Selective removal of dilute arsenic ions (As(III) or As(V)) from acid mine drainage (AMD) containing ferrous irons was investigated using a sulfide precipitation method. 10 mg/dm3 of As(III) could be removed as As2S3 at pH ≥ 2.0 in a NaSH solution, which S/As ratio was higher than6.5 after 0.5 hr of reaction time. H2S gas generation correlated with the residual As(III) concentration in AMD. Below pH 2.5, ferrousions were not found in the sludge and thus selective removal of As(III) was achieved. On the other hand, at pH 2.5, a NaSH solution of S/As ≥ 40 and a reaction time of 24 hr were required to remove 10 mg/dm3 of As(V). However, at pH3, 10 mg/dm3 of As(V) could be removed after 0.5 hr of reaction time by using a NaSH solution of S/As ≥ 6. Nevertheless, around 1/3 of ferrousions were oxidized and removed as ferrihydrate together with As(V). EXAFS analysis showed that As(V) was adsorbed on ferrihydrate when S/As ratio was 7, but precipitated as As2S3 at pH3, when S/As ratio was 20. These results showed that reduction of As(V) to As(III) was very slow and needed enormous amount of sulfurising agent and relatively long reaction time. The average primary diameter of the particles in the sludge was under 900 nm, which was increased with increasing dosage of NaSH. Particles exhibited a negative zeta potential of small value, which indicates that they could coagulate to each other. Therefore, sedimentation rate of the particles from sulfide precipitation method was better than the one from conventional hydroxide co-precipitation method using calcite or lime. Sedimentation of As(V) sulfide particles at pH3 occurred at a very fast speed because a small amount of ferrihydrite co-precipitated with As(V) sulfide, which acted as a coagulant.