The arsenic contamination of groundwater in the Bengal Delta plain represents a crucial problem in particular for Bangladesh and West Bengal, India. Even when alternative water sources become available for potable purposes, agriculture remains dependent mostly on groundwater and, to date, scant attention has been devoted to the fate of arsenic reaching soils with irrigation water. Adsorption/desorption reactions on/from soil colloids are one of the main mechanisms controlling the mobility of arsenic in the water-soil system. The adsorption of As in soil depends mostly on the nature of soil colloids, pH, redox conditions. Arsenite, more toxic than arsenate, is also reported to be more mobile and thus more bioavailable. Both As forms can be found in water-soil environments in a wide range of redox conditions and their co-presence is particularly likely in paddy rice fields. The aim of the present work was i) to characterize As fractions from a rice field (Rice) and a soil cultivated with legumes (Legume), both irrigated by As polluted groundwater, before and after partial saturation with arsenite –As(III)- or arsenate –As(V); ii) to study the adsorption of As(III) and As(V) and their potential release from the soils. The arsenic was sequentially extracted before and after 25% saturation of their adsorption capacity with As(III) or As(V), as deduced from adsorption isotherms. The potential As release was studied using Fe oxide impregnated filter papers as infinite sink on the soils saturated at 25, 50 or 75% with arsenite or arsenate. The mean concentration in the irrigation water was 0.3 mg As L^-1. The two soils were calcareous, silty-clay-loam or silty-loam, with scarce organic carbon and medium CEC. The presence of poorly crystalline iron oxides was more important in the surface layer of the rice field. Total As content (8-11 mg kg^-1) was in the average for Bangladesh soils. It was mainly bound to secondary minerals: mainly to poorly crystalline Fe oxides in the Rice soil and to crystalline oxides in the Legume one. The scarce residual As fraction suggested a low presence of primary minerals rich in As. The soluble As fraction was low, while up to 34% was exchangeable with phosphate. Thus, although the already adsorbed As represented less than 1% of the potential adsorption capacity, according to the adsorption isotherms, it showed some potential mobility. More As(III) than As(V) was adsorbed by both soils and the adsorption capacity of the Rice soil was always higher than that of the Legume soil. The added As was mainly accumulated as phosphate-exchangeable fraction (55-67%), and 25-30% was readily soluble. The P-exchangeable fraction was higher in the Legume than in the Rice soil. The distribution of As(V) was quite similar to that of As(III) in the paddy soil, while in the Legume soil As(V) was adsorbed on soil oxides at higher extent than As(III). The high mobility of both As forms was confirmed by the extraction with Fe oxide impregnated filter papers: 60-70% of the added As(III) was desorbed within 24 hours, even at the lowest saturation degree. Similar percentages of the added As(V) were desorbed from the paddy soil, while from the legume field samples As(V) desorption was less than 40%. The studied soils show a relatively high arsenic mobility that further increases after As addition. A high mobility is likely to occur especially in the rice field, where the main adsorbing phases are represented by poorly crystallised Fe oxides, sensitive to the reducing environment. Thus, even if arsenic is added to the soils with the irrigation water in concentrations much lower than their maximum adsorption capacity, it can be relatively easily mobilised, becoming bioavailable. Special attention should be also paid to the P fertilisation, since this anion can exchange a high percentage of the adsorbed As.