340 



NA 7 URE 



[7'^fy^S^ T878 



A carriage moving smoothly on fom wheels travels beneath 

 the lens at such a distance that the sensitised plate laid upon it 

 comes at the focus for actinic rays. A uniform velocity is given 

 to the carri^e by a string fastened to it and passing over a 

 pulley. To this string a lead weight, just sufficient to balance 

 friction, 'is permanently attached, while a supplemental weight 

 acts at the beginning of motion and is removed just before the 

 sensitised plate reaches the spot of light above described. 



The velocity attained by the carriage is determined by placing 

 a sheet of smoked glass upon it and letting it run under a tuning- 

 fork (Ut] 3 — 512 V. s.) provided with a pointed wire. In every 

 case more than 200 vibrations were counted and measured, and 

 careful comparisons made between the earlier and later ones, so 

 as to be certain of the uniformity of the motion. 



From the description it will be evident that when the carriage 

 alone is in motion a straight line will be photographed upon the 

 plate. On speaking into the mouthpiece the disc is set in 

 vibration, each movement causing change of angular position of 

 the mirror, the reflected light moves through twice this angle, 

 and the resulting photograph gives us the combination of its 

 motion with that of the carriage. The carriage should run from 

 1-igA^ to /(/"/. The negative (examined from the g/ass side), and 

 prints taken from it, then give the syllables in their proper order, 

 and show movements of the disc from the speaker by lines going 

 from the observer. The arrangement of my dark room com- 

 pelled me to make my carriage move from /e/t to right ; hence, 

 in the figures given, forward positions of the disc are represented 

 by the lower portions of the curves. 



The general character of the curves obtained is shown in the 

 accompanying figures, which are the actual size of the originals, 

 except that representing the vowel-sounds, which is about one- 

 half (0-56). 



The velocity of the carriage for the vowel-sounds was 214, for 

 BrownUniversity, 40, and iov How do you do, I4inches per second. 



In the mathematical discussion of these curves the abscissas 

 are measured by the known velocity of the carriage, and serve 

 to determine the pitch, the ordinates represent the amplitude of 

 vibration of the centre of the disc, magnified 200 times in the 

 photographs. The reduction of scale makes the magnifying in 

 the woodcuts only 112 times. 



The ordinates are not strictly straight lines, but parts of the 

 vertex of a parabola, and closely approximate to circular arcs 

 whose radius is the focal length of the lens employed. In the 

 figures given, the centres of curvature of these arcs is at the 

 right hand. 



With an ordinary tone of voice an amplitude of nearly an inch 

 is obtained, implying a movement of the centre of the disc of 

 ■005 inches as determined by actual measurement. 



By varying the accelerating weight and its fall, any manage- 

 able velocity may be given to the carriage. Each syllable 

 requires for its articulation about one-fourth of a second, hence 

 the plates must be quite long when the velocity is great. I 

 employ plates two feet in length, and find that velocities from 

 16 to 40 inches per second give good results. The action of 

 the light is, however, inversely as the velocity. To compensate 

 for this, the size of the circular opening admitting the light may 

 be increased. This, of course, causes an enlargement of the 

 luminous image, and apparently involves an injurious widening 

 of the line traced, but, as observed by Dr. Stein in his experi- 

 ments, the effect of velocity is to narrow the line photographed, 

 since the maximum exposure is in that diameter of the circular 

 image which lies in the line of motion. This is a great advan- 

 tage, since a variation of velocity in the vibration is marked 

 by the widening of the line, often more clearly than by the form 

 of the curve. 



I have employed the ordinary photographic process, not 

 attempting to obtain special sensitiveness. The brightest sun- 

 light is required, a slight haziness interfering seriously with the 

 result. My heliostat employs two reflectors of ordinai-y looking- 

 glass, and the loss of light is considerable. 



Are all the audible elements of speech traceable in these 

 records? in other words, is the record complete? I am not 

 prepared as yet to answer this question definitely, but the fol- 

 lowing experiment leads me to doubt whether an affirmative 

 answer can be given, while at the same time it illustrates in a 

 striking manner th:^ sensitiveness of the ear. The mirror was 

 attached to the disk of a receiving telephone and a photograph 

 taken from it while the instrument was talking audibly. The 

 resulting record was almost a smooth line, showing but very 

 slight indications of movement of the mirror. It would there- 



fore appear that there are distinctly audible elements, which 

 are too minute to be recorded by this method. It is to be noted, 

 however, that the width of the line traced where the vibrations 

 are extremely small, is so great as to mask the curvature, so that 

 the experiment just cited is not entirely fair. 



The clearness and beauty of the curves obtained can hardly be 

 appreciated without inspection of the originals. Their com- 

 plexity and variety open a large field for investigation, and they 

 seem to off"er the means of analysis of articulate speech. 



THE PHONOGRAPH AND VOWEL SOUNDS 



I. — The Vowel Sound o. 



T N a recent letter to Nature we gave a short account of what 

 -"• we believed to be the existing theories of vowel-sounds. 

 In the present paper we will state the chief results as to the 

 vowel of our investigations made by means of the phonograph. 



The experiments were made as follows : — The vowel under 

 consideration was spoken or sung at a given pitch, determined 

 by a piano, while the barrel of the phonograph was turned at a 

 definite speed, regulated by means of a metronome. The in- 

 dentations made in the tin-foil were then mechanically tran- 

 scribed, so as to give curves representing a magnified section of 

 the impressions. The curves were magnified by a system of 

 compound levers, and recorded by an arrangement resembling 

 that of Sir William Thomson's siphon-recorder. The details of 

 the apparatus are described in a paper laid by us before the 

 Royal Society of Edinbiu-gh. The vertical ordinates of the 

 curves drawn in ink, as shown below, are about 400 times larger 

 than the corresponding indentations in the tin-foil, while the 

 longitudinal ordinates are multiplied about seven times. The 

 slowness of the motion by which the transcript was made enabled 

 us to avoid all error due to inertia of the working parts, and the 

 total absence of friction between the marking siphon and the 

 paper allowed the transcript to be made without employing such 

 a pressure on the tin-foil as would sensibly alter the indented 

 curves. This fact was in each case tested by making the phono- 

 graph speak the vowel after it had been copied. All transcripts 

 were rejected if the tin-foil did not continue to give the sound 

 clearly after being used to produce these curves. 



We employed various sizes of chamber and of mouth-piece, 

 various thicknesses of tin-foil, and various discs as receivers. 

 The curves now given as transcripts were found to be practically 

 independent of variations in all these conditions. We are 

 therefore of opinion that the curves do really represent the 

 motion of an air-particle when the vowel is spoken, and that 

 these curves may be regarded as sufficiently unaffected by any 

 periods of vibration proper to the disc and springs employed, or 

 to the air in the chamber of the mouth-piece, to constitute a 

 true record of the essential parts of the vowel-sounds. This 

 may be inferred from the remarkable constancy in the results 

 obtained, with great variations in the conditions of the experi- 

 ment, and from the fact that the indentations, after being 

 copied, were in each instance able to give back the vowel-sound 

 distinctly. 



Fig. I gives a series of curves produced by a single baritone 

 voice singing J on a series of notes ranging from G to/'. This 

 series has been selected because the voice was of good quality 

 and considerable range. 



After the curves were obtained they were subjected to har- 

 monic analysis. One period was divided into twelve equal 

 parts, and twelve ordinates were drawn and measured at right 

 angles to a line joining two successive maximums or minimums. 

 The numbers so determined enabled us to calculate the ampli- 

 tudes of the first six partial tones. 



Table I. gives these amplitudes for the above series of o's, obtained 

 by analysing one period chosen out of the hundreds of similar 

 periods which were given by each utterance. An examination 

 of the ciu-ves and of the table will show that the change in 

 character from note to note is fairly gradual and consistent 

 throughout. The figures are arranged so as to show the absolute 

 pitch of each of the six partial tones. 



Voices of very different qualities were tested in the same way 

 throughout the same range or such parts of that range as were 

 within their compass. 



Too much space would be taken up if we were to give here 

 all the results obtained. It may be briefly said that the several 

 voices agreed very fairly in respect of the partials composing 

 the vowel at each pitch, that is to say, throughout the range 



