1896.] Mr Griffiths, On Thernwmetric "Fired Points." 227 



thus a difference of 5 in the last figure represents a change of 

 o, 001, hence the greatest difference from the mean on these 

 occasions was considerably less than 1/1000°. This at first sight 

 appeared satisfactory. 



At the commencement of October I was enabled to make a 

 couple of determinations of the freezing point with the same 

 thermometer in a new apparatus which has been constructed for 

 the purposes of an investigation into the freezing point of dilute 

 solutions. This apparatus (the particulars of which I hope subse- 

 quently to publish) has been so devised that the temperature of 

 the air space surrounding the entire vessel (as also that of the 

 walls surrounding that air space) is at the temperature of the 

 mixture in the inner vessel. I have only as yet been able to make 

 two determinations ; the results are as follows : 



1293947 

 •948 



Hence it would appear that in spite of the agreement in the 

 previous results, the influence of the room temperature was 

 sufficient to cause a rise exceeding '003 in the temperature. 

 The similarity of the conditions throughout the first series accounts 

 for the agreement amongst their results. 



It may be objected that such accuracy is unnecessary. To 

 this I would reply that it is evidently desirable that the limit of 

 error of a " fixed point " should certainly be less than the limit of 

 experimental error due to the instruments by which temperature 

 can be measured. 



I do not feel in a position to make any satisfactory suggestion ; 

 at the same time I think we should be moving in the right 

 direction if we more strictly defined the conditions of a deter- 

 mination. For example, we should obtain more constant results 

 than are possible under existing circumstances if we adopted, as a 

 temporary standard, water distilled at a certain rate from alkaline 

 permanganate within some definite but convenient limit of time, 

 which, having been collected in either copper or Jena glass 

 vessels was then mixed in certain definite proportions with ice, 

 crystallised from it, and placed in a vessel of definite size when 

 the temperature of the laboratory was within certain limits*. 

 The ideal standard being experimentally unobtainable, we require 

 a practical one if uniformity and accuracy are to be attained. 



* As a difference in pressure of 100 rnm. causes a change of about o- 001 in the 

 freezing point, a statement as to the standard pressure should (in anticipation of 

 further advances in thermometry) be added to any definition such as that suggested 

 supra. 



