A wide range of anatomical features are shared by all vertebrates, but absent in our closest invertebrate relatives. The origin of vertebrate embryogenesis must have involved the evolution of new regulatory pathways to control the development of new features, but how did this occur? Mutations affecting regulatory genes, including those containing homeobox sequences, may have been important: for example, perhaps gene duplications allowed recruitment of genes to new roles. Here I ask whether comparative data on the genomic organization and expression patterns of homeobox genes support this hypothesis. I propose a model in which duplications of particular homeobox genes, followed by the acquisition of gene-specific secondary expression domains, allowed the evolution of the neural crest, extensive organogenesis and craniofacial morphogenesis. Specific details of the model are amenable to testing by extension of this comparative approach to molecular embryology.