450 Mr. E. H. Griffiths on 



either observer could secure identity at more than one 

 temperature. 



An inspection of Plate VII. at temperature 10° will show 

 that the agreement between the values of Rowland, Schuster 

 and Gannon *, and my own is there very close. Rowland's 

 result at that temperature is 4*2 x 10 7 , and if we assume that 

 value at 10° it is probable that our error will be of small 

 dimensions. 



Part III. 



Practical Suggestions, 



The ideal thermal unit should 



(A) be a natural, not an arbitrary, one, and have some 

 real relation with other units of energy. 



(In absence of further reasons the educational 

 value of such a relation is sufficient to establish this 

 proposition.) 



(B) It should not be dependent on the observations or 

 conclusions of any one investigator. Any such 

 dependence must certainly result in future correc- 

 tions and changes. 



(C) It should be of a convenient magnitude, and should 

 cause as little disturbance as possible in the nume- 

 rical values resulting from our present system of 

 thermal measurements. 



We cannot, I think, do better in this matter than follow 

 the principles which have governed the choice of our elec- 

 trical units. Let us consider for a moment the steps by which 

 we have arrived at our present value of the ohm. The value 

 of the ideal unit was first defined, that value being some con- 

 venient multiple of an absolute unit. The first attempt at the 

 practical standard, whose resistance should be that of the ideal 

 one, is that of the B. A. Committee of 1863, and, as is well- 

 known, their conclusion, expressed in terms of the length of 

 a column of mercury a square mi Him, in section at temperature 

 0°, was 104-83 centim. 



The work of Rayleigh and Schuster, Rowland, Glazebrook, 

 and others clearly demonstrated the existence of error, which 

 led to a redetermination by the International Congress of 

 1881, when the value was increased to 



106*0 centim. 



Lastly, we have the (so-called) final determination of the 



* Assuming the validity of the reduction to 10 c O. 



