This experiment explored the ideas that energetic masking (EM) limits thebenefit achievable from spatial separation in a competing speech task andthat differences in EM between normalhearing (NH) and hearingimpaired (HI)listeners can explain differences in spatial release from masking (SRM). Targetsentences and similar masker sentences (or in another condition, masker noises)were filtered into four narrow spectral bands and presented simultaneously.The spectral overlap of the target and masker bands was varied from minimal(interleaved bands) to maximal (identical bands). Both NH and HI listenersshowed spectral tuning, such that masking by noise or speech was reduced asspectral overlap was reduced. In fact, this signal processing often improvedintelligibility relative to that seen with unfiltered speech. Moreover, thresholdsin the noise and speech maskers were closely related, suggesting that EM wasthe main factor limiting performance across tasks. More masking was foundin HI as compared to NH listeners for all spectral overlaps and both maskers,and the HI group generally had less SRM. However, variations in spectral overlapdid not lead to dramatic changes in SRM or in the difference in SRM observed between groups.