The objective of the study was to define how could be helpful a nerve biopsy for identification of atypical cases of chronic inflammatory demyelinating polyneuropathy (CIDP). An ad hoc committee in 1991 defined the clinical, electrophysiological and pathological criteria for diagnosis of CIDP. In common with other authors, we regard the rather specific electrophysiological criteria as being too restrictive, and we think that a significant number of patients may therefore not benefit from effective treatment or be excluded from therapeutic trials. The Inflammatory Neuropathy Cause and Treatment (INCAT) group (2001) has proposed new electrophysiological criteria of CIDP, which are more sensitive and do not loose any specificity. Over a period of three years (January 1999 to December 2001), we classified 44 patients into two categories: those presenting the strict criteria of the ad hoc committee and those who we regarded as cases of CIDP who did not meet these strict criteria. All these patients benefited from one or more clinical and electrophysiological examinations; extensive biological workup and genetic study when appropriate excluded other causes of neuropathy. Nerve biopsies were taken from all patients and samples were included in paraffin and epon for systematic light and electron microscopic examination. Out of 44 patients, 24 fulfilled the INCAT electrophysiological criteria with only 12 of these cases fulfilling the criteria of the ad hoc committee. Eight patients did not fulfill any of the widely accepted electrophysiological criteria of CIDP. However, study of nerve biopsies of these eight patients revealed histological features characteristic of CIDP according to histological criteria (AAN-1991). Among these patients, six have been treated and five responded favorably to conventional treatments for CIDP. Without information from the nerve biopsy, these patients would not have been treated effectively because their electrophysiological profile was indicative of axonal impairment interpreted erroneously as primary.