The spontaneous autoimmune diabetes of the nonobese diabetic mouse and the BioBreeding diabetes-prone rat bears a striking similarity to human type 1 diabetes. For this reason it comes as a disappointment that interventions known to prevent the disease in the animal models proved ineffective in human trials. For an explanation I propose that, although the mechanism of beta-cell destruction is the same in the three species, this common mechanism may be the end point of different loss-of-tolerance pathways in different human patients, only two of which are represented by the two animal models. Research strategies to address this problem will require more attention to immune subphenotypes or their corresponding subgenotypes in the human and, possibly, the generation of additional mouse models of spontaneous disease.