Antennapedia class homeobox genes, which in insects are involved in regional specification of the segmented central regions of the body, have been implicated in a similar role in the vertebrate hindbrain. The development of the hindbrain involves the establishment of compartments which are subsequently made distinct from each other by Hox gene expression, implying that the lineage of neural cells may be an important factor in their development. The hindbrain produces the neural crest that gives rise to the cartilages of the branchial skeleton. Lineage also seems to be important in the neural crest, as experiments have shown that the crest will form cartilages appropriate to its level of origin when grafted to a heterotopic location. We show how the Hox genes could also be involved in patterning the mesenchymal structures of the branchial skeleton. Recently it has been proposed that the rhombomere-restricted expression pattern of Hox 2 genes is the result of a tight spatially localised induction from underlying head mesoderm, in which a prepattern of Hox expression is visible. We find no evidence for this model, our data being consistent with the idea that the spatially localised expression pattern is a result of segmentation processes whose final stages are intrinsic to the neural plate. We suggest the following model for patterning in the branchial region. At first a segment-restricted code of Hox gene expression becomes established in the neuro-epithelium and adjacent presumptive neural crest. This expression is then maintained in the neural crest during migration, resulting in a Hox code in the cranial ganglia and branchial mesenchyme that reflects the crest's rhombomere of origin. The final stage is the establishment of Hox 2 expression in the surface ectoderm which is brought into contact with neural crest-derived branchial mesenchyme. The Hox code of the branchial ectoderm is established later in development than that of the neural plate and crest, and involves the same combination of genes as the underlying crest. Experimental observations suggest the idea of an instructive interaction between branchial crest and its overlying ectoderm, which would be consistent with our observations. The distribution of clusters of Antennapedia class genes within the animal kingdom suggests that the primitive chordates ancestral to vertebrates had at least one Hox cluster. The origin of the vertebrates is thought to have been intimately linked to the appearance of the neural crest, initially in the branchial region.(ABSTRACT TRUNCATED AT 400 WORDS)