Context: Ber's Comprehensive Integrative Puzzle aims to assess analytical clinical thinking in medical students. We developed a paediatric version, the MATCH test, in which we added two irrelevant options to each question in order to reduce guessing behaviour. We tested its construct validity and studied the development of integrative skills over time. Methods: We administered a test (MATCH 1) to subjects from two universities, both with a 6-year medical training course. Subjects included 30 students from university 1 who had completed a paediatric clerkship in Year 4, 23 students from university 2 who had completed a paediatric clerkship in Year 5, 13 students from both universities who had completed an advanced paediatric clerkship in Year 6, 28 paediatric residents and 17 paediatricians. We repeated this procedure using a second test with different domains in a new, comparable group of subjects (MATCH 2). Results: Mean MATCH 1 scores for the respective groups were: Year 4 students: 61.2% (standard deviation [SD] 1.3); Year 5 students: 71.3% (SD 1.6); Year 6 students: 76.2% (SD 1.5); paediatric residents: 88.5% (SD 0.7), and paediatricians: 92.2% (SD 1.1) (one-way anova F = 104.00, P < 0.0001). Students of both universities had comparable scores. MATCH 1 and 2 scores were comparable. Cronbach's α-values in MATCH 1 and 2 were 0.92 and 0.91, respectively, for all subjects, and 0.82 and 0.87, respectively, for all students. Conclusions: Analytical clinical thinking develops over time, independently of the factual content of the course. This implies that shortened medical training programmes could produce less skilled graduates.