Tensiometer measurements were carried out on a typic Paleudult in the humid forest zone of south eastern Nigeria in an alley cropping trial using fertilized and unfertilizedDactyladenia (Acioa) barteri andSenna (Cassia) siamea as hedgerow trees and a no-tree control. The interrow space of alley cropped and no-tree control plots were planted to maize/cassava intercrop. Water withdrawal during short dry spells and the dry season occurred fastest in the no-tree control plot and resembled the pattern in the adjacentS. siamea alley cropping. Previous root investigations showed that the whole no-tree control plot was within the range of root propagation of the adjacent hedgerow trees. Installation of a 70 cm deep root barrier led to a retarded water withdrawal in unfertilized no-tree control plots to a depth of 150 cm. In fertilized no-tree control this retardation occurred to a depth of 110 cm, while at 130 and 150 cm water withdrawal with root barriers was faster than without barriers. Results indicate thatS. siamea depleted water resources in the no-tree control plot and shortened the growing season of cassava. Restricting roots to the assigned plot size can reduce competition for water in adjacent plots even in layers below the depth of the barrier but can also induce compensative water withdrawal from layers which were not necessarily affected by the barrier. It appears that currently no standard methodology is available to conduct agroforestry trials without the risk of invalidation through root interference. Methods to determine minimum plot size in order to reduce the risk of invalidation and misinterpretation of results are suggested.