Many modes of action for arsenic carcinogenesis have been proposed, but few theories have a substantial mass of supporting data. Three stronger theories of arsenic carcinogenesis are production of chromosomal abnormalities, promotion of carcinogenesis and oxidative stress. This article presents the oxidative stress theory along with some supporting experimental data. In the area of which arsenic species is causually active, recent data have suggested that trivalent methylated arsenic metabolites, particularly monomethylarsonous acid (MMA(III)) and dimethylarsinous acid (DMA(III)), have a great deal of biological activity. Some evidence now indicates that these trivalent, methylated, and relatively less ionizable arsenic metabolites may be unusually capable of interacting with cellular targets such as proteins and even DNA. Thus for inorganic arsenic, oxidative methylation followed by reduction to trivalency may be a activation, rather than a detoxification pathway. This would be particularly true for arsenate. In forming toxic and carcinogenic arsenic species, reduction from the pentavalent state to the trivalent state may be as or more important than methylation of arsenic.