This study first compares two methods for measuring first order multifocal electroretinogram (mfERG) implicit time abnormalities in eyes with early diabetic retinopathy. Two analysis methods are used: template stretching (multiplicative scaling) of an 80 msec response epoch and template sliding (cross-correlation or additive scaling) of portions of responses containing the major waveform features. The study also compares the relative sensitivities of N1, P1 and N2 implicit time assessed by cross-correlation. The nature of the change in the mfERG waveform associated with diabetes is also assessed. MfERGs were recorded from 15 eyes of 15 individuals with diabetes and early non-proliferative retinopathy and 20 eyes of 20 healthy control subjects of similar age. Implicit time determined by template stretching is more frequently abnormal in the eyes of the diabetic subjects than the implicit time of any of the components assessed by template sliding. This is attributable to the lower variability of the template stretching implicit time measure in normals. Of the components, P1 is most often abnormal in the eyes of individuals with diabetes. Responses recorded from retinal areas with retinopathic signs are more often abnormal than those from other areas. Later components of the response are not delayed more than earlier ones. We conclude that template stretching is a sensitive measurement technique, but that it does not fully capture the effect of diabetes on the first order mfERG well.