Using ratings of appropriateness derived from an expert physician panel, we measured how appropriately physicians in 1981 performed coronary angiography in a randomly selected, community-based sample of cases in the Medicare population. We studied large geographic areas (three sites) in three states, representing regions of high and low use. The high-use site had fewer procedures classified as appropriate (72%) than either low-use site (77% and 81%, respectively). Over all sites, 17% of procedures were classified as inappropriate. Patients in the high-use site were older, had less severe angina, and were less intensively medically treated than patients in either of the low-use sites. Patients without angina who had not undergone exercise testing constituted the most common subgroup of inappropriate cases. Although overall differences in appropriateness were not large, practice differences do exist. This analysis of practice differences among study sites provides the clinical basis for understanding the small, but significant, differences in the appropriateness of use of coronary angiography. The finding of 17% inappropriate use may be cause for concern.