Identifying β cell autoantigen-reactive T cells that are involved in the pathogenesis of type 1 diabetes has been troublesome for many laboratories. Disease-relevant autoreactive T cells should be in vivo Ag experienced. The aim of this study was to test this hypothesis and then use this principle as a strategy for identifying diabetes-relevant autoreactive T cells. In this study, a CSFE dilution assay was used to detect glutamic acid decarboxylase 65 (GAD65)-and insulin-responsive T cells and HLA-0201*-GAD65 114–122 pentamers were used to detect CD8+ GAD-responsive T cells in memory CD45RO+ and naive CD45RO− cell populations from patients with type 1 diabetes and healthy control subjects. T cell proliferative history was evaluated by flow cytometry telomere length measurement. CD4+ and CD8+ T cells specific for GAD65 and insulin were present in patients with type 1 diabetes and control subjects. Within the naive CD45RO− cells, CD4+ and CD8+ T cell responses were similar between patients and controls. Within the memory CD45RO+ cells, CD4+ T cell responses against whole GAD65 and insulin and HLA-0201*-GAD65 114–122 pentamer-positive CD8+ T cells were found in patients with type 1 diabetes, but not in control subjects (p< 0.05 for all). Responding cells from the CD45RO+ T cell population had substantially shorter telomere lengths than responding cells from the CD45RO− cell population. Diabetes-specific autoreactive T cells in the circulation have uniquely undergone sustained in vivo proliferation and differentiation into memory T cells. Prior selection of these cells is possible and is a way to identify diabetes-relevant target Ags and epitopes.