236 J. P. Cooke on the Law of Definite Proportions 
change in the general crystalline form. It is not meant to assert 
that the variations are identical in character with those of Sb Zns 
and Sb Zna2, but only that there is a strong probability that this 
is the case, which in the first two instances amounts almost toa 
certainty. 
If variations in composition of such magnitude are possible 
when the force of chemical affinity is weak, it is highly probable 
that some variation may occur when the force is strong, and, 
whatever view may be taken of the cause of the variation it will 
now become a matter of importance to ascertain whether many 
discrepancies in analyses hitherto referred to imperfections in 
the process, may not be owing to the same cause which influ- 
ences the composition of the crystals of zinc and antimony. For 
_ this purpose, it will be best to make several analyses of the.same 
compound, prepared under circumstances differing as widely as 
possible, and then to apply to the results “ Peirce’s Criterion for © 
the rejection of doubtful observations.” Such investigations 
will be greatly simplified by tables prepared by Dr. B. A. Gould* 
for facilitating the application of this criterion to which I would 
- refer all chemists who are inclined to take up this line of inves- 
tigation. 
J am well aware that in announcing the existence of perturba- 
tions of the law of definite proportions I am calling in question 
one of the most fundamental dogmas of chemical philosophy, 
and that the new doctrine will have to encounter prejudice on 
this very ground. This law is so intimately associated in many 
minds with the atomic theory, that, to such, absolute definiteness 
seems to be its essential characteristic ; nevertheless, I can not 
no case in which there is absolutely none. This same character 
which pervades the other BA a laws of nature, I claim for 
* Astronomical i vol. iv, p. 8 
+ Ihave used the onsale ‘sre: to designate a class of laws of nature 
which are empirical in rane character inasmuch as they are oben? not mye 
although their de tp has not been discovered, but w 
to which empirical is co pnby- epellid: * Arg 
Presid 
* 
