At first it was verified that the major part of the fluorophotometer signal obtained when measuring corneal autofluorescence originated from fluorescence and not from scatter of excitation light at the corneal surface. The minimum percentage of the signal which can be attributed to fluorescence was determined using a fluorescence blocking filter placed in the excitation and fluorescence light path, respectively. The minimum percentages in two healthy controls, two diabetic patients and two patients after penetrating keratoplasty ranged from 75% to 93% (mean 84%). Then the corneal autofluorescence was determined in 22 healthy controls, 18 non-insulin-dependent diabetes mellitus (NIDDM). 23 insulin-dependent diabetes mellitus (IDDM) and 21 penetrating keratoplasty patients in order to detect a possible difference in autofluorescence as a result of diabetes or penetrating keratoplasty. The means of the corneal peak autofluorescence values in the NIDDM, IDDM and penetrating keratoplasty patient groups were significantly higher than that in the healthy control group (mean values in ng equivalent fluorescin ml-1: 18.0 +/- 4.2 S.D., 20.6 +/- 5.1 S.D., 17.9 +/- 5.5 S.D. and 13.7 +/- 3.7 S.D., respectively; P less than 0.01). The mean values in the NIDDM and IDDM patients did not differ significantly (P = 0.09). The autofluorescence values were independent of age in all four groups (linear correlation coefficient: r less than 0.47). The corneal peak autofluorescence was linearly correlated with the diabetes duration in the NIDDM and IDDM patients [r = 0.6, P = 0.02; increase: 0.36 ng equiv. fluorescein ml-1 yr diabetes-1]. Our results show that corneal autofluorescence is an easily obtained parameter which can be of assistance in evaluating corneal metabolism.