A Dynamical Study of the Vowel Sounds 



By I. B. CRANDALL and C. F. SACIA 

 I MROI ACTION 



' I ""HE stuch' of the vowel sounds presents a problem which has 

 ■'■ interested scientists and scholars in varied fields. A knowl- 

 edge of their nature is of fundamental importance not onh- in com- 

 munication engineering but also in acoustic science, phonetics and 

 vocal music. From the earliest theories and the rough experiments 

 of Willis (1829) and Helmholtz (1859) to the later measurements 

 of D. C. Miller (1916) steady progress has been made toward the 

 accurate determination of their characteristics. 



Further progress in this study has been made possible with im- 

 proved facilities now available in the telephone research laboratory. 

 It has been felt that there was need for more accurate records of 

 the spoken sounds and the development of impro\ed transmitters, 

 amplifiers and other devices has made possible recording apparatus 

 of greater accuracy, range and power than any heretofore used. 



In this paper will be given the results of an analysis of spoken 

 vowel sounds based on a set of accurate oscillographic records. The 

 recording apparatus was designed to record the wave forms of the 

 different speech sounds practically free from distortion over the 

 frequency range from 100 to 5000 cycles. A brief description of 

 this apparatus is given in the appendix. The emphasis in the present 

 paper is placed on the composite frequency characteristics of the 

 sounds as repealed by a i)arliriilar inelhnd of anaKzing the records so 

 obtained. 



.Analysis of tiiic 1).\t.\ 



Tlic iliirliin Miwel sounds inxestigated are shown arranged in 

 a triangle in I'ig. 1. The tlii)iuhongs on, K', _v and long / are not 

 included. Fight records of each sound were taken, four by male 

 and four by female speakers. In speaking these sounds the only 

 constraint imposed on the speakers was that the sound should be 

 completely uttered within an inter\-al of one second. The recording 

 mechanism was so arranged that the whole of the .sound from begin- 

 ning to end was recorded in one continuous graph. In practice the 

 average duration of these .sounds w.is about 0.30 second. Each 

 record shows a sequence of growth and decay in amplitude some- 

 what as follows: first a period of rapid growth in amplitude lasting 

 about .04 second during which all components are quickly produced 



232 



