Analytical methods for the determination of arsenite [As(III)], arsenate [As(V)], dimethylarsinate (DMA), monomethylarsonate (MMA), and roxarsone in filtered natural-water samples are described. Various analytical methods can be used for the determination, depending on the arsenic species being determined. Arsenic concentration is determined by using inductively coupled plasma-mass spectrometry (ICP-MS) as an arsenic-specific detector for all methods. Laboratory-speciation methods are described that use an ion chromatographic column to separate the arsenic species; the column length, column packing, and mobile phase are dependent on the species of interest. Regardless of the separation technique, the arsenic species are introduced into plasma by eithe rpneumatic nebulization or arsine generation. Analysis times range from 2 to 8 minutes and method detection limits range from 0.1 to 0.6 microgram-arsenic per liter (ug-As/L), 10 to 60 picograms absolute (for a 100-microliter injection), depending on the arsenic species determined and the analytical method used. A field-generation specciation method also is described that uses a strong anion exchange cartridge to separate As(III) from As(V) in the field. As(III) in the eluate and the As(V) in the cartridge extract are determined by direct nebulization ICP-MS. Methylated arsenic species that also are retained on the cartridge will positively bias As(V) results without further laboratory separations. The method detection limit for field speciation is 0.3 ug-As/L. The distribution of arsenic species must be preserved in the field to eliminate changes caused by photochemical oxidation or metal oxyhydroxide precipitation. Preservation techniques, such as refrigeration, the addition of acides, or the additoin of ethylene-diaminetetraacetic acid (EDTA) and the effects of ambient light were tested. Of the preservatives evaluated, EDTA was found to work best with the laboratory- and field-speciation methods for all sample matrices tested. Storing the samples in opaque polytethylene bottles eliminated the effects of photochemical oxidation. The percentage change in As(III):As(V) ratios for an EDTA-preserved acid mine drainage (AMD) sample and ground-water sample during a 3-month period was -5 percent and +3 percent, respectively. The bias and variability of the methods were evaluated by comparing results for total arsenic and As(III), As(V), DMA, and MMA concentrations in ground water, AMD, and surface water. Seventy-one ground-water, 10 AMD, and 24 surface-water samples were analyzed. Concentrations in ground-water samples reached 720 ug-As/L for As(III) and 1080 ug-As/L for As(V); AMD samples reached 12800 ug-As/L for As(III) and 7050 ug-As/L for As(V); and surface-water samples reached 5 ug-As/L for As(III) and As(V). Inorganic arsenic species distribution in the samples ranged from 0 to 90 percent As(III). DMA and MMA were present only in surface-water samples from agricultural areas where the herbicide monosodium methylarsonate was applied; concentrations never exceeded 6 ug-As/L. Statistical analyses indicated that the difference between As(III) and As(V) concentrations for samples preserved with EDTA in opaque bottles and field-speciation results were analytically insignificant at the 95-percent confidence interval. There was no significant difference among the methods tested for total arsenic concentration. Percentage recovery for field samples spiked at 50 ug-As/L and analyzed by the laboratory-speciation method (n=2) ranged from 82 to 100 percent for As(III), 97 to 102 percent for As(V), 90 to 104 percent for DMA, and 81 to 96 percent for MMA; recoveries for samples spiked at 100 ug-As/L and analyzed by the field-speciation method ranged from 102 to 107 percent for As(III) and 105 to 106 percent for As(V). Laboratory-speciation results for Environment Canada reference material SLRS-2 closely matched reported concentrations. Laboratory-speciation metho