Patients with maturity onset diabetes that is poorly controlled on maximal doses of oral hypoglycaemic agents are difficult to treat. A prospective randomised crossover study was performed in 58 predominantly non-obese patients on maximal doses of glibenclamide or metformin, or both, to find out if insulin would improve control or well being. The patients were given daily injections of up to 48 units of highly purified porcine lente insulin. Glycaemic control was improved by 15% or more in only 18 patients; 14 others felt better but their diabetes was no better controlled. Those whose control was improved by insulin could not be distinguished by age, duration of diabetes, body mass index, or their own treatment preference. C peptide concentrations, however, did help predict the response to insulin, the fasting C peptide to glucose ratio being considerably lower in those patients whose control was better on insulin. These findings suggest that a simple insulin regimen does not necessarily lead to better glycaemic control in maturity onset diabetes. Nevertheless, a trial of insulin is often justified since it poses few practical difficulties and makes some patients feel better even if their control is not improved. A more complex regimen might improve control in more cases, but it might also be less acceptable to older patients.