When we feel confident in a perceptual decision, that decision is more likely to be correct. How does the brain assess the quality of perceptual decisions? Here, we test two connected hypotheses. First, that perceptual confidence relates to the structure of population activity in the sensory cortex. And second, that this relation differs from the one between sensory activity and perceptual decisions. To this end, we studied neural population activity in the primary visual cortex of macaque monkeys performing a task that reveals the animal's subjective sense of perceptual confidence. Specifically, the animal judged the orientation of ambiguous stimuli and simultaneously reported their confidence in this decision. Choices were rewarded in such a way that the most profitable strategy required the animal to take into account the quality of perceptual decisions. Confidence should be informed by the factors which make perceptual decisions more or less reliable. These factors include the strength of evidence for a particular choice derived from a perceptual impression, and the uncertainty of this perceptual impression. To gain insight into the neural computations that implement this introspection process, we recorded population activity in the primary visual cortex with multilaminar electrode arrays and used nonlinear decoders to expose the relationship between this activity and the choice behavior. We found that both perceptual decisions and confidence can be predicted from V1 population activity. Our analysis validated both hypotheses and suggests that perceptual decisions arise from a computation that evaluates the most likely interpretation of a stimulus, while confidence instead reflects a computation that evaluates the quality of the sensory evidence that informed the decision. This work establishes a direct link between neural population activity in the sensory cortex and confidence in perceptually guided decisions.