A simple, economic and sensitive method for selective determination of As(III) and As(V) in water samples is described. The method is based on selective coprecipitation of As(III) with Ce(IV) hydroxide in presence of an ammonia/ammonium buffer at pH 9. The coprecipitant was collected on a 0.45 m membrane filter, dissolved with 0.5 mL of conc. nitric acid and the solution was completed to 2 or 5 mL with distilled water. As(III) in the final solutions was determined by graphite furnace atomic absorption spectrometry (GFAAS). Under the working condition, As(V) was not coprecipitated. Total inorganic arsenic was determined after the reduction of As(V) to As(III) with NaI. The concentration of As(V) was calculated by the difference of the concentrations obtained by the above determinations. Both the determination of arsenic with GF-AAS in presence of cerium and the coprecipitation of arsenic with Ce(IV) hydroxide were optimised. The suitability of the method for determining inorganic arsenic species was checked by analysis of water samples spiked with 4-20 mu g L-1 each of As(III) and As(V). The preconcentration factor was found to be 75 with quantitative recovery (>= 95%). The accuracy of the present method was controlled with a reference method based on TXRF. The relative error was under 5%. The relative standard deviations for the replicate analysis (n = 5) ranged from 4.3 to 8.0% for both As(III) and As(V) in the water samples. The limit of detection (3 sigma) for both As (III) and As(V) were 0.05 mu g L-1. The proposed method produced satisfactory results for the analysis of inorganic arsenic species in drinking water, wastewater and hot spring water samples.