A total of 132 twin pregnancies first seen at less than 36 weeks gestation were studied prospectively to determine the epidemiological and anthropomorphic factors associated with single or dual fetal growth retardation; the clinical, biochemical and ultrasound serial measurements that are predictive of single or dual fetal growth retardation, and to design scoring systems for the prediction of fetal growth retardation. A number of factors were associated with an increased risk of fetal growth retardation, but the highest risk was present when there was an abnormality in both plasma oestriol and placental lactogen, and the biparietal diameter growth rates were divergent. Scoring systems were derived using multi-variant discriminant analysis for three clinical situations: the patient seen for the first or second time; where only clinical facilities exist and the patient was seen on three or more occasions; and where facilities exist for the biochemical tests of placental function and ultrasonic measurement of the biparietal diameters. The scoring system for the last situation had the highest predictive rate and the lowest false positive rate.