A discrimination task was used to assess changes in the loudness of speech that accompanied changes in the spectral tilt of a simulated hearing aid's frequency response. Band-limited (0.25-4 kHz) spondaic words were spectrally shaped at comparison tilt-factor values of -6, 0, and +6 dB per octave and delivered monaurally via insert earphone to each of 10 listeners with normal hearing (NH) and 15 listeners with mild-to-moderate sensorineural hearing impairment (HI). Results for the NH listeners indicated that loudness differences among the tilt factors were generally perceptible and that loudness judgments were highly transitive across different tilt-factor comparisons. Loudness differences were also perceptible to many of the HI listeners when they switched among tilt factors. The HI listeners' data showed some evidence of transitivity, but not so much as was shown by the NH listeners. Intersubject variability in the loudness judgments was found to be comparable for the two subject groups. Results of the study are discussed with regard to their implications for hearing aid fitting, with particular emphasis on the "parameter adjustment and selection" fitting procedure (J. Punch & R. Robb, 1992).