34 Mr. E. H. Griffiths and Miss Marshall on the 



Remarks on Experiments at 40° C. 



Exp. XI. — " Reduced pressure far too much at commence- 

 ment, was unable to obtain thermal balance until near end of 

 experiment." 



Exp. VIII. & IX. — As pointed out (section 10, p. 15), the 

 values of L are here probably in error by about 0*05. This, 

 however, does not seem sufficient reason for rejecting them. 



Conclusion. 

 When <9 =40-045, L = 100*71. 



Remarks on Experiments at 30° C. 



Preliminary Exp. — The oscillations in d x were very violent, 

 as this was our first attempt at balancing when evaporating 

 benzene ; also, as we had no idea when to finish, the value of 

 %q is very great. We have, however, inserted it to show how 

 little such matters affect the result. 



Exp. II. — " After inserting the dropper, had to withdraw it 

 again owing to its sticking in the tube ; dropper exposed for 

 some time." There is little doubt that, owing to the cooling 

 when withdrawn, some air was sucked back, and hence there 

 must have been some expulsion on re-introduction. We 

 regret that we continued the experiment : the result is 

 useless. 



20. It will be noticed that the irregularities in this Table 

 are greater than in preceding ones. These are due to two 

 causes : — (1) The want of practice in the observers ; (2) The 

 value of the galvanometer-swings were not determined with 

 sufficient accuracy until after Experiment VII. , and, as the 

 values of *Zq are large, any error might affect L, but probably 

 not by more than +0*1. 



In Section 10, p. 15, we have given in full our reasons for 

 considering that all these experiments (I. to VII.) are too low 

 by about 0*12. The mean value of L given by them (omitting 

 Preliminary and II.) is 102*14 : thus it is probable that, had 

 the flask been previously saturated with vapour, we should 

 have obtained about 102' 26, which is in fair agreement with 

 the mean of the last six (102*30), where this cause of error 

 was non-existent. In any case, taking all the values as they 

 stand in Col. XXI. (rejecting the f experiments), we get 

 102*22. The number 102*30, however, is the most probable 

 value. 



Conclusion. 



When <9 = 30*066, L= 102*30. 



