In a diabetes survey 630 persons who had glycosuria or were age- and sex-matched were given a 50-g. oral glucose tolerance test (G.T.T.). Five years later the test was repeated in 527 (83·6%) cases. The original G.T.T.s were classed as normal or showing “G.T.T. diabetes,” “lag storage,” “renal glycosuria,” or “miscellaneous abnormalities.” Compared with the normal group in the first test those with G.T.T. diabetes showed a higher rate of conversion to “florid diabetes,” but many others with G.T.T. diabetes remained unchanged or became normal. The more abnormal the first test the more likely it was to progress to florid diabetes. Most lag storage curves remained unchanged or became normal, but a significant number converted to a diabetic abnormality. Renal glycosuria was an innocent peculiarity compared with other categories. The various “miscellaneous abnormalities” tended to change, and showed an excessive conversion to diabetes. The best separation between non-diabetic and diabetic at the follow-up test occurred at 135 mg./100 ml. Those in the diabetic categories at the original test showed a mortality slightly above that in the general population, comparable to that found in diabetics in general. New cases of florid diabetes appeared in the population which screened negative for glycosuria at the rate of 0·3 per 1,000 per annum.