Autism : no evidence of connection between testosterone and reduced cognitive empathy


It’s long been known that autism is far more prevalent in males than in females. What hasn’t been understood is why.

“Of course, the primary suspect when we have something that is sharply differentiated by sex is testosterone,” says Gideon Nave, an assistant professor of marketing in Penn’s Wharton School.

Yet a new study led by Nave implores scientists to keep looking. In two randomized controlled studies of testosterone administration which were the largest of their kind and included nearly 650 men, Nave and colleagues found no evidence of a link with cognitive empathy, the capacity to read the emotions of others, a trait that is characteristically impaired in people with autism.

They report their findings in the journal Proceedings of the Royal Society B: Biological Sciences.

“Several earlier studies have suggested a connection between testosterone and reduced cognitive empathy, but samples were very small, and it’s very difficult to determine a direct link,” says Amos Nadler of Western University, the first author of the study.

“Our results unequivocally show that there is not a linear causal relation between testosterone exposure and cognitive empathy.”

Prior to this work, the strongest evidence for a link between testosterone exposure and reduced cognitive empathy came in 2011 in a study that found administering testosterone to healthy women reduced their performance on a test of reading emotions.

The results suggested the testosterone impaired their performance.

Moreover, the work pointed to the ratio of the length of the participant’s second finger to their fourth finger, known as the 2D:4D ratio, as a proxy for sensitivity to testosterone.

Some believe that the ratio declines with increased in utero exposure to testosterone, though evidence for that connection is mixed.

That study’s authors contended that their findings supported the idea that prenatal testosterone exposure created a more masculinized brain that less readily inferred the emotional state of others.

The study was used as support for the “extreme male brain” hypothesis of autism, which contends that autism is an exaggeration of “male” tendencies toward a cognitive style characterized by systemizing over empathizing.

The earlier investigation, however, relied on a sample size of just 16 subjects.

And most other research investigating the idea that testosterone is linked to reduced cognitive empathy had relied on correlative rather than causative evidence and had also resulted inconclusive findings.

To obtain more rigorous data on the connection, Nave, Nadler, and their colleagues conducted two randomized controlled studies in which 643 healthy men received an application of testosterone gel or a placebo and completed questionnaires and behavioral tasks that measured cognitive empathy.

Participants were then shown a photo of an actor’s eyes and asked to select the emotional state that best described their expression. All participants also had their 2D:4D ratio measured.

While the testosterone gel did increase participants’ levels of the hormone, the researchers found no evidence that testosterone administration affected performance on tests of cognitive empathy.

They also found no relationship between participants’ performance on the tests and their 2D:4D ratio.

“The results are plain,” says Nave. “However, it’s important to note that the absence of evidence is not evidence of absence.

We found that there is no evidence to support this effect of testosterone, but that doesn’t rule out any possible effects.

From what we know, though, it seems that if testosterone does have an influence, the effect is complex, not linear. Reality is typically not that simple.”

Nadler notes that while the 2011 study included women and the current one included men, one would still expect to find differences if the effect of testosterone were real, especially since men were exposed to more testosterone prenatally, which would presumably amplify the effects of administration. And the new study includes more participants than the earlier one by almost two orders of magnitude, augmenting the researchers’ confidence in the results.

This shows the outline of two heads

Prior to this work, the strongest evidence for a link between testosterone exposure and reduced cognitive empathy came in 2011 in a study that found administering testosterone to healthy women reduced their performance on a test of reading emotions. The image is in the public domain.

The extreme male brain theory of autism has received a lot of attention but, Nave notes, “if you look at the literature carefully, there is still not really strong support for it.

“For now, I think we have to embrace our ignorance on this.”

Gideon Nave is an assistant professor of marketing at the University of Pennsylvania’s Wharton School.

Nave coauthored the study with Amos Nadler of Western University, Colin F. Camerer of the California Institute of Technology, David Zava of ZRT Laboratory, Triana L. Ortiz and Justin M. Carré of Nipissing University, and Neil V. Watson of Simon Fraser University.

Funding: The study was supported by Caltech, the Ivey Business School, IFREE, the Russell Sage Foundation, the University of Southern California, INSEAD, the Stockholm School of Economics, the Wharton Neuroscience Initiative, the Natural Sciences and Engineering Council of Canada, and the Northern Ontario Heritage Fund Corporation.


In human social environments, the ability to make sense of and predict other people’s behavior is crucial for physical and social survival (1). To meet this adaptive challenge, humans have a set of evolved cognitive-empathetic mechanisms, enabling them to accurately infer motives, intentions, thoughts, and emotions of others, largely from subtle bodily cues (24). Cognitive empathy is central to social intelligence and occurs automatically and mostly unconsciously (4).

A major source of information providing cues for cognitive empathy is the eye region of the face, which contains subtle facial expression. The ability to “read the mind from the eyes” is sexually dimorphic, with females on average typically outperforming males (46).

The androgen (sex steroid) hormone testosterone is thought to be involved, as testosterone represents the biggest hormonal difference between the sexes and affects sociality (78).

However, testosterone’s action in the brain is both organizational and activational: first, the hormone preprograms the brain during early development, and, in later life, it selectively modifies brain processing to facilitate or inhibit behaviors depending on social context (9).

In humans, the fetal period of prenatal development is considered critical for testosterone’s effects on brain organization (between weeks 12 and 19 of gestation), whereas the hormone’s activational effects come into prominence in adolescence and adulthood (910).

Interestingly, the androgen theory of autism proposes that fetal programming of the brain by testosterone negatively affects social intelligence (11).

Both cognitive empathy deficits typically seen in autism, and the male-bias of autism, are indirect evidence consistent with the theory.

Moreover, recent studies in which fetal testosterone was sampled from the amniotic fluid of pregnant women provide for more direct evidence: in young typically developing children, fetal testosterone is inversely correlated with eye contact at 12 mo (12), social cognition at age 4 y (13), and social intelligence including reading the mind from the eyes at age 6 to 8 y (14).

Fetal testosterone is associated with a fixed somatic marker that can be indexed after birth: the length ratio of the right hand’s second (i.e., index) to fourth (i.e., ring) finger (2D:4D ratio). Males on average have a significantly lower 2D:4D ratio on their right hand and fetal testosterone is thought to underlie this sex difference, including its variability within the sexes (1516).

The reliability of 2D:4D ratio as a marker of fetal testosterone is substantiated by a large amount of correlational evidence in animals and humans (1517). Moreover, meta-analytic data show that 2D:4D ratio is unaffected by later testosterone fluctuations or circulating levels of testosterone in adulthood.

The ratio is therefore considered a useful, noninvasive marker of fetal testosterone (16). A recent hormone manipulation demonstrated the validity of the marker in animals: experimental testosterone elevation in pregnant rats lowered the 2D:4D ratio of the right paw of their offspring measured in adulthood (18). Further strong evidence in humans comes from a study showing that androgen receptor (AR) polymorphisms (i.e., increasing CAG repeats in exon 1), which produce less effective AR protein, predict less masculine 2D:4D ratios (19).

Furthermore, higher (i.e., more feminized) 2D:4D ratios are observed in women with genetic mutations that disrupt AR function as seen in androgen insensitivity syndrome (20). In light of the interrelations between AR polymorphisms and digit ratios, Breedlove recently suggested that, although 2D:4D ratio is typically discussed in terms of its relationship to prenatal levels of androgens, digit ratio more accurately reflects total androgen stimulation in terms of prenatal androgen levels and androgen sensitivity (17). Finally, with respect to the androgen theory of autism, it is important to note that lower (i.e., more masculinized) 2D:4D ratios have been observed in children with autism or Asperger syndrome, and also in their first-degree relatives (21).

There is also strong evidence for the activational effects of testosterone on human social and emotional behavior. Placebo-controlled testosterone administration studies in typical young adult women have shown reductions in mimicry and conscious recognition of emotional facial expressions (2223).

However, to our knowledge, there is no such direct evidence for down-regulatory effects of testosterone administration on social intelligence or cognitive empathy in particular. Moreover, fetal testosterone might critically mediate in the activational effects of testosterone on human social behavior. In nonhuman animals, it has been shown that early organizational effects of testosterone strongly facilitate the activational effects of the very same hormone in adulthood (910).

Testosterone administration in adult humans might thus impair social intelligence, but especially in those most primed by the same hormone prenatally. Accordingly, we conducted an experiment to test whether testosterone administration impairs cognitive empathy, and whether predicted testosterone-induced impairments in cognitive empathy varied according to the 2D:4D ratio marker of fetal testosterone.Go to:


To investigate effects of testosterone on cognitive empathy, we temporarily elevated the levels of testosterone in young adult females by using a validated sublingual 0.5-mg single-dose testosterone administration technique. We used a crossover, double-blind, placebo-controlled, within-subjects design with a computerized adaptation of the validated reading the mind in the eyes task (RMET; as the behavioral measure of social intelligence (5624). To enable measurement of 2D:4D ratio, subjects’ right hands were scanned and the 2D:4D ratio was computed from these scans using Adobe Photoshop as a measurement-precision tool. This was carried out by two experienced raters, who used the Millet and de Witte procedure (25), and who remained blind to the experiment. Statistical analyses are based on nonparametric tests (Wilcoxon rank tests and Spearman correlations), but additional parametric statistics are applied for insight in explained variances. The 2D:4D ratios measured by the two raters were highly correlated [Spearman ρ(14) = 0.96; P < 0.001].

First, we investigated potential activational effects of testosterone on cognitive empathy. As can be seen from Fig. 1A, compared with placebo, testosterone administration significantly impaired the ability to read the mind from the eyes [Wilcoxon repeated-measures nonparametric test, Z(1, 16) = −2.24; P = 0.013, one-tailed], with 75% of the subjects showing a decrease in performance on the RMET after testosterone administration. Next, we addressed the relation between fetal testosterone and cognitive empathy, first at baseline by relating 2D:4D ratio to mind-reading performance after placebo: nonparametric Spearman correlations over these variables were not significant [ρ(14) = 0.30; P > 0.25]. However, Spearman correlations showed that the relation between 2D:4D ratio and the impairment in cognitive empathy induced by testosterone administration was highly significant [ρ(14) = 0.85; P < 0.0001].

As can be seen from Fig. 1B, applied as a regressor (i.e., parametrically), fetal testosterone exposure (as inferred from 2D:4D ratios) explains more than 50% of the variance in the effect of testosterone administration on cognitive empathy. To qualify this effect, we applied a median split on the individual 2D:4D ratio measures to create groups of high and low fetal testosterone. Wilcoxon repeated-measures analyses (Fig. 1C) showed no effects of testosterone administration on cognitive empathy whatsoever in subjects with low fetal testosterone exposure [i.e., high 2D:4D ratio; Z(1,8) = 0.00; P = 1]. However, in line with expectations, subjects with high fetal testosterone exposure (i.e., low 2D:4D ratio) showed a strongly significant reduction in cognitive empathy after testosterone administration [Z(1,8) = −2.54; P = 0.006, one-tailed].

In addition, consistent with the down-regulatory effects of testosterone administration on mind reading, there was a significant negative correlation between salivary testosterone and cognitive empathy in the placebo condition [ρ(14) = −0.45; P = 0.041, one-tailed]. Importantly, however, testosterone levels measured from saliva (before administration of testosterone or placebo) did not differ between the experimental conditions, were unrelated to 2D:4D ratio, and did not mediate in our critical effect of testosterone administration on cognitive empathy (all P values > 0.40). Finally, none of these effects were caused by secondary-generated mood changes (all Pvalues > 0.18), and our data were not confounded by expectations or subjective biases of the participants with respect to the effects of testosterone in general, or effects of testosterone on cognitive empathy in particular (all P values > 0.30; Experimental Procedures).


We have shown that a single administration of testosterone in female subjects leads to a significant impairment in the ability to infer emotions, intentions, and other mental states from the eye region of the face. Our data provide causal evidence for the hypothesis that testosterone levels negatively influence social intelligence (2627). Moreover, the 2D:4D ratio fetal testosterone marker predicted more than 50% of the variance in this effect; i.e., sizable effects of testosterone on cognitive empathy were seen in only subjects who were highly prenatally primed by testosterone, inferred from low 2D:4D ratios. Thus, our data convincingly show effects of testosterone administration on cognitive empathy, and these may depend on fetal testosterone priming.

There was no relation between the fetal testosterone marker and performance on the RMET at baseline, which is consistent with recent correlational data from large groups of adult humans (56), but not with data from children (14). Although correlational studies, which usually include larger subject samples, report associations between the fetal testosterone marker and economic and social behavior in adult humans (528), the chance of finding such effects might theoretically be greater in young children as their brains have not yet been reorganized by the testosterone surges of adolescence (9). Crucially, however, and consistent with rodent research, we show that fetal testosterone comes into prominence when its priming is experimentally activated by testosterone administration in adulthood (10). Indeed, at present, there is also a negative relation between baseline salivary testosterone levels and social intelligence, but, consistent with the animal data (910), it was testosterone’s early organizational effect indexed by 2D:4D ratio that predicted the effects of administration of the very same hormone on behavior.

Recently, some researchers have expressed doubt over the sensitivity of 2D:4D ratio as an individual marker for differences in prenatal androgen exposure (20). Certainly there is variance in 2D:4D ratio that cannot be attributed to prenatal testosterone alone, and sex or certain phenotypes cannot be predicted from individual digit ratios (1729). However, 2D:4D ratio is useful for predicting human behavior when comparing groups, and has proven to be a valuable marker for individual differences in prenatal androgen exposure in correlational studies (1517). This is substantiated by the present findings wherein digit ratio explained more than 50% of the individual variance in the effects of testosterone on cognitive empathy. One study into 2D:4D ratio variance examined the relationship between fetal steroid hormone levels measured in amniocentesis fluid and the 2D:4D ratio. The relation between fetal testosterone and 2D:4D ratio only became significant when fetal estradiol was taken into account (i.e., fetal testosterone:estradiol ratio) (30), suggesting fetal estradiol and testosterone interactively contribute to 2D:4D ratio (31). This is interesting, but seems inconsistent with the evidence from rodents showing that testosterone and estradiol are involved in masculinizing the brain; i.e., in rodents, brain masculinization depends on circulating testosterone acting on AR receptors and conversely on testosterone converted by the enzyme aromatase into estradiol acting on estrogen receptors (32). Crucially, however, in primates, including humans, brain masculinization evidently is accomplished primarily via androgens acting directly on ARs. The stimulatory role of estrogen receptors in masculinizing the human brain is negligible, because individuals with complete androgen insensitivity syndrome (i.e., nonfunctional ARs) show feminized behavior (33), whereas masculine behavior can be observed in men with dysfunctional aromatase (34).

In sum, we show that 2D:4D ratio has strong predictive power in estimating effects of testosterone administration on cognitive empathy in humans. This finding is consistent with animal data (910) and establishes that 2D:4D ratio might be a useful marker for differing effects of testosterone administration in humans.

Opposite effects (i.e., improvements in cognitive empathy) have been shown after administration of the “female-type” peptide hormone oxytocin in healthy young males (24). Furthermore, improvements in cognitive empathy after oxytocin administration were recently also observed in young males diagnosed with autism or Asperger syndrome (35), which suggests that oxytocin may hold some potential for intervention in autism spectrum disorders conditions. It seems that, depending on conditions, testosterone works antagonistically on the neuroendocrine system that expresses oxytocin (3637). Steroid and peptide hormones are known to act interdependently in the brain, and as hyperfunction of testosterone and hypofunction of oxytocin has been suggested in autism (3638), an integrative approach might add to our understanding of this neurodevelopmental condition (26).

In conclusion, we have shown that administration of testosterone in humans leads to significant impairment in the cognitive empathic ability to infer emotions, intentions, feelings, and other mental states from the eye region of the face. Moreover, a proxy of subjects’ fetal testosterone, the right-hand 2D:4D ratio, suggests that prenatal testosterone priming is crucial in this effect. Here we provide evidence in humans that activational effects of testosterone on adult social cognition may depend on early, prenatal organization by the same hormone, testosterone. Further research is necessary to establish whether this finding is specific to social intelligence or can be generalized to other human behaviors in which testosterone plays a role, such as social dominance or sexual motivation (2639). Nonetheless, our data provide unique insights into the psychobiology of social intelligence and open up opportunities for further research in human social neuroendocrinology.

University of Pennsylvania
Media Contacts: 
Katherine Unger Baillie – University of Pennsylvania
Image Source:
The image is in the public domain.

Original Research: The study will be published in Proceedings of the Royal Society B.


Please enter your comment!
Please enter your name here

Questo sito usa Akismet per ridurre lo spam. Scopri come i tuoi dati vengono elaborati.