The undergraduate medical curriculum at Linköping was changed and problem-based learning introduced in 1986. As a consequence the need for new examinations arose. To assess the students' consultation skills, ability to use theory in practice and skills in problem-based learning, a new examination called 'phase examination' was designed and introduced. The examination results of a 4-year student cohort are presented here, including descriptive statistics and correlations between phase examinations 1 and 2, as well as a semester examination consisting of modified essay questions and an objective structured clinical examination. The construct validity of the phase examination was tested by factor analysis. In general, low correlation between the three examinations were found, and interpreted as a result of inter-case and inter-rater variability. The factor analysis identified two factors in both phase examinations--one representing the solving of a learning task, the other consultation skills. The variables concerning the use of knowledge had an intermediate position, more connected to the solving factor in phase 1 and the consultation factor in phase 2. With the close connection of theory and practice in the phase examination, both by its design and the pairing of examiners (one general practitioner and one basic scientist), these examinations can be an important learning experience for both students and teachers. The deficiencies in the reliability of the phase examinations is, in our view, compensated by its directing and supporting effect on students' learning and its face and construct validity.