Depression is believed to hinder one’s ability to reason about oneself (metacognition). This impairment can arise from dysfunctional biases and/or learning processes. However, the relationship between depression, biases and learning in metacognition is not known. Here we combined multi-trial behavioural experiments with computa- tional modelling to explicitly test whether depression impacts biases and/or learning in a metacognitive task. First, using a perceptual estimation task with fixed feedback valence (N=131), we show that depressive symptoms predict negative metacognitive biases but do not impact learning. Next, we tested the generality of our results in a more realistic perceptual estimation task where we varied the valence of the feedback. Using a Rescorla-Wagner model of confidence fitted to behavioural data (N=74), we show that also in this task, depressive symptoms predict negative metacognitive biases but do not impair learning. Overall, our study suggests that depression impacts metacognitive states but not one’s ability to learn while offering a behavioural-computational framework for the study of metacog- nition in depression.