While it is generally accepted that T cells are critical for the development of diabetes in the non-obese diabetic (NOD) mouse, the precise functions of the CD4+ and CD8+ subsets remain ill-defined. Transfer experiments have provided evidence that CD4+ cells are the disease initiators, provoking massive mononuclear leukocyte infiltration into the pancreatic islets, while CD8+ cells play an effector role, responsible for the final destruction of islet beta cells. It was surprising, then, to find that NOD mice carrying a null mutation at the beta 2-microglobulin (beta 2-mu) locus, and thereby lacking major histocompatibility complex class I molecules and CD8+ T cells, developed neither insulitis nor diabetes. Here, we argue that the absence of insulitis in these animals results from their lack of CD8+ cells because islet infiltration is also absent when NOD mice are treated with an anti-CD8 monoclonal antibody (mAb) at a young age. Interestingly, the anti-CD8 effect is only observed when the mAb is injected during a discrete age window--2 to 5 weeks after birth. Transfer experiments indicate that the lack of CD8+ cells during this period somehow alters the phenotype of CD4+ cells, preventing them from expressing their insulitis potential. This is not because they are generally immuno-incompetent nor because they are generally more prone to differentiating into cells with Th2 characteristics. Given that neither the beta 2-mu mutation nor anti-CD8 treatment affect insulitis in a T cell receptor transgenic (tg) mouse strain with a CD4+ T cell repertoire highly skewed for an anti-islet cell reactivity, the most straight-forward interpretation of these observations is that CD8+ cells are required for effective priming and expansion of autoreactive CD4+ cells.