328 M. W. Siemens on the Question of 



in which I can scarcely suppose him, however, to be serious : it 

 is that we have never found a gold chain become brittle. 



It may, however, be conceded that changes in the resistances 

 of normal standards and their copies may be so small as to be 

 without practical importance in our present experiments. But 

 the normal measures of the B. A. are destined to serve in future 

 times, when probably an infinite number of higher claims to 

 exactness in a measure will be set up than we can pretend to. 



On this ground it is certainly significant that the Committee 

 should have made ten normal measures instead of a single one, 

 even supposing them to agree with each other to within 0*03. 

 per cent., as asserted. 



With regard to the agreement between the B. A. unit and 

 10 7 - I ^§ units, even allowing that it is within 0*1 per cent., 



second > O r > 



as stated in the Report of 1864, still it is too small to enable 

 the values to be given out as equivalents ; and if a coefficient 

 must be used, I do not see that it matters in the least whether 

 it differs a little more or a little less from unity. 



Besides, it is by no means satisfactorily proved that this pro- 

 fessed accordance of the-B. A. and the 10" ~^ units really exists. 

 A glance at the Table of results given in the Report of 1864 

 (p. 350) will suffice to show, between the values of the negative and 

 positive observations, combined into pairs, differences amounting 

 to 9 per cent. These differences are partly accounted for by the 

 torsion of the fibre, which entered into the results, adding or 

 subtracting from the observed values, according as the helix of 

 the apparatus was rotated right- or left-handed. But the mean 

 values of these pairs of observations differ still as much as 1*4 

 per cent. By what train of reasoning the Subcommittee holds 

 itself justified, in the face of such differences between even the 

 means of their single observations, in concluding upon a pro- 

 bable error of only 0*1 per cent., I am totally unable to ima- 

 gine. Whatever method is taken to calculate the mean values 

 of the given numbers, if some of the extreme measurements or 

 some of the mean values be left out, very much greater differ- 

 ences are indicated. 



In my opinion, certainty lies only within the limits of those 

 numbers which are not liable to be regarded with mistrust as 

 heing faulty and therefore entitled to be cast out. 



It is, however, impossible, from the numbers given in the 

 Table in question, to conclude upon a degree of exactness so 

 great between the B. A. and 10 7 ^^ units, as is done by the 



c second 3 J 



Subcommittee; and when we recollect that the observations 

 were made with the same apparatus, with the employment of 

 the same correction-coefficients, the probability is considerably 



