The aim of this study was to determine whether identification of glial cells in amniotic fluid samples could form a useful supplementary test in the antenatal diagnosis of neural tube defects (NTDs). In a 5-year study, 1452 samples of middle trimester amniotic fluid were examined blind to the results of other antenatal diagnostic tests and to the outcome of pregnancy. Reason of amniocentesis included raised serum alpha-fetoprotein (329), previous NTD (73), and a family history of NTDs (71). Duplicate cytospin preparations were stained with Giemsa and an antibody to glial fibrillary acidic protein (GFAP), and on this basis a prediction of fetal NTD status was made which was not communicated to clinicians. Subsequent management of pregnancies was influenced only by the results of routine antenatal testing for NTDs. Twenty cases of NTDs occurred among the 1406 cases in which the outcome was subsequently known. Of these 20 cases, only five (four anencephalic, one spina bifida) were correctly predicated by immunofluorescent identification of GFAP-positive cells in the amniotic fluid. The remaining 15 cases (two anencephalic, 13 spina bifida) were not so identified. In a further 18 cases, apparently GFAP-positive cells were identified in the absence of NTDs. We conclude that GFAP immunofluorescence examination of routine amniocentesis samples of amniotic fluid is not a useful predictive test for NTDs.