The toxicology of arsenic is complicated by its ability to convert between oxidation states and organometalloidal forms. These processes cause differences in the relative tissue-binding affinities of the various arsenic species, and they determine both the intoxication and the detoxification mechanisms. In this review, a chemical hypothesis of arsenic biomethylation is developed from an examination of data and observations presented by researchers who conducted numerous in vivo and in vitro experiments. It is likely that a combination of pathways is actually used during methylation of arsenic in vivo, and that the principal mechanism depends on various factors affecting the cellular environment. Despite these uncertainties, several observations can be made: (i) glutathione (GSH) is required for reduction of arsenic(V) to arsenic(III) species in preparation for enzyme-catalyzed oxidative methylation; (ii) GSH is not involved in monomethylation once arsenite is formed, but GSH is involved in dimethylation by reducing methylarsonic acid [MMA(V)] to methylarsonous acid [MMA(III)]; (iii) GSH is also required in the methylation of arsenic by stabilizing the reductive nature of the cell; (iv) a different methyltransferase is used in each methylation step; (v) dithiols (either a cofactor or the methyltransferases) are required for both mono- and dimethylation and (vi) where dithiols are involved, oxidative methylation reduces the stability of the arsenic-sulfur complex and permits dissociation of the arsenic species. This lower affinity of the pentavalent organoarsenic species for dithiols is part of the reason why methylation of arsenic can be a detoxification mechanism when the As(III) intermediates are not permitted to accumulate.