The N. medianus was stimulated electrically in 44 healthy persons and in 30 unselected diabetics. Latencies of the F wave were measured, corrected for different lengths of the upper extremity, and used for computation of nerve conduction velocity. In comparison with conventional motor nerve conduction velocities the F wave latencies showed a more pronounced difference between the normal and the pathological (diabetic) range. The F wave nerve conduction velocities also varied less than conventional nerve conduction velocities. The explanation for this might be that the F wave recordings are the result of a full anti- and orthodromic impulse propagation which travels 5 to 6 times further than the impulse measured with conventional nerve conduction velocity methods. Therefore small delays of nerve conduction will be amplified and can be better recognized by F wave latency measurements. The present findings suggest that F wave latencies may be more sensitive than conventional nerve conduction velocities in evaluating mild cases and in identifying the initial phases of neuropathies.