61 fetuses/newborns who had an aberrant karyotype in amniocentesis (AC) or percutaneous umbilical blood sampling (PUBS) were followed‐up by chorionic villus sampling (CVS) at birth or after interruption.The overall rate of discrepancies is surprisingly high. Among 46 cases with a non‐mosaic numerical aberration in AC or PUBS three had a discrepant finding in placental tissue. This was also true in one of seven cases with non‐mosaic structural aberrations and in three of five cases with mosaic structural aberrations. All three cases with a mosaic numerical aberration in AC or PUBS were not represented by CVS and/or lymphocytes or fibroblasts, demonstrating the general problem of the unpredictable prognostic value of mosaicism. Our data suggest, that in case of prenatal diagnosis by CVS, using a combined procedure of short‐term (STC) and long‐term culture (LTC), in our sample we would have missed one case of 45, X (1·6 per cent). When relying only on STC another two cases, one with 47,+21 and one with 46, XX, der(22) would not have been recognized (4·9 per cent, n=3). All other chromosome aberrations would have been detected by STC alone. On the other hand, one case of 45, X was ‘nearly missed’ because of low‐grade mosaicism in AC (45, X[1]/46, XX[19]), whereas in placental tissues and PUBS only 45, X was represented. This study mimics a false‐negative rate of about 1:3000 (STC plus LTC) or about 1:1000 (STC alone) for an a priori risk group of two per cent (e.g., advanced maternal age). Copyright © 1998 John Wiley & Sons, Ltd.