Same Self-Doubt, Different Causes: How Anxiety and Gender Distort Confidence

Women usually shake off gender-linked self-doubt over time, but anxiety-induced insecurities tend to grow with reflection.

Confidence shapes the risks we take, the opportunities we pursue, and the goals we set for ourselves. So what happens when that confidence doesn’t match reality on certain tasks? Research published in Psychological Medicine suggests people underestimate their abilities for very different reasons depending on whether they’re anxious or influenced by gender expectations.

Scientists from University College London and the University of Copenhagen studied how 1,447 people formed confidence judgments after making decisions. Their discovery challenges conventional wisdom about self-doubt: anxiety makes underconfidence worse over time, while gender-related underconfidence actually improves when people take longer to reflect on their performance.

“We found that underconfidence associated with anxiety symptoms became more prominent the longer individuals took to make metacognitive judgments, suggesting that it is exacerbated by additional time for introspection,” the researchers wrote. “In contrast, gender-related underconfidence decreased with greater metacognitive judgment time, suggesting that additional time for introspection is able to remediate prepotent biases.”

The study analyzed data from four large online experimental datasets where participants completed visual tasks and then rated their confidence in their performance. Researchers measured how confident people felt, and how long they took to arrive at those confidence judgments.

Anxiety and Gender Create Opposite Patterns

The research team, led by Sucharit Katyal and Stephen M. Fleming, wanted to understand whether different types of underconfidence share common roots or emerge from distinct psychological processes. They focused on two well-documented sources: anxiety symptoms and gender.

Previous studies have shown that people with higher anxiety tend to underestimate their abilities, even when their actual performance matches that of less anxious individuals. Similarly, women often report lower confidence than men on identical tasks, despite performing equally well.

When participants with higher anxiety symptoms took longer to report their confidence, their self-doubt deepened. More time appeared to allow negatively biased evidence about performance to accumulate, lowering confidence even further.

Women showed the opposite pattern. When they took more time before reporting confidence, the gap between their confidence and men’s confidence shrank. With additional reflection, women’s confidence rose to match their actual performance, while men’s stayed relatively constant.

Building a Computer Model of Self-Doubt

To understand why anxiety and gender affect confidence so differently, the researchers built a computational model that simulates how confidence judgments may form over time. The model treats confidence formation as an accumulation of evidence, similar to how we make decisions.

After making a decision, our brains continue gathering evidence about whether that decision was correct. This post-decision processing shapes how confident we ultimately feel. The model identified three ways this process can become distorted: the brain might accumulate biased evidence, the threshold for feeling confident might shift, or the relationship between accumulated evidence and reported confidence might change.

When the researchers applied this model to their data, they found anxiety and gender operated through different mechanisms. Anxiety triggered both a shift in confidence thresholds and a steady accumulation of negatively biased evidence over time. Each passing second allowed more negatively biased evidence about performance to contaminate the confidence judgment.

Gender differences stemmed only from shifted confidence thresholds. Women started with lower confidence criteria, but because they accumulated unbiased evidence about their performance, taking more time allowed that evidence to overcome initial doubt.

What This Means for Building Better Confidence

These findings suggest that different approaches may be more helpful for addressing underconfidence depending on its source. For people experiencing anxiety, relying more on early confidence estimates might be beneficial, as additional reflection time allows negative biases to accumulate. For addressing gender gaps in confidence, the opposite approach may apply. Taking time to carefully evaluate performance could help overcome socially learned biases. At the longest response times in the study, gender differences in confidence disappeared entirely.

The research examined participants ranging from young adults to older individuals across multiple online experiments. Each experiment used different visual tasks, including color discrimination, memory tests, and dot counting. Despite these variations, the core findings remained consistent across most experiments.

The computational model represented a significant advance in understanding how people perceive their own thinking. While previous models had examined how confidence accuracy improves with time, this study was the first to model how different psychological factors distort confidence as it forms. Model fits closely matched observed data, successfully capturing the opposite relationships between time and confidence for anxiety versus gender.

In exploratory analyses, the researchers also examined how age and compulsive symptoms affected confidence dynamics. Older participants showed increasing underconfidence over time compared to younger ones. People with compulsive symptoms displayed the opposite pattern of anxiety, becoming overconfident with additional reflection time. These findings point toward confidence distortions during post-decision processing as a general feature of individual differences in mental health.

Understanding these mechanisms opens new possibilities for interventions. Rather than generic advice to “be more confident,” people could receive targeted guidance based on the source of their underconfidence. Anxiety might respond to strategies that prevent negative rumination, while socially learned biases might require structured reflection on actual performance. Confidence is not simply high or low, but dynamically constructed through mental processes that unfold over seconds.

---