1936 年 11 巻 6 号 p. 549-556
Problem: It is well-known that, under noisy circumstances, we are disturbed in the perception of speech sound. The purpose of this study was to investigate the relation between the loudness of noise and the magnitude of disturbance in telephone hearing.
Method: Materials used were composed of (1) 100 one-syllabled nonsense words, (2) 100 two-syllabled nonsense words and (3) 100 two-syllabled words. The experimenter read slowly these words before a telephone. Cautions were paid not to alter the pitch and the loudness of voice. The observer, at another telephone, recorded the words as he heard them. As the source of noise we used electrical bells attached to a stand near by the observer. By controlling the current with a rheostat, we could change the loudness of noise in 3 steps, 3 phon, 4.6 phon and 7.2 phon respectively.
Results : In Fig. I in the Japanese text (See p. 551) is represented the relation between loudness of noise and magnitude of disturbance. The lower curves show the percentage of errors, the upper curves the percentage of cases when the subjects could not respond. The abscissa shows the loudness of noise in phon (Barkhausen phon). The disturbance increases linearly with the loudness of noise and is greatest in one-syllabled nonsense words, slightest in two-syllabled words. If we show this relation for each sound separately we get Fig. 2. (See p. 552 in the Japanese Text). The curves show the composite results of the errors and the cases when the subjects could not respond. As will be seen, the vowels, except for the case when the noise was 7.2 phon, suffer practically no disturbance. Among the consonants P is the most disturbed; and nearly all consonants show almost linear increase of disturbance as the noise increases. Averaging these results for each sound, we obtain Fig. 3 (See p. 553 in the Japanese Text). From it we can see that all sounds can be divided into 3 groups according to the magnitude of disturbance they suffer; (1) P, (2) all other consonants and (3) vowels, semivowels and ending N. In the next place, the mutual relation of affinity among sounds as represented by errors can be shown as in Fig. 4 (See p.554 in the Japanese Text) (for vowels and semivowels) and Fig. 5 (See p.555 in the Japanese Text) (for consonants). As for the former, we can see that (1) neighbouring sounds in vowel-triangle tend to be mistaken for each other, (2) all sounds tend to be heard as O and as U. As for the latter, (1) voiced consonants and unvoiced consonants make separate groups, (2) they tend to be heard as G, B and K respectively. In the last place attempts were made to explain these results with the masking effect.