Research in estimating the water status of crops is increasingly based on plant responses to water stress. Several indicators can now be used to estimate this response, the most widely available of which is leaf water potential (Psi(LWP)) as measured with a pressure chamber. For many annual crops, the predawn leaf water potential (Psi(PLWP)), assumed to represent the mean soil water potential next to the roots, is closely correlated to the relative transpiration rate, RT. A similar correlation also holds for young fruit trees grown in containers. However, exceptions to this rule are common when soil water content is markedly heterogeneous. Two experimental conditions were chosen to assess the validity of this correlation for heterogeneous soil water content: 1) young walnut trees in split-root containers. The heterogeneity was created by two unequal compartments (20% and 80% of total volume), of which only the smaller was irrigated and kept at a moisture content higher than field capacity (permanent drainage). 2) adult walnut trees in an orchard. In this case, soil water heterogeneity was achieved by limiting the amount of localised irrigation (20% of the irrigated control) which was applied every evening. Values of sap flux and of minimum and predawn leaf water potentials with homogeneous and heterogeneous soil water content were compared for trees grown in the orchard and in containers. In spite of intense drought reflected by very low RT or stem water potential, Psi(PLWP) of trees under heterogeneous moisture conditions remained high (between -0. 2 and -0.4 MPa) both in the orchard and in containers. These values were higher than those usually considered critical under homogeneous soil conditions. A semi-quantitative model, based on the application of Ohm's analogy to split-root conditions, is proposed to explain the apparently conflicting results in the literature on the relation between Psi(PLWP) and soil water potential.