Tonal thresholds and consonant recognition were measured in low-pass maskers as a function of masker bandwidth and spectrum level. Thresholds and consonant-recognition scores were obtained for normal-hearing subjects, and for pairs of normal-hearing subjects (who listened in threshold-equating background noise) and hearing-impaired subjects. Consonant-recognition scores were compared to scores predicted by a modified articulation index. Mean thresholds measured in low-pass maskers were higher for hearing-impaired than for normal-hearing subjects for signal frequencies above the masker. Slopes of functions relating thresholds for signals above the masker to masker spectrum level were not significantly different between hearing-impaired and normal-hearing subjects listening in spectrally shaped broadband noise (SSBB), but were shallower than slopes of masking functions for normal-hearing subjects listening without SSBB. Slopes of masking functions for signals within the masker were equivalent for all subjects. Slopes of functions relating consonant recognition to masker spectrum level were similar within subject pairs, whereas, in some cases, slopes of functions relating consonant recognition to speech level were shallower for hearing-impaired subjects than for their normal-hearing counterparts. Although greater improvement in consonant recognition with speech level was predicted for hearing-impaired than for normal-hearing subjects, on average, less improvement with speech level was observed. Shallower slopes of functions relating score to speech level observed for some hearing-impaired listeners may result from more shallow growth of speech sensation levels in spectral regions above the low-pass masker.