AbstractThe spatial and temporal variability of the Palmer drought severity index (PDSI) over the south‐eastern United States during the period 1931–1982 was examined through the use of principal component analysis. Spatial mapping of the component loadings identified six statistically significant modes of variability or PDSI anomaly patterns, which together account for 77.3 percent of the total variance. The first component, which represents the dominant pattern, accounted for 43.4 per cent of the total variance and identified an in‐phase oscillation across the entire study area, indicating a general tendency for all sections to experience either a drought or a wet spell concurrently. The remaining five components, which account for 33.9 per cent of the total variance, each identified an out‐of‐phase oscillation. Application of Kaiser's Varimax orthogonal rotation on the six retained principal components led to the identification of six contiguous subregions of nearly equal size, each of which displays unique drought characteristics.Examination of the time series derived from the principal component scores within each subregion revealed a general trend in which the occurrence of drought was much more prevalent during the first three decades of the study period, culminating during the mid 1950s when most of the subregions experienced their worst drought episode.After removal of the inherent ‘red noise’ persistence found in the monthly PDSI data, spectral and cross‐spectral density analysis performed on the time series indicated that no cyclical patterns or periodiciti