632 



NATURE 



[January 13, 192 1 



Nature of Vowel Sounds 

 By Prof. E. VV. Scripture. 



The Analysis of J'owel Curves. 



SINCE the time of VVheatstone and Helmholtz 

 the vowels have been almost universally 

 supposed to obtain their tones by acting as re- 

 sonators to certain overtones of the larynx tone. 

 Helmholtz even constructed an apparatus of a set 



Rotator 



Near Drum 



Pulley for Rotating 

 Tube 



Pulley for 

 Rotation 



Pulley for side 

 movement 



Fig. I. — Apparatus for tracing gramophone curves. A steel needle near one end of a long lever 

 follows the groove. lis movements are enlarged 500 times and registered on a band of 

 smoked paper. 



of harmonic tuningf-forks by combinations of 

 which he hoped to produce the vowels. Ever since 

 the invention of the speech-recording machine b) 

 Scott and Koenig in Paris the analysis of vowel 

 curves has been expected to solve the problems of 



Fig. 2. — Vowel curves. The waves fall into groups ; the top line contains 

 eight groups, the next line six, the third seven, the fourth eight, and 

 the last seven. Each group corresponds to one vibration from the 

 larynx. 1 he length of a group gives the pitch of the laryngeal tone ; 

 in speech this is always rising or ^ailing. The height of the waves 

 indicates the intensity. This is nearly always small at the beginning of 

 a vowel ; there is a steady rise to a maximum and then usually a fall 

 to the end. The small waves within a group give the characteristics of 

 the vowel sound. The top line is » piece out of the middle of the 

 vowel in " well." The second line is from the vowel in "here." The 

 third is near the beginning of " your." The fourth is the first part of 

 the vowel in "good." The last is from the middle of "health." In 

 the second, third, and fourth cases there is evidently pre.sent a tone 

 more or less nearly the octave of the laryngeal tone. The other tones 

 and the tones in the other cases can be found only by analysis. 



the nature of a vowel and of the differences 

 between different vowels. 



At the present day the vowels can be recorded 

 on talking machines, and their curves can 

 be traced off with an accuracy that leaves nothing 

 NO. 2672, VOL. 106] 



to be desired. The work of Hermann on the 

 curves of the vowels and consonants by means 

 of the phonograph is still unsurpassed. For my 

 own investigations the gramophone was chosen 

 as the most available machine. 



A disc with the desired record was placed on a 

 very slowly revolving plate 

 (Fig. i). A long lever of Japan- 

 ese straw was held in an axle at 

 one end. Near this end a steel 

 point projected downward into the 

 speech groove. \t the other end 

 there was a recording point made 

 of a fine glass thread. As the disc 

 revolved, the movements were 

 magnified — up to 500 times — and 

 traced on a moving band of 

 smoked paper. 



Pieces of vowel curves cut out 

 of a tracing of a record by Joseph 

 Jefferson are shown in Fig. 2. 

 The curves show that in speech 

 the vowels change constantlv in 

 pitch, in intensity, and in char- 

 acter. They also show that the vowels actually 

 used in speaking are often not what the phone- 

 tician supposes them to be. 



The point of interest on the present occasion, 

 however, is the nature of a single wave of a vowel. 

 At the present day there is only one way of 

 analysing a wave — namely, the harmonic analysis. 

 ."Vny wave can be represented as made up of a 

 series of simple sine waves with the relations of 

 frequency of i : 2 : 3 : . . . and with various ampli- 

 tudes. A harmonic analysis of 

 the wave in the top line of 

 Fig. 3 gives the four curves in 

 the lines below. This means 

 that the four curves, if added 

 together, will give a result like 

 that in the top line. 



Suppose, now, that we have 

 a curve that consists of a vibra- 

 tion repeating itself every 3^ 

 times to a wave. The harmonic 

 analysis gives as result a fairly 

 strong fundamental of the fre- 

 quency I, a stronger vibration 

 of the frequency 2, a still 

 stronger vibration of the fre- 

 quency 3, a somewhat less 

 strong vibration of the fre- 

 quency 4, and ever-lessening 

 vibrations of the frequencies 5, 6, 7, etc. Not one 

 of these vibrations was actually present in the 

 original curve. The strength of the original 

 vibration of 3^ could not be directly given, because 

 there was no place for it in the harmonic series. 



The harmonic analysis shows us how a given 

 curve can be represented as made up of a series 

 of harmonic components ; it does not say that it 



Fig. 3. — A curve composed 

 of four sinusoids. 



