Detection of a change in spectral shape, or profile analysis, appears to be mediated by comparisons across widely separated frequency ‘‘channels’’ rather than by local comparisons among adjacent frequency regions [e.g., Greenetal., J. Acoust. Soc. Am.73, 639–643 (1983)]. Two experiments were conducted in order to determine the ‘‘resolution bandwidth’’ of these channels. The first involved detection of an increment to a single component of a multicomponent background as a function of the number of components in the background. Performance improved as the number of components was increased from 3 to 21. Further increases yieldedpoorerperformance and the estimate of the ‘‘resolution bandwidth’’ from these data suggests that this poorer performance was due simply to masking. The second experiment involved discrimination of a multicomponent complex having a flat amplitude spectrum from one having a sinusoidally ‘‘rippled’’ amplitude spectrum. The latter experiment yielded somewhat larger estimates of the ‘‘resolution bandwidth’’ than did the former. Finally, profile analysis was investigated under a dichotic condition that precluded peripheral masking of the signal. Our results, like those of Green and Kidd [J. Acoust Soc. Am.73, 1260–1265 (1983)], suggest that, although spectral analysis can be achieved using information across ears, performance is inferior to that obtained with diotic stimuli.