"Talks a good game about freedom when out of power, but once he’s in – bam! Everyone's enslaved in the human-flourishing mines."

How Do We Get Breasts Out Of Bayes Theorem?

[Epistemic status: I guess instincts clearly exist, so take this post more as an expression of confusion than as a claim that they don’t.]

Predictive processing isn’t necessarily blank-slatist. But its focus on building concepts out of attempts to generate/predict sense data poses a problem for theories of innate knowledge. PP is more comfortable with deviations from a blank slate that involve the rules of cognition than with those that involve the contents of cognition.

For example, the theory shouldn’t mind the existence of genes for IQ. If the brain works on Bayesian math, some brains might be able to do the calculations more effectively than others. It shouldn’t even mind claims like “girls are more emotional than boys” – that’s just a question of how different hormones affect the Bayesian weighting of logical vs. emotional input.

But evolutionary psychologists make claims like “Men have been evolutionarily programmed to like women with big breasts, because those are a sign of fertility.” Forget for a second whether this is politically correct, or cross-culturally replicable, or anything like that. From a neurological point of view, how could this possibly work?

In Clark’s version of PP, infants laboriously construct all their priors out of sensory evidence. Object permanence takes months. Sensory coordination – the belief that eg the auditory and visual streams describe the same world, so that the same object might be both visible and producing sound – is not assumed. Clark even flirts with the possibility that some really basic assumptions might be learned:

Plausibly, it is only because the world we encounter must be parsed for action and intervention that we encounter, in experience, a relatively unambiguous determinate world at all. Subtract the need for action and the broadly Bayesian framework can seem quite at odds with the phenomenal facts about conscious perceptual experience: our world, it might be said, does not look as if it is encoded in an intertwined set of probability density distributions. Instead, it looks unitary and, on a clear day, unambiguous…biological systems, as mentioned earlier, may be informed by a variety of learned or innate “hyperpriors” concerning the general nature of the world. One such hyperprior might be that the world is usually in one determinate state or another.

I realize he’s not coming out and saying that maybe babies see the world as a probability distribution over hypotheses and only gradually “figure out” that a determinate world is more pragmatic. But he’s sure coming closer to saying that than anybody else I know.

In any case, we work up from these sorts of deep hyperpriors to testing out new models and ideas. Presumably we eventually gain concepts like “breast” after a lot of trial-and-error in which we learn that they generate successful predictions about the sensory world.

In this model, the evolutionary psychological theory seems like a confusion of levels. How do our genes reach out and grab this particular high-level category in the brain, “breast”, to let us know that we’re programmed to find it attractive?

To a first approximation, all a gene does is code for a protein. How, exactly, do you design a protein that makes men find big-breasted women attractive? I mean, I can sort of imagine that if you know what neurons carry the concept of “breast”, you can sort of wire them up to whatever region of the hypothalamus handles sexual attraction, so that whenever you see breasts you feel attraction. But number one, are you sure there’s a specific set of neurons that carry the concept “breast”? And number two, how do you get those neurons (and no others) to express a certain gene?

And if you want to posit an entire complicated breast-locating system made up of hundreds of genes, remember that we only have about 20,000 genes total. Most of these are already involved in doing things like making the walls of lysosomes flexible enough or something really boring like that. Really it’s a miracle that a mere 20,000 genes can make a human at all. So how many of these precious resources do you want to take up constructing some kind of weird Rube-Goldbergesque breast-related brain circuit?

The only excuse I can think of for the evo psych perspective is that it obviously works sometimes. Animals do have instincts; it can’t be learning all the way down.

Sometimes when we really understand those instincts, they do look like weird Rube Goldberg contraptions made of brain circuits. The classic example is baby gulls demanding food from their mother. Adult gulls have a red dot on their beaks, and the baby bird algorithm seems to be “The first thing you see with a red dot is your mother; demand food from her.” Maybe “red dot” is primitive enough that it’s easier to specify genetically than “thing that looks like a mother bird”?

The clearest example I can think of where animals clearly have an instinctive understanding of a high level concept is sex/gender – a few gay humans and penguins aside, Nature seems pretty good at keeping its creatures heterosexual. But this is one of the rare cases where evolution might really want to devote some big fraction of the 20,000 genes it has to work with to building a Rube Goldberg circuit.

Also, maybe we shouldn’t set those few gender-nonconforming humans aside. Remember, autistic people have some kind of impairment in top-down prior-based processing relative to the bottom-up evidence-based kind, and they’re about eight times more likely to be trans than the general population. It sure looks like there’s some kind of process in which people have to infer their gender. And even though evolution seems to be shouting some really loud hints, maybe if you weigh streams of evidence in unusual ways you can end up somewhere unexpected. Evolution may be able to bias the process or control its downstream effects, but it doesn’t seem able to literally hard-code it.

Someone once asked me how to distinguish between good and bad evolutionary psychology. One heuristic might be to have a strong prior against any claim in which genes can just reach into the level of already-formed concepts and tweak them around, unless there’s a really strong reason for evolution to go through a lot of trouble to make it happen.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

Leave a Reply

313 Responses to How Do We Get Breasts Out Of Bayes Theorem?

  1. localdeity says:

    There is a theory, put forth by the manga/anime “Prison School”, that the precursors to humans crawled on four limbs, making their buttocks highly visible (and one can imagine evolving the ability to judge mate fitness based on the appearance of someone’s buttocks), and then, after humans started walking on two legs, females developed larger breasts to appeal to the same brain-circuitry that used to judge buttocks. “Boobs are fake butts”.

    Overdramatic anime video here: https://www.youtube.com/watch?v=cKHUgJ5suIA

    • dansimonicouldbewrong says:

      This hypothesis was also offered in Desmond Morris’ “The Naked Ape” (1967).

      • Deiseach says:

        First, I think it’s entirely possible that “men like big boobs” is culturally determined; there has been some suggestion (and it’s entirely possible that this work has been done with an agenda in mind so take it with a pinch of salt) that in societies where female breasts are on display, they are not treated as sexually attractive in the same way; the way that men walking around bare-chested in our society is considered normal (for a certain set of circumstances) and not considered likely to drive women into frothing fits of lust because of the context. (And let me inform you, in the correct context, women can find men’s naked chests every bit as provocative).

        In this model, the evolutionary psychological theory seems like a confusion of levels. How do our genes reach out and grab this particular high-level category in the brain, “breast”, to let us know that we’re programmed to find it attractive?

        Secondly, I think it’s entirely possible that this comes out of breast-feeding – you know, the primitive way babies were fed before the invention of formula? Large breasts are more generally milk-filled breasts, and (A) a woman with a plentiful supply of milk will be able to ensure her infant does not die of starvation at an early and vulnerable period (B) large breasts will clearly be a signal of fertility since the woman has big breasts due to lactating, hence she has successfully reproduced.

        Children of women with plentiful milk supplies surviving over those whose mothers cannot feed them will, given a long enough time, find big breasts attractive because it means survival. (Small-breasted females will trigger the ‘risk of starvation because not enough milk produced’ model). The same selection pressures that give us the peacock’s tail will then give us human females who have fat deposits in their mammary glands so that they emulate milk-filled breasts, even when the females are not lactating.

        Males will find these sexually attractive (with a bit of social pressure along the same lines as the ‘rats in lingerie’ experiment) et voilá, we end up a couple of hundred thousand years down the line inventing cosmetic plastic surgery to give ourselves bigger boobs than nature provided! (Heterosexual women, being in the possession of breasts themselves, will not find them sexually stimulating for the same reason heterosexual men do not go “oh, he’s got a nice penis” and find it sexually stimulating).

        • doubleunplussed says:

          Scott understands the evolutionary pressures, but he’s wondering about the mechanism. How you pre program a concept like ‘breast’ into 20,000 genes? The whole human genome is barely enough data to make a blurry jpeg of a breast.

          • Loris says:

            Hang on. Human haploid genome : 3×10^9 base-pairs.
            A base-pair has the coding content of 2 bits, so the coding potential is about 700 megabytes. Exactly how sharp do you need the detail on this breast image?
            Okay, so you only counted the proteins, which are about 1-2% of the genome. That’s still probably enough capacity for most of the purposes pictures of breasts are used for.

          • doubleunplussed says:

            My bad: I mistakenly read “20,000 genes” as 20,000 base pairs. Which would be insane, but I thought that was the point being made, that it was insane.

          • You would only need to encode the visual outline of body shapes and breasts, if that helps lower the size of information needed. I do remember that baby ducklings react totally differently to shadows with different shapes, cowering when they recognize the shape of a hawk, for example, so it’s within the range of possibility that the basic outline of an averaged female or an averaged male is hard coded in.

          • MugaSofer says:

            Are all blind people bisexual, then?

        • Nyx says:

          > Large breasts are more generally milk-filled breasts,

          This is untrue, even though all mammals lactate, most mammals do not have the same large deposits of fatty tissue around their nipples as human women do.

          • Randy M says:

            Although the reverse is true; lactating increases breast size.

          • Deiseach says:

            I did not mean “large breasts have milk inside”, what I meant was “in our flat-chested period of hominid evolution, having large(r) breasts meant having milk-filled breasts” and thus large permanent breasts that result from fat deposits are mimicking milk-filled breasts due to lactation and so on to the conclusion.

            For feck’s sake, lads, have none of ye seen a cow with swollen udders ready to be milked? Actually, that’s a good example; we’ve bred domesticated cattle for dairying to produce much more milk than a calf needs, so they have large, prominent udders that wild bovines don’t – see this quote from the Wikipedia article on the aurochs:

            Even in carrying cows, the udder was small and hardly visible from the side; this feature is equal to that of other wild bovines.

          • baconbacon says:

            I did not mean “large breasts have milk inside”, what I meant was “in our flat-chested period of hominid evolution, having large(r) breasts meant having milk-filled breasts” and thus large permanent breasts that result from fat deposits are mimicking milk-filled breasts due to lactation and so on to the conclusion.

            The longer the mimicking goes one the weaker the selection pressure becomes.

        • dansimonicouldbewrong says:

          FTR, I wasn’t endorsing Morris’ hypothesis–only providing an old citation. I also find it rather unconvincing–it seems to me to confuse “breasts” with “breasts as displayed in modern Western fashion”.

        • mwengler says:

          The milk theory seems quite wrong to me. 1) I was not breast-fed, I *LOVE* female breasts in terms of getting turned on by them. 2) Clothed breasts are very exciting too. 3) Any hint at all that men with large or small breasted mothers have any difference in how attracted THEY are to female breasts?

          • Deiseach says:

            1) I was not breast-fed,

            YOU weren’t. Gazumpty generations of your ancestors were, where milk from your mother or another female was the only food source in early infancy. This is a bit like you saying “I never learned to walk a tight-rope but I can walk a chalk line on the ground, how come that so?”

          • MugaSofer says:

            But the question Scott is asking isn’t “do men like breasts because of evolution”, but “how does evolution make men like breasts (assuming it does)?”

          • Careless says:


            Did you forget that you were advancing a cultural explanation, which means his ancestors could have nothing to do with his preferences?

        • Speaker To Animals says:

          And let me inform you, in the correct context, women can find men’s naked chests every bit as provocative

          Coke commercials?

        • Winter Shaker says:

          Large breasts are more generally milk-filled breasts

          Going on what I remember from reading Natalie Angier, it’s not merely that most other mammals are perfectly capable of producing enough milk without having human-style breasts, but even among humans, there are no significant differences in quantity of milk produced between large- vs small-breasted women – every gets about the same amount of milk gland apparatus, but gets wildly varying amounts of fatty tissue that the milk glands are embedded in. (And our closest relatives the chimps and bonobos only have swollen breasts while lactating and are flat-chested the rest of the time).

          She makes a good case that the human female breast really is an aesthetic/sexual-attraction adaptation akin to the peacock’s tail feathers, thus leaving it as something of a mystery to be explained by something less utilitarian than milk production.

          • Cliff says:

            You guys all seem to be missing the point of the above comment. It’s not that fat = more milk, it’s that fat is designed to mimic a lactating breast with a large milk supply. If all women were originally flat-chested, large breasts would be a reliable sign of a good milk supply, therefore women evolved to mimic this feature with fat deposits.

          • Winter Shaker says:

            But in our close relatives, and presumably in our proto-human ancestors, swollen breasts would have also signified ‘currently pregnant or lactating, therefore not currently fertile’. The mystery is how something that in other species was a sign that a particular female was not currently ‘worth’ having sex with, in our species came to be mimicked by an adaptation that specifically does create sexual attraction.

          • baconbacon says:

            You guys all seem to be missing the point of the above comment. It’s not that fat = more milk, it’s that fat is designed to mimic a lactating breast with a large milk supply. If all women were originally flat-chested, large breasts would be a reliable sign of a good milk supply, therefore women evolved to mimic this feature with fat deposits.

            A lactating woman has lower fertility while she is breast feeding, and has already found a mate. Women don’t lactate and then have a kid, its the other way around. Additionally women’s breasts often swell late in pregnancy when they are at their least fertile (basically zero chance of getting them pregnant again and having it come to term).

          • John Schilling says:

            Human infants being unusually vulnerable for an unusually long time, much of our sexuality is based around encouraging fathers (or, if not available, any other suitable male) to hang around with new mothers in spite of the screaming babies. It is hardly surprising that human males seem to come hard-wired to find a feature unusually prominent in new mothers to be sexually attractive.

          • Deiseach says:

            Yes, but while “lactating means less likely to get pregnant if we mate now”, it also means “is fertile and has proven it by successful pregnancy”, so that also signals “worth investing my genetic capital in” over “flat chested – may be fertile, may not be, may not successfully carry pregnancy to term, may not be able to feed infant after delivery, etc.”

          • baconbacon says:

            Yes, but while “lactating means less likely to get pregnant if we mate now”, it also means “is fertile and has proven it by successful pregnancy”,

            Then it also means “has already found mate, will probably have major conflict over my attempts with him”, also women’s nutritional requirements are higher than usual right now, will have to provide more calories for her, helping a rivals offspring until she is done feeding.

          • baconbacon says:

            Human infants being unusually vulnerable for an unusually long time, much of our sexuality is based around encouraging fathers (or, if not available, any other suitable male) to hang around with new mothers in spite of the screaming babies. It is hardly surprising that human males seem to come hard-wired to find a feature unusually prominent in new mothers to be sexually attractive.

            I like this as potentially part of the answer, I think the actual mechanism goes further down the line.

            Humans are generally more visual than most animals compared to smell oriented, and this shifted some of our mating practices to physical markers away from pheromone markers. Other heavily visual species, like birds, also have strong visual markers for mating purposes.

            There are a lot of potential advantages here, pheromones cast a wide net. If a female is in estrus she will often attract males from miles around, a visual cue like coloring, or breast size, can be directed at individuals far more easily and allow for more female mate selection. As John Schilling notes, long term bonding is more important for humans than your typical animal (lots of birds also form long term bonds), which enhances the value of female mate selection.

        • Difference Maker says:

          Ridiculous. But you obviously don’t like boobs

      • Johannes D says:

        Obviously he stole it from the anime.

      • MugaSofer says:

        IIRC The Naked Ape suggests that breasts (and red lips) are intended to mimic the presented buttocks and vulva prior to sex, not some quadrupedal protohuman ancestor. Similar idea, though.

    • eclairsandsins says:

      Prison School confirmed rationalist fic

  2. jeff daniels says:

    Interesting. What if we take the genes-can-encode-attraction-to-breasts idea to its logical conclusion, despite the 20,000 genes limitation? We can hypothesize that:

    1. lots of different genes influence the polygenic “attraction-to-breasts” trait
    2. those same genes also influence thousands of other different cognitive processes.

    In other words, to mix creature metaphors, what if the genes that are supposed to encode “attraction-to-breasts” in the relevant neurons are also supposed to encode “red dot = mother” in its respective neurons?

    It could be that the reason evolutionary psych seems to attract so much bunk research is that all of our evolved brain-process traits are connected to each other. This would mean that isolating particular genetically-determined cognitive traits is extremely difficult, so the data might be so unclear that it’s easy to support more or less any hypothesis.

    • Worley says:

      That’s a very good point. IIRC, they’ve done “genome-wide association studies” on “educational attainment”, and there are hundreds or thousands of genes which show noticeable correlation with how much education a person gets. (I don’t know how they get rid of confounding associations.) Supposedly you can see a secular trend in the genetically-predicted educational attainment in the general population. But it would hardly be surprising if thousands of genes had various semi-random effects through the neural network, and increasing the relevance of 1,000 alleles by 1% would significantly change behavior in a selected-for situation.

  3. AnonYEmous says:

    let’s edit this comment up a notch

    here goes: it’s probably likely that “big breasts” is lumped in with all the rest of “mate selection” where evolution does have a strong reason to devote genes to. Also, of course evolution can’t hard-code things in, there are mutations.

    • jeff daniels says:

      Seems a bit of a stretch. Considering that men are also disproportionately attracted to butts, and certain waist-hip ratios or whatever, and women are disproportionately attracted to certain chest size/proportions (and butts), i’m not sure i buy the idea that those attractions are partly based on non-sex-related learned behavior like infants nursing.

      • Ketil says:

        Seems testable: do men who weren’t breast fed more often turn out gay or at least, less interested in breasts?

        My guess would be that we (most of us) have hard coded gender categories, including a sexual identity and a sexual preference. Something turns us (again: heterosexual men) on to women – and anything we associate with women (including breasts or hip-to-waist ratio) will trigger this. (Btw, another thing that appears hard coded by evolution, is our liking for sugar and fat)

        The question is how flexible these mechanisms are. Different cultures present and emphasize gender identities differently, it would be interesting to see to what degree this affects what is perceived as attractive. Again, this should be possible to get some data on.

        • Saint Fiasco says:

          Maybe evolution spent a lot of genes to encode a high-level concept of “opposite gender”. Seems like the kind of thing that may be worth spending a lot of its limited budget on.

          Then the low-level characteristics like breasts and waist ratios become attractive in men because they are used to infer opposite gender.

          • @Saint Fiasco

            So perhaps the sexuality algorithm is something as simple as “check your own physical and behavioral characteristics and then be attracted to the sex that has average characteristics opposite your own (or closest to for homosexuality)”?

        • Ozy Frantz says:

          In communities with a lot of trans people, I’ve noticed a lot of diversity in what straight men consider to be an appropriate target of their sexuality. Only people with estrogen-dominant hormone systems; only people with vaginas, but if you otherwise look like a man that is awesome; only people assigned female at birth who look like girls; only femboys, trans women, and feminine-presenting cis women; only interested in people who identify as women whatever they look like; etc etc. I think it is probably somewhat unlikely that there exists a “only interested in assigned-female-at-birth people” gene.

          • eyeballfrog says:

            That seems like variation in where to draw the line on the edges. The target group is clearly cis women, and different men have different disqualifying factors. That suggests there’s *something* that says “put positive attraction weight on these features that women usually have that men don’t”, but it doesn’t always get the weights the same.

            Perhaps it’d be worth looking into any correlations between the backgrounds of the men who have similar weighting profiles.

          • Skivverus says:

            For a specific example, I’d probably count myself as part of Ozy’s second group, but there are several distinct factors that code to me as “feminine” – breasts aren’t so much on the list, but faces are (half-Asian, so that’s apparently stereotypical), as are voices, personalities (I believe I can understand the “brony” impulse), and handcuffs scents.
            There are distinct qualia associated with each of these, even if each of them contributes to the single category of “person I’d want to date”.

          • MugaSofer says:

            Seems like people start out attracted to “women” or “men” (insert handwave for bisexual people here), but then their brains construct different definitions. One straight man defines [sex-I’m-attracted-to] as “has breasts”, one as “has vagina”, and a third defines it as “dresses a certain way”. (Similar things seems to be happening with non cis-by-default people.)

            And yet somehow evolution has succeeded at arranging for most cis men to target cis women most of the time, and vice-versa. How does evolution reach in and find a high-level concept like “man” or “woman”, that’s seemingly being independently constructed by each person?

            The phenomenon of “prison gays” (and sailors, people stranded on islands etc) could be a possible clue – seems like if you restrict people to a male-only or female-only environment, their mate-selection circuits get hypersensitive. Study of the mechanism people use to find partners in unisex environments could be worthwhile here. Ancient Greek homosexuality, common but with strict roles, seems like another possible clue.

          • Anthony says:

            First: Restriction of range problem. Communities with lots of trans people will not have a representative hetero/cis population.

            Second: That large range of interest happens in communities without trans people, too. The distribution is different, but straight men have a fairly wide variety in their attractions in the communities with which I’m familiar, which mostly have very few trans people. (Most have some visible gay and bi people.)

        • Jaskologist says:

          Approach the problem like a programmer.

          You’re an angelic DNA coder working on the latest iteration of the red-striped pteranodon project. You’ve already built out a beak that is shaped just perfectly for catching the fish in this particular habitat, and you’ve moved on to the reproductive system.

          Now, the stuff about how it should rub its cloaca against the other pteranodon’s cloaca is very old, solid code by now. No need to mess with that. But how do you get it to find those cloacas in the first place? This is tricky. Your critter needs to be able to identify red-striped pteranodons somehow. It also needs to be able to distinguish them from the purple-striped pteranodons in the area (that merger has not yet been approved by the the higher-ups).

          Now, you could hard-code in something that recognizes pteranodon shapes and looks for red stripes. But this is not your first rodeo. As a seasoned coder you know that your design is going to evolve over time. Eventually marketing is going to want the red stripes in green, or branch off into a speckled pteranodon, and you’re going to have to rewrite this subsystem every time. What to do?

          The answer: imprinting. Your product isn’t going to change that much from generation to generation, and it’s going to have parents raising it. Tell the mating subsystem that, whatever the parents look like, that’s what you look like, and that’s the sort of thing you should mate with. Add a few extra bits for distinguishing between those with ZW and ZZ chromosomes and you’re done; you can leave that code untouched for pretty much the next several millennia, no matter what your critter may evolve into.

      • Deiseach says:

        Permit me to inform you, jeff, that “infants nursing” come about as a direct result of “sex-related behaviour”.

        What I find “a bit of a stretch” is that reproduction is treated as distinct from sexual attraction here, but I suppose that is because of the modern culture where sex is for pleasure in the main and very much divorced from “sex makes babies” unless and until a decision to have a baby is made by the parents. So moderns looking at this will automatically think in terms of “what makes me horny” and not “what makes me horny and gets us a baby in nine months time”. In our evolutionary history the two were intertwined.

        • Adam Lein says:

          Also I don’t think that nursing is a learned behavior. WHO recommends breastfeeding within 1 hour of birth; not a lot of time to watch other infants nurse and decide to try it out. My personal experience is that newborns catch their breath and then immediately start smashing their face into anything that looks like a dark-skin-tone-circle against a relatively-light-skin-tone background.

          • eyeballfrog says:

            There’s also the rooting reflex in newborns, which causes them to try to suck on anything that brushes their cheek. This reflex goes away relatively quickly, so it appears babies do learn what breasts look like. I didn’t realize newborns had a visual cue reflex like the one you describe, though.

          • Cliff says:

            They don’t. Newborns will mash their faces into literally anything if they are hungry (and they are). I think they might have a preference for skin though.

  4. gorbash says:

    Let’s not talk about men and breasts, because that’s going to attract comments we don’t need.

    Let’s talk instead about peahens, which are attracted to those peacocks with the giant feather displays.

    I don’t necessarily believe that there’s something in the peahen’s brain that connects “big feathers” to “sexually attractive”. Evolution can do some wacky things when it’s motivated by deciding who gets to reproduce, but as Scott says, it’s not clear how evolution is going to find a “big feathers” neural cluster.

    On the other hand, I think the peahen’s brain has learned that “big feathers” implies “successful and healthy”. And I do believe that “successful and healthy” is connected to “sexually attractive” in the peahen’s mind. That sounds like exactly the sort of thing that evolution would do.

    • moridinamael says:

      I somehow feel like it’s not a coincidence that a peacock with its feathers spread really strongly reminds me of those creepy Google deep dream images. Abstract away what you’re really looking at – it’s an arc of vivid constrasting colored concentric circle patterns. I’ll bet there is a pretty low-level feature detector that goes off when it sees something like that.

      • Douglas Knight says:

        One theory of peacock signalling is that they are signalling a large encumbrance. But a better theory is that they are signalling freedom from parasites by exhibiting a large canvas on which to paint simple patterns in bright colors, in which small errors might be easily visible.

    • engleberg says:

      A peacock with big healthy feathers is less malnourished and disease-ridden than a peacock with shrivelled, mangy, tiny feathers. A woman with big fat knockers is better-fed and less diseased. ‘small errors easily visible’- well put. But maybe a little negative? ‘Small triumphs, easily grasped’, not to go full Richard Blade.

      When we go back to long skirts and full veils guys will trigger off ankles again.

    • Gerry Quinn says:

      Isn’t the problem that you have to encode “big feathers” somehow, whatever it represents?

      This all makes me wonder why all creatures don’t depend more on pheromones, which probably are quite easy to make with a few genes – you just need a factory to make a unique chemical, and a receptor in the nose to detect it and cause the secretion of some sort of hormone that sex-related systems can then glom onto.

      • baconbacon says:

        Lots of species do you pheromones, but they have a major drawback. They require the target to have the specific pheromone receptor. A female can’t just wander out and go “come and get it boys, I am super stinky”, the boys have to be looking for that particular scent. This makes getting on the path tougher, and also getting off the path for a more favorable strategy tougher.

  5. Ashley Yakeley says:

    Newborn babies prefer to look at faces than at other things: surely this must be innate?

    • tentor says:

      There’s a region in the brain specifically responsible for face recognition. How do you encode that in a gene?

      • Amy says:

        The more I learn about interactions between genes, the amount of interactions, genes regulating other genes almost at random through random chemical attractions that turn out to be statistically significant, generic neural structures with a few slight nudges that evolved to just happen to make them settle into the right chaotic attractor … the more I feel that genes are like some string that, by miracle, cryptographically hashes into a valid program. Okay, it’s not like that, not at all. There is order and specific functions and changing one gene doesn’t flip all the outcome measures by 50%. But it’s closer to that than the worst human-made self-modifying spaghetti code we’ve ever made. So it wouldn’t be a gene encoding facial recognition, it would be some gene, or more likely system of genes, that is promoted at the right place at the right time(due to other genes, which you would never guess were even involved in brain development at all), to start a Rube Goldberg chain of events that ends with pushing a developing part of the brain towards an attractor that ends with it being easily wired to recognize faces.


        What we describe as “the gene for X” doesn’t mean that the gene is a mechanism that does X, just that the gene is a necessary part, a weak point if you will, in a much larger process that does X. For instance, the SRY gene might be called “the gene for maleness” – if it’s present (due to a Y chromosome, or a mutation that copied it to an X chromosome, in which case we get an XX male) – it triggers an extremely complicated process that ends up with the person developing as male (of course, any part of the process can fail at any point if there’s lack of a necessary gene or an external influence – see, intersex and trans people). But there are other necessary genes, like the gene coding for the androgen receptor, which, if absent can lead to someone having an SRY gene still developing completely female. Neither of these genes contain any information for building a male body, much less a blueprint. That’s done by a poorly understood, complex mechanism that involves much of the human genome at some point. What they are is on-off switches, choke points in development – we can call them “genes for maleness” but that doesn’t mean they actually give any implementation details about how to build a male body. (And note, that sex is one of the easiest cases – we have SRY because it’s adaptive to have a genetic male/female switch to balance out sexes among children, and we have AR so a cell equipped with mechanisms to develop either male or female could receive hormonal instructions which way to go. For face recognition, there’s no need for either of these. So any genes that change facial recognition will be random mutations in the machinery with a lot of side effects, not a conveniently evolved on/off switch)

      • Ketil says:

        What Amy says, but also that there is hardly ever “a” gene for something, almost everything is highly polygenic, meaning outcome depends on a number of genes interacting (among themselves and with the environment) in interesting ways.

        Think of Legos, there are a lot fewer than 20K pieces, but you can build a lot of different structures with them.

      • Murphy says:

        That’s the tip of the iceberg.

        There is no single gene for an elephants trunk. How do you encode a zebras stripes in genetics, how do you encode for a brain that will be interested in high fat high carb food but find most common environment poisons to taste terrible? how do you encode the desire to drink fresh water but an aversion to drinking salt water?

        how do you encode a fear of heights while looking down but not up? how is that even encoded?

        how do you encode immune cells to move towards infection sites rather than away from them?

        how do you program a single cell to divide in such a manner that you fairly reliably end up with separate fingers with the bones on the inside?

        We have examples of people and animals in which some steps are broken so we have small clues at certain steps that will leave someone with, say, no natural fear of heights who will lean all the way out a window without any part of their brain screaming “Nooo!” or people born with merged fingers but there’s countless incredibly intricate processes that tend to yield quire reliable results.

        There are answers to some of these questions but most of biology, trying to extract what’s encoded from the gene is about a billion times harder than working out how this circuit worked with some vital parts disconnected from the circuit:


        When your eyes manage to work thanks to a set of incredibly intricate nerve growths that somehow manage to yield a functioning image capture and processing system…. the fact that part of your brain matter is particularly good at picking up info about your own species from the captured images is not the most remarkable part of the whole process.

        • Deiseach says:

          there’s countless incredibly intricate processes that tend to yield quite reliable results.

          Going off at a tangent, but this is part of why I’m so blasé about AI existential threat: the assumption from the engineering side seems to be “well, of course once we get hardware/code advanced enough, it’ll just happen that working intelligence will spring forth cleanly like Athena from the head of Zeus” and I can’t help imagining the hollow-eyed biologists looking up from their maps of protein structures and going “Will it? Will it really?”

          • Murphy says:

            That’s the terrifying thing though: that incredible intricacy is almost certainly some local maximum reached by little more than a hill climbing genetic algorithm running massively parallel for a few billion years without even vaguely intelligent guidance.

            If which hints that if a researcher makes some unusually good assumptions then sets their experiment on a similar search it’s entirely imaginable that they could hit on a might more optimal sollution.

            My background is computing but I’m working in biology and while biology astounds me that doesn’t mean it always finds the best sollutions. Gm crops can be given much improved pathways etc.

          • Murphy says:

            I think I should clarify, understanding the designs and networks that evolve is like trying to un-mix tea and milk.

            Looking at an evolved system and trying to figure out how it works is a nightmare but that doesn’t mean we can’t engineer a developing system nor use the same mechanisms to pull the same trick but faster.

          • MNH says:

            @Murphy, I suspect you already know this, but I’d like to point out it’s a bit unfair to characterize natural selection as a hill-climbing algorithm, since genetic drift adds stochastic elements that will, in the long run, allow it to escape some local optima in favor of further, better ones.

          • onyomi says:

            I recently read “Metaphors we Live by,” by Lakoff and Johnson and was struck especially by how much our brains do with so little: almost all our “base” words and concepts have to do with manipulating our physical bodies through a spatial world of people and things. We are very bad at thinking in pure abstraction without relying on heuristic metaphors like “bigger=more significant,” “up=happy; down=sad,” “time is a moving object or something we move through; also an exhaustible commodity,” and so on.

            Yet look at all we do with these “avoid danger while seeking food and sex” machines (interestingly, just learned Lakoff also has a book called “Women, Fire, and Dangerous Things”).

          • Murphy says:


            Very true but you get my point. while there’s lots of local minimima between pond scum and the human brain they’re all escapable.

        • AnthonyC says:

          Everything I know, I know from Sid Meier:

          Remember, genes are not blueprints. This means you can’t, for example, insert “the genes for an elephant’s trunk” into a giraffe and get a giraffe with a trunk. There are no genes for trunks. What you can do with genes is chemistry, since DNA codes for
          chemicals. For instance, we can in theory splice the native plants’ talent for nitrogen fixation into a terran plant.

          —Academician Prokhor Zakharov,
          “Nonlinear Genetics”

      • TheRadicalModerate says:

        By ensuring that all of the various feature-detectors needed to recognize faces all project to a common area. So you don’t grow a “face-detector”. Instead, you grow a region that’s uniquely hooked up to other regions so that the likelihood of detecting faces is extremely high.

    • moridinamael says:

      Maybe it’s because the moving shape of the human mouth appears to sync up with the modulating auditory input of the voices the baby is hearing, and the baby uses this observation to form its initial hyperparameters?

  6. Sticking with the topic of breasts: are there any psychological traits or disorders that impair an infant’s ability to recognize them and/or to instinctively latch on once the recognition happens?

    • Ozy Frantz says:

      Human babies have some instincts but are often taught by their parents how to latch properly. (This is part of the reason breastfeeding failure is so common these days– many parents don’t know how to teach their babies to latch.) Intellectually and developmentally disabled babies and babies with low birth weight often have more trouble. The sucking and rooting reflexes might not be present if the baby is premature or has severe neurological damage.

      • Mary says:

        For mothers, nursing is a learned behavior. A gorilla raised in a zoo could not nurse her firstborn. When she became pregnant, the zoo got a bunch of nursing mothers to come and nurse their babies outside the cage. She could nurse her second.

    • albatross11 says:

      I don’t know about disorders, but I do know that it’s often quite a pain to get nursing to work out. I was quite surprised by that with my kids, particularly my first–my wife had a “lactation consultant”[1] spend quite a bit of time with her, and there were all kinds of common problems and proposed solutions floating around. Now, successfully nursing is absolutely on the critical path, evolution-wise–if you can’t get it to work, you die/lose your offspring. And yet, it’s not all that easy to get to work. (But then, childbirth is even worse.)

      [1] This was always some nurse who’d also breastfed in the past, and probably involved some moderately complicated training and certification, but I’m pretty sure that the same role has been played by moms, older sisters, and older female friends since there have been people. And this makes me wonder if the reason you have complicated advice in books and lactation consultants to get this to work now is because we used to almost always live in close-enough communities that it was pretty-much never a problem for a new mom to find some other woman who’d breastfed in the past and could help her out.

      • mwengler says:

        I think it is entirely resonable that once humanity develops the ability to pass stuff down memetically (i.e. other people helping you learn how to do stuff) that genetic mechanisms helping transmit that info, or even just making it simple enough that you stumble upon it randomly, begin to wither. I hypothesize a certain amount of “genetic energy” that has to be economically used: once we have the technology for glasses we don’t have to kill of people who are smart who are also near-sighted, it becomes worth the genetic energy to keep the smart because people who make glasses will help with survival. Or put another way, the “environment” for which we must be “fittest” includes all the other humans helping us, which is a moving target itself as human technology improves.

        Humans have a hard time with babies compared to a lot of other mammals. Baby heads are really big so they get born way more immature than other mammals, and we can get away with that because our big heads come along with the ability to care for babies in way more complex ways that smaller headed mammals that just rely on biology for everything, or at least for many more things than we do.

    • Randy M says:

      Training babies how to latch is more about making sure the nursing doesn’t discomfort the mother than making sure the baby gets fed, afaik.
      A hungry infant can be temporarily soothed by sucking on a fingertip. I’m not familiar with studies about whether or not infants exhibit suckling responses to abstract pictures of breasts–they’ve probably been done, though.But shape and texture is clearly a part of the suckling reflex.
      Also, newborns have been known to crawl on their mothers belly towards her breasts.

      • Ozy Frantz says:

        Poor latch can, in fact, lead to the child being inadequately fed.

        Newborns crawl towards her breasts based on the smell; they recognize the scent of their mothers’ breast milk.

      • sconn says:

        A bad latch results in not just pain but a plummeting milk supply. So it really is necessary for survival for newborns to latch correctly. I’m into natural parenting and was told that there was no need at all to guide the baby to nurse — that if he was left completely undisturbed, he’d crawl up to the breast and latch on. And I guess this does happen, but … none of mine ever did. If I had left them to figure it out on their own, they’d probably have starved to death. They got the whole “oh this is a breast, let’s suck on it” bit, but not the “you really have to open your mouth REALLY wide” bit, so they weren’t getting milk that way. They needed me to help them out.

    • moridinamael says:

      Setting aside cases where the baby has issues with rooting in the first place, babies will actually seek to latch onto anything that you poke them in the cheek with. They don’t actually need to recognize what they’re sucking on as a breast. This is, in fact, why pacifiers work.

  7. Joe says:

    This seems like such an incredibly basic challenge to the evopsych/modular mind/etc view that I’m sure there must be a standard response to it. I can’t believe that nobody’s ever thought about this before.

    How much design detail can you get from a single gene? How does it scale as the number of genes increases? How do we know?

    • Leonard says:

      Presumably everyone agrees that the non-brain parts of us are encoded in genes? There’s an awful lot of design detail there, too. So I really don’t feel this is much of a challenge for evopsych.

      There’s also the matter that animals do have instincts which include all sorts of “mental” traits. And that with brains that are (seemingly) far less complex than ours. How do you encode “build a spider web” in genes? How, for that matter, do you encode mating instincts in spiders? And yet obviously instincts for building webs and mating are coded in there somehow.

      • petealexharris says:

        It’s certainly a challenge to any specific claims that a human behaviour is more genetic than learned, absent evidence from twin studies.

        We know humans learn lots of things we can’t possibly have needed in the ancestral environment, so we can’t ignore the possibility of learned behaviour in anything humans do, even if we think we have a story for how it might have evolved. Knowing instincts exist but not how they work is not evidence that instincts are responsible for any particular thing.

      • Nancy Lebovitz says:

        We’re not completely encoded by our genes– the pattern of small blood vessels develops in a sensible way rather than being encoded in detail.

      • Aapje says:


        A lot of patterns in nature are clearly based on fairly simple functions though, like fractals. You don’t need too much storage to encode a function.

    • albatross11 says:

      The same challenge works for absolutely any complicated evolved thing–not just behavior, but the functioning of organs, the complex structures of the body, etc.

      • Joe says:

        Yes, it does! So there seem to be two possibilities. One is that the argument “20K genes can’t code much complexity” is wrong. Perhaps it vastly understates how sophisticated each protein actually is, or it ignores that proteins actually form building blocks for a much wider variety of more complex mechanisms, somewhat like Lego bricks, as someone analogised them to elsewhere in the thread. Or maybe both of these.

        Alternatively, perhaps we’re vastly overestimating how complex most organs, behaviours, etc. are. Maybe far more behaviour is learned rather than innate than we might otherwise expect; maybe organs are complex in the same sense that the Mandelbrot set is “complex”, i.e. it isn’t, it just looks like it if you don’t understand how it’s created.

        But in either case, some claim or other is quite wrong. Exciting!

    • Eli says:

      This seems like such an incredibly basic challenge to the evopsych/modular mind/etc view that I’m sure there must be a standard response to it. I can’t believe that nobody’s ever thought about this before.

      Honestly, the correct response is to admit that the genes can’t be coding a gazillion-and-one innate modules, and that we should probably look for simple learning algorithms that can pick up the Evolutionarily Right Thing with only a slight bias.

    • mwengler says:

      I’m on board with this comment. The complexity of bunches of other inherited things is astonishing, binocular vision producing depth perception, recognizing objects in our visual fields under all sorts of variations of coloring, distance, orientation, lighting conditions (including speckled lighting). ATP-ADP cycle. The process of Mitosis and Meiosis making reliably very good copies for millions of generations. Insulin. An immune system that adapts to new threats. Healing of tissues (the ones that heal anyway). Temperature and pressure regulation in bodies. By comparison to the structure of a working wing, is having the ability to feel sexually, uhm, titillated by the slightest suggestion of breasts all that much of a burden on a system?

  8. isotropy says:

    The genetic code has to include fundamentals of the body plan so it can build the body. The sexual attraction code could be using a subroutine with a pointer to the “body plan” section of the genome. That would make it possible to encourage the neurons to build up attraction to MOTAS based on the right body plan without having to hard-code each specific body plan in the attraction function. Then we’ve reduced the task to “recognize members of own species from sensory data”.

    • Charles F says:

      The sexual attraction code could be using a subroutine with a pointer to the “body plan” section of the genome

      I don’t think this actually makes sense. Yes, there is a body plan but it’s in genes, and we don’t look at somebody and associate it with a set of genes. It seems basically impossible to me that a section of our genes could be coded to interpret another section of our genes and build a model of the traits those genes encode out of neurons.* Whereas the idea of two small clusters of built-in models of traits, one of which gets hooked up to sexual attraction by puberty, seems plausible.

      *ETA: some explanation of the intuition behind this. Genes are a programming language based on a really weird niche in chemistry. We can encode a human with 20,000 of them because the stuff involved in DNA is one of the best possible spaces in chemistry for building living things. (or at least one of the best places that was simple enough to find by random chance given a few billion years) A huge portion of the work of building a person is just finding that convenient self-replicating quirk in the laws of physics. And converting a section of DNA into neurons would mean building a gene interpreter (read: chemistry compiler?) in genes, which doesn’t even seem remotely possible.

    • John Schilling says:

      The genetic code has to include fundamentals of the body plan so it can build the body.

      Unfortunately for this theory, biological organisms don’t build themselves the way human engineers build things. They build themselves from the inside out, in a highly localized and decentralized fashion, which requires a peculiar sort of “blueprint” that doesn’t actually include anything resembling a drawing of the finished object.

      You get lots of instructions like, “If you are a developing Type TG678L1 cell and between 40% and 60% of your immediately adjacent environment is either Type TG678L1 or Type TG612X9 cells with the rest amniotic fluid, produce another Type TG678L1 cell at the edge of the amniotic-fluid zone”, which when implemented across a fetus result in a pair of breasts, but that’s not terribly helpful for a separate, “When the following pattern of retinal cells is activated, perceive sexual arousal” routine.

  9. Being heterosexual (or any sexual actually) sort of needs to be composed of more basic instincts to do with things like how male and female bodies are shaped, and how breasts and other sexual characteristics look and so on. The high level understanding that someone is a particular sex is composed of lower level cues, and it must be these cues that are attractive and not simply the knowledge of sex itself. The high level conception is meaningless without these things, because it’s simply the sum of those parts, and you can’t use it as a hyperprior to train in the other stuff because it has no meaning at all without its components. Scientifically we can say that sex is about chromosomes and that gender is about social labels, but humans don’t come with hyperpriors about these things, and then learn they should like breasts because they like women. “Woman” and “man” have no pre-existing meaning to the senses other than the average characteristics that cause us to construct these categories in the first place, so it’s very much the other way round.

    To the senses women and men are are about certain shapes, bodyparts, and behaviors. Boobs aren’t attractive because they are attached to women. Women are attractive because they have boobs (along with other attractive qualities; let’s not get too sexist here!)

    • albatross11 says:

      As an aside for Westerners: a lot of men find womens’ shaved legs quite attractive (say, in a short skirt or dress). But obviously, shaving your legs is cultural, not something that’s inherent. My WAG is that there’s some kind of program set up for finding things sexually attractive, and that this program gets linked up to finding particular signals attractive (shaved legs, long flowing hair, big breasts, the right hip/waist ratio, etc.) in most people. And when that linking process gets messed up or hijacked, you get various fetishes[1]. Maybe homosexuality just has to do with the initial steps in creating that linkage between signals and sexual attraction, or maybe there’s some more fundamental thing that determines how that linkage will be defined.

      I claim no deep knowledge about this, it’s just my intuition.

      [1] And how functional that is depends on social context. If the linkage is toward finding lots of black leather sexy, you can probably manage a pretty good sex life in modern US society, but probably not so well in Amish society. If the linkage is toward finding children sexy, you’re probably not going to do too well anywhere in US society.

      • Yosarian2 says:

        This could be a thing though where you’re exaggerating differences that already exist. If people are programmed to notice certain secondary traits that distinguish men from women, and to find those traits attractive, “women usually have somewhat less body hair then men” is a fairly noticeable one in most people. So if women shave, maybe they’re just exaggerating that natural secondary sex characteristic in order to look even “more female” and therefore more attractive to those instincts.

      • Joyously says:

        This feels right to me. My intuition is that we come with with an instruction something like “find this sex/gender attractive” and then we figure out what that means.

        Another example: I (and a lot of other heterosexual women) find men in suits attractive, but the concept of suit-and-tie is obviously not encoded in our DNA.

        • cmurdock says:

          Or “find [exceptional specimens of] this sex/gender attractive”– whether “exceptional” is in reference to physical fertility/virility or to one’s ranked position within the tribe. Business suits would be a signal for the latter; police and military uniforms for the former. If there were a style of dress that tracked with low prestige and fitness *and* was distinctively male, I would suppose that heterosexual women would not find that style of dress attractive (and, obviously, mutatis mutandis, heterosexual men).

          The problem is that I can’t think of any examples of distinctively male or distinctively female styles of dress which are specifically associated with low status or fitness (the lower ranked we are, the more men and women dress alike), so that hypothesis might be difficult to test. I could say that “there are probably no cultures where people are physically attracted to markers of low status”, but that would possibly be a circular argument depending on how “low status” is defined.

          • sconn says:

            Any ladies here turned on by overweight guys in dirty wife-beaters? Any dudes who like toothless women?

          • sconn says:

            Augh, should have put in more than two seconds’ thought into that comment. It seems obvious to me that if dirty wife-beaters or toothlessness *were* attractive, high-status people would adopt them, because they want to be attractive. Jeans are attractive, used to be low-status, and then people figured out their butts looked good in them so now there are $200 jeans.

    • acrimonymous says:

      “Woman” and “man” have no pre-existing meaning to the senses other than the average characteristics that cause us to construct these categories in the first place, so it’s very much the other way round.

      I think this is precisely what’s at question in Scott’s post. If the brain does have some things encoded, and if PP is true, then it is entirely possible that there is a higher-level “woman” and “man” categories against which individuals are tested.

      • mwengler says:

        My own introspection suggests it is not the category of “women” to which I am attracted (I am a heterosexual male). It is a subclass of (generally) women that are attractive, and they tend to be young (at least as young as me), smooth, female shaped (as opposed to lumpy potato shaped), with pleasant features, breasts are nice, hips are nice, eyelashes, hair on the head, etc etc all the things we associate with attractive. Further, as an heterosexual male I introspected to answer the question of certain of the newer genders, “Are they guys with boobs or chicks with dicks?” My own introspective answer is “chicks with dicks.” Despite my neocortex knowing they are “really” guys and they are lacking that one particular sexual marker (a vagina) that supposedly makes the whole game work in evolutionary terms, a man who gets boobs, smooths his skin, emphasizes hips over waist, uses mascara, grows a full head of hair, etc. is sexually attractive. I mention this to emphasize the point that what we are attracted to is not “things that if I screw will make copies of my genes come out.” They are rather a set of markers collected over millions of years (obviously starting long before we were humans and probably even before we were mammals) as marking, statistically speaking, good prospects for reproduction, WITHOUT US EVER HAVING TO HAVE A CONSCIOUS CLUE ABOUT REPRODUCTION. My own introspection that attraction is almost completely determined in non-neo-cortical mechanisms is that makeup, shaving, and plastic surgery all work on us. Sexual attraction is almost cartoonishly simple, certainly simple enough to work even long before we got smart.

        • Deiseach says:

          They are rather a set of markers collected over millions of years (obviously starting long before we were humans and probably even before we were mammals) as marking, statistically speaking, good prospects for reproduction, WITHOUT US EVER HAVING TO HAVE A CONSCIOUS CLUE ABOUT REPRODUCTION.

          Sure! When “persons with this set of characteristics” is the same list as “persons who will carry my baby/give me a baby”, then you don’t need anything too complicated other than “person looking like this makes me go sproing!” without having to get the brain-as-reasoning/thinking thing involved on any level at all.

          This then gives us a basis for the whole “gender roles are socially determined” and “performing femininity” arguments because yeah, we’re not making any conscious or reasoned decisions about “I’m attracted to persons who look like this, if you look like this I’m attracted”. Which is not to say that biology has nothing to do with it, it does, in a very fundamental way – but a way that can be tricked and deceived and moulded.

        • Conrad Honcho says:

          Are you trying to come up with an evo-psych argument that traps aren’t gay?

    • @acrimonymous

      But those higher level concepts will be composed of lower level ones like “women generally have breasts”, and then you have to have an idea of what an average breast looks like and so on. They’re not anything more than the sum of their parts in other words. Turtles Boobies all the way down etc.


      There are definitely cultural components to fetishes, but unlike high level things like leather and shoes that work on a Pavlovian method, some things are near universal and are not considered fetishes such as attraction to boobs and butts (in different ratios true), so there must be some kind of highly embedded criteria for what an attractive boobbutt looks like (or attractive male body parts too) in order for it to be so robust and resistant to cultural drift.

  10. Null Hypothesis says:

    It’s a little off-topic, but since the comments so far all seem to be about it in the first place: A youtube personality named Lindybeige did a quite good video on “Why do Women have Breasts?”

    And it’s actually a bit more of a paradox than it seems at first glance. The two primary arguments:

    They’re for feeding babies, you idiot!


    They’re sexy, you idiot!

    Are actually insufficient. No other mammals have permanent breasts. In normal mammals, females get pregnant, develop breasts, use breasts to feed said children, and then the breasts go away.

    Meanwhile, while breastfeeding females are generally not very fertile. So in reality breasts should be very unsexy because they are a very strong sign of low fertility.

    So why do women have permanent breasts, and why do men find them sexy? It’s a complicated question. Lloyd (Lindybeige) in the video does a very good job of stepping through some possibilities of how it developed.

    He also expounds on some of the issues Scott brought up, in terms of being limited in how much genes should be able to efficiently encode instincts. Up until some point, proto-human men generally should have found breasts to be a turn off because it conveys negative information about a potential woman’s fertility. When they became permanent, men should have found breasts to be nothing because they present little useful information (about fertility – healthiness and overall fitness could be a different story). Instead they’re obsessed with something that basically provides no information. Lindybeige posits a more parsimonious mutation that shifted from the old model to this sub-optimal new model.

    Nothing iron-clad (what is in this topic?) but very interesting and thought provoking. He also has quite a good sense of humor. I highly recommend watching it.

    As for how babies identify it and know what to look for? That’s probably up to the mother. Babies seek being on their chests, and naturally suckle. Mom should take care of the rest. It’s remarkable how many baby mammals (read: pretty much all of them) can be fed with a bottle.

    • Actually, a youtube personality named Lindybeige actually did a video on “Why do Women have Breasts?”

      I was thinking of this video when I read the title. It’s a fairly compelling explanation.

    • Bean says:

      Re: breasts
      Haven’t seen the video yet, but my first instinct when reading this was that women have permanent breasts to incentivize men to stick around.

      A lot of animals make it really obvious when they’re ovulating, which lets them copulate only when it won’t be a waste of energy. Once that’s done, the males scram and find some more females.

      Animals that need bipaternal care tend to be different. By hiding when they are most fertile, the females force the males to stay for the entire cycle if they want to reproduce. And if they aren’t going to be maximizing the number of offspring, they might as well maximize the survival of these.

      So if permanent breasts in humans are related to hidden ovulation, then I guess breasts are a con???

      Ahhh, I don’t know. I tried to look up if baboons, which supposedly also have hidden ovulation, have permanent breasts. It turns out that gelada baboons have sexual swellings on their chest that correspond to sexual receptivity and fertility – the opposite of the video’s point. And there’s a bunch of variation in primates in general, with some showing obvious signs and others not changing at all.

      Eh, I’ll watch the video and think about this harder when I’m more awake.

    • Deiseach says:

      No other mammals have permanent breasts

      Yup. Human females are also different in that we’re in permanent oestrus (more complicated than that, but the dumb version is that we’re up to mate anytime, anyplace, anywhere rather than being in season/heat at a particular time of the year). So we’re basically signalling via large permanent breasts that we’re lactating (and thus while not fertile now we have been successfully fertile) while we’re also available for mating anyway.

      Human sexuality is very confusing! 🙂

      • aimward says:

        I wonder if sexual maturity feeds in here, as well. In a species where estrus state is generally unknown and unknowable without sophisticated measurements, and where there are few structurally apparent cues between sexual maturity and immaturity, breast development and hip ratio are one of our only reliable ways for easily making the distinction between “young” and “way too young”.

        • sconn says:

          Good point! I was hung up on “how do we tell a male from a female” but of course it’s also important to be able to tell if a female is of reproductive age or not.

    • acrimonymous says:

      The question may be posed incorrectly or in a way that does not attract a good answer.

      When we say “men are attracted to breasts”, we are thinking about the kind of breasts men are attracted to. I, for one, do not find the sagging breasts of many hunter-gatherer women to be attractive at all. On the other hand, the young adult women of hunter-gatherer tribes often have non-sagging breasts, which I find just as attractive as western women’s “breasts,” by which I mean my image of non-sagging breasts, which I, and, I think, many others, seem to assume is the average for women.

      I think it is possible that modern men’s attraction to “breasts” is really an attraction to youthfulness and sexual inexperience.

      EDIT: Another possibility is that, as breasts are fatty deposits, their status signals whether or not the environment is good for mating.

      • sconn says:

        I hear the men of hunter-gatherer tribes think Western men’s obsession with breasts is ridiculous and infantile. Since the women of their tribes go topless, breasts are just a lot less big of a deal. Yet I’m not positive they don’t care at all.

        It makes me wonder if a culture where we always wore shoes would give men a universal foot fetish.

    • bbartlog says:

      You can address most of the objections by assuming that there is a strong imprinting component to human (visual) sexual attraction.
      Since human males, unlike those of other primates, don’t get turned on by some kind of strong signal of estrus, evolution must have hacked together a system of visual attraction in the space of just a few million years. Well, as already pointed out, it hardly seems likely that this would include low-level genetic programming to allow specifically for recognition of breasts.
      But it could include a system of gradual imprinting that was triggered by various environmental cues. Humans would learn what MAS looked like over time because they had various reference models walking around together with some kind of internal genetic bias to favor ones of the appropriate gender.
      In this context, of course men would develop an attraction to breasts, because they’d grow up in some tribal environment where a lot of women had them (and this would apply even if we assume some environment where breasts are solely due to current milk production, not yet evolved in the direction of being larger).
      This would also be one explanation for various paraphilias.

  11. P. George Stewart says:

    There’s actually a fair amount of complexity, and not a little oddness to the topic of breasts, as Lindybeige expounds in the following video. This is from the point of view of “why do women have (permanent) breasts?” But it touches on the topic of why men like breasts too.

    According to Lindy (this is just him tossing some ideas around and suggesting an alternative hypothesis) it’s untenable because in the deep ancestral environment (not quite human yet), breasts would be a sign of lactation, but the lactation period at that time during which females kept their breasts (around 2 years) would have been a period of relative infertility (natural contraception to space out having babies), so if there was any evol psych pressure, it would be to be turned off by breasts.

    He postulates an overlap between the development of “secret ovulation” (nothing like the chimp’s pink bum signalling ovulation) and the development of breasts. How secret ovulation developed isn’t clear, but secretly ovulating mutant females who had a mutation for keeping their breasts after child-rearing would have a cuckolding advantage because their providing mate would be unconcerned (“she’s got breasts=infertile”), and they could shag the band’s “rock star” behind the cave while secretly fertile.

    At first, it would be touch and go whether the mutant females’ cuckoldry would succeed because the breasts would still be a turn-off, but at some point, which Lindy calls Big Bang, there would be a race on for men to avoid cuckoldry, by being turned on by breasts.


    • Deiseach says:

      Gentlemen, I would just like to remind you mildly that sexual attraction is not a one-way street: women are involved the same way as men. It’s not all and only about what men find attractive.

      What are the benefits to women of having permanent breasts? Because women can find them inconvenient and wish they were flatter-chested or didn’t have them at all. Attracting a mate is one thing, but they have to have functions beyond merely that, and I do think you are overlooking the “nurturing an infant” factor. If successful mate selection is about ensuring the spread and survival of your genes, and finding strong genes to combine with them for that success, then having infants which result from that combination survive infancy is a strong part of all that.

      You can be banging the alpha and every other male in the troop and getting pregnant year on year, but if the babies never survive infancy, you are not being reproductively successful. And neither are the males which mate with you.

      • Null Hypothesis says:

        The video itself goes into a lot more detail than that. Breasts maintaining functionality is actually one of his primary points. The permanent breasts are sub-optimal, but still functional.

        Larger breasts don’t directly correlate with more milk (and if they did that’d be wasteful, babies only need so much) and they tend to interfere with the ability to easily feed babies. So you have some very strong counteracting forces that prevent breast sizes from becoming as absurd as a peacock’s tail.

      • albatross11 says:

        I think it’s common for women to accumulate fat on their breasts and hips, and both of those are also very common signals of attractiveness for men. So you could imagine bigger breasts/hips –> healthier, better fed, more likely to succeed at having kids. (I think obesity was basically never a health problem until modern times.)

        The other thing that’s important to remember here is that men and women have really different evolutionary strategies w.r.t. sex. Sex for a woman is a big investment–there’s a substantial chance she’ll end up having a kid, which is somewhat risky for her and which she can only realistically do a few times in her life because of the cost of caring for the kid. Sex for a man, absent some kind of marriage or pair-bond, costs nothing. (Except the cost of convincing the woman / not getting a spear in the back from an outraged husband.)

        So my guess is that breasts and hips and other such signals probably are more about signaling “I’m a good choice to pair-bond with” than “I’m a good choice for a quick one-night-stand.” Women don’t really have to advertise so hard for that second choice.

        • Deiseach says:

          Sex for a woman is a big investment–there’s a substantial chance she’ll end up having a kid, which is somewhat risky for her and which she can only realistically do a few times in her life because of the cost of caring for the kid.

          Depends on how many you define as “a few” – you can have as many kids as physically possible if not using some method to prevent conception, but how many then survive infancy is another matter., and of course maternal death rates indicate a real risk attached to pregnancy.

          Some quick searching finds that the average size of the Irish family over the past hundred years has been around five-six children, only dropping in the 1970s to 2-3. And another source indicates that five-plus children was a common family size:

          In Ireland in 1960, one-third of births were fifth births or higher. This was an extraordinarily large proportion by the standards of the developed world and pointed to an incidence of large families that was quite unique to Ireland at the time. The typical pattern elsewhere was that first births outnumbered fifth-plus births, but in Ireland there were almost one and a-half times as many fifth-plus births as first births. Even in the US, where the overall TFR in 1960 was close to that of Ireland, fifth-plus births were little over half as significant in relative terms as in Ireland.

          By the 1980s, fifth-plus births had fallen to 15 per cent of the total in Ireland, but this was still significantly ahead of the corresponding proportions in Italy and the US, where births of this order had dwindled to insignificance. It was not until the late 1990s that higher order births in Ireland dropped to something approximating normal levels for developed countries. Even then, however, Ireland was at the upper limit for fourth order births or higher (by the mid-1990s, Ireland, Poland and Cyprus were the only European countries where fourth-plus births exceeded 10 per cent of total births – UN, 1997).

          Given that family sizes of seven to ten surviving children were not uncommon, and that some women had given birth to twenty children over the course of their fertile years (granted, not all the children survived), then in terms of early human history “a few times” isn’t all that few!

          • sconn says:

            My social circles consist mainly of conservative Catholics who use no birth control, and it’s common knowledge that if you have sex and don’t do anything to prevent it, you will have a baby about every other year. More often, if you don’t breastfeed. 10 kids is not uncommon among my friends, and if you don’t have kids it’s because you resorted to abstinence for awhile. I have been married eight years and have four children. I am one of six, myself, despite my parents having used birth control for 15 years. I think people vastly underestimate how many kids you’ll have if you really do *nothing* to stop it.

            On the other hand, hunter-gatherer tribes often have 3-4 years between pregnancies, because of much more frequent breastfeeding combined with taboos against having sex when the last child is still nursing. (I also suspect dietary carbohydrates are a factor, but that’s just speculation.)

            But I’m okay with saying a woman can only have “a few,” because 10-20 kids is nothing compared with the number a man can have if he is polygamous. 100 is not even pushing it, really, if he’s got enough mates.

      • Eli says:

        Honestly, if I really had to guess at why human females have such prominent breasts and, on top of those, a nonstop estrus cycle, I’d have to guess this:

        At one point, our population bottlenecked in some natural disaster or another. There really wasn’t much more than a minimum breeding population of humans left. The resultant need to rebuild our numbers exerted a selection pressure for individuals who could spend more time and energy on reproduction, without worrying about silly old things like “appropriate mating seasons”.

        So our females mutated to have their bodies treat all the time as mating season (compared to other mammals, which really can’t get pregnant and don’t bother mating quite so often), which then also created regular periods. But which also may have saved the species. Since they were now spending a lot of time on having children (more than they’d like, but natural selection laughs at our likes), the same genes might have coded for large, active nursing features.

        • P. George Stewart says:

          This is why I’ve always been fond of that supervolcano/population bottleneck idea (although I gather it’s not favoured) – if it were true, it would explain a lot about is in one fell swoop.

          • Nancy Lebovitz says:

            Was there a fast population increase after the bottleneck, or a slow recovery? Either one is possible depending on the environment.

            It could have been an extended period of grimly hanging on if everything is wrecked or a Heinleinian pioneer extravaganza of open land with no competitors. Or some mixture.

        • baconbacon says:

          The flaw here, which may be fixed with some nuance to the hypothesis, is that these changes would have to spread to all women. In the scenario you describe, a bottleneck where the population is unconstrained by resources during the rebound, doesn’t lend itself to a wipe out of the previous paradigm. This is because with all the resources around you have trouble breeding those genes down to zero. At stable or shrinking population rates it is easy to (eventually) push inferior genes out of the pool altogether, but in high population growth scenarios they only are reduced in frequency. Then when the population returns to carrying capacity the old paradigm is once again favored, and that 10, 5, or even 1% of the population left with those genes starts winning out.

      • P. George Stewart says:

        “It’s not all and only about what men find attractive.”

        Coming back to this to comment: the video talks about what women find sexually attractive, i.e. the “rock star.” But it also talks about what women find “attractive” in the sense of the cuckold, i.e. the accommodating guy who, like a chump, will raise the kid they got by the rock star for them.

        Using the “appliances vs ornaments” trope – women are (say) two thirds into objectifying men as appliances, and one third into objectifying men as ornaments, men the other way round – two thirds into objectifying women as ornaments, one third into objectifying women as appliances (maybe it’s thirds or maybe it’s the Golden Ratio 🙂 ).

        Neither is completely free of some level of shrewd unconscious or conscious calculation, but neither is completely free of being prone to bouts of pants-on-head-retarded lust either, it’s just that the weighting is different for men and women.

  12. Jack V says:

    Huh. Good question. But lots of things seem inbuilt. The obvious vision things like edge detection and dividing the world into objects. The baby bird’s red dot algorithm is an interesting example: even “look for the red dot” is hard to design genes for to describe a system of neurons which do it, built out of subcomponents also built out of neurons. But that clearly works somehow. And there’s also a system for “recognising creatures, and telling this particular creature apart from all the others”.

    Peacocks liking giant feathery tails is somewhat more complicated than that, but it feels like the same sort of thing, SOMEHOW, even if we don’t know how.

    Presumably the theory of the mind is less-blank-slatey than it sounded, in some ways, even if it still is in others. That sounds quite promising, in that using existing evidence, we make some specific conclusions about the different parts of the theory, rather than “it can explain anything with enough metaphors”.

  13. jonm says:

    I think Scott is basically right here.

    Simpler rules should be easier to encode and more flexible.

    circuit 1: determine what patterns in sense data most differentiates the sexes
    circuit 2: determine what others around you consider to be attractive traits
    circuit 3: some very deep level rules about detection of disease (i.e. dripping pus etc) as well as the ability to learn what is associated with disease

    These would generally be a pretty effective set of rules that would explain both the conformity in what is attractive and the ways in which it can vary pretty radically across cultures.

    iterating between 2 and 3 will tend to bring society towards an equilibrium that favours healthy body types).

    Iterating between 1 and 2 will tend to lead to an exaggeration of gender/health indicating physical traits such as breasts (because on average they are interpreted in a culture that values them) but none of them need to be hard coded because you can just learn that from the surrounding culture.

    So basically evolutionary psychologists have it half right, those traits did evolve to signal but not because we have innate tendency to value those particular traits just because culture is part of the evolutionary environment. Additionally, as a trait becomes more selected for, it becomes a more easily learned signal of attractiveness and gender so in turn becomes more selected for (breasts are a pretty reliable sign of gender in modern humans).

    • Unirt says:

      There appear to be many innate behaviours that seem pretty complex and high level; e.g. young birds that have never seen an insect in their life show clear avoidance of brightly coloured (yellow-black-red) insects, but don’t fret about brown insects or red and yellow berries. Just specifically coloured insects, with legs and heads and all. Sounds pretty complex, but is nevertheless a well established innate behaviour. Any theory of how mind works needs to somehow take these complex instincts in account. On the other hand, female great tits preference for male great tit song is learned, they can be trained to prefer blue tit song (and blue tit males as mates).

      I don’t know if it’s been studied whether peahen have an innate or learned preference for big tails (would be easy to test), but in some domesticated populations it seems to have been lost (evolutionarily or culturally? Again, easy to test). I’d like to know which proportion of the known sexual and dietary preferences in animals are learned and how many they are born with…

      • J Mann says:

        On the other hand, female great tits preference for male great tit song is learned, they can be trained to prefer blue tit song (and blue tit males as mates).

        Wow – given the original post, that sentence definitely didn’t end where I thought it was going to. Very well done if intended! (

        And a good lesson on the importance of punctuation – a well placed possessive apostrophe probably would have avoided the confusion. 🙂

  14. BlindKungFuMaster says:

    Not an answer, but two remarks:

    The bayesian, multi-level, top-down and bottom-up processing stuff is more or less a description of the cortex. The innate pattern recognition (like snake detection, etc.) happens in older parts of the brain afaik.

    Framing the information content of our genome as “20,000 genes” may be somewhat misleading. These genes are used to create proteins via rna, but >98% of all produced rna is not used to create proteins. Instead they have a multitude of other functions in the cell. And that’s just one example how there may be many ways to functionally change the genome without touching any genes.

    • Jonathan says:

      The bayesian, multi-level, top-down and bottom-up processing stuff is more or less a description of the cortex. The innate pattern recognition (like snake detection, etc.) happens in older parts of the brain afaik.

      I was going to propose something along this line. The cortex might be fully plastic, but the lizard parts of the brain have a bunch of hard coded logic. As infants, when we’re busy molding the cortex, the lizard parts are triggering on the various things they’ve been programmed for and all of that gets folded into the final models.

      If you wanted to test such a theory (in the most unethical way possible), raise a child in a completely sterile environment and see how they respond to those stimuli when released.

  15. summerstay says:

    It’s not just sexual attraction (which seems to be distracting a lot of the commenters), there are a lot of other human instincts that seem innate. I’ve had this exact same question about spiders– how can people have an instinctive fear of them, when the concepts themselves are part of a learned framework that is probably different in each individual? (My concept of spider, for example, is faintly connected to my concept of water spouts.)
    In the 19th century, many scientists (Samuel Butler, for instance) thought that memory and genetics must be the same mechanism, because of this interaction. The fall of Lamarckian thinking and discovery of DNA pretty much ended this speculation, but the discovery of DNA methylation (and the Lamarckian passing of fear responses to children in mice) seems to have reopened the case a bit.
    In fruit flies and C. elegans, we have good evidence that certain mating behaviors are caused by hard-coded arrangements of neurons, even though there are only about 300 neurons to account for (see this answer, for example) but it is quite a jump from these small systems we understand to a human brain capable of representing any concept somehow linking those concepts up with such innate structures. My own guess would be that there is more information coded in “junk” DNA than just those 20,000 protein creating genes that is guiding the development of the embryo in such a way as to put these instincts in the right place in the brain to get attached to the relevant concepts when they come around, and that this process often goes a little wrong (where people get odd -phobias or -philias).

    • tmk says:

      Do we have an innate fear of spiders? Or is it learnt? A lot of things that seem innate may just be learnt early.

      • BlindKungFuMaster says:

        There is good evidence that primates have an innate snake detector and I would be surprised if the same isn’t true for spiders.

          • Nick says:

            I never knew about this and this is fantastic.

          • John Schilling says:

            Quite amusing, but, alternate hypothesis: Cats maintain a mental map of their immediate surroundings, and are alarmed by something mysteriously appearing right behind them while they are eating. Unlike cucumber-aversion, this offers obvious evolutionary advantages for a non-apex ambush predator.

          • Dedicating Ruckus says:

            Yeah, looking at that, there’s the “jump into the air and skedaddle” case that seems to happen where someone put the cucumber behind them while they were eating — easily explained by pure startlement, could presumably duplicate with any object — and the “watch with suspicion, approach slowly, bat with paws while recoiling away after” case, that happens other times — which is a pretty obvious “investigate this possibly-dangerous possibly-edible thing” program, and I’ve seen cats do it with every class of object as well.

            Of course, the video is still hilarious, but many things cats do are hilarious.

          • Charles F says:

            Anybody have a cat they’re willing to scare and a variety of cylindrical, green, and green cylindrical things in order to test @John Schilling’s hypothesis?

          • willachandler says:

            The simultaneous activation of multiple startle reflexes in cats — including fears of (1) unexpected objects, (2) being stalked, (3) snakelike shapes, and (4) reptilian colors — plausibly accounts for the remarkable intensity of feline cucumber-fear.

            Many SSC readers (including me) will be empathically discomfited at the idea of excessive experimental testing of this hypothesis upon one’s own cat. Such experiments may not be illegal, yet surely they are morally wrong. Because yes, cats *do* have feelings (don’t they?). 🙂

            After all, WWBD (what would the basilisk do)? Terrify cats in perpetuity? Definitely not cool! 🙁

            For psychotherapeutic applications of simultaneous startle stimuli, see (e.g.) Gallagher et al. 😉

          • Doctor Mist says:

            Anybody have a cat they’re willing to scare

            I tried it on ours, which I expected to work great, since when she eats she is insanely focused. But no: She finished eating, saw the cucumber, sniffed at it perfunctorily, and ambled off.

      • Ozy Frantz says:

        Phobias of spiders (and other evolutionarily relevant fears, like snakes, heights, social disapproval, enclosed spaces) are classically conditioned more quickly than phobias of evolutionarily irrelevant fears (like cars, guns, cigarettes, toxic substances under the sink).

      • Null Hypothesis says:

        Some theories posit that human’s eyesight developed to the quality it is (we have very good eyes, all things considered) so we could better spot snakes.

        Or, I suppose the more causal statement would be that all the proto-humans that had bad eyesight got killed by snakes.

        It’s funny how perceptions of the bible, and other religious texts, seem to start out so often as perfect truth, then turns to ridiculous, arbitrary philosophical anecdotes, and then wraps back around to being potentially deep, historic metaphors.

      • Worley says:

        For the things that people tend to think of as intrinsically frightening (dogs, snakes, spiders), what studies I’ve seen say that only a minority seemingly have an intrinsic, phobic fear of them. OTOH, as another commentator notes, it’s also easier to train people who don’t fear them to fear them, and irrational phobia levels of fear are far more common. E.g., a lot more people are deeply afraid of dogs than autos even though autos kill a lot more people. (Though Freud had a phobic fear of trains.)

  16. Nancy Lebovitz says:

    My impression is that the sexual importance of breasts varies from culture to culture. There could be a simpler(?) genetic system which tells people to be attracted to what their culture says is sexy.

    • Worley says:

      There’s every reason to suspect that humans have that, too: Someone who posses something that is culturally-valued has more human capital, and there’s value to sucking up to people with a lot of human capital. Similarly, many of these valued signs show access to resources (material or social), and a mate with access to resources is more valuable than one without. As some observer noted, “The one constant about the ideal female figure in any culture is that it’s difficult and expensive to attain.”

  17. Anon. says:

    You’re underselling the instincts of animals. I think the animal world provides us with enough examples of extremely complex sexual selection systems that are obviously not learned, so that we can dismiss the “how could genes possibly do this” objection out of hand.

    Just look at the bowerbird. Somehow there are genes that code for proteins that make them build elaborate mating nests, fill them with colored objects, and then perform a mating dance whenever a female comes to look. The females in turn have the ability to judge how good the nest and the colored objects and the dance are. Different species of bowerbird build different styles of nest. This is high-level stuff, I’d say even higher than breasts which are relatively simple to judge. And clearly not built up Bayesianely from nothing.

    Something interesting in this area would be to look at how developmental paths differ. Gazelles can walk a minute after being born. It takes a human a year! Perhaps you could argue that we have sacrificed instincts for greater learning ability? (Nietzsche would love this) Here’s a paper in this vein: Extraordinary intelligence and the care of infants. Longer helplessness = higher INT.

    • moridinamael says:

      Also: spider webs, beaver dams, ant and termite colonies, wasp and bee hives, bee dances, all bee behavior, the various finely honed hunting strategies of marine vs avian vs quadrupedal predators, you can really go on and on with examples of animals with tiny brains being born knowing how to do amazing things without being taught them.

    • Cerastes says:

      Consider the simple question I expounded upon in the thread about the book – how is a hatchling snake instantly and immediately able to both use *all* modes of locomotion known to it (across WAY WAAAAY more degrees of freedom than any human body) and process the sensory world well enough to navigate and even accurately strike prey and predators? And it even works if prematurely removed from their eggs (within limits, of course, usually only a few days).

  18. nestorr says:

    Do men in societies where women are routinely topless give much of a damn about breasts?

    In one of Robert Sapolsky’s lectures he talks about the mechanics behind instincts and mentions chicks pecking at the ground, the instinct sets up the behaviour by making them peck at their own toes, after some trial and error the chicks find that pecking at food hurts less and tastes better. So it seems instincts only need to point you in the right direction with some hack and then let the individual figure things out via trial and error and external cues.

    • Worley says:

      Well, someone should have a good survey of the known research. My personal guess, based on various statements of preferences that I’ve heard from various cultures is that men tend to categorize breasts in ways that make ev psych sense. Absolute absence of them is taken as unattractive. Large breasts indicate that she can maintain a fairly high body-fat content. Drooping breasts tend to indicate greater age (the net present reproductive value of a human in foraging societies seems to beak around age 15) and especially having nursed previous children (who are likely to remain as claims on her milk and labor).

  19. VirgilKurkjian says:

    I dunno Scott, I think you’re way off. I don’t know why you think genetics can carry enough information to build a breast, but not enough information to find one. These are several orders of magnitude differences in complexity. Exactly how it’s able to do carry this much information is a huge mystery, I agree, but given what it can already do I think we should accept that this is possible.

    Further, there are so many complex behaviors that appear to be innate. Humans, rodents, and prey birds all seem to have innate fear responses to hawk-shaped shadows. Spiders build complex webs with no chance to learn them. Bees have complex parasite-fighting behaviors which seem to depend on a small number of genes. Drosophila males have courtship behaviors that seem to depend on a single gene! Evolution seems to have no problem cramming in ever more weird crap into our code.

    • Cerastes says:

      How is building a breast complex? Hypothetically, you could just have estrogen-activated genes in the nipple which secrete signaling proteins to diffuse outwards to cause hypertrophy in the adipose and connective tissue, and signals growth in the associated glands and activates their milk producing genes. That’s like 6 genes, maybe 12. The diffusion alone will produce the “dome” shape, collagen will remodel to accommodate excessive strain, and the pattern of blood vessels and ducts is just physics plus remodeling to relieve mechanical stress.

      I strongly recommend D’Arcy Thompsons’ “On Growth and Form”. It’s over a century old, but it explains how exceedingly simple physics (e.g. surface tension, pipe flow) and/or simple mechanisms like growth gradients can produce remarkably complex and intricate biological structures.

      • VirgilKurkjian says:

        Very cool perspective, not a way of thinking I’m familiar with! I can see how, at this level of abstraction, building a breast would be really easy. But I think you’re already starting at a high level of complexity. For example, “have estrogen-activated genes in the nipple which secrete signaling proteins” isn’t a primitive action — this sort of thing is exactly where I would expect that complexity to be.

        Lots of hard problems are easy once you assume you have a function which does most of the work. So you’re right that this is easy once you have a complex “library” of functions. Building that library is what I expect to be complex.

        There may be a cognitive library of genetic functions which make it similarly easy to program infants to recognize breasts! We just don’t understand cognition well enough yet to conceive of the details.

        • Cerastes says:

          In nature, design is cheap – we’ve had about 600 million years and uncountably vast numbers of individuals to come up with robust methods and tools. And as far as complex biological structures go, “lump of fat & connective tissue with some glands and ducts” is pretty simple.

          I like to describe the genome as like building a house. Some code makes stuff – metal, wood, plastic, glass. Some code is “tools” for shaping that stuff – power saws, rulers, drills, nailguns. And some is instructions for how to use tools and assemble parts. The same tool is often used for many things, as well. And once a new tool is invented, it exponentially expands the range of construction options, and everything can interact with everything else.

  20. willachandler says:

    In the hilarious MythBusters episode Do Larger Breasts Equal Bigger Tips?

    “Kari Byron puts this theory to the test and the conclusion has a twist that may come as a surprise.”

    The twist is that larger breasts stimulate larger tips even more strongly from female customers than from males.

    Hmmmm … perhaps primates in general are socially wired to treat nursing females with special consideration? Plausibly, this behavior would be a conserved evolutionary trait.

    There is evidence too, that these predilections are not breast-specific. On the tiny outer islands of the equatorial Pacific, even today women’s breasts are bared … it’s simply too hot-and-humid to wear clothing above the waist. The thought-provoking result is that high-school-age males eroticize women’s knees and thighs (which are hidden by traditional island garb).

    These hormonally-charged equatorial males experience “Women’s Tennis Magazine” as supremely erotic … library copies are adorned with ball-point-pen comments like “WOW!” written on the naked shapely legs of players.

    Needless to say, present-day neuroscience is not obviously close to providing neuron-level explanations of these effects. We have only a medical maxim / “clinical pearl” to this effect: “Whatever physicians push down in one place, will pop up in another place.”
    To borrow a passage from Melville

    “So help me Heaven, and on my honour the story I have told ye, gentlefolk, is in substance and its great items, true.

    I know it to be true; it happened on this ball; I trod the bared-breast island; I knew the hormone-charged students; I have held the annotated tennis magazines, the clinical pearl was taught to medical residents in my presence.

    • Sam Reuben says:

      The solution, I imagine, would be to say that there are many given characteristics which cause a judgment of “feminine,” and then of those, the “hard-to-get” ones are especially eroticized. There are different systems for the gender-judgment than there are for the sexual appeal-judgment, and so you see some discrepancy. Nobody, I think, would posit that those Pacific islanders don’t see breasts as a feminine feature. They’re obviously a feminine feature. They would only say that they aren’t an especially sexy feature.

      A more interesting study would be whether or not those women experience the same levels of erotic stimulation from their breasts as women in breast-hiding cultures do, and whether they experience more from their knees and thighs. That, along with a study of foreplay in the region, would tell us a lot about human sexuality.

    • Worley says:

      My guess is that erotization of a part revolves around the fact that if a male can see a part of a female that is normally covered, then he has advanced in her good graces and investing resources in gaining her further good will is more likely than average to pan out. What would be interesting is to go to these islands and study how people judge the relative attractiveness of various breasts. The ev psych prediction is that people would give ratings similar to what people from other cultures do, though they wouldn’t consider answering the survey to be *erotic*.

    • Mary says:

      Here’s a related story: A man and a woman both belonged to the Society for Creative Anachronism. They also did stuff together as friends outside it, so he was, of course, perfectly used to seeing her in shorts. But one day they were doing a dance demo, and he came backstage while she had a foot up on a chair so she could adjust her garter, and he could see her calf — oooo baby, what a turn-on!

      A truly complicated result of the effects of clothes.

      • sconn says:

        Yeah, I remember one time I was wearing an ankle-length dress and hiked it up a little bit to tie my ankle boots. A male friend freaked out and I was like — what gives? You’ve seen me in shorts! But it’s not the same thing.

  21. Lasagna says:

    Dammit! The first time I’ve ever thought I had a “gotcha” comment – but what about animals, Scott? WHAT ABOUT ANIMALS?? – you went and addressed it in the next paragraph. I’ll get you next time.

    Seriously: I actually DO think that you’re kind of brushing aside the fact that most animals appear to be born knowing that a “determinate world is more pragmatic”. Anyone who has ever watched a puppy or kitten grow up from birth can attest to how quickly they’re able to successfully navigate the world – in particular, the predictive processing argument for how motor skills develop seems dubious for these guys, who are born with it in a way that humans aren’t.

    If human babies “see the world as a probability distribution over hypotheses and only gradually “figure out” that a determinate world is more pragmatic,” shouldn’t that be reflected somewhat in the animal kingdom as well? I’m not against the idea that there is something so utterly different between mankind and animals that the division is more biblical than evolutionary, but doesn’t that need to be addressed before we can really embrace the predictive processing model?

    • Dedicating Ruckus says:

      Isn’t it known that humans are born way earlier, developmentally, than animals are? (Something something high encephalization skull size something.) And there are definitely things that human infants basically can’t learn before a certain age, and when they reach that age seem to learn automatically.

      I’d be very sympathetic to the hypothesis that most of this is innate rather than learned, but the development process creating the structure for the innate instinct simply happens outside the womb in humans.

      • Lasagna says:

        That’s an interesting point – I didn’t know that, and it explains a lot – but doesn’t it still argue against the predictive processing model? If extra development – motor skills, processing of sights and sounds, cognition, whatever – is happening in the womb for other members of the animal kingdom, then it isn’t happening through predictive processing, right? PP necessarily requires exposure to the outside world.

        • Dedicating Ruckus says:

          Yes, this still argues against the hypothesis that most of these behaviors are learned via PP rather than innate. Really, the fact of many animals very obviously having complex instinctive behaviors from the moment of birth/hatching makes that point succinctly; in humans the picture is merely complicated by being born at an earlier developmental stage.

          Personally, I don’t follow Scott in saying that PP is some kind of grand unified theory of brain architecture; it’s an elegant model for sensory perception and even some higher-order reason, and to some degree for motor systems where it basically duplicates the insights of PCT, but there are obviously many areas that it doesn’t trivially cover.

    • Worley says:

      Well, as long as your only job is to *choose correctly*, you don’t need a deterministic model of the world. Indeed, many of the things that a human expert judges correctly aren’t based on deterministic assessments — in a lot of cases, what you need an expert for is when the evidence doesn’t make it certain what is happening, and you need a choice that optimizes the cost/benefit results of a probability distribution of factual possibilities. Real Bayesean judgement there. When you need a deterministic model is when you have to *say* “The horse is in the barn.” You gotta collapse the wave function before you can speak.

  22. Witness says:

    To a first approximation, all a gene does is code for a protein. How, exactly, do you design a protein that makes men find big-breasted women attractive?

    You probably don’t. But a collection of genes/proteins that ends up building a system more likely to find big breasts attractive seems more likely to successfully reproduce than collection which doesn’t.

    And number two, how do you get those neurons (and no others) to express a certain gene?

    The same way you get “a gene” that encodes for (say) eye color to only express in the iris. (Which is not to say that I understand that process, only that this is no special objection to PP)

  23. Sam Reuben says:

    Did you know that frogs have a small and very specialized part of their nervous system which handles flycatching? It tracks any small, black object that crosses their field of vision, and then causes the tongue to dart out and snap it up. The frog has no real control over this. You can feed the frog lead pellets by flicking them in front of it, and it will keep on eating them until it dies.

    Did you know that mice have fairly ironed-out social circuits? There are specific parts of the brain that have simplistic functionality that sums to “go hang out with other mice,” and when stimulated, the mice go hang out with other mice. There’s something similar in their brains for mating rituals.

    Did you know that humans have not one, but two dedicated areas in the brain for language? Actually, you probably did. But do you see the pattern going on here?

    Our existing studies of brains and brain architecture show that there are some areas of the brains of different animals which are pre-programmed to handle certain tasks. The fact that PP doesn’t account for this isn’t a failure of those studies, but a failure of PP. As mentioned in the main essay, it’s trivially obvious that animals have instincts, and quite elaborate ones at that. For most of them, it’s probably not too difficult to create with a limited set of genes, because all that’s needed is the construction of one coherent and fairly particular system, which is something we already know that bodies are extremely good at doing. (The power of genes in expressing traits with such little data is still terrifyingly impressive, and we don’t have a great grasp of how it’s done, but it’s not a surprise any longer that it happens.) How crazy is a frog-tongue circuit, when we have working eyeballs and visual interpretation systems? Not very.

    The question here, I think, is how it is that humans will have what I’ll call “soft preferences.” The frog example and the mouse example exhibit highly particular and totally irresistible brain functions, where there’s basically no question about how an input will connect with an output. Humans who are attracted to women find breasts attractive, but this doesn’t mean there’s an instinct to instantly grab them or something silly. There’s not even a simple condition of mind where, upon then seeing breasts, the human will grab them. You could take someone at peak state of arousal, put them in front of an attractive person, insist that they not touch (let’s say, with some reward for it), and despite very much wanting to, the aroused individual could keep their hands to themselves. (They could also give into temptation, but the point is, it’s not a done deal either way.) As such, the desire for breasts is a “soft preference,” where it provides a heavy weight on how decisions are made, but is not considered the only weight in relevant circumstances. (Humans do, for reference, have hard preferences as well. Consider the desire for breath: in a situation where breath is needed, humans simply lose control such that they can take a breath. It’s possible to train away things like the drowning instinct, but you basically can’t hold your breath until you die. Doing it until you pass out is already a remarkable “accomplishment.”)

    Soft preferences, then, indicate that what we’re encoding for in the case of breasts or whatever else is a general tendency to associate. Certain concepts or structures are inherently weighted to associate nicely with other concepts or structures, such as “breast” with “feminine” with “sexually desirable” (again, in the case of those attracted to women – is there a term for that? Gynosexual? It seems there ought to be). So, how is this possible? Gynosexuals aren’t born with the proposition “breasts are associated with femininity, which is associated with sexual desirability” in our minds, so that they can make careful judgments based on it. But what they do have is a highly developed visual processing system, a highly developed gender categorization system, and a working sex drive, all of which are capable of “talking” to one another. The visual processing system, which we already know has extremely high competence in identifying particular visual tropes (e.g. the extreme ability to pick out eyes or faces), is able to positively identify breasts through visual data, and passes on the information about the visualized object (because object distinction is another powerful feature of the visual processing system) to the gender categorization system. The gender categorization system takes that piece of evidence, and adds it to its judgment of masculinity/femininity/neutrality on the visualized object. If the femininity is high, or the sex drive is active for some other reason, this judgment is passed on to the sex drive, which then forms some certain desires based on that. And yet, at any stage in the process, when the main learning and thinking systems receive this data (as we know they do: we know the things we see, the judgments we make about gender, and how aroused we are), they are not bound to them. They can alter and adjust whichever piece of data they choose on the highest level, and determine outcomes that way. For example: “that looks like a breast, but it’s just a painting of one,” or “that person looks feminine, but I know they are/identify with being a man” or “no, we’re not going to make sexual advances in public, bad sex drive.” All of these fairly regimented systems are organized such that the Bayesian learning can go on in tandem: in fact, what is being Bayesed up isn’t so much the real world as it is our built-in systems of judgment. Clearly, there’s no contradiction in saying that there’s instincts and also PP-style learning, unless you’re an Analytic philosopher who insists that we perceive the outside world directly, which is a ridiculous proposition. The Analytics are strong on other fronts, but this is yet another issue that they should have just listened to Kant about.

    So in summary: there’s no contradiction in the first place, except for the one generated by an inane predicate of a certain school of philosophy. So long as what you’re learning from and about are your own sensory interpretation systems, there’s no problem with some of those having more structure and determination than others and with them talking to one another independent of the central learning processes. (It’s also possible for this same framework to take place within a judgment system, which is why it takes a while for babies to “learn to see:” the limited Bayesian judgment systems that operate within the visual system take a little while to get properly coordinated with the built-in equipment.) All of this information is perfectly compatible, and the resulting model is viable and fits nicely with reality.

    • Nancy Lebovitz says:

      “Did you know that frogs have a small and very specialized part of their nervous system which handles flycatching? It tracks any small, black object that crosses their field of vision, and then causes the tongue to dart out and snap it up.”

      Would that select for light-colored insects? Do frogs position themselves so that insects are silhouetted against sunlight? Do insects get selected to be dark– I think most flying insects are dark*– because frog risk is better than not getting the extra heat?

      *I’ve spent my life in a temperate climate, if that matters.

      • Sam Reuben says:

        I really hope this link works; Google links always make me a little nervous. These are the criteria for the reflex. I can’t answer all your questions, having not been there when frogs were evolving this reflex, but this touches on most of the points.

        • toastengineer says:

          I really hope this link works; Google links always make me a little nervous.

          Remember when the IETF was laying out the URL standard and made it very, very clear that URLs are meant to uniquely identify a digital object, not to reflect the configuration of your web server, and should never, never, everchange? Those were good times.

      • Cerastes says:

        Frogs can also perform “background subtraction”, regardless of the color of the bug. Photoreceptors eventually accomodate to light levels, and if your eyes don’t keep moving (they always are), your vision will just fade out. Frogs don’t make these rapid eye movements, and thus can just sit there and let the world “fade out”, but any moving object will suddenly appear as the only visible thing in the world. Makes targeting really easy and more dependent upon movement than color.

    • Toby Bartels says:

      You can feed the frog lead pellets by flicking them in front of it, and it will keep on eating them until it dies.

      Well, that explains The Celebrated Jumping Frog of Calaveras County. Apparently Twain got the story wrong, since he writes ‘he […] prized [its] mouth open and took a tea-spoon and filled [it] full of quail shot’, but that’s straight-up cheating; tricking the frog into eating the shot itself is much more sporting, and clearly what the stranger must have done. (Do I need a spoiler alert for a 19th-century story? the most famous short story by the most famous 19th-century American author?)

    • madrocketsci says:

      Humans being able to have soft-preferences could also be an example of purpose specific circuitry in the brain (though certainly nothing as trivial as the frog-reflex): Isn’t one of the main functions of the prefrontal cortex to interrupt lower-level instinctive circuits, and provide the ability for self-control in the first place? I may be drastically oversimplifying, but it seems the PFC “stands in between” the brain-layers that it has buried and the “output layer”. When it gets injured, self control takes a hit, and people begin acting on their instincts in a more direct stimulus/response manner.

      You would expect some amount of arbitrary variation to the layout of a pure blank-slate learned neural network. And yet, we know that the brain has a lot of very specialized subregions/organs which have the same function (or at least produce the same pathologies when injured) in pretty much every human, and even in other mammals. (Hence our ability to do brain research relevant to humans in the first place on mice.) The high level architecture/layout at least has to be governed by some sort of specification independent of subsequent learning/stimulus.

      (Which isn’t to say that this isn’t very interesting. I have no intuition currently for how unsupervised learning could happen, and yet we know the brain does it on many different degrees of scale. This might be a very productive model. And it is important, because any more general model for unsupervised learning promises to be far more powerful and autonomous than supervised learning.)

    • madrocketsci says:

      Someone I know once said that while training during a military exercise in a swamp, being tortured by the local insects, he noticed that the swamp-hornets would always scan him in the exact same way. Any given swamp hornet would do the exact same orbit about his head before deciding that his hair wasn’t pollen bearing and flying off in the exact same way after the exact same amount of scan-time, provided that he stayed motionless to provide them equivalent input.

      He said it was a little freaky how mechanical the process was: Here you are looking at something that seems almost too hard-coded to work in a real-world environment and result in something as successful and versatile as an insect. (And yet, you would expect that a lifeform that has a simple enough brain would *have* to become elemental and mechanical in its behavior) (He is a programmer and computer science specialist, among many other things. He has participated in robotics challenges, and understands how difficult it is to get a mechanical program to deal with a messy real world when you have inadequate senses.)

  24. Randy M says:

    Has anyone tried to write computer code or even simple human language instruction to determine the minimum number of steps to create an accurate model of the brain? Nevermind not knowing where in the genome “breasts are hot” is, 20,000 seems like a low number just for making sure the axons in the hippocampus leading to the frontal cortex have the right kind and number of neurotransmitters, and so on.

    • Charles F says:

      I like this question. I sort of touched on it in a different reply and I hope somebody who’s more expert in this stuff will chime in and correct all of my half-baked intuitions.

      I think the basic idea is that DNA is one of the best self-replicating configurations in chemistry that’s simple enough to find by random chance given a few billion years. And a lot of our biology is the configuration that requires the least DNA to accurately encode. So DNA is a quirk in a specific area of the properties of chemistry, and most functional biology is a quirk in how DNA self-replicates. Most of the information in biology is encoded not directly in genes, but in the chemical properties that make our specific type of biology the easiest thing to describe in that system. Unless we invoke some sort of creator, it feels like an anthropic principle argument, which bothers me, but it seems mostly unavoidable to me.

      Follow-up: has there ever been any serious work done on what alien, non-DNA-based biologies could possibly exist? Do we know of simpler-than-DNA self-replicating chemicals that might theoretically be programmed into life? Or more complicated ones the could be programmed into other life?

      • Randy M says:

        Yes, I did notice that your response above kind of makes my question silly, or at least a gross simplification; the fact that the instructions themselves are three dimensional, reactive objects means that genes aren’t directly analogous to lines of code.

        However, at the same time it does highlight that our intuitions about what is possible to encode there may not be entirely reliable, and the fact that various animal instincts including visual pattern recognition really do seem innate means it isn’t outside the realm of possibility for humans to associate certain shapes (to be refined through experience, of course) with the arousal systems.

      • madrocketsci says:

        Follow-up: has there ever been any serious work done on what alien, non-DNA-based biologies could possibly exist? Do we know of simpler-than-DNA self-replicating chemicals that might theoretically be programmed into life? Or more complicated ones the could be programmed into other life?

        I seem to recall something (vaguely) from a biology lecture ages ago about certain proto-bacteria having some non-DNA information carrier – some sort of scroll-like roll-up protein that does the same sort of thing – assemble other protiens when acted on by a decoder protein-complex serving the same role as a ribosome.

        • void_genesis says:

          There are probably lots of mechanisms for heredity that don’t rely on DNA base coding. Epigenetics recently became an established field but they are still finding new mechanisms on a fairly regular basis. Immune systems transfer between generations in mammals at least and are intimately linked with neurology. Microbiomes also transmit across generations and they are also linked to brain function.

          DNA gets all the attention because it only recently became possible to sequence it thanks to a conversion of serendipitous lab techniques and the emergence of high powered computers. And yet we still don’t really understand how proteins, cells and organisms work yet, not even close. We can’t adequately describe how the shape of the hand is formed, let alone something much more complex like a brain. I am resigned to the likelihood that we are missing something fundamental, just like the alchemists lacked meaningful atomic theory and Darwin knew nothing of DNA. If only we would spend more energy doing more speculative and exploratory science rather than sequencing another genome that we don’t know what to do with.

          • madrocketsci says:

            We have a glut of scientists (or, at least, more than there are university and research positions), and some of those scientists are supercomputer jockeys with a big-data hammer looking for a pattern-recognition nail, so we (as a civilization) can do both. 😛

    • TheRadicalModerate says:

      You need to break this into three pieces:

      1) What do you need to encode to create a brainstem and limbic system? (Answer is likely to be a lot of highly idiosyncratic things evolved over billions of years.)

      2) What do you need to encode to create a neocortical column? (Answer is likely to be a fairly small set of specific things that evolved a couple of hundred million years ago.)

      3) What do you need to encode the projections between the two, and the inter-cortical projections? (Answer is likely to be a fairly small set of things that evolved over a few tens of millions of years.)

      Human behavior lies in how the various regions learn given the input from projections from other regions, and from projections from the brainstem and limbic system.

    • Cerastes says:

      Development is the key. Sadly, I know little about nervous development, but let’s start with something still complex – the guts.

      All vertebrates are, at an early stage, just a tube within a tube. The outer tube is the body, the inner tube is the “guts”. There’s a fair number of genes responsible for making this tube-in-a-tube, but only a hundred or so, including two that set up signals for an anterior-posterior gradient.

      Now all you need is one gene for blood vessel growth that activates at 4 points along the body (roughly in quarters plus one at the end) to vascularize the inner tube from the outer, and four signals generated by each vessel. At vessel one, nearest the head, the tube portion right at the vessel baloons out and you’ve got a stomach and esophagus – turn it on longer for a bigger stomach. The next two say “tube, get longer” and boom, intestines. The spot where vessel 1 secretitions balance out vessel 2 activates another series of genes that make the liver and pancreas. Now you have a digestive system, with only a handful of genes responsible for organizing it, plus a few hundred for actually doing the metabolic work of various parts.

      The meta-lesson is that regulation of genes is hugely important. Someone recently was able to turn a herbivorous tadpole’s digestive system into that or a carnivore and vice versa, just by up or down regulating one gene, because that gene turned on or off (or up or down) the expression of dozens of other ones, and interacted in complex ways with still more signals. Any given gene may be regulated by dozens of signaling genes of various intensities and relationships.

      • Randy M says:

        Now you have a digestive system, with only a handful of genes responsible for organizing it, plus a few hundred for actually doing the metabolic work of various parts.

        The actual structure of the digestive system is pretty simple, a bit more complex than “a series of squishy tubes of various shapes and some sack like organs that excrete chemicals in response to other chemicals.”

        Brain structure seems very closely determined and specified, although to be honest I don’t know to what extent we could have equivalent functionality if different regions were arranged and shaped radically differently.

  25. Deiseach says:

    As a person in the possession of breasts since the age of eleven, I am sitting back with popcorn a packet of Jacob’s Mikado biscuits to listen to you guys (and I do mean guys) discuss this one.

    Can we get a follow-up post on “How do we get trim and appealing male backsides out of Bayes Theorem?” in the interests of thoroughness, completion, equality, diversity and non-gender/orientation presumption? 🙂

    • albatross11 says:

      It seems like the men should be trying harder (in evolutionary terms) to attract women than the other way around, given the difference in investments for sex.

      • Randy M says:

        Rather, men should be trying harder to attract any woman, and women should be trying harder to attract a particularly good man. Which is pretty much what we see.

        • Andrew Klaassen says:

          Rather, men should be trying harder to attract any woman, and women should be trying harder to attract a particularly good man. Which is pretty much what we see.

          That’s what we say we see. It’s funny, though, that the men I’ve known who were most enamoured of that theory were also the ones who obsessively rated women on a 10-point scale and expressed horror at the fact that they might ever sleep with someone below a 5. They would say, “Men will fuck anybody!”, but then focus their energy on making sure they were never reduced to the shame of actually doing that.

          And the people who actually “fuck [almost] anybody” are almost always women who are working as prostitutes.

          OkCupid found a similarly counter-intuitive pattern in their data: Women are only slightly more likely to contact the most desirable men. Men, on the other hand, are much more likely to contact the most desirable women. IOW, exactly the opposite of the simplified form of the standard theory that you’re expounding. Women try to attract any man (to an approximation). Men compete over a small number of particularly attractive women. It doesn’t fit the theory, does it?

          A slightly less simple form of the theory says that, in many species, the sex which competes amongst itself most for resources – usually the female – is plain, while the sex which competes most for mates – usually the male – focuses on being sexually attractive. In humans, there’s variation between societies in both which sex competes to control the resources and which sex competes to be the most sexually attractive, and it’s not always clear how the two are interacting. When societies have castes or classes piled atop one another, it gets even more complicated. Louis XIV is strutting around in heels at the same time that rich bourgeois men are dressing in the plainest black they can find.

    • willachandler says:

      … commences to practice female-attracting moon-walking and tenor-singing behaviors.

      `Cuz this stuff definitely ain’t easy to learn! 🙂

      The neurons and genes that regulate — also the Bayesian logic (if any) that justifies — the conceiving, performing, and appreciating of beyond-natural male-performed female-attracting human behaviors like these won’t be easy to characterize, eh?

      • Deiseach says:

        Tenor? Pffttt! Try counter-tenor – oh yeah, baby 🙂

        I don’t know if you’ve ever seen the 1994 film, Farinelli, which was probably more on the side of sensationalism than pure historic accuracy, but it was amazing to look at and listen to – though they cheated a bit with the voice, it was a mix of a soprano (Ewa Małas-Godlewska) electronically mixed with a counter-tenor (Derek Lee Ragin).

        While I’m throwing in links to Baroque operatic Youtube videos, here’s Jessye Norman killing it as Dido.

    • acrimonymous says:

      “How do we get trim and appealing male backsides out of Bayes Theorem?”

      I’ve been trying to get them out of the gym, not Bayes Theorem. Am I doing it wrong?

      • Deiseach says:



        BAYES. METHOD. NOW!!!!!

    • Schibes says:

      Can we get a follow-up post on “How do we get trim and appealing male backsides out of Bayes Theorem?” in the interests of thoroughness, completion, equality, diversity and non-gender/orientation presumption?

      I don’t know much about male backsides (aside from the one I’m sitting on now, which seems comfortable enough) but I’m pretty sure a Bayesian study of the male phallus would also satisfy all of your requirements, with the added benefits of a more easily quantifiable dataset (two simple numeric values, AFAIC) and more enthusiastic participation across a broader group of study subjects.

      • Deiseach says:

        Well, I stayed away from genitals as that’s another kettle of fish, but certainly there are things like the Chris Evans, Human Dorito meme which do indicate that a certain shoulder-to-waist ratio is very attractive on a man.

        Very attractive, indeed, especially when combined with a trim backside 😉

    • One thing I’ve noticed is that men are almost universally focused on body parts, whereas women seem all over the place. Some women are interested in chests and butts and genitals as erotic objects and many others (I would say most) are fairly indifferent, outside of some mediating context.

    • Sam Reuben says:

      It’s probably pretty similar – in both cases, it’s a pair of pleasantly round things. One has a marker in the center, and one does not. The standard theory, I believe, is that breasts are just piggybacking off of the backside, and it’s quite clear why the most serious walkers in the history of walking are sexually attracted to musculature which indicates high competence in walking.

      Just eliminate any mention to the nipple and areola and switch out gender references and most of the responses would work pretty well for you. I dig your complaint, though; the dialogue is pretty slanted towards the male perspective when it doesn’t need to be.

      • sconn says:

        But glutes are useful to the owner of them (you can use them to walk!) whereas fat deposits in the breasts (as opposed to the actual mammaries, which aren’t that big) aren’t useful to the woman who’s got them. So it’s a little harder to explain in that case. I can’t think of a single male body part that isn’t useful to the male, whereas I can’t think of any purpose in fatty breasts aside from attracting men. Fat storage is useful, I guess, but hips seem a more convenient place to put it.

        • baconbacon says:

          I can’t think of a single male body part that isn’t useful to the male

          Ball hair.

          • sconn says:

            Huh, I guess so. Though I’m not sure it would serve an attracting-women purpose either. As a female, I find it gross. Then again maybe it’s about increasing smells or something, like somebody suggested below.

            Beards, though. I should have thought of beards. Some women find them attractive. And they are a surefire way to tell someone is male.

  26. Itai Bar-Natan says:

    It’s worth remembering that the number “20,000 genes” is the count for how many protein-coding genes we have. However, brain structure and instinctive behavior is an area where I would expect that a lot of complexity is coded by regulatory regions.

    • TheRadicalModerate says:

      Yup. It’s all in laying down the specific combination of protein markers that will cause a specific set of axonal projections to sniff out where it’s going. I suspect that that constrains what gets learned–and what’s able to be learned–a lot more than your intuition would say it would.

  27. jeqofire says:

    I had no idea people even thought about breasts at all until some point in Jr High / High School (I forget when exactly) when it became impossible not to notice that my classmates would not shut up about them. And hip-towaist-ratio didn’t register for much longer (true, people would not lie about liking big butts, but I took it to mean “some people like fat”, and since other people were blatantly insulting toward fat people, it just struck me as individual preference.)

    I never had depth perception, and was rather sheltered. I knew that it wasn’t uncommon for women to have notable breasts, but it’s not an aspect I thought about unless it was brought to my attention more directly. The Mighty Morphing Power Rangers action figures might have helped, not just because I had no idea the yellow ranger was a boy (named Boy) in Zyuuranger, but also because all of the Pink Ranger toys I could find had the exact same body as the others. And the Planeteers were pretty flat. Toys are relevant because depth perception is not a requirement for getting the idea.

    The insistence that male=mammaphilic idea still irritates me. But I suppose almost every male stereotype I come across bothers me; it’s just that this one is like spider-webs in a disheveled old house. Then, as you have had enough of the face full of web and try to leave, you discover that, while you were inside, a spider decided the front door would be an excellent place for a new web. … umm, if it matters, I hid my gender when I first got to the internet, and was constantly mistaken for female for, like, 3-4 years, at least. Maybe I should try and ask someone who’d remember what that was about…

  28. slofgren says:

    Quick thought re: phenotype space possible with 20k genes. I think you are using a bit of a simplistic model of what neurons and neuronal connections are possible with those 20k genes. First, even if we assume that genes can only be on or off, which obviously isn’t true, and phenotype is only a function of whether those genes are on or off, then you have a phenotype space of 2^20k. Then allow for continuous distribution of coding gene expression, which can change in variable ways to extracellular signaling. Also consider noncoding RNA, which makes up, if I recall correctly, a much larger amount of the human genome than coding RNA, and which is transcribed but not translated but that plays a role in phenotype, like miRNA (and, as an interesting aside, someone told me once, not sure if it’s true, that non-coding RNA expressed in the brain showed some of the greatest rates of change since chimps). What you have is a large possible cellular phenotype space for a given fixed human genome. Not saying that this means that breast neurons that have a special encoded increase in synapse-forming potential with some sexual pleasure neurons in the hypothalamus is true. But as I see it, it certainly is possible with the human genome.

  29. I think that all the stuff about being “embodied” that you seem to dismiss as meaningless has the answer here…

  30. Jiro says:

    Someone said something like this, but I think you can compare this to any non-behavioral trait. Growing an eye or exactly five fingers is just as complicated a thing to be controlled by genes as preference for breasts, and pretty much all of Scott’s objections apply to it. If 20000 genes is not enough to have some of them that make breasts attractive, how in the world is 20000 genes enough to grow fingers, in quantities of exactly 5 on each hand?

    • Charles F says:

      I think you missed the point. It wasn’t that 20k isn’t enough for any of these things, it was that the supply of genes is limited, and if we had sets encoding all of the things evopsych says we’re hardwired for, we’d run out given any reasonable assumption for how many genes it takes to put something in the brain. Our assumption should be that most things are learned, with a very small core of innate knowledge that’s very important to survival and reproduction.

      I do think talking about big breasts was probably a bad choice, since later in the piece he comes right out and says:

      But [sex/gender] is one of the rare cases where evolution might really want to devote some big fraction of the 20,000 genes it has to work with to building a Rube Goldberg circuit.

      And that Rube Goldberg circuit probably assigns a weight to large breasts, which could probably be tweaked without using a whole lot of extra genes.

      But in general, I think the point – doubting a concept that could conceivably be learned is actually built into the brain by genes – stands.

      ETA: Also, nobody I know of still supports anything Lemarckian, so I don’t see how we can fail to accept that physical attributes are built-in. But everybody agrees we can learn things. So there’s a difference in that evopsych stuff could be genes or learned and we can argue sometimes that it wouldn’t fit in genes. With anatomy we have no choice but to accept that evolution found a way to make room for it in the space it had available. So the objection can’t be made in both situations.

      • acrimonymous says:

        I wonder if there is a problem with people conflating or arguing about different kinds of “learning.” At first I thought Scott was back to the nature vs. nurture argument, but after thinking about it some more, maybe I’m wrong. Maybe there are things learned by the brain that could not not be learned–they aren’t encoded but they aren’t “nurture” either.

      • Dedicating Ruckus says:

        > Our assumption should be that most things are learned, with a very small core of innate knowledge that’s very important to survival and reproduction.

        But why should this be our assumption? You can talk about the limited information space in the genome all you like, but the same limited space manages to correctly code for the entire human body in its dizzying unimaginable complexity. I would make a guess that by human means, actually encoding complete instructions to build a human would run in the terabytes or higher, not the ~1GiB that we apparently actually have. Therefore, making the standard assumption that all the relevant data is in fact in the genome alone, there already must be some mechanism in place to encode this information with astounding efficiency; there’s no reason to assume that this mechanism can’t encode complex innate instincts as well.

        • Charles F says:

          Hard to say since neither of us actually quantified how extensive a “very small core” is.

          And obviously, the mechanism can encode complex innate instincts, attraction to the opposite sex is right there.

          But FWIW, based on catherio’s comment I’d significantly increase the amount I expect DNA does to encode associations with concepts. (small numbers, and the differences between concepts aren’t *that* distinct/precise, but still really interesting)

      • Cerastes says:

        Animals violate this all the time. A baby sidewinder rattlesnake is born with a sensory system fully online, ditto for motor – they can see, move, and strike from the instant of birth, as well as a thousand other behaviors like tongue-flicking, shedding their skins, cutting their way out of the egg, finding prey and water, burrowing to set up ambush sites, etc. Learning may be possible (though vipers have a reputation for being exceptionally stupid snakes), but contrary to your statements, basically everything needed, down to the smallest detail, is hard-coded in.

    • bbartlog says:

      Another objection to hard-coded breast attraction, besides the limited encoding space, is the limited time we would have had to evolve it. I think evolution probably *could* hack together a hard-coded attraction to breasts, given enough selection pressure and enough time, even with a limited number of genes; but we are talking about a situation where we only had a few hundred thousand generations of primate evolution, with relatively small numbers of offspring as well as a whole lot of other evolutionary work in progress.

  31. Eli says:

    Remember that Predictive Processing is an incomplete model insofar as it fails to account for valence. We also have some fairly direct evidence from rats and people that something vaguely like time-difference reinforcement learning operates on sensory rewards.

    Ok, so what does that matter? Remember, in TD-RL, the brain doesn’t just learn which states have which rewards, and it mostly doesn’t learn the exact causes of rewards. It also learns to associate some secondary value with strong predictors of eventual sensory reward.

    So, for instance, with regards to sex, the brain doesn’t just learn “wet things rubbing on penises is rewarding”. It learns “strong predictors of upcoming sex-acts are also somewhat rewarding.”

    The big cross-cultural question then becomes: in which environments are which bodily features strong predictors that actual sex will soon take place? So if you live in some African country where women display their breasts all the time, we’d predict right now (retrodict?) that you don’t find breasts very attractive. In fact, maybe if what women cover up there is thighs, you’re really attracted to any hint of thigh.

    There’s definitely a joke in here somewhere about full-body veils and sexy ankles. Or actually, the whole original point of veils was that once they kinda started happening, hair became considered extremely sexy.

    The right learning algorithm, coupled to the right embodied cues, can replace a whole hell of a lot of instinct. And sure, you may want to yell NO FUCK NO THAT’S NOT EMBODIED, but I think it’s pretty uncontroversial to suppose that basic sensory rewards are embodied. Feel hungry, hungry bad, eat food, food tastes good, food makes full, feel better. See girl/boy, be 14, have heard vague rumors about penises and vaginas something something, hormones forcing a lot of heavy breathing and feelings down there, try it out, feels good, figure out how to do this more often.

    We’re still left with the question of how the body knows to release those hormones, but honestly, that could be some very, very simple visual filtering (as for faces). The embodiment of reward will more-or-less take it from there.

    The slate isn’t totally blank, but it only needs some vague smears of chalk on it to get the picture started.

    • Dedicating Ruckus says:

      The problem with this theory is that, if my middle/high school experience is any guide, in a great number of boys the breast obsession begins well before any actual sexual experience.

      Saving the cultural-conditioning hypothesis isn’t hard (and given the example of the Pacific islanders, it seems we must); you just have to assume a lot of internalizing social cues and cultural presentations. But this then raises the question of how the arousal circuitry gets tied in with some big cultural gestalt of The Most Female-Coded Body Parts; that seems to require a different class of mechanism than just another iteration of Bayesian learning.

      • Randy M says:

        Saving the cultural-conditioning hypothesis isn’t hard (and given the example of the Pacific islanders, it seems we must)

        It could go either way. Breasts are an innate arousal cue, but with enough negative reinforcement (breasts with no sexy time) (assuming breast feeding doesn’t count because the sex drive hasn’t kicked in then) this cue is disregarded, or in cultures with breasts uncovered this isn’t ever linked to the arousal systems.

        • Dedicating Ruckus says:

          Alternative hypothesis: admitting the idea of congenital neurological variation between human groups, gene/culture coevolution, and so on, is it possible that the Pacific Islanders don’t have an innate breast/arousal connection that Westerners do? (This seems on the face of it to be pretty unlikely but not impossible.)

          Any studies on Pacific Islanders adopted in infancy and raised in the West? Admittedly, that’d be a weird damn study to do…

          • Randy M says:

            I don’t think adoption is required here, as raised in a western culture is going to have enough social cues/pressure to learn something.

    • If traits were coded on specific and complex combinations of 20,000 genes, then that’s different, but as I understand it (in my limited way) genes are known to be more “linear” than that.

      If gene A coded for trait A and so on, that’s one thing, but if genes ordered A, B, C… coded for some different set of traits than genes ordered B, A, C… then that would be another matter entirely, and allow us to have a colossal amount of highly specified inbuilt traits. It would also make geneticists shoot themselves, I imagine.

      • Randy M says:

        If gene A coded for trait A and so on, that’s one thing, but if genes ordered A, B, C… coded for some different set of traits than genes ordered B, A, C… then that would be another matter entirely, and allow us to have a colossal amount of highly specified inbuilt traits. It would also make geneticists shoot themselves, I imagine.

        Oh, you reminded me of one of the coolest things I learned in genetics! Now, maybe someone will correct me if I’m wrong, but from what I understood and remember, the DNA strands for antibodies can be spliced differently before being translated. So gene “ABCDEFG” might be cut into “ABEFG” sometimes, and other times “ACDEG” and so on. (Each letter here probably representing a codon, and not a single base, but idk). Basically, few long complicated genes encoding for a vast number of antibodies, which allows receptors for the chemicals of a multitude of pathogens to be preprogrammed efficiently.

        Except it wouldn’t be just one gene, but also the gene that codes for the splicer protein/proteins that determine how it is put together. And these proteins can be up or down regulated based on what antibodies are needed at any time and to make specific, rather than general antibodies (or maybe antibody distribution ratios?). It’s all fabulously complicated.

  32. Randy M says:

    Recall the photos in the recent post about Surfing Uncertainty, how very vague images of the spots are able to be resolved readily into dalmations and cows.
    Breasts, and even female shapes in general, aren’t that complicated to abstract. Paired, symmetrical, plump. Put a couple dots in an infinity symbol and you could probably arouse guys. So it isn’t so absurd, once you grant that any preferences more complicated than strictly chemicals are possible to hard code, breasts could be one of them.

    • sconn says:

      I remember a bit in a Michael Pollan book where he made two lumps of dough (because he was baking sourdough boules) and remarked that they were kind of sexy, like breasts. My (female) reaction was “do you ever think of anything else?”

      Dudes may feel free to comment on whether they see boobs everywhere, the way that all humans see faces everywhere.

      • baconbacon says:

        I worked in a bakery kneading bread dough for some time. I would say that the thought “heh, these are like boobs” never actually went away, though I would never have described them as sexy, kind of sexy, mildly sexy or on the verge of sexy. A lot of it probably also had to do with the total repetitiveness of the task which allows the mind to wander.

      • Randy M says:

        It isn’t so much that men have a singular focus as it is that human minds over fit to patterns; all clouds look like something, not always breasts.

  33. baconbacon says:

    To a first approximation, all a gene does is code for a protein.

    Our first approximation was wrong. This is true for some genes, and they were the easiest genes to identify. 1 gene = 1 protein seemed reasonable for a while, but the whole god damn thing is a mess beyond that.

  34. Desertopa says:

    I think it’s a mistake to assume that an “instinct for liking large breasts” would have to manifest in the form of knowledge that’s coded in from infancy. Compare our “instinct for being afraid of snakes.” Humans or monkeys don’t have an unlearned fear of snakes, but we have a predisposition to being much more inclined to learn to fear snakes than things like rabbits or flowers. So most of the instincts we have for finding various things sexually attractive probably take the form of extra weighting in the learning process, rather than hard-coded knowledge or behaviors.

    • Cerastes says:

      Incorrect – many primates have a built-in fear of snakes, even if they’ve never seen one before, and it’s proportional to how dangerous their local snakes are to them. Lemurs display little fear and lots of curiousity, because none of the snakes big enough to eat them in MG are arboreal (even Sanzinia would have a hard time with a juvenile lemur). New World primates show more fear, because of large arboreal boas and vipers. Old-world monkeys and apes freak the fuck out (technical term), because they live in the same habitat as extraordinarily lethal arboreal species like mambas, forest cobras, and even Reticulated pythons (which have been documented eating small bears and *preying* on small humans).

      This seems to be weaker in humans, and may be entirely learned, but even lab monkeys which have never been outside will flip out and try to beat rubber snakes to death.

      • Desertopa says:

        Do you have a link to that research? It contradicts what I’d learned about the subject previously, and the research references I’d found from a quick re-check of the subject.

  35. secondcityscientist says:

    I want to push back a bit on the “only 20,00 genes” bit of this article. A huge amount of the action happens not at the level of protein-coding, but at the level of gene regulation. An individual can have a SNP outside of the coding region that affects the amount of protein produced by a gene, and produces a phenotype that way. Or maybe the regulatory SNP changes the response to a signalling pathway. Or it only changes expression in one part of the organism, not in others. There’s lots of ways to modulate the effects of a gene without messing with the protein, and people have shown a lot of ways to make complex, emergent phenotypes from pretty basic building blocks.

    Additionally, vertebrates have an even “better” variation-generating mechanism in the form of neural crest cells. Most cell populations are pretty deterministic, but for whatever reason this sub-population takes its time to differentiate and as a result is extremely plastic for a long time. Small changes to gene expression in NCCs can therefore have huge impacts on the organism’s phenotype. Facial features and melanocytes, like zebra stripes, are NCC-derived, as are many elements of the peripheral nervous system. I wouldn’t be surprised if they were responsible for some instinctual behaviors as well.

  36. Randy M says:

    I love this post for both combining two of the best organs into one topic, and for having a title that sounds like a brogrammer manifesto.

  37. uau says:

    Something that has bugged me about popularized evo-psych explanations that I consider relevant here is that they assume a direct hardcoded desire to do exactly whatever the fitness-increasing action would be. But evolution could very plausibly instead develop some kludge that usually ends up causing the desired behavior, at least in the overall environment the behavior originally evolved in. What exactly the new feature is could vary a lot while still achieving the same end result.

    For a concrete example, consider how humans avoid sex with close family. This could be achieved in the brain by having a hardcoded concept of “is close family”, ways to determine that, and a negative association from that to sex. And some simple evo-psych explanations seem to assume that this must be how things work. However, we know that this is at least not the only mechanism. Instead, or at least as a significant addition, humans avoid sex with people they spent their early years with (whether those people actually are genetically related or not), and prefer the smell of people who have different immune system genes.

    • Charles F says:

      I think this is an important point that gets missed/forgotten too much, but it’s also hard to think of every seemingly unrelated kludge that could produce the right result before talking about the result.

    • secondcityscientist says:

      But evolution could very plausibly instead develop some kludge that usually ends up causing the desired behavior

      I would go further than this. I would say that, depending on your definition of “kludge”, either evolution is a process composed almost entirely of kludges or evolution is by definition a series of kludges.

      The vast, vast majority of people have very little conception of how evolution works. Even smart people! I recall reading some speculative report about humans colonizing Mars, and they had some astrophysicist speculating about changes to the human form after generations on another planet. The changes he was suggesting would cause any biologist to say “No… no, that’s not how it works at all”.

  38. Mengsk says:

    I don’t think the 20,000 number reflects an upper limit on the complexity of the features that can be genetically determined. Protein coding genes, which is what you count to get the 20,000 figure, only make up 1%-2% of our genome. The function of the remaining 98% isn’t understood in depth, but probably plays an important role in turning genes off and on at certain points in development, which is where most complex behaviors are encoded.

  39. TheRadicalModerate says:

    Instead of thinking about this in terms of genes, it’s better to think in terms of the connectome, which is of course just a bag of various axonal projections that’s been determined genetically. I have no clue how many hox-like genes are involved in controlling projections in brain development, but it’s a safe bet that they’re heavily vertically encoded, i.e., a fairly small number of genes interact to express proteins that say things like “follow the lateral geniculate to V1 road!”, which will be only slightly different than the combination of proteins that drive stuff from the auditory cortex to the various language areas, or any of the other gross neurological wiring.

    From an information-theoretic perspective, you can encode a few hundred gross projections with a few tens of genes, and that’s pretty much all you need. So the question really isn’t, “What gene makes big breasts attractive?” Instead, it’s something closer to, “What combination of projections makes it highly likely that the perceptual correlates of big breasts will tickle some hormone-secreting region that will result in attraction?”

    As for the specific case of the “big breasts = attraction”, you also have to remember that babies are really, really good at finding nipples. Whether they do that by texture or smell or mouthfeel, it’s pretty likely that everything needed comes pre-encoded in some combination of brainstem and/or limbic system structures. From there, it’s pretty likely that any learned behavior about identifying breasts will find some way of projecting back into the basic “nipples are good!” morphology, and from there things become a lot more plausible. Biasing how strong that connection is with a few sex hormones doesn’t seem very unlikely from there.

    The other thing to remember in all of this PP stuff is that, even though the projections are hard-wired, all the learning happens bottom-up. The way you train deep-learning networks is to run the first (bottom) layer unsupervised for a while, then add layers one at a time. So the upper layers basically become the supervision for the lower layers even as they’re doing their own (unsupervised) feature detection. From there, it’s all about what layers you hook to what other layers, which gives you specific behavior.

    So the bottom regions of neocortex, which mostly have explicit projections to/from various (genetically determined) brainstem and limbic structures get very specific sets of patterns to do feature detection on, and the inter-cortical projections from those regions to other regions will derive fairly common sets of features, which will ultimately result in certain concepts being roughly encoded in certain specific regions. From there, a projection between the “nipples are good!” and “oh, look, big breasts!” region seems a lot more plausible.

    • Randy M says:

      Whether they do that by texture or smell or mouthfeel

      Not to take away from your comment, but one of these things here is not like the others. I would expect taste and smell to be a lot easier to hard code than sight, sound, or feel (excepting temperature) as these two senses are chemical detection, which would work by producing chemical receptor proteins, something DNA already codes for extensively just for internal communication. Sight or sound, on the other hand, have to translate an array of sensory stimulus into nerve impulses which need complex and likely trained brain areas to interpret.

      It’s like the difference between a computer program that tries to interpret faces from 2-D images, and one that reads binary signals.

      • TheRadicalModerate says:

        I think that’s mostly right. Instinctive responses are going to rely on a lot less neural circuitry than more complex learned ones. But the question is whether you can roughly hook a region that’s likely to contain a particular learned response back to the instinctive circuitry. The answer to that depends on how likely a particular set of concepts is to reside in a particular cortical region, and that’s a lot more likely than you’d guess.

      • eelcohoogendoorn says:

        That idea doesnt really mesh with the way smell actually works in practice; the olefactory system doesnt encode genes to make proteins to detect specific molecules. Really is more complex and unknown than the following; but it is more like the system codes for proteins that detect particular molecular bonds, or molecular subgroups. We have about a 1000 of such detectors; and picking out any particular smell amongst these amounts to detecting the ‘fingerprint’ this molecule leaves on those 1000 receptors.

        Hence, genetic coding for a particular smell is in practice as much a problem of neurology as coding for, say, a particular sounds.

  40. Nornagest says:

    I’m guessing we don’t innately find (most) sex-linked traits attractive. I’m guessing instead that sexual attraction gets bootstrapped out of some combination of secondary sex characteristics, gender roles, and a very few highly conserved traits that brain development can cue off directly.

    It could look something like this (not saying it does, only that this is a plausible sequence): because of how olfactory reception works, it’s possible to have proteins coding specifically for sensitivity to compounds associated with people of the appropriate sex and of reproductive age. Over some critical period, kids’ developing brains learn to associate those pheromones with a certain set of characteristics, some biological and some culture-bound, which couldn’t be coded for directly: these might include wide hips (or broad shoulders), prominent breasts (or muscles), a tendency to wear dresses (or suits). There’s a strong correlation with inferred gender for most people, but it doesn’t map perfectly to gender, explaining how a particular couple might have “chemistry” or an individual might have a “type”. (Might also have something to do with bisexuality.) Probably this happens pretty early. It stays dormant for a while, but during puberty, a switch gets flipped somewhere and suddenly these characteristics start cueing sexual attraction.

    This accounts for a lot of the stuff we see about sexuality in the wild: different people like different features in different proportions, and culture matters too. If your culture doesn’t put much emphasis on breasts as a sexual cue you probably won’t grow up to be a breast guy. It even explains fetishes: occasionally, something goes weird in the bootstrapping process, and a seemingly random association forms: to shoes, maybe, or to rubber or feathers.

  41. cassander says:

    But evolutionary psychologists make claims like “Men have been evolutionarily programmed to like women with big breasts, because those are a sign of fertility.” Forget for a second whether this is politically correct, or cross-culturally replicable, or anything like that. From a neurological point of view, how could this possibly work?

    The same way female peacocks came to like peacocks with big tail feathers?

    Which is to say, that those capable of investing energy in a secondary trait generally produced stronger more successful offspring than those that did not, so the peacocks that didn’t like big tails were, over time, outbreeded by those that did.

  42. Ryan Baldini says:

    I don’t really care about this topic, or have a strong opinion about the large-breast-attraction idea, but I think you’re hitting on the relevant fact in the end:

    Somehow, the vast majority of humans are heterosexual: they are actually attracted to members of the opposite sex. We can assume that this is at least partly encoded in the genes. 20,000 genes aside (mind you, DNA does more than encode protein-coding genes), *somehow* our DNA manages the feat of reliably making us attracted to the opposite sex. Why wouldn’t it be able to detect breast size, and do something about it?

    “One heuristic might be to have a strong prior against any claim in which genes can just reach into the level of already-formed concepts and tweak them around, unless there’s a really strong reason for evolution to go through a lot of trouble to make it happen.”
    Trouble? What trouble? If a mutation arises that slightly increases fitness, selection will tend to make it increase.

    • Ryan Baldini says:

      “And if you want to posit an entire complicated breast-locating system made up of hundreds of genes, remember that we only have about 20,000 genes total. Most of these are already involved in doing things like making the walls of lysosomes flexible enough or something really boring like that.”

      You’re probably making a mistake here. It’s not that there are “genes for boring cellular function” and, separately, “genes for more interesting, higher-level stuff.” Humans are cellular organisms: it all operates through cellular biochemistry! If there are genes that do influence human males to be more attracted to breasts than otherwise, they probably operate through a complex combination of these boring cellular functions.

      Consider the most recent study (two days old) linking particular genes to intelligence: http://www.biorxiv.org/content/early/2017/09/06/184853.1.full.pdf+html.
      VAMP4 is mentioned as a candidate (http://www.genecards.org/cgi-bin/carddisp.pl?gene=VAMP4). Is VAMP4 “a gene for IQ tests”? Potentially yes, in the sense that one variant of it might cause its holders to do better at IQ tests, somehow. But it also has a specific biochemical function: apparently it is “involved in the pathway that functions to remove an inhibitor (probably synaptotagmin-4) of calcium-triggered exocytosis during the maturation of secretory granules.” If it turns out that variation in VAMP4 really does affect IQ, then it must work through some very complex process that involves this biochemical function. But there’s nothing in VAMP4 that makes it inherently devoted to IQ to the exclusion of other funcions. Hell, even fungi have the VAMP4 gene.

      This speaks to your “only 20,000 genes” dilemma: these genes are all doing some basic chemical function, at the low level. But *variation in them* can produce variation at higher levels. A gene “for intelligence” doesn’t steal a gene away from “boring cellular chemistry.”

  43. catherio says:

    Former neuroscientist here. There’s some primate neuroscience literature on the physical layout of high-level selectivity in the brain that I think can shed some light.

    In essence, the spatial layout of high-level concept-selective areas is roughly the same from animal to animal (or person to person); even for concepts that are completely novel, which no monkey in ancestral history would have had to learn.

    This implies that genes don’t have a challenge quite as hard as “how the hell do you even code for ‘things that look like snakes'”; it only has to code for the *physical area of the brain* where the “snake” concept reliably ends up.

    Marge Livingstone’s lab, results on the spatial consistency of totally novel domain representations, by having a few poor monkeys repeatedly drill a specific weird task all their lives:

    Work suggesting that this “proto layout” is present at birth:

    It’s probably not only true for monkeys. fMRI results show coarseley similar concept layouts in humans, e.g. Uri Hasson making subjects watch movies in brain scanners showing similar spatial responses:

    • Scott Alexander says:

      Why would this happen?

      And how much does it happen? Does the concept “fax machine” exist in the same place in every human brain?

      And if it did, are concepts arranged in some kind of conceptual order, such that the distance between the concept “fax machine” and “otter” is related to how like or unlike those two things are?

      If so, does that suggest some kind of objective (or at least objective-relative-to-humans) ontology of concepts?

      Asking for a friend.

      • HaltingProblem says:

        I posted my theory in the open thread. Short answer: it’s a hash table.

      • catherio says:

        Jack Gallant’s lab has some pioneering work and great visualizations that might help get a foothold on these questions: http://gallantlab.org/brainviewer/huthetal2012/ -> in this interface, you want to start clicking on regions on the cortical surface, especially in interesting areas like FFA, to see which concepts a given voxel’s activity is strongly correlated (large red circles) or anti-correlated (large blue circles) or uncorrelated (small circles) with. You can also click on concepts in the right-hand pane, and see what the associated pattern of activity looks like.

        Concepts like “fax machine” have a distributed representation, not a single-point-localized representation. That is to say, a whole constellation of different parts of the brain vary their activity with the presence or absence of fax machines. For example, this voxel goes crazy for all sorts of manmade things, including indoor places and electrical devices: http://gallantlab.org/brainviewer/huthetal2012/?vox=7326&vert=12456

        Note also that the activation pattern across the brain shifts depending on what you’re paying attention to: http://gallantlab.org/brainviewer/cukuretal2013/

        I linked above to Marge Livingstone’s work, which deals with concepts that the monkeys have become *highly expert* in (they were trained to discriminate three categories – tetris shapes, helvetica characters, and cartoon faces – for hundreds or thousands of hours). Domains of high expertise end up recruiting whole swaths of cortex to specialize just in those domains. But that’s not to say that parts of the brain outside the newly-created “tetris shape area” aren’t contributing anything to the representation tetris shapes; they’re just not responding *specifically only* to tetris shapes.

        I’d guess that *if* two different humans were to both become really really expert at fax machine tasks (recognition, discrimination, etc.) – like, 10,000 hours of expert – *then* the area of the brain that became specialized for fax machines would end up in the same place.

        I’m not sure which of the many flavors of “why” your initial question is. That said, it seems like your own post answers one potential “why”: one good reason why evolution might’ve given rise to consistent spatial patterns of concept formation could be in order to allow genes to manipulate specific *concepts* using *spatial patterns* of gene expression.

  44. moscanarius says:

    And if you want to posit an entire complicated breast-locating system made up of hundreds of genes, remember that we only have about 20,000 genes total. Most of these are already involved in doing things like making the walls of lysosomes flexible enough or something really boring like that. Really it’s a miracle that a mere 20,000 genes can make a human at all

    I am not trying to be facetious, and I know many people already pointed this, but these 20000 genes are more than they look. Alternative splicing is very common in human gene expression, so a single locus does not necessarily encodes just one polypeptide. A few genes vital for morphological development are used over and over and over in different cellular contexts to yeld quite different structures (hox genes, for example). Most proteins can interact with other proteins (and we may be underestimating the extent of this, since it is worksome to characterise in the lab), and the resulting complexes may acquire new, unpredicted functions. And many proteins have moonlight jobs in the cell (glycolytic enzymes being famous for this), so we may have the case where 20 thousand are doing the work of 80 thousand.

    • joscha says:

      Perhaps it might make sense to see DNA as an operating system of the cell, i.e. a set of routines that the cell can execute in response to its current state and the conditions it measures via the receptors at and the chemicals passing through its membrane. The 20000 genes are basically 20000 different subroutines. The space of useful subroutines for single celled organisms has probably been almost exhaustively conquered via massively parallel brute force search in the first few hundred million years of cellular evolution, and new types of genes were usually the results of substantial changes in the environment on earth (which at some point allowed for complex multicellular organisms). Once you see our organism as a giant and slightly fuzzy cellular automaton, then 20000 different (and often quite complex) transition rules gives you quite a lot of complexity, especially once you realize that each cell holds quite a few bits of state (to account for its current differentiation) on top of that.

  45. void_genesis says:

    It seems to me like two issues are getting muddled here.
    The example of focus here is how breasts are sexually attractive. One element I haven’t seen anyone here mention is that recently it was discovered that the small glands around the edge of the areola secrete a pheromone that stimulates baby suckling. I always wondered what those tiny bumps are for. Women with underdeveloped ones are more likely to have trouble breast feeding. It is quite possible there is a pheromonal connection to providing an instinct in males finding breasts attractive.

    The broader topic is how genes encode instincts in animals. In humans this is difficult to study because without controlled trials of babies without normal stimulus (very hard to justify) it is difficult to know what parts of human behaviour are pure instinct. Animals would be a better place to start. And even animals with extremely reduced genomes seem to have the same capacity for instincts. It is also possible to breed out very important instincts, for example many strains of chickens have lost the ability to sit on eggs.

    I think it would be a mistake to assume that genetics is the only factor at play in the creation of brain structures that support instinctive behaviour. We know very little about control of morphogenesis for all structures, though it is starting to look like bioelectricity is a key ingredient. Mitochondria seem to be a key component of brain function as well. How these systems are inherited is currently unknown. Behavioural adaptations may be passed on to offspring through non-genetic mechanisms.

  46. Herman Stone says:

    Shouldn’t the term be “hypoprior”?

  47. Phil Goetz says:

    Experiments have shown that monkeys are “semi-instinctively” afraid of snakes, and there’s plenty of anecdotal evidence that other animals (horses, elephants) are also. You might also find studies on instinctive fear of spiders and scorpions, though I don’t recall for certain.

    I say “semi-instinctively” because the way these studies are typically used in a psych class is to address genes vs. environment: monkeys (so the story goes) with no exposure to snakey stimuli aren’t afraid of snakes, but monkeys can acquire fear of snakes (or sticks), e.g., by seeing other monkeys react to them with fear, much faster than they can acquire fear of supposedly comparable stimuli.

  48. jamaihalem says:

    Well, even if any gene encodes (with few RNA-encoding exceptions) for one proteines, they interact in a bunch of ways as repressing or enhancigng other genes pr regulating its area of expression. All in all, its hierarchical and multifactor-conditioned expression prabably maximises the amount of information it contains. Also, as some other peopld here have boted,genes have a pleiotropic effect in many cases, so its not just “bites” all the time.

    Although I’m by no means knowledgeable in this, I’d propose some mechanism like that: we have this capacity to recognise our keen and its located in some area in our brain and, as long as we interact with humans, it develops correctly (a million people must have said that before). As we humans have some things in common, there’s no necessity to assign them to either sex, but there are some evident differences between both sexes, so there must be two different sub-regions, one for each sex. I bet some basic information is always there, at least potentially, some basic clues, say “curvy, high pitch & light” vs “heavy, square-like and low-pitch”, wich serve to filter information at the most primary level but this information is shaped to mate societal standards and is also heavyly accompanied by only-social markers (such as colour codes or gendered names or hairstyles), wich are also saved in the same sex-region as the natural clues. Even some natural tendencies can be masked by social conventions or directed towards some specific sexual traits rather than others (see modern western standard for women vs. Rubens’ chubby one vs. Roman small-nipples policy), but I don’t think they are totally hidden in any case. Also, probably children fall appart from each group to some degree (after all, aren’t “male” and “female” more or less recently invented words?). Note that I’m just talking about recognising each sex, not about preferring it. Of course, natural clues should work well for healthy individuals, as unhealthy individuals’ comoany usn’t valuable enough (it can pose even a threat) and it’s ideally scare enough, so why bother about them?

    It seems that just one hormone (=one or few genes regulating it) accounts for most sex differences, being testosterone much more prevalent in males and estrogen in females. Probably there’s some common mechanism for primary sexual attraction rooted in our brains and theese hormones just have to “wake-up” it in puberty assigning it to a concrete path towards one sex or another. As for gays or bis, it could happen that the paths are altered by some mutation, as many of them are just as manly or effeminate as any normal folk for all the other sexual traits that are regulated by sex hormones.

    Now, as brain is a very complex organ but it’s probably shaped by a not-so-hughe number of genes wich interact in a very complex way. So those regions and their instinctive paths (latter reformed and ampliated by education in a bayesian fashion) may not need so much genes “invested” in if wisely mixed. Also, sexual preferences could be regulated by a small bunch of genes in chromosome Y deactivating estrigene and enhancing testosterone production. Isn’t that beautyful?

    Finallt, it comes as needless to say that almost any investment could be justified in order to make sexual organisms mate with the opposite sex individuals, specially if tgey are fertile. All in all, a non-sexually-discriminating species would be no more effective than a species where only females ever feeded (or less so), as copulating requires a huge investment so you better invest wisely.

    PD1: It comes without saying that I by no means think non-heterosexuals are worthless in evolution (see keen selection), let alone that they should ve discriminated against.

    PD2: I think chemical differences that play a role in sexual attraction are probably also instinctive, as they aren’t probably the same chemicals a child would perceive from his mother. However, I know a story that demonstrates that the way theese stimulus are perceived is still shaped by live experience. It happened in my country about a century ago or less, when showers weren’t a thing in the countryside yet. A doctor would be visited by a woman who smelled terribly so before the inspection she was made take a shower. Some days aferwards, her husband came protesting thst the doctor had stolen her the “female odor” of hers. In that same village, people now probably wouldn’t stand the woman, as women are expected to have a light, clean odor now, as a result of shower widespresding. However, even now sex-linked odors act as sexual stimulus.

  49. MrApophenia says:

    You want to really scratch your head on the instinct question, look at instinctual behavior in dogs. My dad has an English Shepherd, and he raises goats. He did not train the dog to actually herd the goats. But from puppy-hood on, it herded them, no training needed.

    When I was a kid we had a dog that was kind of a generalized mutt, but somewhere in its ancestry there must have been a hunting hound. Because when he saw a rabbit he wouldn’t just chase it – if he chased it into the weeds, he’d stop at the edge of the weeds right where the rabbit went in, and go into a perfect-form hunting dog point. Again, we never trained him to do this (and we had this dog from very early infancy – no one else did either.) We never took him hunting. But he still knew to do it.

    These are complex behaviors, and they’re things humans first trained into dogs relatively recently, evolutionarily speaking! But that training has somehow already made the leap to hereditary knowledge.

    • Sam Reuben says:

      I’d be interested in seeing which of these behaviors are already exhibited, in some form or another, in wolves. My money’s on “a lot of them.” Herding is fantastic behavior for getting a group of prey into nice formation to be chased, or splitting one off so it can be killed (a critical part of domestic herding, minus the killing). Hunting point is a good way of indicating to pack members where prey is, so that you can gang up on it together. Other behaviors are the same way: fetching is good for feeding pups and the pack, barking alerts is good for protecting the pack, etc etc.

      So what human breeding has likely done is take these complex behaviors, which were bred into wolves by standard evolutionary forces over evolutionary timescales, and chosen one or another to emphasize in different breeds. This kind of selection takes absurdly short periods of time to reach an effective conclusion, and results in marked differences between breeds in appearance and behavior. So not as impressive as it might seem, but definitely support of the whole instinct-is-powerful deal. (Dogs can also be trained to ignore their instinct! That helps out some certain free-will positions a lot.)

  50. Chris Said says:

    How do we get breasts out of Bayes Theorem?
    Bayes Theorem is an anagram of “Breast, Homey!”

  51. S. Aiv says:

    It doesn’t seem particularly difficult to learn attraction to breasts if you assume you can encode for attraction to pheromones of the opposite sex. You grow up associating that “smell” with people of the opposite sex, and cluster that with all the other things that you associate that with. Then you Pavlovian reaction that, and you’re done.
    You can even imagine that the more explicit the characteristic, the more attractive. I.e. if a big colorful peacock tail as compared to your bland peacock tail differentiates you from the opposite sex, then a bigger tail is even better.
    On a side note, I wonder if Pavlov’s dogs salivated more if you used a bigger bell / rang louder / faster

  52. But this is one of the rare cases where evolution might really want to devote some big fraction of the 20,000 genes it has to work with to building a Rube Goldberg circuit.

    I think it’s confusingly worse than that. Think about the range of human emotions and seemingly instinctual behaviours that key on social queues. Like for example jealously, which has to somehow process social standing and/or possessions of another person. Or in-group loyalty. These seem both remarkably consistent (amount varies but almost everyone has them – suspiciously much to be purely learnt thing) and complex. How could you possibly encode something as high level as that without learnt social concepts? Obviously you can’t and need a ‘close enough’ trigger like the red dot, but then I can’t really think of anything simple enough to encode but still close enough to be useful…

    ps> apologies for totally anthropomorphizing evolution there

    • Worley says:

      One thing that makes you acutely aware of an instinct is if everybody has it *but you*. In my case, I don’t get jealous, and it’s kinda hilarious to walk around watching people do objectively quite stupid and counterproductive things under the thrall of that instinct.

      • marvy says:

        Now you made me curious. Example?

      • sconn says:

        Me too. I do not appear to have the instinct for revenge. People talk about wanting to see bad people punished somehow and I’m like “but why?” I feel the lack of this instinct is why I left religion; I couldn’t see how a good God would punish evildoers, yet many of my friends felt that there would be a cosmic injustice in the world if he didn’t.

  53. willachandler says:

    Excerpt from sociobiologist Ed Wilson’s The Creation: an Appeal to Save Life on Earth (2006):

    Is it so strange that at least a residue of habitat selection persists among the human instincts? The programmed search for a correct environment is a universal of animal species for the best of reasons — it is an imperative of survival and reproduction.

    My favorite example, being an entomologist, I suppose, is the behavior of the fairyfly, a tiny [note:~1/2 mm-long] myrmarid wasp that parasitizes the eggs of dytiscid water beetles. After flying about and finding the right places to mate, the female begins her search for prey. She lands on the surface of an appropriate body of water that may hold eggs. She stands there at first, her little body held in place by surface tension. To submerge she digs — she is too light to dive — through the surface tension with her legs. She then swims downward, using her wings as paddles. Reaching the bottom, she searches about like a pearl diver for the eggs of the water beetles, into which she inserts her own eggs.

    All of this is accomplished with a brain no larger than a dot made by a fine pen.

    That core rationalist notions like “Bayes Theorem” commonly appear nowhere in sociobiological works like Wilson’s The Creation will give SSC/LW-style rationalists reason to wonder whether sociobiologists like Ed Wilson grasp what rationalists are talking about (and conversely, give Wilson-style sociobiologists reason to wonder whether SSC/LW-style rationalists grasp what sociobiologists are talking about).

  54. algorizmi says:

    A single gene can have a lot of variety, in fact the majority of genes in animals are alternatively spliced. Some genes such as cadherins (important[1] for dendrite growth pattern in neurons) or immunoglobulins (antibodies) offer a very high degree of alternative splicing. In a given individual one cadherin gene can code for ~30,000 different proteins. So there’s a fair bit more than twenty thousand levers to pull on this rube goldberg machine.

    [1]: Cadherins are transmembrane proteins with variable domains exposed on the external surface. Since each neuron only expresses one variant of the gene it can recognize and grow away from itself on this basis. Allowing for efficient connections to different neurons.

    • Randy M says:

      Since each neuron only expresses one variant of the gene

      How does that work, I wonder? Based on chance, since having such a high number of combinations ensures differences most of the time, or based on some feedback mechanism of nearby neurons?

  55. Joe says:

    I want to point out that the title of this post is wrong. Your real question is “How do we get breasts out of 20,000 genes?” and has nothing to do with predictive processing per se. If tomorrow we discovered incontrovertible evidence that PP is completely wrong, we would still be faced with the same basic puzzle of how complex features and behaviours can be coded with such a limited set of genes.

  56. wildtypehuman says:

    Are breast-attracted people also attracted to large “breasts” on men? (In the large dude sense.) If not, I think that suggests something about the necessary complexity of any breast appreciation algorithm–seems like simple geometry is probably not the driver.

    • Deiseach says:

      I think that is definitely where socialisation comes in; attraction to large breasts then being directed to the “proper” targets i.e. large-breasted women are sexually attractive; male breasts are not to be considered in a sexual context in the same way, and men with visible bosoms are regarded with disdain or at least as mildly humorous, depending on whether the breasts are the result of fat or extreme muscle development – see Groucho Marx’ quip about “Samson and Delilah”:

      Giving his reasons for not enjoying the screen epic Samson And Delilah, Groucho Marx quipped: “No picture can hold my interest when the leading man’s bust is bigger than the leading lady’s.”

  57. Murphy says:

    I remember a while, walking along the street and seeing a bus flash past.
    The bus windows were tinted with sunlight reflecting from them and I was, give or take, 50 yards away.

    In easily less than half a second as the bus passed in and out of view through the window I could

    1: recognise a human face.
    2: recognise the sex of the individual. High confidence.
    3: recognise that the person very likely had downs syndrome. High confidence.

    That’s one hell of a pattern recognition system given I wasn’t trying to do any of those things.

    The thought of trying to train an image clasifier to do a similar task through similar visual noise would be quite daunting.

    It also implies that my pattern recognition system considers some things to be super super important to track.

  58. Worley says:

    It seems like the question boils down to an experimental problem. An example I’d like to see studied, because it’s so simple, is nesting behavior in gulls. It turns out that gulls have an instinct to sit on eggs. The egg-detector isn’t very particular, the object only has to be approximately round (a cube will do) and have the surface visual pattern that the species’ eggs have.

    In a way, it may be hard to construct instinctual behavior using genes. Even wiring an intrinsic behavioral response to a fixed pheromone is more than we understand. But the individual’s problem isn’t actually having the instinctive pattern recognizer work *well*, only having it work at least as well as in other individuals. You don’t have to be perfect, only better than all the other things that can be done with 20,000 genes.

  59. ScarecrowBoat716 says:

    I’m late to the discussion but is it possible autistic people’s problems with motor control relates to predictive processing? If your brain uses predictive processing to move the muscles, perhaps a flaw in being able to make predictions is also responsible. I myself believe I am on the autism spectrum (undiagnosed but I meet many of the classic symptoms) and have issues with fine motor movements.

  60. Lambert says:

    No mention of Fisherian runaway anywhere?
    Traits that are selected for sexually and genes causing the opposite sex to select for it tend to reinforce one another.
    The phenotype ‘preference for large breasts’ coexists with the phenotype ‘large breasts’.
    If this effect can make a peacock grow such a large and impractical tail, it can no doubt cause a peahen to be inclined toward such tails.

  61. Joyously says:

    People have tried to train neural networks to learn languages–has anyone tried to get a neural network to distinguish attractive people from unattractive people? You could make a feminist project out of it–feed the machine advertising and TV shows and see what it thinks a beautiful woman looks like.

  62. Doctor Mist says:

    Please forgive a naive question: How do we know we have only 20,000 genes? What is a gene for the purposes of this discussion?

    From Mendel’s position, “gene” would just be an atomic idea connected to a particular trait. I’m guessing that in modern terms, a gene is a thing that encodes a protein. Are we made up of at most 20,000 distinct proteins? I’m pretty sure I’ve read that a given DNA base pair could participate in the construction of more than one protein (a horrible hack you sometimes see as well in extremely space-limited machine code).

    In any case, it seems odd for people to suggest, as I saw earlier in these comments, that the limited number of genes means that almost all behavior must be learned. Those 20,000 genes must in some sense be ultimately responsible, whether what they code for happens during embryogenesis or during high school? The encoded proteins could in combination (and 20K**n is really high for even small n) lead directly to certain behaviors, and they could in combination lead to a three-pound sloppy general-purpose computer that learns those behaviors.

    I can see arguments either way concerning the efficiency and reliability of these two approaches for any given desirable behavior. But I find it hard to see why the latter would be easier or require fewer genes. Making computers is hard.

    • Murphy says:

      Some genes do indeed code for multiple protiens. In DNA it’s likely less about space and more that it’s easier to evolve 2 versions of X based on cutting out or adding in a middle section sometimes than to evolve an entirely new protein to do the job. Cells have and endless list of “works just well enough most of the time” hacks.

  63. suntzuanime says:

    Evolution has to pack really tightly to fit all the important information into our genome. Packing tightly means using on-site materials whenever possible. That means you don’t encode the whole idea of what breasts look like into the genome, you just encode enough to find a breast and then you refine your idea of what they look like using that model and attach it to whatever hook in the brain is waiting to receive it. This can explain fetishes (oops you found a foot when you were looking for breasts) although probably what’s going on with homosexuality is somewhat different.

  64. Ryan says:

    If evolution can’t make men attracted to women’s breasts how can it make men attracted to women at all?

  65. Worley says:

    A really odd case is transvestism. Reports from transvestites are that it’s a deep-seated, unlearned urge or preference, which suggests that it’s due to an instinctive mechanism that is operating in a not-most-adaptive way. But they also report that there is no component of wanting to change sex/gender or to masquerade. Which suggests that the instinctive mechanism is about clothes, and the gender-specificity of different types of clothes. And that’s very odd, because as opposed to other putative human instincts, which deal with things that primates have been dealing with for millions of years, clothes are at most a few hundred thousand years old. It’s hard to understand not only how you hard-wire such an instinct but also how you do it so quickly.

    • Dedicating Ruckus says:

      From the inside view, many (most? all?) paraphilias also feel deep-seated and unlearned, but I’d be hesitant to conclude from that that they actually are. In particular, an instinct that gets learned during childhood, but doesn’t actually take effect until puberty, will probably feel as if it came straight from your internal nature even if it really didn’t.

  66. sconn says:

    It’s experimentally demonstrated that infants are attracted to circles with darker colored dots in the middle. Which makes sense. They are aimed at their food supply. And that seems simple enough to code.

    However, given the wide variety in people’s tastes regarding the opposite sex, the cultural differences (e.g. some African cultures go for overweight women, the Japanese find necks attractive, etc.) I’d probably put my money on, “men are attracted, in general, to *women* and when a trait is strongly associated with women they’ve been attracted to, it goes in their memory as an attractive trait.”

    But how do they know which women to be attracted to in the first place? I dunno, pheromones maybe? This sounds more difficult to code than breasts, but it’s probably significantly easier, because it’s a single chemical formula. Or it could be some very basic visual cues that help us tell someone is female. Basically rather than “men like big breasts because milk production” we could say, “men like big breasts because the bigger the breasts, the smaller the chance they messed up and it’s actually a dude.” Basically all traits men are attracted to, are ways you can tell someone is female in the first place. The more hourglass-shaped, the more hairless, the more delicate the facial structure — the better the odds that reproducing with this person will work at all, because they’re the opposite sex.

  67. Douglas Knight says:

    Sometimes the development algorithm uses a learning algorithm because it wants individuals to learn different things, and sometimes it uses a learning algorithm as a form of compression. It spends time in life to save space in the genome. Even though cats use learning algorithms, they probably all end up with the same vision system.

    Many animals, such as prey mammals, are born being able to see. Predator mammals, such as cats, can be protected by the mother and don’t need locomotion and vision from birth. They can afford a period of vulnerability in which to learn to see. But the prey mammals demonstrate that sight can be hard-coded. Feline vision is probably almost hard coded, just with a few details to fill in. Outside of experiments, all cats probably fill in the details the same way, and, indeed, the same way as lambs. So you might say that the system is hard-coded, it just takes three months to bake. If cats are afraid of snakes, the shape of the snake could be defined in terms of the low-level concepts, which are the same for all cats outside of experiments.

  68. wiserd says:

    It’s worth noting that not all genetic information is found in genes. So the full size of the genome should be considered. The notion of ‘junk DNA’ has been grossly over-applied.

  69. Bram Cohen says:

    So glad there’s work being done on the very important question of why breasts are so fascinating

  70. Ben Kuhn says:

    If you’re OK with people having gender-related instincts, then maybe the relevant instinct is “be more attracted to things that you’re more confident are female”? This explains attraction to all types of sexual dimorphism (e.g. wider hips also).

  71. void_genesis says:

    It is worth pointing out wide female hips are an adaptation probably driven by increasing baby head size. They probably became universally sexualised after they developed for practical reasons. Breasts are the opposite in that they seem to have been driven by sexual selection alone, though can we realistically picture a tribe where flat chested women fail to get pregnant? The mechanism is more likely to be through resource provision by males leading to higher reproductive success. Pubic hair and beards might also be driven by sexual selection, though it is more likely that they are functional in that they increase surface area in regions that probably secrete pheromones and help evaporate those important chemical signals. Add in kinks and humans seem to be able to associate anything with sex. An interesting question might be- how malleable are human brains with regards to kinks? (Seems to be a critical window prepuberty for making associations). Are kinks heritable? Or maybe just the tendency to form extra-sexual paraphilias has some heritable basis (might correlate with neurotic tendencies for example), regardless of the particular fixation.

    • Skivverus says:

      An interesting question might be- how malleable are human brains with regards to kinks? (Seems to be a critical window prepuberty for making associations). Are kinks heritable?

      For the first question, it varies between individuals: some, at least, have reported acquiring kinks after being introduced to them by significant others, while others have had them since well before puberty (in my case, by age five). I’ve also heard of at least one case where someone lost interest in kinks they previously had. All that’s anecdata, though: I’m unaware of any formal research on the matter, though I haven’t searched for any either.
      As for heritability, I’d expect inheritance of capacity more than of particular interest, but also for particular interests to run in the family anyway, due to shared environment: people don’t move that often.

  72. faucetscomplete says:

    Shallow well pumps ar typically a preferred selection for moving liquids settled at a depth of no quite twenty five feet below ground level.
    get more information visit our website.

  73. justathoughtyouknow says:

    Bayes is one of *the* fundamental rules in statistics, but why is it mentioned just so commonly here? The number of fields of mathematics involved from conception to…err
    ..conception enticing objects and images is quite large. Bayes isn’t the only rule in math in that entire process.

  74. vV_Vv says:

    To a first approximation, all a gene does is code for a protein. How, exactly, do you design a protein that makes [men find big-breasted women attractive]/[women develop breasts]? I mean, I can sort of imagine that if you know what [neurons carry the concept]/[epitelial stem cells specialize into the tissue] of “breast”, you can sort of wire them up to whatever region of the [hypothalamus handles sexual attraction]/[skin that runs over the milk lines], so that [whenever you see breasts you feel attraction]/[they form nipples which later grow into functional breasts]. But number one, are you sure there’s a specific set of [neurons]/[stem cells] that carry the [concept] “breast”? And number two, how do you get those [neurons]/[stem cells] (and no others) to express a certain gene?

    And if you want to posit an entire complicated breast-locating system made up of hundreds of genes, remember that we only have about 20,000 genes total.

    Most genes seem to do multiple things each. It’s a sort of holographic encoding.

    Maybe “red dot” is primitive enough that it’s easier to specify genetically than “thing that looks like a mother bird”?

    I don’t know about gulls, but it’s well known that humans have specific brain regions that respond to faces. If the human brain instinctively knows what a face looks like, then it is conceivable that it also knows what a breast looks like, in particular because before the modern era, the breast is your only source of food for your first months of life.

    • void_genesis says:

      That “specific brain region” for faces is a real phenomenon, but is also very flexible in how it is used. People who are experts at recognising different models of cars use the same region to quickly tell them apart. Recent work on monkey brains in this area has managed to pin down individual clumps of neurons that each judge different aspects of a face such as distance between eyes and between nose and mouth, so much so that they can reconstruct the face being seen just by measuring the activity in each region. I think they reported about 100 different metrics used to measure each face. So we have this paradox where most brains learn to do more or less the same task in the same brain region using the same strategies, but the brain itself is extremely adaptable to being used in different ways. I look at it like this- human legs are structured in a way that means that most people end up optimising the way they use them in pretty much the same way, but people who were born with a missing foot or on a planet made out of marshmallow would go through a different process of optimisation of use in response to the environment around them. Brains are the same. “Breasts” aren’t hard coded as such and more than walking is hard coded in legs, but the structures respond to the stimulus of breasts in a way that is predictable enough to have adaptive utility.

  75. maxkesin says:

    While there’s certainly some complexity in the encoding of breast attraction it’s dwarfed by the complexity of actually building the anatomical breast and other parts of the body. Clearly genes compress information uncannily well, but IMO the question of encoding breast attraction is not very strong in this light

  76. Rusty says:

    But my cat somehow knows about mice so that even when he sees them on TV he goes behind the set to try to find them. He sure seems to have a gene for ‘mice’. Or maybe he learned all about them from his mum? Instincts seem to be pretty hard wired.

  77. Jakub Łopuszański says:

    Recently I was wondering about similar problem: how is the fear for spiders encoded?
    My friend came with suggestion that perhaps we do not fear spiders but simply “do not like high frequency” and fast moving, numerous, thin legs, cause some high frequency pattern in neurons? That could be encoded easily.
    Similarly: round, full breasts, are perhaps some low frequency, gentle curve pattern.
    And perhaps it’s not just coincidence that breasts shape are easily expressed concept, if the correlation runs in opposite direction: breast evolved to match this pattern?

    Another idea has to do with how do two brains (which obviously differ by internal structure) communicate the concept of breasts between each other: by language. Language being the standard serialization protocol of brains. Perhaps, evolution, being a sloppy developer, which doesn’t care about separation of concerns, nor clean architecture, just puts a string “boobs are cool” somewhere in dna, then waits for the child to learn the language from other humans, and finally decodes the message using the same deserialization mechanism which is already built at this age to deal with other humans. I oversimplify here, but just want to pose a possibility that genes and memes can coevolve and interfere to produce cool effects. For example a rule “do whatever your parents did” seems to be short, but to unpack it you need memes.

  78. Jakub Łopuszański says:

    Step one: make each individual shy, and cover parts of itself which are not identical among everyone
    Strip two: make each individual interested in novelty

    Effect: boys which see boobs and haven’t seen them on their own body seem interested in boobs.
    More important effect: individuals prefer opposite sex
    Side effects: fetishes, interest in forbidden fruits, joy in peeping

  79. Andrew Klaassen says:

    Neurons do a lot of crawling during very early development. They’re like hippies on the road, traveling to find out who they really are. Some of them crawl back and forth, back and forth, between the same two areas. Some of them set out on longer adventures. In both cases, the signaling molecules they encounter along the way tell them a little bit more about who they are and where they should go next. They build up a history – some of it via promoters and enhancers and feedback loops, some of it via histone modifications – that builds their unique identity.

    When they finally settle down and start sending out axons, the same process continues: The axons meet signaling molecules along the way that tell them where to go next, and also tell them more about who they are. Some of them learn that they should just talk to their neighbours; some of them learn that they’re destined to connect widely separated parts of the brain, or send long axons down the spinal cord (and don’t forget to cross from left to right when you get to the crossing signal, if that’s your destiny!)

    Where do the signaling molecules come from? Well, that’s where it gets Rube Goldbergian. They, of course, were laid down by other cells, at just the right time, in just the right place – or they were laid down by this neuron itself, as part of its earlier travels. And this neuron, releasing its own signaling molecules at just the right time and place, will guide other neurons in turn.

    There’s some robustness in the system, so that minor failures can be worked around. However, some failures will have a cascading effect that causes multiple additional failures down the line. Having at least some horizontal lines in a kitten’s environment is a completely dependable source of signaling. It’s perfectly safe for the process that builds visual processing in the brain to incorporate it as a guidance signal. …until a demented researcher comes along and creates the only environment in the history of kitten-kind which is missing that signal. The result is a cascading failure.

    I’m familiar with the making of animated films, and there’s a loose analogy there. The making of every part of the film – every drawing, every model, every texture, every animation, every render – depends on all of the upstream steps being completed at the right time and put into the right place. Minor failures can be worked around, but if OMG-the-whole-surfacing-department-quit, the end result can be either an aspect of the film that’s well below the desired quality – the kitten can’t see horizontal lines and has wonky-looking rendered fur – or total failure of the project.

    Failure is only part of the story, of course – there’s also the possibility of alternate developmental pathways – but it’s illustrative: Each step in the process builds on the previous steps. We know that this is true on the level of the developmental history of individual neurons, and I suspect that it’s true on the level of the organism as a whole.

    • Andrew Klaassen says:

      If you find this interesting, the chapter on neuronal migration in Building Brains: An Introduction to Neural Development (by David Price, et al.) is an enjoyable read.

      In all cases, neuronal migration must be tightly regulated – the timing, direction, distance travelled and termination must all be controlled to ensure that appropriate numbers of cells reach the appropriate regions of the nervous system at appropriate times. A variety of strategies to regulate neuronal migration have been uncovered: some neurons migrate individually, others migrate in groups and some are guided on their journey by a scaffold provided by other cells.

  80. Andrew Klaassen says:

    Molecular Principles of Animal Development by Arias and Stewart has one of the best – if not the most clearly written – explanations of the “how do we get so much complexity from only 20,000 genes?” question. The 20,000 genes are not the program. They are merely the system calls. Most of the program lies in the regulatory regions surrounding the genes.

    “Although the repertoire of transcription factors is limited, the number of different combinatorial possibilities is immense…. Under different cellular conditions, different combinations of transcription factors may assemble at the same regulatory region, with a different regulatory effect on transcription. … The presence of a particular set of regulatory sequences is not the only important factor for enhancer function. Equally important is the context of these sequences, in other words, the nature and spatial relationship of the surrounding sequences.”

    The important word there is “combinatorial”. With a handful of regulatory regions around a gene, and a handful of regulatory proteins that can bind to them in cooperative or competitive interactions, you quickly get a large number of possibilities. (If you’re a computer programmer, think of how complicated things can get with a handful of “and” and “or” statements in an “if”.) Add in the fact that these regulatory proteins and regions are part of regulatory networks, and the number of combinations becomes immense. (Now we’ve got a multithreaded application in which a function is called or not depending on the combination of functions which called it, with rules of arbitrary complexity… and then that function in turn can contribute to calling other functions, including itself.)

    So that’s how we get the complexity. It doesn’t tell us how a specific instinct would be encoded, but it does tell us how something as complex as an instinct is encodable.

  81. Difference Maker says:

    Hehehe, if you were not asexual you would not be confused as to whether it is innate

  82. Difference Maker says:

    There seem to be a lot of asexuals here. What I find curious is how such a trait could have been transmitted, even accounting for random variation. Perhaps many are only missing the boob gene but still like girls

Leave a Reply