24 diabetic patients stabilised on conventional bovine insulins and possessing insulin antibodies underwent a study of the immunological and clinical consequences of changes in both purity and species of their insulin preparations. After a 2-month run-in period patients were treated for 3 consecutive 4-month periods with (a) purified bovine insulin (20-40 ppm proinsulin), (b) highly purified porcine insulin, and (c) semisynthetic human insulin, without elective dose changes. Mean insulin antibody levels changed little on purified bovine insulin (22.2 leads to 23.4 micrograms/l) but fell on highly purified porcine (23.4 leads to 12.9 micrograms/l) and remained much the same on Semi-Synthetic human insulin. In contrast, C-peptide antibodies fell significantly and continuously throughout the study. The slower rate of fall in C-peptide antibody levels is likely to be due to the prolonged half-life of circulating exogenous proinsulin in the presence of insulin antibody. Although insulin dose remained constant the incidence of hypoglycaemic episodes did not increase and glycosylated haemoglobin levels rose significantly when patients were on porcine insulin. The deterioration in diabetic control may have been due to greater temporal mismatch between insulin needs and insulin availability with pork or human insulin than with beef insulins, and to reduced insulin antibody levels. The use of purer insulins which more closely resemble the human form can cause a significant reduction in levels of insulin and C-peptide antibodies. Although these changes may have other benefits they do not necessarily produce better diabetic control. Subcutaneous insulin regimens need to be tailored to the individual patient and, indirectly, to his antibody status.