The relationship between self-reported height and cancer risk was investigated in an integrated series of case-control studies including 273 cases of cancer of the oesophagus, 474 of the stomach, 558 of the colon, 352 of the rectum, 227 of the liver, 267 of the pancreas, 110 of the larynx, 2,388 of the breast, 545 of the cervix uteri, 563 of the endometrium, 688 of the ovary, 80 of the prostate, 263 of the bladder, 105 of the kidney, 43 Hodgkin's disease, 152 non-Hodgkin's lymphomas, 109 multiple myelomas, and a total of 5,574 controls admitted to hospital for acute, non-neoplastic conditions. No significant positive trend in risk was observed for any of the cancer sites considered, and some suggestion of elevated risks for the upper quintile of height was observed only for prostate (relative risk, RR = 1.4), kidney (RR = 2.7) and colon (RR = 1.2) in males (but not in females). For breast cancer, all the RRs for subsequent quintiles of height were close to unity. Four neoplasms showed significant inverse trends with height: oesophagus (RR = 0.7 for highest vs. lowest quintile), cervix (RR = 0.4), endometrium (RR = 0.7) and ovary (RR = 0.6). For oesophagus and cervix the trends could be explained, at least in part, in terms of social class correlates (multivariate RR 0.8 and 0.5, respectively), while for endometrium they could possibly be related to an inverse correlation between height and body mass (multivariate RR 0.9). Thus, our study did not support the hypothesis that nutrition in childhood and adolescence (which in this population is a determinant of adult height) is directly related to the subsequent risk of cancer at several major cancer sites. A number of inverse associations emerged, which may be either spurious and incidental, or suggest that poorer nutrition early in life may be an unfavourable indicator of the subsequent risk of selected neoplasms.