The field invites a lot of bad science.


Credit: Liz Henry, “evolutionary psychology bingo”. Used under Creative Commons CC-BY-ND 2.0.

Evolutionary Psychology is an idea that holds a lot of surface appeal. The basic premise is that human behaviors, drives, and instincts can be explained in terms of natural selection: hunger exists because it kept our ancestors from starving to death; fear exists because it kept our ancestors from being eaten by large predators, or trampled by large prey; sex drives exist because they increased the number of children that our ancestors left behind; and so on.

But Evo Psych has a problem. Too many people—definitely among the fans, and perhaps even among the professionals—think that the following is valid reasoning:

  1. Modern humans do X
  2. I can think of a scenario where X would have been beneficial to early Homo sapiens
  3. Therefore, X is 100% genetic

Disclaimer: I’m not a scientist. If you want a scientist’s-eye view of the problem, “The truth has got its boots on” by Erin Giglio, which includes an explanation of why her field, Behavioral Ecology, doesn’t get along with the Evolutionary Psychology field despite the fact that the two fields study overlapping phenomena.

But I do know enough about biology as an interested layperson that I can explain some of the pitfalls that a lot of Evo Psych fans (and even some Evo Psych scientists) don’t habitually look for.

First, though, we need to establish some background.


In biology, an organism’s genotype is the set of genetic variants that it carries, while its phenotype is the body and behavior that was actually produced by the organism’s developmental process.

The relationship between genotype and phenotype is much more complex than one might suspect. Suppose that gene A affects the average height of a population — you might think of gene A as a “height gene” — but it might also turn out that gene A and height have nothing to do with each other for certain individuals: for instance, individuals where gene B has a particular variant, or individuals who didn’t receive enough nutrients in childhood. And in the other direction, gene C might have two common variants that seem to control only one harmless trait, but then a rare third variant appears, and individuals with that rare variant of gene C have a dozen seemingly-unrelated developmental problems compared to individuals with the common variants.

In a sense, there’s no such thing as a gene “for” a trait: all phenotypes are merely side effects of gene transcription, which is chemistry. (A very special type of chemistry, but chemistry nonetheless.) The genes themselves code for proteins or, rarely, for functional RNA; they never code for actual phenotypic traits. The proteins aggregate together, cleave each other, bind each other, warp each other, and the end result is a pile of roughly person-shaped proteins. The fact that some of your proteins are organized into “skin” and other proteins are organized into “muscle” is irrelevant at the gene level; you don’t have a “skin” gene or a “muscle” gene. There are proteins for collagen and myosin, it’s true, but it took thousands of other genes to organize the development of your body, with your embryonic form producing various proteins to measure time, distance, and direction so that your stem cells could orient themselves in the developing body and identify that they were destined to become, say, muscle cells — and therefore to upregulate the myosin gene.

Evolution doesn’t select for single genes. If a gene does not play nicely with its neighbors, that gene will “lose” evolution, no matter how amazingly adaptive the gene could be in another situation. Evolution selects for collections of genes that work well together.

Phenotypic plasticity is the idea that an organism’s phenotype is deeply dependent on the environment in which the organism develops. If the organism is exposed to environment A, it develops trait X; but if exposed to environment B, it develops trait Y instead. For a species that might experience environment A in one generation and environment B in another, this can be quite adaptive.

But there might not be genes for any of that! One might naïvely expect that there is a gene for X, a gene for Y, and a gene for switching the X/Y genes on and off depending on A/B. But that expectation is baseless: it might be that environments A and B produce slight changes in chemistry that naturally lead to either X or Y; it might be a happy side effect of completely unrelated pathway, but evolution will use it anyway.

All this is to say: just because a trait exists, doesn’t mean that the trait arose because of natural selection on gene variants.


Behavior in the animal kingdom falls along a spectrum between two extremes. On one end, you have animals that need very little parenting because they operate mostly on instinct: a spider hatches from its egg knowing how to spin webs and hunt prey. At the other extreme, you have animals that require lots of parenting because they operate mostly on learned behavior: an elephant spends the first 20 years of its life as a child learning from its mother, and then the rest of its life as an adult learning from the herd’s matriarch.

Why would evolution ever produce an animal that spends its life learning instead of doing? Wouldn’t evolution just encode the necessary behaviors into the animal’s genes? Wouldn’t it be more efficient if the elephant could be born self-sufficient and ready to forage on its own?

Well, no, not necessarily. Environments change: weather patterns shift, droughts and floods happen, new predators move into the area, foods that were once plentiful become scarce and vice versa. In short, behaviors that previously worked can suddenly stop working, sometimes so quickly that individuals live long enough to see both extremes. And a behavior, especially a complex one with lots of steps, might be inappropriate or even maladaptive in the new environment; so it’s not always favorable for evolution to hard-code behaviors into genes as instincts.

We would expect that behaviors are more likely to become instincts as they meet these criteria:

  • The species has a short lifespan, such that an individual is unlikely to see major environmental change over its lifetime.

  • The species inhabits an area with very little variation from year to year; the more predictable, the better.

  • The behavior is broadly applicable to many situations; this probably means that the behavior is something simple, like “sudden noises are scary; run away from them”, not “go this way to find the next watering hole”.

But how might a behavior be transformed into an instinct in the first place?

Well, individuals within a population vary from each other. I don’t necessarily mean that they genetically vary: even if they are all clones of one another, the tiny environmental differences during development will produce slightly different brain wirings, giving rise to the individual preferences that we call “personality”. These individuals will succeed or fail; mostly due to random chance, but partly due to their tiny differences in personality.

So now, introduce genetic variation. Each human being has on the order of 100 novel mutations, and since approximately nine tenths of the human genome is non-coding, that means each human has about 10 mutant genes unique to that individual, plus another 20 gene variants that were novel to their parents, 40 gene variants that were novel to their grandparents, and so on. A gene variant affecting some random protein found in all cells might slightly bias the development process, causing the cells in the brain to connect to each other slightly differently, thus resulting in a different range of possible personalities. So now, not only does personality vary depending on tiny effects in the environment, but the population is overlaid with a dappled blanket of genetic variance that affects each individual’s range of possible behavioral phenotypes.

So now natural selection happens. Suppose that the environment is very predictable, at least on the scale of the organism’s lifetime. Lineages that are more likely to produce well-adapted phenotypes — for instance, more likely to produce a fearful response to predators — will be favored by natural selection, so individuals from that lineage will be slightly overrepresented in the next generation’s gene pool. Likewise, lineages that are more likely to produce maladaptive phenotypes — say, more likely to approach predators with calm and curiosity — will be disfavored.

(But remember: a given lineage will produce both adaptive and maladaptive phenotypes! We’re really talking about the net balance of [number of well-adapted individuals × how well-adapted they are] ÷ [number of maladapted individuals × how maladapted they are]. Evolution might favor a genetic ensemble that results in a weak instinct (“that guy’s shady, keep an eye on him”) that expresses itself in every individual, or it might favor a strong instinct (“HOLY CRAP, RUN AWAY!”) that only expresses itself in a fraction of the population. Which one “wins” evolution depends on how the math works out: you can’t just take a qualitative description of the problem and use armchair theorizing to work out which one wins; instead, you need to get quantitative and look at the actual numbers.)

So now we see that, if the environment is consistent enough and the trait is adaptive enough, natural selection will begin to favor lineages that are more likely to exhibit the trait. A behavioral phenotype that originally arose in some individuals due to environment is slowly transformed into a consistent instinct that arises in all individuals due to genetics.

But if the environment is not consistent across generations, then behavioral phenotypes that are adaptive in one generation might well be maladaptive in the next generation. Natural selection always favors the adaptive phenotypes, but “adaptive” is always in flux, so these evolutionary nudges will come out in favor of lineages that produce both presence and absence of the trait. If there is a “tell” that reveals whether the trait will be adaptive or maladaptive for the coming generation, then evolution will favor lineages that already take advantage of that “tell”. But genes are ultimately chemistry, so only factors that are easily detectable by chemical reactions — temperature is an obvious one — are likely to be present in some lineage’s existing repertoire of genetic variation.

If there are no “tells”, or if they are not easily detectable via chemistry, then randomness will be favored: if individuals might develop in one of two environments — say, “wet” and “dry” — then each individual will have a random chance of being adapted for either the “wet” environment or the “dry” environment. If there’s a 3 in 4 chance for the “wet” environment in each generation and a 1 in 4 chance for the “dry” environment, then evolution will favor lineages that produce both the “wet” phenotype and the “dry” phenotype in a 3:1 ratio.


Now that we have that background, we can talk about behavioral sex dimorphism.

Recall what we said about inconsistent environments. Well, “environment” doesn’t just mean “the ecology in which you develop”, or even “the womb in which you develop”. In evolutionary biology, “environment” is a gene’s-eye view of the world, so the body itself is part of the environment, and therefore so are the other genes that contributed to that body.

Well, one of those genes is SRY, short for “sex-determining region of the Y chromosome”. For various reasons that aren’t important right now, evolution has decided that the best strategy is for about 50% of the population to bear SRY, making them likely to develop into the “man” phenotype, and for the other 50% of the population to lack SRY, making them likely to develop into the “woman” phenotype. For all the genes other than SRY, “man” or “woman” is simply one more piece of the environment that will be considered when natural selection judges how adaptive the individual’s gene variants are.

So, suppose there’s a behavior that’s adaptive for those with the “male” variant of SRY and maladaptive for those with only the “female” variant. Let’s get concrete and suppose the trait is “sexual attraction to women”. Well, this aspect of the environment is fairly stable over the individual’s lifetime. And SRY itself is a chemical signal, so we already have a “tell” as to which of the two common environments this individual will likely exist in. And the trait is fairly consistent about being very adaptive for men and slightly maladaptive for women. So if one lineage expresses “sexual attraction to women” unconditionally, and another lineage expresses “sexual attraction to women” only if SRY is present, then (all else being equal) we would expect natural selection to slightly favor the second lineage over the first. So, although this thought experiment is not evidence that “sexual attraction to women” is an evolved sex-dimorphic behavior, we can see that it is a candidate for one.

But carefully note our line of reasoning: the behavior had to be both adaptive for one sex and maladaptive for the other sex.

Let’s consider another behavior that is purported to be a sex-dimorphic difference: spatial reasoning vs linguistic reasoning. Unlike the previous behavior we considered, this behavior is not something that is clearly maladaptive to one sex or the other: we might suppose that spatial reasoning is more important to men and that linguistic reasoning is more important to women… but we have not made the case that spatial reasoning is bad for women, or that linguistic reasoning is bad for men. That’s an important distinction: if there is no maladaptation, then there is no selective pressure to make the behavior conditional on the environment; we would expect both sexes to have both traits. It would be very surprising if there were truly a difference, and surprising findings must be held to a higher standard of evidence than expected findings.

That’s not to say that spatial reasoning vs linguistic reasoning is not an actual sexual dimorphism! A small but non-zero sex difference on this behavior has been found in multiple studies, so this particular difference is one that we need to treat as a serious possibility — but not necessarily as a fact (the studies weren’t that good).

Well, how can we explain this? What might cause linguistic reasoning to be maladaptive in men? Well, suppose that spatial reasoning was historically more adaptive in men than in women, and also suppose that spatial and linguistic reasoning are tradeoffs of one another: above a certain point, a person with strong spatial reasoning might have weaker linguistic reasoning, and vice versa. Now we have a testable hypothesis: if we were to measure spatial reasoning, measure linguistic reasoning, and calculate how strongly correlated the two are, then we would expect there to be a non-trivial negative correlation. (Note: this is not what the existing studies on sex dimorphic behavior have measured.)

Let’s consider one last sex-dimorphic trait, although this one happens to be physical rather than behavioral. Namely: the ratio of ring finger length to index finger length. No one has been able to come up with a plausible explanation for why this trait would be adaptive for either sex. It seems to be tied to pre-natal exposure to testosterone. The best guess is that it’s a completely selection-neutral side effect: some gene gets used in two or more places during development, and one of those places is sex-dimorphic, so all the other places get a slight tweak in the presence of testosterone even though they don’t need it.


Now we can see the problems with Evo Psych.

A lot of people who promote Evo Psych results do not realize the massive caveats that come with Evo Psych results. Much of the Evo Psych literature consists of measuring whether various traits are found in the same proportions across multiple human cultures. As discussed, just because a trait is found evenly does not mean that the trait has been selected for:

  • If the trait has a strong influence on selection and it is consistently (mal-)adaptive across environments, then phenotype variation will be reduced until the trait is universally present (absent).

  • If the trait has a strong influence on selection but it can be either adaptive or maladaptive depending on a changing environment, then phenotype variation will be increased until the prevalence of the trait matches the odds that the trait will be adaptive.

  • If the trait has little or no influence on selection, then the trait may be wholly random or it may be a side-effect of some random gene’s expression patterns.

As you’ll see, simply saying “yup, this is found in all cultures” doesn’t tell you anything about which of the three categories the trait lies in: you would need to do some fairly subtle statistics to tease that out. The only thing that can be ruled out with confidence is the possibility that the trait is consistently maladaptive.

And trying to measure sex-dimorphism is even harder: not only do you need to confirm that the trait is more prevalent in one sex than another, you also need to propose a hypothesis for why the trait is maladaptive in one sex but not the other, and then you need to show evidence of that maladaptation.

And that doesn’t even get into the WEIRD problem. Most studies are conducted against undergraduate students, not a random sample of the population, because undergrads are plentiful on college campuses; and even of the studies that don’t rely on undergrads, most of them rely on respondents who live in Western, Educated, Industrialized, Rich, and Democratic countries. It’s incredibly expensive to conduct your study against, say, the Pirahã or the Mosuo — odds are that your university department doesn’t have a sister department in the Amazon basin or in rural China, so you’ll have to pay someone to fly there, hire an interpreter, and earn the trust of the locals. In Evo Psych, even the most “cross-cultural” studies rarely actually conduct new research against such non-WEIRD populations. Yet without going to those lengths, you can’t truthfully say that you’ve controlled for cultural variation.

Even physiological studies, such as PET and fMRI scans or cadaver dissections, cannot distinguish environment from biology: it’s well-known that environment can affect physiology, including the physiology of the brain. For instance, certain areas of the brain are drastically enlarged in musicians, and the degree of enlargement correlates strongly with how often the musician plays. The brain isn’t a static thing: it grows and changes as you exercise it, making you get better at the things you actually do. Even supposing that the average man’s brain is measurably different from the average woman’s, it’s not currently possible to tell whether those differences were due to practice, to environment (including culture), or to SRY and its consequences.


If there’s one takeaway here, it’s this: Evolutionary Psychology results are tricky to interpret, because the reader often has a preconception of which questions are being answered by the study, and that preconception can easily be wrong. That’s a risk with all science, honestly, but Evo Psych seems particularly prone to it. It is not always obvious what hypothesis an Evo Psych study is actually testing, especially for a reader not well-versed in evolutionary biology. That lends the field to out-of-context quotes and outright misrepresentation. If you’re about to use an Evo Psych study to back up your point, be careful that you actually understand the science behind it well enough to know what its findings mean.