Modifications to the commercially available hydride generator resulted in an improved operating procedure and enhancement of the arsenic signal. Experiments with arsenic(III) and arsenic(V) showed that identical results could be produced from both oxidation states. However, as arsenic(V) is reduced more slowly than arsenic(III), peak areas and not peak heights must be measured when the arsine is immediately stripped from the system (approximately 5 s reaction). When the reduction is allowed to proceed for 20 s before the arsine is stripped, peak heights may be used. For a 200 mg ml^-1 solution, the relative standard deviation is 2.8 percent for arsenic(III) and 3. 8 percent for arsenic(V). The detection limit for arsenic using the modified system is 0.50 mg ml^-1.