Motivated by the need to understand large-scale hydrologic response, significant research has been directed toward the identification of coherent regions using characteristics of streamflow variability. Typically, these regions are delineated using principal component analysis on streamflow. This method does not account for differences in temporal scales of fluctuations embedded in the time series. To capture this, we use wavelet spectral analysis. Wavelet spectra from the specific streamflow series are obtained for outflow binned at 3¿-length segments along the border of the conterminous United States. Rotated principal component analysis is performed on the wavelet spectra to obtain clusters of segments that exhibit similar distribution of variability across scales. Three physically distinct modes explain over 89% of the variability. Two of the modes identified are associated with high variability at seasonal scales, and the third is associated with high variability at small timescales. The runoff generation mechanisms underlying the observed modes of multiscale variability of various regions are also discussed. Each of these coherent modes of multiscale variability indicate the existence of regions with similar scales of fluctuations that are located geographically apart, as well as regions located geographically close with dissimilar scales of fluctuations. ¿ 2000 American Geophysical Union