A computerized search of 276 randomly selected cardiology abstracts published in conjunction with three national meetings revealed that 137 abstracts (49.6 per cent) led to full-length articles in peer-reviewed journals within a period of 37 to 43 months. The median interval between publication of the abstract and of the article was 14 months. Full-length articles were derived from 5. per cent of the abstracts presented at the meetings of the American Federation for Clinical Research, the American Society for Clinical Investigation, and the Association of American Physicians ("clinical societies"); 55 per cent of the abstracts presented at the American Heart Association Scientific Sessions; 49 per cent of the abstracts presented at the American College of Cardiology Scientific Sessions; and 35 per cent of the abstracts submitted to the clinical societies but not presented. During the observation period, each abstract was cited a mean of 1.6 times in a journal reference list; the number of citations was not significantly higher for abstracts that led to articles than for abstracts that did not. These findings emphasize the preliminary nature of abstracts and suggest restraint in their citation as references.