The transforming growth factor β1 (TGFβ1) signalling pathway^1,2 is important in embryogenesis³ and has been implicated in hereditary haemorrhagic telangiectasia (HHT)^4,5, atherosclerosis^6,7, tumorigenesis^8,9 and immunomodulation^10,11. Therefore, identification of factors which modulate TGFβ1 bioactivity in vivo is important. On a mixed genetic background, ∼50% Tgfb1−/− conceptuses die mid-gestation from defective yolk sac vasculogenesis3. The other half are developmentally normal but die three weeks postpartum^3,10,11. Intriguingly, the vascular defects of Tgfb1−/− mice share histological similarities to lesions seen in HHT patients^3–5. It has been suggested that dichotomy in Tgfb1−/− lethal phenotypes is due to maternal TGFβ1 rescue of some, but not all, Tgfb1−/− embryos¹². Here we show that the Tgfb1−/− phenotype depends on the genetic background of the conceptus. In NIH/Ola, C57BL/6J/Ola and F1 conceptuses, Tgfb1−/− lethality can be categorized into three developmental classes. A major codominant modifier gene of embryo lethality was mapped to proximal mouse chromosome 5, using a genome scan for non-mendelian distribution of alleles in Tgfb1−/− neonatal animals which survive prenatal lethality. This gene accounts for around three quarters of the genetic effect between mouse strains and can, in part, explain the distribution of the three lethal phenotypes. This approach, using neonatal DNA samples, is generally applicable to identification of loci that influence the effect of early embryonic lethal mutations, thus furthering knowledge of genetic interactions that occur during early mammalian development in vivo.