Experiments in rodent models of insulin-dependent diabetes mellitus (IDDM) suggest that destruction of pancreatic beta cells can be both initiated and inhibited by certain environmental factors such as dietary constituents. We studied nutritional impact of certain protein sources of natural-ingredient, non-purified (NP) rodent diet on diabetes incidence and insulitis severity in the spontaneous diabetic, non-obese diabetic (NOD) mouse. Long-term ad lib. feeding of diets containing wheat flour (800 g/kg), and to a lesser extent soya-bean meal (400 g/kg), were associated with relatively high diabetes incidence (65 and 45% respectively), whereas a diet based on hydrolysed casein (HC; 200 g/kg) as the only source of protein significantly (compared with the wheat-flour diet) inhibited expression of diabetes (22%). Feeding a hypo-allergenic soya-bean-protein hydrolysate resulted in diabetes incidence and insulitis severity similar to that of the soya-bean-meal-fed group. This may indicate that protein hydrolysis per se may not be necessary for dietary modification of diabetes in the NOD mouse. The window of vulnerability to diabetogenic diets was found to be between weaning and about 70 d of age. In the diabetic mice insulitis was less frequent in the HC-fed group when compared with those fed NP (P = 0.04), soybean meal (P = 0.03), soya-bean-protein hydrolysate (P = 0.012) or wheat flour (P = 0.0002). In the non-diabetic mice the wheat-flour diet was associated with a high insulitis severity in comparison with the HC group (P = 0.004). Early avoidance of NP diet was associated with lower degree of insulitis in both diabetic (P = 0.00003) and non-diabetic mice (P = 0.001) when compared with the mice fed on the HC diet later in life. These findings are contributing to further clarification of diabetes-promoting dietary constituents, which may have some nutritional implications for IDDM-susceptible children.