The monaural free-field speech-reception threshold (SRT) without and with a hearing aid was investigated for conversational sentences presented to 50 hearing-impaired listeners. SRT without a hearing aid was measured in quiet and in noise levels of 40, 55, 70, and 85 dBA. SRT with a hearing aid was obtained in quiet and at noise levels of 25, 40, 55, and 70 dBA. The noise had a long-term average spectrum equal to that of the sentences. The 50 subjects were equally distributed over five degrees of pure-tone hearing loss and five types of hearing impairment (sensorineural high-frequency losses, with or without recruitment; flat audiogram of a sensorineural, mixed, or conductive origin). It is shown that a model of SRT as a function of noise level, developed by Plomp [J. Acoust. Soc. Am. 63, 533-549 (1978)] gives a good description of the SRT values measured, both without and with a hearing aid. The data illustrate that, generally, current hearing aids do not improve speech intelligibility in noise beyond, roughly, 60 dBA.