A new procedure is presented for selecting the gain and frequency response of a hearing aid from pure-tone thresholds. This was developed from research which showed that a previous procedure did not meet its aim of amplifying all frequency bands of speech to equal loudness but that frequency responses which did so were considerably more effective. Measurements of 30 sensorineurally hearing-impaired ears (27 subjects), together with data from other studies, were analyzed to determine the best formula for predicting the optimal frequency response, for each individual, from the audiogram. The analysis indicated that a flat audiogram would require a rising frequency response characteristic of about 8 dB/octave up to 1.25 kHz and thereafter a falling characteristic of about 2 dB/octave. Variations in audiogram slope required about one-third as much variation in response slope. Three frequency average (3FA) gain was calculated to equal the 3FA gain of the previous procedure. Forty-four subjects (67 aided ears) fitted by the new procedure were evaluated by paired comparison judgments of the intelligibility and pleasantness of speech. The prescribed frequency response was seldom inferior to, and usually better than, any of several variations having more, or less, low and/or high-frequency amplification. On the average, used gain was approximately equal to prescribed gain. It is concluded that the new formula should prescribe a near optimal frequency response with few exceptions.