The length of survival and various risk factors were studied utilizing 408 residents (141 men, 267 women) of a large residential home. The subjects, aged 68 years or more at entry, admitted between 1978 and 1983, were physically independent, continent, and non-diabetic. They were followed until December 31, 1988, by which time 78% had died. The multivariate proportional hazard analysis showed the following entry variables to have the indicated effects on relative mortality rate ratios: 5-years' higher age (+10%, NS), persistent bacteriuria (+13%, NS), abnormal ECG (+26%, NS), current smoking (+63%, P less than 0.01), mildly impaired mobility (+96%, P less than 0.001), higher levels of in-study systolic pressure in 10-mm Hg steps (-4%, NS), higher entry serum cholesterol in 1-mmol/L steps (-7%, NS), and higher hematocrit in 5% steps (-14%, P less than 0.02). Female sex was associated with a +25% (NS) rate ratio; socioeconomic status and body weight were without effect. These data highlight the relative importance of specific factors associated with survival of persons within retirement homes and indicate that: (1) the presence of mild impairment of mobility at entry is by far the strongest predictor of early death; (2) smoking and lower hematocrit also exert important adverse effects; and (3) certain "risk factors", ie elevated systolic blood pressure and serum cholesterol, have a minimal protective effect, if any, in this age group.