the science of gender identity (part 3: psychology)

This is the third post in a multi-part series surveying the current science of gender identity, particularly with regard to the transgendered population. My first post on the subject covered proposed genetic associations and corresponding research. The second post on the matter discussed observed differences in brain anatomy between transgendered and cisgendered individuals. Here I survey a subset of the available the psychological science on the matter, starting by describing recent research on the stability of gender cognition in young transgendered children (it is stable). I then discuss research that strongly suggests that the psychological functioning of transgendered individuals improves with commencement of hormone replacement therapy. Finally, I briefly describe two studies addressing questions about whether personality differences exist between transgendered and cisgendered individuals.

Stable Gender Cognition in Transgendered Children

A parent challenged by the emergence of a transgendered child might reasonably hold skepticism, asking questions such as “is this a phase?”, “are they pretending?”, or “are they delayed in gender cognition growth?” Indeed therapists and medical providers struggle with these questions as well, as they severely impact decisions a transgendered child and their parents might want to make such as the application of puberty-blocking drugs and cross-sex hormone treatment. Unfortunately, there is very little research available to answer such questions, and existing research involves evidence dominated by children’s self-reporting of gender-identity. However, self-reporting would be unreliable under the conditions addressed by the questions above, so more precise measurement is needed.

A study [3] described below complements self-reporting (hereafter called an “explicit” measure of gender cognition) with two “implicit” measures of gender cognition. Such implicit measures are less susceptible to the possible bias discussed above. Particularly, implicit measures of identity and preferences were recorded and matched for comparison with the explicit measurements. The study recruited 32 transgendered pre-pubertal children and their cisgendered pre-pubertal siblings as one type of control. 32 additional control pre-pubertal cisgendered children were recruited and matched to the subjects by age and natal sex.

All transgendered study subjects were living in their preferred gender role, a matter I’ll discuss in more detail later. The participants then answered two implicit questions and three explicit questions:

Gender-preference implicit association test

This test evaluates a child’s implicit gender preferences. Subjects taking this test are asked to classify four categories of pictures (“male”, “female”, “good”, and “bad”) by forming key-value pairs such as (“female-good”, and “male-bad”). The pictures of male and female stimuli were photographs of children. The pictures of “good” and “bad” stimuli included things like puppies, ice-cream, snakes, and car accidents. The children then label their key-value sets with smiley faces or frowns for “good” and “bad”, and with additional photographs of a male and a female for gender (so they don’t have to read or write).

Such a design is considered more robust to the issues expressed at the top of this post involving explicit measurement. Scoring for each outcome results in more positive values when a test taker’s natal sex was associated with “good”, and negative values otherwise. In the case of the transgender subjects, they are scored twice, once with “good” associated with their gender identity and once with “good” associated with their natal sex.

Gender-identity implicit association test

This test evaluates a child’s implicit gender identity. The design is similar to the gender-preference implicit association test described above, except that “good” was replaced by “me” (“myself”, “I”, “mine”) and “bad” was replaced by “they” (“them”, “theirs”, “other”). As a result of this replacement some reading skills were required. The scoring followed the same pattern: More positive values indicate association of “myself” with stimuli related to the natal sex, negative values otherwise. Again, the transgendered subjects were scored both by natal sex and by gender identity.

Explicit gender peer preference

In a controlled manner carefully matching age and attractiveness, subjects were shown a series of pictures of boys and girls and asked which they’d most like to be friends with. Scores were coded in the controls by how many times a peer was selected of their own natal sex. For transgendered subjects, scores were calculated in this manner and in the opposite manner.

Explicit object preference

Participants were shown a series of images of children (varied by gender) and told that each had a preferred toy. Subjects were then asked which of the preferred toys they’d prefer playing with (testing for which gender “endorses” a toy). When the endorser and the participant are of the same natal sex among the controls, a count was added to the score. Transgendered subjects’ responses were recorded both by natal sex and by gender identity.


The plots below, taken directly from the study paper [3], illustrate clearly that transgender subjects’ gender cognition follows that of their cisgendered peers when scored by gender identity for each of the four tests described above (p-values > 0.20). When scored by natal sex the departure is significant (p-values < 0.002):

The researchers were looking for three things in the data to answer questions about gender cognition in transgendered children. First, if the results for transgendered children looked confused, the conclusion might be that gender cognition in these children was not mature yet, or they were “confused”. However, the data revealed a clear direction in all measures.

Second, to assess whether the transgendered children were merely pretending, the researchers looked for the pattern of natal-sex congruent response on the implicit measures and gender identity response on the explicit measures. The results do not show this either.

Finally, the observed pattern of gender identity congruent responses on both the implicit and explicit measures strongly suggests that the transgendered children “know who they are”—that their gender identity is stable.

There is one major problem with this study however, but it is not a show-stopper: I mentioned above that the transgendered subjects were living according to their preferred gender identity. This meant they lived in supportive environments that permitted this. Therefore, it is unclear how well the results generalize to the total transgendered experience, given that this kind of familiar support is rare. That being said, this study still provides strong evidence that gender cognition is stable in young children.

Hormone Replacement Therapy and Psychological Functioning

Two studies investigated the impact of hormone replacement therapy on the psychological functioning and well-being of transgendered individuals [1, 2], both finding evidence that psychological functioning improves with hormone administration. The first study [1] administered the Social Anxiety and Distress Scale (SADS) and the Hospital Anxiety and Depression Scale (HADS) tests to 187 transsexual subjects, of which 64% where undergoing hormone replacement therapy while the remaining 36% had not started such treatment. The researchers controlled for covariates such as age, education level, duration of hormone treatment, and whether any cross-sex surgeries have occurred. 60% of the subjects were male-to-female transsexuals and 40% were female-to-male.

The SADS test measures social anxiety and social distress using a true/false questionnaire. Higher scores correlate with greater social anxiety. Both groups (with/without hormone treatment) fell within normal ranges, but the mean score for the group under hormone replacement therapy was significantly lower (p=0.038, F-test).

The HADS test quickly measures depression and anxiety using a questionnaire and is intended for use in non-psychiatric settings. It is not as comprehensive as the MMPI-2 (discussed below), but useful for initial diagnostics. It divides into two subscales: HAD-D for measuring depression and HAD-A for measuring anxiety. Mean HAD-D scores fell in the normal range for both groups, as did mean HAD-A for the group undergoing hormone replacement therapy. However, mean HAD-A for the non-treated group scored in the range suggesting possible mood disorder. The differences in means for both tests between the groups were significant: p=0.001 for HAD-A and p=0.002 for HAD-D (F-test).

There are a few limitations with this study, which the authors carefully describe. First, the subject pool came from Catalonia, which has a very developed transgender care infrastructure. Therefore the results may not generalize well to less accommodating regions. Second, there may be selection bias in that the most socially anxious or depressed may not seek out transition-related care. Finally, a longitudinal study is needed to firmly establish causality.

The second study [2] investigated the psychological improvement in transgendered men after starting testosterone therapy. Baseline measurement of psychological functioning for each study subject was established through administration of the Minnesota Multiphase Personality Inventory (MMPI-2) immediately prior to starting testosterone injections. The MMPI-2 was then re-administered to each participant three months after starting the testosterone treatment to assess changes. Because this was a longitudinal study design stronger conclusions about causality could be made.

The MMPI-2 is a popular, reliable, and well-studied tool for assessing psychopathology. It employs 567 true or false questions to examine clinical conditions such as depression, paranoia, and schizophrenia. Its scoring procedure involves comparison to norms by natal sex—the initial validation of the tool did not separate sex and gender—which is a problem for transgendered individuals. Therefore this study’s researchers employed the male scoring procedure when comparing results to those of male controls, and the female scoring procedure when comparing results to those of female controls. This enabled them to separate the effects of sex and gender in the analysis.

48 transgender men, 62 female controls, and 53 male controls were recruited for the study. The covariates of age, educational attainment, and employment status were controlled for.

At three months, after adjusting for baseline measurements, significant reductions in MMPI-2 scores for hypochondria, depression, hysteria, and paranoia occurred for the transgender subjects relative to the female controls. (Higher scores indicate greater pathology). A similar comparison to male controls yielded no statistically significant results. These two studies suggest that administration of hormone replacement therapy to transgendered individuals improves their psychological condition. However, more research into the matter would help.


Two studies [4, 5] explored personality differences between adult transsexuals and controls, each study applying a similar method but with different ethnic groups to investigate if findings are consistent across culture.

The first study [4] took place in Barcelona with 166 male-to-female transsexuals and 88 female-to-male transsexuals. There were 404 controls (division by natal sex unspecified). All subjects were mentally healthy. The covariates of age, educational attainment, and employment status were controlled for.

The researchers administered the Temperament and Character Inventory (TCI) to the subjects to characterize their personalities. The TCI is a 240 item questionnaire designed to measure each of seven general dimensions of personality: “novelty seeking”, “harm avoidance”, “reward dependence”, “persistence”, “self-directedness”, “cooperativeness”, and “self-transcendence”.

The study produced two findings: First, the personality differences between transsexuals and the controls were not clinically relevant—i.e., the transsexuals have “normal” personalities. Second, the researchers observed that the differences that did exist between female-to-male transsexuals and male-to-female transsexuals mirrored previously observed differences between cisgendered males and cisgendered females using the same measurement procedure. This means that male-to-female transsexual personalities are more congruent with cisgendered female personalities, and likewise for female-to-male transsexuals and cisgendered male controls.

The second study [5] administered a shortened form of the TCI to 187 female-to-male and 72 male-to-female Japanese transsexuals, along with 184 cisgendered male and 159 cisgendered female controls. All subjects were psychologically healthy. These researchers observed high and statistically significant values for the personality traits of “reward dependence” (relative to cisgendered males) and “cooperativeness” (relative to cisgendered males and male-to-female transsexuals) in the female-to-male subject group. They also observed a high and statistically significant value for trait of “self-transcendence” in the male-to-female group, relative to the values for each of the other three groups. These findings concur with [4] only in the “self-transcendence” observation for male-to-female transsexuals ([4] also conducted pair-wise comparisons, which I did not describe above since the differences observed were not considered “clinically relevant” by the authors). This difference in outcome might be explained by cultural differences or by differences in TCI version used (both language and length).

Related Posts

the science of gender identity (part 1: genetics)
the science of gender identity (part 2: brain anatomy)