Introduction: Our study aims at assessing the validity of summative criteria-referenced self-assessment in higher education, and in particular, if that validity varies when the professor counts self-assessment toward the final mark. Method: One hundred and twenty-two first year students from two groups in Teacher Education at the Universidad de Cadiz (Spain) participated in the study, each group following a different course in the same six-month period. The relevant difference between the two courses was that, in one of them, self-assessment was included among the assessment methods and counted for 5 per cent of the final mark. The professor was the only marker in the other course. Once the courses finished, participants completed a self-assessment questionnaire that included competences, learning results and contents, and were asked to give themselves a mark. Self-assessment data were compared with the marks given by the professor, and the 10 students with the highest discrepancies were interviewed. Results: In both groups, the students' self-assessments were higher than the marks given by the professor, with significantly higher differences in the students with poorer results and in the group in which self-assessment counted towards the final mark. In this group, no relationship was found between the professor's and the students' assessments. Discussion: When self-assessment does not count towards the final mark, students' and professor's assessments tend to be highly similar; when self-assessment counts towards the final mark, over and underestimations increase dramatically. The main reasons that we found for this are the desire to obtain the highest possible grades and the stress associated with self-assessment. Implications of implementing self-assessment in higher education are discussed. (Contains 3 tables.)