Many biologists assume, as Darwin did, that natural selection acts mainly on late embryonic or postnatal development. This view is consistent with von Baer's observations of morphological divergence at late stages. It is also suggested by the conserved morphology and common molecular genetic mechanisms of pattern formation seen in embryos. I argue here, however, that differences in adult morphology may be generated at a variety of stages. Natural selection may have a major action on developmental mechanisms during the organogenetic period, because this is when many adult traits are specified. Evolutionary changes in these early developmental mechanisms probably include subtle shifts in the timing of gene expression. Changes of this kind have little or no gross effect on the anatomy of the embryo; they are only phenotypically expressed, or readily detected, when amplified at later stages. The phylotypic stage, the developmental hourglass, modularity, and von Baerian divergence are reassessed in terms of these arguments.