Without a doubt, rodent models have been instrumental in describing pathways that lead to pancreatic beta-cell destruction, evaluating potential causes of type 1 diabetes and providing proof-of-principle for the potential of immune-based interventions. However, despite more than two decades of productive research, we are still yet to define an initiating autoantigen for the human disease, to determine the precise mechanisms of beta-cell destruction in humans and to design interventions that prevent or cure type 1 diabetes. In this Perspective article, we propose that a major philosophical change would benefit this field, a proposition that is based on evaluation of situations in which rodent models have provided useful guidance and in which they have led to disappointments.