The experience after implantation of 3,334 Björk-Shiley valves over a 15 year period is described. With a 99.2% follow-up (covering 17,511 patient-years, mean follow-up time 6.3 years) and an autopsy rate of 75% among all fatalities, altogether 19 cases of mechanical failure were documented. There were no mechanical failures among the standard Delrin Björk-Shiley valve (n = 271), the aortic standard Pyrolyte Björk-Shiley (n = 739), or the Monostrut Björk-Shiley valve (n = 377). One of the mitral standard Pyrolyte valves (n = 430) fractured. Among the 1,461 convexo-concave valves, 18 fractured (6/884 with an opening angle of 60 degrees and 12/577 with an opening angle of 70 degrees). The actuarial incidence of mechanical failure at 5 years was 0.6% (with an upper 95% confidence limit of 1.2%) for the 60 degree convexo-concave valve and 2.8% (upper 95% confidence limit of 4.4%) for the 70 degree convexo-concave valve (p less than 0.01). Two groups of valves were especially affected by this complication; the 23 mm aortic 60 degree convexo-concave valve (5 year actuarial incidence 2.2%, upper 95% confidence limit 4.7%) and the 29 to 31 mm mitral 70 degree convexo-concave valve (8.3%, upper 95% confidence limit 14.2%). The hazard function presently indicates a constant (60 degree convexo-concave) or decreasing (70 degree convexo-concave) tendency for mechanical failure. The time interval between the first symptom of mechanical failure and circulatory collapse was significantly (p less than 0.01) shorter after aortic failure than after mitral failure, and no patient with a fractured aortic prosthesis survived long enough to undergo reoperation. The incidence of mechanical failure among patients dying suddenly (but with an autopsy) was 9.6% (95% confidence limits 4.9%-16.6%), and most cases of sudden death were unrelated to the prosthesis. The management of patients with suspected mechanical failure is described. Prophylactic re-replacements are discussed but cannot be generally recommended at present.