A leaching test conducted in field lysimeters for the purpose of pesticide registration is evaluated, particularly in terms of factors such as the effects of soil type, variability in leaching between replicate lysimeters, and simulation of worst-case scenarios. Two herbicides, dichlorprop and bentazon, were chosen as test compounds due to their documented high mobility in laboratory tests. Four different soil types (sand, loam, clay, peat) and two irrigation treatments were included. Both herbicides were applied at rates representing normal doses (1.6 and 0.6 kg ai ha−1 of dichlorprop and bentazon, respectively). 36Cl was also applied to sand and clay lysimeters to follow the pattern of water movement. Leaching of dichlorprop for the varying soil type/treatment combinations ranged from 0.02 to 1.8% of the amount applied. Leaching losses of bentazon reached up to 0.07% of that applied. Leaching of both herbicides was greater mostly in clay monoliths than in sand monoliths, which was explained in terms of macropore flow. A more effective macropore flow was also suggested to be the main reason why more dichlorprop leached in clay and peat monoliths treated with a small water input. Detectable, and in some cases large, concentrations of dichlorprop were found in the first drainage water in early autumn in all soil/treatment combinations, indicating the occurrence of preferential flow in all soils tested, including sand. A rapid breakthrough of 36Cl was also found in clay and low-irrigation input sand, providing additional confirmation of the role of preferential flow processes in these soils. It is concluded that field mobility tests for pesticide registration are a necessary complement to measurements of physical/chemical properties of a compound and that these should be performed in a range of soil types, including at least one structured soil. Other factors identified to be of importance when evaluating lysimeter studies such as this were the analytical detection limits of the pesticides and the need for replication.