Although the molecular lesions underlying a substantial proportion of all cases of permanent and neonatal diabetes have now been defined, in as many as 30% of cases the defect is unknown. Three complementary approaches may help to define further disease-causing changes: (1) the molecular dissection of the insulin secretory process itself using cellular, physiological and imaging techniques; (2) measurements of the level of expression of beta-cell genes in islets from rodents or humans with type 2 diabetes (T2D) by oligonucleotide micro-array or proteomic analysis, and (3) population-wide whole-genome association studies of T2D. Here, I survey recent published data in this context.