A cross-sectional survey and subsequent longitudinal study among diabetic residents of Rochester, MN--The Rochester Diabetic Neuropathy Study (RDNS)--is population-based and uses quantitative, validated, and unique end points to detect, classify, and stage neuropathy. Nondiabetic persons, drawn from the same population, serve as controls. For patients 10 to 70 years old, the RDNS cohort is representative of diabetics living in Rochester, MN. We assessed reproducibility of tests used to characterize and quantitate severity of neuropathy in 20 diabetic subjects without neuropathy and with varying severities of neuropathy. Using intraclass correlation coefficient (rI) as a measure of test reproducibility, we found high rI (usually 0.9 or better) with small confidence intervals for the Neurologic Disability Score (NDS); weakness subset of NDS (W-NDS); vibratory and cooling detection thresholds (using computer-assisted sensory examination [CASE] IV); compound muscle action potentials; sensory nerve action potentials; and motor nerve conduction velocities. There was good agreement among three trained observers for NDS and the W-NDS.