To assess the rate of metabolic deterioration and potential risks of failure of continuous subcutaneous insulin infusion during basal insulin delivery, we deliberately stopped infusion in nine insulin dependent diabetics. Plasma glucose, blood 3-hydroxybutyrate and plasma free insulin were measured for 9 h whilst the patients remained supine and fasting. Mean plasma glucose remained unchanged at normal fasting levels for the first hour, then rose to plateau at about 10 mmol/l until the end of the experiment. The final plateau level of glucose varied from patient to patient; two C-peptide secreting diabetics plateaued at low glucose levels. In contrast, blood 3-hydroxybutyrate rose progressively, without plateauing. PLasma free insulin concentrations fell during the withdrawal period and there was a highly significant negative correlation between free insulin and 3-hydroxybutyrate. No patient was more than mildly unwell after 9 h of insulin deprivation. We conclude that under these experimental conditions there is glycaemic autoregulation and that ketones may sometimes be a more appropriate monitor of insulin deficiency or loss of diabetic control than is glucose. Accidental failure of continuous subcutaneous insulin infusion and interruption of basal delivery in resting and fasting diabetics will probably not cause dangerous metabolic or clinical deterioration.