The empathy penalty: a startling study on girls, math, and social sensitivity

Examining a 2016 study that set out to test Baron-Cohen’s popular but flawed autism theory and ended up uncovering something more troubling: a hidden cost to empathy in young girls

Odilon Redon, Ophelia among the flowers

As I dig into autism research, I keep running into this thing called the Empathizing-Systemizing Theory (E-S for short), first proposed by psychologist Simon Baron-Cohen in the early 2000s.

Given how often it’s discussed (one paper on the theory has been cited 1,555 times), I assumed it must offer something of value for autism: a useful lens, maybe even a foundation.

But as I’ve looked closely at one study grounded on the E-S theory, I learned how very wrong my assumption was.

What I found instead was that the E-S model not only fails to deliver insight into autistic cognition, but also quietly reinforces some of our most persistent gender stereotypes. I’ll show you how I got here.

The Empathizing–Systemizing Theory, in brief (and why it’s due for retirement)

Simon Baron-Cohen’s empathizing-systemizing (E-S) theory has been a staple of autism discourse for over two decades.

The original theory went like this: empathizing and systemizing are two distinct cognitive traits. Baron-Cohen described them as separate dimensions. You could score high or low on either one, independently.

• Empathizing is the ability to identify what someone else is thinking or feeling and to respond appropriately. It includes both understanding others’ perspectives and emotionally resonating with them, as seen when someone is skilled in reading facial expressions, grasping social cues, or offering comfort.

• Systematizing is the drive to understand and predict rule-based systems, whether mechanical, mathematical, linguistic, or abstract. It’s the impulse behind spotting patterns, building models, or figuring out how things work.

Baron-Cohen created tests to score how much E and S a person has. He then measured the difference in a person’s scores (S minus E) to slot them into five “brain types”:

• Type B (balanced)

• Type E (more empathizing than systemizing)

• Type S (more systemizing than empathizing)

• Extreme Type E (much more empathizing than systemizing)

• Extreme Type S (much more systemizing than empathizing)

He hypothesized that autistic individuals fall into the last category: the Extreme Type S brain, where systemizing vastly outweighs empathizing.

Over time, though, the model was increasingly misinterpreted and simplified. Rather than being understood as two separate traits, E and S came to be treated as opposites on a single continuum: more empathizing meant less systemizing, and vice versa.

At the same time, the theory’s gendered framing intensified. Baron-Cohen had always linked systemizing to the “male brain” and empathizing to the “female brain,” but this became central. Autism, in this view, came to represent an Extreme Male Brain.

His logic went like this:

• Autism = extreme systemizing + impaired empathizing

• Men systemize more than women, on average

• Therefore, autism is an exaggeration of male cognitive traits

This gendered framing was speculative from the start. Still, it became a cornerstone of the theory’s appeal. Baron-Cohen often cited male overrepresentation in math and engineering as evidence that systemizing, and therefore the autistic brain itself, is aligned with male traits.

However, as we’ll soon see, one of the few studies to directly test this connection uncovered something surprising.

The 2016 study that used the E-S Theory to predict math abilities

That’s all important prelude to what I really want to focus on today: a startling discovery on the link between empathy in girls and math ability.

That link was uncovered in a 2016 study that set out to rigorously test some of the E-S theory’s central claims.¹

The researchers started with a long-standing Baron-Cohen hypothesis: that better math performance stems from stronger systemizing abilities. Yet despite years of speculation and citation, only one study had ever directly tested this and found no evidence of a link.

So, these researchers set out to try again, with sharper tools and a more comprehensive approach.

The researchers studied 112 typically developing children, ages 7 to 12, about half of them girls. Here I want to underline that although the E-S theory is closely tied to autism, this was not a study about autism.

They gave the children math tests, as well as validated questionnaires measuring systemizing (the S score), empathizing (the E score), social responsiveness, reading ability, IQ, and math anxiety. (Math anxiety measures whether a child experiences feelings of tension, worry, or fear when engaging with math-related tasks.)

Their hypotheses, drawn from Baron-Cohen’s work, were straightforward:

• Children who scored high in systemizing on an absolute basis (predicted to include more boys than girls) would perform better in math

• Those with a Type S brain (high systemizing, low empathizing) would outperform those with a Type E brain (the reverse) in math

• If both of those predictions held, then the study would reveal that systemizing tendencies contribute to gender differences in math achievement

• The researchers also set out to test whether E and S were on a continuum, such that a high score in E meant a low score in S, and vice versa

Instead, the study dismantled these assumptions one by one.

What they found undermined the E-S theory at nearly every turn

First, E and S were not on a continuum for the study group. Being strong in one area didn’t predict being low (or high) in the other. They are independent traits. This already chips away at the idea that girls are generally stronger in empathizing relative to systemizing, and vice versa for boys.

Second, although it was true that boys had slightly higher systemizing scores (an insignificant difference), and girls had meaningfully higher empathizing scores, these results related to math ability in surprising ways.

The first surprise was that systemizing scores didn’t correlate with math performance. This held true even when controlling for IQ and reading ability. Kids who scored high on systemizing weren’t any better at math calculations (like arithmetic problems). There was a slight positive link to more conceptual math reasoning tasks, but even that wasn’t strong enough to be statistically significant.

The idea that systemizing predicts mathematical skill, a frequently cited implication of the E–S framework, simply didn’t hold.

Indeed, IQ and reading ability explained much of the variation in math reasoning scores. The systemizing score (S score), by contrast, added no meaningful information.

It’s worth underlining here that Baron-Cohen’s theory was explicitly gendered: he argued that men are, on average, more systemizing than women, and that this may explain male overrepresentation in math-heavy fields like engineering. But this study found no link between systemizing and math performance in children, and no math advantage for systemizing “brain types,” axing that logic at its root.

The second surprise is a real twist: higher empathy scores were linked to slightly lower performance in basic math tasks, even after adjusting for general cognitive ability. This negative link between empathy and basic math was statistically significant for the girls in the sample, but not the boys.

This wasn’t just a null result for systemizing: it was a reversal of expectations. The very trait we often praise in children and especially girls – the ability to tune into others – was the one most closely linked to lower math scores in this group.

In the end, and despite the researchers’ hypothesis, it wasn’t the gap between systemizing and empathizing that mattered. It was empathy alone that predicted math ability.

The researchers dug deeper to figure out why. They found that girls with high E scores were also more socially responsive, as measured by the separate Social Responsiveness Scale (SRS).²

By excavating more layers of linkages, researchers found that empathy alone didn’t directly impair math skills. Instead, it predicted higher social sensitivity scoring on the SRS, and that score was linked to lower math performance in girls.³

It’s worth pausing here. The study didn’t just fail to support Baron-Cohen’s theory; it actively undermined it. It challenged the assumption that systemizing is a cognitive strength linked to mathematical reasoning. It showed that systemizing and empathizing are not diametrically opposed. And it hinted at something more concerning: that empathy, and specifically, social awareness, may actually impair girls’ math performance under certain conditions.

A theory of stereotype absorption, with earlier data to back it up

Why would this happen?

Why would high empathy predict lower math ability, even as other measures of intelligence like IQ and reading ability hold constant?

The researchers offered a compelling hypothesis: girls who are especially socially aware may be more attuned to implicit messages about what girls are (and aren’t) supposed to be good at. In classrooms where teachers harbor anxiety about math or subtly doubt girls’ abilities in it, or where peers assume boys are naturally better at STEM, these girls pick up on those cues and internalize them.

I was surprised – horrified – to discover that this hypothesis builds on established research. Previous studies have shown that when female teachers are anxious or uncertain about their own math abilities, it can negatively affect girls’ (but not boys’) performance, likely by sending subtle signals about who is expected to excel.⁴

This study adds a sharper edge: it suggests that the very capacity to pick up on those signals might itself be a vulnerability in this context, allowing stereotypes to take hold earlier and more deeply in the girls who are most attuned to their environment.

The idea that empathy can backfire in this highly specific way is both fascinating and unsettling. It challenges our usual assumption that a surplus of empathy is always a strength, and a deficit always a weakness. For the socially gifted school girl, the ability to pick up on subtle social cues can be a real asset. But in certain environments, it may also make her more vulnerable to internalizing harmful, unspoken messages.

Where this leaves us

Baron-Cohen’s E–S theory continues to circulate in the literature, but studies like this make it hard to take the theory at face value. At best, it’s an oversimplified model. At worst, it may have distracted researchers for years from more productive questions, especially those about how social cognition interacts with environment, context, and stereotype.

Indeed, studies show that the gender gap in math achievement begins as early as kindergarten, emerging at the same time as teacher expectations that boys are naturally better at math than girls. The linkage to empathy discussed above underscores the urgent need for early classroom interventions, before these subtle but damaging beliefs take hold in young minds.

This isn’t a story about autism, even though that’s how I came across the E-S theory initially. It’s a story about the seductive power of tidy cognitive theories and how easily cultural assumptions, especially gendered ones, get folded into scientific frameworks.

It’s also a story about empathy: not as a universal good, but as a context-sensitive tool, one that can either support or sabotage us depending on what it’s tuned to.

The better question is not: “Why aren’t girls better at math?” but: “Who is most attuned to what they’re being told – explicitly or not – about who they’re supposed to be?”

There’s no reason this phenomenon would be limited to math. What other messages are being internalized by socially-attuned girls about the limits of female achievement?

Did you appreciate this research deep dive? If so, please consider supporting my work here at Strange Clarity.

1. Escovar, E., Rosenberg-Lee, M., Uddin, L., et al. (2016). The empathizing–systemizing theory, social abilities, and mathematical achievement in children. Scientific Reports, 6, 23011. https://doi.org/10.1038/srep23011

2. The Social Responsiveness Scale (SRS) is a parent-report questionnaire designed to assess a child’s social functioning across a range of real-world contexts. It includes subscales measuring:

• Social Awareness, the ability to notice social cues (e.g., recognizing when someone is bored or upset)

• Social Cognition, the ability to interpret those cues and understand others’ thoughts or emotions

• Social Communication, how effectively a child uses language and nonverbal signals in social interactions

• Social Motivation, the drive to engage in social behavior and sustain relationships

• Restricted Interests and Repetitive Behavior, behaviors commonly associated with autism such as rigidity, repetitive actions, or intense focus on narrow interests

3. How did they figure this out? That’s a question I had too. To test whether social sensitivity explains the link between empathy and lower math scores, the researchers used a statistical tool called mediation analysis. The idea behind mediation is that one variable (in this case, empathy) may influence another (math performance) indirectly, through a third variable that sits in the middle.

The study showed that higher empathy predicted lower math calculation scores, especially in girls. The researchers then looked for possible mediators: traits that might explain how empathy affects math. For this, they used subscales from the Social Responsiveness Scale (SRS) (see footnote above), which measures various aspects of social ability.

They ran a series of regression analyses to see if these social subscales reduced the strength of the link between empathy and math. That would suggest that the empathy–math relationship runs through the subscale.

The results: in girls, Social Cognition, Social Awareness, and Social Communication all mediated the link. These traits help explain why empathic girls performed slightly worse in math. In boys, the same patterns weren’t statistically significant. The only gender difference that reached significance was in Social Awareness, which had a stronger negative impact on girls’ math scores than on boys’.

These subscale distinctions are nuanced and I struggled for a while to understand the implications. What it means is that multiple forms of social sensitivity matter for girls, one in particular stands out. For Social Cognition and Social Communication, girls' scores predicted lower math, but the gender difference wasn't strong enough statistically to say the effect was truly different in girls versus boys. We can’t say with confidence that they mattered more for girls than boys, only that they didn’t show up in boys.

The results are different for Social Awareness. Even if boys and girls had the same measure of Social Awareness, the girls who were more socially aware tended to do worse in math than the boys. This means that something specific to girls’ heightened Social Awareness accounted for the worse math ability.

Recall that Social Awareness refers to the ability to pick up on social cues. The logical conclusion is that through their heightened Social Awareness, the girls were picking up on social cues that either were visible only to them, or meaningful only to them. Marrying this result with prior research on stereotyping, the researchers concluded that they were picking up on gender-based stereotyping cues.

In plain terms: the girls who were most attuned to social cues may have been the most likely to notice and absorb subtle classroom messages about who’s “supposed” to be good at math.

4. Beilock, S. L., Gunderson, E. A., Ramirez, G., & Levine, S. C. (2010). Female teachers’ math anxiety affects girls’ math achievement. Proceedings of the National Academy of Sciences, 107(5), 1860–1863. https://doi.org/10.1073/pnas.0910967107