Math-anxious people consistently underperform in math. The most widely accepted explanation for why this underperformance occurs is that math-anxious people experience heightened anxiety when faced with math, and this in-the-moment anxiety interferes with performance. Surprisingly, this explanation has not been tested directly. Here, using both self-report and physiological indices of anxiety, we directly test how much in-the-moment anxiety explains math-anxious underperformance. Results indicate that in-the-moment anxiety indeed explains why math-anxious people underperform-but only partially, suggesting a need to seriously consider alternative mechanisms. Results also showed that while some highly math-anxious individuals-those with high levels of heart rate variability-experienced less in-the-moment anxiety, they nevertheless performed no better at math. For these individuals, math-anxious underperformance must occur for reasons unrelated to in-the-moment anxiety. More broadly, our findings point to substantial individual heterogeneity in the mechanisms underlying math-anxious underperformance. Accounting for this mechanistic heterogeneity may prove vital for optimally boosting math performance in math-anxious individuals.

Anxieties that are specific to a particular kind of thinking have been demonstrated for a variety of cognitive domains. One hypothesized consequence of these anxieties is reduced interest in pursuing activities and, consequentially, careers that involve the type of thinking in question in an effort to avoid engaging in that type of thinking. There is little research addressing this avoidance hypothesis, possibly because it is difficult to categorize pursuits as objectively creative or spatial. Here, we measured the perceptions that participants, themselves, hold about how much pursuits (careers and activities) involve different types of thinking. We developed a novel framework for calculating affinity coefficients, within-person associations between perceived cognitive involvement and interest across several pursuits. Having a negative creative affinity coefficient, for instance, means being less interested in pursuits the more they are perceived as involving creative thinking. Results across three separate cognitive domains (creativity, mathematics, and spatial reasoning) reliably showed that higher anxiety in a domain uniquely predicted a lower affinity coefficient in that domain, providing consistent evidence of avoidance tendencies linked to cognition-specific anxieties. These findings suggest that feeling anxious about particular types of thinking may play a significant role in shaping our interests, both big and small.

Anxiety within the domains of math and spatial reasoning have consistently been shown to predict performance within those domains. However, little work has focused on how specific these associations are. Across two studies, we systematically tested the degree of specificity in relations between anxiety and performance within math and spatial reasoning. Results consistently showed that anxiety within a cognitive domain predicted performance in that domain even when controlling for other forms of anxiety, providing evidence that cognition-specific anxieties are valuable for understanding cognition-specific performance. We also found that general trait anxiety did not explain a significant portion the anxiety-performance link in either math or spatial reasoning, suggesting that these anxiety-performance associations are not due to the propensity to feel anxious generally. Interestingly, while spatial anxiety did not explain any of the anxiety-performance association in math, math anxiety did explain a significant portion of the anxiety-performance link in spatial reasoning. These results suggest that, while links between anxiety and performance cannot be reduced to a single underlying general anxiety construct, there may nevertheless be overlap between domain anxieties. We end by calling for a more detailed examination of the unique and shared mechanisms linking anxiety and performance across disparate cognitive domains.