In the present paper, we examine how arsenic species accumulate in fish liver and explore the hypothesis that sublethal arsenic concentrations in fish hepatic tissue interfere with stress-mediated gene expression. We exposed killifish (Fundulus heteroclitus) to 787 or 0 microg/L arsenic in tank water for 2 weeks. Arsenic exposure elevated total liver arsenic from 3.4 microg/g wet weight (control fish) to 9.6 microg/g wet weight, and resulted in a higher relative proportion of toxic (e.g. monomethylarsenous acid, dimethylarsenous acid, arsenic V) versus benign (arsenobetaine) arsenic species in this tissue. Following the exposure period, arsenic-treated and control fish were then subjected to a stress protocol: confinement and mechanical chasing for 15 min every 3 h. Liver tissue and blood were sampled from fish not exposed to the stressor at time 0, and at 8, 12, 24 and 40 h following the first stressor. Concentrations of the stress hormone cortisol increased significantly, and glucocorticoid receptor mRNA levels increased and then decreased in both groups, but patterns were nearly identical between arsenic pre-treated and arsenic untreated fish. Prior arsenic exposure prevented the stress-induced increases in stress-responsive LDH-B mRNA levels and enzyme activity observed in fish that had not been exposed to arsenic. However, in another stress-responsive gene, PEPCK, arsenic did not interfere with the stress-induced increase in gene expression, suggesting that the effects of arsenic on stress-mediated gene expression are complex and may involve regulatory pathways that differ between these two genes.