Gender differences in mathematics anxiety and the relation to mathematics performance while controlling for test anxiety

Why do two students walk out of the same math test with identical scores but completely different relationships to the subject — one feeling fine and the other quietly dreading the next one? That gap isn't about ability. It's about something researchers can measure precisely: mathematics anxiety, a specific state of discomfort triggered by mathematical tasks. When Amy Devine, Kayleigh Fawcett, Dénes Szűcs, and Ann Dowker took the time to measure it carefully — carefully enough to separate it from general exam nerves — a clear pattern emerged about who carries it and what it costs them. Mathematics anxiety, or MA, is not just about disliking math. The background of Devine and colleagues' paper describes it as a state of apprehension, tension, worry, and sometimes outright fear that is specifically elicited by mathematical tasks. It can emerge in the early school years and appears to become more common with age. The consequences are real: people high in MA tend to avoid elective math classes and careers requiring quantitative skills. In the UK, only about seven percent of pupils take mathematics to advanced levels, and disliking the subject is one of the most commonly cited reasons. Here's the methodological problem that has plagued this research area. MA overlaps with test anxiety — the general dread of being evaluated. The two are correlated, typically between 0.30 and 0.50. However, measures of MA correlate much more strongly with each other, between 0.50 and 0.80, showing they're tracking something distinct. The trouble is that most studies haven't controlled for test anxiety when studying MA. So when a researcher finds that anxious students score lower on math tests, they can't easily determine whether it's math-specific anxiety causing the damage or just general exam nerves. Devine and colleagues designed their study to address that problem. The sample consisted of four hundred thirty-three British secondary school students from years seven, eight, and ten — ages roughly eleven to sixteen — at a single rural comprehensive school in England. They took a custom mental mathematics test, then filled out two questionnaires: the nine-item Abbreviated Math Anxiety Scale, known as the AMAS, which has strong internal consistency with a Cronbach's alpha of 0.90, and Sarason's thirty-six item Test Anxiety Scale, which in this sample returned an alpha of 0.86. The math tests were designed carefully — problems in addition, subtraction, multiplication, and division, but deliberately excluding rote-learned times-table items. This ensured students had to actually use working memory rather than just recall. They had five minutes to complete the test. Then came the questionnaires. The sequence matters: math performance first and anxiety measures second, so the test results weren't contaminated by having just been primed to think about anxiety. To analyze the data, the team standardized scores within each year group — important because you'd expect a fifteen-year-old to perform differently than an eleven-year-old. Then they ran correlations, partial correlations controlling for test anxiety, and simultaneous multiple linear regression, entering both MA and test anxiety as predictors of math performance at the same time. That last technique is the key move. Regression separates the unique contribution of each predictor, letting you ask: does math anxiety predict performance above and beyond whatever general test nerves are already explaining? Now for the findings. Girls reported substantially higher math anxiety than boys — zero point thirty-three standard deviations higher on average. Girls also reported higher test anxiety, with a gap of zero point fifty-five standard deviations. Both differences were highly statistically significant. And then the twist: math performance showed no significant gender difference at all. Girls averaged slightly below zero in standardized terms, while boys averaged slightly above, yet that gap did not reach significance. Girls felt more anxious about math, felt more anxious about tests in general, and yet scored the same. That parallel result — equal performance with unequal anxiety — is the first hint that something important is being hidden inside the averages. Drilling into the correlations, the picture sharpens. In girls, math anxiety correlated negatively with performance at a correlation of negative 0.35. Test anxiety also correlated negatively with performance in girls, at a correlation of negative 0.21. In boys, the MA-performance correlation was only a correlation of negative 0.11 and was marginal — not quite reaching conventional significance at a p-value of 0.07. A formal test of the difference between those two correlations showed they were significantly different from each other, with a p-value of 0.023. Not only are girls more anxious, but their anxiety is more tightly linked to their scores. Then comes the critical step: controlling for test anxiety. When Devine and colleagues held test anxiety constant using partial correlations, the negative link between math anxiety and performance in girls remained strong — a partial correlation of negative 0.30, significant at a p-value below 0.001. In boys, the partial correlation was negative 0.11, still marginal. Math anxiety retained its predictive grip on girls' performance even after accounting for the shared variance with general test nerves. For boys, once you consider test anxiety, math anxiety adds essentially nothing to the prediction. The regression analysis confirmed this. In the girls' model, with math anxiety and test anxiety entered simultaneously as predictors, the model was highly significant and explained nearly twelve percent of the variance in math scores. Within that model, math anxiety was a significant negative predictor — a standardized beta of negative 0.32 — while test anxiety was not. In the boys' model, the regression barely reached significance overall and explained only one point five percent of variance; neither predictor was significant on its own. The numbers are telling: nearly twelve percent versus one point five percent. That is not a subtle difference. This is where the methodological contribution earns its keep. Prior studies that didn't separate MA from test anxiety may have been measuring a blend of the two and attributing the effects to math-specific anxiety without justification. By running partial correlations and simultaneous regression, Devine and colleagues isolate the unique signal. This signal turns out to be gendered in a very specific way: math anxiety predicts performance for girls independently of general test nerves, but not for boys. The finding is only visible once you conduct the analysis carefully enough to look for it. The authors situate this within broader literature. Two meta-analyses had previously reported average negative correlations between math anxiety and performance of around negative 0.27 and negative 0.34. The current study's correlations fall within that range for girls, and the regression results extend those findings by specifying the conditions under which the relationship holds. Interestingly, boys in this sample did show a significant negative MA-performance correlation in a separate analysis — a correlation of negative 0.18 — but boys' math anxiety and test anxiety were themselves very strongly correlated, at a correlation of 0.76. That high overlap means the two predictors are nearly redundant for boys in the regression, which is precisely why neither emerges as significant when they're entered together. Girls show the same positive relationship between MA and TA, but at a lower correlation of 0.36, which is enough separation to let math anxiety's unique effect emerge. Pull back to the big picture, and the implications are uncomfortable in a specific way. Girls perform equally to boys on the test. But they carry significantly more anxiety while doing it. If math anxiety is suppressing their performance — and the regression evidence suggests it is — then their observed scores may actually understate what they're capable of. The performance gap we're not seeing could be the one that anxiety erased. The downstream consequences extend beyond a single test session. High math anxiety is associated with avoiding advanced math courses, which closes off career paths requiring quantitative skills before students even know what they're giving up. Since there is evidence that math anxiety can develop during the primary school years, Devine and colleagues argue that longitudinal research needs to start early — tracking students from the beginning of formal math education to understand how anxiety develops, whether it causes performance deficits or results from them, and when intervention might actually help. They also recommend supplementing self-report questionnaires with implicit measures like arithmetic-affective priming and physiological indicators of anxiety, which would reduce reliance on students' own ability to accurately report their internal states. Equal test scores can coexist with unequal emotional burdens. That is the finding this paper keeps returning to. The careful separation of math anxiety from test anxiety, the attention to gender, and the regression that asks what uniquely predicts what — all of it converges on a single provocative conclusion: for girls, math anxiety is a real and independent predictor of math performance. For boys, in this sample, it is not. Understanding why that asymmetry exists and how early it takes hold is the work the authors say still needs to be done. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.