People speak more loudly in a noisy room or when momentarily deafened and more softly in a quiet room or when sidetone is artificially increased. The effort to compensate for these changes in the signal-to-noise ratio, or to match directly changes in the intensity of a model, typically falls about halfway short (in decibel units). This is probably because a speaker considers that he has doubled his own vocal level in half as many decibels as it takes to double the loudness of the signal or the noise. More concisely, the Lombard-reflex, sidetone-penalty and cross-morality matching functions have exponents of about one-half because the exponent of the loudness scale is haft that of the autophonic scale of voice level. This amounts to saying that the speaker matches changes in signal or in noise to keep the signal-to-noise ratio nearly constant, but he is misled by the disparity in the sensory operating characteristics of speaking and listening.For a subject to perceive a doubling in loudness, the sound-pressure level mnst be more than tripled if generated by an external source, but less than doubled if generated by the subject himself, vocally. The former relation is well known as the sone scale (Stevens 1); the latter is becoming known as the autophonic scale, a term coined in the pages of this JOURNAL by Lane, Catania, and Stevens," who described the speaker's perception of his own vocal level in some detail.When a speaker judges the dynamic characteristics of his own speech, the possible sources of cues include airborne sound (air sidetone), head sidetone, and proprioception. When the speaker stops talking and listens to someone else instead, he is deprived of most of these cues and, as a listener, he must base his judgments differently. Since the sensoD' characteristics of speaking and listening axe so different structurally (yon B•k•ssa), it is not surprising to learn that they are quite different functionally. After establishing that the exponent (slope, in log-log coordinates) of the antophonic scale is approximately 1.2 whereas that of the sone scale is 0.6, Lane, Catania, and Stevens went on to confirm this disparity by asking subjects to match their vocal levels to changes in the level of a criterion sound. Given In complementary experiments, speakers were instructed to compensate for, rather than match, the changes in the loudness of a criterion sound, so that the loudness would be held constant. The criterion sound was the speaker's own voice, fed back to him at vaxious levels in an interphone system, and his task was to vocalize so as to compensate for changes in amplification introduced into the sidetone channel. Again the