Connexion between the Atomic Weights. 113 



PI. II. is drawn on paper ruled logarithmically. The 

 •graph of an equation of the form y=-x m is a straight line on 

 this paper, so that all the computed points lie on a straight 

 line. The actual distances of the points (#) showing the 

 experimentally determined values above or below the straight 

 line, are proportional to the percentage differences between 

 the computed and the experimental numbers. 



It will be seen that the actual values of the atomic weights 

 lie very close to the straight line in PI. II. 



An equally good straight line would of course be obtained 

 by plotting the cube roots of w instead of w. 



Thus the formula is of a different type to Stoney's, from 

 which it would follow that the actual values (not the logarithms) 

 of the cube roots of the atomic weights plotted against the 

 logarithm of the order would give a straight line. 



On the Determination of the Power Constant and the 

 Use of the Rule. 



The whole of the preceding leads to the conclusion that a 

 relationship of the form 



expresses to a close degree of accuracy some fundamental 

 ■connexion which exists between the masses of different kinds 

 of atoms. 



The equation has been thrown into the form 



W=N 121 



in the previous pages so as to render the whole as definite as 

 possible. But no claim is made as to 1*21 being the best value 

 to give to g. The value of g should be obtained from the list of 

 elements from lithium to samarium, augmented or not as may 

 appear most just to anyone who wishes to redetermine the 

 value of q. If there is any truth in the views explained in 

 this paper, the constant g is of primary importance, and its 

 exact determination becomes a matter of interest. The value 

 which will be obtained depends on whether or not an element 

 analogous to manganese is to be assumed between molybdenum 

 and ruthenium ; g will be slightly less or greater according 

 us a gap is or is not left. 



The method might, however, be used to test this point. 

 By considering only elements as far as molybdenum the best 

 value of g could be found, and then the sum of the deviations 



Phil. Mag. S. 6. Vol. 4. No. 19. July 1902. I 



