PHYSICS. 225 



an apparatus by which he obtains vertical sections of the 

 impressions made on the tin-foil of the phonograph, magni- 

 fied 400 diameters, and called " speech-curves." In the word 

 tah, for example, intoned, there is first the " preparation," 

 the curve gradually but irregularly rising; then the "at- 

 tack" a bold serrated precipice, with numerous rather sud- 

 den valleys; next the "glide" a perfect tumult of curva- 

 tures, which gradually settle down into the " vowel " proper. 

 This remains constant for a considerable number of periods, 

 and vanishes aw.ay gradually to silence. This curve Jenkin 

 lias submitted to analysis, reducing it to its separate pendu- 

 lar curves, and lias succeeded in tracing out as many as five 

 partial tones. The results differ materially for different 

 speakers, and Jenkin is endeavoring to classify these speech- 

 curves into genera. Thompson proposes to improve the sib- 

 ilants in the phonograph by placing a strip of card or watch- 

 spring across the opening edgeways, so that the voice im- 

 pinges on the edge of the strip. The aspirates are also well 

 spoken by such an instrument. 



Jenkin and Ewirig have studied elaborately the form of 

 the sound tracings produced by Edison's phonograph. With 

 reference to the vowels, these observers note that if a set of 

 vowel sounds be spoken to the phonograph, and then it be 

 made to speak at several different velocities of rotation, no 

 difference can be detected in the quality of the sounds. This 

 they regard as contradictory of Helmholtz's statement that 

 each vowel sound has a characteristic note of definite pitch. 

 Moreover, they have observed that the wave form of the 

 markings produced by any vowel sound does not remain un- 

 changed at all pitches; but whether these changes are due 

 to alterations in the amplitudes of the constituents or to va- 

 riations of phase is not determined. Subsequently, in a pa- 

 per before the Royal Society of Edinburgh, they announced 

 the curious fact that both vow r els and consonants are unal- 

 tered by being spoken backward. Words such as ada, aba, 

 aja, ete, could be readily identified whichever way the cylin- 

 der was turned, even by persons ignorant of what had been 

 said. Moreover, they find that ab said backward becomes 

 ba, thus proving that a reversible part really constitutes an 

 element of speech. Thus putting the word noshdeesossa on 

 the cylinder, and turning it backward, it repeats association 



K2 



