Three sets of electrodiagnostic criteria for establishing primary demyelination in chronic polyneuropathy are evaluated. Sensitivity is assessed in 70 patients with clinically established chronic inflammatory demyelinating polyradiculoneuropathy (CIDP). The criteria use different abnormal values, one adjusts for the effects of axonal loss, while another relies only on conduction velocity. However, even when consideration is given to sufficient number of nerves tested, there is no significant difference (P = 0.37) in diagnostic sensitivity among them, with 48% to 64% of CIDP patients fulfilling criteria for primary demyelination. Specificity is assessed by applying the criteria to 47 patients with motor neuron disease and 63 patients with diabetic polyneuropathy. No patients meet any of the criteria. Further analysis shows that as sensitivity increases specificity decreases, because of overlapping distributions of nerve conduction abnormalities in these neuropathic disorders. A sensitivity of approximately 66% is a practical limit for electrodiagnostic criteria in CIDP.