Fifteen centres collaborated in a study to determine the value of amniotic fluid acetyl cholinesterase (AChE) electrophoresis in the diagnosis of fetal open neural-tube defects (NTDs) between 13 and 24 weeks of gestation among women with positive amniotic fluid alpha-fetoprotein (AFP) results. Pregnancies with positive amniotic fluid AFP results were identified from about 34 000 tested; on average they had values greater than the 99.6th percentile for non-NTD pregnancies. Amniotic fluid samples were available from 1099 pregnancies with positive AFP results, and the percentages which yielded positive AChE results were as follows: anencephaly 99.6% (476/478); open spina bifida 99.4% (333/335); non-NTD pregnancies (i) without serious malformation and not ending in miscarriage 6% (8/125), (ii) without serious malformation but ending in miscarriage 47% (34/73), (iii) exomphalos 75% (47/63), (iv) congenital nephrosis 0% (0/11), (v) other serious malformations 50% (7/14). The gel-AChE test was thus found to be very effective when applied to amniotic fluid samples with positive AFP results, substantially reducing the number of false positives with only a very small loss in the detection of open neural-tube defects. Among women with positive amniotic fluid AFP results, the risk of having a fetus with an open NTD compared to having one without a serious malformation or a miscarriage is raised about 16-fold if the AchE test is also positive. The actual risk will depend on the birth prevalence and the reason for amniocentesis. For example, in the U.K. the risk of open spina bifida (expressed as an odds ratio) when both the amniotic fluid AFP result and the gel AChE results are positive would be as follows for women referred for amniocentesis in different categories; approximately 288:1 for those referred because of single raised serum AFP level (greater than or equal to 2.5 time the normal median at 16-18 weeks of gestation, 80:1 for women referred because of having previously had an infant with an NTD, and 8:1 for women referred for other reasons. It the amniotic fluid sample is stained with blood which is mainly fetal in origin, the corresponding odds are much lower (36:1, 10:1, and 1:1 respectively): if it is not blood-stained or is stained mainly with maternal blood, they are higher (954:1, 265:1, and 26:1 respectively).