As part of a screening project for detection of Toxoplasma gondii infection among pregnant women in Norway, nested polymerase chain reaction (PCR) aimed at the detection of T. gondii in amniotic fluid samples was included in the diagnostic routine. The results were compared with the routine criteria for congenital infection: i) T. gondii detected in amniotic fluid or cord blood by mouse inoculation, ii) specific IgM or IgA in serum collected after birth, and/or iii) specific IgG persisting beyond one year of age. The PCR was based on the B1 gene with an internal control gene amplified together with the B1 gene. One hundred and two amniotic fluid samples collected during pregnancy and/or at delivery from 67 pregnant women with serological evidence of primary T. gondii infection were available for examination by both B1-PCR and mouse inoculation. Six samples were positive and 86 samples were negative by both methods (90% concordance). One sample was mouse inoculation positive and B1-PCR negative while nine samples were B1-PCR positive and mouse inoculation negative, of which five were associated with four infants without proven infection. 59%, and 41% of samples associated with infected infants were positive by B1-PCR and mouse inoculation, respectively. The difference was mainly due to a lower detection rate by mouse inoculation after antiparasitic treatment. The specificity of B1-PCR was 94%. Even though B1-PCR performed on amniotic fluid samples did not detect all infected infants, it represented a valuable tool in addition to conventional methods in the diagnosis of congenital T. gondii infection.