Intakes of starch, non-starch polysaccharides (NSPs), protein and fat have been compared with colorectal cancer incidence in 12 populations worldwide. There were strong inverse associations between starch consumption and large bowel cancer incidence (large bowel r = -0.70, colon r = -0.76). There was no significant relation with NSPs, although the association with large bowel cancer incidence was still significant when NSP was combined with resistant starch (RS) to give an estimate of fermentable carbohydrate (large bowel r = -0.52, colon r = -0.60). The relationships between starch, RS and NSPs and cancer incidence remained statistically significant after adjusting for fat and protein intakes. The strong inverse associations found here suggest a potentially important role for starch in protection against colorectal cancer and correspond with the hypothesis that fermentation in the colon is the mechanism for preventing colorectal cancer. Measures of both starch and NSPs need to be included in future epidemiological studies of diet and bowel cancer.