we 
TRANSACTIONS OF SECTION B. 705 
research, might easily be brought about; not by any exercise of authority on the 
part of the committee, but by mutual consent of the investigators, working in con- 
Aerence, and aided by the suggestions which the international body might develop. 
In short, the committee could not order, but it might persuade; and in this direc- 
‘tion its influence ought to be decidedly beneficial. Even if it did no more than to 
point out the essential problems to be solved, it would fully justify its existence. 
There is one more general problem which the proposed committee should 
consider: that of methods. What are the best experimental methods for the 
determination of atomic weight ratios, and how shall the data be handled mathe- 
matically? On each division of this question there is something to be said. The 
existing methods, the methods which are commonly employed, are somewhat 
conventional in their character, and need exhaustive scrutiny. They are not 
sufficiently varied in their details to eliminate all danger of constant or cumula- 
tive errors, and new lines of attack, new points of view, ought to be considered 
-and developed. L : 4 
At present, the data relative to atomic weights are treated like successive 
Jinks in a series of chains, each link to be considered separately ; while in reality 
they form an interlacing network of interdependent quantities which should be 
“liseussed in some broadly general way. 
To illustrate my meaning, let us consider a specific ratio, 
CaF, : CaSO,, 
which has been repeatedly measured in order to determine the atomic weight of 
fluorine. In this case a series of measurements is made, involving of necessity 
some error which may be great or small. From these data the atomic weight in 
‘question is computed, with the assumption that the atomic weights of calcium 
and sulphur are known. But these antecedent values are themselves in error by 
small but unknown amounts, and these errors are superimposed upon the experi- 
mental error of the ratio itself, so that all three appear in the final result of the 
calculation. The errors may be compensatory in part, or they may be cumula- 
tive; and which is the case we cannot certainly know. In a proper reduction 
-of the data the ratio should contribute to our knowledge of all three of the atomic 
weights represented, and its error should be distributed among them instead of 
being piled, with others, upon one. That is, the ratio should not be discussed by 
‘itself, but should be combined with other ratios in such a manner that several 
related atomic weights might be determined simultaneously. This, I believe, will 
be the method of the future; and ultimately all trustworthy evidence, concerning 
all atomic weights, will be put into one set of normal equations with simultaneous 
solution for all. First, the experimental errors will be made as small as possible ; 
after that they will be so uniformly distributed as to become inappreciable. For 
this procedure the mathematical method is well known, but the existing data 
are too incomplete for its present application. ‘The final reductions will not be 
possible for many years to come. Work, comparable with that of Stas and 
Morley, needs to be done for all the chemical elements, and done with a broad 
purpose in view; after that the mathematician can contribute his share to the 
solution of the general problem. 
It is greatly to be hoped that at some future time some of the great labora- 
tories may undertake systematic work of the kind I have indicated. The 
fundamental constants of chemistry are surely of equal importance with the value 
of the ohm, the form of the earth, or the solar parallax ; and institutions like the 
Reichsanstalt in Berlin, the International Bureau of Weights and Measures at 
Paris, or the Davy-Faraday laboratory of the Royal Institution, might well con- 
tribute to their determination. In this direction an international committee 
conld exert an influence far beyond that of any individual], or even of any one 
‘society, but the problems at issue must first be clearly understood and formulated. 
To clear the ground, to arouse interest, and to stimulate systematic research, are 
importart functions of the proposed body. 
Yours very truly, 
F. W. CLarke. 
1899. Zi 
