Circulating antibodies to pancreatic beta-cell antigens are markers of islet autoimmunity. In first-degree relatives of persons with type 1 diabetes, the levels and range of antigen specificities of these islet antibodies reflect the risk for clinical diabetes. However, in the general population, in which the disease prevalence is up to 30-fold lower, the predictive value of islet antibodies is correspondingly less. Islet antibody assays are primarily research tools to identify 'prediabetic' individuals for secondary prevention trials, but can also discriminate type 1 diabetes in several clinical situations. Loss of first-phase insulin response (FPIR) to intravenous glucose signifies imminent diabetes, but FPIR is normal in most islet-antibody-positive individuals. The contribution of a single FPIR measurement to risk assessment is therefore limited, but rate of fall of FPIR may be a useful predictor. Although beta cells are destroyed by autoreactive T cells, the assay of islet antigen-reactive T cells is not routine. Genetically, the major histocompatibility complex encoding human leukocyte antigen (HLA) alleles accounts for about 50% of familial clustering of type 1 diabetes. HLA typing is not diagnostic, but can be used to differentiate high- from low-risk individuals, e.g. at birth. While 'preclinical' diagnosis raises important medical and ethical questions, an optimized screening strategy provides a basis for counselling and follow-up. Recent knowledge of disease mechanisms and 'proof-of-principle' in the non-obese diabetic (NOD) mouse model justify expectations that type 1 diabetes is preventable, and even intervention that only delays onset of clinical diabetes is likely to be cost-effective.