1882.] 
Analyses of Books. 
419 
whenever exceptions appeared it was argued, not without plausi- 
bility, that with greater correctness these anomalies would dis- 
appear. This hope did not, however, hold good. Even when 
Dumas suggested the important modification that the atomic 
weights of the elements might be multiples of that of hydrogen 
by halves or quarters, as well as by whole numbers, the discre- 
pancies were numerous, and the hypothesis fell more and more 
into discredit. 
The recent researches of Marignac, and especially those of 
Stas, which have justly been regarded as models of accuracy, 
were generally recognised as fatal to Prout’s law, since the atomic 
weights as thus determined agreed neither with whole, half, nor 
quarter multiples of that of hydrogen, and the variations seem 
wholly outside the range of experimental errors. 
Recently, however, Dumas pointed out a source of error, pre- 
viously overlooked, in the experiments of Stas. Many of the 
determinations of the latter chemist had necessitated the use of 
pure metallic silver, which was melted under a layer of borax and 
saltpetre. Dumas found that such silver absorbed and retained 
oxygen in ponderable quantities, thus rendering a correction 
necessary. On applying this correction Prof. Clarke finds that 
out of seven elements one has its atomic weight unaffected, one 
is rendered less accordant with Prout’s hypothesis, but five are 
brought more closely than before to even multiples or half mul- 
tiples of hydrogen. Thus in five cases out of seven the increased 
accuracy springing from the correction of Dumas tells in favour 
of Prout. It is also very possible that other metals, beside silver, 
when used in the determination of atomic weights, may also 
occlude gases. 
The following argument in favour of Prout is put forward by 
Prof. Mallet : — Out of eighteen of the elements whose weights 
are best determined, ten have values which fall within one-tenth of 
a unit of whole numbers. Now if we examine what is the mathe- 
matical probability of such an approximation to the requirements 
of Prout’s law being a mere coincidence, it will be found to be 1 
as against 1097-8. In short, if half multiples are considered 
admissible, it seems more probable that the few apparent excep- 
tions should be due to undetected constant errors than that the 
great number of close agreements should be accidental. 
Prof. Clarke admits that he began his re-calculation of the 
atomic weights under a strong prejudice against Prout’s hypo- 
thesis, but that the facts forced him to give it a respectful con- 
sideration. 
It would of course be rash and premature to pronounce the 
question decided either way, but, having regard to the evidence 
collected and arranged by Prof. Clarke, the balance of proba- 
bility seems to us to be in favour of Prout’s hypothesis. 
