A pot experiment was conducted in a net-house of the Department of Soil Science at Bangladesh Agricultural University (BAU), Mymensingh, in 2006. Soil was collected from the BAU farm. The soil is classified as Aeric Haplaquept (US Soil Taxonomy) and contains vermiculite, mica and kaolinite as the dominant clay minerals. The soil had the following characteristics: texture - silt loam, pH - 6.5, organic C - 0.93%, total N - 0.1%, available P - 10.2 mg kg−1 (0.5 M NaHCO3 extraction); exchangeable K - 0.12 cmol kg−1 (1 M NH4OAc extraction), available S - 13 mg kg−1 (0.15% CaCl2 extraction), total soil Fe oxide — 7300 mg Fe kg−1 (citrate-dithionite extraction), and total As - 4.1 mg kg−1. The experiment consisted of a total of 36 earthen pots, each containing 12 kg soil. Arsenic was added to the pots at the rates of 0, 15 and 30 mg kg−1, combined with P and Fe at 0 and 50 mg kg−1. Thus there were twelve treatment combinations, i.e., As0P0Fe0, As0P0Fe50, As0P50Fe0, As0P50Fe50, As15P0Fe0, As15P0Fe50, As15P50Fe0, As15P50Fe50, As30P0Fe0, As30P0Fe50, As30P50Fe0 and As30P50Fe50, in three replications. Arsenic, Fe and P were added as Na2HAsO4·7H2O, FeSO4·7H2O, KH2PO4 in solutions. The initial soil-As concentration was 4.1 mg kg−1; therefore, the total As-treatment concentrations were 4.1, 19.1 and 34.1 mg kg−1, respectively. Doses of fertilizer elements were determined as per the Bangladesh National Fertilizer Recommendation Guide. Nitrogen, K and S were applied to each pot at total concentrations of 100 mg N, 60 mg K and 30 mg S per kg soil. Nitrogen was applied in three equal splits, the first split before planting and the remaining two splits at tiller- and panicle-initiation stages of crop growth. Sources for N, K and S were CO(NH2)2, KH2PO4 and CaSO4·2H2O, respectively. The amounts of P, K and S were adjusted with applications of KH2PO4 and CaSO4·2H2O. All fertilizers were added as solutions and thoroughly mixed with the soil at 1- day before transplantation. Three 35-day old rice seedlings (cv. BRRI dhan 33) were transplanted into each pot. Pots were watered regularly to maintain 3–4 cm standing water in each pot. The crop was harvested at 120 days following transplantation, when it had attained maturity. Root biomass, grain yield, straw yield, plant height, tillers hill−1, and grains panicle−1 were determined. Ironoxide plaque, visible as a reddish coating, was formed on the root surfaces. Soil was removed from the roots by careful and repeated washing in water. During this treatment, the Fe-oxide plaques remained largely intact on the root surface. Iron-oxide plaques were dissolved from the roots and analyzed for As and Fe concentrations (as discussed below). Arsenic, Fe, and P concentrations were measured in the rice-grain (dehulled whole grain), straw, and root samples. Total N content of soil was determined by the Kjeldahl method, and available P content by 0.5 M NaHCO3 (pH 8.5) extraction, exchangeable K by 1 M NH4OAc, pH 7.0 extraction and available S content by 0.15% CaCl2 extraction. Extractable P and S were analyzed colorimetrically, and extractable K was analyzed directly by flame photometry. Iron plaque on the root surface was removed by dithionite-citrate-bicarbonate (DCB) and NH4-oxalate (in the dark) extractions prior to analyses of the extracts for Fe and As. Arsenic in the DCB extracts was not analyzed because of problems of this extract with the hydridegeneration system for As analysis. Plant samples (root, straw and grain) were digested using an open vessel HNO3-H2O2 digestion method, followed by analysis of total Fe, As and P. Total As was determined by flow-injection hydride-generation flame-atomic-absorption spectrophotometry (FI-HGFAAS) (Samanta et al. 1999), using a UNICAM 969 FAAS (England). NIST SRM 1568a Arkansas, a long grain rice flour was used as a certified reference material (CRM). Total As recovery by the HNO3- H2O2 digestion method was 88±5% (n=10) of the certified reference value. The data were statistically analyzed following the principle of F statistics, and the mean values were compared using Duncan’s Multiple Range Test (DMRT) with the software package MSTATC.