MEASUREMENT OF NOISE KAYE 171 



tion level. For example, the standard frequency chosen for much of 

 the work at the National Physical Laboratory is 800 cycles, while 

 in the United States 1,000 cycles has been largely used. 



SPEECH AND NOISE 



As already remarked, the masking effect of a background of noise 

 on the ease of conversation is one of primary interest. Conversation 

 begins to be difficult when the background of noise reaches 70 or 80 

 decibels, while at 90 decibels conversation, even by shouting, is 

 virtually impossible. 



It should be remembered that the greater part of the energy of the 

 human voice is in the low-frequency region. Some 60 per cent of tlie 

 energy lies in frequencies below about 500 cycles per second, and about 

 85 per cent below 1,000 cycles per second. It is known, however, that 

 the intelligibility of speech rests largely on the high-frequency con- 

 sonants — say, above 1,000 cycles per second — rather than on the low- 

 frequency vowels (say, 120 cycles for the male voice and 240 cycles 

 per second for the female), despite the fact that the low-frequency 

 components carry the major part of the energy. 



Davis and Evans at the National Physical Laboratory have ob- 

 served that conversation in an inclosure is facilitated if high-pitched 

 noises are excluded from without. It is fortunate that such notes 

 can be more readily excluded than low notes, and particularly so, 

 where limits are set to the massiveness of the walls of the inclosure, 

 as in an aeroplane cabin. Furthermore, such higli notes as gain 

 entrance are more readily absorbed by mounting absorbent on the 

 inner walls. (Davis, Journ. Roy. Aer. Soc, 1931.) 



As regards the effects of noise on the hearing of speech, Knudsen 

 found in 1925 that if the interfering sound is a ijure note at about 

 speech level, the interference with speech is almost independent of 

 frequency, but that for greater intensities low-pitched notes interfere 

 more than high. He also states that the interfering effect of noise is 

 greater than that of a pure note whatever the pitch. Figure 4 (due 

 to Mr. Fleming) summarizes Knudsen's results on the effect of extra- 

 neous noise on the intelligibility of speech as interpreted by articu- 

 lation tests with speech of normal loudness (50 db.). As will be 

 seen, even a little noise affects speech reception adversely, while a 

 noise level of some 30 or 40 decibels reduces the intelligibility by an 

 intolerable amount. 



ANNOYANCE AND NOISE 



We have alreadj^ referred to the association of annoyance and 

 noise, and the question has recently been the subject of experiment. 

 Precise measurement could scarcely be expected, perhaps, but it is 

 clear that both frequency and loudness are among the factors which 



