In an examination designed by a medical specialty board, six tests were administered to groups of candidates for certification to investigate the feasibility of using videotapes of doctor-patient interactions to assess candidates' ability to note signs of potential physical disorders during a patient's physical examination and history taking, and their ability to draw reasonable conclusions and make appropriate decisions based on the available evidence. Common-item links, derived from the questions associated with selected interactions, were used to collect the data needed to equate the tests, and the Rasch model was the basis for scoring the tests and analyzing the equating data. Two indices were derived to help identify triads and links that lacked coherence. Investigation of the distributions of these indices is suggested so that firm guidelines for identifying triads and links that lack coherence can be established. Studies also should explore the relationship between coherent and incoherent triads and links, with the aim of clarifying the relationship and explaining the apparent contradiction detected in these data in which apparently coherent triads contained one or more seemingly incoherent links. (Author/MH)