The objective of this study was to derive and test clinical scoring system in order to predict the presence of gallstones and to identify inappropriate referrals. The design of this scoring system comprised a prospective audit of all referrals with suspected gallstones and a prospective validation of the system on new patients. The subjects used in the study consisted of 300 inpatient, outpatient and general practitioner referrals to a radiology department with suspected gallstones during the period 1984-1986 (1st cohort) and 200 similar referrals from 1987 to 1988 (2nd cohort). The main outcome measures were gallstones as determined by oral cholecystogram and/or ultrasound of the gallbladder. 32.7% and 24.0% of referrals in the first and second cohorts, respectively, had gallstones. A scoring system to predict gallstones was derived on the first cohort based on age, Murphy's sign, duration and type of pain, presence or absence of flatulence, and source of referral. This scoring system discriminated well between those with and without stones in both cohorts. Low risk groups, comprising about 15% of the referrals in each cohort, were identified with a prevalence of gallstones of 4.5% (1st cohort) and 3.0% (2nd cohort). If these patients had not been referred the radiological workload would have dropped by 15%, with a very small reduction in diagnostic yield. We conclude that prior to radiological assessment it is possible, on the basis of source of referral and symptoms, to identify a group of patients with a low prevalence of gallstones. These cases, with a low diagnostic yield, contribute substantially to the workload of a radiological department.