METHODS EMPLOYED IN CALIBRATION OF MERCURIAL THERMOMETERS. 155 



vernier. The errors of reading would thus be reduced from four to two, 

 and the maximum possible inaccuracy halved. 



(11) In order to compare experimentally the different methods of 

 correction, the Committee have applied them to the same thermometer. 

 The instrument selected was thermometer C, and the methods applied to 

 it were those of Gay-Lussac, Hallstrom, Rudberg, Thiesen, Miirek, and 

 Bessel — the latter both as modified by von Oettingen, and also with the 

 additional changes introduced by Professors Thorpe and Riicker. The 

 measurements were made in the Physical Laboratory of the Yorkshire 

 College. 



Those for Bessel's method, and for the first example of Gay-Lussac's 

 method, were performed by Professor Riicker. The remainder, together 

 with much of the laborious calculation required, were kindly undertaken 

 by Mr. "W. Heaton, M.A., Demonstrator in the Clarendon Laboratory of 

 the University of Oxford, to whom the warm thanks of the Committee 

 are due for the labour and skill which he has expended on the investiga- 

 tion. It is owing to his help that they are able to present the Report 

 without further delay. 



In the first place the thermometer was corrected by Gay-Lussac's 

 method. The thread used was about 19 m.m. (l°-6) long. When it was 

 moved the lower end was always brought veiy near to, though not into 

 absolute coincidence with, the position previously occupied by the upper. 

 An inspection of the curve wheit completed showfed that the errors intro- 

 duced by these small deviations from the thedtetical position were abso- 

 lutely negligible. The length of the thread, "together with that of the 

 scale divisions in which its extremities lay, was measured twice for each 

 position of the thread. As when the thread was displaced, the lower end 

 lay in the same division as that which previously contained the upper, 

 each division was really measured four times, and the mean of these 

 values, which rarely differed by more than 0°-003, was taken as the true 

 value. 



The standard correction curve obtained is exhibited in fig. I, 

 Plate I. 



All the calculations, together with those required in the other calibra- 

 tions to be described, are given in Part II. 



In the second application of Gay-Lussac's method to the thermometer 

 the thread employed was somewhat longer, viz. about 22 m.m. (2°). 

 Instead of measuring each division in which the ends lay, the scale of the 

 thermometer was previously studied, and was found, as is usual in ther- 

 mometers calibrated by the maker, to consist of groups of divisions of 

 equal length. The values of these were determined and used as above 

 described. Another difference in the application of the method was that 

 it was now used as a principal point method to determine the correction 

 at every second degree. As, however, the length of the thread was very 

 nearly 2°, the divergence from the theory of the step-by-step method was 

 small, though considerably larger than in the last case. 



To eliminate any error a preliminary curve was drawn, and the 

 method of approximation applied. The distance between the positions of 

 the upper and lower ends (assumed in the first approximation coincident) 

 was never greater than 2-2 m.m. (0°-2), and the maximum correction 

 introduced in the second approximation was 0°'007. 



In what follows the correction curve obtained by the use of the thread 

 2° long was used as a first approximation curve to apply the transference 



