The interpretations of 314 thoracic aortograms obtained over a 4 1/2-year period for possible traumatic rupture were reviewed to determine the frequency and causes of false-positive examinations. The radiographs and clinical records of all patients whose aortograms were abnormal or equivocal were examined. Two hundred consecutive thoracic aortograms that had been interpreted as normal were reviewed to determine the frequency and character of normal morphologic variations at the aortic isthmus, with particular attention to ductus diverticula, in an attempt to identify means of reducing the number of examinations that were equivocal or false-positive for aortic rupture. Aortic rupture was diagnosed with confidence in 12 patients (4%), and all were surgically proved. Another nine aortograms (3%) were equivocal when standard aortographic criteria for rupture were used. For four of the nine, further imaging studies indicated normal aortas, but five patients underwent thoracotomy. Two of these had aortic rupture, two had ductus diverticula, and one had an ulcerated plaque at the site of aortographic abnormality. Thus, false-positive aortograms led to unnecessary surgery in three cases--1% of all 314 aortograms and 14% of the 21 aortograms that were not clearly normal. In summary, two (14%) of 14 aortic ruptures in our series could not be distinguished at aortography from anatomic variants or disease not requiring surgery. To avoid false-negative diagnosis, we must accept occasional false-positive diagnoses, necessitating exploratory thoracotomy with its attendant risks.