In an effort to develop referral criteria for the ordering of abdominal radiographs for patients presenting with abdominal symptoms, we prospectively studied the relation between clinical data and radiographic abnormalities. Of 1780 examinations, 179 (10.0%) showed some radiographic abnormality. If abdominal radiographs would have been limited to those patients who had moderate or severe abdominal tenderness, or to patients with a high clinical suspicion of bowel obstruction, renal or ureteral calculi, trauma, ischemia, or gallbladder disease, regardless of the degree of tenderness, 956 (53.7%) examinations would not have been done. All radiographic abnormalities reflecting a serious pathologic process would have been identified. Only 33 (3.5) abnormalities of limited significance, almost all localized or generalized ileus, would have been undetected. The adoption of these referral criteria would result in minimal loss of clinicall useful information, large financial savings, and a reduction in radiation exposure.