NATURE OF LANGUAGE — JONES 505 



The curves of Figure 9 show the articulation of some typical 

 speech sounds when the frequency regions below or above the given 

 point are suiDpressed. The ordinate gives the number of times the 

 sound was correctly observed per 100 times called ; the abscissa, the 

 frequency of cut-off. In each figure the effect of suppressing the 

 frequencies below the cut-off is shown by the curve at the left, the 

 effect of suppressing those above it by the one at the right. The 

 diphthong " i," the long vowel " e," and the semi-vowel " 1 " are each 

 perceived with an error less than 3 per cent when either half of the 

 frequency range is used. The intersections of the two curves, the 

 cut-off frequency where the articulation is the same with either 

 low-pass or high-pass filters, are at different points in each of the 

 three cases, however. 



In the cases of the short vowels " u," " o," and " e," the frequencies 

 below 1,000 cycles are important to good articulation, but those 

 above 2,000 may be suppressed with little effect. 



In the cases of the fricatives "" s," " z," and " th " quite different 

 effects are observed than with the former two classes. Some of 

 the peculiar results shown have not yet been explained. Even 

 if all frequencies up to 5,000 cycles are correctly transmitted, these 

 sounds are noticeably impaired by the suppression of those above. 

 The lower frequencies up to 1,500 cycles contribute practically noth- 

 ing to the articulation of " s " and " z." It has been observed, in 

 the case of a system which supj^resses all frequencies above 2,500 

 cycles, that about 82 per cent of the syllables were heard correctly 

 in an articulation test, and that the errors were made up principally 

 of failures in the three sounds " s," " z," and " th." 



In conclusion then, we have seen that the ordinary ear is an 

 exquisitely developed organ for sensing minute and rapidly re- 

 peated variations in air pressure. It can perceive sound waves 

 ranging in pressure amplitude from less than 0.001 dyne to over 1,000 

 dynes, and in frequency of vibration from about 20 cycles per sec- 

 ond to about 20,000 — a range of about 10 octaves. Human speech 

 employs frequencies from a little below 100 cycles per second to 

 about 6,000 cycles — a range of about 6 octaves. The intensities 

 and frequencies used most in conversation are those located in the 

 central part of the area of audition. The energy of speech is 

 carried largely by frequencies below 1,000, but the characteristics, 

 which make it intelligible, largely by frequencies above 1,000. Under 

 quiet conditions good understanding is possible with undistorted 

 speech having an intensity anywhere from one hundred times 

 greater to a million times less than that at exit from the mouth. 

 On the whole, the sounds " th," " f ," " s," and " v " are hardest to 

 hear correctly and they account for over half the mistakes made in 



