Summary. Many patients with Type 2 (non-insulin-dependent) diabetes mellitus are treated with insulin in order to control hyperglycaemia. We studied fasting plasma C-peptide, glucagon stimulated plasma C-peptide, and 24 h urinary Cpeptide in relation to clinical type of diabetes in 132 insulin treated diabetic subjects. Patients were classified clinically as Type 1 (insulin-dependent) diabetic subjects in the presence of at least two of the following criteria: 1) significant ketonuria, 2)insulin treatment started within one year after diagnosis, 3) age of diagnosis < 40 years, and 4) weight below 110% of ideal weight of the same age and sex. Eighty patients were classified as Type 1 and 52 as Type 2 diabetic subjects. A second classification of patients into 6 C-peptide classes was then performed. Class I consisted of patients without islet Bcell function. Class II-VI had preserved islet B-cell function and were separated according to the 20%, 40%, 60% and 80% C-peptide percentiles. The two classifications of patients were compared by calculating the prevalence of clinical Type I and Type 2 diabetes in each of the C-peptide classes. This analysis showed that patients with a fasting plasma C-peptide value < 0.20 nmol/1, a glucagon stimulated plasma C-peptide value < 0.32 nmol/1, and a urinary C-peptide value < 3.1 nmol/1, or < 0.54 nmol/mmol creatinine/24 h, or < 5.4 nmol/24 h mainly were Type I diabetic patients; while patients with Cpeptide levels above these values mainly were Type 2. At these limits the percentage, predictive value of positive tests as indicators of Type 2 diabetes were as follows: fasting C-peptide 83%, stimulated C-peptide 86%, and urinary C-peptide expressed as nmol/1 76%, as nmol/mmol creatinine/24 h 79%, and as nmol/24 h 78%. Similarly, the percentage predictive value of negative tests as indicators of Type 1 diabetes were as follows: fasting C-peptide 86%, stimulated C-peptide 88%, and urinary C-peptide expressed as nmol/1 79%, as nmol. mmol creatinine-24 h 81%, and as nmol/24 h 80%. If patients without detectable C-peptide were excluded, the predictive value of negative tests were as follows: fasting C-peptide 81%, stimulated C-peptide 88%, urinary C-peptide expressed as nmol/161%, as nmol/mmol creatinine/24h 69%, and as nmol/24 h 64%. In conclusion, post glucagon C-peptide gives a good distinction between Type 1 and Type 2 diabetes mellitus in insulin treated diabetes while 24 h urinary C-peptide gives a less sensitive distinction between the clinical types of diabetes.