the Thermal Unit, 447 



were reasons for suspecting that the temperature of the 

 Cavendish Standard, at the time of comparison, slightly 

 exceeded that of our cells. 



In ' Science Progress,' April 1894, I entered fully into this 

 matter, and a recomparison with the Cavendish Standard made 

 during the spring of this year (when greater attention was 

 paid to securing the equality of the temperature) confirms 

 my opinion that 1*4342 was the more probahle mean value of 

 my Clark's cells. This would diminish the value of J by 1 in 

 4000, and it would thus remain 4*198 x 10 7 *. 



Schuster and GUnnon. — I regret that at the time of 

 writing I have only been able to obtain an abstract of the im- 

 portant investigation by Professor Schuster and Mr. Gannon. 

 This abstract, however, is sufficient to indicate the value of 



* There is one criticism of Professor Schuster and Mr. Gannon's 



regarding my work to which I am glad to have this opportunity of 



replying. The following is a quotation from the Abstract of their 



paper : — " We may reasonably estimate the uncertainty due to the 



cooling correction by calculating what the observed rate of cooling, 



either at the beginning or the end of the experiment, must have been in 



order to produce a difference of one part in a thousand in the final 



results. 1 ' They remark later: — "In the experiments quoted by Mr. 



Griffiths at p. 482, the loss of heat due to radiation and conduction 



at the end of his experiments amounts to about 9 per cent, of his 



heat-supply. If such loss had been wrongly estimated to 1 per cent. 



an error of one-tenth per cent, would result in the final value." This 



criticism may be just when the rate of rise at the exterior temperature is 



determined by the time of rising over the whole range, as was the case in 



Professor Schuster and Mr. Gannon's experiments ; in my own, however, 



the rising rate at each part of the range was determined by separate and 



independent observations. If their criticism were sound, the uncertainty 



as to the rate of rise at any given temperature would depend on the 



temperature-range of the experiments, whereas, had I doubled the ranges 



dd 

 my conclusions as to the value of — , when the temperature of the 



calorimeter crossed the exterior temperature, would remain unchanged. 

 True, the rate of change in the heat-capacity of water would be affected 

 by errors in the cooling-rate at the extremities of the range, but my 

 resulting value of J would remain unaltered when expressed in terms of 

 the heat-capacity of water at the exterior temperature. 



I think, therefore, that Professor Schuster and Mr. Gannon over- 

 estimate the possible errors in my results due to errors in the deter- 

 mination of the " cooling rate." With regard to this " cooling cor- 

 rection " I am in agreement with their conclusion that it probably varies 

 with the rate of rise, and I had, before the publication of Professor 

 Schuster and Mr. Gannon's work, expressed my views on this point in a 

 paper on the " Specific Heat of Aniline " communicated to this Section in 

 18*44 l . I have there shown that f> (the "cooling coefficient") is a 

 function of the rate of rise in temperature, and I described some 

 experiments which led to the conclusion that the error, under cir- 

 cumstances similar to those prevailing in my J experiments, was of 

 small importance unless the variations in the rate of rise were extreme. 



1 See also Phil. Mag., January 1895. 

 2 I 2 



