Graphite furnace-atomic absorption spectrometry (GF-AAS) is a sensitive, precise, and accurate technique that can be used to determine arsenic and selenium in samples of water and sediment. The GF-AAS method has been developed to replace the hydride generation-atomic absorption spectrometry (HG-AAS) methods because the method detection limits are similar, bias and variability are comparable, and interferences are minimal. Advantages of the GF-AAS method include shorter sample preparation time, increased sample throughput from simultaneous multielement analysis, reduced amount of chemical waste, reduced sample volume requirements, increased linear concentration range, and the use of a more accurate digestion procedure. The linear concentration range for arsenic and selenium is 1 to 50 micrograms per liter in solution; the current method detection limit for arsenic in solution is 0.9 microgram per liter; the method detection limit for selenium in solution is 1 microgram per liter. This report describes results that were obtained using stop-flow and low-flow conditions during atomization. The bias and variability of the simultaneous determination of arsenic and selenium by GF-AAS under both conditions are supported with results from standard reference materials--water and sediment, real water samples, and spike recovery measurements. Arsenic and selenium results for all Standard Reference Water Samples analyzed were within one standard deviation of the most probable values. Long-term spike recoveries at 6.25, 25.0, 37.5 micrograms per liter in reagent-, ground-, and surface-water samples for arsenic averaged 103 plus or minus 2 percent using low-flow conditions and 104 plus or minus 4 percent using stop-flow conditions. Corresponding recoveries for selenium were 98 plus or minus 13 percent using low-flow conditions and 87 plus or minus 24 percent using stop-flow conditions. Spike recoveries at 25 micrograms per liter in 120 water samples ranged from 97 to 99 percent for arsenic and from 82 to 93 percent for selenium, depending on the flow conditions used. Statistical analysis of dissolved and whole-water recoverable analytical results for the same set of water samples indicated that there is no significant difference between the GF-AAS and HG-AAS methods. Interferences related to various chemical constituents were also identified. Although sulfate and chloride in association with various cations might interfere with the determination of arsenic and selenium by GF-AAS, the use of a magnesium nitrate/palladium matrix modifier and low-flow argon during atomization helped to minimize such interferences. When using stabilized temperature platform furnace conditions where stop flow is used during atomization, the addition of hydrogen (5 percent volume/volume) to the argon minimized chemical interferences. Nevertheless, stop flow during atomization was found to be less effective than low flow in reducing interference effects.