Direct evaporation from soil is an important component of crop water balances in semi-arid environments. The effects of a crop and of crop management on this water loss from the soil have been estimated in the past using combinations of field measurement and simple models, but there are inconsistencies in the conclusions reached. This paper presents data from water balance studies on millet crops in Niger during the 1991 and 1993 seasons. Evaporation from soil (E(s)) was measured under two contrasting cropping intensities in both years using the microlysimeter method, Small seasonal reductions in E(s) from the higher intensity crop were recorded (12% and 16% in 1998 and 1993, respectively). Significant reductions in daily E(s) were: (1) nearly all recorded within a limited period in the season when there were large differences in transpiring leaf area; (2) recorded for both high (> = 2 mm day(-1)) and low (<0.8 mm day(-1)) values of E(s). These data indicate that soil drying by root water uptake contributed to the reduction of E(s). Increased shading of the soil by the crop canopy does not result in a proportional reduction of E(s). Two simple models for estimating E(s) beneath crops (Ritchie, 1972; Cooper et al., 1983) are compared with field data and an improvement to the Ritchie model is suggested. Two new parameters are introduced to estimate the relative importance of (1) the atmospheric vapour pressure deficit to potential evaporation and (2) root water uptake to soil drying. The brief description of environment and crop included in the new approach allows identification of the environments in which there is scope for substantial reduction in E(s) through crop management.