The prevalence of mental disorders (DSM-IIIR criteria) among 107 neurological inpatients was estimated, as well as the extent to which disorders were detected by neurologists. The validity of the scaled version of the General Health Questionnaire (GHQ-28) was evaluated using Receiver Operating Characteristic (ROC) analysis and DSM-IIIR as external criteria. Of the 107 patients who submitted to a structured psychiatric interview (SCID-R), 56 (52.3%) showed evidence of a mental disorder. Major depressive episode (n = 16), generalized anxiety disorders (n = 13) and dysthymia (n = 12) were the most frequent diagnoses. The neurologists recognized only 13/107 cases (12.1%). Significantly more women than men exhibited some form of mental disorder. The validation of GHQ-28 in the series of 107 neurological inpatients indicated that the best trade-off between sensitivity and specificity was the cut-off score of 5/6. The high occurrence of mental disorder, in association with the low rate of detection by the neurologists, points to the need for special attention to be paid to this problem by staff and experts.