The effect of physical activity on all-cause mortality was examined using a representative random sample of 1,536 persons (15 years and older) in three communities in Upper Bavaria. 27.0% of the respondents reported regular and 26.2% occasional physical exercise. During the 5-year follow-up 5.1% (n = 79) of the original sample died. Using a logistic regression model, the relation between physical activity and mortality was measured by the odds ratio, with subjects reporting regular physical activity as the reference group. Crude mortality risk was significantly higher among the physically inactive (men: 3.97; women: 4.36) but not among respondents practising occasionally (men: 1.67; women: 1.24). After adjustment for potential confounding variables (age, social class, physical and mental health), the mortality risk was elevated but not statistically significant for the physically inactive (men: 1.76; women: 1.51) and for the group practising occasionally (men: 1.50; women: 1.14).