We developed two separation methods using CZE with UV detection for the determination of the most common inorganic and methylated arsenic species and some phenylarsenic compounds. Based on the separation method for anions using hydrodynamic sample injection the detection limits were 0.52, 0.25, 0.27, 0.12, 0.37, 0.6, 0.6, 1.2 and 1.0 mg As/L for phenylarsine oxide (PAO), p-aminophenylarsonic acid (p-APAA), o-aminophenylarsonic (o-APAA), phenylarsonic acid (PAA), 4-hydroxy-3-nitrobenzenearsonic acid (roxarsone), monomethylarsonic acid (MMA), dimethylarsinic acid (DMA), arsenite or arsenious acid (As III ) and arsenate (As V ), respectively. These detection limits were improved by large-volume sample stacking with polarity switching to 32, 28,14, 42, 22, 27, 26 and 27 mg As/L for p-APAA, o-APAA, PAA, roxarsone, MMA, DMA, As III and As V , respectively. We have applied both methods to the analysis of the arsenic species distribution in aqueous soil extracts. The identification of the arsenic species was validated by means of both standard addition and comparison with standard UV spectra. The comparison of the arsenic species concentrations in the extracts determined by CZE with the total arsenic concentrations measured by inductively coupled plasma-atomic emission spectroscopy (ICP-AES) indicated that CZE is suited for the speciation of arsenic in environmental samples with a high arsenic content. The extraction yield of phenylarsenic compounds from soil was derived from the arsenic concentrations of the aqueous soil extracts and the total arsenic content of the soil determined by ICP-AES after microwave digestion. We found that 6-32% of the total amount of arsenic in the soil was extractable by a one-step extraction with water in dependence on the type of arsenic species.