[Epistemic status: So, so speculative. Don’t take any of this seriously until it’s replicated and endorsed by other people.]


If you’ve ever wanted to see a glitch in the Matrix, watch this spinning mask:

Source: http://hearingthevoice.org/2013/11/14/predictive-coding-masterclass/

Did you see it? As the face started to turn away from you, your brain did…something, and then you were seeing a normal frontwards-facing mask again. It turns out your visual system has really strong views about whether faces should be inside-out or not, and it’s willing to execute a hard override on perception if it doesn’t like what it sees.

But not always. Some people get glitchier glitches than others; a few seem almost immune. Studies find schizophrenics and autistic people to be consistently less glitchy than the rest of us. The correlation’s not perfect. But it’s definitely there. Something about these people’s different cognitive processing styles lets them see through the illusion.

I wanted to replicate this result myself. So a few months ago, when I surveyed readers of my blog, I included some questions about perceptual illusions (including a static version of the hollow mask). I got five thousand responses, including a few from schizophrenic and autistic readers. Sure enough, the effect was there.

Schizophrenic readers were about twice as likely to report a weak reaction to the mask illusion as non-schizophrenics (28% vs. 14%, p = 0.04). They were also more likely to have a weak reaction to a similar illusion, the Spinning Dancer (58% vs. 81%, p = 0.01). Readers with a family history of schizophrenia landed in between schizophrenics and healthy controls (16% for mask, 63% for dancer, ns).

Autistic readers were only slightly more likely to report a weak reaction to the mask illusion than neurotypicals (17% vs. 14%), but thanks to our big sample size we could be pretty confident that this was a meaningful difference (p = 0.004). There was no different between autists and neurotypicals on the Spinning Dancer, not even a weak trend (58% vs. 60%, p = 0.4).

Looking deeper, I found a few other anomalies on illusion perception. Most were small and inconsistent. But one stood out: transgender people had an altered response pattern on both illusions, stronger than the alteration for autism and almost as strong as the one for schizophrenia (mask: cis 14% vs. trans 21%, p = 0.003; dancer: cis 58% vs. trans 71%, p = 0.001). These results are very tentative, and need replication. My mass survey isn’t a very sensitive instrument, and I place low confidence in any of this until other people can confirm.

But for now, it sure looks like a signal. Something seems off about transgender people’s perception, something deep enough to alter the lowest-level components of visual processing. If it’s real, what could it be?


A few days ago, trans blogger Zinnia Jones asked me if there might be any neurochemical reason trans people dissociate so much.

Dissociation is a vague psychiatric symptom where you feel like you’re not real, or the world isn’t real, or you’re detached from the world, or something like that. It sounds weird, but if you explain it to someone who’s had it, they’ll say “Oh yeah, that thing!” It’s usually unpleasant, and tends to occur in PTSD, borderline personality, and extreme stress.

And in transgender people. The only formal study I can find on this describes it as “greatly prevalent”, and suggests that up to 30% of trans people may have dissociative conditions (compared to less than 1% of the general population). This matches trans people’s self-reports (1, 2, 3, 4, 5). Anecdotally (according to Zinnia’s impression of the trans community) and formally (see Costa & Colizzi 2016) hormone replacement therapy is an effective treatment for dissociative problems.

Intuitively this makes sense. Trans people feel like they’re “trapped in the wrong body”, so of course they feel detached from their bodies / like their bodies aren’t real / like their bodies aren’t theirs. Hormone therapy helps solve the “wrong body” problem, so it also solves the dissociative symptoms.

We aim to bridge psychosocial and biological levels of explanation. We can say that someone is stressed out because their boss overworks them, but also because they’re secreting high levels of cortisol. We can say that someone is depressed because they broke up with their boyfriend, but also because they have decreased synaptogenesis in their hippocampus. Causation gets tricky, and this is a philosophical minefield for sure, but overall these two levels should be complementary rather than competitive. So what’s the biological correlate to trans people having dissociation problems?

Practically all searches for the biological basis of dissociation end up at the NMDA glutamate receptor, one of the many neurotransmitter systems in the brain. Even though its cousins dopamine and serotonin usually get top billing, glutamate is probably the brain’s most important neurotransmitter, and NMDA glutamate receptors in particular are involved in all sorts of interesting things.

Drugs that block NMDA receptors cause dissociation. The most famous dissociative anaesthetic, ketamine, is an NMDA antagonist. So is DXM, a recreational drug that causes dissociation in abusers. Wikipedia’s list of dissociative drugs is basically just fifty-five NMDA antagonists in a row. The only other category they list are kappa opioid agonists, and kappa opioid agonism probably – you guessed it – antagonize NMDA. If we take this result seriously, every substance we know of that causes dissociation is an NMDA antagonist in some way.

Does anything improve NMDA function – an effect we might expect to alleviate dissociation? Yes, and among a list of intimidating research chemicals called things like “aminocyclopropanecarboxylic acid” is one familiar name: estrogen. See eg El-Bakri et al, which finds that “estrogen modulates NMDA receptors function in the brain…enhancing NMDA function”. McEwen et al: “One of the long-term effects of estradiol [estrogen] is to induce NMDA receptor binding sites in the CA1 region of the hippocampus.” Bi et al: “17-B-estradiol [estrogen] enhances NMDA receptor phosphorylation and function.” I don’t fully understand this research, but it seems to point to estrogen promoting NMDA activity in some way.

So transgender people dissociate a lot, a state usually associated with hypofunctioning NMDA receptors. And trans women get better when they take estrogen, a hormone that improves NMDA function. That’s interesting. But what does this have to do with those optical illusions?


The Hollow Mask illusion and its cousins may depend on NMDA function.

To oversimplify: the brain interprets the world through Bayesian calculations. In Corlett et al’s model, it communicates top-down priors (ie assumptions based on previous knowledge about the world) through NMDA receptors and bottom-up new evidence through AMPA receptors. They write:

In a hierarchical cortical system in which representations become more abstract with increasing distance from the primary input, higher levels of the hierarchy specify top-down predictions through NMDA receptor signaling and any mismatches between expectancy and experience are conveyed upward through the hierarchy via rapid AMPA and GABA signaling

When you see a hollow mask, the brute facts of how the mask looks are your bottom-up sensory evidence. Your top-down prior is that every other face you’ve seen for your entire life has been normal, not inside-out. Given the strength of the prior, the prior wins, and your brain interprets the mask as a normal face.

Unless your brain is bad at applying priors, ie its NMDA receptors aren’t working that well. Then it just sticks with the bottom-up sensory evidence showing that the mask is hollow.

Schizophrenia and autism both probably involve decreased NMDA function in different ways. For schizophrenia, see eg Olney, NMDA receptor hypofunction model of schizophrenia, and Coyle, NMDA receptor and schizophrenia: a brief history. Ketamine seems to replicate the symptoms of schizophrenia pretty well and is commonly used as a model for the disorder. For autism, see eg Lee, NMDA receptor dysfunction in autism spectrum disorders and this study where screwing with NMDA receptors in mice seems to turn them autistic.

From this we would predict that estrogen would help treat schizophrenia and autism. It does. Schizophrenia is more common and more severe in men than women, with researchers noting that “gonadal steroids may play a role in buffering females against the development of schizophrenia”. Women are known to sometimes get schizophrenia triggered by menopause when their estrogen levels decrease. Estrogen supplementation is an effective schizophrenia treatment, and there’s some interest in developing estrogen receptor modulators that can help schizophrenic men without making them grow breasts. Meanwhile, autism continues to be about four times more common in men than women, autistic women tend to have more “male-typical brains”, and although it’s considered unethical to treat autistic boys with estrogen, it works in mice and fish. Once again, doctors are looking into estrogen analogues that don’t turn people female as possible autism treatments.

We might also predict that estrogen would increase glitching on the hollow mask. I can’t study this directly, but on the survey, 15% of biological males had weak reactions to the illusion, compared with only 11% of biological females, p = 0.01. Since women have more estrogen, that looks good for the theory.

Transgender people have higher rates of autism and schizophrenia. The Atlantic actually had a good article about this recently : The Link Between Autism And Trans Identity. They cite one study showing 8% autism rate in trans people (compared to 1-2% in the general population), and another showing that autistic people were 7.5x more likely to express “gender variance”. Apparently a lot of trans people have problems getting hormone therapy because their doctors think the gender issues are “just” because of their autism. Some might say that denying people estrogen because they have a condition which studies suggest estrogen can successfully treat is a bit, I don’t know, crazy and evil, but I guess people get really weird around this stuff.

My survey broadly confirms these numbers. Autism rates were sky-high in every category – it’s almost as if the sorts of people who like reading blogs about how gender is all just NMDA receptors skew more autistic than average – but there was a remarkable difference across gender identities. 15% of cisgender people were autistic, but a full 52% of trans people were.

The survey also finds that about 4% of non-schizophrenic people were transgender, compared to 21% of schizophrenics and self-suspected schizophrenics. Other people have noticed the same connection, and I’ve met more schizophrenic transgender people than I would expect by chance given the very low rates of both conditions.

If this is right, we end up with this rich set of connections between schizophrenics, autistics, ketamine, dissociative experiences, estrogen, gender identity, and the hollow mask. Anything that decreases NMDA function – schizophrenia, autism, ketamine – will potentially cause dissociative experiences and decreased glitching on the mask illusion. Estrogen will improve NMDA function, treat dissociative experiences, and bring back hollow-mask glitching.

So I wonder: is NMDA hypofunction related to transgender? That would explain the autism and schizophrenia connections. It would explain the hollow mask numbers. It would explain the dissociation. It would explain why estrogen helps the dissociation. And it would explain a lot of internal connections between all of these different conditions and factors.


I’m going to stop here, even though there’s a lot more worth saying on this, because I’ve already gotten so far into Speculation Land that trying to chain any more conclusions on would probably be premature. So let’s switch to some reasons for skepticism.

First, the research into NMDA receptors is too interesting. People argue that NMDA is key to depression, key to anxiety, OCD, chronic pain, and borderline personality (my guess is the depression claims are mostly overblown, the borderline claims are 100% absolutely right and revelatory, and I’m agnostic on the others). On the one hand, explaining everything sounds sort of good. On the other hand, it also sounds like what would happen if a field was getting kind of overhyped and slipping into methodology loose enough to prove anything it wanted. Maybe a vague link between a receptor which is literally everywhere in the brain and some psychiatric disease isn’t that interesting. A theory that can explain absolutely everything should always cause suspicion.

Second, I’m still not sure what to make of the Hollow Mask results on my survey. Although the transgender results were unusually strong, I did get mildly statistically significant results on about half of the thirty-or-so things I looked at, including seemingly-unrelated items like political affiliation. Some might argue that this means something is wrong with my survey. Others might argue that hey, we know political attitudes are about 50% genetic, and the last time people tried trace the genes involved all the strongest results were genes for NMDA receptors. Have I mentioned that NMDA receptors are really interesting?

Third, I included a second illusion I asked about on the survey, the Spinning Dancer. It also had an odd response pattern among transgender people. But it didn’t correlate at all with the Hollow Mask Illusion, and it doesn’t seem to be elevated among autists. I don’t know what’s going on here, and the whole thing makes me more suspicious that all of this is some weird artifact.

Fourth, all this predicts that ketamine will cause reduced glitching on the Hollow Mask. It doesn’t. Corlett argues that this is because chronic, but not acute, NMDA dysfunction is required to stop the hollow mask glitches “because [keatmine] has a predominant impact on bottom-up AMPA signaling”. I don’t really understand this and it seems like a prediction failure to me. On the other hand, chronic marijuana use does prevent mask glitching, which might be because of marijuana causing NMDA hypofunction over time, which I guess is a point in favor of the chronicity theory.

Fifth, although trans women dissociate less when they take estrogen, trans men dissociate less when they take testosterone. I can’t find whether testosterone has similar NMDA-promoting properties in the brain, although it sometimes gets aromatized to estrogen so that might be relevant. Also, I’ve never heard of any trans woman taking testosterone or trans man taking estrogen. If that makes dissociation worse – and from the psychosocial perspective it probably should – then that would be a strike against this theory.

Sixth, although I played up the transgender/autism and transgender/schizophrenia links, the truth is that transgender people have higher rates of every mental illness, to the point where it may just be some general factor. I think I’m justified in focusing on these two results because transgender people’s higher rates of depression and anxiety are probably just related to being transgender being depressing and anxiety-provoking in this society. But schizophrenia and autism are 80+% genetic, and so harder to explain away like that. Still, somebody could question the relevance of worrying about these two conditions in particular.

I hope some of this can be sorted out in the near future. A first step would be for someone official to replicate the transgender Hollow Mask pattern and prove that it’s not just confounded by autism and schizophrenia rates in that population. A very tentative second step would be to investigate whether chronic use of the supplements that improve NMDA function in schizophrenia – like glycine, d-serine, and especially sarcosine – can augment estrogen in improving gender dysphoria. Remember to consult your doctor before trying any weird supplements since they may cause unintended side effects, like becoming a Republican.

It could also be worth trying to understand more explicitly why gender identity and NMDA should be linked. This post is long enough already, but I might write more on this in the future. If you want a preview, check out The Role Of Neonatal NMDA Receptor Activation In Defeminization And Masculinization Of Sex Behavior In The Rat and draw the obvious conclusions.

New Comment
4 comments, sorted by Click to highlight new comments since:

As a 23-year sufferer of DP I was hoping to get in contact with you. I love this article! I've never seen another one like it. I don't have gender problems, but I've come to realize that I do have Aspberger's tendencies.

I am a nutritionist who began taking l-serine 6 months ago and have received about 25 health benefits from it! Love this stuff! I try to take l-glycine, but it messes with my breathing by clogging my sinuses. I still have DP, but know that I'm on the right track...along with following a ketogenic diet. Taking Benfothiamine, along with other quality B's has also been essential in helping my brain heal.

I just have questions for you, but it's hard for me to put them all down here. It would be easier to have a conversation as it would probably stimulate even more questions. The DP community is in dire need of this information! I try to tell them, but they're usually so busy freaking out that they don't listen well.


This is not Scott, but I'm curious what DP means. Depersonalization?

Yes. Sorry...should have spelled it out.

It seems to me that the effects you find for the categories you examine are likely not strongly related. In particular, I am utterly unsurprised that Autistic persons showed an effect on the spinning face but not the , given that facial recognition is a critical social skill that Autistic people tend to struggle with:


It seems more likely that this is the cause of the effect you find than any problems with a single neurotransmitter, which to me sound suspiciously like the old idea that lack of serotonin was the only cause of depression.