Composite scores may be more sensitive and reproducible than single attributes of nerve conduction for detection of peripheral neuropathy, but this requires validation in large patient cohorts. Also, the concordance of individual attributes versus composite scores with clinical measures of severity has not been tested. Here, we study these issues in prospectively studied cohorts: diabetic patients from Rochester, Minnesota (RDNS; n = 396); chronic inflammatory demyelinating polyneuropathy (CIDP) patients (n = 55); and multifocal motor neuropathy (MMN) patients (n = 18). With specificity fixed at the 97.5 percentile, we found that, in generalized polyneuropathies (diabetic and CIDP), composite scores (especially ones including conduction velocity, distal latencies, and F-waves) of individual or multiple nerves tended to be more sensitive than individual attributes. By contrast, for multiple mononeuropathies, some individual attributes or composite scores of individual nerves were more sensitive than composite scores. In diabetic polyneuropathy, composite scores tended to be more reproducible than individual attributes of nerve conduction. Highly significant correlations were found between individual attributes or composite scores and neurologic impairment in diabetic polyneuropathy and in CIDP; in general, correlation coefficients were higher for composite scores. These correlations were higher for amplitudes than for conduction velocities or distal latencies. We conclude that, with the availability of microprocessors and normative databases, electromyographers may increasingly seek to express nerve conduction abnormality also as composite scores of individual or several nerves.