ENERGY DISTRIBUTION IN SPEECH \27 



seems to show that the choice of test material does not require especial 

 consideration, provided it is of sufficient length. It seems to be a 

 matter of rather greater importance to increase the number of 

 observers. 



In the second place, it seems that for the actual energy distribution, 

 the results previously obtained from the vowel analyses are definitely 

 in error, in that they show relatively little energy associated with the 

 lower voice frequencies. 



Criticism of the Results 



The foregoing treatment provides a curve showing the frequency 

 distribution of the square of the excess pressure on the diaphragm. 

 In an undisturbed field of sound energy we have for the intensity 



2 pa 



in which p is the mean density of the medium, a the velocity of sound 

 in the medium and P the maximum excess pressure. 



It remains for us to consider in how far the results obtained represent 

 the frequency distribution of sound energy in speech. 



Due to the fact that at frequencies where the sound wave-length is 

 short and comparable with the diameter of the transmitter, consid- 

 erable reflection takes place, and the pressure on the diaphragm is 

 proportionately greater for these frequencies than for those which are 

 not accompanied by strong reflection. In this respect again the 

 higher frequencies provoke the greater response in the system. 



The following experiment was tried to investigate this variation. A 

 wall six feet square, with a central hole to fit over the condenser trans- 

 mitter, was brought up to make the transmitter a part of a plane wall. 

 The clearance around the periphery of the transmitter was tightly 

 closed, and reflection was to be expected at all frequencies. Where 

 total reflection takes place, a given quantity of sound energy result? 

 in twice the alternating pressure on the diaphragm as when no reflection 

 occurs. That is, the resulting electrical energy observed should be four 

 times as great for total reflection as for no reflection. The wall was 

 expected to cause reflection at all frequencies, and the experiment 

 consisted in reading the electrical response, with and without the wall, 

 the condenser transmitter being -exposed to tones of frequencies from 

 200 to 10,000 cycles per second under definite adjustments of the 

 supply circuit of a receiver producing this tone. When the frequency 

 is low, little reflection takes place from the transmitter standing alone, 

 and bringing up the wall should cause a great increase in the response 



