Paired grain, shoot, and soil of 173 individual sample sets of commercially farmed temperate rice, wheat, and barley were surveyed to investigate variation in the assimilation and translocation of arsenic (As). Rice samples were obtained from the Carmargue (France), Doñana (Spain), Cadiz (Spain), California, and Arkansas. Wheat and barleywere collected from Cornwall and Devon (England) and the east coast of Scotland. Transfer of As from soil to grain was an order of magnitude greater in rice than for wheat and barley, despite lower rates of shoot-to-grain transfer. Rice grain As levels over 0.60 microg g(-1) d. wt were found in rice grown in paddy soil of around only 10 microg g(-1) As, showing that As in paddy soils is problematic with respect to grain As levels. This is due to the high shoot/soil ratio of approximately 0.8 for rice compared to 0.2 and 0.1 for barley and wheat, respectively. The differences in these transfer ratios are probably due to differences in As speciation and dynamics in anaerobic rice soils compared to aerobic soils for barley and wheat. In rice, the export of As from the shoot to the grain appears to be under tight physiological control as the grain/shoot ratio decreases by more than an order of magnitude (from approximately 0.3 to 0.003 mg/kg) and as As levels in the shoots increase from 1 to 20 mg/kg. A down regulation of shoot-to-grain export may occur in wheat and barley, but it was not detected at the shoot As levels found in this survey. Some agricultural soils in southwestern England had levels in excess of 200 microg g(-1) d. wt, although the grain levels for wheat and barley never breached 0.55 microg g(-1) d. wt. These grain levels were achieved in rice in soils with an order of magnitude lower As. Thus the risk posed by As in the human food-chain needs to be considered in the context of anaerobic verses aerobic ecosystems.