The results are presented of a diabetic survey in Zagreb where, in a typical population sample from five General Practice Units, 88.9% of the total number of inhabitants aged over 15 (9176) were examined. The examination in the screening phase comprised the determination of blood sugar 1–2 h after the largest daily meal. — The results are analysed from technical, sociomedical, and epidemiological aspects. 1. For standardized epidemiological comparisons, blood sugar measurements after glucose loading should be adapted and so arranged that the validity of the screening phase may be assessed. 2. All blood sugar distributions proved unimodal, continuous, and positively skewed. Consequently, all critical diagnostic values are necessarily arbitrary and of greater importance to group than individual diagnosis. The OGTT results in so-called borderline cases proved unreliable in the course of time. Classifications based upon single determinations are inadequate. 3. In the epidemiological studies of diabetes, a quantitative approach to the analysis of relationship between the continuously varying blood sugar level and other epidemiological characteristics of the population, has been neglected. The use of this approach in the Zagreb Diabetes Survey promises further refinements in the understanding of the importance of glucose tolerance in complete populations.