In the absence of external cues, neurons in vitro polarize by using intrinsic mechanisms. For example, cultured hippocampal neurons extend arbitrarily oriented neurites and then one of these, usually the one nearest the centrosome, begins to grow more quickly than the others. This neurite becomes the axon as it accumulates molecular components of the apical junctional complex. All the other neurites become dendrites. It is unclear, however, whether neurons in vivo, which differentiate within a polarized epithelium, break symmetry by using similar intrinsic mechanisms. To investigate this, we use four-dimensional microscopy of developing retinal ganglion cells (RGCs) in live zebrafish embryos. We find that the situation is indeed very different in vivo, where axons emerge directly from uniformly polarized cells in the absence of other neurites. In vivo, moreover, components of the apical complex do not localize to the emerging axon, nor does the centrosome predict the site of axon emergence. Mosaic analysis in four dimensions, using mutants in which neuroepithelial polarity is disrupted, indicates that extrinsic factors such as access to the basal lamina are critical for normal axon emergence from RGCs in vivo.