METHODS EMPLOYED IN CALIBRATION OF MERCURIAL THERMOMETERS. 159 



Table V. 



Ill terms of 0°001. 



0°"01, when the third approximation is partly carried out, that obtained 

 by Professors Thorpe and Riicker's method remains virtually unaltered, 

 while Yon Oettingen's is brought into agreement with it. The former, 

 therefore, though somewhat the longer, is taken as the more accurate 

 method. 



(17) The discrepancies between the Bessel and the second Gay- 

 Lussac curve were next investigated. The extremely small effect pro- 

 duced on the correction at any one point by any one measurement of a 

 thi-ead length in Bessel's method, makes it possible to use these threads 

 as means of comparison. The corrections for the upper and initial points 

 of all the threads measured in Bessel's method were taken out both from 

 the Bessel's curve, fig. 3, Plate I., and the Gay-Lussac curve, Example 2, 

 fig. 1, Plate I. The thread lengths were corrected by these, and then, 

 by subtracting from them the corrected mean thread lengths, the apparent 

 errors of the measures were deduced. 



As ten threads were measured at each of the initial points, if the cor- 

 rection for any one of these points was too high or too low, the threads 

 measured at it would be too short or too long. 



This fact can be used to test the relative accuracy of the two curves. 

 For instance, at 108°, the mean apparent error of the threads, when 

 corrected by Bessel's curve, was + 0°-001, which would show that the 

 correction of the initial point was too low, and should be changed from 

 0°"151 to 0°'152. Applying similar reasoning to the other initial points, 

 and to both the Bessel and Gay-Lussac curves, the following Table 

 results : — 



Table VI. 

 In terms of 0°-001. 



