International comparisons show strong inverse protective associations with starch, NSP (fibre, non-starch polysaccharides) and vegetable intakes, and positive associations with meat consumption in large bowel cancer. Estimates of relative risk from cohort investigations are in the same direction although generally weak, and red and processed meat, rather than white meat seem to be associated with elevated risk. A protective effect of starch and NSP probably arises from their marked effect on bacterial metabolism in the large bowel, which can be postulated to affect gene expression and DNA repair via increased butyrate production. Stool weight is also increased and pH reduced leading to alterations in secondary bile acid production, and mucosal cell proliferation. In rodent models, 'insoluble' sources of NSP are generally protective, although butyrate, resistant starch and soluble NSP may not reduce tumorigenesis. High levels of meat increase faecal ammonia and N-nitrosocompound (NOC) concentration. Some of the chromosomal mutations found in human colorectal cancer are consistent with effects of NOC and heterocyclic amines. More data are required from human experimental studies linking alterations in diet with known stages in carcinogenesis in the large bowel, and from large cohort studies which have collected biological samples in order that interaction between diet, biomarkers of diet, and different genotypes that may determine risk can be examined.