160 G. Smeal: 



divided into two groups, in which the humidities were all vibove and 

 all below 50 per cent, respectively. In the earlier series all humidities 

 were above 50 per cent. The application of the formula to each' 

 group separately gave 77 = 0.9589 below 50 per cent, humidity and 

 0.9715 above. Both these values are less than that obtained for the 

 two groups combined, instead of being one greater and one less, as 

 we might expect. Such a result seems to indicate very clearly that 

 the value of the second constant -q is almost entirely dependent on 

 the nature and distribution of the accidental errors of the first 

 constant A, so that it will vary arbitrarily with the particular group 

 of observations chosen. In other words, there is no physical justifi- 

 cation for the insertion of a second constant. It is probable that its 

 value, determined by trial as above, would always be less than unity, 

 but this does not indicate the existence of any phenomenon which 

 is not implicitly allowed for by the simple formula. The following 

 tables will show that the insertion of it is not attended by any 

 increase in accuracy, provided the observations are good. It may 

 possibly be permisisible to say that a series of observations which 

 yields a value of 7) markedly different from unity is unsatisfactory in 

 some respect. 



In Table III. the values of x observed are compared with those 

 deduced from the two formulae, according to the equations 

 a:i=/-0.0007232B(/-/') 

 and x, = 0.9877/- 0.0006967B(/-/''), 

 and the differences, or errors, S^Xi = x - x\ and AX2 = x-Xo, are also 

 given. 



