The aim of the research is to study the capacity for self-evaluation of University students undergoing tests involving mathematics, linguistic and formal reasoning. Subjects were asked to estimate the number of correct answers and subsequently to compare their performance with that of their peers. We divided the subjects into three groups on the basis of performance: poor, middle and top performers. The results demonstrate that all the subjects in all tests showed good awareness of their level of actual performance. Analyzing comparative assessments, the results reported in literature by Kruger and Dunning were confirmed: poor performers tend to significantly overestimate their own performance whilst top performers tend to underestimate it. This can be interpreted as a demonstration that the accuracy of comparative self-evaluations depends on a number of variables: cognitive and metacognitive factors and aspects associated with self-representation. Our conclusion is that cognitive and metacognitive processes work as "submerged" in highly subjective representations, allowing dynamics related to safeguarding the image one has of oneself to play a role.