We evaluated the effectiveness of arsenic (As) hyperaccumulator Pteris vittata to continuously remove As from three contaminated-soils containing 26-126 mg kg(-1) As over 7 harvests in 3.5 years. Changes in As speciation in soils, amended with P fertilizer (P-soil) or insoluble phosphate rock (PR-soil), were assessed via sequential fractionation. Arsenic in available (soluble + exchangeable), non-labile (bound to amorphous + crystalline Fe/Al oxides), and residual fractions constituted similar to 12%, similar to 80%, and similar to 8% of soil As. Soluble As declined while exchangeable As was unchanged, likely due to replenishment from non-labile As, which accounted for similar to 87% of decline in total soil As. Although plant-available As is important, the non-labile As better predicted the frond As concentration in P. vittata, with the correlation being r=0.90 and 0.64 for PR-soils and P-soils. P. vittata removed 44% of soil As from PR-soils compared to 33% from P-soils, suggesting the low-soluble P from PR was more effective than P fertilizer in enhancing As uptake by P. vittata. To facilitate acquisition of P from PR, P. vittata produced larger root biomass to solubilize non-labile As, allowing for more efficient phytoextraction. Published by Elsevier B.V.