258 CARNEGIE INSTITUTION 



records great weight should be laid on those of poets and orators. 

 In the first place, the printed verse or speech can convey only a part 

 of the intention of the author ; a good gramophone record by 

 Patrick Henry or Longfellow would give something that is now for- 

 ever lost. In the second place, from accurate speech curves traced 

 from such plates we can obtain for the first time reliable data con- 

 cerning the essential factors of oratorical speech and of verse. 



The plans on both sides of the Atlantic are complete. I am ready 

 to carry them out privately or under the direction of the Carnegie 

 Institution, according as the Trustees may direct. In the former 

 case the matrices and records will remain my personal property. 

 There will be no expense except for traveling. 



5. Provision should be made for the construction of the following 

 apparatus : 



One piece, suggested by Dr. Billings, is designed to turn the rec- 

 ords of the speech curves back into sound. The difficulties with 

 the design have been great, but have now been overcome. The 

 success of a speech record depends on allowing time for the point 

 of the reproducer to follow the fluctuations of the groove. The 

 larger the disc or cylinder the better the result. The speech waves 

 are thus very long and flat, and often cannot be distinguished by 

 the eye from a straight line. In my tracing machine the amplitude 

 of the fluctuation is greatly increased, while the length of a wave 

 remains the same, or is even shortened. Thus the fluctuations 

 become visible. Such waves cannot be used for producing sounds 

 directly. They must be greatly drawn out in length and reduced in 

 amplitude. For this purpose the speech curve is in the new appa- 

 ratus to be cut on a wooden cylinder and reproduced as a depression 

 in wax on a phonograph cylinder. The former cylinder is rotated 

 at a very slow speed in comparison with that of the latter, while the 

 amplitude is appropriately decreased. The minor technical diffi- 

 culties are, of course, numerous, but the machine can presumably 

 be finished before next spring. 



The value of such a machine lies not only in the fact that it can 

 be used to test the accuracy of a tracing by turning it back into 

 sound (it can be more readily tested in other ways), but particu- 

 arly in the fact that it opens up an entirely new method of study- 

 ing the voice. A vowel consists of a series of vibrations that are 

 not of constant character. The form of the curve always (in my 

 results for American English) changes from beginning to end, wave 

 by wave. The sound must therefore be different at each instant. 



