The speech-recognition ability of 6 listeners with permanent noise-induced hearing losses above 1 kHz was evaluated with the Pascoe High-Frequency Word List spoken by a female talker and presented in the field. Experiment 1 presented the words at 5 intensity levels through 5 different amplification systems (1 with a uniform reponse and 4 with increasing amounts of high-frequency emphasis). Experiment 2 modified the frequency response of the system associated with the highest scores in experiment 1 in 4 ways: elimination of frequencies above 6.3 kHz, addition and deletion of 1/3-octave emphasis at the low-frequency boundary of the hearing loss and addition of 6-dB high-frequency emphasis. The frequency response associated with the highest word-identification score for each listener had from 20-33 dB more gain in the frequency region of the hearing loss than the uniform response. Three major factors affected the word-identification scores: audibility of the speech energy, separation of the 3rd-octave-band levels of the words from the discomfort threshold and balance between the low- and high-frequency levels of the speech signal.