The aim of our study was to compare three protocols for second-trimester maternal serum screening for Down's syndrome in the same serum samples, using two triple tests [total human chorionic gonadotrophin (HCG), alpha-fetoprotein, unconjugated oestriol; and free beta-HCG, alpha-fetoprotein, unconjugated oestriol] and a double test (free beta-HCG and alpha-fetoprotein). The three protocols were compared in a series of 23 serum samples from Down's syndrome pregnancies and in a cohort of 2516 pregnant women receiving routine antenatal care between June 1992 and June 1993. Among the 23 affected cases, at a cut-off risk of 1:380, the detection rate of Down's syndrome was comparable with the double test (74%; 17/23) and the triple tests (65%; 15/23) (not significantly different). At the same cut-off risk, in the cohort of 2516 pregnant women screened between 15 and 18 weeks gestation, both protocols using free beta-HCG achieved a significant reduction of the number of false positive cases (P = 0.013 and 0.004 for double and triple tests respectively). We conclude that, compared to total HCG, alpha-fetoprotein and unconjugated oestriol, use of free beta-HCG and alpha-fetoprotein represents a better second-trimester screening test for Down's syndrome, because it significantly decreases the false positive rate at a lower running cost. The addition of unconjugated oestriol to the double test adds no further advantage.