MH. W. N. SHAW ON HYf .KOMKTRIC METHODS. 75 



freely exposed without any provision for a definite circulation of the air cannot be 

 expected to give results which are accurate to within 2 per cent. ; in fact, no 

 satisfactory formula of reduction can be found. 



The errors are especially serious when the air is nearly or quite saturated and the 

 wet bulb coated with ice. 



The most accurate results with this method are obtained when an artificial 

 circulation of air is maintained, and in that case the highest accuracy claimed for 

 the method allows an error of 2 per cent, with temperatures above zero, and twice 

 that amount with lower temperatures, the comparisons being made with ALLUARD'S 

 form of REGNAULT'S dew-point instrument. 



6. The case of the hair hygrometer is very perplexing ; opinions are very conflicting, 

 and it seems to be a question upon which meteorologists take sides. An immense 

 amount of experimental work has been done, and it is possible that an observer 

 might make very useful observations with the instrument, but whether the inferences 

 from his observations would be regarded with any degree of confidence by others 

 seems to be still an open question. 



With the exception of REGNAULT'S researches on the chemical method (p. 121), the 

 experimental work that has hitherto been published has consisted in the simultaneous 

 observation of the vapour pressure in free air by two different hygrometric methods. 

 This plan is open to uncertainties arising from the two following causes : 



(i.) The air is taken from two different positions for the two instruments ; in 

 other words, the observations are made upon two different specimens of air, whose 

 hygrometric states are assumed to be identical, not upon the same air. Perhaps the 

 assumption is sufficiently justified when the air has immediate access to the reading 

 parts of each instrument, but I have pointed out in the summary of results certain 

 reasons for considering this uncertainty to be serious when the access is not direct, as 

 in the case of the comparison of a dew-point instrument with a ventilator-psychro- 

 meter, in which case the air may suffer alteration in passing over the vanes of the 

 ventilator. And, indeed, in any case, in an experiment designed to compare the 

 efficiency of different methods, the uncertainty arising from this cause is a dis- 

 advantage, for the liability to error in the instruments themselves is quite sufficient to 

 make the comparisons difficult ; and, moreover, if we could rely upon the identity of 

 the results of two instruments, a similar arrangement might be employed to determine 

 whether there is any local variation in the b} grometric state of the air, an independent 

 question not without interest. 



(ii.) The observations obtained with different instruments differ in character. The 

 chemical method gives the mean value of the vapour pressure during the period of 

 the experiment, and does not indicate any small variations from time to time. The 

 wet bulb takes a certain finite, though it may be short, time to reach its equilibrium 

 state, and therefore does not give the vapour pressure at any particular instant. The 

 dew-point instrument, on the other hand, may be taken as giving the vapour pressure 



L 2 



