Six subjects have participated in a study to determine the relationship between actual and predicted performance-intensity functions for nine conditions of high-pass, low-pass, and band-pass filtered speech. Three subjects have bilateral sensorineural hearing loss of varying degrees and etiology, one subject has a unilateral sensorineural loss and two subjects have normal hearing. All subjects were trained and tested on 72 consonant-vowel syllables recorded by two male and two female talkers. Each of the nine filtered conditions (as well as the unfiltered speech) was tested at a minimum of five different levels. Hearing loss in the normal-hearing subjects was simulated by using spectrally shaped wide-band noise to create masked thresholds similar to those of the hearing-impaired subjects. Measured intelligibility scores are compared with the predictions of Articulation Theory using modifications developed by Dugal, Braida, and Durlach [in Acoustical Factors Affecting Hearing Aid Performance and Measurement, edited by Studebaker and Hochberg, (1980)] [Research supported by NIH.]