A new forced-choice procedure for measuring comfortable loudness for speech was evaluated on a group of 30 normal-hearing subjects. The test-retest reliability was determined for the forced-choice ascending and descending approach and found to be high (r = 0.83 and 0.84). Clinical reliability indicated that 80 to 83% of the subjects varied no more than 5 dB upon retest. In a second experimental session, comparisons between the forced-choice procedure and the method of limits (ascending and descending approaches) were evaluated on a group of 14 sensorineural subjects. Results indicated the forced-choice procedure eliminated the measurement differences found with the method of limits (ascending and descending approaches) and the forced-choice measurements remained stable across varying degrees and configurations of hearing loss. The procedure is fast, reliable, and has several advantages over the method of limits procedure.