To document the duration of fetal cardiac time intervals in uncomplicated singleton pregnancies using a novel non-invasive fetal electrocardiography (fECG) system and to demonstrate this technique's ability to acquire recordings in twin and triplet pregnancies. Prospective cross sectional observational study. Antenatal wards and clinics, day assessment unit and centre for fetal care at a tertiary referral hospital, London, UK. Three hundred and four singleton and multiple pregnancies, 15-41 weeks of gestation. Using electrodes sited on the maternal abdomen, a fetal electrocardiography (fECG) system was developed and tested on 304 pregnant women from 15 to 41 weeks of gestation, of whom 241 were uncomplicated singletons, 58 had twin and 5 had triplet pregnancies. The composite abdominal signals were stored on a laptop computer and the fECG derived off-line using a digital signal processing technique. For singletons, linear regression was used to analyse PR, QRS, QT and QTc intervals, and construct time-specific reference ranges. Duration of fECG time intervals as a function of gestational age. Success of signal separation in singleton, twin and triplet pregnancies. For singletons, a total of 250 recordings was obtained from 241 individuals with a signal separation success rate of 85% (213/250). Success rates were significantly poorer between 27 and 36 weeks of gestation (2 x k chi(2), P < 0.0001), with 84% (31/37) of separation failures occurring during this period. P, Q, R and S waves were seen in all cases where fetal signals were separated and were used to generate fECG time interval reference ranges. In 22% (43/199) of analysed cases, no T waves were identified, 63% (27/43) of whom were < or =24 weeks of gestation. In twins and triplets, separate fetal signals were obtained in 78% (91/116) and 93% (14/15), respectively; P, Q, R and S waves were evident in all averaged fECGs, while T waves were identified in 59% (54/91) and 57% (8/14). This study provides reference ranges with gestation for fECG intervals derived non-invasively from normal singleton pregnancies and demonstrates the feasibility of obtaining complete fECG recordings non-invasively across a wide gestational range in pregnancies of all pluralities. The fECG time intervals described will enable the identification of pathological fECG recordings from high risk pregnancies where fECG abnormalities are suspected.