250 CARNEGIE INSTITUTION 



It was found that a combination of soft walled resonators with a 

 yielding source of tone was adequate to produce all the vowels. 



These experiments led to two conclusions: first, that one element 

 in the nature of a vowel is a modification of the action of the vocal 

 cords themselves, and, second, that it is possible to manufacture a 

 machine that will sing the vowels. 



With the first conclusion at hand the study of the curves was re- 

 newed from a different point of view and new methods of analysis 

 were devised. The curves were found to be explainable on this 

 theory. We thus conclude that in speech the vocal cords do not 

 execute their vibrations independently of the adjustment of the cavi- 

 ties above them, but that the form of vibration differs for different 

 vowels. Thus, when the vowels a, e, i, etc., are spoken at the same 

 pitch, the vocal cords vibrate so as to emit puffs of air of different 

 force and consequently of different timbre. One element [in the 

 nature of a vowel therefore consists of a modification of the vibration 

 of the vocal cords — a result that, as far as I am aware, has never 

 been suggested before. The consequences for phonetics, laryngol- 

 ogy, and vocal culture have not been deduced. 



The element of soft resonators must be considered at this point. 

 The vocal cavities have hitherto been treated like resonators of brass 

 or glass and conclusions have been deduced on this basis. L,ord 

 Rayleigh, for example, states that for very high tones the Hermann 

 view of the independence of the cavity tone from the cord tone may 

 be correct, but that for tones within an octave or two of the cord 

 tone the Helmholtz overtone theory is certainly valid — a statement 

 that would hold good for cavities made of brass or other hard mate- 

 rial, but that is contradicted by every vowel curve obtained. The 

 explanation is simple : the flesh forming much of the walls of the 

 vocal cavities is soft, and nearly all the rest is covered by a moist 

 membrane. The phenomena of resonance are quite different. As 

 stated before, a brass resonator responds only to a tone or tones of 

 definite pitch, whereas a resonator with walls of water responds to 

 any tone. A series of artificial heads made of metal with the proper 

 adjustments of the cavities could not be made to speak the vowels, 

 but a series made of gelatine or similar material should do so. A 

 mathematical treatment of the laws of resonance for soft cavities has 

 not yet been developed, although it would be of value. 



The theory based on the experiments in making vowels and the 

 conclusive evidence for the speech curves shows how we must pro- 

 ceed in order to make an efficient vowel machine. The materials I 



