The median maternal serum unconjugated oestriol level between 13 and 27 weeks gestation in 77 pregnancies associated with Down's syndrome was lower than the median level in 385 unaffected control pregnancies matched for maternal age, gestational age, and duration of serum sample storage (P less than 0.001). The median level for the affected pregnancies was 73% of that in the controls. Low unconjugated oestriol levels can be used to detect fetal Down's syndrome; at cut-off levels selected to detect at least 35% of affected pregnancies, unconjugated serum oestriol was a better screening test than either maternal age or serum alpha-fetoprotein (AFP). The use of all three variables in combination to select women with a 1:250 or greater risk of a Down's syndrome term pregnancy would yield a 45% detection rate with a false-positive rate of 5.2%. The same detection rate using maternal age alone or using age and serum AFP in combination would yield higher false-positive rates, 15% and 9.8% respectively. The addition of unconjugated oestriol to a Down's syndrome screening programme would therefore be more efficient than the use of age and AFP alone; for a given detection rate fewer women would need an amniocentesis or, for a given percentage of women having an amniocentesis, more pregnancies with Down's syndrome would be detected.