Anxious-depressed individuals underestimate themselves even when they’re right

New research published in Nature Communications sheds light on why individuals with anxiety and depression often remain underconfident in their abilities—even when their actual performance is intact. The study found that people with more anxious-depressive symptoms are less responsive to moments when they feel confident, which may prevent them from building a more accurate and positive overall belief in their capabilities. Interestingly, these individuals still responded normally to external feedback, suggesting that the underconfidence stems from how they integrate their own confidence experiences.

The researchers conducted the study to better understand how people with anxious-depression symptoms form beliefs about their own abilities—a process known as metacognition. Although earlier studies had shown a link between depression, anxiety, and underconfidence, they had not clarified the mechanisms behind this bias. The key question was whether these individuals misinterpret their moment-to-moment confidence, ignore successes, or overweight negative feedback.

“Previous work has shown that people with subclinical anxiety and depression symptoms can have unreasonably low self-beliefs in their abilities despite being able to perform as well as others. We wanted to study how such self-beliefs are maintained despite evidence to the contrary, as this can help provide insight into how we can potentially mitigate such low self-beliefs,” said study author Sucharit Katyal, a postdoctoral researcher at the University of Copenhagen.

To investigate, the researchers ran two large experiments with a total of over 500 participants recruited online. They measured two types of confidence: “local” confidence, or how sure participants were in their decisions on individual trials, and “global” confidence, which referred to how well they thought they were performing across a block of trials. They also manipulated feedback during tasks to test how participants updated their global confidence based on either internal cues (local confidence) or external evaluations (feedback).

Participants completed perception and memory tasks in a gamified setting. In the perception task, they judged which type of fruit appeared more often on the screen. In the memory task, they identified which fruit had been in a previous array. After each decision, they rated their confidence in that choice. Occasionally, they received feedback from a fictional “auditor” about whether their choice had been correct. At the end of each block, participants estimated how many answers they thought they had gotten right—this was the global self-performance estimate.

Free daily newsletter

The researchers designed feedback blocks where correct responses were more or less likely to be followed by positive feedback, allowing them to manipulate participants’ perceived success independently of actual performance. This setup mimicked real-life scenarios where a person may receive more encouragement or criticism, regardless of how well they are doing.

Both experiments showed that feedback had a strong influence on global confidence. Participants who received mostly positive feedback in a block felt more confident about their performance, even though their accuracy had not changed. Those who received mostly negative feedback felt less confident. This demonstrated that global confidence could be shifted without altering task performance. Interestingly, these effects even carried over to new blocks without feedback and sometimes across different tasks, suggesting that feedback had a lasting influence.

But when the researchers looked at how local confidence influenced global confidence, a striking pattern emerged. People with higher anxious-depression symptoms were less likely to let moments of high confidence improve their overall self-assessment. While they still reported how confident they felt on individual trials, they didn’t seem to take that into account when judging how well they had performed overall. This blunted sensitivity to high confidence seemed to maintain their global underconfidence, even when their performance was good.

To better understand this pattern, the researchers used computational modeling. They tested different models that could explain how people with anxious-depression symptoms update their beliefs about themselves. One model suggested they might ignore their own local confidence; another suggested they might respond more strongly to negative feedback; a third proposed a general bias to rate themselves lower regardless of experience.

The results pointed clearly to the first model. The best-fitting explanation was that people with anxious-depression symptoms failed to incorporate their high-confidence experiences into their broader self-beliefs. They were not more affected by negative feedback than others, nor did they show a general tendency to underestimate themselves across the board. The problem appeared specific to the internal learning process—how they used their own confidence signals to build a bigger picture of their abilities.

“We had expected that people with low self-beliefs would be disproportionately sensitive to negative feedback, but we did not find this to be the case,” Katyal told PsyPost. “Instead, they were more sensitive to only their own past low confidence when they did the task.”

To test whether these effects extended beyond the task itself, the researchers also looked at participants’ self-beliefs in a separate task. They asked participants to rate how well a set of positive and negative adjectives described them. Those who had received mostly negative feedback during the earlier tasks were more likely to endorse negative self-descriptions afterward. This suggests that the effects of skewed feedback and internal confidence processing may shape how people view themselves more broadly, not just within a specific task.

“We also found that receiving negative feedback, even on an online gamified type task, can hamper people’s more global self-esteem – not just related to cognitive task performance but how they feel about themselves more generally,” Katyal said.

The findings help explain why people with anxiety and depression often hold negative views of their competence and may avoid new challenges—even when they have the skills to succeed. Their global underconfidence may persist not because they perform poorly or receive too much criticism, but because they do not fully recognize or internalize the moments when they feel capable.

The study also suggests that feedback interventions could help. Because people with anxious-depressive symptoms responded normally to external feedback, well-calibrated positive reinforcement might help shift their global confidence in a more accurate direction. The effects of feedback were strong, extended across domains, and even influenced broader self-beliefs, making it a promising tool for future interventions.

“Some people hold unreasonably low self-beliefs in their abilities despite performing well, sometimes known commonly as the ‘imposter syndrome.’ Here, we find that such low self-beliefs are maintained by underconfident people due to reduced sensitivity to past situations when they had high confidence on performing tasks of those abilities and higher sensitivity to situations when they had low confidence,” Katyal told PsyPost.

“However, underconfident people were not differentially sensitive to past instances of positive and negative feedback about their abilities. This implies that it can be helpful for people with low self-beliefs to rely on external feedback from peers who can give them accurate assessment of their abilities and not rely solely on their internal self-assessments.”

But the study has limitations. Although the sample included individuals with elevated symptoms, it did not include people with formal clinical diagnoses. Future research is needed to confirm whether these findings apply to clinical populations. Additionally, while feedback helped shift confidence in the short term, it is unclear how long these changes last or how best to sustain them. More work is needed to explore how feedback-based interventions could be designed to produce lasting improvements in self-beliefs and mental health.

“The study was conducted on large online participant samples, so it needs to be replicated in other populations to be globally applicable,” Katyal said.

The study, “Distorted learning from local metacognition supports transdiagnostic underconfidence,” was authored by Sucharit Katyal, Quentin JM Huys, Raymond J. Dolan, and Stephen M. Fleming.