We describe the incidence and explore the quality of registered diagnosis for Guillain-Barré syndrome (GBS) in a population-based retrospective study in South-West Stockholm (SWS), Sweden, during the period 1973-1991. We used data on registered hospital discharges and open-care visits, and from medical records. Medical records were available for 83 (80.6%) of 103 patients with registered GBS diagnosis at discharge from hospitals. For 69 (83.1%) of such patients, the information from the hospital record fulfilled the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) clinical criteria for GBS. One out-patient presented with suspected GBS. In contrast, none of 40 patients discharged with diagnosis of unspecific polyneuropathy fulfilled the above mentioned criteria. During the period 1973-1991, the mean annual incidence of GBS per 100,000 was 1.74, 95% CI 1.41-2.12, from register data and 1.49, 95% CI 1.19-1.85, after excluding those not fulfilling criteria for GBS. The age-adjusted incidence after validation was higher for men, 1.64, 95% CI 1.19-2.21, than for women, 1.46, 95% CI 1.05-1.99. An increasing incidence with age was found, presenting the age-specific curve a bimodal shape. The results of this study confirm that hospital discharge diagnostic data in Sweden can be used for purposes of epidemiological research and surveillance for GBS, and show that the incidence of GBS in SWS compares to those described in other populations.