Unfortunately, because as Vox has coined the phrase MPAI (Most People Are Idiots) there needs to be a couple of disclaimers regarding this post.
- It is not a “takedown” of Vox or his ideas. It is what used to happen in the halcyon days of the internet open-source code equivalent. Vox has definite skills that I do not have and vice versa. His willingness to look into the minutiae of things that I find obvious at a higher resolution, and therefore not interesting for me, is far superior to mine. At least unless something genuinely grabs my attention, then I can be as obsessive and detailed as any weaponised autist, as my soon to be published book will show without a doubt in various instances (such as certain aspects of astronomy, or details of WWII). With regard to the details of evolution, and the minutiae of various philosopher’s ideas that I find absurd to begin with, he’s definitely in the lead though.
- His SSH is definitely useful, and valid. Of course, people do tend to make it both more and less than what it is. Human beings are complex and while the generic term of Gamma, or Sigma, or Delta, etc, is generically correct for a man overall, there are isolated or specific instances where that label might not be the most accurate, and this is something that Vox himself has stated on multiple occasions. To a certain extent, the average midwit takes the SSH and tries to make it into some ironclad box into which all men must fit, at all times, in all circumstances, and are reduced to that in toto. It is the Red Pill equivalent of “all women are whores” and so on.
- I wholly agree with his books Probability Zero and the Frozen Gene. Insofar as I have looked into them, they are correct. I am aware that he is also continually updating these works and as a general rule his theory on it all is reinforced more with each iteration.
In fact, the publication of his Probability Zero was instrumental in having me review and update from scratch a theory that was already obvious to me 31 years ago when I first published it, but is now an absolute certainty that can no longer be denied, so it’s not as if I have anything against Vox’s ideas concerning the SSH or Evolution in general.
My point here is one that may or may not be of interest to him, as I am also aware he is busy and may not get to this, or, alternatively, being as I am busy too, I may have missed some update that covers what I am critiquing below. If so, being made aware of it would be welcome.
Right, now to the critique. There is a bit of a TL;DR section at the end that starts with the heading The Conclusions of Epigentics
In his latest post on Sigma Game, he characterises Evolutionary Psychology as “fake science squared” and his SSH as proto-science.
I would tend to agree with the description of the SSH. After all, as readers here or of Caveman Theory are aware, I had developed my own version of the SSH long before I knew of Vox at all. So the observation of the traits he ascribes to different types of men is absolutely founded in reality and proper observation.
That said, while I don’t necessarily put a lot of stock in evolutionary psychology, as a model, it can and does predict certain things, if at a different scale than the SSH.
Furthermore, logically, you cannot pretend to have one set of measurements for the competing theory and excuse your own from it:
Regarding the Environment of Evolutionary Adaptation (EEA), he says:
“Verification of the causal mechanism: zero. Nobody has ever observed a Pleistocene male choosing a mate. Nobody has identified the neural substrate of the alleged module. Nobody has demonstrated the heritability of the preference. Nobody has shown the selection coefficient that would have driven its fixation. The entire causal chain from ancestral selection to modern preference is inferred, not observed, and the inference rests on assumptions about ancestral environments and reproductive payoffs that are themselves unobserved. What’s actually being measured is a correlation between a stimulus and a rating. The evolutionary scaffolding is decoration.”
But is the SSH put under similar testing? Short answer: No.
No one has identified the neural substrate of a Gamma compared to a Delta either.
No one has demonstrated the heritability of any part of the SSH either. Nor any other mechanism that would explain why someone becomes a Gamma instead of Lambda. There has been no empirical study of this any more than anyone has demonstrated heritability in EEA, and arguably quite a bit lees in fact.
As a model that predicts certain behaviours, the SSH works, and at a granularity that is higher resolution than EEA, but EEA also predicts things accurately probably at a similar rate as the SSH just at a different granularity. We could posit that the EEA model is mostly accurate at a macro scale and the SSH is mostly accurate at a micro scale. The REASON as to why the SSH or the EEA “work”, that is the details of the model , are basically opaque in both instances. The why , if you like, that Vox studiously avoids looking at. At least as far as his public posting on the issue goes, as he has repeatedly stated that the does not concern himself with the WHY, only the WHAT.
In which case, whether a model is real or not, becomes academic. If it works use it and if not, ignore it. If you look at the EEA as a large model predictor, it certainly appears to work.
I use Claude to check my theories with regard to sourcing research papers that are relevant to my hypothesis, then also ask Claude to find counter-examples, and I think overall the first part of the argument I am making is fairly solidly proven.
I asked Claude to find the papers related to EEA that not only are statistically significant, but that were risky predictions BEFORE the evidence was in, and predicted it correctly. These are the ones it pulled up.
Strongest cases
Kenrick & Keefe (1992), age preference asymmetry. The naive social-norms account predicts males prefer younger females at every age. Fertility-tracking predicts something odd: teenage males should prefer slightly older females, because peak fertility is above their own age. That’s what the personals-ad and marriage-record data show, across US, Dutch, Indian, and Filipino samples and across decades. The prediction is non-obvious, contradicts the rival theory’s direction, and holds.
Lieberman, Tooby & Cosmides (2003, 2007), kin detection. The model predicts that duration of childhood co-residence — not stated belief about genetic relatedness — is the operative cue for sexual aversion and for moral opposition to third-party incest. It won out over conscious belief in the data, with co-residence duration predicting the aversion in siblings who knew they were unrelated and failing to predict in siblings raised apart who knew they were related. Backed independently by Shepher’s kibbutz cohort data and Wolf’s Taiwanese sim-pua marriage records (elevated divorce, depressed fertility). [Kurgan Note: This study is in my opinion rather weak as it’s essentially quite theoretical in methodology, nevertheless, it has SOME validity]
Euler & Weitzel (1996), grandparental investment ordering. Paternity uncertainty predicts a specific rank order of solicitude: maternal grandmother > maternal grandfather ≈ paternal grandmother > paternal grandfather. Confirmed in German data and replicated in Greek, American, and South African samples. Four-way ordinal prediction — many ways to fail, it doesn’t.
Walter et al. (2020), 45 countries, N≈14,000, preregistered. This one matters because it was a genuine head-to-head. The sociostructural account predicts sex differences in mate preference should attenuate in more gender-egalitarian societies. They didn’t. Preregistered, large, cross-national — and the rival prediction failed. [Kurgan Note: this one seems pretty solid to me in methodology as well as results and is precisely the type of prediction at a larger granular scale that I expect of EEA when the model works]
Öhman & Mineka; Cook & Mineka . Lab-reared rhesus acquire snake fear by observational conditioning but do not acquire equivalent fear of flowers or rabbits from identically-constructed stimuli. Content-specific learning preparedness, demonstrated experimentally rather than inferred from ratings. [Kurgan Note: there are at least two relevant papers on this. The linked one is to three experiments, bit of a rabbit hole, but it produces results that are in keeping with my view of epigenetics, and I was not aware of this paper before toady.]
As stated, I also ask for counter-examples, which do exist, but nevertheless I don’t think invalidate my view that epigenetics plays some role.
Where it failed
- WHR itself. Singh’s 0.7 optimum replicated in Western samples but broke down where it counted. Yu & Shepard found Matsigenka preference for higher WHR; Marlowe’s Hadza data went the same direction. The cross-cultural universality claim — the part that requires the EEA — is the part that failed. [Kurgan note: This is the example Vox used in his post and it is indeed a fail, which frankly, I am not surprised by, I think the original postulation was rather speculative to begin with. However, a subsequent study seems to support the thesis depending on cultural norms (which again would go to my epigenetics theory) ]
- Ovulatory shift. Gildersleeve’s meta-analysis said yes, Wood et al. said no, and Jones et al. (2018), a large within-subject longitudinal design, found no cycle-contingent shift in masculinity preference. Effectively dead.
The caveat that bears on your Vox Day argument
Every confirmed case above is consistent with EEA reasoning but underdetermined by it — none of them observes the ancestral selection episode, which is exactly his complaint, and the complaint is not answered by any of this. What the good cases show is something narrower and still useful: that EEA reasoning generated non-obvious predictions which then beat their rivals in the data. That’s a real epistemic achievement and a different one from demonstrating the causal history.
Which cuts back at him. If the standard is “made a risky prediction, could have failed, didn’t, and beat a competitor theory head-to-head” — Kenrick & Keefe and Lieberman et al. clear it. The SSH has no equivalent: no preregistered head-to-head, no rival framework it has outperformed on a shared dataset, no case where its prediction and a competitor’s diverged and it won. He’s comparing his best-case reading of SSH against evopsych’s worst-case output.
That last paragraph is in essence my view too, however, I also grant that Vox has a valid point even so, but one that I believe diminishes considerably once you bring epigenetics into it.
In Layman’s Terms:
Here is a simple definition of epigenetics:
The literal sense first: epi- means “on top of” or “in addition to.” Epigenetics is what sits on top of the DNA sequence — changes in whether and how strongly a gene gets used, without any change to the sequence itself.
The problem it solves is this. Every cell in your body carries the same genome. A liver cell and a neuron have identical DNA and behave nothing alike. So the sequence alone can’t be what makes them different. What differs is which genes are switched on, how strongly, and when. Epigenetics is the machinery that does that switching — and, crucially, that keeps a cell remembering what it is after it divides.
The methodology it works by:
The main mechanisms
DNA methylation. A methyl group — one carbon, three hydrogens — gets chemically attached to a cytosine base, usually where a C sits next to a G. When these accumulate in the control region at the front of a gene, that gene is typically silenced. Enzymes add the tags; other enzymes strip them off. This is the mechanism people usually mean and the one that’s easiest to measure.
Histone modification. DNA isn’t loose in the cell; it’s wound around spool-like proteins called histones. Chemical tags on the protruding tails of those spools change how tightly the DNA is wound. Loosely packed DNA is accessible to the machinery that reads genes; tightly packed DNA isn’t. Acetyl tags generally loosen, and removing them generally tightens.
Chromatin architecture and non-coding RNA round it out, and whether the RNA side properly belongs under the heading is one of the things specialists argue about.
Now, in principle, my argument is that given that epigenetics has demonstrably been proven to affect the next generation even when the skill or event happened for the first time in the parents, and happens without the Darwinian aspect of selection, I think it plays a significant role that sits on top of, and separate from, the proposed (erroneous) Darwinian evolution aspects of “evolution”.
To try to make a couple of (imperfect but useful) analogies to explain my view of things:
If you assume DNA is a page of text, Darwin says that when you combine two pages to produce a third page, over time, the resulting third page that reads best survives to produce increasingly good literature. Vox essentially proved this is nonsense and that the random mixing of those two pages tends to produce increasingly less fit literature, and only extremely strong survival (legibility) pressures tend to produce better third pages, and even then, the time required is far longer than the universe has been around. He is right in that, and has essentially dragged Darwinian Evolution out into the open street and executed it with a sawn off shotgun.
Epigenetics however is akin to someone highlighting the better parts of each page, and those highlighted parts have a stronger tendency to survive into the third page created by the blending of the two parent pages. So the highlight does not change the baseline text (DNA) but it highlights/activates certain aspects of it. Those aspects are generally positive given the environment in which the parents exist, otherwise such a strong relevance of the skill would not be done by the parents to the degree it activates or deactivates various parts of their DNA. And this advantage is passed on (to some degree at least) to their offspring. But… just as it can be learned in one generation, (and then added to and built upon by succeeding generations) it can also be lost in one generation. This tends to support the idea that families who continue a specific trade or skill over generations become increasingly good at it and learn and are better at that skill than others in general terms, all without darkening selection existing at all.
The empirical studies on epigenetics are:
Meaney’s rat work : variation in maternal licking-and-grooming produces differential methylation of the glucocorticoid receptor promoter in pups, which produces lifelong differences in stress reactivity and dominance behavior. McGowan et al. (2009) found the homologous NR3C1 methylation difference in human post-mortem hippocampus among suicide victims with childhood abuse histories — smaller and noisier, but real.
There are others, which you can source if you re interested.
So, epigenetics is a mechanism that produces exactly what SSH describes. Recognizable, stable, cross-cultural adult male behavioral patterning, arising without any evolutionary process at all. Vox’s descriptive taxonomy might be substantially accurate and still have nothing to do with types.
The same literature concerning epigenetics demolishes the concept of fixity, which then would also apply to the SSH if it is related to it.
The Meaney phenotype is reversible. Cross-fostering, methionine infusion, HDAC inhibitors all shift it. Rank-and-neurochemistry work (vervets, and crudely in crayfish) shows the arrow running from social position to physiology, not only the reverse.
If SSH ranks are epigenetic-developmental phenotypes, which I think they may very well be, then a man’s “rank” is substantially the residue of his traumatic childhood, status history, and includes that of his father in all likelihood, and is at least potentially revisable when the history changes through a concerted and persistent effort. That converts the framework from a typology into what I would classify as a (very strong in most cases) state description, and a state description cannot underwrite the predictive claims he’s resting the whole case on to an absolute degree.
In fairness, as I said at the start, Vox does not claim the SSH is as rigid as some of his fans/exponents make out, but he does tend to assume the rank is generally irreversible. And I agree that it generally is very strongly exhibited, but I do not agree that it is written in stone.
There is a further valid point that Claude highlights,
Where the predictions actually fail
An epigenetic reading generates risky predictions SSH doesn’t make and hasn’t tested: rank distribution should track early-life status contingencies — paternal absence, birth order, early nutritional stress, adolescent status shocks — and should shift measurably after sustained status change. It should show dose-response. It should show partial reversal under intervention. Test any of that and you have science. “Anyone can run the test on people they know” is not that test: unblinded, unpreregistered, classification made post hoc, categories elastic enough that a Gamma who succeeds gets reclassified as a secret Sigma. That’s the structure that made Galenic humors and Kretschmer’s somatotypes look predictive for centuries. His criterion — everyday confirmation by the observer who already holds the theory — is the specific procedure that manufactures confirmation, which is why blinding exists.
Now, there is a weakness in the epigenetic transgenerational theory I am proposing, and that is that it too lacks the kind of solid evidence required in a scientific context to make it watertight. Nevertheless, epigenetic provides at least half of the answer I believe does produce the results I have observed over the last 40 years or so of noticing various things.
The things that CAN be measured by epigenetics are not necessarily relevant to the transference of skills, even if we DO observe the transference of skills. The fact a fencer or a rower tends to produce better fencers or rowers in their family line of fencers or rowers is an observed fact, but other factors can account for it, like skill learning process starting at a young age, including in subtle ways that turn up even before the actual training takes place. So it can’t be attributed solely to epigenetics and once again we would have a working model with a rather unfounded model, which however, is still stronger than the current SSH model, which, as far as I know, is pretty much absent an explanation of why and how it works as it does.
But there is a further aspect I think plays a part.
The Conclusions of Epigentics
What I am saying in more basic language is in essence: the man you meet at forty is largely the settings that got installed in him as a child.
Not the genes he inherited or a type he instantiates (the mechanisms of which would in any case be absolutely mysterious). So he is a Gamma or a Sigma or whatever, mostly by the configuration his early environment wrote into him.
The rat experiment
Some mother rats lick and groom their pups a lot. Some don’t. The pups of the attentive mothers grow into calm adults — low stress hormones, quick recovery, confident exploration. The pups of the inattentive mothers grow into jumpy, reactive adults. This lasts their whole lives.
The obvious objection is that anxious mothers pass on anxious genes. So the experiment: swap the pups at birth. Give the biological pups of a neglectful mother to an attentive one.
They grow up calm. Swap the other way, they grow up jumpy. The genes went one way, the temperament went the other. The mothering did it, not the inheritance.
Then the mechanism. In the calm rats, a particular gene involved in shutting down the stress response is switched more fully on — chemically tagged differently in the brain. And this isn’t just correlation: give a drug that alters those chemical tags and the behaviour changes with them. You can push the phenotype in either direction. So the causal chain is visible end to end — mothering, chemical marking, gene activity, adult behaviour.
So now you have a behavioural pattern that:
- appears reliably and recognisably in adults
- is stable across the whole lifespan
- runs in families, generation after generation
- looks for all the world like an inherited type
But none of it is a type, and none of it is inherited. Attentive mothers raise daughters who become attentive mothers, so the pattern flows down the line looking hereditary while being nothing of the kind.
Everything Vox’s socio-sexual hierarchy observes —recognisable patterns, stable across a man’s life, showing up in every culture, running in families— is exactly what developmental programming produces. His descriptions could be largely accurate and still have nothing to do with types.
The same experiments that explain the stability also demonstrate the reversibility. Swapping the pups works. The drug works. Even in adult animals, not just pups, the intervention shifts the settings back.
So if a man’s rank is a developmental phenotype, it is a record of what happened to him , not a fact about what he is. And a record can be revised when the circumstances change.
That is the difference between a state and a type. A type is what you are; a state is where you currently sit. The SSH sells prediction: identify the category, forecast the man across situations and across years. Only a type supports that. A state doesn’t. Or rather, it can, and it can appear to, and yes I do think changing your state is extremely hard, but absolutely NOT impossible. And my personal experience both in myself and others I think is valid, given I have trained literally hundreds of men (mostly) and some women in at least martial arts, and also have rather extensive work done in hypnotherapy and an understanding of the human mind that in practical terms is superior to a lot of people that work in tat field professionally. I know that sounds “arrogant” but my long experience in this field is a reality, regardless of any false modesty one might try to have (pointlessly).
So, the final point is that with regard to Vox’s SSH, the mechanism that would give his framework a real scientific footing is the same mechanism that says his central practical claim is wrong. He needs the biology to make the taxonomy respectable, and the biology dissolves the taxonomy into a snapshot.
The observations and predictions the SSH makes are real and valid, but pretending they are fixed on some unstated biological or other basis is wholly unsupported.
The SSH is largely a theory of permanent kinds without a supporting underlying structure that can wholly be explained by current states (hard as these may be to change).
And there is a further issue I think plays a part:
Morphic Resonance
The work of Rupert Sheldrake on this has been known to me for over 30 years, at a time when his ideas where considered heretical in the absolute, and are today vindicated by multiple replicated experiments. Explaining the whole thing would make this already long post too long, but I suggest you read up on it yourself if interested.
So… in my opinion, with regard to the SSH, both epigenetics and morphic resonance play a part, and in my life experience I also believe that the so-called SSH rank of a man CAN change, even quite drastically, but usually only due to very extreme conditions. This is not a new theory, I have stated many years ago already in writing (see the OG blog) that only three things can generally help a man change drastically:
- Continued relentless effort over a period of decades, usually at least 20 years.
- A well-controlled LSD trip with positive and well done guidance (recidivist in criminality dropped hugely after a single trip before these experiments were banned)
- Repeated deep hypnosis paired with an honest intent to change.
Options 2 and 3 are faster but option 2 is probably not available without considerable risk. So option 3 is the fastest paired with a good deal of option 1 too.
Anyway, that is my critique and I hope Vox eventually has the time to respond if he sees fit to do so.
The intent, as always, being to improve the overall usefulness of a theory, not critiquing the individual who in any case created something useful even if it does not become any more defined or predictive than it already is.
This post was originally published on my Substack. Link here
