Development of definitive (fetal liver-derived) red cells is blocked by a targeted mutation in the gene encoding the transcription factor GATA-1. We used in vitro differentiation of GATA-1- mouse embryonic stem (ES) cells to reveal a requirement for GATA-1 during primitive (yolk sac-derived) erythropoiesis and to establish a rescue assay. We show that the block to development includes primitive, as well as definitive, erythroid cells and is complete at the level of globin RNA expression; that the introduction of a normal GATA-1 gene restores developmental potential both in vivo and in vitro; and that efficient rescue is dependent on a putative autoregulatory GATA-motif in the distal promoter. Use of in vitro differentiated ES cells bridges a gap between conventional approaches to gene function in cell lines and analysis of loss of function mutations in the whole animal.