In Expt 1, 34 individually-penned Finn Dorset ewes of mean live weight 68 kg were synchronized in oestrus and mated to Suffolk rams. From mating until day 28 of pregnancy each received daily 15 MJ of metabolizable energy (ME) and 225 g crude protein (CP). From day 28 to slaughter on days 34, 41, 48 or 55 half of the ewes continued on this feeding regime and half had their daily intake reduced abruptly to 7.5 MJ of ME and 112 g CP. The mean number of ovulations per ewe was 4.03 (range 2–8) and the mean number of viable foetuses at time of slaughter 3.35 (range 2–6). The combined loss of ova (fertilization failure and early embryonic death) was 14.6% and detectable foetal deaths 2.2%. Level of feeding had no significant effect on these measures or on foetal growth. Foetal growth from 34 to 55 days was described by the equation Inω = 0.962 — 18.613e-00272t—0.00091t(f—3), where w = foetal weight (kg), t = age (days) and f = litter size. Within-litter variability measured as the S.D. of In ω (kg) was 0.081 for twins, 0.108 for triplets and 0.106 for quadruplets and higher multiples. In a second experiment Suffolk x Finn Dorset embryos were transplanted at the rate of two per uterine horn into 15 recipient Finn Dorset ewes. Embryo survival was 72% and foetal weights at 60 days varied from 67 to 146 % of the mean value of 66 g. Within-litter variation in foetal size was only about 70 % of that expected for foetuses developing from the variable distribution in their initial positioning that occurs naturally. The correlation between foetal weight and placental weight at day 60 was 0.72 (P < 0.001) indicating that the association between foetal weight and placental weight in prolific ewes is not confined to late pregnancy. The results of both experiments are consistent with the hypothesis that the greater within-litter variability in birth weight in large litters is controlled by events in early pregnancy.