Surface water balance computations were applied to historical data from the central and southeastern United States. The focus of the study was to characterize soil water variability across spatial scales in these two different parts of the United States. The regions studied in each area span the range of spatial scales from 106 km2 to 103 km2. Routinely collected precipitation, temperature, pan evaporation, and discharge data were utilized together with widely used parameterizations to reconstruct soil water (depth-integrated and areal-averaged soil moisture) for the period from January 1960 through December 1988. The results suggest that in the central United States, larger regions possess a greater range of temporal soil water variability than smaller embedded regions and that extreme soil water anomalies possess scales of spatial coherence that range from 105 km2 to 103 km2, with stronger anomalies characterized by smaller scales. In the Mississippi-Missouri region, variability of aggregate soil water cannot be fully characterized by soil water variability in a few “representative” catchments. Remarkable similarity of the spatiotemporal variability of soil water across scales (106 km2 to 103 km2) was found for the southeastern United States. In this area the variability across scales was characterized by a frequency scaling ratio which decreased linearly with soil water anomaly strength and which remained substantially higher than that corresponding to random occurrence, even for extreme soil water anomalies. In both regions of the United States and for all spatial scales, extreme positive anomalies show smaller spatial coherence than negative anomalies of the same magnitude.