In semi-arid areas under rainfed agriculture water is the most limiting factor of crop production. To investigate the best way to perform fallow and its effect on soil water content (SWC) and root growth in a barley (Hordeum vulgare L.) crop after fallow, an experiment was conducted on two soils in La Segarra, a semi-arid area in the Ebro Valley (Spain). Fallow was a traditional system used in these areas to capture out-of-season rainfall to supplement that of the growing period, usually lasting 16 months, from July to October of the next year. Soil A was a loamy fine Fluventic Xerochrept (Haplic Calcisol, FAO) of 120 cm depth and Soil B was a loamy Lithic Xeric Torriorthent (Calcaric Regosol, FAO) of only 30 cm depth. The experiment was continued for four fallow-crop cycles in Soil A and for two in Soil B. In Soil A, three tillage systems were compared: subsoil tillage (ST), minimum tillage (MT) and no-tillage (NT). In Soil B, only MT and NT were compared. In the fields cropped to barley, SWC and root length density (LV) were measured at important developmental stages during the season, lasting from October to June. In the fallow fields SWC was also monitored. Here, evaporation (EV), water storage (WS) and water storage efficiency (WSE) were calculated using a simplified balance approach. The fallow period was split in two 8-month sub-periods: July–February (infiltration) and March–October evaporation (EV). In Soil A, values of WSE were in the range 10–18% in 1992–1993, 1993–1994 and 1994–1995 fallow, but fell to 3% in 1995–1996. Among tillage systems, NT showed significantly greater WSE in the July–February sub-period of 1992–1993 and 1993–1994 fallow, but significantly lower WSE in the March–October sub-period, due to greater EV under NT. Consequently, no differences in total WSE were found between tillage systems. In Soil B, WSE was low, about 3–7%, and there were no difference between tillage systems. During the crop period, the differences in SWC and LV between tillage systems were small. Regarding yields, the best tillage system depended on the year. NT is potentially the best system for executing fallow, but residues of the preceding crop must be left spread over the soil.This work was funded by the Comisión de Investigacion Cientı́fica y Técnica (CICYT), AGR91-312 and AGF94-198 projects. We also thank the Ministerio de Educación y Cultura (MEC), which funded the doctorate studies of J. Lampurlanés and P. Angás.