We studied the incidence of reversibility of coronary obstructions in a consecutive series of 313 patients with nonoperated coronary artery disease catheterized twice 3 to 118 (mean 38) months apart. Recanalization was observed in three patients and regression from an initial less than 100% obstruction in six patients. Progression in a different location occurred in six of the nine patients who demonstrated one recanalized or one regressive lesion. We conclude that true regression is an infrequent event in the natural history of medically treated patients with coronary artery disease; moreover, the pathophysiology and clinical relevance of angiographic regression remain poorly defined.