates who are toward the top of the male status hierarchy, but that they
just respond negatively to the connotations of the concept of dominance
employed in these experiments. So let’s look at the battery of studies that
have used the other concrete indicator of high status, high SES.
The most fun studies purporting to demonstrate that females prefer high-
SES males were conducted by the anthropologist John Marshall Townsend
and the psychologist Gary Levy. The procedures and results of these studies
were almost identical, so I will focus primarily on one of them, since it is
representative of the others and discussion of it applies equally to the
others.
In Townsend and Levy’s study, 112 white female undergraduates from
Syracuse University were shown photographs of two male models, who
were selected for this purpose because an independent group had rated one
as very handsome and the other as very homely. In the photographs, each
male model was dressed in each of three different “costumes.” One was
the uniform of a Burger King employee, intended to depict low SES.
Another was a plain off-white shirt, to depict medium SES. And the third
was “a white dress shirt with a designer paisley tie, a navy blazer thrown
Mating 233
over the left shoulder, and a Rolex wristwatch showing on the left wrist,”
to depict high SES.25
The female subjects were asked to indicate their degree of willingness to
enter six types of relationship, ranging from very casual to very serious,
with “someone like” the models. The six types of relationship were
described as “coffee and conversation,” “date,” “sex only,” “serious
involvement, marriage potential,” “sexual and serious, marriage poten-
tial,” and “marriage.” Subjects were instructed to indicate their degree of
willingness on the following five-point scale: (1) very willing, (2) willing,
(3) undecided, (4) unwilling, and (5) very unwilling.
To simplify discussion, I’ll focus on three representative relationship
types: “date,” “serious involvement, marriage potential,” and “marriage.”
The average ratings from the female subjects for these three relationship
types are presented in table 5.1.
Townsend and Levy found two aspects of the results in table 5.1 to be
significant. First, at each level of involvement (from date to marriage), each
model got better ratings in the high-SES costume than in the other cos-
tumes. The sole exception was that the handsome model was a slightly
more desirable date in the medium-SES costume than in the high-SES
costume. As the level of involvement became more serious, however, even
Table 5.1
Average Female Willingness to Enter a Relationship as a Function of Male “Costume”
Handsome model
Costume
Homely model
Costume
the handsome model scored better in the high-SES costume than in either
of the other costumes. Second, high status appears to compensate for
homeliness, since the high-SES costume raised the homely model’s accept-
ability at every level of involvement over that of the handsome model in
the low-SES costume. Townsend and Levy conclude that females prefer
high-status mates. And the Evolutionary Psychologist Bruce Ellis interprets
the experiment as showing that “status and economic achievement are
highly relevant barometers of male attractiveness, more so than physical
attributes.”26
But, if we look at the data in a different way, we see a different barom-
eter. For, at every level of involvement and at every status level, the hand-
some male is preferred over the homely male. In fact, the handsome male
in the medium-SES costume scored better than the homely male in the
high-SES costume, despite the fact that the latter can presumably provide
more resources than the former. So the results could equally well be inter-
preted as showing that physical attributes are “highly relevant barometers
of male attractiveness,” more so than male status.
In addition, although Townsend and Levy claim that status compensates
for homeliness, since the homely model in the high-SES costume scored
higher than the handsome model in the low-SES costume, the homely
model scored 3.20 and 3.17 for serious involvement and marriage respec-
tively. Both of these scores lie between “undecided” and “unwilling” in the
scale used to indicate degree of willingness to enter a relationship with
someone like the model. So high SES doesn’t appear to make females willing
to enter a serious relationship with someone they find homely. (Also, the
homely model inexplicably scored better in the low-SES costume than in
the medium-SES costume, despite the fact that medium-SES males can
provide resources that low-SES males can’t. There is no doubt some fashion
advice lurking here about who should wear plain off-white shirts.)
While Evolutionary Psychologists typically deemphasize the role of male
physical attractiveness in female mate preferences, both of these objections
could easily be accommodated simply by claiming that women weigh both
physical attractiveness and status in choosing a mate. The data, it could
be claimed, demonstrate that both have significant effects on female mate
preferences. Thus, the data don’t have to be a perfect fit to the prediction
that females prefer high-status males, they only need to show that status
plays a role in female mate preferences. And the data do seem to demon-
strate this.
But there is a problem with the data nonetheless, which is due to the
fact that Townsend and Levy’s subjects were all university students. As we
Mating 235
have seen, homogamy is the most robust mating phenomenon, and status
homogamy is second only to race homogamy in the strength of its effect
on mate choice. There are two possible explanations for this. First, it may
be that everyone competes for high-status mates, but only high-status indi-
viduals succeed in attracting high-status mates. This effect could trickle
down to other status levels, with the result that everyone ends up with a
mate of comparable status. Second, it may be that people actually prefer
mates of comparable status, and status homogamy results from this pref-
erence. The sociological evidence actually favors the second explanation.
People mate with those of comparable status primarily out of preference,
rather than settling for a mate of comparable status because of an inabil-
ity to get a mate of higher status.
But why would anyone prefer to mate with a fellow medium-status indi-
vidual, say, rather than a high-status individual? The reason could well be
the one that Kenrick and Keefe give for why age similarity is important in
mate choice. Recall that Kenrick and Keefe argued that all forms of simi-
larity between mates may facilitate long-term parental cooperation, since
homogamously mated individuals have “similar expectations, values,
activity levels, and habits.”27 Consequently, they suggested, selection may
have favored homogamous mating—a “birds of a feather mate together”
principle—among our ancestors. If this is right, then we should expect
people to prefer mates of comparable status in addition to mates of similar
age. As a consequence, assortative mating by status would be fairly robust,
which in fact it is.
Studies of status homogamy have considered four different dimensions
of status: education level, cultural status of occupation, income level, and
social-class origins. Of these, educational homogamy is the most robust.
Overwhelmingly, people select mates who have achieved (or will achieve)
a comparable level of education. Indeed, one of the boundaries that is very
rarely crossed in mating relationships is the boundary between those who
have some college education (like Townsend and Levy’s female subjects)
and those who have only a high-school education (presumably like
Townsend and Levy’s Burger King employee). Consequently, the fact that
Townsend and Levy’s subjects preferred medium- to high-SES males over
low-SES males could simply be an artifact of assortative mating by status.
The results could simply be due to the fact that the female subjects per-
ceived the low-SES model as uneducated and tended to consider only males
of probable comparable education level as prospective mates.
This explains why the models would score so low in the low-SES cos-
tumes, but why would they score higher in the high-SES costumes than in
236 Chapter 5
a year. Guess what. The female law students preferred the doctor. But this
doesn’t necessarily indicate a preference for high SES per se. For this pref-
erence is precisely what we should expect if female law students were
choosing mates of comparable education level, comparable cultural status
of occupation, and comparable projected income level.
Further, the female law students rated the high-SES models higher than
the college students had. This is also precisely what we should expect from
assortative mating by status, since the law students form a more homoge-
neously high-status sample than the undergraduates. Because the under-
graduate sample was less status homogeneous, containing a higher ratio
of medium-SES respondents, they gave a lower average score to the high-
SES model. In short, the Townsend and Levy data fit the preference ratings
that would result from status homogamy alone.
Thus, given the composition of the subject groups in these experiments,
none of the experiments can distinguish whether female respondents were
indicating a genuine preference for a mate with high SES or whether their
ratings were a product of simple assortative mating by status. Given the
independently documented robustness of status homogamy, we already
know that, if you ask medium- and high-SES females what they want in a
mate, they will show a preference bias against low-SES males. In order to
distinguish a genuine preference for high status from assortative mating
by status, we would need data on the preferences of low-SES females. We
would also need a subgroup analysis by status of female preferences; that
is, we would need to see female preference orderings broken down by SES
of female respondent. Evolutionary Psychologists have provided no such
data, but it would be needed to substantiate their claim that females desire
high-SES males. For without this data, there is no evidence that medium-
SES females don’t prefer the medium-SES model over the ostentatious high-
SES model, or that low-SES females don’t find the handsome model in the
Burger King uniform most desirable of all. Only if low-SES females sys-
tematically indicate a preference for high-status males can we infer a real
preference for high status. However, the independent evidence of status
homogamy suggests that the latter preference pattern would not be found.
Of course, it is entirely possible that women actually have an evolved
preference for high-status men, but that the experiments conducted by
Evolutionary Psychologists have simply failed to reveal that fact because
their results are confounded by status homogamy. This could be the case,
for example, if dominance and high SES, as those are articulated in the
experimental instruments used by Evolutionary Psychologists, are poor
measures of the kind of status that women have evolved to prefer. After
238 Chapter 5
all, this preference is supposed to have evolved long before people had
education levels and incomes or had careers as doctors and lawyers. And
the ability to easily dominate opponents in tennis matches, which weren’t
played in the Pleistocene, may not set off alarms in women’s evolved
status-detection modules. So maybe Evolutionary Psychologists have
simply used experimental procedures that are incapable of detecting
female preference for high-status mates, though it is there waiting to be
detected.
Although this is a possibility, I doubt it’s true. As Evolutionary Psychol-
ogists standardly define the concept, status “refers to an individual’s
relative position in a social group; it is a measure of where one stands among
one’s peers and competitors.”28 If selecting a high-status mate had, and
continues to have, significant fitness consequences for women, as Evolu-
tionary Psychologists claim, we should expect women to have evolved
techniques of detecting social relations among males. They should have
evolved to be sensitive to the structural features of male interactions,
detecting how males form a hierarchy in their interactions with one
another, regardless of which particular intrinsic male qualities are corre-
lated with high or low status. And, if women have evolved to be sensitive
in this way to where a male stands in relation to other males, and to prefer
those who are in the upper strata of male hierarchies, the kinds of exper-
iment we’ve discussed should detect that preference. So, the fact that the
results of these experiments are confounded by status homogamy proba-
bly isn’t due to the experiments’ making use of cues that didn’t indicate
status in the Pleistocene.
At this point Evolutionary Psychologists could respond that assortative
mating by status cannot explain the sex differences they consistently find.
For studies of mate preferences consistently find that women place a
greater emphasis on a potential mate’s status than do men. If female pref-
erences for high status are simply a by-product of the fact that the female
experimental subjects are relatively high status themselves and are merely
indicating a preference for males of comparable status, then we should
expect male experimental subjects (who tend to be from the same social
classes as female subjects) to place a comparable emphasis on a potential
mate’s status. For, if people prefer mates of similar status, that should be
evident not only in female preferences, but in male preferences as well.
But study after study finds that a potential mate’s status doesn’t matter to
men in the way that it matters to women.
The studies showing a sex difference with respect to preferences for high-
status mates have focused on the income dimension of status, finding that
Mating 239
women care more about a potential mate’s earning capacity than men do.
One such study was conducted by Douglas Kenrick in collaboration with
the psychologists Edward Sadalla, Gary Groth, and Melanie Trost. Kenrick
and his collaborators provided 64 female and 29 male undergraduates with
a list of characteristics and asked them how they would weigh those char-
acteristics in choosing a partner for a date, for steady exclusive dating, and
for marriage. “Participants were asked to give the minimum and maximum
percentiles of each characteristic that they would find acceptable in a
partner at each level of involvement. Several examples were given to clarify
any questions about the percentile concept, e.g., ‘A person at the 50th per-
centile would be above 50% of other people on [the characteristic] kind
and understanding, and below 49% of the people on this dimension.’ ”29
The primary evidence supporting the claim that females desire high-
status mates came from responses regarding minimum acceptable earning
capacity. With respect to this characteristic, average responses for each sex
were as presented in table 5.2, and these results do appear to show that
male earning capacity plays a strong role in female mate choice. Indeed,
in accordance with Evolutionary Psychologists’ expectations, as the level
of involvement becomes more serious, and consequently a male’s ability
to provide resources becomes increasingly important, females appear to
place increasing weight on earning capacity. And this result appears to
confirm that women prefer high-SES males.
Again, however, the female subjects were all American undergraduates,
hence, on average, of medium SES or higher. (Education level alone would
make them of medium SES, but some may be higher because of higher
social-class origins, anticipated high-status occupations, or anticipated
high income levels after college.) Thus, again, it is not clear whether the
results actually demonstrate a preference for high status per se, or whether
they merely reflect status homogamy. For, again, if the subjects are pre-
dominantly from the upper half of the socioeconomic continuum, and if
Table 5.2
Average Minimum Acceptable Earning Capacity of a Potential Partner
hypothesis doesn’t entail that medium-SES women should have lower stan-
dards for income in a potential mate than high-SES women.
However, third, if women prefer males in their own socioeconomic
group, then medium-SES women will, on average, exhibit a preference for
medium-SES males, which will appear as a preference for a moderate ability
to provide resources. High-SES women, on the other hand, will exhibit a
preference for high-SES men, which will make them appear to desire an
even greater ability to provide resources than is desired by medium-SES
women. Such preference patterns are precisely those that assortative
mating by status should lead us to expect. So, again, the data are unable
to demonstrate a female preference for high status per se rather than simple
assortative mating by status.
Of course, all the studies so far considered have involved only American
female subjects. So perhaps the results of these studies are confounded by
status homogamy simply because the studies made use of unrepresentative
samples. If so, a much larger, cross-cultural study should be able to provide
results that are not confounded by status homogamy.
This larger study would obviously be Buss’s cross-cultural study, in which
he had his male and female respondents rate the importance of eighteen
characteristics in choosing a mate. Respondents used a four-point scale,
ranging from 0 (irrelevant or unimportant) to 3 (indispensable). In thirty-
six of Buss’s thirty-seven samples, females valued the characteristic good
financial prospect significantly more than did males. In the remaining
sample (Spain), females valued it more than males, but the difference
was not significant. In the entire study, the average female rating of good
financial prospect was 1.76 and the average male rating was 1.51. This
appears to show that female preference for high-SES mates is not an
artifact of unrepresentative American samples, but is in fact a robust
universal preference.
Despite the size of Buss’s sample, however, and despite the number
of cultures sampled, the sample is still not representative in the way
that would be needed to distinguish a preference for high SES per se from
simple assortative mating by status. As Buss himself admits: “The samples
obtained cannot be viewed as representative of the populations in each
country. In general, rural, less-educated, and lower levels of socioeconomic
status are underrepresented.”32 But these underrepresented groups are pre-
cisely those who might, on average, place much less emphasis on earning
capacity because of status homogamy. In fact, the preference ratings of
these groups could significantly lower the average rating of good financial
prospect and thereby remove the appearance of a female preference for
244 Chapter 5
high-SES mates. For, as we have seen, in samples that are skewed toward
the upper half of the socioeconomic continuum, assortative mating by
status biases preference ratings in favor of males in the upper half, and
against males in the lower half, of the socioeconomic continuum.
Further, the psychologists Alice Eagly and Wendy Wood reanalyzed
Buss’s data and compared it with transnational data on economic inequal-
ity between the sexes gathered by the United Nations. They found that the
sex difference in ratings of good financial prospect by Buss’s subjects was
greater in societies with greater economic inequality between the sexes.
Where there was less economic inequality between the sexes, the sex dif-
ference in ratings of good financial prospect was smaller.
If, as I have argued, income is a better predictor of other dimensions of
status among males than among females under conditions of economic
inequality between the sexes, this is precisely the result we should expect.
For, under conditions of strict economic equality between the sexes,
income will be as good a predictor of other dimensions of status for one
sex as for the other, and the sex difference in emphasis on earning power
will disappear. The fact that Eagly and Wood found a correlation between
degree of economic inequality across societies and the strength of the sex
difference in emphasis on good financial prospect is thus fully consistent
with assortative mating by status. Therefore, since low-SES groups are
strongly underrepresented in Buss’s study, and since the sex difference in
emphasis placed on good financial prospect diminishes as economic inequal-
ity between the sexes diminishes, even Buss’s study fails to demonstrate a
female preference for high-SES mates per se.
So far we have considered studies that ask females to indicate the qual-
ities they prefer in a mate, and I have argued that none of these studies
demonstrates a preference for high-status males, since all their results are
confounded by assortative mating by status. There is, however, another
way in which evidence of female preference for high-status mates can be
gathered. Rather than asking females what their preferences are, we could
ask males how they are doing in their mating efforts. If we find that high-
status males enjoy greater mating success than lower-status males it would
presumably be evidence of a female preference for high-status males.
This approach was taken in a study by the anthropologist Daniel Pérusse.
Pérusse had students at two major French-speaking universities in the
Montreal area distribute questionnaires to their native French-speaking
friends and relatives, thereby apparently avoiding the problem that plagues
samples of college and university students. Respondents were asked to
report education level achieved, occupation, and income, which were used
Mating 245
as the measures of status. They were also asked to report the number of
coital partners during the previous twelve months and the number of coital
acts per partner. The information about coital partners and coital acts per
partner was used to construct a measure of mating success, which Pérusse
called number of potential conceptions (NPC). Mating success was thus rep-
resented by the formula
NPC = [1 - (1 - p P1 )] + [1 - (1 - p P2 )] + . . . + [1 - (1 - p Pn )]
where Pi is the number of coital acts with the ith partner, and p is an esti-
mated probability of conception per coital act, which for a variety of
reasons Pérusse set at 0.03.
Pérusse found a very weak correlation between status and mating success
(that is, the higher a male’s status, the higher his NPC) in the whole sample
and no correlation whatsoever between status and mating success for men
over forty. The latter was a particularly surprising result, since men older
than forty have typically achieved higher occupational and income levels
than younger men have; so, if higher-status males enjoy higher mating
success, men over forty should show a strong correlation between status
and mating success. Given that the correlation between status and mating
success for the entire sample was lower than Pérusse expected, he surmised
that the over-forty group was responsible for weakening the correlation in
the entire sample. Pérusse also speculated that the lack of correlation
between status and mating success in the over-forty group was due to the
fact that the majority of men over forty were married, hence presumably
monogamous, thereby lowering their overall mating success regardless of
status.
To test these speculations, Pérusse divided the over-forty sample into
married and unmarried groups. Since there were too few unmarried males
over forty on which to base a comparison of the two groups (there were
8), Pérusse expanded the two groups to include all married and unmarried
respondents between the ages of thirty and forty-nine. Pérusse then com-
pared the correlation between status and mating success for the unmarried
males in this age group with the correlation for married males. He found
a strong correlation between status and mating success for unmarried
males and no correlation for married males.
This seemed to confirm his suspicion that the lack of correlation between
status and mating success among married males was due to their presumed
monogamy. Married males, Pérusse suggested, “may not be in a position
to translate socioeconomic advantages into mating success as freely
as uninvolved men would.”33 Thus, he argued, the hypothesis that
246 Chapter 5
Beatty and Jack Nicholson, frequently select women two or three decades
younger. . . . Men who are high in occupational status are able to marry
women who are considerably more physically attractive than are men who
are low in occupational status.”34 Presumably, the attractive young women
are in very high demand, so the greater ability of high-status men to marry
them attests to the fact that women desire them more. Indeed, women
with the attractiveness that men desire appear to use it to bag the high-
status husbands that every woman wants.
There is, however, a problem with this argument. For Elder’s study
obtained attractiveness ratings for the female subjects, but not for their
husbands. Indeed, the only measure of the husbands’ desirability that Elder
used was status level. This raises the possibility that Elder’s findings were
confounded by male physical attractiveness. To test this idea, the sociolo-
gists Gillian Stevens, Dawn Owens, and Eric Schaefer analyzed the wedding
announcements of 129 couples that appeared in the major daily newspa-
per of a small city. Each wedding announcement contained information
about the education levels and occupations of the bride and groom, plus
a wedding photograph. The images of bride and groom in these photo-
graphs were separated, so that spouses didn’t appear together, and a mixed-
sex panel rated each bride and each groom for physical attractiveness. Since
all couples were dressed in formal wedding attire in the photographs, dif-
ferences in attractiveness judgments weren’t affected by differences in the
everyday attire of the brides and grooms. So there were no “Burger King
uniform” effects in the attractiveness judgments.
Stevens and her colleagues found a strong correlation between the attrac-
tiveness ratings of spouses. Very attractive females married very attractive
males, average-looking females married average-looking males, and unat-
tractive females married unattractive males. Indeed, there was a much
stronger correlation between the attractiveness of spouses than between
female attractiveness and male social status. They concluded that “physi-
cal attractiveness plays a large and an approximately equal role in marriage
choices for men and for women.”35 In short, mating is strongly assortative
by degree of attractiveness.
Further, a number of independent sociological studies have demon-
strated a strong positive correlation between attractiveness and both
income level and occupational achievement.36 Highly attractive people
have better jobs and make more money, on average, than do average-
looking people, who in turn have better jobs and make more money than
people who are judged to be homely. Perhaps contrary to what you’d
expect, this effect is even stronger among men than among women. Thus,
Mating 249
Table 5.3
Mating Success of Males by Class
Social class
Now, if the evidence for the claim that females prefer high-status males
is as weak as I’ve made out, why is the claim so widely accepted? I think
the reason is that we are captivated by a particular picture of the relation
between sex and status among our primate relatives, and this picture affects
our perception of human mating. It is widely accepted that among non-
human primates high-status males have greater mating success than males
lower in the status hierarchy. This belief is due partly to the popularity of
the engaging work of the primatologist Frans de Waal, who has been one
of the main purveyors of this idea. Once we’re convinced of the strength
of the correlation between status and mating success among our primate
relatives, the standards of evidence that are required to convince us of a
correlation in humans get lowered considerably. As de Waal says: “In
monkeys and apes there is a clear link between power and sex. High-
ranking males enjoy sexual privileges, and are more attractive to the oppo-
site sex. We need only look at recent events in the White House (and at a
television spectacular like ‘Who Wants to Marry a Multimillionaire?’) to
see how much the link exists in us too.”37
But there are two problems with this viewpoint. First, to determine
whether there is a correlation between male status and mating success in
humans, we need do much more than simply look at the Lewinsky scandal
and a single sensationalistic television show on the Fox network. We need
the same amount and quality of evidence concerning humans that has
been gathered about nonhuman primates before we can conclude that we
humans are like our primate relatives in this respect. Of course, Evolu-
tionary Psychologists have presented much more than just two items of
anecdotal data. But, as I’ve argued, their evidence fails to demonstrate that
females prefer high-status males. It appears more convincing than it is
because a conviction about a link between status and sex in primates leads
us to think that it must also be true that human females prefer high-status
mates.
Second, the primate literature is far from definitive in showing a corre-
lation between status and mating success among nonhuman primates. The
sociologist Lee Ellis did a literature survey of all studies published on
the correlation between status and sex among nonhuman primates. The
studies reviewed employed two primary measures of mating success:
number of copulations and number of copulations with a female in estrus.
Ellis found that two-thirds of all published studies reported a positive cor-
relation between status and one of these measures of mating success, while
one-quarter reported no correlation, and the remainder reported a nega-
Mating 251
indicating that “high status in men leads directly to increased sexual access
to a larger number of women,” and he implies that this is due to the greater
desirability of high-status men.39
But, among the Turkmen, women were sold by their families into mar-
riage. The reason that higher-status males enjoyed greater reproductive
success among the Turkmen is that they were able to buy wives earlier and
more often than lower-status males. Other studies that clearly demonstrate
a reproductive advantage for high-status males are also studies of societies
or circumstances in which males “traded” in women. This isn’t evidence
that high-status males enjoy greater reproductive success because women
find them more desirable. Indeed, it isn’t evidence of female preference at
all, just as the fact that many harem-holding despots produced remarkable
numbers of offspring is no evidence of their desirability to women. It is
only evidence that when men have power they will use it to promote their
reproductive success, among other things (and that women, under such
circumstances, will prefer entering a harem to suffering the dire conse-
quences of refusal).
Thus, although we have long been held captive by a picture in which
high-status primate males are the preferred mates of primate females, there
are reasons for thinking that the picture has been largely an illusion. If we
let go of that picture, we will also let go of the reflexive expectation that
the link between male high status and female preference exists in us, too.
And, if we let go of that reflexive expectation, we will not be so easily con-
vinced by impoverished evidence that human females prefer high-status
males. We will then be able to look at the evidence with eyes unclouded
by an antecedent conviction regarding what we’ll find. When we do, as I
have argued, we will see that there is no convincing evidence of a robust
female preference for high-status males. Just as male mate preferences will
turn out to be more complex than Evolutionary Psychologists have
claimed, female mate preferences will no doubt turn out to be more
strongly tied to physical attributes of males (physical attractiveness, bodily
symmetry, or chemical signaling of histocompatibility) than Evolutionary
Psychologists have claimed. Indeed, evidence of this association is already
beginning to accumulate.
I have argued that there isn’t good evidence for Evolutionary Psychology’s
two core claims regarding human mate preferences. Although my argu-
ments have pointed out a number of specific problems with those claims,
Mating 253
adults quickly gives us a sense of our own mate value, and we then tend
to pursue those with comparable mate value, having learned that pursu-
ing those of higher mate value is futile. In short, you come to realize that
you’re a six, and you confine your mating efforts to other sixes, occasion-
ally taking a stab at a seven in the hope that you’ll bag a mate with slightly
higher mate value and thereby enhance your reproductive prospects,
however slightly.
In this vision of human mating, tens are the target of evolved human
desires, but actual matings involve compromising one’s desires in the light
of realistic assessments of what one is likely to get on the mating market.
But here’s the catch. As even Evolutionary Psychologists admit, a pair of
well-matched fives or sixes (an average-looking woman with a working-
class husband) can have just as many offspring as a pair of nines or tens
(a supermodel with a rock-star husband). This can happen in different
ways. They could, of course, have just as many children. But they could
also have fewer children who nonetheless provide them with the same
number of grandchildren as are enjoyed by the pair of nines or tens. Sim-
ilarly, they could have fewer children and grandchildren, but nonetheless
have just as many great-grandchildren.
This entails that a six can have just as much reproductive success by
“compromising” and mating with another six as by mating with a ten.
One doesn’t necessarily compromise one’s reproductive success by “com-
promising” one’s desires for a mate of maximal mate value. There is
nothing about modern environments, however, that should make this fact
an evolutionary novelty, an unprecedented quirk of our modern age. Our
ancestors who were sixes should have had available the option of “com-
promising” their desires for tens and mating with fellow sixes, yet nonethe-
less producing just as many offspring as they could have produced by
mating with nines or tens. But, if this is the case, there could not have
been strong selection for a desire for tens in the first place. For selection
could not have distinguished between sixes who desired—whether requited
or not—to mate with tens and sixes who simply desired and mated with
other sixes. In other words, desires for members of the opposite sex over
a range of “mate values” would have had equal fitness. Thus, selection
would not have favored and driven to fixation a desire for mates of the
highest “mate value.”
Of course, there is undoubtedly a limit to how far down the scale of puta-
tive mate value one can trade before encountering a precipitous decline in
reproductive success. It may be that males at the very bottom of the status
hierarchy, who can provide absolutely nothing in the way of resources,
Mating 255
or older women are inherited—only if sons exhibit the same mating pat-
terns as their fathers. If sons of men who mated with older women them-
selves mate with younger women, and sons of men who mated with
younger women themselves mate with older women, then selection will
not drive any particular preference to fixation. Since equal reproductive
success can be achieved by matings with women in a range of ages in each
generation, as I have argued, selection will not have severely constrained
the available choices in each generation. Each generation will have enjoyed
a range of flexibility in choosing mates, with equal reproductive success
within that range.
Thus, since it should have been possible throughout human evolution-
ary history to achieve as much reproductive success by mating with a six
(a medium-status male or a slightly older, less than stunningly beautiful
woman), say, as with a ten, selection will not have designed human mate
preferences to be targeted quite so laser-like on the so-called tens. However,
we can expect selection to have designed mate preferences that discrimi-
nate against those at the bottom of Evolutionary Psychology’s scale of mate
value.
Of course, this argument is highly sensitive to which characteristics one
takes as defining mate value. The argument only works for cases in which
one can “trade down” the scale of putative mate value without thereby
negatively affecting one’s reproductive prospects. There are certainly some
characteristics of members of the opposite sex that one cannot trade down
very far while still achieving similar reproductive success. If mate value
were defined by pathogen resistance, for example, with higher mate value
reflecting a greater ability to fight off pathogens, we would certainly find
those who mated with sixes leaving fewer offspring than those who mated
with nines or tens.
The real question, then, is whether male status and female youth are
characteristics that females and males respectively can “trade down” while
still achieving comparable reproductive success. My skeptical argument
presupposes that, within limits, they are. Evolutionary Psychology’s view
of human mate preferences presupposes, in contrast, that male status and
female youth are characteristics that couldn’t have been traded down by
our ancestors without a corresponding decline in reproductive success.
Part of the reason to be skeptical of Evolutionary Psychology’s view of
human mate preferences is that Evolutionary Psychologists have no evi-
dence for this presupposition. In fact, Evolutionary Psychologists have
priced themselves out of the market for the evidence they’d need. Typi-
cally, the way to test the idea would be to gather evidence about the repro-
Mating 257
Leaving behind the issue of which characteristics people seek in their long-
term mates, let’s focus now on what long-term mate preferences are for.
According to Evolutionary Psychology, the function of long-term mate
preferences is to select a partner for marriage. Of course, this concept of
marriage doesn’t involve white flowing dresses, tuxedoes, Wagner’s
“Wedding March,” churches, or the county clerk’s office. Even in societies
with codified laws, there are legally recognized marriages between people
who have never taken vows before a judge or minister. In the United States,
these are known as common-law marriages, which are recognized in several
states and which, once recognized, must also be recognized as legal
marriages by all other states under the full faith and credit clause of the
Constitution. Further, in most cultures without systems of codified laws,
long-term mateships are ritually sanctioned by the community. If we are
not to have too provincial a conception of marriage, these mateships
should also count as marriages. But, given the diversity of cultural and legal
practices in which long-term mateships are embedded, what makes them
all marriages?
What Is Marriage?
to the offspring of another male, then the care-giving male pays the costs
of providing care without receiving any of its fitness benefits, since those
benefits go to the male whose offspring receive the care. So, Evolutionary
Psychologists argue, a male can enhance his fitness by providing parental
care only if the children he’s caring for are in fact his own. Thus, in order
for males to accrue fitness benefits from providing care for their offspring,
they first needed to solve the problem of paternity uncertainty, to maximize
the likelihood that the offspring for whom they provided care were in fact
their own.
According to Evolutionary Psychologists, males “agreed” to marriage
because it helped as much as possible to solve the problem of paternity
uncertainty. Through marriage, with its mutual sexual obligations, males
achieved round-the-month sexual access to a female, and by the implicit
marital contract this sexual access was typically exclusive. To put it in more
sinister terms, marriage allowed ancestral males to monopolize the repro-
ductive careers of their wives. They would have been content, perhaps, to
simply attempt to monopolize females during their fertile periods each
month, as chimpanzees attempt to do, but concealed ovulation denied
them that possibility. So they had to establish a long-term monopoly over
their mates’ sexuality, and marriage provided that long-term monopoly. By
entering marriages—that is, by forming implicit contracts with females for
sexual access and exclusivity—ancestral males were able to reduce their
paternity uncertainty and thereby reap the fitness benefits of sowing
parental care among only their own offspring.
Thus, whereas the traditional view sees ancestral males as having
“agreed” to marriage simply in order to accrue the fitness benefits of
biparental care of their offspring, Evolutionary Psychology sees ancestral
males as having “agreed” to marriage only on the condition that the invest-
ment of parental care that they cast upon children would return unto them
rather than benefiting some other male. On both accounts, the fitness
benefit to males is the same at the end of the day, but according to
Evolutionary Psychology the bargain struck between the sexes was more
complicated than the traditional view acknowledges.
The anthropologist Kristen Hawkes and her colleagues Alan Rogers and
Eric Charnov argue, however, that both accounts are mistaken about the
fitness payoff to males of paternal care. Hawkes and her colleagues con-
structed a number of game-theoretic models in which males have to
“decide” how much of their reproductive effort to allocate to parental care
of their offspring and how much to allocate to effort to secure as many
matings as possible with as many women as possible. In all of the models,
266 Chapter 6
males evolve to allocate very little effort to parental care and to allocate
the vast majority of their effort to mating, because the potential fitness
payoff of parental care is never high enough to offset the potentially very
high fitness payoff of increased mating opportunities. Even in those
models in which the payoff of paternal care is very large, the evolved allo-
cation to paternal care remains small, because the potential payoff for
promiscuous mating is always larger.
The most interesting finding by Hawkes and her colleagues is that, even
in a model of a pair-bonded population in which males are assured of
paternity, and hence assured that their paternal care isn’t misspent, males
still allocate very little effort to parental care and the vast majority of their
effort to promiscuous mating. To see why this would be so, suppose that
a pair-bonded population in which males are assured of paternity was
invaded by a mutant “Cad” male who eschewed all parental care and
devoted all of his time to widely playing the field of already-mated females.
This male would pay the absolute minimum for every offspring he pro-
duced, since those offspring would be cared for by the males he cuckolded,
while the average payoff of parental care to other males in the population
would decrease, since some of them would now be providing parental care
to the Cad’s offspring rather than their own. So, the overall fitness payoff
to the mutant Cad would be greater than the average fitness payoff of the
pair-bonded “Dads.” As a result, Cads would increase in the population.
In order to protect their investments, this would force Dads to allocate
much of their time to “guarding” their mates against the charms of the
Cads, thus reducing their allocation to parental care. This, however, would
reduce the overall biparental care received by offspring in the population,
which would in turn reduce the fitness payoff that Dads had been enjoy-
ing for being dads, causing Dads to lose further ground against the Cads.
In order to keep up, Dads would have to begin allocating some of their
effort to pursuing promiscuous matings, which would further reduce their
allocation to parental care. Thus, Hawkes and her colleagues found, even
in a population in which males are assured of paternity, the evolutionar-
ily stable state is for males to allocate very little reproductive effort to
parental care. Consequently, they argue, the benefit that accrues to a male’s
offspring because of paternal care, hence the fitness benefit that accrues to
the male for providing that care, is insufficient to account for the observed
level of parental care that human males provide. To put it crudely, during
our evolutionary history males didn’t “agree” to marriage because of the
fitness benefits they accrued from caring for their offspring.
Marriage 267
If Hawkes and her colleagues are right, however, what accounts for the
evolution of marriage? For, in order for marriage to evolve, both sexes had
to have “agreed” to the long-term biparental obligations that characterize
marriage. If Hawkes and her colleagues are right, it’s puzzling why males
would have ever “agreed” to the institution of marriage. But the psychol-
ogists Barbara Smuts and David Gubernick have a solution to this puzzle.
According to Smuts and Gubernick, long-term male parental care evolved
because of sexual selection by females for males who provided child care.
As in the traditional view, females stood to benefit by choosing care-giving
males because of the enhanced survivability and subsequent reproductive
success of their offspring. So it paid females to select for care-giving males.
But why did males go along with this deal and become care-giving? Because
females rewarded care-giving males with ongoing mating opportunities,
hence ongoing opportunities for paternity of offspring. Thus, males evolved to
provide long-term parental care not because of the direct fitness benefits
that accrued to their offspring because of that care, but because of the
increased opportunities for paternity that they earned from the mothers
of the offspring to whom they provided care. In short, according to Smuts
and Gubernick, long-term male parental care actually evolved as a form of
mating effort, not as a pure form of parenting effort. Accordingly, Smuts
and Gubernick call this hypothesis regarding the evolution of marriage the
mating effort hypothesis.
Smuts and Gubernick’s mating effort hypothesis doesn’t deny that males
accrue fitness benefits from providing parental care to their own offspring.
In fact, if a female accrues fitness benefits from the biparental care her off-
spring receive, then the father of those offspring thereby accrues the same
fitness benefits. Smuts and Gubernick merely argue, like Hawkes and her
colleagues, that those benefits are insufficient to explain why males would
have ever “agreed” to the institution of marriage. The deal-making benefit
for males, they claim, was the increase in paternity opportunities provided
by the mothers of the children for whom they provided care. Of course,
once a female preferentially mates with a male who cares for her offspring,
that increases that male’s odds of paternity of her next offspring, which
increases the odds that his continued parental care will eventually be
directed to his own offspring. So, even if a male begins the process by
caring for another male’s offspring, he can eventually reap the fitness
benefits of caring for his own offspring. But, Smuts and Gubernick argue,
this fitness benefit was actually a by-product of the evolution of male
parental care, not the ultimate cause of it.
268 Chapter 6
why humans have motives to marry in the first place. While Evolutionary
Psychology acknowledges the role of these proximate causes in human
marriage, its conception of the nature of marriage seeks the ultimate cause
of human marriage. And, insofar as we have motives that impel us to
commit to marriages, Evolutionary Psychology seeks the ultimate causes
of those motives as well. If the traditional view of the evolution of mar-
riage is right, the ultimate cause of marital motives in both sexes is
increased survivability and subsequent reproductive success of offspring. If
Evolutionary Psychology is right, these were operative ultimate causes, but
paternity certainty was also among the ultimate causes of marital motives
in men. If the mating effort hypothesis is right, survivability and subse-
quent reproductive success of offspring is the ultimate cause of marital
motives in women, but the ultimate cause of marital motives in men is
increased paternity opportunities. The hypotheses about the evolution of
marriage that we have just considered suggest the ultimate causes of why
people have the desires and emotions that proximately cause them to
become and remain married even when they don’t want children. But
none of the hypotheses implies that any of these desires or emotions are
themselves directed at increased survivability of offspring or increased
mating opportunities.
To make this point very clear, suppose we just had a single, simple desire
to marry, to form a long-term, committed, sexual relationship with
someone for whom we feel deep affection and friendship and who recip-
rocates those feelings. Suppose also that this simple desire evolved because
those with the desire tended to act on it and thereby enjoyed greater repro-
ductive success, on average, than those without the desire. In that case,
the desire to marry has the function or purpose of reproduction. This does
not mean, however, that we possess that desire as a means to achieving
the desired end of reproduction (in the way that I often desire to go to the
supermarket because it is a means to obtaining the desired end of having
ice cream). It also doesn’t mean that our desire to marry is merely the con-
scious manifestation of an unconscious desire to reproduce (in the way
that Freud thought that a girl’s desire to play boys’ baseball, for example,
is perhaps merely the conscious manifestation of the unconscious desire
to have a penis). Indeed, it doesn’t necessarily mean that we desire to
reproduce at all. Rather, the desire to marry is nothing more than the desire
to marry. Reproduction is the purpose of the desire to marry only in the
functional sense, according to which that desire enhanced the reproduc-
tive success of our ancestors. Reproduction is then the ultimate cause of the
desire to marry, but this doesn’t mean that the desire to reproduce is among
272 Chapter 6
typically done at the behest of one of the principals (as among the Kgatla
of South Africa) or it is the culmination of a successful “blind date”
arranged by the parents. As Helen Fisher says: “in the vast majority of cul-
tures, the views of both the boy and girl are sought before wedding plans
proceed. Modern Egyptians provide a good example. Parents of potential
spouses design a meeting between the youths; if the two like each other,
parents begin to plan the marriage.”7
There are, nonetheless, some societies in which parents or other relatives
arrange marriages without consulting the principals about their desires. But
these arranged marriages are actually the exceptions that prove the rule
that individuals have their own very strong, and arguably evolved, desires
regarding whom they marry. The most poignant example of this is the
practice of sim-pua marriage in Taiwan, in which a family gives a young
daughter, typically under the age of three, to the family of a young boy to
whom she will later be married. The two are then raised by the boy’s family
until they are old enough to form a conjugal union, at which point the
boy’s father announces one night over dinner that the two are henceforth
husband and wife. The anthropologist Arthur Wolf found that, upon
hearing this declaration, many of these young men and women simply
refuse to go through with the planned marriage and run away. When they
do proceed with the marriage plans, Wolf found that the large majority of
these couples never consummate their marriages, that sim-pua marriages
are characterized by far more extramarital affairs than other marriages in
the culture, that they produce far fewer children, that children who are
born in such marriages are nearly always known by the community to have
been fathered by a man other than the husband, and that they have a far
higher divorce rate than other marriages in the culture. And failure is not
specific to this extreme type of arranged marriage. As Fisher points out,
where arranged marriages “can be dissolved, as in New Guinea, on atolls
in the Pacific, in much of Africa and Amazonia, people regularly divorce
and remarry mates they choose themselves.”8 While certain cultural rep-
resentations of romantic love may be specifically Western and recent, the
reality is that people everywhere are strongly motivated by their own sense
of whom they want as a mate, and these desires are not easily manipu-
lated by those who would arrange their marriages for them. This is why,
in the vast majority of societies practicing arranged marriage, the princi-
pals are actively involved in the process of determining whom they are to
marry.
Now, one could continue to play the game of arguing that Evolutionary
Psychology’s conception of marriage and its hypotheses about the motives
274 Chapter 6
partner represents enhanced fitness benefits for the spouse who is dissolv-
ing the marriage. Further, Buss argues, infidelity is sometimes a means of
shopping for a new mate of higher “mate value” before actually paying
the cost of divorcing one’s partner. In such cases, infidelity is actually part
of a long-term mating strategy, since it is a means of seeking a “new and
improved” long-term mate while continuing to accrue benefits from the
current long-term mateship. Whatever the details of the case, however, dis-
solving a marriage in order to remarry is an analogue of breaching a con-
tract when the benefit to be gained from openly dissolving the contract
outweighs the cost of losing the services provided by the “jilted” party.
But infidelity isn’t always a long-term mating strategy, a means of shop-
ping for a new long-term mate. Many unfaithful spouses desire to preserve
their marriages, but also to surreptitiously pursue extramarital sex, or what
Evolutionary Psychologists artfully call extrapair copulations. For often a
married couple may have children to care for, or have excellent reproduc-
tive prospects together, so that the fitness benefits of the marriage are high
for both parties and a divorce would be costly for both. Nonetheless, there
may be certain circumstances under which an undetected short-term infi-
delity would have potential fitness payoffs for a married individual. In such
cases, infidelity is a short-term mating strategy pursued in conjunction
with the long-term strategy of remaining married. Thus, according to Evo-
lutionary Psychologists, when short-term infidelity occurs it is an attempt to
reap the fitness benefits of both extrapair copulations and a stable mar-
riage. But, if short-term infidelity is analogous to a contractual breach, we
should find that it is typically committed when the benefits of an extra-
pair copulation outweigh the cost of marital dissolution weighted by the
probability of detection (and the consequent dissolution of the marriage
at the hands of the cheated-on spouse). And this, Evolutionary Psycholo-
gists claim, is what we do in fact typically find. So short-term infidelity is
the marital analogue to the second type of contractual breach.
As I pointed out earlier, however, if marriage is an implicit reproductive
contract, we should find not only breaches of the contract in order to
accrue fitness gains, but also some method of detecting when breaches
occur or are likely to occur. Since infidelity is one of the most common
ways in which an implicit reproductive contract is breached, we should
find that people are relatively adept at detecting it. As Buss says: “It’s
unlikely that love, with the tremendous psychological investment it
entails, could have evolved without a defense that shielded it from the
constant threat from rivals and the possibility of betrayal from a partner.”9
The defense that evolution has equipped us with, according to Evolution-
Marriage 277
of pregnancies she can carry to term during her reproductive years, whereas
a man’s maximum reproductive output is limited only by the number of
women he can impregnate. A woman, then, can achieve her maximum
reproductive output with a single mate, whereas a man can achieve his
theoretical maximum only with multiple mates.
This means, argue Evolutionary Psychologists, that even a man in a sex-
ually active marriage with a fertile mate can always increase his number
of offspring by having extrapair copulations—indeed, at the theoretical
limit, he can increase his number of offspring by one for every extrapair
copulation with a new partner. “A married man with two children,” as Buss
says, “could increase his reproductive success by a full 50 percent by one
short-term copulation that resulted in conception and birth.”11 Thus, Evo-
lutionary Psychologists argue, we can expect that men have an evolved
psychological mechanism that inclines them to pursue short-term infi-
delity whenever the potential costs of the infidelity are low. In short, men
have an evolved tendency to cheat with as many women as possible as
often as they think they can get away with it.
Of course, this only tells us what everyone already thinks that they know
about men. Conventional wisdom has long held that men, married or not,
will pursue sex with virtually any willing woman. As Mark Twain remarked,
in a wry commentary on the seventh commandment, “by temperament,
which is the real law of God, many men are goats and can’t help com-
mitting adultery when they get a chance; whereas there are numbers of
men who, by temperament, can keep their purity and let an opportunity
go by if the woman lacks in attractiveness.”12 This witticism was subse-
quently matched by the following ditty, variously attributed to William
James and to Dorothy Parker:
Hogamous, higamous,
Men are polygamous.
Higamous, hogamous,
Women monogamous.
thinking about this question is that this gender gap parallels a gender gap
in the reported number of lifetime sexual partners. A consistent result of
sex surveys is that males report an average of two to four times as many
lifetime (opposite-sex) sex partners as women. The problem with these
results is that every new (opposite-sex) sex partner for a man is a new sex
partner for a woman. So, the average number of lifetime (opposite-sex) sex
partners has to be the same for both sexes. There have been a variety of
explanations of this gender gap in survey data, but the consensus among
sex researchers is that the gender gap is due to a combination of male over-
reporting and female underreporting of number of lifetime sex partners
(although there is disagreement about the mechanism that gives rise to
male overreporting and female underreporting). So, it is possible that male
overreporting and female underreporting similarly account for the gender
gap in the reported incidence of extramarital involvements.
One possible problem with this explanation, however, is that males may
actually underreport extramarital affairs despite overreporting lifetime
number of sex partners. Kinsey and his colleagues found that male sub-
jects were very reluctant to answer questions about extramarital sex: “There
is probably nothing in the histories of older married males who belong to
better educational and social levels that has more often been responsible
for their refusal to contribute to the present research. Many of the persons
who have contributed only after some months or years of refusal to do so,
prove to have nothing in their histories that would explain their original
hesitancy except their extra-marital intercourse. Even those who have con-
tributed more readily have probably covered up on this more often than
on any other single item.”15 So the incidence of extramarital affairs among
men is probably at least as high as that reported in surveys. Whether that
real incidence is as high as that reported by Kinsey or only about half that,
as reported in later studies, or whether the incidence varies with era or
culture, is something we will probably never know with any degree of
certainty.
Thus, if the large majority of the women with whom males have extra-
marital affairs are themselves married (as Fisher argues), and if the gender
gap in the reported incidence of extramarital involvements is not due to
male overreporting (as Kinsey’s work implies), then the bulk of the gender
gap has to be due to underreporting of extramarital liaisons on the part of
women. The actual incidence of extramarital involvements among women
is no doubt significantly higher than is revealed by self-reports in survey
data. Indeed, the actual incidence of extramarital affairs among women is
surely well more than half the incidence of extramarital affairs among
men.
Marriage 281
additional sex partners did not translate into additional children.”16 Given
this apparent fact, how could selection ever have favored the female
pursuit of short-term infidelity?
One possibility is that the assumption that females can’t increase their
total number of offspring by adding extramarital partners is mistaken. It
is possible, for example, that females can increase their odds of getting
pregnant during a menstrual cycle by having multiple mates. Conception
is a very tricky business in even the best of circumstances. A fertile female’s
ovum can fail to be fertilized even when she has frequent unprotected sex
with a fertile male throughout her menstrual cycle; and, if her egg does
get fertilized, it can often fail to implant properly in the uterus. It is pos-
sible that having two or more partners with different genetic profiles
during a cycle would increase the odds of successful conception. If so, that
would get a female pregnant sooner, which in turn would allow her to
shorten the times between pregnancies and thereby increase her total
number of pregnancies. In fact, studies of a number of animal species have
found that “multiple mating” does increase the odds of a successful preg-
nancy. A study of Gunnison’s prairie dogs, for example, found that females
who mated with three or more males while in estrus had a pregnancy rate
of 100 percent, whereas females who mated with one or two males had a
pregnancy rate of only 92 percent. Of course, humans aren’t prairie dogs;
we don’t even have a similar reproductive physiology or mating system.
And no studies have been done of humans to see whether females can
increase their odds of pregnancy through multiple mating. So, while it’s
an interesting possibility that females could increase their number of chil-
dren by taking extramarital partners, there is currently no evidence that
they can do so.
Another possibility is that, while females can’t increase the quantity of
their offspring through extramarital affairs, they can increase the quality
of their offspring. This is the possibility that has been central to Evolu-
tionary Psychology’s understanding of female short-term infidelity. David
Buss has most fully articulated Evolutionary Psychology’s theory about the
role that offspring quality plays in female short-term infidelity, so I will
focus on his account. Here is how the story goes.
The quality of offspring is a function of their fitness, and offspring fitness
is largely determined by parental fitness. For parents have genes that influ-
ence their ability to survive and reproduce, and parents pass their genes
on to their offspring, in whom those genes influence the ability of off-
spring to survive and reproduce. Individuals with higher-than-average
fitness have “good genes,” and by inheriting those “good genes” their off-
Marriage 283
But securing good genes for her child is only part of the way in which
a married woman can benefit from an extramarital affair. For, as a married
woman, she also has a long-term mate who will invest in her offspring.
So, as long as she can keep her extramarital affair concealed, a woman who
has an extramarital affair with a male with good genes gets the fitness ben-
efits of both worlds: She obtains superior genes for a child who can then
be reared on the secured paternal investment provided by her cuckolded
long-term mate with inferior genes. Thus, Buss says: “Some women pursue
a ‘mixed’ mating strategy—ensuring devotion and investment from one
man while acquiring good genes from another.”18
Of course, this “mixed” mating strategy was available to ancestral
women as well. Presumably some ancestral women pursued this strategy
while others didn’t. And, since ancestral women who had a propensity to
pursue short-term infidelities with males with good genes would have pro-
duced higher-quality offspring, on average, than ancestral women who
lacked that propensity and remained faithful to their long-term mates, over
the course of human evolution selection “forged a female psychology of
infidelity.”19 That is, not only did selection design the female mind to
pursue the attraction and retention of a long-term mate, but selection
designed the female mind to pursue extrapair sex under specific condi-
tions—in particular, when paternal investment has already been secured
from a long-term mate, when the extrapair sex is likely to go undetected,
and when the extrapair partner has better genes than the long-term mate.
The mind of the married woman, then, is designed to ascertain when these
conditions are satisfied and to feel sexual desire for an available high-
quality male. As Buss says: “Women’s sexual psychology, including their
desire to stray, exists today solely because that’s what benefited ancestral
women.”20 In other words, the female propensity toward short-term infi-
delity is an adaptation.
Buss claims that several lines of evidence converge to support the claim
that female short-term infidelity is an adaptation. First, there is the previ-
ously mentioned fact that, by standard estimates, the rate of paternity dis-
crepancy in humans is at least 10 percent. So a significant portion of
children born within marriages are sired by women’s extrapair partners. Of
course, if female short-term infidelity is an adaptation for producing
higher-quality offspring, it is not enough that contemporary women
engage in significant extrapair coupling; adaptation requires that ancestral
women pursued extrapair copulations as well. And there is indirect evi-
dence that they did. Comparative studies of several nonhuman primate
species have shown a correlation between male testes size in a species and
Marriage 285
In this experiment, Thornhill and Gangestad made use of the fact, which
has been known for some time, that both sexes respond to subliminal
scents (or pheromones) emitted by members of the opposite sex. Thornhill
and Gangestad issued their male participants T-shirts that had been washed
in unscented laundry detergent. Male participants were instructed to
refrain from using scented soaps or colognes for a two-day period, during
which they were also to refrain from eating spicy foods, drinking alcohol,
smoking, and having sex. They were required to wash their bedsheets in
unscented laundry detergent at the beginning of the two-day period and
to sleep in the T-shirt each of the two nights of the experimental period.
Each male subject then returned his T-shirt in a sealed plastic bag, to be
sniffed by the female participants (who were not taking hormone-based
contraceptives that prevented natural ovulation). Females smelled all T-
shirts in a random order both during ovulation and during the infertile
phase of their menstrual cycles, and they rated each T-shirt on a scale
of 1 to 10 for “pleasantness” and “sexiness” of smell. Thornhill and
Gangestad found that scents of the T-shirts of highly symmetrical men
were rated highest—but only by women who were ovulating. During the
infertile phases of their menstrual cycles, women did not prefer the scent
of symmetrical males’ T-shirts over that of relatively asymmetrical males’
T-shirts.
Thus, it appears that women do prefer males with good genes, but only
during that phase of their menstrual cycles when they stand to benefit
from sex with males with good genes. This, of course, seems to make sense
from an evolutionary cost-benefit standpoint. For women can reap the
benefits of extrapair sex with men with good genes only when they are
fertile, but they can pay the costs of extrapair sex (loss of the cuckolded
long-term mate) at any time throughout their menstrual cycles. So,
Gangestad and Thornhill say, “selection may have shaped female interest
in men who possess indicator traits of good genes such that it changes
across the cycle: increases when women are fertile . . . and decreases when
not.”21 Putting it more bluntly, Buss says: “Women detect the scent of sym-
metry, prefer that scent when ovulating, and choose more symmetrical
men as affair partners.”22
Another source of evidence that Buss cites to support the claim that
female short-term infidelity is designed to tip the odds of paternity in favor
of the extrapair partner derives from the work of the spermatologists Robin
Baker and Mark Bellis. In a nationwide survey of 3,679 British women who
were not taking hormone-based contraceptives, Baker and Bellis asked
women to report whether they were having extrapair copulations, to
288 Chapter 6
provide information about the frequency and timing of their in-pair and
extrapair copulations, and to provide details about their last copulation.
The women who reported extrapair involvements also reported that
approximately 60 percent of their copulations during the fertile phase of
their menstrual cycles took place with their extrapair partners, whereas
about 60 percent of their copulations during the infertile phase of their
cycles took place with their in-pair partners. So, when women have affairs,
their sexual activity with their affair partners is higher during the fertile
phase of their menstrual cycles than during the infertile phase, and more
of their sexual activity during the fertile phase occurs with their affair part-
ners than with their long-term mates. Further, of the women who reported
extrapair involvements, those whose last copulation was with their extra-
pair partner reported a higher incidence of “copulatory orgasms” than did
those whose last copulation was with their in-pair partner (where Baker
and Bellis define a female “copulatory orgasm” as one that occurs between
one minute before and forty-five minutes after male ejaculation). This is
significant, according to Baker and Bellis, because a copulatory orgasm
increases sperm retention, since the contractions of orgasm cause the
cervix to dip down into the pool of semen ejaculated at the back of the
vagina and “suck” sperm up into the uterus where they have greater access
to the oviducts. Thus, Buss concludes: “Women have more ‘high sperm
retention’ orgasms with their affair partner than with their regular partner.
. . . Furthermore, women seem to time their orgasms with their affair part-
ners to coincide more closely with when they ovulate.”23
Finally, in a study done with their colleague Randall Comer, Thornhill
and Gangestad performed their standard measurements of seven bilateral
traits (to compile an index of overall degree of asymmetry) on 86 hetero-
sexual couples and had each member of each couple complete a ques-
tionnaire about the female’s orgasms. They found a moderate correlation
between the male and female reports of female orgasms, except in a
handful of cases in which the women reported that they always faked
orgasm and the men took them to be the real thing. Combining the male
and female reports for the correlated cases, and using only the female
reports for the other cases, they constructed an orgasm profile for the
female of each couple. And they found that women whose partners had a
high degree of bodily symmetry had more copulatory orgasms than
women whose partners had a lower degree of bodily symmetry. Further,
they found that this wasn’t because women whose partners were sym-
metrical had more orgasms overall, since the symmetry of a woman’s
partner didn’t correlate with the total number of orgasms she experienced
Marriage 289
(including those from oral sex or manual manipulation). It was simply that
women whose partners were highly symmetrical experienced more copu-
latory orgasms, more of the allegedly “high-sperm-retention” orgasms,
than other women.
This is a lot of merely circumstantial evidence, but it is a lot of merely
circumstantial evidence, and Buss thinks that the combined evidence pro-
vides rather definitive support for the following picture. Women have a
long-term mating psychology, which leads them to seek a long-term mate
who will provide parental investment (which, according to Buss, is desired
in the form of a high level of resources). But, once they have secured a
long-term investing mate, women become motivated by “an evolved EPC
[extrapair-copulation] psychology that is distinct from their long-term
mating psychology.”24 This evolved extrapair-copulation psychology con-
sists in “a psychological mechanism in women specifically designed to
promote short-term mating.”25 So, if a woman’s husband is a little weak in
the genes, and if an opportunity for a discreet short-term infidelity with a
male with good genes (a symmetrical male who smells sexy) presents itself,
this psychological mechanism will cause strong sexual desire for that male
and the woman will likely act on the desire. If an affair is begun with a
male with good genes, this evolved mechanism will cause intensified desire
for the extrapair partner during the fertile phase of the menstrual cycle,
since the extrapair partner will then be especially sexy smelling, and this
will lead the unfaithful woman to pursue more copulations with her
extrapair partner than with her husband while she is fertile. Since the
evolved mechanism will have caused her to choose a symmetrical male as
an extrapair partner, she will also experience more copulatory orgasms
with her affair partner than with her husband; and, since the majority of
these copulatory orgasms with the affair partner will occur while she is
fertile, she will be more likely to become pregnant by her extrapair partner
than by her husband. If she does become pregnant by her extrapair partner,
she will have a child with better genes than those her husband could
have provided, and she will be able to raise that child on the already
secured resources provided by her cuckolded husband. This, in turn, will
enhance her reproductive success. And, as it is for modern women, so it
was for ancestral women. Thus, the desire for and pursuit of short-term
infidelity with males with good genes is a human female psychological
adaptation.
Now, I think that the claim that women have a psychological adapta-
tion for short-term infidelity goes well beyond the evidence. In fact, I will
argue that the pattern of female short-term infidelity described above is
290 Chapter 6
tional dissatisfaction, and they asked those of their subjects who cited
extramarital involvements to indicate which form of marital dissatisfac-
tion they felt led to the extramarital involvement. Glass and Wright found
a strong correlation between the type of marital dissatisfaction experienced
by women and the type of extramarital relationship they became involved
in. Women who developed extramarital emotional involvements reported
emotional dissatisfaction in their marriages. But women who were emo-
tionally dissatisfied in their marriages were not significantly more likely to
develop extramarital sexual involvements than were women who reported
being happy in their marriages. Similarly, women who reported extramar-
ital sexual involvements reported sexual dissatisfaction in their marriages.
And women who were sexually dissatisfied in their marriages were far more
likely to develop extramarital sexual involvements than women who
reported happy marriages. Thus, Glass and Wright found, it is not simply
marital dissatisfaction that prompts the vast majority of women’s extra-
marital sexual affairs, but specifically sexual dissatisfaction in marriage.
Glass and Wright’s findings replicate the findings of a number of previ-
ous studies. In a 1938 study of 1,250 California couples, the psychologist
Lewis Terman found that 27.2 percent of women who rated their marriages
as unhappy reported that they “sometimes” or “very frequently” desired
extramarital sex, whereas only 5.2 percent of the women who rated their
marriages as happy reported similar levels of desire for extramarital sex.
Further, he found that 19.4 percent of women who “never” or “sometimes”
experienced orgasm with their husbands reported that they “sometimes”
or “very frequently” desired extramarital sex, while only 7.8 percent of
women who “usually” or “always” experienced orgasm with their hus-
bands reported the same level of desire for extramarital sex. In a 1956
survey of 6,251 English women, the sociologist Eustace Chesser found that
25 percent of women who “always or frequently” had orgasm during sex
with their husbands “occasionally” desired extramarital sex, and only 3
percent of them “frequently” desired extramarital sex. In contrast, of
women who “rarely or never” had orgasm with their husbands, 38 percent
reported “occasionally” desiring extramarital sex, and 10 percent reported
“frequently” desiring it. In a 1974 study of 2,372 married American
women, the sociologists Robert Bell and Dorthyann Peltz found that, of
women who reported having had extramarital intercourse, 48 percent said
that sex with their husbands was “poor” or “very poor,” whereas 18 percent
said that sex with their husbands was “very good.” Further, they found
that, of women who reported having had extramarital sex, 40 percent
reported “never” having orgasm during sex with their husbands, while 24
292 Chapter 6
percent reported having orgasm with their husbands “at least sometimes”
(which, of course, is a widely inclusive category). Finally, in a 1975 survey
of more than 100,000 female readers of Redbook magazine, the psycholo-
gists Carol Tavris and Susan Sadd found that only 22 percent of women
who rated sex with their husbands as “very good” reported having had
extramarital sex, but that 48 percent of women who rated their marital sex
as “poor or very poor” reported having had an extramarital affair. While
each of these samples is limited in one way or another, the findings are
remarkably consistent: Sexually dissatisfied married women are far more
likely than sexually satisfied married women to both desire and have extra-
marital sex.
Two studies by Buss and his colleague Heidi Greiling corroborate all of
these findings. In one study, 101 women were given a list of forty-seven
different life circumstances and were asked to rate how likely it would be
that they would develop an extrapair sexual involvement in each of those
circumstances. The results showed that the leading factor that would moti-
vate female extrapair sexual involvements was retaliatory. Women said that
discovering a partner’s infidelity would make it somewhat likely that they
would have extrapair sex themselves. But, in very close second and third
places were motivations related to sexual dissatisfaction. Women rated
“current partner is unwilling to engage in sexual relations” and “sexual
relations with current partner have been unsatisfying for a long time”
almost as high as discovery of a partner’s infidelity. In the other study,
Greiling and Buss asked 90 women to imagine a woman who was in a com-
mitted long-term relationship but who chose to have a short-term sexual
relationship with another man. The participants were then given a list of
twenty-eight possible benefits that the woman might derive from her
short-term infidelity. The top-rated benefit, which was rated significantly
higher than the second-rated benefit, was “likelihood of receiving sexual
gratification.” Given that sexual dissatisfaction in the current relationship
ranked second and third among the leading motivations for extrapair
sexual involvement in the previous study, it is reasonable to assume that
the participants in the second study imagined the hypothetical unfaithful
woman to be motivated by sexual dissatisfaction. If so, the results of the
second study indicate that women perceive that a woman who is sexually
dissatisfied in her marriage is likely to be able to find sexual gratification
in an extramarital sexual involvement and that sexual gratification is a
worthwhile benefit of an extramarital sexual involvement.
Thus, there is exceptionally strong evidence that, for women, sexual dis-
satisfaction in marriage is a leading factor in the occurrence of short-term
Marriage 293
recently did a group of researchers turn from asking about when women
are “receptive” to sex to asking questions about variations throughout the
menstrual cycle in female desire for sex, female fantasy about sex, female
masturbation, and female initiation of sex. This shifted away from a focus
on sexual acts involving females, which could often be initiated by a male
partner at a time when the female is not particularly interested, to a focus
on female-determined sexual activity. And this shift in focus engendered
some very interesting research, which consistently found a peak in the
levels of female desire for sex, fantasy about sex, masturbation, and initi-
ation of sex during the fertile phase of the menstrual cycle.26 The rise in
female-initiated sexual activity during the fertile phase is probably an adap-
tation, since it is too well designed for reproduction to be an accident or
by-product of something else. So, the increase in sexual activity with extra-
pair partners during the fertile phase of the menstrual cycle is really just a
by-product of a generalized increase in female-initiated sexual activity
during that period.
Evolutionary Psychologists are aware of this research, and they grant the
reality of an increase in female sexual desire and female-initiated sex
during the fertile phase of the menstrual cycle. They even grant that this
is an adaptation, but they argue that it can’t explain the facts about female
short-term infidelity. For Baker and Bellis didn’t find an increase in overall
level of sexual activity during the fertile phase or an increase in sexual
activity with in-pair partners. They found a more targeted increase in
sexual activity. It was only sexual activity with extrapair partners that
increased during the fertile phase. “Hence,” Gangestad argues, “it appears
that women’s heightened sexual desire midcycle is not an increase in
sexual desire in general. Rather, the data suggest that it motivates sexual
behavior toward specific kinds of partners, in particular (pending further
research) those who may have promised genetic benefits ancestrally.”27 So,
if the by-product explanation is correct, why is there a selective increase in
sexual activity with extrapair partners during the fertile phase?
The reason for this is relatively mundane. First, as we’ve seen, the vast
majority of women who have extramarital sexual affairs are sexually
dissatisfied in their marriages. Second, if a sexually dissatisfied married
woman is seeking sexual gratification through an extrapair involvement,
she will be unlikely to repeat extrapair sexual encounters with men who
fail to provide her with the desired sexual gratification. As we’ve seen, there
are potential costs to short-term infidelity, and if the desired benefit of
sexual gratification isn’t to be had with a particular male, a woman is not
going to continue a sexual involvement with that male. So, any male who
296 Chapter 6
women are unable to land a husband who is both symmetrical and invest-
ing. This entails that symmetry is one of the things that women seek in a
long-term mate because of the “good genes” signaled by symmetry. So, the
fact that women have affairs with symmetrical men doesn’t support Buss’s
claim that their choice of symmetrical men is produced by “an evolved
EPC [extrapair copulation] psychology that is distinct from their long-term
mating psychology.”28 For suppose, to simplify, that women seek long-term
mates who have only two qualities—symmetry and a propensity for high
parental investment. When a woman seeks a partner for a short-term
infidelity, the same preference structure can be operative in her choice of
partner, but parental investment becomes irrelevant. (This would not be
the case if the infidelity is actually part of a long-term-mate replacement
strategy, but we are considering only female short-term infidelity at the
moment.) When parental investment drops out of the equation, the pref-
erence for symmetry is the only preference left. Thus, even if women have
an evolved preference for symmetrical men, this preference can be part of
a single set of mate preferences, which is operative in choosing both long-
term and short-term mates. Symmetry becomes more significant in con-
texts of short-term infidelity only because women don’t need to strike the
best trade-off among their preferences in the way they need to when select-
ing a long-term mate.
In sum, then, here is the by-product explanation of the discovered
behavioral pattern of female short-term infidelity. Suppose, as the evidence
seems to indicate, that the minds of many women contain the following
three adaptations, among others: the “sex drive” (the desire for a regular
and fulfilling sex life, together with the patterns of planning and acting so
as to ensure the satisfaction of that desire), a peak in sexual desire during
the fertile phase of the menstrual cycle (with its attendant increase in
female-initiated sexual activity), and a preference for symmetrical males
(which may be just one of several mate preferences). These three adapta-
tions can causally interact in the following way. When a woman is sexu-
ally dissatisfied in her marriage, there is increased probability that she will
begin an extramarital sexual involvement if a desirable opportunity for
such an affair presents itself. This is due to the standard operating proce-
dures of the “sex drive.” Since the potential costs of infidelity are high,
however, she will be selective in choosing an extrapair partner. In select-
ing an extrapair partner, the same preferences that were operative in choos-
ing a long-term mate will also be operative, but some of them (for example,
a propensity for parental care) will be irrelevant to the specifically sexual
298 Chapter 6
role for which the extrapair partner is being selected. As irrelevant prefer-
ences drop out of the selection process, the preference for symmetry will
come to loom large in the selection of an extrapair partner. As a result, if
a woman begins an extramarital affair, there is increased probability that
she will select a high-symmetry male as her extrapair partner. But this is
due to the standard operating procedures of her unitary set of mate pref-
erences, rather than to a set of preferences specifically tailored to short-
term infidelities. Then, once a woman begins an extramarital affair, her
peaking desire for sex during the fertile phase of her menstrual cycle will
cause an increase in the number of sexual encounters that she initiates.
Since, by this hypothesis, she began the affair because sex with her
husband is less gratifying than sex with her affair partner, the sexual
encounters that she initiates will be directed almost exclusively toward her
affair partner. As a result, the copulations with her highly symmetrical
affair partner will tend to be concentrated during the fertile phase of her
menstrual cycle. But this will be due to the standard operating procedures
of her fertile-phase peak in sexual desire. Thus, the three adaptations can
conspire, under circumstances of sexual dissatisfaction in marriage, to
produce a pattern of behavior that appears to be the direct result of an
adaptation specifically for short-term infidelity.
Two of the adaptations postulated in this alternative scenario (the fertile-
phase peak in sexual desire and the preference for symmetrical men) are
explicitly accepted by Evolutionary Psychologists, and the third (the sex
drive) is a commonplace and nowhere explicitly rejected by Evolutionary
Psychologists. And the data can be explained by appeal to these three
adaptations without the need for Evolutionary Psychology’s postulation of
an adaptation specifically for female short-term infidelity. In short, the
by-product explanation is simpler, since it postulates fewer adaptations
overall, and it accounts for the same facts; therefore it is to be preferred to
Evolutionary Psychology’s explanation on grounds of parsimony.
The by-product explanation does, however, appear to leave a thread
loose. For one strand of the web of evidence woven around Evolutionary
Psychology’s adaptationist hypothesis is Thornhill and Gangestad’s results
that appear to show that women have more “high-sperm-retention” cop-
ulatory orgasms with highly symmetrical men than with less symmetrical
men. Those results seemed to lend further credence to the idea that women
are designed to reap genetic benefits for their offspring through the pursuit
of extrapair copulations. If extrapair copulations with highly symmetrical
men are simply the by-product of interacting adaptations, rather than
the result of an adaptation specifically for extracting good genes, what
Marriage 299
accounts for the higher incidence of copulatory orgasms with highly sym-
metrical men?
The by-product explanation does, in fact, leave this thread loose, but it’s
not at all clear that it’s a thread that needs to be woven into any adequate
explanation of female short-term infidelity. For there are good reasons for
skepticism regarding the claims made for the role of female orgasm in Evo-
lutionary Psychology’s account of female short-term infidelity.
First, although Baker and Bellis claimed to find that female copulatory
orgasms resulted in higher sperm retention relative to copulation without
female orgasm, all of their evidence indicated that this allegedly higher
level of sperm retention did not in fact increase the probability of concep-
tion. So, even if women have more copulatory orgasms with highly sym-
metrical men, there is no evidence whatsoever that that translates into
higher levels of paternity for highly symmetrical men. In order to preserve
the claim that female copulatory orgasm is facilitating paternity by sym-
metrical males, Thornhill and Gangestad are forced to argue that female
copulatory orgasm merely functions to draw more of the extrapair partner’s
sperm into the uterus than the in-pair partner’s sperm. Thus, while not
directly increasing the odds of conception, female copulatory orgasm is
supposed to give the extrapair partner more “lottery tickets,” in effect, than
are given to the in-pair partner. She thereby increases the odds that, if she
becomes pregnant, he will be the father. But, if sucking up sperm into the
uterus through copulatory orgasm doesn’t increase the odds of conception
over not sucking up sperm into the uterus, it’s hard to see how sucking up
the extrapair partner’s sperm, while not sucking up the in-pair partner’s
sperm, is supposed to increase the odds that any resulting pregnancy will
be due to the extrapair partner’s sperm. In addition, if the extrapair partner
does enjoy an advantage in the competition to impregnate a woman
having an extramarital affair, this is no doubt due to the fact that she has
more sex with him during her fertile phase than she does with her
husband, as Baker and Bellis’s data show. Thus, there’s no evidence that a
female’s copulatory orgasms play any role in determining the paternity of
her offspring.
Second, there are problems in the way that “copulatory orgasm” is
defined in the study by Thornhill and Gangestad. Baker and Bellis claimed
to find that any female orgasm that occurred between one minute before
and forty-five minutes after ejaculation into the vagina sucked sperm up
into the uterus and thereby resulted in higher sperm retention. Following
this finding, Thornhill and Gangestad asked their subjects to report the
incidence of any female orgasms that occurred in one of three time
300 Chapter 6
segments during copulation: before the male’s ejaculation, at the same time
as ejaculation, and within forty-five minutes after ejaculation. Thornhill
and Gangestad then considered only the simultaneous and postejaculation
orgasms to be high-sperm-retention “copulatory orgasms.” And, as noted,
they claimed to find that their female subjects reported having more of
these orgasms with highly symmetrical men than with less symmetrical
men.
But the biologist Nicholas Pound and the Evolutionary Psychologist
Martin Daly point out that there are two significant problems in this pro-
cedure. First, the study only compared the orgasm profiles of women whose
partners were highly symmetrical with those of women whose partners
were less symmetrical. It didn’t compare a woman’s orgasm profile with a
highly symmetrical partner with that same woman’s orgasm profile with a
less symmetrical partner. So, their study failed to show that individual
women varied their number of copulatory orgasms in accordance with the
degree of symmetry of their sex partner; and that is what would need to
be shown to support Evolutionary Psychology’s adaptationist explanation
of female short-term infidelity. Second, by asking only about postejacula-
tion orgasms that occurred during copulation, Thornhill and Gangestad
ignored a potentially significant number of high-sperm-retention orgasms.
For, if any orgasm within forty-five minutes after ejaculation is a high-
sperm-retention orgasm, as Baker and Bellis claim, women could have had
many high-sperm-retention orgasms that were not “copulatory orgasms”
as Thornhill and Gangestad defined them. Even women with extremely
asymmetrical partners could have had manually or orally induced orgasms
within forty-five minutes after ejaculation, and these would have been
high-sperm-retention orgasms as well. So, Thornhill and Gangestad in fact
failed to show that the partners of highly symmetrical men had more
high-sperm-retention orgasms, as opposed to more high-sperm-retention
orgasms during copulation, than the partners of less symmetrical men.
There is, then, no good evidence of a connection between male sym-
metry and female high-sperm-retention orgasm, and there is no good evi-
dence of a connection between so-called high-sperm-retention orgasms
and conception. Thus, no account of female short-term infidelity has to
explain why women have more copulatory orgasms with highly symmet-
rical men than with less symmetrical men. The by-product explanation
accounts for all the evidence that truly need be accounted for. And by pos-
tulating a female adaptation specifically for short-term infidelity, Evolu-
tionary Psychology goes well beyond the available evidence.
Marriage 301
rivals for their mates’ affections (criticizing the rivals’ appearance or intel-
ligence), to enhance one’s own personal appearance, to threaten or become
violent toward the rivals, to increase the frequency of public displays of
possession (holding hands with or putting an arm around their mates), to
shower their mates with affection or gifts (saying “I love you” more fre-
quently), to turn up the sexual heat in the relationship (initiating sex more
frequently and making love more passionately), and to threaten or enact
violence toward their mates. Most of these actions are designed to counter
perceived threats to a sexual relationship by minimizing a mate’s oppor-
tunities for infidelity, diminishing a mate’s interest in the rival, or rein-
vigorating a mate’s interest in oneself. Further, Buss found that each sex
rated most of these tactics, when employed by the opposite sex, to be at
least moderately effective in preventing a mate from straying.
Jealousy, then, appears to be triggered by perceived threats to a rela-
tionship, and it appears to motivate a number of actions that are effective
in retaining a mate who may be straying or contemplating straying.
Since jealousy is so closely tied to, and effective in promoting, our repro-
ductive interests in these ways, Evolutionary Psychologists argue that jeal-
ousy shows significant signs of being an emotional adaptation designed to
detect and thwart threats to our reproductive interests.
Evolutionary Psychologists further argue that, since men and women
faced different threats to reproductive interests throughout human evolu-
tionary history, the sexes have evolved distinct jealousy mechanisms. As
Buss and his colleagues say: “The evolutionary hypothesis about sex dif-
ferences in jealousy is domain-specific—it proposes that the psychological
mechanisms of each sex will contain dedicated design features, each correspon-
ding to the specific sex-linked adaptive problems that have recurred over thou-
sands of generations of human evolutionary history. From an evolutionary
perspective, the odds that the sexes will be psychologically identical in
domains where they have recurrently confronted different adaptive prob-
lems over the long expanse of human evolutionary history are essentially
zero.”30
In particular, the psychological mechanisms of the sexes will differ with
respect to “design features” that are dedicated to looking out for and
responding to different kinds of cue about when reproductive interests are
in jeopardy. As Buss says: “From an ancestral man’s perspective, the single
most damaging form of infidelity his partner could commit, in the cur-
rency of reproduction, would have been a sexual infidelity. A woman’s
sexual infidelity jeopardizes a man’s confidence that he is the genetic father
of her children. A cuckolded man risks investing years, or even decades,
304 Chapter 6
sis” of jealousy and with its claim that there are sex differences in the
“design features” of the psychological mechanisms of jealousy. Let’s begin
examining these problems by considering Evolutionary Psychology’s “evo-
lutionary analysis” of jealousy.
Evolutionary Psychology’s theory of jealousy, with its hypothesis of
evolved sex differences in the triggers of jealousy, is frequently called “the
evolutionary theory of jealousy.” In one sense, this is right: It is the only
evolutionary hypothesis regarding jealousy that has been articulated and
tested in the literature. In another sense, however, it is deeply misleading:
There are other genuinely evolutionary theories that one could hold about
jealousy, none of which has even been considered by Evolutionary Psy-
chologists, let alone tested. For example, Kristen Hawkes has suggested to
me that the ultimate cause of male sexual jealousy may be mere paternity
competition, rather than paternity uncertainty. This suggestion doesn’t
deny an evolved sex difference in the triggers of jealousy, but it offers a
different evolutionary analysis of that difference, a different account of
what drove the evolution of the sex difference. Let’s explore this alterna-
tive for a moment.
According to Evolutionary Psychology, marriage is an implicit repro-
ductive contract according to which parental care (given by the male) is
exchanged for assurance of paternity (given by the female). The assurance
of paternity is crucial, according to Evolutionary Psychology, since that is
what ensures a male that his parental care won’t be misspent on the off-
spring of another male. Jealousy, then, has evolved as an emotional alarm
that signals that one’s mate is, or could be contemplating, violating the
implicit reproductive contract. Since a female violates the contract by nul-
lifying the assurance that her mate will have paternity of her offspring,
Evolutionary Psychology predicts that the proximate mechanism underly-
ing male jealousy will monitor contexts in which parental care is at issue
for signs of potential sexual infidelity, since sexual infidelity entails the
potential for misspent parental care. In contrast, since a male violates the
implicit reproductive contract by withdrawing his parental care, female
jealousy is supposed to function to detect the potential for withdrawal of
parental care, which is supposedly signaled by the development of other
emotional involvements.
As we saw in our discussion of marriage, however, there are both theo-
retical and empirical reasons for believing that the mating effort hypoth-
esis—according to which marriage evolved as a contractual exchange of
parental care (given by the male) for paternity opportunities (given by the
female)—provides a better explanation of the evolution of marriage than
308 Chapter 6
ity of interactions. The payoff to As of the “trust no one” policy will consist
in avoiding the costs of interactions with Bs, and the payoff to Bs will
consist in avoiding the costs of interactions with As. In short, the same
“trust no one” policy will have different functions for As and Bs. (If you’re
tempted to think that the policy has the same function for each—namely,
that of avoiding being injured—just vary the thought experiment, to
suit your imagination, so that the kind of cost that As impose on Bs is dif-
ferent from the kind of cost that Bs impose on As in their interactions,
although the costs are equal. Then the “trust no one” policy functions in
As to avoid one kind of cost, while it functions in Bs to avoid a different
kind of cost.)
Second, suppose that As and Bs do have some way of telling one another
apart prior to engaging in interactions. If As and Bs develop a way to dis-
tinguish one another, then they can selectively interact only with their
own kind, thereby accruing the benefits of interacting with their own kind
while avoiding the costs of interacting with the other kind. To do so,
however, each has to develop some kind of detection mechanism, which
functions to distinguish As from Bs. Adding this detection mechanism to
their anatomy or psychology will, however, entail some form of develop-
mental costs. Now here’s the rub. If the benefits from interacting with
one’s own kind are sufficiently small, they won’t cover the costs of devel-
oping the detection mechanism required to distinguish the individuals
from whom one will benefit from the rest. In such an event, it will be
cheaper overall for the individual not to develop a detection mechanism
and to simply stick with a “trust no one” policy. For As, this policy is not
as precise in its effects as a “don’t trust Bs” policy would be, since it fore-
closes on the benefits of interactions with other As. But the added preci-
sion that a “don’t trust Bs” policy would bring doesn’t cover the costs of
implementing that policy. Sometimes it doesn’t pay to be too specific and
precise.
Now let’s apply this point to the evolution of jealousy. Let’s suppose that
infidelity imposes different kinds of cost on the two sexes. If a man’s wife
contemplates straying, he risks losing paternity opportunities or risks
investing in another male’s offspring. If a woman’s husband contemplates
straying, she risks losing the parental care and resources he does or could
provide for her children. But, although the costs of infidelity are different
for the two sexes, it doesn’t necessarily follow that the two sexes must have
evolved mechanisms with different “design features” that specialize in track-
ing only events that are correlated with the costs specific to one’s sex. It’s
possible that both sexes possess the same mechanism, which operates at a
slightly more general level than any possible mechanisms concerned only
Marriage 315
with sex-specific costs. In particular, suppose that both sexes have the same
jealousy mechanism, which monitors the environment for, and is then
triggered by, any event that poses a threat to any relationship in which one
has invested one’s reproductive effort (so that this would include the three
contexts mentioned in the discussion of the paternity competition hypoth-
esis). If this were the case, this single mechanism would nonetheless have
somewhat different functions in the two sexes, since it would serve to
protect each sex against the particular kind of cost that sex suffers from
infidelity or abandonment. In males, it would function to protect against
losing paternity opportunities or resources, whereas in females it would
function to protect against losing a male’s parental care and resources. And,
like the “trust no one” policy, this single mechanism would be simpler to
evolve in the human lineage than two sex-differentiated mechanisms,
since it wouldn’t require separate developmental mechanisms in the two
sexes that have been perfected by selection over evolutionary time to
perform highly sex-specific functions.
Let me put all this another way. The fact that there is a sex difference in
the potential costs of infidelity doesn’t entail that selection must have
created a corresponding sex difference in the “design features” of the mind.
The fact that men and women might stand to lose different things from a
partner’s infidelity or abandonment doesn’t entail that selection must have
designed the minds of the two sexes to work differently, to function in ways
that are highly focused only on what each sex has to lose. Rather, selec-
tion could have designed the jealous mind to function exactly the same
in both men and women: to become jealous in response to any event, such
as a sexual infidelity, that provokes anxiety that one might lose a (poten-
tial) partner to another. Evolutionary Psychologists believe that each sex
has its own sex-specific module for jealousy, which contains a lot of innate
information about sex-specific threats to reproductive interests and
domain-specific procedures for operating with that information. I’m sug-
gesting, in contrast, a very minimal conception of the jealousy mechanism
that selection has designed, according to which there is no “built-in”
information about sex-specific threats to reproductive interests, but instead
only a more general responsiveness to situations and events that pose a
threat to a relationship in which one has invested one’s reproductive effort.
This doesn’t mean, however, that this minimal jealousy mechanism can’t
generate some sex differences in the way that people respond to particu-
lar circumstances. It only means that those sex-differentiated responses
aren’t built in to the way that the minds of the two sexes work.
This proposal, of course, is highly theoretical and abstract at this point.
We’ve already seen evidence that seems to favor Evolutionary Psychology’s
316 Chapter 6
theory that there are evolved sex differences in the psychological mecha-
nisms of jealousy, and it’s not at all clear how this alternative suggestion
is supposed to explain, or explain away, those differences. That will become
clearer, however, as we take a closer look at the evidence that Evolution-
ary Psychologists have accumulated in support of their theory. So we turn
now to a closer examination of that evidence.
Let’s begin by examining the questionnaire data that Evolutionary Psy-
chologists have gathered. In one question that appeared in the question-
naires, subjects were presented with the following “dilemma”:
Please think of a serious committed romantic relationship that you have had in the
past, that you currently have, or that you would like to have. Imagine that you dis-
cover that the person with whom you’ve been seriously involved became interested
in someone else. What would distress or upset you more (please circle only one):
(A) Imagining your partner forming a deep emotional attachment to that person.
(B) Imagining your partner enjoying passionate sexual intercourse with that other
person.39
Table 6.1
Infidelity Dilemma 1: Percentage of Respondents Choosing Sexual Infidelity as More
Upsetting by Survey Sample
Survey sample
Male 60 53 61 55 76 73
Female 17 23 18 32 32 4
Male 21 51 28 59 38
Female 5 31 16 18 13
Marriage 317
(A) Imagining your partner trying different sexual positions with that other person.
(B) Imagining your partner falling in love with that other person.41
This question, in which the order of the sexual infidelity and emotional
infidelity was switched from the original dilemma, was then used by a
number of other researchers in several subsequent studies in a total of five
societies. The results of the second set of studies are presented in table 6.2,
where the numbers are the percentages of respondents who chose (A), the
sexual infidelity.42
Notice that there is a significant sex difference in the results of these
studies. In no study did more women than men report sexual infidelity to
be more upsetting than emotional infidelity. Indeed, averaged across the
results of all studies, in response to both dilemmas many more men than
women reported sexual infidelity to be more upsetting than emotional infi-
delity—51 percent of the men versus 22 percent of the women in response
to the first dilemma (table 6.1), and 38 percent of the men versus 13
percent of the women in response to the second dilemma (table 6.2). It is
this sex difference that Evolutionary Psychologists have emphasized in
their description of the results and that they have cited as providing strong
support for Evolutionary Psychology’s theory of jealousy.
But the simple existence of a sex difference is insufficient to support Evo-
lutionary Psychology’s theory. For the existence of a sex difference is actu-
ally an indirect consequence of Evolutionary Psychology’s theory, being
entailed by Evolutionary Psychology’s primary claims regarding the adap-
tive problems that jealousy evolved to solve for each of the sexes. Accord-
ing to Evolutionary Psychology, the adaptive problem that male jealousy
evolved to solve was that of protecting against misspent parental invest-
Table 6.2
Infidelity Dilemma 2: Percentage of Respondents Choosing Sexual Infidelity as More
Upsetting by Survey Sample
Survey sample
Male 44 44 47 43
Female 12 12 12 11
Male 23 30 53 32
Female 12 8 22 15
318 Chapter 6
sexual infidelity in response to the second dilemma (table 6.2) are as high
as 47 percent in a U.S. sample and 53 percent in the Korean sample, they
are as low as 23 percent in the Dutch sample and 30 percent in the German
sample. These results hardly support Buss’s claim that “men’s jealousy
appears to be more sensitive to cues of sexual infidelity” and that this is
true “across cultures.”43 And, again, even though more men than women
reported sexual infidelity to be more upsetting than emotional infidelity
for both dilemmas and in all samples, this sex difference in itself does not
support Evolutionary Psychology’s claim that males have evolved to be
especially sensitive to cues of sexual infidelity. The sex difference certainly
does require explanation. But the fact that male responses do not consis-
tently indicate the focused concern with sexual infidelity that Evolution-
ary Psychology predicts raises serious doubt about whether Evolutionary
Psychology’s explanation of the sex differences is correct.
So the very data that Evolutionary Psychologists cite as confirmation
of their theory do not, in fact, provide clear support for it. But there are
other data that provide further difficulties for Evolutionary Psychology.
Buss’s questionnaire dilemmas were administered to homosexual men and
women in several studies, and the results from homosexual men are par-
ticularly puzzling from the perspective of Evolutionary Psychology’s
theory. The psychologists Virgil Sheets and Marlow Wolfe found that only
24 percent of homosexual men chose sexual infidelity as more upsetting
than emotional infidelity in the first dilemma (choosing between a
partner’s forming an emotional attachment versus having passionate sex)
and only 5 percent chose sexual infidelity as more upsetting in the second
dilemma (choosing between a partner’s falling in love with someone else
versus trying different sexual positions). The psychologist Christine Harris
administered only the second dilemma to a sample of homosexual men,
of whom only 13 percent chose sexual infidelity as more upsetting than
emotional infidelity. Finally, the psychologist Michael Bailey and his col-
leagues administered Buss’s two dilemmas to both homosexual and het-
erosexual men and women and found that homosexual men were even
less likely than heterosexual women to report sexual infidelity to be more
upsetting than emotional infidelity. In sum, all of these studies found
homosexual men to be far less likely than heterosexual men to find sexual
infidelity more upsetting than emotional infidelity. Indeed, homosexual
men rather overwhelmingly report emotional infidelity to be more upset-
ting than sexual infidelity.
This fact is difficult to reconcile with Evolutionary Psychology’s claim
that there are sex differences in the evolved “design features” of the
320 Chapter 6
they differ in their beliefs about how closely the two forms of infidelity are
linked in the minds and behavior of members of the opposite sex. Men,
according to this hypothesis, believe that women are unlikely to have sex
without being in love, but that they can be in love without having sex.
Consequently, when a man suspects his partner of a sexual involvement
with another man, he infers that she must be in love with him as well. In
contrast, if he suspects his partner of an emotional involvement with
another man, he doesn’t necessarily infer a sexual involvement as well. For
a man, then, a partner’s sexual infidelity represents a “double shot” of infi-
delity, since the sexual infidelity is believed to be accompanied by an emo-
tional infidelity, whereas her emotional infidelity does not represent a
“double shot” of infidelity. Similarly, according to this hypothesis, women
believe that men can easily have sex without being in love, but that they
can’t be in love without wanting sex. Consequently, when a woman sus-
pects her partner of an emotional involvement with another woman, she
infers that a sexual involvement with that other woman is likely or forth-
coming as well. In contrast, if she suspects her partner of a sexual involve-
ment with another woman, she won’t automatically infer that he must
also be in love with that woman. For a woman, then, a partner’s emotional,
but not his sexual, infidelity represents a “double shot” of infidelity. Thus,
when faced with one of Buss’s forced-choice questionnaire dilemmas,
males choose sexual infidelity as more distressing because a female’s sexual
infidelity is more likely than her emotional infidelity to signal a “double
shot” of infidelity, whereas females choose emotional infidelity as more
distressing because a male’s emotional infidelity is more likely than his
sexual infidelity to signal a “double shot” of infidelity.
Of course, if the double-shot hypothesis is correct, we should find evi-
dence that men and women differ in their beliefs about the connection
between sex and love in the minds and behavior of members of the oppo-
site sex. So DeSteno and Salovey and Harris and Christenfeld conducted a
series of studies designed to determine whether the sexes’ beliefs do differ
in the predicted ways. The studies differed slightly in their details, but they
all involved presenting subjects with two types of question. In one type,
subjects were asked to suppose that their mates, or other members of the
opposite sex, had become sexually involved with someone else. They were
then asked to indicate (on a nine-point or five-point scale) how likely they
thought it was that their mates, or those other members of the opposite
sex, had become emotionally involved with that other person as well. In
the other type of question, they were asked to imagine an emotional
involvement and then asked how likely it was that that involvement would
324 Chapter 6
Imagine that you discover that the person with whom you’ve been seriously
involved became interested in someone else. What would upset or distress you more
(please circle only one):
(A) Imagining your partner forming a deep emotional (but not sexual) relationship
with that person.
(B) Imagining your partner enjoying a sexual (but not emotional) relationship with
that person.45
Marriage 325
(A) Imagining your partner having sexual intercourse with that person, but you are
certain that they will not form a deep emotional attachment.
(B) Imagining your partner forming a deep emotional attachment to that person,
but you are certain that they will not have sexual intercourse.47
Imagine that your partner both formed an emotional attachment to another person
and had sexual intercourse with that other person. Which aspect of your partner’s
involvement would upset you more?
Table 6.3
Infidelity Dilemma 3: Percentage of Respondents Choosing Emotionless Sexual Infi-
delity as More Upsetting by Survey Sample
Survey sample
Male 43 45 62
Female 18 19 37
326 Chapter 6
Table 6.4
Infidelity Dilemma 4: Percentage of Respondents Choosing Emotionless Sexual
Infidelity as More Upsetting by Survey Sample
Survey sample
Male 65 54 75
Female 31 30 75
Table 6.5
Infidelity Dilemma 5: Percentage of Respondents Choosing the Sexual Aspect of
Infidelity as More Upsetting by Survey Sample
Survey sample
Male 61 47 33
Female 13 27 21
With the exception of the Japanese response to the second of these new
dilemmas (table 6.4), there is a significant sex difference in subjects’
responses, with more males than females reporting the sexual aspects
to be more upsetting than the emotional aspects, as predicted by Evolu-
tionary Psychology. However, the fact that a majority of American men
reported a partner’s “deep emotional (but not sexual) relationship” to be
more upsetting than a “sexual (but not emotional) relationship” (table 6.3)
is certainly contrary to Evolutionary Psychology’s general predictions
about the nature of male jealousy. Nonetheless, the fact that there is a con-
sistent sex difference in these results does appear to falsify the double-shot
hypothesis, since that hypothesis seems inconsistent with a sex difference
in responses when only a “single shot” of infidelity is guaranteed.
Buss and his colleagues also raise a couple of theoretical objections to
the double-shot hypothesis. First, they argue, the double-shot hypothesis
relies heavily on the fact that the sexes differ in their beliefs about the con-
nection between sex and love in the minds and behavior of the opposite
sex, but it offers no explanation of this sex difference. In contrast, they
argue, Evolutionary Psychology can explain the sex difference in beliefs.
In particular, women believe that men frequently have sex without an
emotional involvement, whereas men believe that women typically don’t
Marriage 327
there is any reason to suppose that the particular beliefs in question are
likely to be an innate part of our psychological equipment. And there is,
in fact, no reason to suppose that they are. Indeed, since the beliefs in
question are relevant only to an individual’s decisions about sexual rela-
tionships, and since most humans don’t embark upon sexual relationships
until their midteens or later, individuals have ample opportunity to acquire
the relevant beliefs through learning before those beliefs are implicated in
life decisions. Human social life is brimming with information about
romantic and sexual relationships, and adolescents are eager observers of
those aspects of human social life, so it is hardly necessary to have such
simple beliefs built in to the innate structure of our minds. This is an
instance of what the cognitive scientist Andy Clark calls “the 007 princi-
ple.” As Clark says: “In general, evolved creatures will neither store nor
process information in costly ways when they can use the structure of the
environment and their operations upon it [instead]. . . . That is, know only
as much as you need to know to get the job done.”52 Since the informa-
tion about sex differences in sexual strategies is easily extractable from the
environment, by Clark’s 007 principle we are unlikely to have evolved to
store that information innately. Indeed, since the benefit of having those
beliefs is easily obtainable through learning, it isn’t necessary to pay the
developmental costs of building them in to the innate structure of our
psychological equipment.
There is, however, an additional problem with the double-shot hypoth-
esis that Buss and his colleagues don’t mention. According to the double-
shot hypothesis, each sex picks as more distressing the form of infidelity
that is most likely to signal a “double shot” of infidelity. So, according to
the hypothesis, men are more upset by sexual infidelity because they actu-
ally take sexual infidelity to indicate that their partners have also been
emotionally unfaithful. But the double-shot hypothesis doesn’t explain
why a double shot of infidelity should be of greater concern than the par-
ticular “single shot” of infidelity that is actually chosen as the more dis-
tressing of the two. In other words, the double-shot hypothesis doesn’t
explain why men, for example, should actually be focused on whether
their partners have been both emotionally and sexually unfaithful rather
than just on whether their partners have been sexually unfaithful. Perhaps
it’s supposed to be obvious that being cheated on both emotionally and
sexually is worse than just being cheated on sexually. But I don’t see that
this is obvious, in much the way that I don’t see that it is obviously worse
to die from two gunshot wounds than to die from one gunshot wound
alone.
Marriage 329
sionate, many male subjects could find it harder to discount the threat of
a passionate sexual involvement. Given the descriptions of the two
options, then, and given the supposition that subjects are reacting to cues
of relationship jeopardy, it’s easy to see how male subjects could be pretty
much evenly divided in their interpretations of which of the two options
is a more reliable indicator of relationship jeopardy.
Things become even clearer when we examine the second dilemma
(table 6.2), in which subjects are asked to choose between “imagining your
partner trying different sexual positions” with another person and “imag-
ining your partner falling in love with that other person.” “Falling in love”
is not ambiguous in the way that “deep emotional attachment” is. As a
result, if subjects are reacting to cues of relationship jeopardy, we should
expect that they should find “falling in love” more threatening to a rela-
tionship than a “deep emotional attachment.” And, in fact, averaged over
all trials with the second dilemma, the percentage of males reporting
“falling in love” to be more distressing than “different sexual positions” is
62 percent, an increase of 13 percent over the number of male subjects
choosing the emotional infidelity as more distressing in response to the
first dilemma. Interestingly, the percentage of female subjects choosing the
emotional infidelity in the second dilemma also increases by 9 percent over
the number choosing emotional infidelity in the first dilemma, which con-
firms that the language “falling in love” doesn’t lend itself to being dis-
counted as nonthreatening in the way that “deep emotional attachment”
does. But, in that case, why isn’t the percentage of male respondents who
choose “falling in love” even higher than it is? Why isn’t it as high as the
percentage of female respondents choosing it? Because males believe that
a female’s sexual infidelity is still a strong signal of relationship jeopardy.
But, since “falling in love” is a stronger signal of relationship jeopardy,
we find more males reporting it to be more distressing than the sexual
infidelity.
There is a similar effect in the responses to the third and fourth dilem-
mas reported above. In the third dilemma, subjects are asked to choose
between “a deep emotional (but not sexual) relationship” and “a sexual (but
not emotional) relationship.” And in the fourth dilemma, subjects are asked
to imagine their partners forming relationships and then to choose
between “sexual intercourse . . . , but you are certain that they will not form
a deep emotional attachment” and “a deep emotional attachment . . . , but
you are certain that they will not have sexual intercourse.” In the third
dilemma, then, the choice is between emotional or sexual relationships
that are currently not characterized by an additional sexual or emotional
336 Chapter 6
American males. And, in fact, averaged across all studies in the United
States, 61 percent of American males reported sexual infidelity to be more
distressing than emotional infidelity in response to the first dilemma (table
6.1), and 44 percent reported sexual infidelity to be more distressing in
response to the second dilemma (table 6.2). In contrast, when German and
Dutch responses are averaged, only 40 percent of German and Dutch males
chose sexual infidelity in the first dilemma (table 6.1), and only 26 percent
chose sexual infidelity in the second dilemma (table 6.2).
So far, however, in comparing the relationship jeopardy hypothesis with
Evolutionary Psychology’s theory of jealousy I’ve focused primarily on the
heterosexual and homosexual responses to Buss’s questionnaire dilemmas.
We’ve seen that the relationship jeopardy hypothesis accounts for all the
data that Evolutionary Psychology accounts for and for some of the data
that Evolutionary Psychology can’t account for (such as the jealousy data
from homosexual males). But Evolutionary Psychologists have cited more
evidence than this in support of their theory, and if the relationship jeop-
ardy hypothesis is to be a contender it must explain or explain away this
other evidence as well. So let’s turn our attention now to some of the other
evidence that Evolutionary Psychologists have cited in support of their
theory.
Recall that Evolutionary Psychologists also appeal to findings regarding
a sex difference in physiological arousal in response to imagining sexual
and emotional infidelities. These results were obtained in a 1992 study in
which Buss and his colleagues hooked up subjects to several gadgets and
asked them to “imagine you find out that your partner is having sexual
intercourse” with another person and to “imagine that your partner is
falling in love and forming an emotional attachment to that person.” Buss
and his colleagues measured subjects’ physiological responses to these
imagined scenarios and found that male subjects showed a greater physi-
ological response to the sexual imagery than to the emotional imagery,
whereas females showed a greater physiological response to the emotional
imagery than to the sexual imagery. This sex difference has been taken to
indicate that males are cued in to sexual infidelity, whereas females are
cued in to emotional infidelity.
Buss and his colleagues concluded their report with a few cautions that
the study was “limited in ways that call for additional research.” In par-
ticular, they said: “future studies could test the alternative hypotheses that
the current findings reflect (a) domain-specific psychological adaptations
to cuckoldry versus potential investment loss or (b) a more domain-general
mechanism such that any thoughts of sex are more interesting, arousing,
Marriage 339
and perhaps disturbing to men whereas any thoughts of love are more
interesting, arousing, and perhaps disturbing to women, and hence that
such responses are not specific to jealousy or infidelity.”56 Despite recog-
nizing the need for this “additional research” before a victory could be
declared for Evolutionary Psychology’s theory, in the decade since this
caution was issued Evolutionary Psychologists have yet to conduct any
studies designed to rule out the more domain-general hypothesis. This lack
of evidence, however, has not prevented Evolutionary Psychologists from
declaring victory for their theory that the sexes have distinct domain-
specific psychological mechanisms adapted to sex-differentiated problems.
Christine Harris has done the called-for follow-up study, however, and
her results tell against Evolutionary Psychology’s theory. Harris instructed
one group of male subjects to imagine the same scenarios that Buss and
his colleagues used, and she instructed a second group of male subjects to
“imagine that you and your partner are having sexual intercourse” and to
“imagine that you and your partner are falling in love and forming an
emotional attachment to one another.”57 For the group that was asked to
imagine the infidelity scenarios, Harris’s results replicated those of Buss and
his colleagues. She found that males had a significantly greater physiolog-
ical response to imagined sexual infidelities than to imagined emotional
infidelities. But Harris also found that the males in the group that was
asked to imagine sex with their partners and falling in love with their part-
ners also exhibited a significantly greater physiological response to imag-
ining sex than to imagining falling in love. Indeed, Harris found that there
was no significant difference between the one group’s physiological arousal
in response to imagined infidelity and the other group’s physiological
arousal in response to imagined sexual intercourse.
Harris’s results indicate that the results obtained by Buss and his col-
leagues are unquestionably confounded by the fact that males become
more physiologically aroused by imagining events with sexual content, in
general, than by imagining events with emotional content. This is in line
with some findings by the Evolutionary Psychologists Bruce Ellis and
Donald Symons regarding sex differences in sexual fantasy. Ellis and
Symons found that men are more likely than women to have sexual fan-
tasies and to have them more frequently. More interestingly, they found
that women’s sexual fantasies tend not to be so overtly sexual. When fan-
tasizing about a sexual encounter, women tend to focus on details about
the context of and emotions involved in the encounter. The fantasies of
men, on the other hand, tend to focus on body parts and their interac-
tions, while ignoring any emotional context of the fantasized encounter.
340 Chapter 6
This sex difference in sexual fantasy, Ellis and Symons argue, accounts for
sex differences in the consumption of sexual and romantic literature and
imagery. Males are the principal consumers of hard-core pornographic
videos and magazines that feature body parts in varying degrees of mag-
nification, whereas women are the principal consumers of romance novels
and films that provide rich descriptions of the emotional context of sexual
encounters. Given this sex difference in the nature and content of sexual
and romantic fantasies, it is not surprising that males react far more
strongly to sexual imagery than to emotional imagery. Sexual imagery
comes easily to most males in a way that emotional imagery does not.
Similarly, emotional imagery comes easily to most females in a way that
explicitly sexual imagery does not. So, males show greater physiological
reactivity to imagined sexual infidelity than to imagined emotional
infidelity not because of differences in the implications of the form of
infidelity per se, but because sexual imagery is more vivid for males
than emotional imagery.
With this in mind, now, reconsider the fifth dilemma. Recall that Buss
and his colleagues posed a dilemma that guaranteed subjects that a double
shot of infidelity had occurred and asked them which aspect of it they
found most distressing. The results (table 6.5) show the usual sex differ-
ence, with males reporting the sexual infidelity to be more distressing than
the emotional infidelity. But this is a “dilemma” in which there is really
no dilemma. Subjects are presented with a situation in which their partner
is in love and having a sexual relationship with someone else, which guar-
antees the maximum likelihood of abandonment. So, by the relationship
jeopardy hypothesis, neither aspect of the infidelity should register as more
distressing than the other in such a situation, since the principal object of
concern, abandonment, is already signaled with the highest degree of like-
lihood by the description of the situation. But, in that case, why should
there be a sex difference in the responses? Because, when subjects are asked
which aspect of the imagined extrapair involvement is most distressing,
they choose the aspect that presents itself with greatest vividness to their
imaginations. As a result, males tend to choose the sexual aspect, since that
is more vivid in male imagination than the emotional aspect, and females
tend to choose the emotional aspect, since that is more vivid in female
imagination than the sexual aspect. So, this particular result is likely due
to the sex difference in sexual versus emotional imagination. Each sex is
simply responding most strongly to the aspect to which they are best
attuned. In short, the results from the fifth dilemma are also confounded
by the fact that males are more responsive to sexual than emotional
Marriage 341
imagery in general and that females are more responsive to emotional than
sexual imagery in general.
Indeed, it is quite likely that all of the questionnaire results are con-
founded to some degree by the fact that, for males, the act of imagining
an event with sexual content elicits greater physiological arousal than the
act of imagining an event with emotional content, and vice versa for
females. For, even if males are more strongly inclined to find their part-
ners’ emotional involvements with other men to be more threatening to
their relationships, and hence more jealousy inducing, than sexual-but-
not-emotional involvements, the fact that subjects are asked to imagine
their partners’ forming both sexual and emotional involvements may bias
the results. Even if all subjects have a jealousy mechanism that monitors
signals of possible desertion by their partners, and even if males take emo-
tional involvements to be stronger signals of possible desertion than sexual
involvements, the experimental design used in the studies by Buss and his
colleagues may not be eliciting unadulterated jealousy responses from male
subjects. In fact, when Buss and his colleagues issued their caution regard-
ing their results, they did not confine the scope of the caution to their
physiological study. They claimed that all of their results, questionnaire
results included, could be called into question by the discovery that “any
thoughts of sex are more interesting, arousing, and perhaps disturbing to
men whereas any thoughts of love are more interesting, arousing, and
perhaps disturbing to women.” Thus, the sex difference in imagination and
reactivity to scenarios with sexual and emotional content could affect all
the questionnaire results.
While it is perhaps impossible to determine the extent to which the ques-
tionnaire results are affected by the sex difference in imagination, and
thereby determine the extent to which the results reflect solely on the jeal-
ousy reactions of subjects, it is doubtful that the apparent sex difference
in the results would disappear if the sex difference in imagination could
be controlled for. There are good reasons for thinking that a female’s sexual
infidelity signals a greater threat to her relationship than a male’s sexual
infidelity does to his. And, if both sexes have an evolved “capacity to learn
to distinguish threatening from nonthreatening” infidelities, as per the
relationship jeopardy hypothesis, we should expect male jealousy to be
triggered by sexual infidelities to a greater extent than female jealousy. But,
none of the evidence so far considered has shown this difference to be due
to a sex difference in the “design features” of the psychological mecha-
nisms of jealousy, rather than to a difference in the acquired (though
perhaps accurate) beliefs about the sexual strategies of the opposite sex.
342 Chapter 6
This brings us, then, to the final form of evidence that Evolutionary Psy-
chologists have cited in support of their theory of jealousy, the evidence
concerning cross-cultural similarities in laws concerning adultery. Daly and
Wilson found that “cross-cultural and historical reviews of adultery law
reveal a remarkable consistency of concept: sexual intercourse between a
married woman and a man other than her husband is an offence.”58 In
fact, in every one of the societies they reviewed, which were globally rep-
resentative, they found a wife’s adultery to be sufficient grounds for her
husband to be granted a divorce. In contrast, in only a few of the societies
reviewed was a husband’s adultery considered either an offense or suffi-
cient grounds for his wife to be granted a divorce. Further, they found, in
a vast majority of societies, a wife’s adultery is considered so severe an
offense against her husband that the law accords her husband diminished
culpability for any violent acts, including murder, that he commits upon
finding his wife in flagrante delicto with a lover. As Wilson and Daly
remark: “Throughout the English-speaking world, the common law recog-
nizes three kinds of acts as sufficiently provoking to reduce murder to
manslaughter, and they constitute a virtually exhaustive list of fundamen-
tal threats to fitness: assaults upon oneself, assaults upon close relatives, and
sexual contact with one’s wife. Several American states had statutes or rulings
that made killing upon the discovery of wifely adultery no crime at all;
although these were finally abolished in the 1970s, jury acquittals and
discretionary refusals to prosecute persist.”59 Commenting on these find-
ings, Buss says: “Lawmakers and everyday jurors apparently believe that
stumbling upon carnal evidence of adultery is a provocation so severe that
many ‘rational’ men would resort to extreme violence.”60 And this is
because lawmakers and everyday jurors implicitly recognize that “a sexual
infidelity may have inflicted such a severe cost on a man in the currency
of paternity uncertainty and the associated misdirection of his invest-
ments, that killing the woman may have been a viable means of stanch-
ing the costs.”61
What Evolutionary Psychologists find interesting about all these laws is
the fact that they are so one-sided. Laws consistently grant men divorces
from their wives for wifely sexual infidelity, but rarely do they grant
women divorces from sexually unfaithful husbands on the ground of
sexual infidelity alone. Laws consistently accord diminished culpability to
men who murder the wives they have found in flagrante delicto with a
lover, but they do not extend the same treatment to women who catch
their husbands with a lover. This “double standard,” Evolutionary Psy-
chologists argue, is due to an implicit and universal recognition of a dif-
Marriage 343
ference in the costs to men and women whose spouses are sexually unfaith-
ful. While a wife’s sexual infidelity imposes the “severe costs” of “pater-
nity uncertainty and misdirection of investments” upon her husband, a
husband’s sexual infidelity in itself imposes no fitness costs on his wife,
since he has an endlessly renewable sperm supply with which to provide
his wife with offspring, and his sexual infidelity needn’t necessarily dimin-
ish his investment in his and his wife’s joint offspring. This is why Wilson
and Daly’s “virtually exhaustive list of fundamental threats to fitness”—a
list of fundamental threats to fitness tout court, notice, not a list of funda-
mental threats to male fitness—includes female sexual infidelity but not
male sexual infidelity. Thus, Evolutionary Psychologists argue, the cross-
cultural legal double standard exists simply because laws are made by
people who recognize that, as a result of a sex difference in the fitness costs
of spousal sexual infidelity, a spouse’s sexual infidelity inflicts greater psy-
chological pain on men than on women. The fact that our universal “folk
psychology”—our everyday understanding of the minds and behavior of
others—recognizes that a spouse’s sexual infidelity causes men greater psy-
chological pain than women is evidence of an evolved sex difference in
the emotional weight that male and female minds place on a partner’s
sexual infidelity.
There is, of course, a different explanation of the double standards
concerning adultery in laws the world over. Throughout recorded history
the world over, men have made the laws, and they have made them to
promote their own interests, not women’s interests. Thus, rather than
reflecting a universal recognition that female sexual infidelity is more
costly and distressing to men than male sexual infidelity is to women, laws
merely reflect the self-serving interests of those who’ve made them. If
women had made the laws, laws would either contain none of the double
standards concerning adultery or they would contain a double standard
that served women’s interests.
It may seem that there is no way to test this explanation against Evolu-
tionary Psychology’s explanation of the cross-cultural similarities in
adultery laws, but there is. The psychologists Luci Paul, Mark Foss, and
MaryAnn Baenninger have conducted precisely such a test. Paul and her
colleagues argue that, if legal double standards merely reflect a universal
recognition that there is a sex difference in the pain caused by sexual infi-
delity, due to a sex difference in the fitness costs of being cheated on, men
and women should endorse the same double standards. That is, not only
should men find female sexual infidelity to be a greater offense than male
sexual infidelity, but women should concur that their infidelity imposes
344 Chapter 6
greater pain or costs on their partners than their partners’ sexual infidelity
imposes on them. In other words, if the double standards merely reflect a
real sex difference in psychological pain or fitness costs, then both men
and women should find male anger in response to female sexual infidelity
to be more justified than female anger in response to male sexual infidelity.
Paul and her colleagues tested this prediction among 92 female and 80
male subjects. To both male and female subjects, Paul and her colleagues
posed the following questions: “How angry should a guy be at a girlfriend
who has cheated on him?” and “How angry should a girl be at a boyfriend
who has cheated on her?” Subjects were asked to indicate an appropriate
degree of anger on a seven-point scale, in which 1 was labeled “not at all,”
4 was labeled “somewhat,” and 7 was labeled “very.” Female subjects indi-
cated that women should be angrier at a cheating boyfriend (an average
anger rating of 6.8) than men should be at a cheating girlfriend (an average
of 6.5). And male subjects indicated that men should be angrier at a cheat-
ing girlfriend (an average of 6.6) than women should be at a cheating
boyfriend (an average of 6.4). Thus, rather than both sexes’ endorsing a
double standard that reflects the alleged sex difference in degree of distress
in response to being cheated on, both sexes endorsed self-serving double
standards.
These results indicate that, if laws the world over had been made by
women, the double standards they embody would be reversed. This means,
however, that the existence of cross-cultural double standards in laws
regarding adultery can’t be taken as evidence of a universal understanding
that males suffer greater costs than females as a consequence of a partner’s
sexual infidelity, and thereby as evidence of a sex difference in the “design
features” of the psychological mechanisms underlying jealousy. The cross-
cultural double standards are evidence of nothing more than the fact that
men the world over have made self-serving laws and that men through-
out history have had the power to impose those laws on women.
To conclude, then, we have seen that cross-cultural legal double stan-
dards provide no evidence for Evolutionary Psychology’s claim that there
are evolved sex differences in the “design features” of the psychological
mechanisms underlying jealousy in humans. In addition, we have seen
that the physiological evidence for a sex difference in the “design features”
of jealousy is confounded by the fact that males exhibit greater physio-
logical arousal in response to imagining any event with sexual content
than to imagining any event with emotional content and females exhibit
greater arousal when imagining events with emotional content than events
with sexual content. As a result, the physiological study by Buss and his
Marriage 345
colleagues tapped more general features of the minds of its subjects, rather
than features specific to those subjects’ jealousy.
Further, we have seen that the vast array of questionnaire data actually
poses some problems for Evolutionary Psychology’s theory of sex-linked
differences in the “design features” of jealousy mechanisms. Evolutionary
Psychology’s theory is unable to account for why males, on balance, place
a greater emphasis on emotional infidelity than on sexual infidelity, why
there are such widely ranging cultural differences in the degree to which
males emphasize sexual infidelity, and why there is such a striking differ-
ence between homosexual and heterosexual males in this regard. Finally,
we have seen that there is a genuinely evolutionary alternative to Evolu-
tionary Psychology’s theory of jealousy—the relationship jeopardy hypoth-
esis—that accounts for all of the data that Evolutionary Psychology is
unable to account for in addition to accounting for all of the data that
appear to support Evolutionary Psychology’s theory. Thus, jealousy may
well be a human psychological adaptation, but there is simply no good
evidence that men and women possess distinct psychological mechanisms
that have been tailored by selection to perform different functions.
7 Parenthood
“First comes love, then comes marriage, then comes baby in the baby car-
riage.” So the old ditty goes. Following its logic, chapters 5 and 6 dealt
with love and marriage respectively, so naturally this chapter deals with
the baby in the baby carriage. In chapter 5 we examined Evolutionary Psy-
chology’s claim that humans possess evolved preferences for long-term
mates with high “mate value.” In chapter 6 we considered Evolutionary
Psychology’s claim that these mate preferences exist for the purpose of
selecting a partner for marriage. Marriage, Evolutionary Psychologists
argue, is an implicit reproductive contract, which evolved along with a
host of psychological adaptations specific to it: needs and desires that
impel us to enter and remain in long-term unions, strategic desires for
extrapair copulations, and emotional alarms that serve to detect signs of
possible infidelity and to motivate actions designed to protect one’s invest-
ment in one’s mate. If we have psychological adaptations for love and mar-
riage, which exist for the sake of reproduction, selection must also have
designed psychological adaptations for caring for the baby in the baby car-
riage. And Evolutionary Psychologists claim that we do, indeed, have an
evolved psychology of parental care. The most important and influential
work on the evolutionary psychology of parental care has been done by
the Evolutionary Psychologists Martin Daly and Margo Wilson, and that
work will be the focus of this chapter.
probability of 0.5 that a copy of that allele is present in its mother and a
probability of 0.5 that a copy of it is present in its father. Consequently,
the relatedness (r) between human parents and their offspring is 0.5.
For full siblings, also, relatedness is 0.5. The sibling relationship is a little
more complicated than that of parent to offspring, so we’ll have to go back
to some basics to see why this is the case. Suppose that the maternal geno-
type at a particular locus is A1A2 and that the paternal genotype is A3A4.
(In this example, the numerical subscripts simply indicate distinct allele
tokens on different chromosomes; they don’t necessarily indicate alleles
with different DNA sequences. So, A1A2 could be a homozygous genotype.)
There will then be four possible offspring genotypes at that locus: A1A3,
A1A4, A2A3, and A2A4. Suppose that one offspring has A1A3. Let’s select one
of the alleles from this individual at random—A1, say. Now, what is the
probability that a full sibling of this individual also has a copy of A1? Since
the sibling will have one of the above four genotypes, two of which contain
A1, the probability that the sibling also has A1 is 0.5. Similar reasoning
would reveal that it doesn’t matter which of the four genotypes we assume
to be possessed by the target offspring or which allele we select from the
assumed genotype in order to determine the probability that a sibling has
a copy of that allele. Thus, relatedness between full siblings is 0.5.
I won’t go through the reasoning involved in calculating relatedness for
each of the other types of familial relationship. But it is worth pointing
out the values of r for some other human familial relationships. For half
siblings, who share only one parent, relatedness is 0.25. The relatedness
between grandparents and their grandchildren is also 0.25, as is the relat-
edness between aunts or uncles and their nieces or nephews. The related-
ness between first cousins and between great-grandparents and their
great-grandchildren is 0.125. And the relatedness between second cousins
and between great-great-grandparents and their great-great-grandchildren
is 0.0625. Finally, relatedness between two individuals whose most recent
common ancestor was many generations ago is effectively zero.
The late English biologist William D. Hamilton showed that the coeffi-
cient of relatedness has significant and interesting implications with
respect to how selection operates. As we saw in chapter 1, selection is a
process whereby fitness-enhancing genes increase in frequency, and to say
that a gene increases in frequency is just to say that it increases the number
of its copies relative to the number of copies of alternative alleles. We also
saw that genes can increase in frequency by affecting their bearers’ phe-
notypes in ways that enhance the ability to reproduce. But Hamilton
showed that this is not the only way that genes can affect their bearers so
Parenthood 353
as to increase their own frequency. For suppose that I have the allele A. If
I reproduce, there is a 50 percent chance that my child has a copy of A.
But, if my parents reproduce and give me a sibling, there is also a 50 percent
chance that my sibling has a copy of A. So, although A could increase its
number of copies by enhancing my reproductive ability, it could also
increase its number of copies by influencing me to act in ways that increase
the chances that my parents will give me a sibling. From the “perspective”
of A, it doesn’t really matter whether I or my parents reproduce.
Indeed, in general, from the “perspective” of A, it doesn’t matter whether
it is I who reproduce or any other individual with a copy of A. This brings
us back to the coefficient of relatedness, which gives the probability that
another individual has copies of my genes. As we have seen, the other indi-
viduals who are likely to have copies of my genes are my kin—my parents,
my siblings, my children, my nieces and nephews, my grandchildren, my
cousins, and so on. Thus, if A influences me to aid the reproductive efforts
of my kin, it can be just as successful in increasing its number of copies as
it would be if it influenced my own reproduction. This led Hamilton to
conclude that selection favors not only genes that enhance their bearers’
ability to reproduce, but also genes that influence their bearers to aid the
reproductive efforts of kin.
I can aid the reproductive efforts of my kin by performing any act that
provides some fitness benefit to them. This could involve feeding or caring
for them as children, feeding or caring for their children, providing them
with essential food or resources as adults, or even serving as matchmaker.
By performing an act that provides a fitness benefit to a relative, I increase
the chances that they will successfully reproduce, and I thereby increase
the odds of producing more copies of my genes. Thus, I can contribute
copies of my genes to future generations directly, by producing offspring,
or indirectly, by helping my relatives reproduce. Accordingly, my fitness—
my expected genetic contribution to future generations—has both direct
and indirect components. The direct component is a measure of my ability
to contribute copies of my genes to future generations by reproducing,
whereas the indirect component is a measure of my ability to contribute
copies of my genes to future generations by enhancing the reproductive
abilities of my kin. Hamilton referred to the sum of the direct and indirect
components of fitness as inclusive fitness.
Of course, the acts that I perform to enhance the reproductive abilities
of my kin exact fitness costs from me, the actor. This involves a rather dif-
ferent kind of fitness cost-benefit analysis than we considered in chapter
1, where the fitness benefit of an act accrues directly to the actor, the same
354 Chapter 7
individual who pays the cost of the act. When the actor both pays the cost
and accrues the benefit of an act, we can expect the actor to perform an
act as long as the cost of the act is less than the benefit of the act (weighted,
of course, by the probability that the benefit will ensue). As long as that
condition is met, the act, on balance, enhances the fitness of the actor. But
under what conditions should we expect individuals to perform acts that
are costly to themselves and beneficial to others?
Hamilton argued that we should expect individuals to perform acts that
benefit others as long as the cost of the act (to the actor) is less than the
benefit of the act (to the recipient of the benefit) weighted by the related-
ness of the actor to the recipient of the benefit (since r gives the probabil-
ity that the benefit to the recipient will in fact redound to the actor’s
genes). To make this intuitive, let’s again employ the technique of repre-
senting fitness costs and benefits by whole numbers. Let’s suppose that a
1,200-calorie meal provides 12 fitness points to the person who eats it. As
long as I don’t have to spend more than 11 fitness points obtaining and
eating the meal myself, it pays me to obtain the meal and eat it. But under
what conditions would it make sense for me to obtain the meal and give
it to a sibling to eat? The meal would enhance my sibling’s fitness and
thereby increase the odds that my sibling’s genes will leave copies of them-
selves. But, of course, for any particular gene I have, there is a probability
of 0.5 that my sibling has a copy, hence a probability of 0.5 that my
sibling’s reproducing will increase the number of copies of any particular
gene that I possess. Thus, the inclusive fitness benefit that accrues to me by
feeding my sibling a 12-point meal is 6 points (the 12-point benefit to my
sibling weighted by the relatedness of 0.5). Consequently, as long as
obtaining the meal for my sibling doesn’t cost me more than 5 fitness
points, it pays me—that is, my genes—to give the meal to my sibling. Sim-
ilarly, as long as it doesn’t cost me more than 2 fitness points to obtain
the meal, it pays me to give the meal to a niece, since the inclusive fitness
benefit to me of feeding my niece a 12-point meal is 3 points (the 12-point
benefit to my niece weighted by the relatedness of 0.25).
As the above examples show, if the fitness benefit to the recipients of
my acts is assumed to be constant, my willingness to provide that benefit
to a relative should vary as a function of my relatedness to that relative.
This is because the more distant the relative, the lower the cost it is worth-
while for me to absorb in order to help that relative. Other things being
equal, if an act of mine would benefit all my relatives equally, I should be
willing to pay twice the cost to provide that benefit to a parent, sibling, or
child as I would be to provide it to a niece or nephew, and I should be
Parenthood 355
willing to pay twice the cost to provide that benefit to a niece or nephew
as I would be to provide it to a cousin. This reasoning reportedly prompted
the English biologist J. B. S. Haldane to quip that he would risk his life to
save two siblings or eight cousins.
An interesting implication of Hamilton’s theory is that parental care is
just an instance of the broader phenomenon of selection for aiding kin.
The care that I provide my children increases my inclusive fitness by
enhancing the fitness of individuals to whom my relatedness is 0.5. Of
course, in the typical case, it pays me more to care for my children than
to care for my cousins, since I am more closely related to my children, so
I get a higher return on the care that I invest in them. But, it can be just
as beneficial to me to feed and care for an infant sibling as to feed and care
for my infant child, since my relatedness to both is 0.5.
It is now but a short step to Daly and Wilson’s General Formula—namely,
that the strength of A’s love of B is proportional to the product of B’s repro-
ductive value and B’s relatedness to A. For, if familial love is the motiva-
tional mechanism that causes us to perform acts that benefit kin, and
if that motivational mechanism has been designed by selection, then, by
Hamilton’s theory, strength of familial love should vary (at least partly) as
a function of relatedness. Other things being equal, I should love my
parents, siblings, and children more than I love my aunts, uncles, nieces,
and nephews, and I should love my grandchildren more than I love my
cousins. Of course, one thing that isn’t always equal is the reproductive
value of each of my family members. My children have greater reproduc-
tive value than my parents, so although my relatedness to both is the same,
by Daly and Wilson’s General Formula I should love my children more
than my parents. Similarly, since the average reproductive value of chil-
dren is greater than that of their parents, parents should love their chil-
dren more than their children love them. But, if reproductive value is
equal, I should love members of my immediate family more than I love
distant relatives, and I should love equally all family members to whom
my relatedness is the same.
With this understanding of Daly and Wilson’s General Formula regard-
ing parental love, let’s turn to the evidence that Daly and Wilson claim
supports their General Formula. Let’s first consider the General Formula
with respect to the relationship between a parent and its genetic offspring.
Since relatedness between parents and their genetic offspring is always 0.5,
Daly and Wilson’s General Formula entails that parental love of genetic
offspring should vary as a function of the offspring’s reproductive value.
This has a couple of implications. First, since reproductive value increases
356 Chapter 7
Daly and Wilson tested the first of these predictions with data on Cana-
dian child maltreatment fatalities between 1974 and 1983. They found that
the filicide rate for children under the age of one year was thirty-four fili-
cides per million children in the population (per year).6 The rate dropped
precipitously to nine filicides per million one-year-old children, and the
rate declined steadily from there with increasing age of the child, until the
rate was less than one filicide per million pubescent children. This pattern
contrasted sharply with the homicide rate for children killed by nonrela-
tives, which hovered around five homicides per million children from
infancy through pubescence. Thus, the declining risk of filicide from
infancy to pubescence accords with the first of Daly and Wilson’s
predictions.
With respect to the second prediction, Daly and Wilson argue that we
should expect children “with a variety of imperfections predictive of poor
prospects for survival or reproduction” to be at greater risk of maltreat-
ment than children without such “imperfections.”7 These “imperfections”
would include “such congenital handicaps as spina bifida, fibrocystic
disease, talipes, cleft palate, and Down’s syndrome.”8 In a survey of studies
of child abuse in Australia, England, and the United States, Daly and
Wilson found that “those children who are severely abused include any-
where from two to ten times as many of these congenital problems as one
would expect on the basis of their incidence in the population-at-large.”9
In addition, in a study of thirty-five societies represented in the Human
Relations Area Files, a large database of ethnographic information on soci-
eties around the world, Daly and Wilson found that “deformity” or severe
illness was the second most frequently cited reason for infanticide. Thus,
children who exhibit phenotypic cues that are potentially predictive of low
reproductive value do appear to be at greater risk of maltreatment than
children who do not exhibit those phenotypic cues, which accords with
the second of Daly and Wilson’s predictions. These two findings appear to
support the hypothesis that parental solicitude varies partly as a function
of the reproductive value of offspring.
In deriving these two specific predictions from Daly and Wilson’s
General Formula regarding parental love, we held the degree of relatedness
between parent and child fixed at 0.5 because that is the relatedness
between parents and their genetic children. But many children are cared
for by “substitute parents,” adults other than genetic parents who are in
loco parentis to a child. The most common substitute parents are steppar-
ents, adoptive parents, grandparents, and other relatives (for example,
358 Chapter 7
aunts and uncles). If Daly and Wilson’s General Formula is correct, the
strength of love that substitute parents feel for their children should not
be as great as that felt by genetic parents for their children. This leads Daly
and Wilson to what they call “the most obvious prediction from a Darwin-
ian view of parental motives”: “Substitute parents will generally tend to
care less profoundly for children than natural parents, with the result that
children reared by people other than their natural parents will be more
often exploited and otherwise at risk. Parental investment is a precious
resource, and selection must favor those parental psyches that do not
squander it on nonrelatives.”10 In short, substitute parents should be more
likely than genetic parents to maltreat their children.
But not all substitute parents are genetically equal. Grandparents and
aunts and uncles have a relatedness of 0.25 to the children for whom they
are substitute parents. So, while they should not feel as much love as
genetic parents for those children, they are still rather closely related to
the children for whom they are substitute parents. Thus, while Daly and
Wilson’s “most obvious prediction” should lead us to expect grandparents
and aunts and uncles to be more likely than genetic parents to maltreat
the children for whom they provide care, we should not expect the risk of
maltreatment at the hands of these other relatives to be that much greater
than the risk at the hands of genetic parents, other things being equal.
Stepparents and unrelated adoptive parents, however, have a relatedness
to their stepchildren and adopted children that is effectively zero. As a
result, according to Daly and Wilson’s General Formula, such parents
should feel little to nothing by way of true parental love for their adop-
tive children or stepchildren (since the reproductive value of the child is,
in these cases, multiplied by a relatedness of zero). Nonetheless, steppar-
ents and unrelated adoptive parents are in roles in which they are expected
to pay the same fitness costs of parental care that are paid by genetic
parents. But, unlike genetic parents, they do not stand to reap any inclu-
sive fitness benefit from the care they provide to their children. This vio-
lates the essential logic of inclusive fitness. And Daly and Wilson argue
that this should be particularly problematic in the case of stepparenthood.
“The stepparent has, after all, usually entered into the relationship out of
an attraction to the new mate; the stepchild must frequently enter into
the remarriage decision as a cost, not a benefit. Whereas satisfying rela-
tionships with nonrelatives ordinarily involve careful reciprocity, parental
investment is exceptional: parents tolerate a cumulative imbalance in the
flow of resources. With all the good will in the world, stepparents may
strive to feel the altruism of a natural parent, but they do not always—
Parenthood 359
Paul Glick, of the U.S. Census Bureau, had estimated that 10 percent of
these children lived with a stepparent. Wilson, Daly, and Weghorst accord-
ingly estimated that 70 percent of American children lived with two
genetic parents. They made similar estimates for the percentages of chil-
dren living with a genetic mother only or a genetic father only. These per-
centages allowed population estimates of the numbers of children living
in each household type, and this made it possible to compute rates of mal-
treatment for each household type.
Wilson, Daly, and Weghorst then compared these rates with the rate of
maltreatment for the population at large (combining all household types).
They found that the rate of maltreatment of children in all age groups
living with both genetic parents was markedly lower than the rate of mal-
treatment within the population at large. In contrast, the rate of mal-
treatment for children under the age of three living with a genetic parent
and a stepparent was 4.6 times the rate for children under three within the
population at large. The rate of maltreatment for children living with a
genetic parent and a stepparent declined with increasing age of the child,
but even children aged fourteen to seventeen living with a stepparent were
maltreated at a rate 1.6 times that of the population at large. In addition,
they found that children under the age of three who lived with a genetic
mother only were maltreated at nearly 3 times the rate of maltreatment
for that age group within the population at large, and children in the same
age group living with a genetic father only were maltreated at a rate more
than 7 times that within the population at large.
Daly and Wilson were dissatisfied with these results for a couple of
reasons. First, the population estimates of the numbers of children living
in various household types, on which the calculation of maltreatment rates
was based, were unreliable. In particular, Daly and Wilson found reasons
for doubting Glick’s estimate that as many as 10 percent of all children
lived with a stepparent. And, if that estimate was too high, the calculated
rates of maltreatment of children living with a stepparent were too low.
Second, the case reports to the American Humane Association came from
a number of different states in the U.S., and the criteria for maltreatment
varied significantly from one state to the next. Consequently, Daly and
Wilson undertook to do a better controlled test of their hypothesis, in
which they could eliminate these two problems.
In the study that has become a classic within Evolutionary Psychology,
Daly and Wilson analyzed cases of child maltreatment in the municipal-
ity of Hamilton-Wentworth in Ontario, Canada, during a one-year period
from 1982 to 1983. Their sample consisted of 99 maltreated children under
Parenthood 361
the age of eighteen, who were active cases for the two children’s aid soci-
eties in Hamilton-Wentworth and who were registered as victims of mal-
treatment with the Ontario Child Abuse Registry. The children’s aid
societies were able to provide Daly and Wilson with information about the
living arrangements of all 99 children. Grouping abuse victims by age and
household type, these 99 cases broke down as shown in table 7.1.13
In order to calculate rates of maltreatment for each household type,
in 1983 Daly and Wilson conducted their own telephone survey of
Hamilton-Wentworth residents, which allowed them to collect very
specific and detailed information about 1,286 households in Hamilton-
Wentworth. Those 1,286 households included 841 children under the age
of eighteen, and each respondent with a child in the household was asked
about the relationship of each child in the household to each adult in loco
parentis to the child. Daly and Wilson considered any adult who coresided
with a child and had responsibilities for caring for that child to be a parent,
regardless of the marital status of the adults in the household. If two
genetic parents of a child lived with that child, the household was classi-
fied as a two-genetic-parent household, regardless of whether the parents
were legally married or had a common-law union. Similarly, a male who
coresided with a female and one or more of her genetic children, and who
shared parental responsibilities for those children, was considered a step-
father of those children regardless of whether he was legally married to
Table 7.1
Numbers of Maltreated Children by Household Composition in Hamilton-
Wentworth, 1982–1983
their mother. From their survey data, Daly and Wilson estimated the fre-
quency of different living arrangements of children in the Hamilton-
Wentworth area. From these estimates and population estimates derived
from the data in table 7.1, Daly and Wilson derived the rates of maltreat-
ment of children in each of the three age groups for each of the different
types of living arrangement. They then calculated the risk of maltreatment
to children in each of the other types of living arrangement relative to the
risk of maltreatment to a child living with two genetic parents. The results
are in table 7.2.14
As these results show, children aged from birth to age four living with a
single genetic parent were 12.5 times more likely to be victims of mal-
treatment than similarly aged children living with both genetic parents.
Children up to age four living with a genetic parent and a stepparent, in
contrast, were 40.1 times more likely to be victims of maltreatment than
children living with both genetic parents. From age five to ten, the rela-
tive risk to stepchildren dropped sharply, with children living with a
genetic parent and a stepparent being 19.4 times more likely to be victims
of maltreatment than children living with both genetic parents. And, for
children aged eleven to seventeen living with a genetic parent and a step-
parent, the relative risk of maltreatment dropped sharply again, reduced
effectively by half. In all age groups, however, children living with a genetic
parent and a stepparent were at a significantly greater risk of becoming
victims of maltreatment than children living with both genetic parents.
Two subsequent studies have corroborated these findings. In one of these
studies, Daly and Wilson analyzed a British report on child abuse in
England and Wales between 1983 and 1987. They found that children of
all ages who lived with a stepparent were approximately 19 times more
likely to be registered victims of physical injury than were children who
Table 7.2
Relative Risk of Maltreatment by Household Composition in Hamilton-Wentworth,
1982–1983
lived with both genetic parents.15 In the other study, the psychiatrists
Kwang-iel Kim and Bokja Ko administered a questionnaire to 1,142 third
and fourth graders in two elementary schools in Seoul, Korea, that asked
about experiences of being battered by family members and about family
structure, among other things. Although Kim and Ko did not obtain pop-
ulation data that allowed them to calculate rates of battery for each house-
hold type, they nonetheless found the incidence of battery reported in
stepfamilies to be higher than should be expected given the presumed fre-
quency of stepfamilies in the population as a whole.
When Daly and Wilson focused on filicide, rather than maltreatment
generally, they found the picture to be even more bleak for children living
with a substitute parent. In their 1976 U.S. data, they found that a child
living with at least one substitute parent was “approximately 100 times as
likely to be fatally abused as a child living with natural parents only.”16 In
addition, in a study of 147 cases of filicide in Canada between 1974 and
1983 in which the perpetrator had been identified, Daly and Wilson found
that children under the age of three were roughly 70 times more likely to
be killed by a stepparent than by a genetic parent.17 As in the Canadian
child maltreatment data, the relative risk of filicide decreased with increas-
ing age of the child. But Daly and Wilson found that teenage children were
still roughly 15 times more likely to be killed by a stepparent than by a
genetic parent.18 Daly and Wilson take all of these data to provide strong
confirmation of their prediction that children who live with substitute
parents—in particular, stepparents—are at a greater risk of maltreatment
and filicide than children who live with their genetic parents.
Two subsequent studies by other researchers, however, produced results
that apparently disconfirm Daly and Wilson’s “most obvious prediction.”
In one study, the psychologists Catherine Malkin and Michael Lamb ana-
lyzed cases of physical and fatal abuse reported to the American Humane
Association in 1984. Malkin and Lamb compared the frequencies with
which genetic parents and stepparents perpetrated minor physical abuse,
major physical abuse, and fatal abuse. They found that “the risk of major
physical abuse or fatal abuse by biological parents was greater than the risk
of major physical abuse or fatal abuse by nonbiological parents. Descrip-
tive data revealed that nonbiological parents were proportionately more
likely (93%) to engage in minor physical abuse than were biological parents
(87.8%), whereas a greater proportion of biological parents (11%) engaged
in major physical abuse than did nonbiological parents (6.5%).”19 Malkin
and Lamb concluded that “biological parents were more rather than less
likely than nonbiological parents to abuse severely and to kill rather than
364 Chapter 7
Daly and Wilson have responded by pointing out that Temrin and
his colleagues calculated homicide rates for all children under the age of
sixteen, rather than for children of different age groups as Daly and Wilson
had done in their studies. Daly and Wilson argue that this method ignores
“the fact that the average child in the population at large was substantially
older (and therefore more likely to have had time to acquire a stepparent)
than the average homicide victim.”23 As a result, they claim, the method
used by Temrin and his colleagues artificially reduces the homicide rate for
children living in stepfamilies. Daly and Wilson obtained the Swedish data
and calculated the homicide rates for children aged one through four only,
an age group that accounted for 57 of the 139 homicides. Within that age
group Daly and Wilson found the rate for children living with both genetic
parents to be 3.8 homicides, and the rate for children living with one
genetic parent and one stepparent to be 31.7 homicides, per million chil-
dren (per year). Thus, Daly and Wilson conclude, when the homicide rates
are calculated properly, the Swedish study actually supports their “most
obvious prediction.”
Daly and Wilson thus argue that neither study undermines their claim
that children living with a stepparent are far more likely to be abused than
children living with both genetic parents. In their book The Truth about
Cinderella, Daly and Wilson argue that the reason that children who live
with stepparents are at greater risk is that stepparents are more likely than
genetic parents to maltreat or kill their children. Stepparents, they argue, “don’t
want to do what they feel obliged to do, namely to make a substantial
investment of ‘parental’ effort without receiving the usual emotional
rewards.”24 Stepparents suffer a “resentment of pseudo-parental obliga-
tion,” and this resentment frequently boils over in violent outbursts during
conflicts with their stepchildren, whereas the parental love that genetic
parents feel for their children typically inhibits violent outbursts when par-
enting becomes stressful.25
To further test the idea that stepparental abuse and filicide is the result
of violent outbursts precipitated by resentment, Daly and Wilson did a
study of the methods of killing children under the age of five in the Cana-
dian cases of filicide mentioned earlier. Daly and Wilson focused on
paternal filicide because, for children under five, “stepparental abuse is
overwhelmingly steppaternal abuse, not necessarily because stepfathers are
more dangerous than stepmothers but because small children scarcely ever
reside with stepmothers.”26 Daly and Wilson found that stepfathers were
much more likely than genetic fathers to kill their children by hitting
them, kicking them, or striking them with a blunt object, whereas genetic
366 Chapter 7
fathers were more likely to kill their children by shooting them, suffocat-
ing them, strangling them, or asphyxiating them with exhaust fumes. In
addition, genetic fathers were vastly more likely to commit suicide after
killing their children, and significantly more likely to kill their wives in
the same violent incident, than were stepfathers. In a study of filicides in
England and Wales between 1977 and 1990, Daly and Wilson found the
same pattern. Daly and Wilson take these results to indicate that step-
fathers typically kill their children in impulsive, violent rages, whereas
genetic fathers typically kill their children in premeditated acts of deliber-
ately wider scope than mere filicide. “Thus,” they say, “some considerable
proportion of men’s killings of their genetic offspring appear to have been
undertaken as parts of suicidal and/or familicidal projects, in which
despondency may be of greater motivational relevance than hostility to
the victims. . . . The same cannot be said of cases in which men killed their
stepchildren.”27 In other words, stepfathers killed out of unbridled hostil-
ity toward their stepchildren.
This study appears to confirm Daly and Wilson’s claim that stepparental
abuse is the consequence of violent, impulsive outbursts precipitated by
simmering resentment over having to fill an unwanted parental role. Since
Daly and Wilson argue that parental love serves to inhibit tendencies or
impulses to react violently in conflicts with children, their finding that
stepparents are more likely than genetic parents to abuse and kill their chil-
dren out of anger appears to show that stepparents lack the parental love
for their stepchildren that genetic parents have for their own children.
Daly and Wilson thus conclude that these results support their “most
obvious prediction” that substitute parents care less profoundly for their
children than do genetic parents. And their results appear to show, more
generally, that the strength of parental love varies with the relatedness
between parent and child, which supports Daly and Wilson’s General
Formula that the strength of parental love is proportional to the product
of the reproductive value of the child and the child’s relatedness to the
parent.
The sociologists Jean Giles-Sims and David Finkelhor have been reluc-
tant to blame Daly and Wilson’s findings on stepparenthood per se, and
they have offered an argument that may have already occurred to readers
with a sociological orientation. Giles-Sims and Finkelhor argue that divorce
and remarriage rates are higher among lower socioeconomic classes, so that
stepfamilies are overrepresented among lower socioeconomic classes. In
addition, the rate of violent crime in general is higher among lower socioe-
conomic classes, and the majority of child maltreatment reports come from
Parenthood 367
lower socioeconomic classes. Thus, they argue, the underlying cause of the
elevated risk of child maltreatment could be poverty, and the apparent cor-
relation between stepfamilies and an elevated risk of child maltreatment
could simply be a by-product of the overrepresentation of stepfamilies
among lower socioeconomic classes.
But Daly and Wilson argue that poverty, or low socioeconomic status,
was not a confounding variable in their studies. First, in an analysis of data
compiled by the National Center for Health Statistics regarding the living
arrangements of children in the United States in 1976, the sociologist
Christine Bachrach found no significant socioeconomic differences
between households with two genetic parents and those with one genetic
parent and a stepparent. Since these data were collected the same year as
the American Humane Association’s maltreatment data used in Wilson,
Daly, and Weghorst’s first study, Daly and Wilson argue that the results of
that study could not have been confounded by a correlation between
poverty and stepparenthood. Second, in their Canadian study, Daly and
Wilson found that stepfamilies were not, in fact, overrepresented among
lower-income households. Stepfamilies constituted 6.4 percent of the low-
income households and 5.3 percent of the high-income households they
surveyed. Thus, Daly and Wilson conclude, the correlation between step-
parenthood and an elevated risk of maltreatment cannot be explained
away as a by-product of a correlation between poverty and stepparenthood.
So far we’ve considered the evidence that Daly and Wilson offer in
support of their claim that patterns of parental love and care accord with
predictions derived from general evolutionary considerations. That is,
we’ve considered evidence that parental motivational systems produce
effects that accord with Daly and Wilson’s General Formula. But what kinds
of process in those motivational systems cause those effects? How, in other
words, does the parental mind function so as to produce the patterns of
behavior that Daly and Wilson claim to have documented? How do parents
make “decisions” about parental investment that manage to vary invest-
ment according to the reproductive value of a child and the relatedness
between parent and child? Do parental psychologies calculate the repro-
ductive value of children and then moderate parental investment accord-
ingly? Do they have mechanisms that process information about
relatedness and then calculate the degree to which parental investment
will enhance inclusive fitness?
The answer to these last two questions is, of course, no. Daly and Wilson
in no way think that parental psychologies have been designed to process
information about reproductive value and relatedness. As we saw in our
368 Chapter 7
sexual fidelity and his assessment of the phenotypic similarity of the child
to himself and his blood relatives.”29 Consequently, Wilson and Daly con-
clude, an “evolutionary psychological view of paternal bonding suggests
that perception of paternal resemblance would be correlated with paternal
bonding.”30 The process of paternal attachment, then, begins with a post-
natal assessment of phenotypic resemblance between father and child
(although there may also be an assessment of the child’s quality), during
which a perceived resemblance presumably triggers the establishment of
an individualized love for the child. This stage is then followed by the
extended period during which there is a gradual deepening of paternal
love.
According to this theory, a positive assessment by a mother or father
during a “critical period” in the first several days after birth serves to trigger
the establishment of an individualized love for a child. When this love is
triggered, it serves as an “inhibition against the use of dangerous tactics in
conflict with the child.”31 But a negative assessment by a mother (of a
child’s quality) or a father (of a child’s quality or resemblance to him) may
result in a failure to trigger the establishment of an individualized love for
a child. According to Daly and Wilson, child maltreatment is a by-product
of this failure to “engage the evolved psychology of parental feeling,” since
that failure involves a failure to engage the inhibition against violent reac-
tions to conflict.32
But a negative assessment during the “critical period” for the establish-
ment of individualized parental love isn’t the only factor that can result
in failure to trigger that love. If parents are not exposed to their child
during the “critical period,” the mechanisms for the establishment of indi-
vidualized love can also fail to be engaged. According to Daly and Wilson,
the evolved mechanisms of parental attachment are designed for genetic
parenthood, where in the normal course of events parents are exposed to,
and have contact with, their children immediately after birth. We are
designed to develop feelings of deep, individualized love for healthy new-
borns who are born to us (for women) or resemble us (for men). Since
stepchildren are typically older when they are first exposed to their step-
parents, the “critical period” for the establishment of individualized love
has been missed, and consequently stepchildren don’t trigger parental love
in their stepparents. As a result, stepparents don’t develop the same inhi-
bition against violent reactions to conflict with their stepchildren that
parents normally do with respect to their genetic children. And that,
according to Daly and Wilson’s theory, explains why stepparents are more
likely than genetic parents to maltreat their children.
370 Chapter 7
With this understanding of Daly and Wilson’s theory in place, let’s turn
to a critical appraisal of it. In the next section, I’ll discuss evidence
pertaining to Daly and Wilson’s theory, including the evidence they have
presented in support of it. In the final section, I’ll examine just how much
Daly and Wilson’s work really tells us about the nature of parental
psychology.
The previous section reviewed the evidence Daly and Wilson have offered
in support of their General Formula that the strength of a parent’s love for
a child is proportional to the product of the child’s reproductive value and
the child’s relatedness to the parent. In doing so, it focused primarily on
the implication that parental love varies partly as a function of relatedness
between parent and child. For this entails Daly and Wilson’s “most obvious
prediction from a Darwinian view of parental motives”—namely, that
“substitute parents will generally tend to care less profoundly for children
than natural parents, with the result that children reared by people other
than their natural parents will be more often exploited and otherwise at
risk.”33 As we saw, Daly and Wilson claim that this prediction is confirmed
by their finding that children who live with a stepparent are at far greater
risk of maltreatment than children who live with both genetic parents.
In this section, I will analyze the available evidence concerning the
“most obvious prediction.” The analysis will focus primarily on evidence
concerning stepparental maltreatment, since that is the evidence that Daly
and Wilson claim confirms their hypothesis. But the “most obvious pre-
diction” claims that children who live with substitute parents, not simply
those who live with stepparents, will be at greater risk of abuse and neglect
than children who live with both genetic parents. If Daly and Wilson’s pre-
diction is right, children who live with two genetically unrelated adoptive
parents should also be at greater risk of maltreatment than children who
live with both genetic parents. So I will also present and discuss some data
concerning abuse of adopted children. I will argue that there is no reliable
evidence to support the “most obvious prediction.”
Before delving into analysis of the evidence, however, I’d like to address
a couple of misrepresentations of Daly and Wilson’s findings. It’s not
uncommon for the dissemination of scientific results to resemble the game
of telephone, in which an original message is modified radically by the
time it reaches the end user. Scientific results published in relatively tech-
nical scientific journals often get simplified, distorted, or exaggerated by
Parenthood 371
the time they reach a broader audience. Usually the news media are
responsible for such misrepresentations, but popular science writers also
sometimes distort the scientific theories and results on which they report.
Daly and Wilson’s research regarding the risk to stepchildren provides an
example of how original results get misrepresented to a broader audience
by third-party writings. In this case, however, Evolutionary Psychologists
themselves have distorted Daly and Wilson’s findings in works intended
for broad audiences. Two specific examples are worth pointing out and
correcting before we examine the evidence in greater detail.
First, in his undergraduate textbook, Evolutionary Psychology: The New
Science of the Mind, David Buss presents the data reported here in table 7.2,
then says: “These data show that children living with one genetic parent
and one stepparent are roughly 40 times more likely to be physically abused
than children living with both genetic parents.”34 In fact, if you check table
7.2 again, you’ll see that it is only children under five, not children of all age
groups, of whom this claim is true. The increased risk to children aged five
to ten was less than half, and to children aged eleven to seventeen less
than one quarter, of the increased risk to children under five. Of course,
even when Buss’s exaggeration is corrected, in Daly and Wilson’s sample
children of all age groups who live with a genetic parent and a stepparent
are at greater risk of maltreatment than children who live with both genetic
parents, and this does appear to confirm their hypothesis. However, Buss
also presents Daly and Wilson’s statistics as though they are facts about
whole populations, not just facts about their sample. Yet Daly and Wilson’s
sample was relatively small and perhaps not representative of large popu-
lations. So how confident should we be that the rates Daly and Wilson
found in their Canadian sample reflect the rates of child maltreatment in
large populations? In what follows, I will present results of a study with a
far larger and more representative sample than Daly and Wilson’s, and it
will reveal an increased risk to stepchildren that is of far lesser magnitude
than the increased risk claimed by Daly and Wilson, let alone that claimed
by Buss’s exaggeration. Misrepresentations such as the one in Buss’s text-
book have great shock value, and they make it appear that Evolutionary
Psychology has shown us an obvious and deep truth about human
psychology. But finding a lower relative risk to stepchildren will open the
door to another understanding of the data that show stepchildren to be at
greater risk of maltreatment.
The second misrepresentation occurs in Steven Pinker’s popular book
The Blank Slate: The Modern Denial of Human Nature. Pinker says: “The
psychologists Martin Daly and Margo Wilson have documented that
372 Chapter 7
stepparents are far more likely to abuse a child than are biological
parents.”35 In fact, Daly and Wilson have “documented” no such thing.
Daly and Wilson found that children who live with a stepparent and a genetic
parent are more likely to be victims of maltreatment than children who live
with both genetic parents. Daly and Wilson inferred that the elevated risk to
stepchildren is due to maltreatment at the hands of stepparents; but they
did not document that stepparents perpetrated the maltreatment that
resulted in this elevated risk. And this is a difference that makes a differ-
ence. One of the parents in a stepfamily is a genetic parent. If the elevated
risk to children in stepfamilies is due to an elevated risk of maltreatment
by a genetic parent, then Daly and Wilson’s “most obvious prediction” is
not confirmed. That prediction is confirmed only if stepparents (and other
substitute parents) are more likely than genetic parents to maltreat a child.
Another issue to be examined in this section, then, is the extent to which
stepparents, rather than genetic parents, are responsible for the elevated
risk to children in stepfamilies.
These issues are intertwined with several others, and the entire thicket
of issues will have to be wrestled with in what follows. The best way to
begin is by taking a closer look at Daly and Wilson’s classic Canadian study.
The first thing to get clear about is precisely what counted as maltreat-
ment in Daly and Wilson’s study. Since Daly and Wilson analyzed cases
that were active with local children’s aid societies, they adopted the
definition of maltreatment employed by those societies. Accordingly, they
considered behavior toward a child to be maltreatment as long as “the care
being provided by those in loco parentis is, in the opinion of child welfare
professionals, so poor or unreliable as to imperil the child.”36 One notable
aspect of this criterion is that it includes not only neglect and physical
abuse, but sexual abuse as well. In fact, of the 99 cases of maltreatment
that constituted Daly and Wilson’s sample, 28 were cases of sexual abuse.37
The earlier study with American Humane Association data also included
cases of sexual abuse, although Daly and Wilson nowhere indicate how
many of those cases were sexual abuse.
But sexual abuse and physical abuse appear to be different phenomena
with different underlying causes. In a review of available studies of child
sexual abuse, the social scientists Hilda and Seymour Parker found that
“intrafamilial child sexual abuse is generally not accompanied by physical
abuse.”38 Further, nearly all parental sexual abuse is perpetrated by fathers
or father substitutes, and the majority of victims of parentally perpetrated
sexual abuse are daughters. Parker and Parker also found that stepfathers
are overrepresented among child sexual abusers. Indeed, the sociologist
Parenthood 373
Michael Gordon has reported that stepfathers are perhaps seven times
more likely than genetic fathers to sexually abuse one of their children.
Interestingly, however, Parker and Parker found that stepfathers who
coresided with their stepdaughters during the first three years of their step-
daughters’ lives were no more likely to sexually abuse their daughters than
genetic fathers. In addition, they found that fathers who coresided with
their daughters during the first three years of their daughters’ lives and par-
ticipated in child care and nurturant tasks were far less likely than other
fathers to sexually abuse their daughters. To explain this fact, Parker and
Parker hypothesize that caring for a very young child triggers a mechanism
that inhibits the subsequent development of sexual desire for that child.
This hypothesis is a variant of the Westermarck hypothesis, which has sig-
nificant empirical support. The Westermarck hypothesis claims that living
in close proximity during the first five years of life triggers a mechanism
that inhibits the subsequent development of sexual desire between sib-
lings. According to both hypotheses, the postulated inhibitory mechanism
has the function of reducing the likelihood of incest, which is important
since inbreeding greatly increases the odds that a harmful mutation will
be passed on with fatal or debilitating consequences.
As we have seen, Daly and Wilson argue that maltreatment data can
test their theory about the nature of the motivational systems underlying
parental care. In particular, they argue that such data can test their hypoth-
esis about strength of parental love because parental love functions to
inhibit “the use of dangerous tactics in conflict with the child.”39 Assum-
ing that parental love does have this inhibitory function, the impulses that
fail to be inhibited in cases of sexual abuse undoubtedly originate in dif-
ferent motivational systems than the impulses that fail to be inhibited in
cases of physical abuse. As a result, the mechanisms of inhibition too are
undoubtedly different. A parent who strikes a child in a fit of rage is in a
much different psychological state than a stepfather who sexually molests
the pubescent stepdaughter who has recently begun sharing his home.
Calling these both “lapses of parental love” obscures the fact that cases of
sexual abuse are confounded by the troubled sexual motivation of the
abuser in a way that cases of physical abuse are not. Humbert Humbert’s
sexual obsession with Lolita didn’t originate in a simple “lack of concern
for Lolita’s welfare,” but in Humbert Humbert’s troubled sexuality. It’s
simply not the case that the desires of the sexual abuser are “normal” and
widely shared among parents, but inhibited in most parents by a “concern
for the particular child’s welfare.” This is evidenced in Parker and Parker’s
finding that sexual abuse is typically not accompanied by physical abuse.
374 Chapter 7
For, if “lack of concern for the child’s welfare” underlay sexual abuse, that
same lack of concern would manifest itself in physical abuse as well. Thus,
if the objective is to understand the motivational systems underlying
parental care, conflating sexual abuse and physical abuse is problematic.
For purposes of understanding parental motivation, as it is manifested
in “lapses of parental love” that result in “the use of dangerous tactics in
conflict with the child,” the focus should be exclusively on nonsexual
maltreatment.
If the 28 cases of sexual abuse are removed from Daly and Wilson’s
Canadian data, however, the sample consists of 71 cases of nonsexual
maltreatment. Since stepparents—in particular, stepfathers—probably
accounted for a disproportionate share of the 28 cases of sexual abuse, the
removal of the sexual abuse cases would probably lower the rate of
maltreatment for children living in stepfamilies more than it would lower
the rate of maltreatment for children living with both genetic parents. As
a result, the relative risks to stepchildren wouldn’t be as high as those given
in table 7.2.
In addition, the 71 cases of nonsexual maltreatment in Daly and
Wilson’s sample are cases in which the care was “so poor or unreliable as
to imperil the child.” Daly and Wilson are explicit that this includes cases
of neglect, not simply cases of physical abuse, but they don’t report
the full range of acts or omissions that were included under this criterion
by the children’s aid societies that cataloged the cases. This definition,
however, is very similar to the standard definition of child maltreatment
endorsed by the U.S. National Institute of Child Health and Human Devel-
opment and employed in most American studies of child maltreatment.
According to that standard definition, child maltreatment is any behavior
that “(a) is outside the norms of conduct, and (b) entails a substantial risk
of causing physical or emotional harm. Behaviors included will consist of
actions and omissions, ones that are intentional and ones that are unin-
tentional.”40 As implemented by local agencies reporting maltreatment
under this standard, maltreatment includes neglect, and neglect includes
such omissions as failure to put a child in a car seat and failure to secure
a child with a seat belt while driving.
I am certainly no advocate for the neglectful. But, by any criterion
according to which failure to use a car seat or secure a child with a seat
belt constitutes maltreatment, every child growing up in America when I
did was maltreated. While it is unquestionably desirable to raise con-
sciousness about matters of child safety, and even to penalize those who
Parenthood 375
fail to protect their children to the best of their ability, failure to use a car
seat or secure a child with a seat belt should not be conflated with such
acts as hitting or kicking a child or stubbing out a cigarette on a child’s
back. There are, of course, forms of neglect that are effectively abusive and
that clearly do betray an utter lack of concern for a child’s welfare or life.
But the class of unintentional omissions that are considered neglectful
changes over time within a society and is different across societies, and it
will include unintentional omissions that many reasonable and caring
people at some time or in some place do not recognize as endangering
their children. The class of actions that are designed to inflict suffering,
however, is an entirely different matter. If we want to understand the
“lapses of parental love” that result in “the use of dangerous tactics in
conflict with the child,” as Daly and Wilson claim, we should use data
regarding physical abuse, rather than the amorphous category of mal-
treatment, to test whether substitute parents are more abusive than genetic
parents.
While Daly and Wilson’s study employed a criterion of maltreatment
that included unintentional omissions considered serious enough to
imperil a child, there is no indication of how many of the 71 cases of non-
sexual maltreatment in Daly and Wilson’s study were cases of uninten-
tional omission. If many were, then the appropriate sample to test their
hypothesis would be even smaller than 71 cases. In the absence of defini-
tive information, let’s assume that all 71 cases were cases of physical abuse.
Although a sample of 71 cases of abuse in Hamilton-Wentworth is large
enough to obtain a significant test of whether stepchildren are overrepre-
sented in cases of abuse, it is not large or representative enough to allow
confident extrapolation of abuse rates to the population at large (in the
manner of the quote from Buss). One question, then, is whether the results
that Daly and Wilson obtained in their Canadian sample can be replicated
with another, preferably larger and more representative, sample that is
composed exclusively of cases of physical abuse, cases of acts that harmed
a child.
Another thing to note about Daly and Wilson’s Canadian study is
the fact that, of the 99 cases of maltreatment, only one was a case of
maltreatment of a child living with unrelated adoptive parents, whereas
there were three cases of maltreatment of a child living with a “biological
relative” (table 7.1). This, in itself, should raise suspicion about Daly and
Wilson’s “most obvious prediction,” since that prediction claims that chil-
dren reared by nongenetic parents should be at greater risk than children
376 Chapter 7
Table 7.3
Numbers of Physically Abused Children by Household Composition in NIS-3, 1993
Table 7.4
Estimated Numbers of Children by Age and Household Composition in the United
States, 1993 (in thousands)
Child’s age
Note: Entries are rounded to the nearest thousand, but nonrounded estimates were
used to calculate the abuse rates appearing in subsequent tables.
Table 7.5
Estimated Rates of Physical Abuse (per thousand children) by Household Composi-
tion in the United States, 1993
* The data included no cases of physical abuse in the youngest age group for these
household types.
380 Chapter 7
Table 7.6
Estimated Rates of Physical Abuse (per thousand children) by Average Relatedness
of Parents in Household in the United States, 1993
* The data included no cases of physical abuse in the youngest age group for this
household type.
382 Chapter 7
children living with only nongenetic parents (0.15), despite the fact
that 13.3 percent (or 78,283) of the children living with only nongenetic
parents lived with a single stepparent. Thus, contrary to the “most obvious
prediction,” it appears that only children living with a genetic parent and
a stepparent, rather than children living with a nongenetic parent, are at
greater risk relative to children living with genetic parents. The elevated
risk of physical abuse does not appear to be correlated with relatedness per
se, but with living in a stepfamily. And this appears to falsify Daly and
Wilson’s “most obvious prediction.”
But perhaps appearances deceive. Daly and Wilson argue that the low
rate of abuse in adoptive households should not be taken as a falsification
of their “most obvious prediction” because there are mitigating factors that
lower the rate of abuse in adoptive households. “Nonrelative adoptions,”
they say, “are primarily the recourse of childless couples who are strongly
motivated to simulate a natural family experience; rather than having their
position in loco parentis thrust upon them, they have actively sought it.
Applicants to adopt are screened by agencies, and many are rejected as
unsuitable. . . . Finally, if the adoption (or the marriage) fails, the couple
can return the child, which happens more often than is generally real-
ized.”41 The implication is that these factors serve to ensure that adoptive
households present a very low risk for child abuse and, consequently, serve
to mitigate the natural discrimination against nongenetic children that
would otherwise manifest itself in adoptive households. So, Daly and
Wilson imply, if just any “unsuitable” couple could adopt, if unwanted
children made surprise appearances in the lives of adoptive parents, and
if adoptive parents could not simply return unwanted children, the rate of
abuse in adoptive households would be far greater than it is and probably
greater than the rate in two-genetic-parent homes.
There is no doubt some truth in this argument, and it does appear to
explain away the fact that the rate of abuse in adoptive households is much
lower than the “most obvious prediction” entails that it should be. But
Daly and Wilson’s explanation of the low rate of abuse in adoptive house-
holds opens a hornet’s nest of questions about how we should interpret
the available data.
First, Daly and Wilson claim that the rate of abuse in adoptive house-
holds is low because adoptive parents “actively sought” parenthood, rather
than having it “thrust upon them.” In other words, adoptive parents
rarely abuse their adopted children because adoptive parents wanted their
adopted children. In contrast, we have seen Daly and Wilson argue, the
relationship between stepparent and stepchild is “thrust upon” the
Parenthood 383
of child abuse, as the unit of analysis. That is, the data concern the risk of
abuse to children living in households of varying parental composition,
rather than a child’s risk of being abused by a stepparent or genetic parent.
As a result, none of the data directly confirm Daly and Wilson’s prediction
that stepparents are more likely than genetic parents to abuse their chil-
dren. Only data concerning the rates of physical abuse for various types of
relationship between perpetrator and victim are capable of providing a
direct test of the prediction that stepparents are more likely than genetic
parents to abuse their children.
One significant study that took as its unit of analysis the relationship of
perpetrator to abused child, rather than the composition of the abused
child’s household, was conducted by the sociologist Richard J. Gelles and
the psychiatrist John W. Harrop. Gelles and Harrop analyzed data gathered
by the Second National Family Violence Survey, which was an anonymous
telephone survey of 6,002 households across the United States in 1985.
Households were contacted using random-dialing procedures, and the
6,002 households that made up the survey database were then selected
from among those contacted to ensure demographic representativeness. Of
the 6,002 households contacted, 3,232 included at least one child under
eighteen years of age. Interviewers asked adult respondents about the rela-
tionships among members in the household, and they asked how often in
the previous year various “conflict tactics” were used by the adult respon-
dent in dealing with a child in the household. The conflict tactics fell under
three broad categories: use of rational discussion and agreement, use of
verbal and nonverbal expressions of hostility, and use of physical force or
violence. The category of “physical force or violence” included a number
of items that Gelles and Harrop extracted to form a category of “severe vio-
lence.” This category of severe violence, on which Gelles and Harrop based
their analysis, included “the items that have a high probability of causing
an injury—kicked, bit or hit with a fist; hit or tried to hit the child with
something; beat up the child; burned or scalded the child; threatened the
child with a gun or knife; used a knife or fired a gun.”45
Gelles and Harrop analyzed the resultant data in order to calculate the
rates (per thousand children) at which tactics of severe violence were used
on genetic children and stepchildren, and their results appear in table 7.7.46
As table 7.7 shows, the rate of abuse by acts of severe violence for
stepchildren from infancy to age six is higher than that for genetic chil-
dren the same age, yet only 1.2 times higher. But the overall rate of severe
violence committed by genetic parents is 1.2 times the rate of severe vio-
lence committed by stepparents. On the whole, then, there is no substan-
Parenthood 387
Table 7.7
Rates of Severe Violence against Children (per thousand children) in the United
States, 1985
likely than genetic parents to abuse their children, yet they are no more
likely than genetic parents to admit to having abused their children. This
conjunction, however, entails that stepparents must be far more likely than
genetic parents to lie about whether they’ve abused their children. And
there is no clear reason why this should be so. Genetic parents should be
no less ashamed than stepparents to admit to child abuse, and they should
thus be no less motivated to lie about it. Yet genetic parents must be less
likely than stepparents to lie about abuse in order for Daly and Wilson to
be justified in dismissing Gelles and Harrop’s finding.
In addition, the rates of severe violence that Gelles and Harrop found
are astonishingly high—indeed, they are vastly higher than the rates of
abuse found in NIS-3 (table 7.5). Since the Second National Family
Violence Survey simply asked people about whether they’ve abused their
children, so that all respondents had the opportunity to lie about it, one
would think that a high rate of lying on the part of respondents to the
survey would result in rates of severe violence that were not substantially
greater than the rates of abuse found in official case reports of physical
abuse. Yet the overall rate of severe violence reported by stepparents in
Gelles and Harrop’s study was 10.6 times greater than the overall rate of
physical abuse at the hands of stepparents in NIS-3. If respondents to the
Second National Family Violence Survey were lying with the impunity that
Daly and Wilson imply, it’s inexplicable why they would admit to as much
severe violence as they did admit to. So, while the possibility of lying by
respondents to the Second National Family Violence Survey means that
Gelles and Harrop’s results should be met with some skepticism, there is
no compelling reason why they should dismissed wholesale. Perhaps
Gelles and Harrop’s finding that stepparents are no more likely than
genetic parents to abuse their children is fairly accurate even if the rates
they found are not. And, if their finding is accurate, it falsifies Daly and
Wilson’s “most obvious prediction.”
But Gelles and Harrop’s results don’t actually contradict Daly and
Wilson’s finding that there is an elevated risk to children living in step-
families. For, if children in stepfamilies are sufficiently more likely to be
abused by a genetic parent than are children who live with both genetic
parents, it could be the case both that stepparents are no more likely than
genetic parents to abuse their children (as per Gelles and Harrop) and that
children in stepfamilies are more likely to be abused than children living
with both genetic parents (as per Daly and Wilson). Gelles and Harrop’s
results just aren’t consistent with Daly and Wilson’s interpretation of their
finding—namely, that stepparental abuse accounts for the elevated risk to
Parenthood 389
and his colleagues also gathered information about each of the following
categories of investment: the amount of financial support for college pro-
vided to the children listed, the amount of other financial expenditures on
children aged from birth to twenty-four, and the amount of time spent in
activities with children aged five through twelve. Finally, they classified
the children of their male subjects according to the relatedness between
father and child and the relationship between the father and the child’s
genetic mother, as shown in table 7.8. Class 1 children were the genetic
children of the male and his current mate. Class 2 children were the genetic
children of the male and his previous mate. Class 3 children were the
male’s stepchildren from a current relationship (the genetic children of his
current mate). And Class 4 children were the male’s stepchildren from a
previous relationship (the genetic children of a previous mate).
Each of the four “classes” of children corresponds to a different form of
male reproductive effort. Since Class 2 children are genetic children with
a former mate, investment in Class 2 children yields no return in the form
of paternity opportunities provided by the mothers of those children.
Thus, investment in Class 2 children must be strictly a form of parenting
effort, whereby a male invests in his genetic child so as to enhance his own
fitness by enhancing that of his child. In contrast, investment in Class 3
children must be strictly a form of mating effort, whereby a male invests
in his stepchild in order to maximize his paternity opportunities with the
child’s mother. Investment in Class 1 children, however, can combine both
parenting and mating effort, while investment in Class 4 children involves
neither parenting nor mating effort.
Anderson and his colleagues found that, in each of the categories of
investment, fathers invested the most in Class 1 children (genetic children
from current relationships), and they invested the least in Class 4 children
(stepchildren from previous relationships). More tellingly, Anderson and
his colleagues found that, overall, fathers invested more in Class 3 chil-
dren (stepchildren from current relationships) than in Class 2 children
Table 7.8
Classification of Father-Child Relationships
Of course, the fact that fathers were found to invest more in genetic chil-
dren from current relationships than in stepchildren from current rela-
tionships does show that paternal care is, in part, parenting effort. But, the
fact that fathers were found to invest more in genetic children from current
relationships than in genetic children from previous relationships shows
that paternal care is also mating effort. And, since fathers were found to
invest more in stepchildren from current relationships than in genetic chil-
dren from previous relationships, paternal care appears to be primarily a
form of mating effort. If paternal care is primarily mating effort, however,
the fitness payoff of paternal care isn’t captured by a narrow calculus con-
cerning the relatedness between father and child. Consequently, we should
not expect that selection has designed the motivational systems of pater-
nal psychology to allocate paternal solicitude strictly as a function of relat-
edness between father and child, as Daly and Wilson’s view of paternal
care as parenting effort would lead us to expect.
If we follow Daly and Wilson in arguing that incidence of abuse is a neg-
ative measure of parental solicitude, we can find additional evidence from
NIS-3 that paternal solicitude is more closely tied to paternity opportuni-
ties with a child’s mother than to relatedness between father and child. So
far I’ve presented the data from NIS-3 in terms of the household compo-
sition of the abused child in order to facilitate comparison with Daly and
Wilson’s research. But, in all of the abuse cases in NIS-3, the perpetrators
were identified. So the NIS-3 data can also be mined for comparisons
of abuse rates between various types of perpetrator-victim relationship.
Table 7.9 presents the rates of abuse (per thousand children) at the hands
of genetic fathers and stepfathers acting alone within four types of
household.
Table 7.9
Estimated Rates of Physical Abuse (per thousand children) by a Father Acting Alone
in the United States, 1993
* The data included no cases of physical abuse in the youngest age group for this
household type.
394 Chapter 7
Some of the results in table 7.9 conform to Daly and Wilson’s “most
obvious prediction.” For, overall, stepfathers are 4.9 times more likely to
abuse their children than genetic fathers in two-genetic-parent homes. But,
children younger than five, whom Daly and Wilson believe to be the most
likely to be abused by a stepfather, are in fact 1.8 times more likely to be
abused by a single genetic father than by a stepfather. In fact, children of
all age groups are most likely to suffer paternal abuse at the hands of a
single genetic father, and, overall, single genetic fathers are 1.7 times more
likely than stepfathers to abuse their children. If child abuse is the flip side
of parental solicitude, these results indicate that stepfathers are more solic-
itous toward their stepchildren than single fathers are toward their genetic
children.
These facts don’t sit well with the idea that paternal solicitude is par-
enting effort. For, if it were, stepfathers should exhibit higher rates of abuse
than genetic fathers, regardless of whether a genetic father is (still) married
to his child’s genetic mother, since the relationship with his child’s mother
doesn’t affect relatedness between father and child. The fact that single
genetic fathers are more likely than stepfathers to abuse their children,
however, does conform to the idea that paternal investment is primarily a
form of mating effort. For single genetic fathers are no longer receiving
paternity opportunities from the mothers of their children, whereas step-
fathers are receiving paternity opportunities in exchange for the parental
care they provide to their stepchildren. Thus, paternal abuse is more closely
(negatively) correlated with a father’s paternity opportunities with a child’s
mother (hence his potential returns on mating effort) than with related-
ness between father and child (hence the father’s potential returns on
parenting effort).
The evolutionary reason that Daly and Wilson have for suspecting that
children are at risk from stepfathers is that paternal care is parenting effort
and stepfathers are not related to their stepchildren. However, since the
evidence favors the mating effort hypothesis regarding the evolution of
paternal care, and since patterns of paternal investment and paternal abuse
support viewing paternal solicitude as primarily mating effort, we should
reject Daly and Wilson’s particular evolutionary view of paternal solicitude.
As a result, we should not be led to infer, in the absence of concrete evi-
dence, that the elevated risk to children in stepfamilies is due primarily to
steppaternal abuse. Of course, as table 7.9 shows, stepfathers are more
likely to abuse their children than genetic fathers in two-genetic-parent
households. We will return to this datum, but first let’s examine whether
Parenthood 395
there are reasons for thinking that some of the elevated risk to children
living with a stepfather might be due to abuse by genetic mothers.
The possibility that children living with stepfathers would suffer an ele-
vated risk of abuse at the hands of their genetic mothers is clearly contrary
to Daly and Wilson’s particular “Darwinian view of parental motives.” But
it is not without an evolutionary explanation, since women who take their
children into a marriage to a man other than the father of those children
find themselves facing a conflict of reproductive interests. First, the new mar-
riage represents a potential reproductive venture, so mating effort devoted
to the new marriage has to detract from the parenting effort devoted to a
child from a former marriage. Second, and more important, any children
from the new marriage will also demand parenting effort, which will force
a mother to make a decision about how to allocate parenting effort among
all of her children. From a broadly evolutionary standpoint, we should
expect her to allocate parenting effort in a way that will maximize the
return on her invested parental care. But what kind of allocation would
achieve this?
As Anderson’s study and the data about paternal abuse show, a genetic
mother can expect a greater investment in her children from a genetic
father to whom she is married than from a stepfather. So, if a woman takes
her genetic children into a marriage to a man other than their genetic
father, she can expect her new husband to invest more in any “new chil-
dren” they might have together than in any “old children” she brings from
her previous relationship. If she invests equally in all her children, her new
children will nonetheless flourish more than her old children, since they
receive a greater investment from her husband. The return on her invest-
ment in her new children will thus be greater than the return on her invest-
ment in her old children, despite her equal investment in all. Since this
makes her investment in her old children a worse investment than her
investment in her new children, equal investment in all her children
doesn’t appear to be the optimal allocation.
To address this problem, a mother could adjust her allocation in two
ways. First, she could invest more in her old children than in her new chil-
dren, counterbalancing the differential investment by her husband in an
effort to make all of her children flourish equally. This tactic, however,
would undoubtedly generate conflict with her husband, who is likely to
expect at least an equal investment in the children from their marriage,
and this could jeopardize the mating effort she is investing in the new mar-
riage. So this option is potentially too costly. Second, she could invest more
396 Chapter 7
in her new children than in her old children. This would have the effect
of “throwing good money after good,” since her new children, who are
already flourishing more than her old children because of the higher pater-
nal investment, would flourish even more with the increased investment
from her. Of course, her old children would suffer from this allocation.
But, depending on the number and “quality” of her children from each
marriage, a mother could actually enhance the average fitness of her brood
by investing more in her new children at the expense of her old children.
Thus, selection could have designed maternal psychology to decrease
investment in old children when an increased investment in new children
would result in greater average fitness of a woman’s brood. And, if an
increased risk of abuse is correlated with decreased investment, then, under
the conditions in which this allocation of maternal care results, a mother
would be more likely to abuse her old children than her new children.
Finally, if this were the case, children living with a stepfather would be at
an increased risk of abuse at the hands of their genetic mothers relative to
children living with both genetic parents.
But, given this reasoning, a mother’s remarriage wouldn’t be strictly nec-
essary in order for children to suffer an increased risk of abuse at the hands
of their genetic mother. As Anderson’s study shows, children of single
mothers receive lower investments from their genetic fathers than children
living with married genetic parents. Consequently, a single mother’s
investment in her child will not see the returns that it would see if the
mother were (still) married to the child’s father. Further, if a single mother
is seeking a new mate, she is already headed down the path toward the
decreased investment in her child that is described above. Of course, a
single mother doesn’t yet have new children from a new relationship.
Thus, while we should expect that a single mother might invest less in her
children than a mother who is married to her children’s genetic father, she
should nonetheless invest more than a remarried mother with new chil-
dren (since her current children are, to date, her only live prospects for
transmitting her genes to the next generation). And, again, if an increased
risk of abuse is correlated with decreased investment, a single mother
should be more likely to abuse her genetic children than a mother who is
married to her children’s genetic father, and a remarried mother should be
more likely than a single mother to abuse her children from a previous
relationship. In short, relative to children living with both genetic parents,
a child should be at increased risk of abuse at the hands of a single genetic
mother and at even greater risk of abuse at the hands of a remarried genetic
mother.
Parenthood 397
There is significant evidence that the first part of this prediction is true—
namely, that genetic children of single mothers are at an elevated risk of
maltreatment relative to children living with married genetic parents.
Some of the support for this prediction comes from research by the anthro-
pologists Eckart Voland and Peter Stephan, who studied parish records,
which spanned the period from 1655 to 1939, in two German communi-
ties. First, Voland and Stephan found that “illegitimate” children whose
mothers subsequently married men who were not their genetic fathers died
in infancy at a rate that was approximately six times greater than that
of “illegitimate” children whose mothers subsequently married their
acknowledged fathers. Through an examination of the cases, Voland and
Stephan were unable to find any explanation other than “maternal manip-
ulation” (infanticide or severe neglect) for this differential incidence of
infant mortality. Second, Voland and Stephan found that women who did
not go on to marry the fathers of their illegitimate children had a better
chance of marrying at all if their illegitimate children died in infancy than
if they lived. They further found that women whose illegitimate children
lived, and who did not go on to marry at all, had an average of 1.6 chil-
dren throughout life, whereas women whose illegitimate children died in
infancy, and who went on to marry men other than the fathers of those
children, had an average of 1.9 children throughout life. These two sets of
data suggest that single mothers who do not go on to marry the fathers of
their illegitimate children have greater lifetime reproductive success if their
illegitimate children die and they marry another man than if their illegit-
imate children live and they remain unmarried. Since Voland and Stephan
found “maternal manipulation” to be the only plausible explanation of
the difference in infant mortality, they “interpret the willingness of women
to underinvest in their illegitimate children as an adaptive outcome of a
sexually selected maternal algorithm that weighs current reproduction
against future mating success.”50
In addition, Daly and Wilson found children of single mothers to suffer
an elevated risk of maltreatment. First, Daly and Wilson found that chil-
dren under the age of five living with a single genetic mother were 12.5
times more likely to be maltreated than same-aged children living with
married genetic parents (table 7.2). Similarly, they found that children aged
five through ten living with a single genetic mother were 11.8 times more
likely to be maltreated, and children aged eleven through seventeen were
8.3 times more likely to be maltreated, than same-aged children living with
married genetic parents (table 7.2). Second, in their cross-cultural study of
infanticide, Daly and Wilson found that half of all circumstances in which
398 Chapter 7
infanticides occurred were tied to the mother’s inability to deal with the
demands of child rearing and that nearly half of these circumstances con-
cerned a mother’s being unmarried or lacking assured paternal support for
the child.51 Third, in their study of infanticide in Canada between 1977
and 1982, Daly and Wilson found that unmarried mothers were between
2.5 and 7 times more likely to kill their infant children than married
mothers, depending on maternal age.52 In fact, they found that, overall,
unmarried mothers accounted for 60 percent of all infanticides while
accounting for only 12.7 percent of all live births.53
Finally, the NIS-3 data show that genetic children of single mothers
suffer a greater risk of physical abuse than children living with married
genetic parents. Table 7.10 shows the rates of abuse (per thousand chil-
dren) at the hands of genetic mothers acting alone within three types of
household. Overall, children living with a single genetic mother were 5.9
times more likely than children living with both genetic parents to be
physically abused by their mothers.
While the results in table 7.10 do support the prediction that children
living with single genetic mothers suffer an elevated risk of maltreatment
relative to children living with both genetic parents, they do not appear
to support the prediction that a child should be at greater risk of mal-
treatment at the hands of a remarried genetic mother than at the hands
of a single genetic mother. Children under the age of five were slightly
more likely to be physically abused by a remarried genetic mother than by
a single genetic mother, but overall children living with a single genetic
mother were 2.7 times more likely than children living with a remarried
genetic mother to be abused by a mother acting alone. But, somewhat in
keeping with the rationale behind the prediction that children of remar-
ried genetic mothers should be at greatest risk of genetic-maternal abuse,
children living with remarried genetic mothers were 2.1 times more likely
Table 7.10
Estimated Rates of Physical Abuse (per thousand children) by a Genetic Mother
Acting Alone in the United States, 1993
to be abused by a mother acting alone than were children living with both
genetic parents.
Table 7.10, however, doesn’t reveal all the facts about genetic-maternal
abuse rates. First, children under the age of one year living with a genetic
mother and a stepfather were 13.9 times more likely to be abused by a
genetic mother acting alone (19.5 children per thousand) than same-aged
children living with a single genetic mother (1.4 children per thousand).
Second, many victims of abuse were abused by both parents acting in
concert. If we turn our attention from cases of abuse at the hands of a
genetic mother acting alone to cases of abuse in which a genetic mother
was involved, then we do indeed find that children who live with a genetic
mother and a stepfather are the most likely to be abused by their genetic
mothers. Table 7.11 reports the rates of abuse (per thousand children)
within the same three types of household, but here I have divided the
under-five age group into three smaller age groups in order to show the
very high risk to the youngest children.
As table 7.11 shows, children younger than three suffer a dramatically
elevated risk of being abused by a genetic mother if they live with a step-
father. And, overall, the results do conform to the prediction that children
should be at greater risk of being abused by a genetic mother if they live
with a genetic mother and a stepfather than if they live with a single
genetic mother. This, in turn, supports the hypothesis, suggested by Voland
and Stephan, that mothers exhibit reduced investment in their genetic
children when they have married a man who is not the father of those
children.
We thus have two conclusions that conspire to suggest that some of the
elevated risk to children living with stepfathers is due to abuse by genetic
mothers. First, we saw that paternal solicitude is not strongly correlated
Table 7.11
Estimated Rates of Physical Abuse (per thousand children) Involving a Genetic
Mother in the United States, 1993
Genetic mother + genetic father 1.2 1.5 0.4 1.9 0.8 1.2
Genetic mother only 1.4 2.3 0.5 6.1 4.1 4.1
Genetic mother + stepfather 19.5 16.6 n.d.* 2.7 5.6 4.7
* The data included no cases of physical abuse in this age group and household type.
400 Chapter 7
with relatedness, but is more closely correlated with the paternity oppor-
tunities a male can get in exchange for paternal care. This undermines the
particular evolutionary rationale that leads Daly and Wilson to suspect that
steppaternal abuse is responsible for the elevated risk to children living
with a stepfather. Second, we saw that maternal investment in genetic chil-
dren decreases with single motherhood and decreases even further with
marriage to a man who is not the father of those children. This suggested
a plausible evolutionary explanation for why children who live with a step-
father might be at greater risk of abuse by a genetic mother than children
who live with both genetic parents. But could abuse by genetic mothers
in stepfamilies account for the elevated risk of abuse suffered by children
living with a genetic mother and a stepfather?
The answer appears to be no. If we look more directly at abuse in step-
families, the data do not confirm that genetic mothers are primarily
responsible for the elevated risk to children who live with a genetic mother
and a stepfather. Table 7.12 shows the rates of abuse (per thousand chil-
dren) according to perpetrator within households comprised of the victim’s
genetic mother and stepfather. As these results indicate, children under the
age of one year are slightly more likely to be abused by a genetic mother
than by a stepfather, and children aged one and two are slightly more likely
to be abused by both parents acting in concert than by a stepfather acting
alone. But, children three and older are significantly more likely to be
abused by a stepfather acting alone than by a genetic mother, whether
acting alone or in concert with her husband. Thus, the NIS-3 data do show
that the elevated risk to children living with a stepfather is due primarily
to abuse by the stepfather. Given this result, there appear to be only two
possible conclusions. First, Daly and Wilson could simply be right that
Table 7.12
Estimated Rates of Physical Abuse (per thousand children) in Genetic Mother–
Stepfather Households in the United States, 1993
* The data included no cases of physical abuse for these age groups and household
types.
Parenthood 401
stepparents are more likely than genetic parents to abuse their children.
Second, the results from NIS-3, like Daly and Wilson’s own results, which
appear to support Daly and Wilson’s “most obvious prediction,” could be
unreliable.
As far-fetched as this second option might appear at first glance, there
are arguments to be made in its favor. Both Daly and Wilson’s sample and
the NIS-3 sample consist entirely of officially reported cases of child mal-
treatment. This means that the samples consist entirely of cases of mal-
treatment that were brought to the attention of a professional who worked
in a capacity concerned with child welfare, were investigated in some way
by that professional or others to whom the cases were referred, and were
determined to be genuine cases of child maltreatment by those who inves-
tigated the cases. Studies that rely on official reports of child maltreatment
are thus dependent on the judgments of both those who report the cases
to child welfare professionals and the child welfare professionals them-
selves. Consequently, the factors that go into the judgments about whether
apparent harm to a child was genuinely a case of maltreatment can bias
the results of the studies that rely on official case reports.
Gelles and Harrop, as well as the sociologist Jean Giles-Sims, point out
that child welfare professionals sometimes take the presence of a steppar-
ent in the household into consideration in deciding whether a bruise or
broken bone resulted from an accident or from abuse. That is, many child
welfare professionals take the presence of a stepparent in a household to
be partly diagnostic of maltreatment. Accordingly, Gelles and Harrop argue,
“injuries to children with non-genetic parents are more likely to be diag-
nosed and reported as abuse.”54 Thus, the detection and reporting of child
maltreatment may be biased in a way that increases the proportional rep-
resentation of stepfamily cases in the data set of official case reports and
decreases the proportional representation of genetic-family cases of
maltreatment.
Daly and Wilson are fully aware of the potentially confounding effects
of diagnostic bias on their studies. Indeed, they take the issue of diagnos-
tic bias seriously enough to argue repeatedly that it cannot have had a sig-
nificant effect on their results. They offer the following single argument
against the confounding effects of diagnostic bias, which they repeat in a
number of different publications: “Such biases surely exist, and it is almost
impossible to estimate their magnitude, but they cannot begin to account
for the facts. The reasoning behind this assertion is as follows. If reporting
or detection biases were responsible for the overrepresentation of steppar-
ents among child abusers, then we would expect the bias, and hence the
402 Chapter 7
One study that demonstrates the extent to which official records can
underreport abuse fatalities was conducted by the pediatrician Katherine
Christoffel and her colleagues Nora Anzinger and David Merrill. Christof-
fel and her colleagues examined death certificates and medical examiner’s
records concerning 437 deaths of children under the age of fifteen in Cook
County, Illinois, between 1977 and 1982. In these records, deaths “are
assigned one of five ‘manners of death’: homicide, suicide, accident,
natural, or undetermined. Cases that are ruled of an undetermined manner
include those in which information is lacking to prove intentional injury
or neglect, but in which there is evidence that precludes assigning the
death a natural or accidental manner.”59 Of course, deaths for which infor-
mation was lacking to prove any intentional injury or neglect were typi-
cally not prosecuted, for lack of evidence, and were not subsequently
recorded as homicides in police records. But they were also cases for which
there was physical evidence to preclude attributing the death to sudden
infant death syndrome (SIDS), so they were not the purportedly rare cases
of disguising a child homicide as SIDS to which Daly and Wilson refer. And
there were a not insignificant number of undetermined deaths. Of the 437
deaths that Christoffel and her colleagues examined, 206 (or 47 percent)
were coded as having an undetermined manner. Further, Christoffel and
her colleagues found that “rates for the undetermined deaths exceeded
those for the deaths that were ruled as homicides through age 4 years.”60
Thus, within the class of children whom Daly and Wilson claim to be most
at risk, a child’s death was more likely to be recorded as undetermined than
as a homicide.
A much more thorough investigation of the degree of underreporting of
child maltreatment fatalities was conducted by the pediatricians Bernard
Ewigman, Coleen Kivlahan, and Garland Land. Ewigman and his col-
leagues studied all injury fatalities of children under the age of five in
Missouri from 1983 to 1986. By classifying a death as an injury fatality if
the death certificate listed any external cause as the cause of death, they
identified 384 injury fatalities in that age group during the study period.
Assuming that official records may underreport the number of child mal-
treatment fatalities because of insufficient investigation of cases and inad-
equate communication among law enforcement, departments of family
services, and medical professionals, Ewigman and his colleagues gathered
information about the 384 decedents from nine different sources of infor-
mation. “Sources used included birth certificates; law enforcement sources;
autopsy reports; fire investigation reports; DFS-substantiated abuse or
404 Chapter 7
neglect events that occurred before death; homicide reports to the Federal
Bureau of Investigation–Uniform Crime Report (FBI-UCR) system; medical
records, including emergency department and inpatient records; and the
National Highway Traffic Safety Administration’s Fatal Accident Reporting
System. The Aid to Families With Dependent Children (AFDC) status of
the child was also determined.”61
By collating all available information from these sources about each
injury fatality, Ewigman and his colleagues classified each of the 384 injury
fatalities into one of five categories: definite maltreatment, probable mal-
treatment, possible maltreatment, non-maltreatment, and inadequate
information. They classified an injury fatality as being due to “definite mal-
treatment” if it met any of the following criteria: “Substantiated as child
abuse or neglect by the Division of Family Services, perpetrator convicted
of homicide, death reported to Federal Bureau of Investigation–Uniform
Crime Report, or death coded on death certificate as a maltreatment fatal-
ity.”62 And they classified an injury fatality as being due to “probable mal-
treatment” if there were “findings that strongly suggest maltreatment
caused or contributed to the death,” where the relevant findings derived
from “the clinical history, legal or social services investigations, physical
examination, autopsy, radiological examination, toxicology, or death scene
investigation.”63
Of the 384 injury fatalities, Ewigman and his colleagues found that 121
met their definition of “definite maltreatment.” Of the 121 definite mal-
treatment fatalities, only 96 had been substantiated as abuse or neglect
fatalities by the Missouri Division of Family Services, and only 47 were
recorded as homicides in the Federal Bureau of Investigation–Uniform
Crime Report. In addition, only 58 of the 121 definite maltreatment fatal-
ities were coded as homicides on the death certificates, while 27 were coded
as accidental deaths, 20 as natural deaths, and 16 as undetermined. Thus,
a study of child maltreatment fatalities that relied solely on law enforce-
ment records would have missed a full 60 percent of the definite mal-
treatment fatalities, and it would have missed more than half of even those
cases that were substantiated as abuse or neglect fatalities by the Division
of Family Services. Further, a reliance on any single one of these sources
of evidence would result in a failure to detect at least 20 percent of the def-
inite maltreatment fatalities.
These disturbing facts clearly belie Daly and Wilson’s claim that “we can
be reasonably confident that child murders are usually detected and
recorded.”64 And the reasons why there are many probable child homicides
that are not detected and recorded derive from mundane facts about gov-
Parenthood 405
This 3-year-old died in a residential fire. Accidental manner of death was marked
on the death certificate. There were no autopsy, law enforcement, DFS, or medical
406 Chapter 7
records. The fire investigation report included photographs documenting that the
fire’s point of origin was in the child’s bedroom, where his charred body was found
under a bed with a cigarette lighter nearby. The door to the child’s bedroom was
locked with a chain. The child’s father, who was in charge of the child at the time,
escaped without injury. There was no autopsy and no indication in the fire report
that DFS or law enforcement had been notified of the case.67
Of course, it’s far from certain that cases like the above are homicides. But,
given the facts of the above case, “probable homicide” is not an unrea-
sonable classification.
Unfortunately, this problem is not confined to Missouri. While Ewigman
and his colleagues investigated all injury fatalities with an eye to deter-
mining which were definitely or probably due to maltreatment, the
medical examiner Marcia Herman-Giddens and her colleagues investigated
all fatalities of children aged ten and younger that were actually coded as
homicides in medical examiner records in North Carolina from 1985 to
1994. Herman-Giddens and her colleagues found that, during that period,
there were 259 homicides in which the victim was aged ten or younger.
Of these, 220 were identified by medical examiners’ records as child-abuse
homicides, where a homicide was considered to be due to child abuse if a
child was killed by an injury purposely inflicted by a person who was
responsible for the child’s welfare. But, when Herman-Giddens and her col-
leagues checked these 220 cases against North Carolina’s vital records
system, maintained by the State Center for Health Statistics, they found
that only 68 (or roughly 31 percent) of the cases were coded as child abuse
homicides in vital records. To put this fact in another light, 58.7 percent
of all 259 child homicides in North Carolina during the study period were
due to child abuse but not coded as abuse homicides in the vital records
system.
Similar results were obtained in a Colorado study by the epidemiologist
Tessa Crume and her colleagues. In 1989, Colorado formed a statewide
Child Fatality Review Committee (CFRC), which has subsequently
reviewed every death of a child under the age of eighteen that occurs in
Colorado. The CFRC is a multidisciplinary effort, consisting of profession-
als from coroner’s offices, medicine, social services, law enforcement, crim-
inal justice, mental health, and public health. For every child fatality that
an expert physician suspects may not be due entirely to natural causes, the
CFRC obtains and reviews any available autopsy report, medical records,
law enforcement report, district attorney report, motor vehicle accident
report, and social services history. In this respect, the CFRC functions much
as Ewigman and his colleagues did in their study, reviewing multiple
Parenthood 407
relied does cast serious doubt on any findings based on those records. With
as much as 60 percent of child abuse fatalities not recorded as such in offi-
cial homicide records, there is ample room for diagnostic bias to cloud
homicide data. In the absence of clear evidence of abuse, genetic parents
will likely not be investigated in cases of child fatality, and the death of
the child will be assigned an accidental, natural, or undetermined manner
of death, depending on whether the method causing death more closely
resembles an accident, death by natural causes, or neither. But, as we have
seen, even inflicted injury fatalities may be classified as having an unde-
termined manner of death if there is insufficient evidence to pursue a
charge of homicide against the parents.
More important, Crume and her colleagues actually found direct evi-
dence of a potential diagnostic bias against stepfathers. Crume and her col-
leagues found that maltreatment fatalities at the hands of “other relatives,”
which in their analysis included legally married stepfathers, were 1.37
times more likely to be recorded as such on death certificates than mal-
treatment fatalities at the hands of genetic parents. Moreover, they found
that maltreatment fatalities at the hands of “other unrelated” individuals,
which in their analysis included “live-in boyfriends” of victims’ mothers,
were 8.71 times more likely to be recorded as maltreatment fatalities on
death certificates than maltreatment fatalities at the hands of genetic
parents.69
This last fact is particularly important. In both Daly and Wilson’s studies
and the NIS-3 study, males who lived in a home with a woman and her
genetic children were classified as stepfathers, regardless of whether they
were legally married to the woman in the household. That is, the category
“stepfather” included both legally married stepfathers and common-law
stepfathers (which itself included live-in boyfriends). So, many members
of the group that accounts for some of the highest child abuse and filicide
rates were found by Crume and her colleagues to be far more likely than
genetic parents to get caught. Further, when Daly and Wilson examined
their Canadian filicide data regarding stepfather-perpetrated filicides, they
found that common-law stepfathers accounted for a full 89 percent of the
filicides that were attributed to stepfathers in police records.70 In terms of
the categories used in Daly and Wilson’s studies and the NIS-3 study, then,
Crume and her colleagues actually found that common-law stepfathers,
who alone almost accounted for the higher rate of filicide among stepfa-
thers, are in a group that is 8.71 times more likely than genetic parents to
have a perpetrated child maltreatment fatality recorded as such. To put this
fact in some perspective, we saw in the NIS-3 data that children living with
Parenthood 409
a genetic mother and a stepfather were 4.5 times more likely to be abused
by a stepfather than by a genetic mother. Thus, the degree of diagnostic
bias exposed by Crume and her colleagues is more than sufficient to
account for the greater rate of abuse by stepfathers in official case reports.
This has been a long and wide-ranging section, so let’s take stock of
where we’ve arrived. If Daly and Wilson’s “most obvious prediction” is
correct, then children who live with nongenetic parents should be at
greater risk of abuse than children who live with genetic parents. But we
have seen several results that don’t conform to that prediction. First, we
found that children who live with unrelated adoptive parents are even less
likely to be abused than children who live with genetic parents. Although
Daly and Wilson try to explain away this fact, their very explanation
undermines their comparison of rates of abuse in genetic families with
those in stepfamilies, which they cite in support of their “most obvious
prediction.” Second, we saw that children are at far greater risk of being
abused by a single genetic father than by a stepfather, a finding that is
inconsistent with Daly and Wilson’s supposition that paternal solicitude
is a function of relatedness. Third, we saw that children who live with step-
fathers are more likely to be abused by their genetic mothers than are chil-
dren who live with both genetic parents, so that some of the elevated risk
to children in stepfamilies is due to genetic-maternal abuse. However,
fourth, we also saw that, according to official case reports of child abuse,
a child living with a stepfather and a genetic mother is 4.5 times more
likely to be abused by the stepfather than by the genetic mother (a factor
of increased risk that is nowhere near that found by Daly and Wilson in
their small sample of maltreatment).
There are, however, reasons to suspect a diagnostic bias against steppar-
ents in official case reports of child abuse, which results in proportionately
more cases of abuse by stepparents than by genetic parents being reported
and confirmed. Daly and Wilson argue that any effects of such a diagnos-
tic bias must be negligible, since virtually all child homicides should be
detected and recorded, and the data on child homicide show that step-
parents are far more likely than genetic parents to fatally abuse their chil-
dren. But several studies have shown that a great many cases of fatal abuse
do not appear in official homicide records, so it is far from clear that the
comparative rates of homicide at the hands of stepparents and genetic
parents are reliable. In addition, one study found that a case of fatal mal-
treatment is more than eight times more likely to be recorded as such if
perpetrated by a (common-law) stepfather than if perpetrated by a genetic
parent. Thus, since all of our evidence to date concerning stepparental
410 Chapter 7
abuse derives from official case reports, we simply don’t know whether step-
parents are more likely than genetic parents to abuse their children. Like
many claims in Evolutionary Psychology, Daly and Wilson’s claim that
stepparents are more likely than genetic parents to abuse their children
goes well beyond the available reliable evidence.
Let’s stand back now from the foregoing arguments and consider some
other issues. Daly and Wilson’s stated objective is to provide us with an
understanding of the evolved motivational systems that make up the
parental mind, not simply to provide us with an evolutionarily based pre-
diction about how parents will tend to behave. But suppose, contrary to
the arguments of the previous section, that Daly and Wilson actually had
convincing evidence that stepparents are more likely than genetic parents
to abuse their children. Even that, I will now argue, would not constitute
good evidence for their theory about the workings of the evolved motiva-
tional systems of the human parent.
It has been extensively and graphically documented (albeit doubted by
a few) that male lions who take over a pride typically kill the suckling cubs
who were sired by other males. Killing the sucklings brings their mothers
into estrus sooner, and this enables the males to sire offspring with those
females sooner than they could if they simply waited for the sucklings to
be weaned. This behavior among lions is so well documented, and its adap-
tive advantages are so apparently clear, that it is widely accepted among
those who study animal behavior that male infanticide among lions is an
adaptation.
Daly and Wilson recount these facts in chapter 2 of their book The Truth
about Cinderella, right before launching into chapter 3, “Human Stepfam-
ilies,” in which they recount the alleged risks of abuse and filicide at
the hands of human stepfathers. Alas, the discussion of infanticide-
as-adaptation turns out to be nothing but a red herring in the end. For in
chapter 5, after they’ve presented all their evidence of stepparental abuse
and homicide, Daly and Wilson set the record straight with the theory I
sketched in the first section of this chapter. “Human beings are not like
langurs or lions,” they inform us. “We know that ‘sexually selected infan-
ticide’ is not a human adaptation because men, unlike male langurs and
lions, do not routinely, efficiently dispose of their predecessors’ young.
. . . Child abuse must therefore be considered a non-adaptive or maladaptive
byproduct of the evolved psyche’s functional organization, rather than an
Parenthood 411
adaptation in its own right. . . . All told, we see little reason to imagine that
the average reproductive benefits of killing stepchildren would ever have
outweighed the average costs enough to select for specifically infanticidal
inclinations.”71
But, Daly and Wilson argue, “although sexually selected infanticide is
clearly not a human adaptation, discriminative parental solicitude just as
clearly is.”72 The mechanisms of discriminative parental solicitude, they
claim, are designed to cause parents to feel deep love for a child who, first,
parents have reason to believe is their genetic offspring and, second, has
sufficiently high reproductive value. When these conditions are met, Daly
and Wilson argue, parents develop deep parental love for the child, which
creates a “parental inhibition against the use of dangerous tactics in con-
flict with the child.”73 Child abuse emerges as a by-product of the adapta-
tion of discriminative parental solicitude when the mechanisms of
discriminative parental solicitude are not activated so as to inhibit violent
reactions to annoying and conflictual behavior on the part of children.
Further, it is the putative fact that child abuse results from a lack of inhi-
bition of violent reactions to conflict with a child that is supposed to justify
using evidence about the incidence of child abuse as “reverse assay” evi-
dence about the strength of parental love, which is what Daly and Wilson’s
General Formula and “most obvious prediction” are actually about. For, if
there weren’t a connection between parental love and child abuse, via an
inhibition of the tendency to abuse, there would be no clear reason why
evidence about child abuse could serve to test Daly and Wilson’s evolu-
tionary predictions about parental love.
The above account of the psychological mechanisms underlying both
parental love and parental child maltreatment presupposes that the
natural, default psychological state of adults would lead them to react to
annoying children with physical violence. Parents only take such good care
of their children, when they do, because their deep love for their children
inhibits this natural tendency. Daly and Wilson confirm that this presup-
position underlies their theory of discriminative parental solicitude in a
comment on their discovery that the rate of filicide decreases with increas-
ing age of a child. Daly and Wilson say: “When we consider the con-
spicuous, tempestuous conflicts that occur between teenagers and their
parents—conflicts that apparently dwarf those of the preadolescent
period—it is all the more remarkable that the risk of parental homicide
continues its relentless decline to near zero.”74 In other words, it’s remark-
able that more parents don’t kill their annoying and conflictual teenagers,
since the natural impulse is to want to do so. According to this picture of
412 Chapter 7
period” for their stepparents to establish individualized love for them, they
should be at greater risk than the 20 percent who began living with their
stepparents when they were under the age of one. So, the number of chil-
dren who began living with their stepparents as infants can’t possibly
account for the lowered risk to stepchildren aged one and two. Similar rea-
soning would show that the number of stepchildren aged three and four
who began living with their stepparents as infants can’t possibly account
for the further reduction in risk to stepchildren of that age group.
Of course, the fact that the risk of abuse by a stepparent declines with
increasing age of stepchildren does conform to Daly and Wilson’s claim
that parental love gradually deepens over time. But, parental love is sup-
posed to gradually deepen over time only if it was established in the first
place. The problem with stepfamilies, according to Daly and Wilson, is that
they typically form when stepchildren are beyond the age at which the
parental mind is designed to form an individualized love for a child. Since
the vast majority of stepchildren do not begin living with their steppar-
ents in infancy, the purported gradual deepening of parental love over time
can’t account for the declining risk of stepparental abuse as stepchildren
get older. If anything, Daly and Wilson’s theory entails that stepchildren
who begin living with their stepparents when they are older, but still young
enough to be relatively defenseless, should be at greater risk than stepchil-
dren who have the opportunity to trigger the mechanisms of individual-
ized parental love in their stepparents. Yet, as we have just seen, the data
show just the opposite pattern.
At this point you might be wondering how my arguments in this section
can appeal to precisely the data (drawn from official case reports of child
abuse) that I argued in the last section are unreliable. It’s true that there is
evidence that official case reports of child abuse severely underreport the
number of cases of child abuse. But the evidence also showed that abuse
by genetic parents is more likely to go unreported than abuse by steppar-
ents. So, of the abuse rates derived from official case reports, the rates of
stepparental abuse are the least likely to be affected by any diagnostic bias
that results in the underreporting of child abuse. The rates of abuse by
genetic parents are the most likely to be affected by underreporting—in a
way that underestimates the true rates—and that is what renders suspect
a comparison of the rates of genetic-parental abuse with the rates of step-
parental abuse. Taken in themselves, however, the rates of stepparental
abuse derived from NIS-3 are probably the most reliable of all the rates
derived from it.
Parenthood 417
The last five chapters critically examined a number of the specific claims
Evolutionary Psychologists make about the evolution and nature of human
psychology. In this final chapter, I will move away from examination of
specific claims about human psychology in order to engage some broader
theoretical issues related to Evolutionary Psychology’s advertisement that
it is “the new science of human nature.”
Some of the theoretical issues examined in this chapter are absolutely
central to Evolutionary Psychology’s claim that there is a universal human
nature. That is, the very idea of a universal human nature stands or falls
with some of the theoretical arguments considered here. Other theoretical
issues engaged in this chapter are more properly “philosophical,” since
they concern the broader conceptual framework in which the idea of a uni-
versal human nature is situated and interpreted. While these issues may
be less central to Evolutionary Psychology’s narrowly focused scientific
project of discovering universal psychological adaptations and underst-
anding how they function, they are nonetheless significant. For, in
developing and promoting their account of human nature, Evolutionary
Psychologists have often endorsed positions on broader philosophical
issues, and the positions they’ve endorsed form part of a widely held,
“commonsense” understanding of the idea of human nature. Conse-
quently, it is important to understand both why their philosophical
positions are wrong and how those positions help motivate the quest
for human nature.
Throughout the discussion of these various theoretical issues, I will be
focused on a single theme—that the idea of a universal human nature is
deeply antithetical to a truly evolutionary view of our species. Indeed, I
will argue, a truly evolutionary psychology should abandon the quest for
human nature and with it any attempt to discover universal laws of human
psychology. As the evolutionary biologist Michael Ghiselin so pithily puts
420 Chapter 8
it: “What does evolution teach us about human nature? It tells us that
human nature is a superstition.”1 In other words, the idea of human nature
is an idea whose time has gone.
Let’s begin by examining what it means to talk of human nature. One pos-
sibility is that the concept of human nature could refer to the totality of
human behavior and psychology. In this broad sense, human nature would
simply be whatever humans happen to do, think, or feel, regardless of
whether different humans do, think, or feel differently. If one person is
violent, violence is part of human nature, even if another person is not
violent. If one person is kind, kindness is part of human nature, despite
another person’s inveterate unkindness, which is also part of human
nature. In this very broad sense, the concept of human nature has no par-
ticular theoretical meaning; it is merely an abbreviated way of talking about
the rich tapestry of human existence. And, if this is what one means by
human nature, no one can quibble about the existence of human nature,
since the mere existence of humans guarantees the existence of human
nature.
But, traditionally, the concept of human nature has never meant simply
whatever people happen to do, think, or feel. Regardless of the details of
the theory of human nature in which it featured, the concept of human
nature has traditionally referred to some of the things that people do but
not to others, to some of the things that people think and feel but not
to others. Theories of human nature have differed over precisely which
aspects of human behavior and psychology constitute human nature, but
they have all used the concept of human nature to pick out only a small
part of everything about humanity that meets the eye. That is, regardless
of the theory of human nature in which it featured, the concept of human
nature has traditionally designated only a proper subset of human behav-
ior and mentation, which was claimed to belong to human beings by their
nature as opposed to behavior and mentation that was claimed not to be
owing to or in accordance with that nature. And there are three notewor-
thy features of this traditional concept of human nature.
First, the concept of human nature has always refered to what is dis-
tinctively human about us, to what distinguishes humans from the other
animals on the planet. This aspect of its meaning put the human in the
concept of human nature, and it is what David Buss alludes to when he
“Human Nature” 421
tions can be drawn between aspects of human life that are genetically
transmitted across generations and aspects of human life that are trans-
mitted in other ways, just as we can draw a biological distinction between
genotype and phenotype. Accordingly, the biologist John Bonner defines
culture as “the transfer of information by behavioral means, most particu-
larly the process of teaching and learning,” which he distinguishes from
“the transmission of genetic information passed by the direct inheritance
of genes from one generation to the next.”8 In this sense, culture is present
in a vast array of species, and its evolution predated the emergence of
modern humans. Thus, culture is a biological phenomenon, in the very broad-
est sense of the word biology, despite not being a genetically determined
or genetically transmitted phenomenon. Consequently, the traditional
arguments that there is no human nature, because humans are what they
are due to cultural socialization rather than biology, rest upon a false
dichotomy.
Although Evolutionary Psychology’s conception of human nature
doesn’t involve the traditional dichotomy between human biology
(nature) and human culture, it is highly dependent on a dichotomy
between different biological characteristics of humans. As Tooby and
Cosmides say, “the concept of a universal human nature,” as employed
in Evolutionary Psychology, is “based on a species-typical collection of
complex psychological adaptations.”9 Evolutionary Psychology’s conception
of human nature is thus restricted to universal adaptations, which consti-
tute only a proper subset of the biological characteristics to which the tra-
ditional concept of human nature has applied. If there are universal
psychological characteristics that evolved under genetic drift, for example,
these would not count as part of human nature for Evolutionary Psy-
chologists, although they would for traditional theories that include in
human nature all universal biological traits. Consequently, the contrast
between nature and culture that is part of the meaning of the traditional
concept of human nature is replaced within Evolutionary Psychology’s
conception of human nature by the contrast between traits that are uni-
versal adaptations and traits that aren’t. In sum, then, according to Evo-
lutionary Psychologists, human nature consists of a set of psychological
adaptations that are presumed to be universal among, and unique to,
human beings.
In the remainder of this chapter, I will argue that Evolutionary Psychol-
ogy’s theory of human nature is multiply problematic. For the most part,
these problems are shared by the traditional concept of human nature. So,
while my arguments will be directed at Evolutionary Psychology, they will
424 Chapter 8
coarser scales of description, typifies the body plan of our order (primate),
class (mammal), and subphylum (vertebrate)—consists primarily of fea-
tures that have persisted down lineages and through speciations for tens to
hundreds of millions of years. Although selection probably played a role
in designing the basic body plan that now characterizes humans, it did not
design that structural plan during human history, but rather during the
history of the common ancestor of humans and other primates, mammals,
or vertebrates. Consequently, even though all humans may have two eyes,
two hands, one nose, and a mouth, it doesn’t follow that similarly
universal adaptations emerged during comparatively recent human
evolutionary history.
Finally, strictly speaking, there is no single human anatomy and physi-
ology possessed by all humans around the world of which Gray’s Anatomy
provides a “detailed” and “precise” description. Approximately 0.25 per-
cent of all humans are born with only one kidney, rather than two, yet
nonetheless live reasonably healthy lives. Others are born with three
kidneys, yet still live healthy lives (although there are no solid estimates
of the incidence of this phenomenon). In addition, somewhere between
“one in every 8,000 to 25,000 people is born with a condition known as
situs inversus, in which the positions of all the internal organs are reversed
relative to the normal situation (situs solitus): the person’s heart and
stomach lie to the right, their liver to the left, and so on. (The organs are
also mirror images of their normal structures.)”14 There is no more precise
estimate of the incidence of situs inversus because it creates no medical
complications, so it is typically discovered only incidentally to routine
physical examination (if sought) or medical treatment for some other con-
dition. At the physiological level, there are four main blood types in
humans (A, B, AB, and O), which are genetically coded for at a single locus.
If we move from the four blood types coded for at that one locus to
examine broader categories of blood type, there are more than twenty addi-
tional blood types in humans. And, moving to the outside of the body,
approximately one in every fifteen hundred infants is born with ambigu-
ous genitalia, which do not allow the assignment of a sex. Thus, the idea
that Gray’s Anatomy provides a single “detailed” and “precise” picture of
the anatomy and physiology of every human on earth is plausible only if
one ignores known facts about human anatomical and physiological vari-
ation. Although most of us are pretty much the same in a lot of “coarse”
details, we are not all cast from the same anatomical and physiological
mold, so there is no reason to think that there is a single psychological
mold from which we are all cast. Despite its intuitive appeal, the argument
428 Chapter 8
“By virtue of being members of the human species, all humans are
expected to have the same adaptive mechanisms.”17 In other words, mem-
bership in the same species entails the shared possession of the essential
characteristics definitive of the species. Elsewhere, in a clear expression of
the essentialist view that species are natural kinds, Cosmides and Tooby
say, “the species-typical genetic endowments of species, and the common
ancestry of larger taxa do cause an indefinitely large set of similarities to
be shared among members of a natural kind, as does a common chemical
structure for different instances of a substance.”18 Finally, tying essential-
ism directly to the concept of human nature, the Evolutionary Psycholo-
gist Donald Brown writes: “Universals of essence at the level of the
individual collectively constitute human nature.”19
But how can essentialism regarding species be reconciled with the exis-
tence of organisms that appear to belong to Homo sapiens even though they
don’t possess all of the “qualities that define us as a unique species”? If
species are natural kinds, so that an organism is a member of a species if
and only if it possesses the characteristics essential to that species, and
if some people don’t actually possess all the characteristics that define
human nature, which is the essence of Homo sapiens, aren’t those people
not actually human beings? Isn’t essentialism committed to claiming that
people who lack a characteristic essential to the human species simply
aren’t human? And, if so, how can essentialism integrally involve a dis-
tinction between “normal” and “abnormal” human beings? Aren’t “abnor-
mal” humans not actually human, so that, strictly speaking, there is no
such thing as an “abnormal” human?
Throughout the history of essentialism there has been a tension between
essentialism regarding species and apparent variation within species. The
usual way of resolving this tension is to conceive of a species’ essence as a
causal mechanism that produces the phenotypic characteristics considered
definitive of membership in that species. This involves what Sober calls
the “Natural State Model.” According to the Natural State Model, “there is
a distinction between the natural state of a kind of object and those states
which are not natural. These latter are produced by subjecting the object
to an interfering force. . . . The cause for this divergence from what is natural
is that these objects are acted on by interfering forces that prevent them
from achieving their natural state by frustrating their natural tendency.
Variability within nature is thus to be explained as a deviation from what
is natural.”20 When applied within biology, the Natural State Model entails
that “there is one path of foetal development which counts as the real-
ization of the organism’s natural state, while other developmental results
“Human Nature” 431
(or genotypes) and the environment; but they in fact do not. For selection
never “designs” traits for particular environments in isolation from com-
peting traits. To say that a trait is “adapted to” or “designed for” a partic-
ular environment is simply shorthand for saying that the trait was selected
over alternative traits in that environment. And that, in turn, simply means
that individuals with that trait had higher average fitness in that environ-
ment than individuals with alternative traits. Thus, to say that a trait is
“adapted to” or “designed for” a particular environment emphatically does
not mean that the trait is a perfect “fit” for that environment, that the trait
is the fittest of all possible traits in that environment, or that the trait has
higher fitness in that environment than in any other.
If the motivation for identifying a genotype’s “natural environment”
with its EEA is that the EEA is the environment in which the genotype
made the greatest contribution to fitness (by producing a trait that
enhanced fitness), then there are undoubtedly other environments that
would be better candidates for a genotype’s “natural environment.” For
example, the EEA of a genotype is simply the environment in which that
genotype had higher fitness than available alternative genotypes in the popu-
lation. In a different environment, the genotype may have had an even
greater fitness advantage over those alternatives. So why not identify the
“natural environment” of a genotype with the environment in which the
genotype has its highest fitness? Similarly, had a genotype competed in its
EEA against a different set of alternative genotypes, one of those alterna-
tives may have had higher fitness than the genotype that was actually
selected. Why should a genotype’s EEA be the “natural environment”
for that genotype rather than for some other genotype that would have
had higher fitness in that environment? Had a mutation occurred that
improved the human eye so that it could see as well at night as during the
day, for example, the genotype for that supereye would have been selected
over the genotype for the typical human eye in the EEA of the human eye.
Why should the EEA of the human eye be the “natural environment” for
the human eye rather than for the supereye that would have been selected
in that environment had it actually been present in our ancestral popula-
tion? If a genotype’s “natural environment” is defined in terms of a geno-
type’s fitness, there are no principled grounds on which to identify as a
genotype’s natural environment its EEA rather than an alternative envi-
ronment in which it would have higher fitness, or to identify a genotype’s
EEA as its natural environment, rather than that of an alternative geno-
type that would have had higher fitness in that environment. Thus, it is
arbitrary to call a genotype’s EEA its “natural environment.”
436 Chapter 8
doesn’t label some of the genetic variants “normal” and others “abnor-
mal.” From the standpoint of population genetics, there are simply a
variety of genotypes that change in frequency across generations. A new
mutation, which may or may not increase in frequency under selection, is
no more or less normal than a statistically more frequent allele at the same
locus. Any distinction between “normal” and “abnormal” genotypes must
be imposed on genetic theory from a nonbiological perspective.
Therefore, the Natural State Model, on which any distinction between
“normal” and “abnormal” human characteristics must rely, has no basis
in biology. Nothing in biology justifies viewing certain phenotypes, but
not others, as the “normal” phenotypes for a genotype, and nothing in
biology justifies viewing certain genotypes, but not others, as the “normal”
genotypes for humans. There is substantial variation in human popula-
tions at both the phenotypic and genetic levels, and our best biological
theories to date simply do not partition that variation into “normal” and
“abnormal” variants. As Sober so nicely puts it: “Our current theories of
biological variation provide no more role for the idea of a natural state
than our current physical theories do for the notion of absolute simul-
taneity.”30 To the extent that Evolutionary Psychology’s theory of a uni-
versal human nature relies on the Natural State Model for a distinction
between “normality” (which exemplifies human nature) and “abnormal-
ity” (which does not), its theory of human nature has no foundation in
biology.
The problems with the Natural State Model, however, are merely symp-
toms of deeper problems with essentialism itself. The distinction between
“normal” and “abnormal,” which characterizes the Natural State Model, is
necessary only if one is antecedently committed to the view that there are
certain characteristics that all and only humans share. For, since the claim
that there are characteristics that literally all and only humans share is an
obvious empirical falsehood, it becomes necessary to retreat to the less
robust claim that there are characteristics that all and only normal humans
share. But, if we are not driven to formulate our understanding of species
in terms of what all and only members of a species have in common, we
don’t need a category of “abnormal” to which to relegate the individuals
in a species that happen to lack one or more of the characteristics we take
to be essential to a species, and we then don’t need a category of “normal”
to contain the individuals that do happen to possess those characteristics.
“Human Nature” 439
happen to grow further up the tree will be the same branch, regardless of
whether it perfectly resembles the lower, terminated branch. This is the
significance behind the slogan “extinction is forever.” For species are not
individuated by their characteristics; they are individuated as segments
in the tree of life. If species were individuated by their characteristics, as
natural kinds are, then even if a species ceased to exist it could reemerge
later, provided that organisms evolved later that possessed the same char-
acteristics as those that had died earlier. Thus, again, as biologists under-
stand them, species don’t exhibit the features of natural kinds.
Third, species evolve. In fact, one and the same species may evolve so
significantly that characteristics that typify a species at one time period
cease to typify it at a later time, and another set of characteristics may
become typical of that species. If species were natural kinds, however, a
species could not undergo such significant change. A lineage undergoing
such significant change would have to be classified as one species before
the change and another species after it, since the different sets of typical
characteristics would constitute the essences of different species. By
analogy, given the right chemical intervention, a volume of carbon
monoxide could be transformed into carbon dioxide. But it would not be
the same kind of gas through the change. That is, the kind carbon monox-
ide itself wouldn’t become the kind carbon dioxide, but rather a volume
of gas would be transformed from an instance of the natural kind carbon
monoxide into an instance of the natural kind carbon dioxide. The natural
kinds themselves would remain unchanged. Similarly, if species were
natural kinds, a sufficient degree of evolution would simply transform a
species into another, distinct natural kind. But, as biologists understand
them, species can be radically overhauled by evolution, yet nonetheless
remain one and the same species. Provided that the evolutionary change
occurs within a single branch of the tree of life, the lineage is classified
as the same species, no matter how radical the evolutionary change.
Evolutionary change creates new species only if the change results in the
branching of a lineage (the reproductive isolation and splitting of two pop-
ulations). So, again, as biologists understand them, species don’t exhibit the
features of natural kinds.
Indeed, this last point generates something of a dilemma for the essen-
tialist view that species are natural kinds. Consider the dilemma with
respect to Evolutionary Psychology’s view of Homo sapiens. According to
Evolutionary Psychologists, there are “qualities that define us as a unique
species,” but these qualities evolved during our species’ history. Indeed,
as we saw in chapter 2, Evolutionary Psychologists maintain that our
“Human Nature” 443
a biological context this means that an organism never comes back into
existence once it is dead.”34 In this respect, individuals differ from kinds.
The individual members of a kind are located at particular regions of space-
time, but the kind itself has no particular location in spacetime. Further,
since kinds are constituted by their members, kinds are not discrete. The
same individuals can belong to more than one kind, in which case the
kinds to which they belong overlap rather than having discrete bound-
aries. Indeed, two different kinds can have precisely the same members, in
which case they overlap one another completely.
Second, each individual is spatiotemporally continuous. Each individual
exists continuously between its beginning and end in time, and at every
moment of its existence it occupies the same or contiguous regions
of space. Given its spatiotemporal continuity, an individual’s existence
can be plotted as a “spacetime worm,” a single unbroken line, however
squiggly, through the three dimensions of space and the fourth of time.
For example, we often identify an organism as the same organism solely
because of its spatiotemporal continuity, since in many cases the same indi-
vidual organism undergoes radical change over time. As Mayr points out,
“that caterpillar and butterfly are the same individual is inferred not from
any similarity in their appearance but from this continuity.”35 In this
respect, also, individuals differ from kinds or classes. A kind is not spa-
tiotemporally continuous, since a kind is constituted by its individual
members, and those members are frequently scattered across disparate
regions of spacetime. Indeed, kinds are potentially unlimited, in that
members of a kind can come into and go out of existence in remote reaches
of the universe at any time. Due to some bizarre chemical catastrophe, for
example, all water could cease to exist today, but tomorrow we could syn-
thesize more water in a lab. The kind water would thus not exhibit tem-
poral continuity. Similarly, even if the only water in existence today were
in Brazil, and the only water in existence tomorrow were in Scotland, the
Brazilian and Scottish substances would both be water despite the fact that
the kind water would not exhibit spatial contiguity. This is because all that
matters with respect to whether liquids are water is that they possess the
right chemical structure, and individual samples of liquid can share that
structure without being continuous with one another in time or contigu-
ous with one another in space.
Third, each individual is a cohesive whole. For example, although each
individual organism is composed of parts (organs, cells, and so on), and
can be broken down into its parts, those parts are not a mere collection,
but are organized and functionally integrated. Indeed, what makes the parts
446 Chapter 8
of an organism parts of that organism is the fact that they are function-
ally integrated with other parts of the organism, the fact that they con-
tribute to the organization that makes up that organism. The functional
integration of an organism’s parts consists in the fact that those parts
causally interact with one another, on a local level, in ways that help to
sustain the organism over time and in ways that they do not causally inter-
act with the parts of any other organism. In addition, the parts of an organ-
ism need not resemble one another in any respect in order to be parts of
the same organism and contribute to its functional organization. Your left
lung doesn’t resemble your right femur in any interesting respect, and they
don’t have to share any particular properties in order to be parts of your
body. In this respect, again, individuals differ from kinds. The individual
members of a kind are not members of that kind because they are func-
tionally integrated or organized in any particular fashion. Rather, individ-
uals are members of the same kind simply by virtue of their similarity to
one another.
As Mayr, Ghiselin, and Hull have shown, given the role that the species
concept plays in biological theory, species exhibit each of the three char-
acteristics definitive of individuals, just as organisms do. First, each species
is spatiotemporally localized, occupying the region of spacetime that is cir-
cumscribed by its temporal beginning and end and its spatial borders. More
important, each species has a definite location in the tree of life, a definite
segment of the tree, with a definite beginning and end. No two species can
occupy the same segment of the tree of life, and no one species can occupy
two distinct segments. For, as we have seen, when a species goes extinct,
numerically the same species cannot come into existence later. Even if
other, identical organisms were to come into existence later, they would
be classified by biologists as a new species, not as a continuation of the
earlier species. Species, then, are spatiotemporally localized and discrete.
Second, each species is spatiotemporally continuous. Each species exists
continuously from its temporal beginning to its end, and each species as
a whole is spread over the same or contiguous regions of space for every
moment of its existence. In this respect, like an organism, a species’ exis-
tence can be plotted as a “spacetime worm.” Further, as Hull points out,
the organisms that make up a species are related by descent. “But descent
presupposes replication and reproduction, and these processes in turn pre-
suppose spatiotemporal proximity and continuity. When a single gene
undergoes replication to produce two new genes, or a single cell undergoes
mitotic division to produce two new cells, the end products are spa-
tiotemporally continuous with the parent entity. In sexual reproduction,
“Human Nature” 447
the propagules, if not the parent organisms themselves, must come into
contact. The end result is the successive modification of the same popula-
tion.”36 Thus, species are spatiotemporally continuous.
Third, species are unified, cohesive wholes, held together by the organi-
zational glue of reproduction. For species consist of interbreeding po-
pulations, and both individual populations and groups of interbreeding
populations are united by the reproductive interactions of organisms. As Mayr
points out, this is due to the fact that the organisms that compose a species
develop “from the joint gene pool of the species, and that they jointly con-
tribute their genotypes to form the gene pool of the next generation.”37
The contribution of genotypes to the next generation, however, involves
a great many causal interactions among organisms. The organisms in a
population must structure a great many of their activities around the
pursuit of sex with conspecifics, the act of sex with conspecifics, the incu-
bation or gestation of the embryonic products of sex, and the care and pro-
tection of live offspring. These causal interactions on a local level between
the organisms involved in reproductive activities produce a cohesiveness
within populations and species that is much like the functional organiza-
tion of an organism (which derives from local causal interactions between
its parts). Thus, species are unified, cohesive wholes.
Species, then, exhibit all the properties that are definitive of individuals.
But, if species are individuals, just like organisms, how are we to under-
stand the relation between organisms and species? According to essenti-
alism, the only individuals are organisms, and species, as natural kinds, are
classes of individuals that are united by a shared set of essential proper-
ties. Organisms are thus members of the classes that are their species. In
this respect, essentialists see the relation between organisms and species as
just like the relation between organisms and higher taxa such as orders
and phyla. In the essentialist’s view, higher taxa are also classes of the same
individuals that are members of species, but those individuals are united
in orders, and so on, by sharing increasingly more inclusive sets of essen-
tial properties. In the view that species are individuals, however, organisms
are parts of species in precisely the way that cells are parts of organisms.
In other words, organisms compose a species in precisely the way that cells
compose a body.
The parallel between cell/organism and organism/species is worth bela-
boring for a moment. Cells are clearly individuals: They are spatiotempo-
rally localized (discrete), spatiotemporally continuous, and cohesive. Yet
these individuals are unproblematically parts of another, larger individual
(an organism). But what makes the cells in an organism all parts of the
448 Chapter 8
same, larger individual? It is not shared properties that makes cells all parts
of the same organism. The cells in your body, for example, aren’t cells of
the same body because they have the same genetic makeup. For, in fact,
many of them don’t. In the process of mitosis, which created all the cells
in your body, mutations occur. As a result, there are genetic differences
among many of the cells in your body. They are, nonetheless, all cells of
your body. Conversely, the cells in the bodies of identical (monozygotic)
twins are genetically identical, with the exception of the cells in each twin
that contain mutations. But two genetically identical cells from the bodies
of two twins are not cells of the same body, despite their genetic identity.
So, the genetic makeup of a cell, and its genetic similarity to other cells, is
not what determines which body a cell belongs to. Rather, the cells in your
body are cells of your body because they satisfy two conditions. First, they
all descended, via iterated rounds of cell division, from the same zygote.
For every cell in your body, there is an unbroken chain of descent via cell
division that links it with the same zygote. And, second, those cells that
are parts of your body are so because they are causally integrated into the
overall organization that makes up your body.
In the same way, organisms that belong to the same species need not
share any properties. Sharing properties is not what determines whether
two organisms belong to the same species, even if those organisms do share
a significant number of properties. In fact, in many cases, organisms that
belong to the same species do not resemble one another much at all. In
chapter 1, we encountered Paracerceis sculpta, a species in which males
come in three “morphs” that pursue different mating strategies. Large
males are many times the size of small males, and they possess spiked
“horns” where small males have only little nubs. Judging by shared prop-
erties, the two would be classified as different species, yet they belong to
the same species. In addition, in some species in which developmental
plasticity is common, individual organisms develop to mimic the appear-
ance of other species. In such cases, different organisms in the species can
develop to mimic distinct species, thereby having more observable char-
acteristics in common with those other species than with one another.
Thus, similarity is only incidental to belonging to the same species; it is not
a criterion of it.
Indeed, not only need there be no shared properties among the organ-
isms in a species, but the fact that species are reproductively organized
individuals ensures the maintenance of variation among the organisms in
a species. For, in meiosis, the early stage of sexual reproduction, gametes
are created that contain only half of an organism’s genes, and two gametes
“Human Nature” 449
are homologous. Indeed, all of the traits that you and I share and that are
described in “precise anatomical detail” by Gray’s Anatomy are homologies.
Homologous traits, however, are not classified together by virtue of
shared characteristics, let alone by virtue of shared essential characteristics.
The human brain and the rat brain are homologous despite many struc-
tural differences, and the hind limbs of the crocodile and those of the star-
ling are homologous despite sharing virtually no interesting properties.
The same is true of homologous traits within species. The eyes of each indi-
vidual human are not human eyes because they share properties essential
to being a human eye, but because they are homologies, traits derived from
an equivalent eye in a common ancestor. Indeed, “deformed” eyes, which
lack some of the properties of eyes detailed in Gray’s Anatomy, are nonethe-
less eyes. And the eyes of the blind are human eyes despite not perform-
ing the typical visual function of eyes. Further, male nipples and females
nipples are all nipples because they are homologous traits, not because of
shared morphological or functional properties (which, in fact, they do not
share). This is because, as Wagner says, “homology is assessed regardless
of shape or function.”45 In fact, homology is assessed in precisely the way
that the species classification of two organisms is assessed—genealogically.
Traits of two organisms are homologies if they were derived from an equiv-
alent trait in a common ancestor, regardless of whether they share prop-
erties, just as two organisms belong to the same species if they descended
from a common ancestor in that species, regardless of whether they share
properties. In short, homologies, like the organisms of a species, are unified
by descent, not by shared properties.
Two individual instances of a trait (in two distinct organisms), then, are
not classed together as homologous by the same logic as two samples of
platinum are classed together as the same substance. Instances of natural
kinds, like platinum, are classed together because of their intrinsic prop-
erties, regardless of their provenance. If we froze the universe at a partic-
ular moment of time, for example, we could identify every instance of
platinum simply by determining whether objects were composed of atoms
with atomic number 78. But, in that frozen instant, we would not be able
to identify every instance of a particular homology. For history is every-
thing with respect to determining whether two individual instances of a
trait are homologous. Your eyes and my eyes are homologous (“the same”)
not because of properties they share at this instant, but because of chains
of descent that reach back from each of us into the past and converge
upon a common ancestor. Our eyes are not “the same” because they are
connected by common properties at this moment, but because they are
456 Chapter 8
achieve the status of laws of nature, since laws of nature apply only to
instances of natural kinds. Insofar as psychology concerns itself with dis-
tinctively human cognition and emotion, it must begin to conceive of itself
as being in the business of providing descriptions of homologous characteris-
tics, rather than being in the business of providing laws of thought in the
way that physics provides laws of mechanics or chemistry provides laws
of chemical bonding.
To conclude, then, since Homo sapiens is an individual, not a natural
kind, there is no such thing as human nature. And, since human psycho-
logical mechanisms are homologies, human psychological mechanisms do
not form natural kinds. Consequently, there are no laws of nature that
pertain exclusively to human minds, so Evolutionary Psychology can never
fulfill its promise to be the “new science of human nature” by discovering
the psychological laws that govern the functioning of evolved psycholog-
ical mechanisms. A truly evolutionary science of human psychology will
not only abandon the quest for human nature, but, with it, the quest to
be a science in the model of physics or chemistry.
“Human Universals”
they are cultural universals (that is, phenomena that are present in every
culture), not psychological universals. Although every culture may prac-
tice incest avoidance, obviously not every individual human is psycho-
logically designed to do so. And things like division of labor and oligarchy
are, by their very nature, properties of groups and not of individual
humans.
Pinker’s appeal to cultural universals to support an argument for
psychological universals no doubt stems from Evolutionary Psychology’s
attempt to synthesize aspects of anthropology, which deals largely with
cultures or populations, with aspects of psychology, which deals largely
with the properties of individual human minds. Indeed, Brown is an
anthropologist, and his intent in compiling his list of universals was
to capture what “all societies, all cultures, and all languages have in
common.”46 But when anthropological and psychological agendas are in
play simultaneously, it becomes easy to shift imperceptibly from talk of
cultural universals to talk of psychological universals. And Evolutionary
Psychologists’ arguments for the existence of a universal human nature
often trade on this ambiguity.
The ambiguity is at its most misleading when Evolutionary Psycholo-
gists cite cultural universals as evidence of psychological universals, as
Pinker does in The Blank Slate. Sometimes the purported cultural univer-
sals cited as evidence of psychological universals are not well documented
cultural universals. For example, as we saw in chapter 5, Evolutionary Psy-
chologists were quick to claim, on the basis of a single study in the United
States, that a male preference for females with a 0.70 waist-to-hip ratio is
culturally universal. But, when the anthropological field work was actually
done to find out whether this preference truly is a cultural universal, it was
discovered that it isn’t. At other times, however, Evolutionary Psycholo-
gists cite truly well documented cultural universals as evidence of the
existence of psychological universals. But, even if there are many cultural
universals, as I believe there are, cultural universals do not actually provide
evidence of a universal human nature when that is understood as con-
sisting of psychological universals.
To see why, consider a prominent example of how (putative) cultural
universals are interpreted as evidence of psychological universals. As we
saw in chapter 5, Buss hypothesizes that selection designed the male mind
to prefer nubility in a mate and the female mind to prefer status, and he
argues that these preferences should consequently be universal among
individual men and women. To test these hypotheses, Buss conducted
mate-preference surveys in thirty-three countries. But, in these surveys,
“Human Nature” 459
Buss did not find that every individual man and woman in every culture
around the world possessed the hypothesized mate preferences. In fact,
Buss’s surveys revealed significant numbers of individual men and women
whose mate preferences did not conform to his hypotheses. What Buss
actually found was that the averages of individual survey responses in
(almost) every culture conformed reasonably closely to the hypothesized
responses. That is, Buss found that the average male in (almost) every culture
expressed a preference for a younger mate and that the average female
in (almost) every culture expressed a preference for a mate with a higher
socioeconomic status. Although Buss interpreted these results as providing
confirmation for his hypotheses, averages are actually properties of popu-
lations, not properties of individual humans. In other words, even if they
are taken at face value, what Buss’s survey results show is not that partic-
ular preferences are psychological universals, but that particular average
preferences are cultural universals. Taking those survey results as evidence of
a universal human nature involves inferring psychological universals from
cultural universals, which is a fallacy analogous to that of inferring that
every human is five feet, eight inches tall on the basis of discovering that
the average height in every culture is five feet, eight inches.
While the inference from cultural universals to psychological universals
(hence human nature) is indeed fallacious, in the hands of Evolutionary
Psychologists the inference isn’t quite the bald confusion that my height
analogy makes it out to be. For, in the hands of Evolutionary Psycholo-
gists, the inference from cultural universals to psychological universals is
aided and abetted by Tooby and Cosmides’ theory of culture and the
origins of cultural universals. To properly appraise the role of cultural uni-
versals in Evolutionary Psychologists’ discussions of human nature, then,
we should examine this theory of culture.
According to Tooby and Cosmides, culture consists of “any mental,
behavioral, or material commonalities shared across individuals, from
those that are shared across the entire species down to the limiting case of
those shared only by a dyad.”47 Tooby and Cosmides claim that cultural
contents—specific mental, behavioral, or material commonalities—can be
generated by one of three causal mechanisms.
First, Tooby and Cosmides argue that our (allegedly) universal psycho-
logical adaptations “not only constitute regularities in themselves but
they impose within and across cultures all kinds of regularities on human
life, as do the common features of the environments we inhabit.”48 For
example, since every (“normal”) human’s mind is “preequipped” with the
same presuppositions and expectations about the mental functioning and
460 Chapter 8
Third, Tooby and Cosmides argue, some cultural contents have their
origins within the mind of a single individual, who then transmits to
others a novel idea or behavior. Since these cultural contents originate by
spreading from one individual to others, like a contagion, Tooby and Cos-
mides call such cultural contents epidemiological culture. As Tooby and Cos-
mides put it: “This subset of cultural phenomena is restricted to (1) those
representations or regulatory elements that exist originally in at least one
mind that (2) come to exist in other minds because (3) observation and
interaction between the source and the observer cause inferential mecha-
nisms in the observer to recreate the representations or regulatory elements
in his or her own psychological architecture.”50 For example, if someone
comes up with a new idea for how to catch fish, and that novel idea proves
an effective method for catching fish, that individual’s interaction with
others can result in their adopting the same method of catching fish, and
this new idea can begin to spread in the individual’s population. Epi-
demiological culture thus encompasses such phenomena as baseball, bell-
bottoms, the hokey-pokey, and television.
Interestingly, once an aspect of epidemiological culture becomes wide-
spread in a population, it becomes an aspect of the local conditions to
which that population responds. This change in local conditions, in turn,
precipitates a change in the mental and behavioral responses that are
“evoked” from the psychological mechanisms in the members of that
population. Thus, epidemiological culture can feed into, and precipitate
changes in, evoked culture. For example, once radio or television becomes
a stable aspect of the cultural landscape, it further transforms culture by
evoking novel reactions from evolved psychological mechanisms. Simi-
larly, Evolutionary Psychologists would argue, although pornographic
magazines and videos originated through epidemiological culture, once
they became a fixture of the cultural landscape they became an aspect of
the local conditions to which new generations responded. The patterns of
consumption and use of pornographic materials by new generations thus
became aspects of evoked culture, as evolved sexual responses encountered
a sexual environment populated not only by fellow humans, but by a
wealth of pornographic representations of fellow humans as well.
The above theory of culture underwrites the frequent inference by Evo-
lutionary Psychologists from cultural universals to psychological univer-
sals. For, according to this theory of culture, cultural universals are aspects
of metaculture, which is generated by interactions among psychological
universals and interactions between psychological universals and universal
462 Chapter 8
ening displays, but never attack, and withdraw when attacked. Given
the fitness payoffs in the game we examined, a stable polymorphism
consisting of 75 percent Hawks and 25 percent Doves would evolve in a
population.
Now, suppose that we took a simple statistical survey of a population
consisting of 75 percent Hawks and 25 percent Doves. Suppose, in partic-
ular, that we asked the individuals in the population which tactic they
employed in contests over resources, and respondents were to answer
with a 2 if they played Hawk and a 1 if they played Dove. We would find
the average response to be 1.75, an average that is very much toward the
Hawkish end of the scale. There are two points to consider regarding this
example.
First, suppose that prior to our survey we had reasoned that, given the
fitness payoff of winning a resource, selection should have favored aggres-
sive pursuit of resources. After all, we could have reasoned, there are only
individuals who win resources and those who don’t. Those who win
resources enjoy enhanced fitness, while those who don’t win resources
have nothing. So, we could have argued, selection should have favored
those who aggressively pursued resources, since they would have had
higher average fitness than those who didn’t. Thus, we would have pre-
dicted, individuals should have evolved to pursue the Hawk strategy. When
we then looked at the statistical results of our survey, and found the
average response to be far closer to Hawk than Dove, we could have con-
vinced ourselves that our hypotheses about the past action of selection and
the present natures of the individuals in the population were confirmed.
In fact, however, our hypothesis about the past action of selection would
be mistaken, since there had actually been frequency-dependent selection
for a stable polymorphism of Hawks and Doves, rather than selection for
Hawks. Our hypothesis about the present natures of the individuals in the
population would also be mistaken, since 25 percent of the individuals in
the population are Doves “by nature.” In other words, the Doves aren’t
simply “abnormal” Hawks, but were actually selected to be Doves. The fact
that the results of our survey appeared to confirm our hypotheses was a
spurious consequence of our taking the statistical properties of the popu-
lation as evidence for the phenotypic properties of the individuals in the
population.
Second, suppose that Hawk and Dove are the only available strategies to
pursue in competitions for a resource. Suppose further that the resource
for which individuals compete is the same in every population, as is the
fitness payoff of acquiring that resource. Given these suppositions, the
464 Chapter 8
if the fitness payoff of winning the resource were higher or lower, selec-
tion would produce and maintain a different ratio of Hawks to Doves. This
doesn’t mean that the resource for which there is competition must be the
same in every population, but only that the fitness benefit of winning a
resource must be the same. Second, in every population, individuals would
have to compete for a resource by “choosing” a competitive strategy from
the same set of alternative strategies. For example, frequency-dependent
selection could produce the same stable polymorphism of Hawks and
Doves in every population only if Hawk and Dove were among the com-
petitive strategies from which the individuals in every population had
to “choose.” Third, the fitness costs associated with each of the available
strategies would have to be the same in every population. For example, if
Hawk were a costlier strategy in one population than in another, the payoff
for playing Hawk would be different in those two populations, and the
evolutionarily stable ratio of Hawks to Doves would differ accordingly.
These conditions would have to be met in order for precisely the same
ratio of alternative strategies to be evolutionarily stable in different popu-
lations. But, as long as individuals in different populations “choose” from
among the same set of alternative strategies in competing for resources,
different populations could have similar stable ratios even if the fitness
costs and benefits of the resources and the competing strategies differed
somewhat. To illustrate, consider again the Hawk-Dove game. If the fitness
value of the resource for which Hawks and Doves compete varied from
population to population, the stable ratio of Hawks to Doves would vary
from population to population as well. If we conducted surveys of the com-
petitive strategies in each population, where Hawk is 2 and Dove is 1 as
described above, we might find that population averages ranged from,
say, 1.62 to 1.84. Each of these populations would still be predominantly
Hawkish, so that Hawkishness would still be a cultural universal (that is,
in every culture aggressiveness would dominate competitive interpersonal
interactions). Indeed, we might find that the population averages all clus-
tered around a central tendency, so that despite some variation in stable
ratios across populations there would be a discernible overall tendency of
all populations toward a specific degree of Hawkishness. In such a case,
we would detect some degree of cultural variation in competitive inter-
personal interactions, but Hawkishness would still be a fairly robust
cultural universal.
The real question, of course, is whether the mechanism I have just
described underlies any human cultural universals. To that question, I
think, the answer is a resounding “maybe.” For example, individuals in
466 Chapter 8
every culture compete to mate with members of the opposite sex, and the
fitness-benefit structure of matings with members of the opposite sex is
likely very similar across cultures. In addition, the set of behavioral strate-
gies that individuals can employ to attract and retain mates is probably
pretty much the same for everyone; in every culture, individuals probably
have the same small set of tactics available to them to woo and keep
members of the opposite sex. Thus, mating behavior, which includes mate
preferences, is an aspect of human life with respect to which it is quite
possible that selection would produce and maintain similar stable poly-
morphisms across cultures. Indeed, as I argued in chapter 5, mate prefer-
ences probably are polymorphic. But, polymorphic male and female mate
preferences are nonetheless compatible with average population prefer-
ences that cluster around a central mate-preference tendency (which is
actually what Buss claims to have found).
Bear in mind that it need not be the case that the resources for which
individuals compete in every culture be precisely the same; it need only
be the case that individuals “choose” from the same set of strategies to
compete for resources of equal fitness value. In addition, the fitness costs
of the strategies from which individuals “choose” need not be identical
across cultures in order for the population averages from all cultures to
cluster closely around a central tendency. Thus, cultural universals could
be generated by psychological polymorphisms, provided that individuals
across cultures competed for any resources of roughly equal fitness value
by “choosing” from the same set of strategies, which could vary somewhat
in their fitness costs across cultures. Given the centrality of resource acqui-
sition to survival and reproduction, and given that all humans require
some of the same resources in order to survive and reproduce (where
members of the opposite sex are reproductive resources), it is a distinct pos-
sibility that frequency-dependent selection has produced or maintained
some similar stable polymorphisms across cultures. Such polymorphisms
would entail psychological variation within cultures, but they would
nonetheless generate cross-cultural universals. Consequently, cultural
universals are not actually evidence of psychological universals. For each
cultural universal discovered, only additional empirical investigation can
reveal whether that universal is generated by psychological universals or
by a stable polymorphism of psychological variation. Whenever Evolu-
tionary Psychologists infer the existence of psychological universals from
the discovery of a cultural universal, they are substituting armchair rea-
soning for the necessary empirical investigation.
“Human Nature” 467
As the anthropologist William Durham points out, the incest taboo varies
widely from culture to culture with respect to the degree of relatedness of
the individuals between whom sex is taboo. In some cultures, only sex
between siblings is taboo, whereas in other cultures the taboo extends to
sex between fifth cousins. Moreover, Durham has shown that this varia-
tion is not explained by variation in the likelihood of intimate childhood
association of relatives between whom sex is taboo. For example, first
cousins are no more likely to be reared together in cultures where sex
between first cousins is taboo than in cultures where sex between first
cousins is not taboo. Since an inhibition of sexual desire for those with
whom one was reared can only work to avert sex with those with whom
one was reared, even the universality of such a mechanism can’t explain
why there would be taboos against sex with relatives with whom children
are consistently not reared. Thus, the incest taboo as a cultural phenomenon
can’t simply be generated by an evolved aversion to sex with intimate
childhood associates. While that psychological mechanism may be per-
fectly real, and while it may also be an adaptation, it is simply insufficient
to account for the extent and complexities of the incest taboo as a cultural
phenomenon.
Durham argues, instead, that in all cultures the incest taboo is based on
minimizing the sum of the costs of inbreeding and the costs of outbreed-
ing. Durham argues that in each culture there is a recognition of the
deleterious effects of inbreeding, so that every culture recognizes certain
reproductive costs associated with inbreeding. In each culture, however,
there are also social and reproductive costs involved in outbreeding. These
costs vary from culture to culture as a function of such things as degree of
access to unrelated mates. As a result, the sum of the costs of inbreeding
and the costs of outbreeding varies from culture to culture. Where the costs
of outbreeding are low, the sum will be low, and the incest taboo will
extend to include a wider range of relatives, such as third and fourth
cousins. Where the costs of outbreeding are high, the sum will also be high,
and the incest taboo will be restricted to siblings or first cousins. Not only
did Durham find that this pattern held in a study of sixty cultures in the
Human Relations Area Files, but he found that all sixty of these cultures
were related by descent. Thus, Durham concludes, the evidence suggests
“that all existing incest taboos are related by descent to one, or a few,
ancestral prohibitions,” so that existing incest taboos “may all be cultural
homologs—locally refined variations on the same ancestral theme.”52
Evolutionary Psychologists also point out that myths are culturally
universal, and they argue that myth is an expression of the functioning of
470 Chapter 8
populations diverged and spread across the globe. As a result, cultural uni-
versals do not provide evidence of a universal human nature, if human
nature is understood to consist of universal psychological adaptations,
which is how Evolutionary Psychologists understand it. There are several
processes that could have produced any given cultural universal, and only
extensive, and often very difficult, empirical investigation can reveal how
a cultural universal was produced and whether it has any connection to
evolved psychological mechanisms, let alone universal psychological adap-
tations. In the absence of such empirical investigation, the existence of
cultural universals provides no evidence for Evolutionary Psychology’s
claims about the existence of psychological universals.
intentions in creating the universe and the life forms within it. In short,
complex design is what God created.
For the majority of educated Westerners, the argument from design, in
one form or another, provided the intellectual framework for understand-
ing nature throughout the better part of the nineteenth century. More
important, the argument from design set the terms of the debate between
creationists and naturalists, those who believed that all of nature came to
be through the operation of laws of nature, without any intervention by
an intelligent being. For the argument from design posed the challenge
that any naturalistic theory had to overcome: Provide a convincing expla-
nation of how complex design could have emerged without the interven-
tion of an intelligent being. Thus, throughout the nineteenth century, the
problem that naturalistic theories had to solve was the problem of complex
design.
This was the intellectual context within which Darwin developed his
theory of evolution. Indeed, Darwin had read Paley and had been
impressed by Paley’s argument from design. As a result, the problem of
complex design—the problem of what Darwin called “organs of extreme
perfection and complication”—became Darwin’s own litmus test for his
theory of evolution. In an extended argument, Darwin took on Paley’s own
example of the eye in order to demonstrate that the process of natural
selection was capable of creating “organs of extreme perfection and com-
plication.” The process by which it does so, as we saw in chapter 1, con-
sists of iterated rounds of modification to a preexisting trait followed by
retention of modifications that prove beneficial and elimination of those
that prove detrimental. Darwin thereby provided a naturalistic solution to
the problem of complex design, meeting Paley’s challenge, and his theory
of evolution accordingly became a viable explanation of the origins of life.
In fact, a forceful demonstration of how decisively Darwin’s theory of
natural selection answered Paley’s challenge is provided by the evolution-
ary biologist Richard Dawkins in his book The Blind Watchmaker, the title
of which is an allusion to Paley’s argument.
But, while natural selection was the mechanism that met Paley’s chal-
lenge, there has always been much more to evolutionary theory than
explaining how “organs of extreme perfection and complication” arose
by natural selection. For one thing, the process of selection itself doesn’t
result only in complex adaptations. Selection also eliminates traits from
populations and, arguably, eliminates entire groups or populations. Since
Darwin’s time, it has also become clear that selection can sometimes
prevent a population from becoming optimally adapted to its environ-
“Human Nature” 475
ment. For another thing, there is much more to evolution than natural
selection. Selection is just one force among many that drive the process of
evolution, and these forces are all implicated in having generated the diver-
sity of life forms on earth. Further, all of these evolutionary forces can affect
the form and characteristics of organisms as well as the diversity of life.
Explaining the diversity of life, past extinctions, and both adaptive and
nonadaptive evolution within populations are all among the problems
with which evolutionary theory deals. Thus, there are actually numerous
problems concerning organic form and the diversity of life to which evo-
lutionary theory provides, and aspires to provide, solutions.
Given that the problem of complex design is but one among many prob-
lems in evolutionary theory, why should it be accorded a central status?
The answer, of course, is that it shouldn’t be accorded a central status,
because it doesn’t occupy a central place within the array of problems that
evolutionary theory addresses. This can be seen by examining any intro-
ductory textbook in evolutionary biology. The problem of complex design
was actually Paley’s problem. It was the problem that nineteenth-century
theologians used to challenge naturalistic accounts of the origins and
complexity of life, and they chose that problem because they thought it
to be unsolvable by naturalistic theories. There is nothing in the nature of
things that dictates that the problem of complex design is central to under-
standing life on earth. There is nothing in the nature of things that man-
dates that we should consider explaining complex design to be more
important than explaining, say, the Cambrian explosion, the unprece-
dented and since unequaled proliferation of species between 535 and 525
million years ago. Rather, a focus on the problem of complex design reflects
the explanatory interests of Paley and the nineteenth-century natural
theologians. Thus, to see natural selection, and the creation of complex
design, as the centerpiece of evolutionary theory is to retain the problem
that was created by Paley and the theologians and merely to replace their
solution to the problem with a naturalistic solution. That is, to retain a
focus on the problem of complex design is to adopt naturalism, but only
within the framework of natural theology. It is to view the significance
of evolutionary theory through a lens that was ground by natural
theologians.
But evolutionary theory has triumphed over Paley and the natural the-
ologians. No one committed to allowing the evidence to decide between
evolutionary theory and natural theology can see creationism as a viable
alternative to evolutionary theory. And once evolutionary theory is
accepted as an account of the origins and diversity of life, there is no longer
476 Chapter 8
any reason to draw our problems from the framework of natural theology
to which Darwin responded in the middle of the nineteenth century. There
is no longer any reason to view evolutionary theory through the lens of a
preevolutionary, theological worldview. Indeed, to truly accept evolution-
ary theory as an account of the diversity of life is to allow evolutionary
theory itself to determine which are the interesting and important prob-
lems about organic form and the diversity of life. However, when evolu-
tionary theory itself is taken as the source of problems about the nature of
life, rather than as simply the source of solutions to preevolutionary prob-
lems, the problem of adaptation becomes merely one among, and on a par
with, many problems with which evolutionary theory deals. Once we take
evolutionary theory on its own terms, rather than on Paley’s terms, we can
see that there is no sense in which evolution is all about adaptation in the
way that God’s creation was all about complex design.
Thus, to see adaptations as central to human “nature” in a way that
nonadaptations are not—to see human “nature” as consisting exclusively
of adaptations—is to view the human organism through the theoretical
prism of natural theology. It is to replace God with Natural Selection as
the Creator, but to still maintain that the Creator’s “intention,” as mani-
fested in what was selected for, represents the “nature” of our species,
departure from which is “abnormal.” But this exclusive focus on adapta-
tions in Evolutionary Psychology’s theory of human nature does not derive
from evolutionary theory itself, and it receives no justification from
evolutionary theory. Rather, this focus is a preevolutionary interpretation
of the human organism; it is a theoretical vestige of natural theology.
Consequently, Evolutionary Psychology’s adaptation-centered theory of
human nature is in no true sense evolutionary.
In order to forestall any potential misunderstanding, let me be very clear
about what I am claiming. I am not claiming that humans possess no psy-
chological adaptations, and I am not claiming (as we saw Gould claim) that
human psychological adaptations constitute a small and insignificant part
of human psychology. Rather, I am claiming only that there is no basis in
evolutionary theory for maintaining that psychological adaptations are con-
stitutive of human “nature” in a way that psychological traits that aren’t
adaptations are not. There undoubtedly are adaptations in every species,
but adaptations do not represent the “order of nature” or the raison d’être
of evolution. The belief that they do is an imposition upon evolutionary
theory that derives from a preevolutionary worldview.
Let’s turn now to the second respect in which Evolutionary Psychology’s
theory of human nature represents a distortion of a truly evolutionary view
“Human Nature” 477
Introduction
1. Beynon 1994.
4. Gould 1997.
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
16. Kenrick, Keefe, Gabrielidis, and Cornelius 1996, p. 1506; emphasis added.
17. Kenrick, Keefe, Gabrielidis, and Cornelius 1996, p. 1505; emphasis added.
36. Umberson and Hughes 1987; Hamermesh and Biddle 1994; Mulford, Orbell,
Shatto, and Stockard 1998.
Chapter 6
26. Adams, Gold, and Burt 1978; Hill 1988; Stanislaw and Rice 1988; Regan 1996.
40. Data from Buss, Larsen, Westen, and Semmelroth 1992; Geary, Rumsey,
Bow-Thomas, and Hoard 1995; Buunk, Angleitner, Oubaid, and Buss 1996; DeSteno
and Salovey 1996a; Buss et al. 1999; Pietrzak, Laird, Stevens, and Thompson 2002.
42. Data from Buss, Larsen, Westen, and Semmelroth 1992; Buunk, Angleitner,
Oubaid, and Buss 1996; Harris and Christenfeld 1996a; Buss et al. 1999.
46. Data from Buss et al. 1999; Wiederman and Kendall 1999.
Chapter 7
Chapter 8
1. Ghiselin 1997, p. 1.
3. Loptson 1995, p. 1.
5. Trigg 1988, p. 4.
8. Bonner 1980, p. 9.
Introduction
Barrett, Louise, Robin Dunbar, and John Lycett (2002). Human Evolutionary Psychol-
ogy. Princeton: Princeton University Press.
Beynon, Mike, producer (1994). Desmond Morris’ The Human Animal: The Biology of
Love (television program). London: BBC.
Buss, David M. (1994). The Evolution of Desire: Strategies of Human Mating. New York:
Basic Books.
Buss, David M. (1999). Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn and Bacon.
Buss, David M. (2000). The Dangerous Passion: Why Jealousy Is as Necessary as Love
and Sex. New York: Free Press.
Daly, Martin, and Margo Wilson (1999). The Truth about Cinderella: A Darwinian View
of Parental Love. New Haven, CT: Yale University Press.
Dawkins, Richard (1989). The Selfish Gene (2nd ed.). Oxford: Oxford University Press.
Gould, Stephen Jay (1997). Evolution: The Pleasures of Pluralism. New York Review
of Books 44(11): 47–52.
Kennair, Leif Edward Ottesen (2003). Essay Review: An Alternative Paradigm After
All? Human Nature Review 3: 24–35.
498 Bibliography
Kuhn, Thomas S. (1996). The Structure of Scientific Revolutions (3rd ed.). Chicago: Uni-
versity of Chicago Press.
Maynard Smith, John (1995). Genes, Memes, and Minds. New York Review of Books
42(19): 46.
Pinker, Steven (1997). How the Mind Works. New York: Norton.
Pinker, Steven (2002). The Blank Slate: The Modern Denial of Human Nature. New York:
Viking.
Rose, Hilary, and Steven Rose (2000). Introduction. In H. Rose and S. Rose (eds.),
Alas, Poor Darwin: Arguments against Evolutionary Psychology (pp. 1–13). London:
Jonathan Cape.
Thornhill, Randy, and Craig T. Palmer (2000). A Natural History of Rape: Biological
Bases of Sexual Coercion. Cambridge, MA: MIT Press.
Wright, Robert (1994). The Moral Animal: Evolutionary Psychology and Everyday Life.
New York: Vintage.
Chapter 1
Alcock, John (1993). Animal Behavior: An Evolutionary Approach (5th ed.). Sunder-
land, MA: Sinauer Associates.
Bergstrom, Carl T., and Peter Godfrey-Smith (1998). On the Evolution of Behavioral
Heterogeneity in Individuals and Populations. Biology and Philosophy 13: 205–231.
Darwin, Charles (1859/1964). On the Origin of Species. Cambridge, MA: Harvard Uni-
versity Press.
Dawkins, Richard (1989). The Selfish Gene (2nd ed.). Oxford: Oxford University Press.
Endler, John A. (1986). Natural Selection in the Wild. Princeton: Princeton University
Press.
Futuyma, Douglas J. (1998). Evolutionary Biology (3rd ed.). Sunderland, MA: Sinauer
Associates.
Godfrey-Smith, Peter (1998). Complexity and the Function of Mind in Nature. New York:
Cambridge University Press.
Bibliography 499
Gross, Mart R., and Joe Repka (1998). Stability with Inheritance in the Conditional
Strategy. Journal of Theoretical Biology 192: 445–453.
Hofmann, Hans A., Mark E. Benson, and Russell D. Fernald (1999). Social Status Reg-
ulates Growth Rate: Consequences for Life-History Strategies. Proceedings of the
National Academy of Sciences 96: 14171–14176.
Kaplan, Jonathan, and Massimo Pigliucci (2001). Genes “for” Phenotypes: A Modern
History View. Biology and Philosophy 16: 189–213.
Lewontin, Richard (1995). Human Diversity. New York: Scientific American Library.
Maynard Smith, John (1976). Evolution and the Theory of Games. American Scien-
tist 64: 41–45.
Maynard Smith, John (1982). Evolution and the Theory of Games. New York:
Cambridge University Press.
Maynard Smith, John, and G. A. Parker (1976). The Logic of Asymmetric Contests.
Animal Behaviour 24: 159–175.
Mayr, Ernst (1988). Cause and Effect in Biology. In Toward a New Philosophy of
Biology: Observations of an Evolutionist (pp. 24–37). Cambridge, MA: Harvard Uni-
versity Press.
Ridley, Mark (1996). Evolution (2nd ed.). Cambridge, MA: Blackwell Science.
Shuster, Stephen M. (1992). The Reproductive Behavior of a-, b-, and g-male Morphs
in Paracerceis sculpta, a Marine Isopod Crustacean. Behaviour 121(3–4): 231–258.
500 Bibliography
Shuster, Stephen M., and M. J. Wade (1991). Equal Mating Success among Male
Reproductive Strategies in a Marine Isopod. Nature 350: 608–610.
Sober, Elliott (1984). The Nature of Selection. Cambridge, MA: MIT Press.
Sterelny, Kim, and Philip Kitcher (1988). The Return of the Gene. Journal of Philos-
ophy 85: 339–361.
West-Eberhard, Mary Jane (1989). Phenotypic Plasticity and the Origins of Diver-
sity. Annual Review of Ecology and Systematics 20: 249–278.
Chapter 2
Cosmides, Leda, and John Tooby (1987). From Evolution to Behavior: Evolutionary
Psychology as the Missing Link. In J. Dupré (ed.), The Latest on the Best: Essays on
Evolution and Optimality (pp. 277–306). Cambridge, MA: MIT Press.
Cosmides, Leda, and John Tooby (1994). Origins of Domain Specificity: The Evolu-
tion of Functional Organization. In L. A. Hirschfeld and S. A. Gelman (eds.), Mapping
the Mind: Domain Specificity in Cognition and Culture (pp. 85–116). New York:
Cambridge University Press.
Cosmides, Leda, and John Tooby (1997). The Modular Nature of Human Intelli-
gence. In A. B. Scheibel and J. W. Schopf (eds.), The Origin and Evolution of Intelli-
gence (pp. 71–101). Sudbury, MA: Jones and Bartlett.
Cosmides, Leda, John Tooby, and Jerome H. Barkow (1992). Introduction: Evolu-
tionary Psychology and Conceptual Integration. In J. H. Barkow, L. Cosmides, and
J. Tooby (eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture
(pp. 3–15). New York: Oxford University Press.
Crook, John H., and Stamati J. Crook (1988). Tibetan Polyandry: Problems of Adap-
tation and Fitness. In L. Betzig, M. Borgerhoff Mulder, and P. Turke (eds.), Human
Reproductive Behaviour: A Darwinian Perspective (pp. 97–114). Cambridge, UK:
Cambridge University Press.
Dawkins, Richard (1989). The Selfish Gene (2nd ed.). Oxford: Oxford University Press.
Draper, Patricia, and Henry Harpending (1982). Father Absence and Reproductive
Strategy: An Evolutionary Perspective. Journal of Anthropological Research 38:
255–273.
Draper, Patricia, and Henry Harpending (1987). Parent Investment and the Child’s
Environment. In Jane B. Lancaster et al. (eds.), Parenting across the Life Span: Bioso-
cial Dimensions (pp. 207–235). New York: Aldine de Gruyter.
Fodor, Jerry A. (1983). The Modularity of Mind. Cambridge, MA: MIT Press.
Futuyma, Douglas J. (1998). Evolutionary Biology (3rd ed.). Sunderland, MA: Sinauer
Associates.
Jones, Steve, Robert Martin, and David Pilbeam (eds.) (1992). The Cambridge Ency-
clopedia of Human Evolution. Cambridge, UK: Cambridge University Press.
Lewin, Roger (1998). Principles of Human Evolution. Malden, MA: Blackwell Science.
Pinker, Steven (1997). How the Mind Works. New York: Norton.
Raff, Rudolf A. (1996). The Shape of Life: Genes, Development, and the Evolution of
Animal Form. Chicago: University of Chicago Press.
Symons, Donald (1987). If We’re All Darwinians, What’s the Fuss About? In C.
Crawford, M. Smith, and D. Krebs (eds.), Sociobiology and Psychology: Ideas, Issues,
and Applications (pp. 121–146). Hillsdale, NJ: Lawrence Erlbaum.
Symons, Donald (1990). Adaptiveness and Adaptation. Ethology and Sociobiology 11:
427–444.
Symons, Donald (1992). On the Use and Misuse of Darwinism in the Study of
Human Behavior. In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted
Mind: Evolutionary Psychology and the Generation of Culture (pp. 137–159). New York:
Oxford University Press.
Symons, Donald (1995). Beauty Is in the Adaptations of the Beholder: The Evolu-
tionary Psychology of Human Female Sexual Attractiveness. In P. R. Abramson and
S. D. Pinkerton (eds.), Sexual Nature, Sexual Culture (pp. 80–118). Chicago: Univer-
sity of Chicago Press.
Tooby, John, and Leda Cosmides (1990a). On the Universality of Human Nature and
the Uniqueness of the Individual: The Role of Genetics and Adaptation. Journal of
Personality 58: 17–67.
502 Bibliography
Tooby, John, and Leda Cosmides (1990b). The Past Explains the Present: Emotional
Adaptations and the Structure of Ancestral Environments. Ethology and Sociobiology
11: 375–424.
Tooby, John, and Leda Cosmides (1992). The Psychological Foundations of Culture.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 19–136). New York: Oxford University Press.
Tooby, John, and Leda Cosmides (1995). Foreword. In Simon Baron-Cohen, Mind-
blindness: An Essay on Autism and Theory of Mind (pp. xi–xviii). Cambridge, MA: MIT
Press.
Chapter 3
Betzig, Laura (1998). Not Whether to Count Babies, but Which. In C. Crawford and
D. L. Krebs (eds.), Handbook of Evolutionary Psychology: Ideas, Issues, and Applications
(pp. 265–273). Mahwah, NJ: Lawrence Erlbaum.
Buss, David M. (1998). Sexual Strategies Theory: Historical Origins and Current
Status. Journal of Sex Research 35: 19–31.
Buss, David M., Martie G. Haselton, Todd K. Shackelford, April L. Bleske, and Jerome
C. Wakefield (1998). Adaptations, Exaptations, and Spandrels. American Psychologist
53: 533–548.
Cavalli-Sforza, Luigi Luca (2000). Genes, Peoples, and Languages (Mark Seielstad,
trans.). New York: North Point Press.
Cosmides, Leda, and John Tooby (1994). Origins of Domain Specificity: The Evolu-
tion of Functional Organization. In L. A. Hirschfeld and S. A. Gelman (eds.), Mapping
the Mind: Domain Specificity in Cognition and Culture (pp. 85–116). New York:
Cambridge University Press.
Cosmides, Leda, John Tooby, and Jerome H. Barkow (1992). Introduction: Evolu-
tionary Psychology and Conceptual Integration. In J. H. Barkow, L. Cosmides, and
J. Tooby (eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture
(pp. 3–15). New York: Oxford University Press.
Davies, Paul Sheldon (1996). Discovering the Functional Mesh: On the Methods of
Evolutionary Psychology. Minds and Machines 6: 559–585.
Diamond, Jared (1991). The Rise and Fall of the Third Chimpanzee. London: Vintage.
Foley, Robert (1996). The Adaptive Legacy of Human Evolution: A Search for the
Environment of Evolutionary Adaptedness. Evolutionary Anthropology 4(6): 194–203.
Francis, Richard C., and George W. Barlow (1993). Social Control of Primary Sex Dif-
ferentiation in the Midas Cichlid. Proceedings of the National Academy of Sciences 90:
10673–10675.
Gangestad, Steven W., and Jeffrey A. Simpson (1993). Development of a Scale Mea-
suring Genetic Variation Related to Expressive Control. Journal of Personality 61:
133–158.
Gittleman, John L., C. Gregory Anderson, Mark Kot, and Hang-Kwang Luh (1996).
Comparative Tests of Evolutionary Lability and Rates Using Molecular Phylogenies.
In P. H. Harvey, A. J. Leigh Brown, J. Maynard Smith, and S. Nee (eds.), New Uses for
New Phylogenies (pp. 289–307). New York: Oxford University Press.
Gould, Stephen Jay (1991). Exaptation: A Crucial Tool for an Evolutionary Psy-
chology. Journal of Social Issues 47: 43–65.
Gould, Stephen Jay (1997). Evolution: The Pleasures of Pluralism. New York Review
of Books 44(11): 47–52.
Gould, Stephen Jay, and Richard C. Lewontin (1979). The Spandrels of San Marco
and the Panglossian Paradigm: A Critique of the Adaptationist Programme. Pro-
ceedings of the Royal Society of London B 205: 581–598.
Gould, Stephen Jay, and Elisabeth S. Vrba (1982). Exaptation: A Missing Term in the
Science of Form. Paleobiology 8: 4–15.
Griffiths, Paul E. (1997). What Emotions Really Are: The Problem of Psychological Cat-
egories. Chicago: University of Chicago Press.
504 Bibliography
Jacob, François (1982). The Possible and the Actual. Seattle: University of Washington
Press.
Laland, Kevin N., John Odling-Smee, and Marcus W. Feldman (2000). Niche Con-
struction, Biological Evolution and Cultural Change. Behavioral and Brain Sciences
23: 131–175.
Lewontin, Richard C. (1983b). The Organism as the Subject and Object of Evolu-
tion. Scientia 118: 65–82.
Odling-Smee, F. John, Kevin N. Laland, and Marcus W. Feldman (1996). Niche Con-
struction. American Naturalist 147: 641–648.
Oyama, Susan (2000). The Ontogeny of Information: Developmental Systems and Evolu-
tion (2nd ed.). Durham, NC: Duke University Press.
Pinker, Steven (1997a). How the Mind Works. New York: Norton.
Pinker, Steven (1997b). Letter to the Editors. New York Review of Books 44(15): 55–
56.
Pinker, Steven, and Paul Bloom (1992). Natural Language and Natural Selection. In
J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 451–493). New York: Oxford University
Press.
Raff, Rudolf A. (1996). The Shape of Life: Genes, Development, and the Evolution of
Animal Form. Chicago: University of Chicago Press.
Bibliography 505
Reznick, David N., Mark J. Butler, F. Helen Rodd, and Patrick Ross (1996). Life-
History Evolution in Guppies (Poecilia reticulata) 6. Differential Mortality as a Mech-
anism for Natural Selection. Evolution 50: 1651–1660.
Reznick, David N., Frank H. Shaw, F. Helen Rodd, and Ruth G. Shaw (1997). Evalu-
ation of the Rate of Evolution in Natural Populations of Guppies (Poecilia reticulata).
Science 275: 1934–1937.
Ross, Robert M., George S. Losey, and Milton Diamond (1983). Sex Change in a
Coral-Reef Fish: Dependence of Stimulation and Inhibition on Relative Size. Science
221: 574–575.
Sterelny, Kim (1995). The Adapted Mind. Biology and Philosophy 10: 365–380.
Sterelny, Kim (2003). Thought in a Hostile World: The Evolution of Human Cognition.
Malden, MA: Blackwell.
Sterelny, Kim, and Paul E. Griffiths (1999). Sex and Death: An Introduction to Philos-
ophy of Biology. Chicago: University of Chicago Press.
Symons, Donald (1992). On the Use and Misuse of Darwinism in the Study of
Human Behavior. In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted
Mind: Evolutionary Psychology and the Generation of Culture (pp. 137–159). New York:
Oxford University Press.
Symons, Donald (1995). Beauty Is in the Adaptations of the Beholder: The Evolu-
tionary Psychology of Human Female Sexual Attractiveness. In P. R. Abramson and
S. D. Pinkerton (eds.), Sexual Nature, Sexual Culture (pp. 80–118). Chicago: Univer-
sity of Chicago Press.
Tooby, John, and Leda Cosmides (1990a). On the Universality of Human Nature and
the Uniqueness of the Individual: The Role of Genetics and Adaptation. Journal of
Personality 58: 17–67.
Tooby, John, and Leda Cosmides (1990b). The Past Explains the Present: Emotional
Adaptations and the Structure of Ancestral Environments. Ethology and Sociobiology
11: 375–424.
Tooby, John, and Leda Cosmides (1992). The Psychological Foundations of Culture.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 19–136). New York: Oxford University Press.
Tooby, John, and Leda Cosmides (2000). Toward Mapping the Evolved Functional
Organization of the Mind and Brain. In M. S. Gazzaniga (ed.), The New Cognitive
Neurosciences (pp. 1167–1178). Cambridge, MA: MIT Press.
506 Bibliography
Wilson, David Sloan (1994). Adaptive Genetic Variation and Human Evolutionary
Psychology. Ethology and Sociobiology 15: 219–235.
Wilson, David Sloan (1998). Game Theory and Human Behavior. In L. A. Dugatkin
and H. K. Reeve (eds.), Game Theory and Animal Behavior (pp. 261–282). New York:
Oxford University Press.
Chapter 4
Adams, David B., Alice Ross Gold, and Anne D. Burt (1978). Rise in Female-Initiated
Sexual Activity at Ovulation and Its Suppression by Oral Contraceptives. New
England Journal of Medicine 299: 1145–1150.
Allman, John (1999). Evolving Brains. New York: Scientific American Library.
Bloom, Paul, and Tim P. German (2000). Two Reasons to Abandon the False Belief
Task as a Test of Theory of Mind. Cognition 77: B25–B31.
Buller, David J., and Valerie Gray Hardcastle (2000). Evolutionary Psychology, Meet
Developmental Neurobiology: Against Promiscuous Modularity. Brain and Mind 1:
307–325.
Changeux, Jean-Pierre (1985). Neuronal Man: The Biology of Mind (Laurence Garey,
trans.). New York: Oxford University Press.
Cheng, Patricia W., and Keith J. Holyoak (1985). Pragmatic Reasoning Schemas. Cog-
nitive Psychology 17: 391–416.
Cheng, Patricia W., and Keith J. Holyoak (1989). On the Natural Selection of Rea-
soning Theories. Cognition 33: 285–313.
Cosmides, Leda (1989). The Logic of Social Exchange: Has Natural Selection Shaped
How Humans Reason? Studies with the Wason Selection Task. Cognition 31: 187–276.
Cosmides, Leda, and John Tooby (1987). From Evolution to Behavior: Evolutionary
Psychology as the Missing Link. In J. Dupré (ed.), The Latest on the Best: Essays on
Evolution and Optimality (pp. 277–306). Cambridge, MA: MIT Press.
Bibliography 507
Cosmides, Leda, and John Tooby (1992). Cognitive Adaptations for Social Exchange.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 163–228). New York: Oxford University
Press.
Cosmides, Leda, and John Tooby (1994). Origins of Domain Specificity: The Evolu-
tion of Functional Organization. In L. A. Hirschfeld and S. A. Gelman (eds.), Mapping
the Mind: Domain Specificity in Cognition and Culture (pp. 85–116). New York:
Cambridge University Press.
Cosmides, Leda, and John Tooby (1997). The Modular Nature of Human Intelli-
gence. In A. B. Scheibel and J. W. Schopf (eds.), The Origin and Evolution of Intelli-
gence (pp. 71–101). Sudbury, MA: Jones and Bartlett.
Cowie, Fiona (1999). What’s Within? Nativism Reconsidered. New York: Oxford Uni-
versity Press.
Cummins, Denise, Robert Cummins, and Pierre Poirier (2003). Cognitive Evolu-
tionary Psychology without Representational Nativism. Journal of Experimental and
Theoretical Artificial Intelligence 15: 143–159.
Davies, Paul Sheldon, James H. Fetzer, and Thomas R. Foster (1995). Logical Rea-
soning and Domain Specificity: A Critique of the Social Exchange Theory of Rea-
soning. Biology and Philosophy 10: 1–37.
Deacon, Terrence W. (1997). The Symbolic Species: The Co-evolution of Language and
the Brain. New York: Norton.
Duchaine, Bradley, Leda Cosmides, and John Tooby (2001). Evolutionary Psychol-
ogy and the Brain. Current Opinion in Neurobiology 11: 225–230.
Edelman, Gerald M. (1992). Bright Air, Brilliant Fire: On the Matter of the Mind. New
York: Basic Books.
Fodor, Jerry (2000). The Mind Doesn’t Work That Way: The Scope and Limits of Com-
putational Psychology. Cambridge, MA: MIT Press.
Futuyma, Douglas J. (1998). Evolutionary Biology (3rd ed.). Sunderland, MA: Sinauer
Associates.
Gierer, Alfred, and Christian M. Muller (1995). Development of Layers, Maps, and
Modules. Current Opinion in Neurobiology 5: 91–97.
Gigerenzer, Gerd, and Klaus Hug (1992). Domain-Specific Reasoning: Social Con-
tracts, Cheating, and Perspective Change. Cognition 43: 127–171.
Goldman-Rakic, Patricia S., M. Isseroff, M. Schwartz, and N. Bugbee (1983). The Neu-
robiology of Child Development. In F. Plum and V. Mountcastle (eds.), The Hand-
book of Child Development (pp. 282–331). Bethesda, MD: American Physiological
Society.
La Cerra, Peggy, and Roger Bingham (1998). The Adaptive Nature of the Human
Neurocognitive Architecture: An Alternative Model. Proceedings of the National
Academy of Sciences 95: 11290–11294.
Manktelow, K. I., and D. E. Over (1990). Deontic Thought and the Selection Task.
In K. J. Gilhooly, M. T. G. Keane, R. H. Logie, and G. Erdos (eds.), Lines of Thinking:
Reflections on the Psychology of Thought: Representation, Reasoning, Analogy and Deci-
sion Making (pp. 153–164). New York: Wiley.
Pascalis, Olivier, Michelle de Haan, and Charles A. Nelson (2002). Is Face Process-
ing Species-Specific during the First Year of Life? Science 296: 1321–1323.
Peterson, Candida C., and Michael Siegal (2000). Insights into Theory of Mind from
Deafness and Autism. Mind and Language 15: 123–145.
Pinker, Steven (1997). How the Mind Works. New York: Norton.
Pinker, Steven (2002). The Blank Slate: The Modern Denial of Human Nature. New York:
Viking.
Popper, Karl (1992). Science: Conjectures and Refutations. In Conjectures and Refu-
tations: The Growth of Scientific Knowledge (pp. 33–55). New York: Routledge.
Quartz, Steven R. (2003). Innateness and the Brain. Biology and Philosophy 18: 13–40.
Shapiro, Lawrence, and William Epstein (1998). Evolutionary Theory Meets Cogni-
tive Psychology: A More Selective Perspective. Mind and Language 13: 171–194.
Sperber, Dan, Francesco Cara, and Vittorio Girotto (1995). Relevance Theory
Explains the Selection Task. Cognition 57: 31–95.
Sperber, Dan, and Vittorio Girotto (2002). Use or Misuse of the Selection Task?
Rejoinder to Fiddick, Cosmides, and Tooby. Cognition 85: 277–290.
Stanislaw, Harold, and Frank J. Rice (1988). Correlation between Sexual Desire and
Menstrual Cycle Characteristics. Archives of Sexual Behavior 17: 499–508.
Sterelny, Kim (2003). Thought in a Hostile World: The Evolution of Human Cognition.
Malden, MA: Blackwell.
Stone, Valerie E., Leda Cosmides, John Tooby, Neal Kroll, and Robert T. Knight
(2002). Selective Impairment of Reasoning about Social Exchange in a Patient with
Bilateral Limbic System Damage. Proceedings of the National Academy of Sciences 99:
11531–11536.
Sugiyama, Lawrence S., John Tooby, and Leda Cosmides (2002). Cross-cultural Evi-
dence of Cognitive Adaptations for Social Exchange among the Shiwiar of Ecuado-
rian Amazonia. Proceedings of the National Academy of Sciences 99: 11537–11542.
Symons, Donald (1992). On the Use and Misuse of Darwinism in the Study of
Human Behavior. In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted
510 Bibliography
Mind: Evolutionary Psychology and the Generation of Culture (pp. 137–159). New York:
Oxford University Press.
Tooby, John, and Leda Cosmides (1992). The Psychological Foundations of Culture.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary
Psychology and the Generation of Culture (pp. 19–136). New York: Oxford University
Press.
Wong, Rachel O. L. (1999). Retinal Waves and Visual System Development. Annual
Review of Neuroscience 22: 29–47.
Wright, Robert (1994). The Moral Animal: Evolutionary Psychology and Everyday Life.
New York: Vintage.
Chapter 5
Boyd, Robert, and Joan B. Silk (2000). How Humans Evolved (2nd ed.). New York:
Norton.
Burger, Jerry M., and Mica Cosby (1999). Do Women Prefer Dominant Men? The
Case of the Missing Control Condition. Journal of Research in Personality 33: 358–
368.
Buss, David M. (1994). The Evolution of Desire: Strategies of Human Mating. New York:
Basic Books.
Buss, David M. (1999). Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn and Bacon.
Buss, David M., Max Abbott, Alois Angleitner, Armen Asherian, Angela Biaggio,
Angel Blanco-Villasenor, M. Bruchon-Schweitzer, et al. (1990). International Prefer-
ences in Selecting Mates: A Study of 37 Cultures. Journal of Cross-Cultural Psychology
21: 5–47.
Bibliography 511
Buston, Peter M., and Stephen T. Emlen (2003). Cognitive Processes Underlying
Human Mate Choice: The Relationship between Self-Perception and Mate Preference
in Western Society. Proceedings of the National Academy of Sciences 100: 8805–8810.
Buunk, Bram P., Pieternel Dijkstra, Douglas T. Kenrick, and Astrid Warntjes (2001).
Age Preferences for Mates as Related to Gender, Own Age, and Involvement Level.
Evolution and Human Behavior 22: 241–250.
Daly, Martin, and Margo Wilson (1983). Sex, Evolution, and Behavior (2nd ed.).
Belmont, CA: Wadsworth.
De Waal, Frans (1998). Chimpanzee Politics: Power and Sex among Apes (rev. ed.).
Baltimore: Johns Hopkins University Press.
De Waal, Frans (2000). Survival of the Rapist: Review of A Natural History of Rape,
by Randy Thornhill and Craig T. Palmer. New York Times Book Review (April 2), pp.
24–25.
Doyle, Rodger (2000). By the Numbers: Women and the Professions. Scientific Amer-
ican 282(4): 30.
Dupré, John (1993). Scientism, Sexism and Sociobiology: One More Link in the
Chain. Behavioral and Brain Sciences 16: 292.
Eagly, Alice H., and Wendy Wood (1999). The Origins of Sex Differences in Human
Behavior: Evolved Dispositions versus Social Roles. American Psychologist 54:
408–423.
Ellis, Lee (1995). Dominance and Reproductive Success among Nonhuman Animals:
A Cross-Species Comparison. Ethology and Sociobiology 16: 257–333.
Euler, Harald A., and Barbara Weitzel (1996). Discriminative Grandparental Solici-
tude as Reproductive Strategy. Human Nature 7: 39–59.
512 Bibliography
Ford, W. C. L., Kate North, Hazel Taylor, Alexandra Farrow, M. G. R. Hull, Jean
Golding, and ALSPAC Study Team (2000). Increasing Paternal Age Is Associated with
Delayed Conception in a Large Population of Fertile Couples: Evidence for Declin-
ing Fecundity in Older Men. Human Reproduction 15: 1703–1708.
Gangestad, Steven W., and Randy Thornhill (1997). The Evolutionary Psychology
of Extrapair Sex: The Role of Fluctuating Asymmetry. Evolution and Human Behavior
18: 69–88.
Hamermesh, Daniel S., and Jeff E. Biddle (1994). Beauty and the Labor Market. Amer-
ican Economic Review 84: 1174–1194.
Hawkes, Kristen, James F. O’Connell, and Nicholas G. Blurton Jones (1997). Hadza
Women’s Time Allocation, Offspring Provisioning, and the Evolution of Long Post-
menopausal Life Spans. Current Anthropology 38: 551–577.
Hawkes, Kristen, James F. O’Connell, Nicholas G. Blurton Jones, Helen Alvarez, and
Eric L. Charnov (1998). Grandmothering, Menopause, and the Evolution of Human
Life Histories. Proceedings of the National Academy of Sciences USA 95: 1336–1339.
Hrdy, Sarah Blaffer (1997). Raising Darwin’s Consciousness: Female Sexuality and
the Prehominid Origins of Patriarchy. Human Nature 8: 1–49.
Hyde, Janet Shibley, and John D. DeLamater (2000). Understanding Human Sexuality
(7th ed.). Boston: McGraw-Hill.
Jankowiak, William R., Elizabeth M. Hill, and James M. Donovan (1992). The Effects
of Gender and Sexual Orientation on Attractiveness Judgments: An Evolutionary
Interpretation. Ethology and Sociobiology 13: 73–85.
Kalmijn, Matthijs (1991). Status Homogamy in the United States. American Journal
of Sociology 97: 496–523.
Kenrick, Douglas T., and Richard C. Keefe (1992). Age Preferences in Mates Reflect
Sex Differences in Reproductive Strategies. Behavioral and Brain Sciences 15: 75–133.
Kenrick, Douglas T., Richard C. Keefe, Cristina Gabrielidis, and Jeffrey S. Cornelius
(1996). Adolescents’ Age Preferences for Dating Partners: Support for an Evolution-
ary Model of Life-History Strategies. Child Development 67: 1499–1511.
Kenrick, Douglas T., Edward K. Sadalla, Gary Groth, and Melanie R. Trost (1990).
Evolution, Traits, and the Stages of Human Courtship: Qualifying the Parental
Investment Model. Journal of Personality 58: 97–116.
Laumann, Edward O., John H. Gagnon, Robert T. Michael, and Stuart Michaels
(1994). The Social Organization of Sexuality: Sexual Practices in the United States.
Chicago: University of Chicago Press.
Leonard, Janet L. (1989). Homo sapiens: A Good Fit to Theory, but Posing Some
Enigmas. Behavioral and Brain Sciences 12: 26–27.
Marlowe, Frank, and Adam Wetsman (2001). Preferred Waist-to-Hip Ratio and
Ecology. Personality and Individual Differences 30: 481–489.
Mulford, Matthew, John Orbell, Catherine Shatto, and Jean Stockard (1998). Phys-
ical Attractiveness, Opportunity, and Success in Everyday Exchange. American
Journal of Sociology 103: 1565–1592.
Popenoe, David, and Barbara Dafoe Whitehead (2000). The State of Our Unions:
The Social Health of Marriage in America, 2000. A Report of The National Marriage
Project (http://marriage.rutgers.edu/SOOU/NMPAR2000.pdf).
Rasmussen, Jeffrey Lee, D. W. Rajecki, Anita A. Ebert, Kathy Lagler, Candice Brewer,
and Elizabeth Cochran (1998). Age Preferences in Personal Advertisements: Two Life
History Strategies or One Matching Tactic? Journal of Social and Personal Relationships
15: 77–89.
Sadalla, Edward K., Douglas T. Kenrick, and Beth Vershure (1987). Dominance and
Heterosexual Attraction. Journal of Personality and Social Psychology 52: 730–738.
Simpson, Jeffrey A., and Steven W. Gangestad (1992). Sociosexuality and Romantic
Partner Choice. Journal of Personality 60: 31–51.
514 Bibliography
Singh, Devendra (1993). Body Shape and Women’s Attractiveness: The Critical Role
of Waist-to-Hip Ratio. Human Nature 4: 297–321.
Singh, Devendra, and Suwardi Luis (1995). Ethnic and Gender Consensus for the
Effect of Waist-to-Hip Ratio on Judgment of Women’s Attractiveness. Human Nature
6: 51–65.
Stevens, Gillian, Dawn Owens, and Eric C. Schaefer (1990). Education and Attrac-
tiveness in Marriage Choices. Social Psychology Quarterly 53: 62–70.
Symons, Donald (1979). The Evolution of Human Sexuality. New York: Oxford Uni-
versity Press.
Symons, Donald (1995). Beauty Is in the Adaptations of the Beholder: The Evolu-
tionary Psychology of Human Female Sexual Attractiveness. In P. R. Abramson and
S. D. Pinkerton (eds.), Sexual Nature, Sexual Culture (pp. 80–118). Chicago: Univer-
sity of Chicago Press.
Tassinary, Louis G., and Kristi A. Hansen (1998). A Critical Test of the Waist-to-Hip-
Ratio Hypothesis of Female Physical Attractiveness. Psychological Science 9: 150–155.
Taylor, Patricia Ann, and Norval D. Glenn (1976). The Utility of Education and
Attractiveness for Females’ Status Attainment through Marriage. American Sociologi-
cal Review 41: 484–498.
Townsend, John Marshall (1998). What Women Want—What Men Want: Why the
Sexes Still See Love and Commitment So Differently. New York: Oxford University Press.
Townsend, John Marshall, and Gary D. Levy (1990a). Effects of Potential Partners’
Costume and Physical Attractiveness on Sexuality and Partner Selection. Journal of
Psychology 124: 371–389.
Townsend, John Marshall, and Gary D. Levy (1990b). Effects of Potential Partners’
Physical Attractiveness and Socioeconomic Status on Sexuality and Partner Selec-
tion. Archives of Sexual Behavior 19: 149–164.
Umberson, Debra, and Michael Hughes (1987). The Impact of Physical Attractive-
ness on Achievement and Psychological Well Being. Social Psychology Quarterly 50:
227–236.
Van den Berghe, Pierre L. (1992). Wanting and Getting Ain’t the Same. Behavioral
and Brain Sciences 15: 116–117.
Weinberg, Martin S., and Colin J. Williams (1980). Sexual Embourgeoisment? Social
Class and Sexual Activity: 1938–1970. American Sociological Review 45(1): 33–48.
Wetsman, Adam, and Frank Marlowe (1999). How Universal Are Preferences for
Female Waist-to-Hip Ratios? Evidence from the Hadza of Tanzania. Evolution and
Human Behavior 20: 219–228.
Wiederman, Michael W., and Elizabeth Rice Allgeier (1992). Gender Differences in
Mate Selection Criteria: Sociobiological or Socioeconomic Explanation? Ethology and
Sociobiology 13: 115–124.
Yu, Douglas W., and Glenn H. Shepard (1998). Is Beauty in the Eye of the Beholder?
Nature 396: 321–322.
Chapter 6
Adams, David B., Alice Ross Gold, and Anne D. Burt (1978). Rise in Female-Initiated
Sexual Activity at Ovulation and Its Suppression by Oral Contraceptives. New
England Journal of Medicine 299: 1145–1150.
Bailey, J. Michael, Steven Gaulin, Yvonne Agyei, and Brian A. Gladue (1994). Effects
of Gender and Sexual Orientation on Evolutionarily Relevant Aspects of Human
Mating Psychology. Journal of Personality and Social Psychology 66: 1081–1093.
Baker, R. Robin, and Mark A. Bellis (1993). Human Sperm Competition: Ejaculate
Manipulation by Females and a Function for the Female Orgasm. Animal Behaviour
46: 887–909.
Baker, R. Robin, and Mark A. Bellis (1995). Human Sperm Competition: Copulation,
Masturbation, and Infidelity. New York: Chapman and Hall.
Bell, Robert R., and Dorthyann Peltz (1974). Extramarital Sex among Women.
Medical Aspects of Human Sexuality 8: 10–31.
Bell, Robert R., Stanley Turner, and Lawrence Rosen (1975). A Multivariate Analysis
of Female Extramarital Coitus. Journal of Marriage and the Family 37: 375–384.
Bellis, Mark A., and R. Robin Baker (1990). Do Females Promote Sperm Competi-
tion? Data for Humans. Animal Behaviour 40: 997–999.
Brewer, Devon D., John J. Potterat, Sharon B. Garrett, Stephen Q. Muth, Jr., John
M. Roberts, Danuta Kasprzyk, Daniel E. Montano, and William W. Darrow (2000).
516 Bibliography
Prostitution and the Sex Discrepancy in Reported Number of Sexual Partners. Pro-
ceedings of the National Academy of Sciences 97: 12385–12388.
Brown, Norman R., and Robert C. Sinclair (1999). Estimating Number of Lifetime
Sexual Partners: Men and Women Do It Differently. Journal of Sex Research 36:
292–297.
Buckle, Leslie, Gordon G. Gallup, and Zachary A. Rodd (1996). Marriage as a Repro-
ductive Contract: Patterns of Marriage, Divorce, and Remarriage. Evolution and
Human Behavior 17: 363–377.
Buss, David M. (1994). The Evolution of Desire: Strategies of Human Mating. New York:
Basic Books.
Buss, David M. (1999). Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn and Bacon.
Buss, David M. (2000). The Dangerous Passion: Why Jealousy Is as Necessary as Love
and Sex. New York: Free Press.
Buss, David M., Randy J. Larsen, and Drew Westen (1996). Sex Differences in
Jealousy: Not Gone, Not Forgotten, and Not Explained by Alternative Hypotheses.
Psychological Science 7: 373–375.
Buss, David M., Randy J. Larsen, Drew Westen, and Jennifer Semmelroth (1992). Sex
Differences in Jealousy: Evolution, Physiology, and Psychology. Psychological Science
3: 251–255.
Buss, David M., Todd K. Shackelford, Lee A. Kirkpatrick, Jae C. Choe, Hang K. Lim,
Mariko Hasegawa, Toshikazu Hasegawa, and Kevin Bennett (1999). Jealousy and
the Nature of Beliefs about Infidelity: Tests of Competing Hypotheses about Sex
Differences in the United States, Korea, and Japan. Personal Relationships 6: 125–
150.
Buunk, Bram P., Alois Angleitner, Viktor Oubaid, and David M. Buss (1996). Sex Dif-
ferences in Jealousy in Evolutionary and Cultural Perspective: Tests from the Nether-
lands, Germany, and the United States. Psychological Science 7: 359–363.
Chesser, Eustace (1956). The Sexual, Marital and Family Relationships of the English
Woman. London: Hutchinson’s Medical Publications.
Clark, Andy (1989). Microcognition: Philosophy, Cognitive Science, and Parallel Distrib-
uted Processing. Cambridge, MA: MIT Press.
Daly, Martin, and Margo Wilson (1983). Sex, Evolution, and Behavior (2nd ed.).
Belmont, CA: Wadsworth.
Daly, Martin, and Margo Wilson (1988). Homicide. New York: Aldine de Gruyter.
Bibliography 517
Daly, Martin, and Margo Wilson (1996). Evolutionary Psychology and Marital Con-
flict. In D. M. Buss, and N. M. Malamuth (eds.), Sex, Power, Conflict: Evolutionary and
Feminist Perspectives (pp. 9–28). New York: Oxford University Press.
Daly, Martin, Margo Wilson, and Suzanne J. Weghorst (1982). Male Sexual Jealousy.
Ethology and Sociobiology 3: 11–27.
DeSteno, David A., and Peter Salovey (1996a). Evolutionary Origins of Sex Differ-
ences in Jealousy? Psychological Science 7: 367–372.
DeSteno, David A., and Peter Salovey (1996b). Genes, Jealousy, and the Replication
of Misspecified Models. Psychological Science 7: 376–377.
Einon, Dorothy (1994). Are Men More Promiscuous than Women? Ethology and
Sociobiology 15: 131–143.
Ellis, Bruce J., and Donald Symons (1990). Sex Differences in Sexual Fantasy: An
Evolutionary Psychological Approach. Journal of Sex Research 27: 527–555.
Fisher, H. E. (1992). Anatomy of Love: The Natural History of Monogamy, Adultery and
Divorce. New York: Norton.
Gangestad, Steven W., and Randy Thornhill (1997). The Evolutionary Psychology
of Extrapair Sex: The Role of Fluctuating Asymmetry. Evolution and Human Behavior
18: 69–88.
Gangestad, Steven W., Randy Thornhill, and Christine E. Garver (2002). Changes in
Women’s Sexual Interests and Their Partners’ Mate-Retention Tactics across the Men-
strual Cycle: Evidence for Shifting Conflicts of Interest. Proceedings of the Royal Society
of London B 269: 975–982.
Geary, David C., Michael Rumsey, C. Christine Bow-Thomas, and Mary K. Hoard
(1995). Sexual Jealousy as a Facultative Trait: Evidence from the Pattern of Sex Dif-
ferences in Adults from China and the United States. Ethology and Sociobiology 16:
355–383.
Glass, Shirley P., and Thomas L. Wright (1985). Sex Differences in Type of Extra-
marital Involvement and Marital Dissatisfaction. Sex Roles 12: 1101–1120.
Glass, Shirley P., and Thomas L. Wright (1992). Justifications for Extramarital Rela-
tionships: The Association between Attitudes, Behaviors, and Gender. Journal of Sex
Research 29: 361–387.
Gough, E. K. (1959). The Nayars and the Definition of Marriage. Journal of the Royal
Anthropological Institute of Great Britain and Ireland 89: 23–34.
Greiling, Heidi, and David M. Buss (2000). Women’s Sexual Strategies: The Hidden
Dimension of Extra-Pair Mating. Personality and Individual Differences 28: 929–963.
518 Bibliography
Harris, Christine R., and Nicholas Christenfeld (1996a). Gender, Jealousy, and
Reason. Psychological Science 7: 364–366.
Harris, Christine R., and Nicholas Christenfeld (1996b). Jealousy and Rational
Responses to Infidelity across Gender and Culture. Psychological Science 7: 378–379.
Hawkes, Kristen, Alan R. Rogers, and Eric L. Charnov (1995). The Male’s Dilemma:
Increased Offspring Production Is More Paternity to Steal. Evolutionary Ecology 9:
662–677.
Hill, Elizabeth M. (1988). The Menstrual Cycle and Components of Human Female
Sexual Behaviour. Journal of Social and Biological Structures 11: 443–455.
Hoogland, John L. (1998). Why Do Female Gunnison’s Prairie Dogs Copulate with
More than One Male? Animal Behaviour 55: 351–359.
Kinsey, Alfred C., Wardell B. Pomeroy, and Clyde E. Martin (1948). Sexual Behavior
in the Human Male. Philadelphia: W. B. Saunders.
Kinsey, Alfred C., Wardell B. Pomeroy, Clyde E. Martin, and Paul H. Gebhard (1953).
Sexual Behavior in the Human Female. Philadelphia: W. B. Saunders.
Laumann, Edward O., John H. Gagnon, Robert T. Michael, and Stuart Michaels
(1994). The Social Organization of Sexuality: Sexual Practices in the United States.
Chicago: University of Chicago Press.
Macintyre, Sally, and Anne Sooman (1991). Non-Paternity and Prenatal Genetic
Screening. Lancet 338: 869–871.
Marlowe, Frank (2000). Paternal Investment and the Human Mating System. Behav-
ioural Processes 51: 45–61.
Marlowe, Frank (2001). Male Contribution to Diet and Female Reproductive Success
among Foragers. Current Anthropology 42: 755–760.
Paul, Luci, Mark A. Foss, and MaryAnn Baenninger (1996). Double Standards for
Sexual Jealousy: Manipulative Morality or a Reflection of Evolved Sex Differences?
Human Nature 7: 291–324.
Pietrzak, Robert H., James D. Laird, David A. Stevens, and Nicholas S. Thompson
(2002). Sex Differences in Human Jealousy: A Coordinated Study of Forced-Choice,
Bibliography 519
Pound, Nicholas, and Martin Daly (2000). Functional Significance of Human Female
Orgasm Still Hypothetical. Behavioral and Brain Sciences 23: 620–621.
Sagarin, Brad J., D. Vaughn Becker, Rosanna E. Guadagno, Lionel D. Nicastle, and
Allison Millevoi (2003). Sex Differences (and Similarities) in Jealousy: The Moder-
ating Influence of Infidelity Experience and Sexual Orientation of the Infidelity. Evo-
lution and Human Behavior 24: 17–23.
Sheets, Virgil L., and Marlow D. Wolfe (2001). Sexual Jealousy in Heterosexuals,
Lesbians, and Gays. Sex Roles 44: 255–276.
Stanislaw, Harold, and Frank J. Rice (1988). Correlation between Sexual Desire and
Menstrual Cycle Characteristics. Archives of Sexual Behavior 17: 499–508.
Symons, Donald (1979). The Evolution of Human Sexuality. New York: Oxford Uni-
versity Press.
Tavris, Carol, and Susan Sadd (1975). The Redbook Report on Female Sexuality. New
York: Delacorte Press.
Thornhill, Randy, Steven W. Gangestad, and Randall Comer (1995). Human Female
Orgasm and Mate Fluctuating Asymmetry. Animal Behaviour 50: 1601–1615.
Twain, Mark (1991). Letters from the Earth. New York: HarperPerennial.
Van Schaik, Carel P., and Andreas Paul (1996). Male Care in Primates: Does It Ever
Reflect Paternity? Evolutionary Anthropology 5: 152–156.
Wiederman, Michael W., and Elizabeth Rice Allgeier (1993). Gender Differences in
Sexual Jealousy: Adaptionist or Social Learning Explanation? Ethology and Sociobiol-
ogy 14: 115–140.
Wiederman, Michael W., and Erica Kendall (1999). Evolution, Sex, and Jealousy:
Investigation with a Sample from Sweden. Evolution and Human Behavior 20:
121–128.
Wiederman, Michael W., and Lisa LaMar (1998). “Not with Him You Don’t!”: Gender
and Emotional Reactions to Sexual Infidelity during Courtship. Journal of Sex
Research 35: 288–297.
Wilson, Margo, and Martin Daly (1992). The Man Who Mistook His Wife for a
Chattel. In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolu-
tionary Psychology and the Generation of Culture (pp. 289–322). New York: Oxford Uni-
versity Press.
Wilson, Margo, and Martin Daly (1996). Male Sexual Proprietariness and Violence
against Wives. Current Directions in Psychological Science 5: 2–7.
Wolf, Arthur P. (1966). Childhood Association, Sexual Attraction, and the Incest
Taboo: A Chinese Case. American Anthropologist 68: 883–898.
Wolf, Arthur P. (1970). Childhood Association and Sexual Attraction: A Further Test
of the Westermarck Hypothesis. American Anthropologist 72: 503–515.
Chapter 7
Anderson, Kermyt G., Hillard Kaplan, David Lam, and Jane Lancaster (1999). Pater-
nal Care by Genetic Fathers and Stepfathers II: Reports by Xhosa High School Stu-
dents. Evolution and Human Behavior 20: 433–451.
Anderson, Kermyt G., Hillard Kaplan, and Jane Lancaster (1999). Paternal Care by
Genetic Fathers and Stepfathers I: Reports from Albuquerque Men. Evolution and
Human Behavior 20: 405–431.
Buss, David M. (1999). Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn and Bacon.
Cashell, Alan W. (1987). Homicide as a Cause of the Sudden Infant Death Syndrome.
American Journal of Forensic Medicine and Pathology 8: 256–258.
Christoffel, Katherine K., Nora K. Anzinger, and David A. Merrill (1989). Age-Related
Patterns of Violent Death, Cook County, Illinois, 1977 through 1982. American
Journal of Diseases of Children 143: 1403–1409.
Christoffel, Katherine K., Peter C. Scheidt, Phyllis F. Agran, Jess F. Kraus, Elizabeth
McLoughlin, and Jerome A. Paulson (1992). Standard Definitions for Childhood
Injury Research: Excerpts of a Conference Report. Pediatrics 89: 1027–1034.
Crume, Tessa L., Carolyn DiGuiseppi, Tim Byers, Andrew P. Sirotnak, and Carol J.
Garrett (2002). Underascertainment of Child Maltreatment Fatalities by Death Cer-
tificates, 1990–1998. Pediatrics 110(2): e18.
Daly, Martin, and Margo Wilson (1981). Abuse and Neglect of Children in Evolu-
tionary Perspective. In R. D. Alexander, and D. W. Tinkle (eds.), Natural Selection and
Social Behavior: Recent Research and New Theory (pp. 405–416). New York: Chiron
Press.
Daly, Martin, and Margo Wilson (1985). Child Abuse and Other Risks of Not Living
with Both Parents. Ethology and Sociobiology 6: 197–210.
Daly, Martin, and Margo Wilson (1987). Children as Homicide Victims. In R. J. Gelles
and J. B. Lancaster (eds.), Child Abuse and Neglect: Biosocial Dimensions (pp. 201–214).
New York: Aldine de Gruyter.
Daly, Martin, and Margo Wilson (1988a). Evolutionary Social Psychology and Family
Homicide. Science 242: 519–524.
Daly, Martin, and Margo Wilson (1988b). Homicide. New York: Aldine de Gruyter.
Daly, Martin, and Margo Wilson (1991). A Reply to Gelles: Stepchildren Are Dis-
proportionately Abused, and Diverse Forms of Violence Can Share Causal Factors.
Human Nature 2: 419–426.
Daly, Martin, and Margo Wilson (1994a). Some Differential Attributes of Lethal
Assaults on Small Children by Stepfathers versus Genetic Fathers. Ethology and Socio-
biology 15: 207–217.
Daly, Martin, and Margo Wilson (1994b). Stepparenthood and the Evolved Psy-
chology of Discriminative Parental Solicitude. In S. Parmigiani and F. S. v. Saal (eds.),
Infanticide and Parental Care (pp. 121–134). Chur, Switzerland: Harwood.
Daly, Martin, and Margo Wilson (1995). Discriminative Parental Solicitude and the
Relevance of Evolutionary Models to the Analysis of Motivational Systems. In M. S.
Gazzaniga (ed.), The Cognitive Neurosciences (pp. 1269–1286). Cambridge, MA: MIT
Press.
Daly, Martin, and Margo Wilson (1996). Violence against Stepchildren. Current Direc-
tions in Psychological Science 5: 77–81.
522 Bibliography
Daly, Martin, and Margo Wilson (1999). The Truth about Cinderella: A Darwinian View
of Parental Love. New Haven, CT: Yale University Press.
Daly, Martin, and Margo Wilson (2001). An Assessment of Some Proposed Excep-
tions to the Phenomenon of Nepotistic Discrimination against Stepchildren. Annales
Zoologici Fennici 38: 287–296.
Ewigman, Bernard, Coleen Kivlahan, and Garland Land (1993). The Missouri Child
Fatality Study: Underreporting of Maltreatment Fatalities among Children Younger
than Five Years of Age, 1983 through 1986. Pediatrics 91: 330–337.
Gelles, Richard J. (1991). Physical Violence, Child Abuse, and Child Homicide: A
Continuum of Violence, or Distinct Behaviors? Human Nature 2: 59–72.
Gelles, Richard J., and John W. Harrop (1991). The Risk of Abusive Violence among
Children with Non-genetic Caretakers. Family Relations 40: 78–83.
Giles-Sims, Jean (1997). Current Knowledge about Child Abuse in Stepfamilies. Mar-
riage and Family Review 26: 215–230.
Giles-Sims, Jean, and David Finkelhor (1984). Child Abuse in Stepfamilies. Family
Relations 33: 407–413.
Herndon, Joy, Lilo T. Strauss, Sara Whitehead, Wilda Y. Parker, Linda Bartlett, and
Suzanne Zane (2002). Abortion Surveillance—United States, 1998. In Surveillance
Summaries, April 13, 2002. Morbidity and Mortality Weekly Report 51 (no. SS-3), pp.
1–32.
Malkin, Catherine M., and Michael E. Lamb (1994). Child Maltreatment: A Test of
Sociobiological Theory. Journal of Comparative Family Studies 25: 121–133.
Marlowe, Frank (1999). Showoffs or Providers? The Parenting Effort of Hadza Men.
Evolution and Human Behavior 20: 391–404.
Bibliography 523
Parker, Hilda, and Seymour Parker (1986). Father-Daughter Sexual Abuse: An Emerg-
ing Perspective. American Journal of Orthopsychiatry 56: 531–549.
Pinker, Steven (2002). The Blank Slate: The Modern Denial of Human Nature. New York:
Viking.
Schaik, Carel P. van (2000). Infanticide by Male Primates: The Sexual Selection
Hypothesis Revisited. In C. P. van Schaik and C. H. Janson (eds.), Infanticide by Males
and Its Implications (pp. 27–60). New York: Cambridge University Press.
Sedlak, Andrea J., and Diane D. Broadhurst (1996). Executive Summary of the Third
National Incidence Study of Child Abuse and Neglect. Retrieved 2003 from the
U. S. Department of Health and Human Services, Administration for Children and
Families, National Clearinghouse on Child Abuse and Neglect Information
(http://nccanch.acf.hhs.gov/pubs/statsinfo/nis3.cfm).
Stiffman, Michael N., Patricia G. Schnitzer, Patricia Adam, Robin L. Kruse, and
Bernard G. Ewigman (2002). Household Composition and Risk of Fatal Child Mal-
treatment. Pediatrics 109: 615–621.
Temrin, Hans, Susanne Buchmayer, and Magnus Enquist (2000). Step-parents and
Infanticide: New Data Contradict Evolutionary Predictions. Proceedings of the Royal
Society of London B 267: 943–945.
Temrin, Hans, Johanna Nordlund, and Helena Sterner (2004). Are Stepchildren Over-
represented as Victims of Lethal Parental Violence in Sweden? Proceedings of the Royal
Society of London B 271: S124–S126.
Voland, Eckart, and Peter Stephan (2000). “The Hate That Love Generated”: Sexu-
ally Selected Neglect of One’s Own Offspring in Humans. In C. P. van Schaik and
C. H. Janson (eds.), Infanticide by Males and Its Implications (pp. 447–465). New York:
Cambridge University Press.
Wilson, Margo, and Martin Daly (1987). Risk of Maltreatment of Children Living
with Stepparents. In R. J. Gelles and J. B. Lancaster (eds.), Child Abuse and Neglect:
Biosocial Dimensions (pp. 215–232). New York: Aldine de Gruyter.
Wilson, Margo, and Martin Daly (1994). The Psychology of Parenting in Evolu-
tionary Perspective and the Case of Human Filicide. In S. Parmigiani and F. S. vom
Saal (eds.), Infanticide and Parental Care (pp. 73–104). Chur, Switzerland: Harwood.
524 Bibliography
Wilson, Margo, Martin Daly, and Suzanne J. Weghorst (1980). Household Compo-
sition and the Risk of Child Abuse and Neglect. Journal of Biosocial Science 12:
333–340.
Wolfner, Glenn D., and Richard J. Gelles (1993). A Profile of Violence toward Chil-
dren: A National Study. Child Abuse and Neglect 17: 197–212.
Chapter 8
Amundson, Ron (1994). Two Concepts of Constraint: Adaptationism and the Chal-
lenge from Developmental Biology. Philosophy of Science 61: 556–578.
Bonner, John T. (1980). The Evolution of Culture in Animals. Princeton, NJ: Princeton
University Press.
Buss, David M. (1999). Evolutionary Psychology: The New Science of the Mind. Boston:
Allyn and Bacon.
Cavalli-Sforza, L. Luca, Paolo Menozzi, and Alberto Piazza (1994). The History and
Geography of Human Genes. Princeton: Princeton University Press.
Cosmides, Leda, and John Tooby (1987). From Evolution to Behavior: Evolutionary
Psychology as the Missing Link. In J. Dupré (ed.), The Latest on the Best: Essays on
Evolution and Optimality (pp. 277–306). Cambridge, MA: MIT Press.
Cosmides, Leda, and John Tooby (1992). Cognitive Adaptations for Social Exchange.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 163–228). New York: Oxford University
Press.
Cosmides, Leda, and John Tooby (1994). Origins of Domain Specificity: The Evolu-
tion of Functional Organization. In L. A. Hirschfeld, and S. A. Gelman (eds.),
Mapping the Mind: Domain Specificity in Cognition and Culture (pp. 85–116). New York:
Cambridge University Press.
Cosmides, Leda, and John Tooby (1997). The Modular Nature of Human Intelli-
gence. In A. B. Scheibel and J. W. Schopf (eds.), The Origin and Evolution of Intelli-
gence (pp. 71–101). Sudbury, MA: Jones and Bartlett.
Deacon, Terrence W. (1997). The Symbolic Species: The Co-evolution of Language and
the Brain. New York: Norton.
Bibliography 525
Futuyma, Douglas J. (1998). Evolutionary Biology (3rd ed.). Sunderland, MA: Sinauer
Associates.
Ghiselin, Michael T. (1997). Metaphysics and the Origin of Species. Albany: State Uni-
versity of New York Press.
Godfrey-Smith, Peter (1999). Adaptationism and the Power of Selection. Biology and
Philosophy 14: 181–194.
Hull, David L. (1980). Individuality and Selection. Annual Review of Ecology and Sys-
tematics 11: 311–332.
Izpisúa Belmonte, Juan Carlos (1999). How the Body Tells Left from Right. Scientific
American 280(6): 46–51.
Lewontin, Richard (1995). Human Diversity. New York: Scientific American Library.
Lincoln, Bruce (1975). The Indo-European Myth of Creation. History of Religions 15:
121–145.
Mayr, Ernst (1976). Evolution and the Diversity of Life. Cambridge, MA: Harvard Uni-
versity Press.
Mayr, Ernst (1982). The Growth of Biological Thought: Diversity, Evolution, and Inheri-
tance. Cambridge, MA: Harvard University Press.
Mayr, Ernst (1988). The Ontology of the Species Taxon. In Toward a New Philosophy
of Biology: Observations of an Evolutionist (pp. 335–358). Cambridge, MA: Harvard
University Press.
Sterelny, Kim, and Paul E. Griffiths (1999). Sex and Death: An Introduction to Philos-
ophy of Biology. Chicago: University of Chicago Press.
526 Bibliography
Stevenson, Leslie, and David L. Haberman (1998). Ten Theories of Human Nature (2nd
ed.). New York: Oxford University Press.
Symons, Donald (1979). The Evolution of Human Sexuality. New York: Oxford Uni-
versity Press.
Tooby, John, and Leda Cosmides (1989). Evolutionary Psychology and the Genera-
tion of Culture, Part I: Theoretical Considerations. Ethology and Sociobiology 10:
29–49.
Tooby, John, and Leda Cosmides (1990a). On the Universality of Human Nature and
the Uniqueness of the Individual: The Role of Genetics and Adaptation. Journal of
Personality 58: 17–67.
Tooby, John, and Leda Cosmides (1990b). The Past Explains the Present: Emotional
Adaptations and the Structure of Ancestral Environments. Ethology and Sociobiology
11: 375–424.
Tooby, John, and Leda Cosmides (1992). The Psychological Foundations of Culture.
In J. H. Barkow, L. Cosmides, and J. Tooby (eds.), The Adapted Mind: Evolutionary Psy-
chology and the Generation of Culture (pp. 19–136). New York: Oxford University Press.
Trigg, Roger (1988). Ideas of Human Nature: An Historical Introduction. Oxford, UK:
Blackwell.
Wilson, David Sloan (1994). Adaptive Genetic Variation and Human Evolutionary
Psychology. Ethology and Sociobiology 15: 219–235.
Index
Humans. See also Human behavior; Infidelity, 261. See also Emotional
Human nature infidelity; Extramarital sex; Female
environment of evolutionary infidelity; Sexual infidelity
adaptedness of, 59–60 breaching implicit reproductive
evolution of, 20, 60–61, 93–103 contract, 275–276, 301–302
psychological evolution of, “arms costs-benefits of, 277–278, 305–306,
race” in, 100–103 314–316
sexuality of, 1–2, 6 detrimental to reproductive effort,
universal anatomy of, 424–428 301–302
Hunter-gatherer populations double-shot hypothesis of, 322–333,
adaptive subtasks of, 110 340–341
lifestyles of, 59–60, 83, 94–95 fitness costs of, 342–344
Pleistocene, 59–60 frequency of, 218
variations in male contributions in, gender gap in, 278–280
229 to increase total number of offspring,
Hypothesis testing, logic of, 153–154 280–282
increasing fitness of offspring,
Illegitimate children, maltreatment of, 282–283
395–398 male propensity toward, 279
Imagination, sex differences in, psychology of, 201
340–341 sex differences in, 277–279, 305–306,
Immune system, 140-142 314-316
Incest short-term, 276, 277–278, 280–282,
avoidance of, 62 289–290
likelihood of, 372–374 women’s psychological adaptation for,
taboo against, 468–469 289–290
Inclusive fitness, 350, 353–356, 389- Information processing
390 blurring streams of, 138–139
Income level design of, 69
attractiveness and, 248–249 domain-dominant, 199
mate’s potential for, 238–239 domain-general, 197
projected, 236–237 domain-independent rules of,
status and, 235 149–150
Individuals evolved modules in, 70
cells as, 447–448 of human mind, 64–65
as cohesive whole, 445–446 Information-processing structures,
laws of nature applying to, 451–453, 199–200
453 addition of, 134–135
versus natural kinds, 450–457 Input problem, 159–160
species as, 443–447 Instincts, 128–129
Industrial revolution, 109 Instincts-as-modules, proliferation of,
Infanticide, 397–398 128–129
sexually selected, 411 Intelligence, general-purpose, 86
538 Index