HARMONIC ANALYSIS OF CERTAIN VOWEL SOUNDS. 747 
The tube was secured to a triangular brass plate 7, with three rounded feet, 
two of which stood in a V-groove in another brass plate 7, fixed on the sole- 
plate of the instrument, while the third foot rested against a plane surface on a 
part of the plate 7. These feet gave five points of support, and left the stand 
one degree of freedom of motion, namely, to slide along a horizontal line at 
right angles to the axis of the cylinder c. The sixth degree of constraint was 
given by the screw s, which abutted against the plate 7, and which served to 
adjust the pressure of the pointer against the tinfoil. The stand holding the 
mouthpiece was kept pressing against the end of the screw s by a couple of 
india-rubber bands. This arrangement enabled the mouthpiece to be drawn 
back, or even completely removed with perfect facility, and then replaced at 
any time in absolutely the same position as before. 
_ Our first aim was to secure well-defined curves of such dimensions as would 
allow them to be subjected to harmonic analysis, and to obtain these under 
such conditions as should leave no doubt that each curve did truly represent 
the sound spoken in, at least, its essential vowel quality, if not in all its minor 
characteristics. 
Our method of obtaining these curves was as follows :—-The sound to be 
examined was spoken or sung with the mouth of the speaker close to, but not 
touching, the end of the tube m, and while the sound was being uttered the fly- 
wheel was turned as steadily as possible by an operator who kept time with a 
metronome. The metronome gave rather more than three beats in a second, 
and the fly-wheel was turned so as to make one revolution for each beat. This 
gave a movement to the cylinder regular enough to enable the pitch of the 
sound, when not otherwise known, to be afterwards determined to within one 
semitone by measuring the records. Special care was taken to turn the phono- 
graph regularly when the determination of the pitch was known to depend on a 
measurement of the marks. We employed two kinds of tinfoil; in the earlier 
experiments the weight of the foil was 268 grains per square foot, and in later 
experiments 314 grains per square foot. Our best results were obtained with 
the thicker tinfoil, the resistance of which seemed in no way to interfere with 
the accuracy of the curves; on the contrary, this resistance served to augment 
the directing force under which the disc vibrated, and allowed higher tones to 
be correctly registered. 
The embossed or indented tinfoil record was transcribed so as to give a 
magnified curve drawn in ink on paper by means of the appliances shown in 
Plate XXXV. The pointer of the speaking disc was connected by a fibre 4, 
of fine glass to a lever /, consisting (for the sake of lightness and rigidity) of a 
triangular framework of straws; this frame was pivoted in a fixed support at 
nm. A silk thread y, joined the bottom end of / with a fine glass siphon, which 
formed a second lever pivoted at 2, the upper end of which was bent round so 
