A total of 101 reparative and 389 valve replacement operations, isolated or combined with tricuspid annuloplasty or operations for coronary artery disease, were done for mitral incompetence (1975 to July 1, 1983). The patients undergoing repair as a group were younger and had less hemodynamic and functional derangement than those undergoing replacement. The prevalence of repair was less (p less than 0.001) for two surgeons than for the other four, even when possible differences in patient populations were taken into account by multivariate analysis. Five-year survival rate, including hospital deaths, was 76% after valve repair and 56% after valve replacement (p = 0.005). However, by multivariate analysis, valve replacement rather than repair was only possibly (p = 0.14) a risk factor. (Multivariate analysis in all patients undergoing mitral valve repair in the period 1967 to 1985 [n = 210] did not find the type of annuloplasty to be a risk factor.) The incidence of reoperation was no different after repair or replacement and there was no increase in the risk of reoperation late after repair. Endocarditis early or late after operation occurred in 11 of the 389 patients undergoing mitral replacement and in none of those undergoing repair (p = 0.08). The functional status of the patients was not different between the two groups. These data, and the experience of others, indicate the advantages of repairing rather than replacing the incompetent mitral valve whenever possible.