A least squares method for analyzing multiple, ground-based measurements of electric field changes produced by lightning has been applied to seven small thunderstorms that occurred on July 19 and August 9, 1991, during the Convection and Precipitation Electrification (CaPE) experiment. Two of the storms produced little or no cloud-to-ground (CG) lightning, and a third produced CG lightning only during its early stages. A total of 79 flashes were analyzed on July 19 and 315 on August 9. About 58% of these discharges could be fitted to either a point charge (Q) or a point dipole (P) model, and in this sample, the spatial pattern of the Q and P solutions was consistent with a tripole or double dipole charge pattern in the cloud. In cases where there was little or no CG lightning, the inferred region of upper positive charge was quite close to the inferred negative region. Comparisons of the locations of upper P solutions with measurements of radar reflectivity at S band show that the P solutions tended to cluster at altitudes where the reflectivity was between 25 and 35 dBZ. Comparisons of Q (and P) model solutions with the locations of CG flash strike points, determined using a network of wideband direction finders, showed an average horizontal displacement of 3.9 km with a standard deviation of 3.3 km. This displacement is consistent with the expected random development of a stepped-leader channel from the altitude of the negative Q region to the ground. ¿ American Geophysical Union 1996