1912-13.] The Absorption of Light by Inorganic Salts, No. XI. 159 
applied it to all the well-defined absorption bands on which I could get 
measurements, and found : — 
(1) For certain aqueous solutions of anilin colouring matters pe/m was 
of the order 1*8 10 7 . These bands are therefore caused by the vibrations of 
electrons, and in their case the assumptions of the theory are justified. 
(2) For the same colouring matters in certain other solvents pe/m was 
much smaller. 
(3) The values for aqueous solutions of cobalt chloride and uranyl 
nitrate suggested ions. 
It is somewhat arbitrary to assume, as is usually done, that in the 
medium there are only electric doublets or single electrons and ions, and 
not more complicated systems. I felt that progress could not be made 
unless more complicated systems were constructed. Drude’s results for 
fluorite, etc., seemed to supply a clue. Simply by taking the chemical 
formula and assuming that the chemical bonds were electric forces and 
acted something like spiral springs, I thought it might be possible to get 
systems giving the spectra of some simple molecules. But it was no use 
constructing such systems until there were observations available to test 
them with. My purpose in starting the experiments recorded in this 
series of articles was to get such observations. The salts experimented on 
were chosen so as to make the working out of a dynamical constitution 
for them as easy as possible. For example, if the fluorides, chlorides, 
bromides, and iodides of cobalt and nickel had dynamical constitutions of 
the nature I supposed, there would be some sort of progression in the 
spectrum as the weight of the acid radical increased. I hoped before 
starting the observations that water of hydration and of solution would 
not have much influence on the spectra — that in the case of cobalt 
ohloride, for example, the chief absorption bands would depend only on 
the bonds connecting the cobalt and chlorine atoms. 
The results obtained in the first four articles of the series and repre- 
sented in the curves given on pp. 540 and 541 of the third article completely 
shattered this view. These curves showed that the effects of the acid 
radical and base were roughly additive. In the six nickel salts we had 
three bands occurring throughout at the same places in the spectrum. In 
the six cobalt salts we had two bands occurring at the same places in the 
spectrum. The concentration for which these curves were taken was much 
too great to explain the similarity by ionic dissociation. 
When the concentration was increased almost to saturation, however, 
changes occurred, particularly in the case of the chlorides and bromides. 
Also, on heating the crystallised salts so as to drive off the water of 
