244 CARNEGIE INSTITUTION 



black surface. On account of its delicacy, the machine requires daily 

 inspection and care by a mechanic. The renewal of paper occurs 

 once every 12 or 24 hours, according to the length of the strip of 

 paper, a factor that depends on the length of the room. Each re- 

 newal requires work for 1% to 2^ hours. 



The strips of paper are varnished before they are removed from 

 the machine. A narrow band containing the curve is then cut out 

 and mounted in the form of plates on pasteboard. Each plate is 

 covered with a sheet of celluloid and is delivered to the computing 

 bureau. 



The records referred to in these investigations had been traced off 

 by an apparatus known as Machine A. This machine was allowed 

 to run at Yale University till March i , at which date it was dis- 

 mounted and turned over to the university. Owing to the fact that 

 it was the first of the machines constructed, it could be successfully 

 used only under my constant personal supervision ; owing to the 

 many changes in its construction and to the wear of four years 

 running (often day and night), it had become somewhat deteriorated. 

 It was decided therefore to build two new machines on principles 

 learned by previous experience. These are now finished. One, 

 known as Machine D, has been at work at Yale University under 

 care of the mechanic who built it, and has completed the tracing of 

 a plate containing a record of Dr. S. Weir Mitchell's voice. The 

 other, known as Machine C, was finished in Europe and is about to 

 be set up in Berlin. 



Still another machine, of a different kind, had been constructed 

 by a grant from the Elizabeth Thompson Science Fund several years 

 ago. It traces off curves from a French phonograph with celluloid 

 cylinders. Forty special cylinders of French prose and verse were 

 made for it in 1902. It ran constantly from October, 1902, to May, 

 1903, and furnished the first curves of French speech ever obtained. 

 These curves can be used for solving the problems of French vowels, 

 French melody, and French verse, just as the curves for English 

 were used for the problem of English speech. I have with me the 

 complete tracing of a record of Lc Roi d' Yvetot. It awaits funds 

 for assistance in measurement and time for study. 



The work of the tracing station was carried on at Yale University 

 until September 15. 



At the computing bureau the first work consisted of measuring 

 abscissas and ordinates of the curves on the plates and in com- 

 puting results according to certain formulas and methods. The 



