Normally sown in March in the region of Meknès (Morocco), rainfed sunflower suffers from a severe water deficit from anthesis which seriously affects grain filling. Increasing the stored soil water by appropriate management during the long period of bare soil preceding sunflower planting could be an opportunity which has not been explored for this spring-sown crop. Five methods for autumn soil tillage (mouldboard ploughing, chiselling, paraploughing, disc harrowing, no tillage) and four fallowing methods (chemical weed control, mechanical weeding, allowing weeds and volunteer crops, sowing barley) were compared in Meknès between 1994 and 1998 on calcimagnesic soils with vertic behaviour. Two additional experiments were carried out in 1997 and 1998 to create a range of leaf area indexes and transpiration requirements for sunflower. This was obtained (i) in 1997, by four levels of plant density (2.5–10.0 plants/m2) and three levels of soil water at planting (89, 37, and 29% of total available soil water); (ii) in 1998, by six levels of sunflower defoliation at star bud stage. Simulations with the EPIC-Phase model were performed to explore a wider range of weather conditions (1960–1998) than experienced. The differences in water storage at planting were explained partly by the mode of action of each of the implements tested and partly by the weather conditions which prevailed during the fallow period. After a very dry fallow period (with a frequency less than 1 year in 10), water storage was maximal after disc harrowing and paraploughing (including straw mulching) because soil layers were only marginally exposed to evaporation. Conversely, in a year with a wet fallow period (with a frequency of 4 years in 10), mouldboard and chisel ploughing gave the largest water reserves at planting because of better infiltration at depth with increased porosity. When the fallow period was initially wet, but dry in early spring (with a frequency of 2 years in 10), minimum and no tillage gave the best water storage but the differences between tillage methods were small. In spite of differences in soil water content at planting and clear differences in rooting systems, sunflower yield and seasonal water use were not significantly affected by soil tillage provided that the plant population was the same and weed control was adequate in reduced tillage systems. However, chisel ploughing was a good compromise for maximising stored water at sunflower planting on the clay soils of Meknès. Surprisingly, maximizing soil water content at sunflower planting was not systematically the best solution for maximizing sunflower yield and water use efficiency under the semi-arid conditions of Meknès. A high soil water content at planting leads to excessive leaf area index at the bud stage and consequently to rapid water depletion and yield reduction, especially when seasonal precipitation is low. A 50% refilling of the soil water reserve is sufficient for spring-sown sunflower as was confirmed by the simulation study. Soil moisture in the uppermost layer which governs seedling establishment is a more limiting factor for sunflower yield than total soil water content at planting.