162 REPORT — 1882. 



Lussac's method, and taking the mean, would give an equally good 

 result with much less trouble in calculation. 



(20) On the whole, then, the Committee think that, if the following 

 rules are obeyed, an accuracy of 0°*05 m.m. in the correction of a 

 thermometer scale can be attained. 



I. If a dividing engine is used, select principal points about 40 m.m. 

 apart. ^ 



Correct by measures on a thread about —m.m. long, employing Gay- 



Lussac's as a principal point method. If necessary apply a second 

 approximation. 



This will probably be sufficiently accurate. It may, however, be well 

 to check by using a thread 40 m.m. long to redetermine the principal points 

 accordino" to Gay -Lussac's method, applying the method of transference 

 by means of the curve obtained with a short thread. Combine the results 

 of the two sets of measures at the principal points, giving to the values 

 obtained by the two threads weights proportional to their lengths. 



Use the shorter thread to determine secondary points. Examine 

 further any maxima or minima in the neighbourhood of which the 

 curves slope steeply. 



The total number of measures which would be required to apply this 

 method to Thermometer C would be ten on the long and thirty on the 

 short thread. 



II. If an instrument such as Mr. Brown's is used, reading to 0"1 m.m. 

 on the vernier and to less by estimation. 



Select principal points about 40 m.m. apart. 



Apply Gay-Lussac's method twice by measurement on threads 20 and 



— m.m. long respectively. Use the method as a principal point method, 

 o 



and make the points originally selected principal poinis for both correc- 

 tion curves. 



If necessary, apply a second approximation in each case. 



If the two curves obtained agree to within O'l m.m. their mean will 

 probably be as accurate as is desired. In taking the mean the weight 

 assigned to each should be proportional to the thread-length. If, however, 

 greater certainty is desired, it may be well to determine the principal 

 points, either by Gay-Lussac's method, using a thread 40 m.m. long, or 

 by Thiesen's. 



In the former case combine the three curves at the principal points, 

 giving the weights 6, 3, and 2 to the corrections obtained by the longest, 

 intermediate, and shortest threads respectively. 



If Thiesen's method is employed and there are n principal points, the 

 weights assigned to the three corrections of the r*** principal point will be 



3 nr {n - r), 3 (w - 1), and 2 {n - 1) 

 to that given by Thiesen's method and those obtained by Gay-Lussac's 

 with the longer and shorter thread respectively. This gives so great a 

 preponderance to Thiesen's method that unless the discrepancies are very 

 great the corrections at the principal points practically depend upon it 

 alone. 



' The numbers given in these rules refer to a thermometer about 400 m.m. long, 

 on which ten principal points are to be found. They can readily be modified so as to 

 apply to instruments of other dimensions. 



