Neonatal hyperinsulinism is characterized by severe hypoglycaemia which can cause serious neurologic effects. Pancreatic morphological abnormalities involve either focal or diffuse lesions. The former can be cured by resection, whereas the latter, of uncertain pathogenesis, often require subtotal pancreatectomy. We investigated various hypotheses in an effort to explain the origin of this latter form of hyperinsulinism. We determined that nesidioblastosis, long considered to be the basic structural lesion of the diffuse form of hyperinsulinism, is not specific and does not correspond to a continuous proliferation of endocrine cells. We found that an increase in beta-cell mass can be excluded since the volume density of beta cells is not systematically higher in hyperinsulinemic infants than in controls. The hypothesis of a decrease in D cells is attractive but should be considered with due caution since the decrease of the D-cell volume density observed in hypoglycaemic infants is inconstant. Finally, the notion of beta-cell functional abnormality seems the most likely explanation since a higher quantity of proinsulin was detected within the Golgi area by a specific antibody and abnormal nuclei with abundant cytoplasm were observed in some cells. These histological abnormalities can be observed during intraoperative morphological examination. Functional activity might also be evaluated by studying the messenger RNA of proinsulin.