Farmers in arsenic (As) contaminated areas of West Bengal, India grow rice during dry months (January to April) and use underground water for irrigation with As concentration above WHO defined critical (0.01 mg l −1) limit. In each season they add 50–150 mg As per m 2 soil area. Thus growing rice under deficit irrigation in these areas will reduce As load in soil–root–shoot–leaf–grain continuum of rice ecosystem. Suitable deficit irrigation system has to be screened so that As load will decrease with insignificant reduction in grain yield. With this objective, rice grown under four irrigation regimes (i) continuous ponding (CP), (ii) intermittent ponding (IP), (iii) saturation (SAT) and (iv) aerobic (AER) was tested to assess the arsenic load in soil and various parts of rice on 45 and 80 days after transplanting (DAT). Conditions described in treatments ii, iii and iv were imposed during 15–45 DAT. Highest value (18.18 and 18.74 mg kg −1) of soil arsenic was attained under CP followed by IP, SAT and AER. Root arsenic content under AER at 45 and 80 DAT was at the lowest level (6.14 and 20.54 mg kg −1) and this was 31 and 7.0% lower as compared to CP. As content in leaf and grain attained the lowest values under IP. Grain yield insignificantly differed under IP (4.33 Mg ha −1) over CP (4.69 Mg ha −1). Compared to soil As, As added through irrigation showed stronger relationship with As status of various plant parts. Imposition of IP only during vegetative stage was found to be optimum in terms of reduction of As content in straw and grain respectively by 23 and 33% over farmers irrigation practice with insignificant decrease in grain yield.