Experiments were performed to determine the ability of human listeners to discriminate between a sound with a large number of spectral components in a band, of given characteristic frequency and bandwidth, and a sound with a smaller number of components in that band. A pseudorandom placement of the components within the band ensured that no two sounds were identical. The data suggested that discrimination is primarily based upon the perception of temporal fluctuations in the amplitude envelope of the sound and secondarily upon a steady-state tone color. Experiments using clusters of complex sounds showed that listeners are able to use the information in harmonic bands to discriminate spectral density. [Work partially supported by NIH, the NSF, and the CNRS.]