A method of reverberation suppression and characterization, called the principal components inverse (PCI) method, has been used to separate strong components of the reverberation from the observed data of an active sonar reverberation experiment. The number of principal reverberation components needed to represent these strong components seldom exceeds three, illustrating the fact that, locally in time, segments of the reverberation waveform that have durations much shorter than a transmitted pulse can be represented as different linear combinations of the same two or three basis functions. These basis functions change as we move in time to some other local region. Because of the interest in the physical structure of reverberation, it is wished that the reverberation modeled by these basis functions be explained in terms of the physical components of reverberation models. The PCI method separates a beamformed acoustic signal into what the PCI method regards as a ‘‘strong reverberation component’’ and ‘‘other components.’’ Therefore both the received signal and the corresponding PCI estimates of the strong reverberation have been approximated by linear combinations of delayed, scaled, and Doppler-compressed replicas of the transmitted signal. It was found that the strong signal components of the observed reverberation and the PCI estimated strong reverberation match well.