in a Recalculation of the Atomic Weights. Ill 



without. The latter are chlorine, iodine, potassium, thallium, 

 glucinum, mercury, strontium, tellurium, chromium, moly- 

 bdenum, rubidium, copper, indium, vanadium, antimony, 

 tantalum, erbium, ytterbium, cerium, lanthanum, silicon, zir- 

 conium, platinum, rhodium, ruthenium, and gold. Of these, 

 chlorine, rubidium, and strontium agree closely with half 

 multiples of hydrogen ; while tellurium, molybdenum, iridium, 

 tantalum, erbium, cerium, silicon, zirconium, lanthanum, 

 rhodium, and ruthenium may be dismissed from consideration 

 as not sufficiently well determined to bear upon the problem 

 before us. The exceptions in the cases of potassium, iodine, 

 thallium, glucinum, mercury, chromium, vanadium, antimony, 

 ytterbium, platinum, and gold remain to be considered. 



- For potassium and iodine, we must remember that both of 

 these elements involve the constant error due to the occlusion 

 of oxygen by silver. This error is probably great enough to 

 throw the values for both elements outside the limit of varia- 

 tion above assigned. For thallium it has already been shown 

 that when its atomic weight is calculated with N0 8 =62, 

 Crookes's data give Tl= 204-008. The atomic weights of 

 glucinum and ytterbium, as given in the Tables, are calcu- 

 lated from analyses of the sulphates. If SO 3 =80,then Gl = 

 9*096 ; and Yb = 173-016. Both fall within, and one narrowly 

 within the limit of one tenth of a unit variation. In the case 

 of platinum, I need only say that Seubert's figures give values 

 ranging both above and below the even number 195 ; while 

 as for antimony, although the general mean is Sb = 120*231, 

 Crookes's analyses of the bromide, reckoned with Br = 80, give 

 almost exactly 120 for the value. For mercury, chromium, 

 vanadium, and gold, new determinations are desirable. 



Enough has been said to show that none of the apparent 

 exceptions to Prout's hypothesis are absolutely inexplicable. 

 As the figures actually stand, thirty-nine out of sixty-five 

 elements vary less than a tenth of a unit each from even 

 multiples of the atomic weights of hydrogen. Of the remain- 

 ing twenty-six, three conform to half multiples, three more 

 are legitimately incalculable so as to fall within bounds, and 

 eleven have been so defectively determined that the assigned 

 values can carry scarcely any weight. The remaining nine are 

 still subject to slight revision. In short, the many agreements, 

 which include three fourths of the weH-deteTmined atomic 

 weights, renders Prout's hypothesis very highly probable. It 

 is more likely that the seeming exceptions are due to unde- 

 tected constant errors, than that the great number of coinci- 

 dences should be accidental. The mathematical probability 

 in favour of Prout's hypothesis I have not yet calculated ; 



