It has been proposed elsewhere that the teratogenic effects of retinoids on craniofacial morphogenesis are caused by a disturbance of the migration of cranial neural crest cells. The effects of 3.5 X 10(-5) M and 3.5 X 10(-6) M-retinol on the migration of avian neural crest cells in vitro have been investigated by monitoring cell morphology, locomotory behaviour, fibronectin distribution and actin-microfilament organization. Retinol retards migration by affecting cell-to-substratum adhesiveness. Cells exposed to medium containing retinol are less adherent to the substratum, and although the cell surface is very mobile, are unable to extend or maintain lamellipodia. As a consequence the cells do not actively translocate. Fibronectin distribution at the cell surface is sparse, possibly as a result of shedding, and actin distribution remains diffuse. At the retinol molarities used all these effects are reversible. Thus cells allowed to recover in normal medium flatten out, display lamellipodia and commence active translocation. Fibronectin becomes organized into a fibrillar array and actin microfilaments become organized into cables. The period needed for this recovery is directly related to the molarity of retinol during the initial exposure; after recovery the retinol-treated cells are virtually indistinguishable from control cells. We propose that in vivo the effects of retinoids might be to impair cell-extracellular matrix interaction, thus impeding a cell's ability to migrate through that matrix. Contrary to previous suggestions, the in vivo effects are probably not in any way 'specific' to neural crest cells but are more accurately considered as 'selective', in that any cell undergoing migration would be similarly affected.