We investigated pancreatic polypeptide (PP) release by chew and spit sham feeding as a test of autonomic function in 33 patients with diabetes mellitus (DM) and 24 age-matched healthy controls. Plasma samples were taken at nine 5-min time intervals, and PP plasma levels were determined by radioimmunoassay. Autonomic neuropathy (AN) was diagnosed and staged according to four standardized tests of cardiovascular autonomic function. The integrated PP response during sham feeding was decreased in DM patients (1067 ± 397 vs. 2670 ± 394 pg/ml/30 min; P < 0.05). The maximum increase of PP plasma levels was significantly different between the DM subgroups with AN (36 ± 19 pg/ml; n = 13) and without AN (132 ± 36 pg/ml; n = 20) but varied widely in healthy controls (219 ± 29 pg/ml; range 20–460 pg/ml). A cutoff of maximum PP increase < 20 pg/ml (specificity 100%) resulted in only 46% sensitivity, while a cutoff < 160 pg/ml (sensitivity 100%) yielded 38% specificity for diagnosis of AN. The diagnostic value of the PP test for identifying individual patients against the variable reference range is therefore limited. Maximum PP increase < 20 pg/ml indicates AN, while maximum PP increase > 160 pg/ml excludes AN in diabetic patients.