Twenty seven gastroenterologists (15 physicians and 12 surgeons) were studied to capture their policy in the usage and weighting of cues in making a diagnosis. This was a relatively small sample and on account of possible sampling bias only tentative generalisations will be made. Five case vignettes, each consisting of four to eight cues, were used. Subjects were asked to give their percentage likelihood estimates of various diagnostic possibilities after each cue. This gave an indication of their perception of the significance of various cues in relationship to each diagnosis. There were marked variations in cue weighting by these experts, and in particular, most individuals were far off the mean regarding the weighting of certain clinical features, thus displaying idiosyncratic behaviour in these instances. As may be expected, there were differences in disease prevalence estimates between physicians and surgeons. Early information had an overwhelming effect on the final diagnosis. In most cases the expert relied on a few critical cues rather than on a pattern to make a diagnosis. In view of the strong influences of early diagnostic formulations, these findings confirm the need for doctors to learn to use and collect accurate factual information on prevalence rates and on the most significant critical cues for various disease processes. The study highlights some of the problems faced by novices in learning from experts who may teach them contradictory information about what are the most significant factors in coming to a diagnosis.