Flow injection-differential pulse anodic stripping voltammetry (FI-DPASV) for As(III) and As(V) determination at parts-per-billion (ppb) level was developed with gold film coated glassy carbon as a working electrode. As(III) was determined first by FI-DPASV. Then, As(total) was determined by reducing As(V) to As(III) using 6.5 mM KI for 45 min, and As(V) was quantified as the difference. Parameters for the pre-plated gold film electrode in As(III) and As(V) determination were optimized to obtain linear calibration graphs in the range of 1.0-30 µg/l and 2.0-100 µg/l for As(III) and As(total), respectively. The limits of detection (based on 3 S y/x /m) of 0.81 µg/l As(III) and 2.03 µg/l As(total) were obtained with a precision within 4% (at 2 and 20 µg/l, n = 10) for As(III) and within 6% (at 10 and 100 µg/l, n = 10) for As(total). Analysis of the Standard Reference Material 1640 natural water standard with the certified value of 26.67 ± 0.41 µg/l yielded total arsenic concentration of 25.2 ± 0.7 µg/l (n = 3). For cations commonly found in natural water, Sb(III) was found to interfere with As(III) analysis whereas Hg(II) and Cu(II) interfere with As(total) determination. The developed method was successfully applied to natural water samples collected from arsenic contaminated areas in Ron Phibun Subdistrict, Ron Phibun District, Nakhon Si Thammarat Province in Thailand. The recoveries of the method by spiking were in the range of 85-105%. The results of As(total) in natural water samples analysed by the developed method (FI-DPASV) were in good agreement with those obtained by inductively coupled plasma-optical emission spectrometry.