

HARMONIC ANALYSIS OF CERTAIN VOWEL SOUNDS. 749 
spoken sounds would be transferred to the paper. To complete the apparatus 
it was only necessary to establish a mechanical connection between the wheel 
W and the cylinder c of the phonograph, so that the two might revolve with a 
constant velocity ratio. This was done very simply by putting a wooden drum d 
on the axle of the phonograph, another drum w on the axle of the wheel, and 
then connecting the two bya long string. The magnified ink curves were 
then obtained by putting the mousemill in action, and very slowly turning the 
wheel W, which pulled round with it the cylinder of the phonograph; the 
speed was always kept down sufficiently to prevent the inertia of the moving 
parts from affecting in the smallest degree the truth of the transcribed 
record. 
The figures which are given (Plates XX XVI. to XL.) are exact copies of 
wave-forms traced out in this manner. The length of a complete period in the 
figures is about seven times its length on the tinfoil. Each figure gives only 
a few consecutive periods, chosen out of perhaps the two or three hundred 
which went to make up the complete utterance. These specimens are, how- 
ever, in the case of sounds swng so as to preserve both pitch and vowel quality, 
really representative of the whole tracing. As a general rule, the tracings 
were remarkable for the constancy with which the same wave-form repeated 
itself over hundreds of periods ; while, on the other hand, in a few cases there 
was a greater roughness and irregularity in the forms than could fairly be laid 
to the charge of the method of copying on the original state of the sheet of 
tinfoil. This irregularity may perhaps in some cases have been due to the 
presence of inharmonic constituents in the sounds, but at other times was 
certainly due to wavering in the voice. These cases of irregularity were rare, 
and on no occasion have we observed periodic variation of the wave with a 
recurrence to an old form, even in the most extended utterances. 
The excursions of the end of the siphon were never so large as to make the 
obliquity of the levers a practical source of error. 
After the record on the tinfoil had been used to produce the magnified 
curves, the phonograph was withdrawn from the multiplying apparatus, and 
the record was made to reproduce the sounds originally spoken. In no case 
was the transcribed curve accepted as satisfactory unless the tinfoil proved 
still able to give the original sound satisfactorily. This was done to guard 
against the acceptance of faulty curves, in which the forms impressed on the 
tinfoil had been partly obliterated by the process of transcribing. The adjust- 
ment of the pressure between the pointer and the tinfoil during that process, 
so that it should be neither so great as to obliterate the record nor so small as 
to prevent the pointer from going to the bottom of the indentations in the foil, 
was a matter of the greatest delicacy, and required constant attention. The 
precaution mentioned above of making the tinfoil record repeat the sounds 
VOL. XXVIII. PART III. 9k 
