OF AETS AND SCIENCES. 35 



Tlie variable errors attending the determination of K are those 

 affecting jo^ p.^, and p., and already discussed; and from the measure- 

 ments of Table VI. just referred to, it appears that the variable errors 

 arise about equally from instrumental measurements and from fluctu- 

 ations in p^. 



III. A A. 



As shown by the column headed aA in the tables, the accuracy 

 necessary in the value of A, the coefficient of linear expansion of glass 

 used in the computations, increases as the temperature rises, an error 

 of 0.1 per cent in y corresponding, however, to an error of 20 per cent 

 in A at the highest temperature attiiined. The value of SA used was 

 as follows : — 



These values were selected as corresponding quite closely to the 

 probable change of the coefficient with change of temperature. The 

 absolute value from 0"^ to 100° is based on measurements made on the 

 same tube used, but the variation with the temperature was not 

 measured. The error in this coefficient at 100° cannot, I think, ex- 

 ceed five per cent, and is not likely to reach that amount ; and a con- 

 sideration of the results obtained by many observers, besides those 

 which I have myself made on similar kinds of glass, renders it certain 

 that the value of .3.-1 at 200° cannot be much more in error than at 

 100°. Even had a constant coefficient been employed throughout, its 

 error introduced into A could hardly have exceeded 15 percent at any 

 temperature, and had this coefficient been arbitrarily assumed from the 

 results of other observers, an error exceeding 20 per cent would have 

 been unlikely. Since the value of A A in Table YIII. is 19 per cent 

 at 225°, it appears that, with the values used, the results of the com- 

 putations of y cannot have been sensibly in error from this source, 

 even at the highest temperatures attained. 



IV. A a. 



The most important quantity whose value is necessarily assumed in 

 the computation of i/ is the coefficient of expansion of the gas used. 

 This (luantity and the temperature of transpiration, as may be seen 

 from the columns headed An and A^, or from the expression giving y, 



