Islet amyloid polypeptide (IAPP, amylin) is secreted from pancreatic islet beta-cells and converted to amyloid deposits in type 2 diabetes. Conversion from soluble monomer, IAPP 1-37, to beta-sheet fibrils involves changes in the molecular conformation, cellular biochemistry and diabetes-related factors. In addition to the recognised amyloidogenic region, human IAPP (hIAPP) 20-29, the peptides human or rat IAPP 30-37 and 8-20, assume beta-conformation and form fibrils. These three amyloidogenic regions of hIAPP can be modelled as a folding intermediate with an intramolecular beta-sheet. A hypothesis is proposed for co-secretion of proIAPP with proinsulin in diabetes and formation of a 'nidus' adjacent to islet capillaries for subsequent accumulation of secreted IAPP to form the deposit. Although intracellular fibrils have been identified in experimental systems, extracellular deposition predominates in animal models and man. Extensive fibril accumulations replace islet cells. The molecular species of IAPP that is cytotoxic remains controversial. However, since fibrils form invaginations in cell membranes, small non-toxic IAPP fibrillar or amorphous accumulations could affect beta-cell stimulus-secretion coupling. The level of production of hIAPP is important but not a primary factor in islet amyloidosis; there is little evidence for inappropriate IAPP hypersecretion in type 2 diabetes and amyloid formation is generated in transgenic mice overexpressing the gene for human IAPP only against a background of obesity. Animal models of islet amyloidosis suggest that diabetes is induced by the deposits whereas in man, fibril formation appears to result from diabetes-associated islet dysfunction. Islet secretory failure results from progressive amyloidosis which provides a target for new therapeutic interventions.