In an attempt to develop a measure of the referral behaviour of general practitioners, 21 simulated case histories were constructed which presented between three and six stages at which a decision was required whether or not refer a 'patient' to hospital. Twenty general practitioners completed the case histories and their responses were compared with their referral rates to outpatient departments. No significant correlations were found between the doctors' responses to the vignettes and their actual referral rates, and the repeatability of the vignettes was disappointing. The fact that the case histories appeared realistic to the doctors who completed them was not reflected in objective measures of their validity. Simulated case histories should not be used as a method of measuring doctors' behaviour without establishing their validity and reliability.