The possibility that people who differ with respect to the A, B, and O blood groups might differ in their susceptibility to such diseases as carcinoma of the stomach, duodenal and gastric ulcer, pernicious anemia, and diabetes mellitus has been much studied of late. A critical analysis of the information so far collected shows, however, that much of it cannot be accepted without serious reservations and that some of it is probably worthless. Since blood groups are inherited, their distribution in small communities is sometimes determined by the predominance of certain family or ethnic units. It is extremely difficult to find a series of healthy people that will be truly comparable to a series of patients with a given disease. The statistical methods used can yield valid results only when sound methodology provides reliable data. Failure to observe this axiomatic principle explains much of the present confusion in research on blood groups and disease.