Level of knowledge was tested in 137 volunteer registered nurses in three hospitals in the Midwest. A 34-item multiple choice Diabetes Knowledge Test (DKT) was developed by the first author (S.D.S.) working with a panel of expert nurse diabetes educators. Reliability using coefficient alpha was 0.61. Staff nurses obtained a meanvscore of 25.3, or 74% correct. Thirty-four nurses (25%) scored above 80% correct; 62 (45%) scored between 70% and 79% correct; and 41 (30%) scored below 70% correct. Higher-scoring nurses did well on items requiring use of exchange lists, but they missed items on physiologic actions of insulin, side effects of sulfonylureas, where insulin should be stored, and areas of the body suitable for injections of insulin. Nurses who scored below 60% on the test consistently missed items requiring use of exchange lists, effects of regular exercise and illness on blood glucose, and symptoms and causes of hypoglycemia and hyperglycemia. A strong trend (P = 0.055) was found for surgical nurses to score lower than medical unit nurses, and significant differences (P < 0.05) were found among scores between hospitals. Staff nurses employed in a hospital where clinical specialists also were employed to teach diabetic patients had the lowest scores of the three hospitals. Test content subareas should be developed in the future to test knowledge reliably about urine testing, use of oral medications, home blood glucose monitoring, effects of exercise, and Somogyi phenomenon. A curriculum for regular inservice should be developed for nurses to assure a high level of knowledge for patient teaching with diabetes patients.