During a 6-year period 19,647 consecutively received amniotic fluid samples were analyzed for the alpha-fetoprotein (AFP) concentration. In samples with an AFP concentration above the 95% upper normal limit the fraction of AFP non-reactive with concanavalin A (con A) was determined by crossed affinity immunoelectrophoresis. Of 436 samples from normal pregnancies with a moderate elevation of AFP (between the 95% limit and 2.5 times the median) the con A fraction was normal in 92 to 98% of the cases, depending on the gestational age. At significantly elevated AFP concentrations (above 2.5 times the median), the fractions were normal in 78% of the 50 cases with a normal outcome. In 80 of 89 cases of fetal malformations, abnormally low fractions were found, and in all 19 cases of abdominal wall defects and congenital nephrosis the fractions were low. No significant difference was found in the sensitivity for detection of fetal malformations by con A analysis compared to quantitative analysis for acetylcholinesterase (AChE). However, significantly more false positive results were found in normal pregnancies for the con A analysis compared to the AChE analysis. Both analyses performed significantly better than ultrasound scanning in terms of sensitivity. It is concluded that AChE is the test of choice in prenatal diagnosis of fetal neural tube defects. However, a sensitivity of 1.00 and a predictive value of a negative test of 1.00 were found for the con A analysis for fetal abdominal wall defects. The AchE analysis performed less well in this context.