The Intelligibility of Interrupted Speech

Abstract

The paper concerns the effect of intermittent interruptions of the speech wave upon intelligibility as measured by word articulation tests. The principal experimental variables are: the interruption rate (the number of on-off cycles per second) and the speech-time fraction (the fraction of the time that the communication circuit is turned on). When the interruption rate is very low (on and off once in ten seconds), the percentage of the words heard correctly is, of course, equal approximately to the product of (a) the percentage understood in the absence of interruptions and (b) the speech-time fraction. At the other extreme, when the interruption rate is very high (on and off 5000 times a second), intelligibility is essentially unimpaired regardless of the speech-time fraction. In the intermediate range, each curve of the family relating intelligibility to interruption rate for various speech-time fractions rises to a maximum at about 15 interruptions per second, then falls to a minimum in the neighborhood of 300 interruptions per second. With a sound-time fraction of 0.50 and 15 interruptions per second, the listeners hear almost all of the words despite the fact that the speech is, and sounds, intermittent. A curious effect is produced when, with the same speech-time fraction and the same interruption rate, the speechless intervals are filled with white noise. No longer does the speech sound intermittent; instead, it is as though the speech were the scene behind a picket fence, viewed by a moving observer. The speech sounds continuous, and the words are understood just as well as they were before the noise was introduced. In the experiments just described, the interruptions occurred at regular intervals and the speech was turned on and off abruptly. Other experiments were conducted with iterruptions at irregular intervals and with gradual transition from “on” to “off” and vice versa. The disturbance produced by irregular interruption sounded very much like static and was therefore readily distinguishable from the effect produced by regular interruption. (At high rates, the latter has a tonal quality.) The intelligibility scores, however, were approximately the same with irregular as with regular interruptions. Changing from abrupt to gradual modulation (holding constant the speech-time fraction as measured at a level corresponding to one-half the peak speech amplitude) improved both the quality and the intelligibility of the interrupted speech.