177 



are about equal to TfrW, the real probable error of M must be about 

 1 part in 2,500. The number of observations is however too small for 

 an exact estimate of the probable errors. 



Taking the results with currents of different strengths, we find 

 For strongest current .................... 34-716 



For medium current ...................... 34-715 



For weakest current ...................... 34-727 



which are almost perfectly accordant. Taking the results from the 

 method of recoil and the ordinary method, we find 



For ordinary method .............. 34-726 -010 



For method of recoil .............. 34-705 -006 



If the probable error is subtracted from the first and added to the 

 second they will very nearly equal each other. Hence the difference is 

 probably accidental. Indeed, by the combination of the results it does 

 not seem possible to find any constant source of error, and therefore 

 the errors should be eliminated by the combination of the results. 

 In the final result 



= 34-7192 -0070 



the probable error, -0070, includes all errors except the ratio of G 

 to G". We may estimate the probable error of G at ^jVff and of G" 



Hence the final probable error of R, including all variables, is 

 or -04 per cent, 



or # = 34-7 19 '015. 



The probable error of the British Association determination was -08 

 per cent, not including the probable error of the constants; and of Kohl- 

 rausch's determination db -33 per cent, including constant errors. 



COMPARISON WITH THE OHM 



The difficulty in obtaining proper standards for comparison has been 

 explained above and I shall have to wait until the arrival of the new 

 standard before making the exact comparison. At present I give the 

 following results, which seem to warrant the rejection of Messrs. Elliott 

 Bros'. 10-ohm standard and to make that of Messrs. Warden, Muirhead 

 and Clark correct. I shall designate the coils by the letter of the firm 

 and by the number of ohms. Experiment gave the following results: 



W (10) = 1-00171 X E (10), experiment of June 8, 1877. 



W (10) = 1-00166 X E (10), experiment of Feb. 23, 1878. 



W (1,000): W (100):: W (10): -999876 E (I), experiment of Febru- 

 ary 23, 1878. 

 12 



