54 THE STUDY OF SPEECH CURVES. 



these only the "upper pulley" appears in the figure. A belt from the 

 upper pulley turns the recording disc. A belt from the lower pulley 

 turns a " rotating tube," in which there is a screw that moves the recording 

 disc sidewise; in this way the recording point is made to trace a spiral 

 line from the outside to the center. To permit the side-movement of 

 the recording disc the belt passes through a " belt spanner" with weight, 

 which allows the belt to lengthen and yet keeps it always at the same 

 tension. The "belt tightener" for the ])elt from the lower pulley is pro- 

 vided with a spring that keeps the tension constant. For many purposes 

 it is desirable to have the recording point trace one revolution in a circle 

 and not make a spiral; for this the belt tightener is loosened so that the 

 rotating tube remains still. 



With this apparatus a single wave can be selected from the curve 

 and can be repeated indefinite^ on the zinc strip; the gramophone disc 

 then produces continuously the sound of that wave. Beginning with 

 the first wave, we thus make a disc, one line of which will produce indefi- 

 nitely the sound of the first wave, another line the sound of the second 

 wave, etc. In this way we have an acoustic analysis — for the ear! — 

 of each element in the vowel. For example, we are thus able to hear 

 separately the 25 different vowels that are present in the record of [e] 

 in "get" with 25 waves. Innumerable debated problems can in this 

 way be settled immediately. For example, what is the vowel in "not" 

 in the case of a certain speaker? The word is recorded on a gramophone 

 disc in natural conversation, and the curve is traced off; then a single 

 wave or a group of waves is etched on a zinc strip and traced repeatedly 

 on the gramophone disc; the disc then speaks the vowel continuously 

 as long as desired and affords an opportunity for deciding its resem- 

 blance to [o], [a], or [a]. When this has been done for a number of persons 

 whose speech is recognized as having the standard pronunciation, the 

 proper phonetic spelling of "not" can be settled. We are quite safe in 

 asserting that very many — or most — of the short vowels are incorrectly 

 indicated in the dictionaries. By the speech curves and by this appara- 

 tus it will be possible to settle the correct pronunciations. 



There is still another application of interest, namely, an inquiry con- 

 cerning the sounds of arbitrary curves. For example, what is the sound of 

 a zigzag line or of a zigzag with one element shorter than the others? What 

 is the sound of a curve composed of alternating positive and negative semi- 

 circles? Just as each musical instrument and each vowel records its own 

 pecuhar curve, so each pecuHar curve will produce a special sound; from 

 curves not like those of known musical instruments or vowels we may 

 expect sounds representing musical instruments that do not exist. 



