The fate of cells in the epiblast at prestreak and early primitive streak stages has been studied by injecting horseradish peroxidase (HRP) into single cells in situ of 6.7-day mouse embryos and identifying the labelled descendants at midstreak to neural plate stages after one day of culture. Ectoderm was composed of descendants of epiblast progenitors that had been located in the embryonic axis anterior to the primitive streak. Embryonic mesoderm was derived from all areas of the epiblast except the distal tip and the adjacent region anterior to it: the most anterior mesoderm cells originated posteriorly, traversing the primitive streak early; labelled cells in the posterior part of the streak at the neural plate stage were derived from extreme anterior axial and paraxial epiblast progenitors; head process cells were derived from epiblast at or near the anterior end of the primitive streak. Endoderm descendants were most frequently derived from a region that included, but extended beyond, the region producing the head process: descendants of epiblast were present in endoderm by the midstreak stage, as well as at later stages. Yolk sac and amnion mesoderm developed from posterolateral and posterior epiblast. The resulting fate map is essentially the same as those of the chick and urodele and indicates that, despite geometrical differences, topological fate relationships are conserved among these vertebrates. Clonal descendants were not necessarily confined to a single germ layer or to extraembryonic mesoderm, indicating that these lineages are not separated at the beginning of gastrulation. The embryonic axis lengthened up to the neural plate stage by (1) elongation of the primitive streak through progressive incorporation of the expanding lateral and initially more anterior regions of epiblast and, (2) expansion of the region of epiblast immediately cranial to the anterior end of the primitive streak. The population doubling time of labelled cells was 7.5 h; a calculated 43% were in, or had completed, a 4th cell cycle, and no statistically significant regional differences in the number of descendants were found. This clonal analysis also showed that (1) growth in the epiblast was noncoherent and in most regions anisotropic and directed towards the primitive streak and (2) the midline did not act as a barrier to clonal spread, either in the epiblast in the anterior half of the axis or in the primitive streak. These results taken together with the fate map indicate that, while individual cells in the epiblast sheet behave independently with respect to their neighbours, morphogenetic movement during germ layer formation is coordinated in the population as a whole.