Five computer-synthesized broadband noises, each having the same average spectrum and the same unweighted Leq of 100 dB SPL but very different temporal structures, were used to produce hearing loss in chinchillas. Despite the same exposure energies and spectra, each noise exposure produced a different magnitude and frequency distribution of hearing loss and sensory cell loss. The results indicate that the statistical properties of a signal are important in the determination of hearing loss. When the audiometric and histological results are compared to a metric based upon kurtosis measured in the time and the frequency domain for each exposure, there is a clear indication that these statistical metrics are good predictors of the relative magnitude and frequency distribution of the acoustic trauma.