(Bartram & Wynder, 1989), and obesity appears at most to be weakly related to colorectal cancer (Willett, 1989a). Interpretation of the weak association with obesity is complicated by the fact that weight loss may be a sign of the disease. To obviate the possible direct effect of the tumour on a variety of possible risk factors, studies of adenomatous polyps have been advocated. Physical activity has been considered only in two previous studies, in one of which no association was found (Stemmermann et al., 1988) and in the other an inverse association (Kono et al., 1991). The studies of the relationship between adenomas and energy intake (Hoff et al., 1986;Macquart-Moulin et al., 1987;Stemmermann et al., 1988) and body size (Mannes et al., 1986;Sandler et al., 1988;Stemmermann et al., 1988;Kono et al., 1991) (Little et al., 1993).During the interview, a complete occupational history was recorded. This information was classified according to the Registrar General's classification of occupations (Office of Population Censuses and Surveys, 1970). These data were used to derive measures of socio-economic status based on the subject's current job at interview, their last job before retirement, the job they held for the longest period (if this was 'housewife' or 'househusband', that held for the second longest period was considered) and the job of the highest class.Subjects were asked to state, for the year prior to interview, how much of the day they spent sitting, standing, walking or in heavy work (with categories none, less than half, about half or more and practically all) and how often in this period they took the following forms of exercise -sport/ keep-fit, hard labour such as heavy gardening, housework, brisk walking for half an hour continuously, running or cycling for half an hour continuously and other exercise (with the categories of none, less than once a week, once a week, twice a week or more).Repeatability Thirty-four repeat interviews were completed. The correlation coefficient for energy intake between the two interviews was 0.57. Regarding physical activity, the agreement was better for variables relating to the frequency of undertaking different specific activities (kappa 0.64-0.76) than for the proportion of the day spent in different activities (kappa 0.34-0.48 The method of calculating energy intake has been described in the accompanying paper (Little et al., 1993). In addition to considering associations with each specific component of physical activity in the year prior to interview, summary scores were also derived. Scores for daily activity level in the year prior to interview were calculated as the product of a value for the proportion of the day spent on the activity times the intensity of the activity. The values for the proportion of the day spent on the activity were taken as 1 for less than half, 2 for about half or more and 3 for