Three locally collected gasoline samples were analyzed for arsenic content using factorial experimental design (FED) with hydride generation atomic absorption spectrometry (HGAAS). Five factors (including the reduction agent flow, the acid flow, the sample flow, the reductor concentration and the acid concentration) in two levels (low and high) were used in this optimization process. In order to design the surface response, five levels of central composite with 15 central points were down-selected. The results were compared with those determined by graphite furnace atomic absorption spectrometry (GFAAS) and a 95% confidence interval (regression procedure) was obtained. The HGAAS and GFAAS measurement sensitivities determined arsenic levels of 0.02 mu g L-1 and 0.08 mu g L-1, respectively. (C) 2006 Elsevier Ltd. All rights reserved.