114 BELL SYSTEM TECHNICAL JOURNAL 



the mean dimension of the adjacent volume, in order that the usual 

 resonator theory may apply. 



The model when gently blown with a slow current of air through the 

 small hole in the back gave a good whispered a; but some difficulty 

 was experienced in exciting it correctly for a voiced a. It was first con- 

 nected, at the rear, to an artificial larynx,^ keeping the connecting hole 

 small in order to preserve the dynamical characteristics of the main 

 system. When the artificial larynx was blown (though it did not 

 function well with the output orifice so small) , a recognizable voiced a 

 was produced by the apparatus; but this was not as good as the 

 whispered sound first described. (We have here the point made at 

 the beginning: that the driving system, to imitate the vocal cords 

 successfully, must give a low pitched tone, very rich in partials.) The 

 artificial larynx was then replaced by a telephone receiver excited by 

 the (rip) saw-toothed A.C. wave of 100 fundamental frequency, ar- 

 ranged by Mr. Sacia. A rather poor sustained a sound resulted, quite 

 deficient in volume, because of weak driving through the small hole in 

 the back. Altogether, the artificial larynx, with its intermittent or 

 variable excitation, came the nearest to producing a voiced a; and the 

 sound was similar to that produced by a person actually using the 

 artificial larjmx inserted in the side of the mouth, in the usual manner, 

 for this sound. 



Very fair results were also obtained with a model, built according to 

 specifications, for the sound long o. Models were next attempted 

 for short a and short e. First, a model was made with two volumes 

 Fi = 80 cu. cm., F2 = 45 cu. cm., and having the three openings 

 Ki, Ko, and the hole in the rear of V2 (for a cork fitting connecting the 

 larynx) each about 2.5 cm. in diameter. It was thought that, when 

 blown from the rear of V2, it would give a recognizable short a sound; 

 and that when reversed, i.e., when the cork fitting was inserted in Ki 

 so that Vi and V2 became interchanged, it would give short e. The 

 result was that the sounds produced were nearly alike, and quite un- 

 satisfactory in both cases! However, when the conductivities were 

 modified, so that Ki = 2.5, K2 = 2.0, for short a, and Ki = 2.0, 

 K2 = 2.5 for short e, the volumes being interchanged as before, the 

 results were much better. As described here, the model for shoit e 

 approximates in dimensions the data entered in Fig. 5, but the model 

 for short a {Vi = 80, V2 = 45 cu. cm.) does not quite have the the- 

 oretical division of total volume (namely, Vi = 100, V2 = 23 cu. cm.) 

 entered in the chart. The partition was therefore moved back, until 

 this condition was obtained, with the result that the short a sound 

 was given at least as well as before. 



" Previously described by H. Fletcher and C. E. Lane. 



