Some notes on desistance research

Epistemic status: To make sure I got this post right, I asked Michael Bailey for comments. I edited my initial draft based on his feedback, and he agrees completely with this version.

Also, I guess this post means that the followup that I promised for my last post will have to wait a bit.

“Desistance”, in the context of research by Zucker et al, refers to a very specific phenomenon. The topic is still an open research area, so it is difficult to say anything for sure. However, desistance research specifically applies to early-onset gender dysphoria, which is strongly associated with gender nonconformity. The association with gender-nonconformity makes it easy to assume that highly GNC kids will eventually turn out transgender, but in practice they are usually observed to outgrow this and end up happy with their sex. This specifically is the phenomenon that is covered in desistance research. Not much is known about why they are so strongly correlated, but some researchers believe that the gender-nonconformity can be a reflection of the dysphoria, and that encouraging it may cause the dysphoria to persist. This specific pattern of desistance is observed only for early-onset gender dysphoria, and it likely relies on specific characteristics of the dysphoria that are not seen in other forms. As a result, there is little a priori reason to expect it to generalize to other forms of gender dysphoria, unless these forms of gender dysphoria also have specific characteristics that would imply a high likelihood of desistance.

In particular, desistance research does not apply to autogynephilic gender dysphoria. Not much is known about the specifics that distinguishes dysphoric autogynephiles from non-dysphoric ones, but equivalent phenomena to desistance have not been consistently observed for them (and is unlikely to ever be observed, even if it does exist, as properly identifying and evaluating autogynephiles in the relevant period is very difficult). Autogynephilic gender dysphoria has, however, been observed to disappear in some cases, but this is not a consistent and reliable phenomenon in the same way that desistance from gender nonconformity seems to be.

The research also does not apply directly to “rapid-onset gender dysphoria”. There is currently no research on the desistance rates of ROGD. However, if the model that rapid-onset gender dysphoria is to a large degree caused by social contagion is true, then this creates a large possibility for desistance under the right circumstances, as the gender dysphoria is not purely due to permanent characteristics of the person in question, but instead also due to mutable social environment. If the social contagion model is incorrect, there is very little reason to expect desistance to happen. There is almost no research on the validity of the social contagion model for rapid-onset gender dysphoria, but there are many anecdotes that make it seem somewhat plausible.

The specific ways that desistance might work in early-onset gender dysphorics also means that if children who *could* desist instead go on to transition, they will likely not experience regret from transition, but will instead simply not desist. This has a number of implications. It means that the harm of transition will be limited to the problems associated with being a transgender person rather than a gay person, rather than implying regret and future detransition. However, it also means that if someone who could’ve desisted ends up transitioning, it becomes easy for people to mistakenly think that the transition was a huge success because the patient is very happy with the outcomes, when actually the patient could have been just as happy if they didn’t transition. And lastly, it also implies that interventions which do not permit unlearning gender-nonconformity and atypical gender identity – for example ones where gender-nonconformity is highly encouraged – might very well prevent desistance.


Rates of autoandrophilia among FtMs

Let’s talk about the rates of autoandrophilia I’ve observed among FtMs. I’ve had a few different approaches for assessing this, so it’s probably worth going through things in detail. Arguably, the most direct was asking about it directly:


Rates of transmasc autoandrophilia in the Transmasculine Sexuality Survey.

For counting this properly, I exclude the third question, since it could conceivably by interpreted as assessing gynephilia, rather than autoandrophilia. Doing this, the total autoandrophilia rate in the survey was 74% (n=31). However, one problem with this survey is that I shared it to a nonbinary subreddit, and so it is possible that some NBs responded too, which might make the data hard to interpret. If I exclude those who didn’t pick “male” as one of their gender responses (I allowed multiple, including “nonbinary”, “genderfluid” and “questioning”), I get a rate of 71% (n=14), essentially the same.

It may be informative to also consider the frequency of autoandrophilic fantasies from this group. I won’t include this when counting the rates, though.


Transmasc pre-identification AAP fantasy rates according to Transmasculine Sexuality Survey.

Another survey with a very straightforward methodology was in my Survey of Traits You Are Attracted to or Would Like to Have. Unlike the previous survey, this was targeted at reddit in general, not just transmascs, and so it may have a better accuracy distribution-wise. In this survey, I also asked rather directly whether they found being male hot:


Autoandrophilia in transmasc participants of Survey of Traits You’re Attracted to or Would Like to Have.

Again, the rate of affirmative answers was 74% (n=27). If we exclude the nonbinary participants, the rate increases to 89% (n=9). One nice thing about this survey is that I also asked about “sexual fantasies about being the opposite sex”, so this gives us a way of assessing whether these yield the same sorts of results. The problem is that some trans men will interpret “sexual fantasies about being the opposite sex” as referring to sexual fantasies about being female, so it’s important to figure out how big this problem is.

For the record, the equivalent MtF numbers are 95% (n=19) and 90% (n=10).

In this survey, arousal to fantasies about being the opposite sex wasn’t very strongly associated with autoandrophilia (r~0.28). In fact, it was more closely associated with the autogynephilia measure (r~0.39), despite the fact that the concept of autogynephilic transmascs is… weird. I found that 70% of the ones who got positive scores on the AAP scale also got positive scores on the scale assessing A*P by asking about “opposite sex”, but 43% of those who got zero scores on the AAP scale still got positive scores on the A*P scale. Not great.

Another survey with a slightly less direct approach was my Can you look at some porn For Science? survey, which was about exactly what it says in the title. Before presenting the erotic material, I asked a number of questions about paraphilias, including “Picturing myself as the opposite sex, or with certain physical features of the opposite sex” and “Picturing an attractive man and imagining *being* him”. The latter seemed to work for assessing autoandrophilia in women; assuming the former question assesses AAP perfectly (dubious), the latter had r~0.72 correlation with AAP, 15% false positive rate, and 21% false negative rate.


Amount of autoandrophilia in the transmascs in the porn survey.

68% of transmascs (n=22) and 75% of FtMs (n=8) reported AAP according to this measure. This survey also allows us to revisit the approach of asking about “opposite sex”. It had a correlation of r~0.5, and the rate of positive answers to this among those who had given positive answers to the other AAP question was 80%. Meanwhile, the rate of negative answers to this among those who had given negative answers to the other AAP question was 14%. This is a far higher validity than was observed in the other survey. On the other hand, the equivalent numbers when comparing with the AGP question are 0.37, 91% and 29%, which is still a pretty strong connection, so asking about “opposite sex” probably isn’t great.

The MtF rates of autogynephilia here were 76% (n=33) and 82% (n=23).

So far, this suggests an AAP rate of 68% to 74% (weighted average: 72%, n=80) for transmascs in general and 71% to 89% (weighted average: 77%, n=31) for FtMs. Of course, these numbers should be taken with a grain of salt; for example, there may be some misreporting, where some AAPs report not experiencing AAP. In addition, there is a troubling problem that needs to be taken into account, and which I will address in my next post.

Gender information collaboration project

Right now, a lot of politics about issues relating to transgender and gender nonconforming people are intense and not very forgiving. There’s very little trust, and understandably so, as a lot of people have been hurt in many ways by the current situation.

I don’t think there’s any chance of creating this trust on a big level any time soon. However, this leads to a problem that lots of highly important information will be inaccessible to all the parties who need it. This information might involve the different ways gender issues evolve and present, and how different people cope or treat them. It might involve all sorts of other things of relevance to people. The important problem right now is that this is dispersed across different groups who often don’t trust each other, and who are only interested in sharing it with people they feel are certain to take care of their concerns properly.

The goal here should be to create a win/win situation, where all parties feel that they are treated fairly and benefit from participating. Nothing else can work. For this reason, I propose creating a group of people with diverse, relevant perspectives, and allowing vetos from all the perspectives involved. The group would have to come to a consensus about what information to collect, and how to collect it, before it could even start asking questions. Before making any official opinions about how to interpret the collected information, the group would have to reach consensus on what the interpretation should be. This way, anything done by the group is guaranteed to be a win for all sides, as any problematic suggestions can be vetoed by people who’re concerned.

Example project

For all of this to make sense, it might be helpful to look at an example project that could be worked on. One that I am personally very interested in is documenting the experiences of detransitioners and how they compare to trans people.

If done right, this could help everybody. Detransitioners generally want their experiences to be heard. Trans people would want a fair comparison and contrast of their experiences and those of detransitioners, rather than FUD. If things work out perfectly (and they might not), studying this could also help questioning people decide whether transition is for them.

The problem here is that there are conflicts of interest in mistrust. Detransitioners don’t want their experiences misrepresented, and might reasonably fear that the only goal would be to dismiss their experiences as irrelevant (or worse). Many detransitioners may have strong reservations about transgender issues, and believe in political ideas that trans people find dangerous. This leads to issues about whether trans people feel misrepresented, or feel that the wrong things are getting examined.

All of these issues are solved by having people on each side collaborate. The varied perspectives will allow people who have concerns to make sure that these concerns are taken properly into account, and to ensure that things are only done if everyone benefits. This will allow us to reap the benefits of cooperation despite low amounts of trust.

This is not the only project that could be done. Other possibilities would be to better document what is happening around gender nonconforming people, to process the science and controversy of transness into more-digestible summaries with whatever consensus we can find, and likely a number of other things that I haven’t even thought of.

Proposed structure

Achieving total consensus about everything will only be practical if the groups are sufficiently small, which would lead to problems due to lack of perspectives. An alternate approach would be to create a number of subgroups, each of which is responsible for taking care of one set of concerns. For example, trans people may have one broad subgroup, whose responsibilities would be to ensure that the total group avoids hurting trans people. I expect that it will also be useful to work with radical feminists who have, ehm, strong concerns about trans issues, so they would also have a subgroup, whose job it would be to take care of their concerns. If other groups turn out to be relevant, these groups would also be able to get a subgroup of their own.

We could set up a number of structures, such as a blog, a chatroom or forum for discussion, and maybe also a wiki or something. Depends on what becomes relevant, but at the very least we need some summary of what the groups does, who the group is, and why the group should be trustworthy.

The group might gather and publish raw data. If we can achieve consensus, we might publish some articles on how to interpret our data, various scientific findings, or other things that may be of interest.

The goal would be to try to improve the group quality over time. Initially, it’s likely going to consist of random people on the internet who find the project appealing, but hopefully we can work to establish some respect for the group and start collaborating with people who are more trusted in various ways.


In reality, I’m doing this for rather selfish reasons. I want to examine various things, but I have views that make people not trust me very much. For example, my views on trans policy are aggressively pro-transition to the point where certain groups of detransitioners whose experiences are highly relevant knowledge do not feel comfortable with my beliefs, while my views on the causes of transness (something similar to Blanchard’s typology) lead to many trans people not trusting me. Of course, I would claim that my policies are based on what’s likely to give good results, and that my beliefs are based on the best evidence available, but I don’t think I can convince enough people who disagree.

So, rather than try to force the information out of people, why not collaborate? Hopefully this can be a more productive approach than what I’ve tried until now.

What’s Up With Bi Men?

For some time I’ve noticed that bisexual men in my surveys tend to score higher on measures of gender issues than other men. Initially, the explanation for this was easy: studies have had trouble finding truly bisexual men (see e.g. here or here), so there’s probably some third factor that makes them think they are bi without truly being bi. One possible factor is autogynephilia, and the bisexual men in my surveys are indeed often autogynephilic. Case closed?

… almost. I don’t have many bi men in my surveys, and most of them are autogynephilic, so it’s not clear whether the non-autogynephilic bi men also score higher. However, today I combined the data from a number of surveys (Broader Gender Survey, Thorough Genderbending Survey, Gender and Psychology Survey, Amazon Mechanical Turk Gender Survey, Sexuality and Attitudes to Gender Survey, Survey on Gender and Valued Experiences) to get a total of 40 non-AGP bi men, along with 469 non-AGP straight men, 761 AGP straight men, 157 AGP bi men, 71 AGP gay men and 64 non-AGP gay men. The different surveys used different ways of assessing gender issues, so I standardized them by dividing by the standard deviation and subtracting the average value from the non-AGP straight men. Here’s the results for the surveys separately:


Results from different surveys wrt. gender issues in different groups of men. The x-axis shows the difference in average gender issues between the group and straight non-AGP men. The y-axis shows the significance of the survey, roughly 1/√N, where N is the number of people in the group. Closer to zero in the y means more significant.

This is a bit messy, so I’ve also made a separate chart with the average gender issues across the surveys:


Weighted average of last diagram, with the weights equal to the number of men in the groups.

To me, those results suggest that bisexual men don’t simply score higher on gender issues due to AGP, but also due to some other factor. Unless… maybe some of these men are AGP, but don’t realize it? Though in that case, it’s hard to explain why AGP bi men score higher than other AGP men. This might be attributed to the fact that interpersonal autogynephilia is associated with gender dysphoria beyond the other forms of autogynephilia, but this should also imply that AGP gay men score higher.

I don’t know.

DSM on the Typology

Transness is bimodally distributed:

In both adolescent and adult natal males, there are two broad trajectories for development of gender dysphoria: early onset and late onset. Early-onset gender dysphoria starts in childhood and continues into adolescence and adulthood; or, there is an intermittent period in which the gender dysphoria desists and these individuals self-identify as gay or homosexual, followed by recurrence of gender dysphoria. Late-onset gender dysphoria occurs around puberty or much later in life. Some of these individuals report having had a desire to be of the other gender in childhood that was not expressed verbally to others. Others do not recall any signs of childhood gender dysphoria. For adolescent males with late-onset gender dysphoria, parents often report surprise because they did not see signs of gender dysphoria during childhood. Expressions of anatomic dysphoria are more common and salient in adolescents and adults once secondary sex characteristics have developed.
Adolescent and adult natal males with early-onset gender dysphoria are almost always sexually attracted to men (androphilic). Adolescents and adults with late-onset gender dysphoria frequently engage in transvestic behavior with sexual excitement. The majority of these individuals are gynephilic or sexually attracted to other posttransition natal males with late-onset gender dysphoria. A substantial percentage of adult males with late-onset gender dysphoria cohabit with or are married to natal females. After gender transition, many self-identify as lesbian. Among adult natal males with gender dysphoria, the early-onset group seeks out clinical care for hormone treatment and reassignment surgery at an earlier age than does the late-onset group. The late-onset group may have more fluctuations in the degree of gender dysphoria and be more ambivalent about and less likely satisfied after gender reassignment surgery.