In this paper I have tried to bring together work that highlights the role of homeobox genes in generating craniofacial form. I review both normal and disrupted embryogenesis and ask whether mis-expression of the homeobox genes outside their normal domains could be contributing to congenital facial abnormalities arising from either genetic or teratogenic actions. Experimentally generated transgenic mice carrying loss- or gain-of-function mutations in homeobox genes, in combination with their normal expression patterns, have allowed us to compile and test models of embryonic specification based around a Hox/homeobox code. These models form the basis on which the functional questions are considered. There are four major sections covering different experimental approaches designed to ectopically induce homeobox genes in the head. Transgenic mice, where heterologous promoters drive a given Hox gene in the head, have shown that the more posteriorly expressed Hox genes tend to have a significant effect only on the skull bones of mesodermal origin whereas those normally expressed more anteriorly, in the hindbrain and branchial arches, can affect more anterior branchial arch and neural crest-derived structures. Manipulation experiments which can induce homeobox genes in small, localised regions of the facial precursors show clear and dramatic effects of this expression on facial development. Null mutations in predicted repressors of Hox gene expression, however, do not appear to give rise to substantial craniofacial abnormalities. Retinoic acid, on the other hand, is well known for its teratogenic actions and its ability to induce Hox gene expression. Evidence is now accumulating that at least some of its teratogenic actions may be mediated by its regulation of the Hox and other homeobox genes in the head.