256 EXPERIMENTAL PBONETICS. 



liy tuning forks, the pitch of which is that of the partials of the funda- 

 mental tone, oii^ 0, and a may be badly reproduced, it has l)een found 

 impossible to reproduce e and i. He then o1)]ects to the theory of 

 Hermann, namel}^, that the vowel is an oral intermittent and oscillating 

 tone; first, that the method of recording the vowel on the wax cylinder 

 of the phonograph causes grave errors, because the mouthpiece, tube, 

 air chamber, and vibrating disk all profoundl}' modify the vowel; 

 second, that the method of analj'sis by Fourier's theorem assumes 

 that the vowel curves are constituted by superposed simple curves, 

 which is precisely the question at issue, and therefore the arginuent is 

 a petitio principii; and third, that the data obtained by his method 

 have not enabled Hermann to I'econstruct the vowels with greater 

 success than Helmholtz. Marage then enters upon his own method, 

 which consists essential 1}^ of using a special apparatus constructed on 

 Konig's principle of manometric flames, but so simple as to bo practi- 

 cally free from sources of error; that is to say, there is no mouthpiece, 

 tube, or lever. The pictures of the flames were yjroduced photo- 

 graphically by feeding the flame with acetjdene gas, and chronojihoto- 

 metrical records were taken with each experiment. He then finds that 

 the flame pictures of i, u, and ou show one flame, e, eu, and o two flames, 

 and A three flames. So that the classification of the vowels by flames is 

 exactly that of Grassmann. P^ach vowel, when all errors have thus been 

 got rid of by simplifying the apparatus, always gives the same picture 

 for any given note. The picture is that of a continuous periodic curve, 

 and the number of periods in a second corresponds to the laryngeal 

 note, while the form of the period characterizes the vowel. With the 

 same vowel the period changes with the note. When the note is near 

 the pitch of ordinary speech the period varies very little. This is 

 not so when the vowel is sung; the period then disappears until there 

 is onl}^ the laryngeal note. Marage has also b}^ synthesis reproduced 

 the vowels with remarkable success. His first experiments with 

 resonators were not quite satisfactory; he could reproduce ou, o, and 

 A, but not E and i. He ascertained, however, that to reproduce a the 

 resonator must be tuned to the third harmonic or partial of the note 

 on which a was sung; that to reproduce e, eu, and o the best result 

 was ol)tained when the resonator gave the second partial; and i, u, 

 and ou were imitated (but not successfully) when the resonator was in 

 unison with the lar3^ngeal tone. 



Marage finally devised a syren rotated by an electric motor and 

 consisting of a disk having in it a triangular window representing the 

 glottis. The air is driven under pressure through this aperture and 

 then falls on another disk having windows cut out of it in groups 

 according to the nature of the vowel to })e synthetically reproduced. 

 Thus the disk for a has four groups, each group consisting of three 

 triangular slit-like windows; for o and e the disk shows five groups, 



