The capacitance probe is an attractive device for monitoring soil moisture automatically. However, its sphere of influence is rather small (a few cubic centimetres only). We have analysed the possibility of monitoring moisture at the field scale using only a few probes (less than or equal to 3). We calibrated each probe by establishing a direct relation between the field average soil moisture theta(f) and the signal given by the probe. As in earlier studies, we found that a linear relation is generally suitable. A classical statistical analysis was performed to assess the error of a single probe. When replicate probes were installed, we obtained replicate estimates of theta(f). We proposed an estimator theta(f) that combines all replications optimally. Three experiments each lasting several months were carried out on bare tilled fields to evaluate the probe against gravimetric measurements. Our results show that the calibrations differ significantly from one probe to another. Once calibrated, the capacitance probe provided accurate soil moisture measurements (70% of the calibration relations had residual standard deviations < 0.02 m(3) m(-3)), but it is advisable to have at least two replicate probes. Soil water storage was well estimated by combining four to seven probes to establish the moisture profile, despite the error induced by each probe, Moreover, the temporal variations in water storage were accurately measured by the probes. We found an error of 0.6 mm day(-1) (standard deviation) in daily variation of the water storage, which partly involved the error made on the reference measurements (gravimetric method).