THE VOLCANIC ROCKS OF THE ANDES. 1 67 



development of the rock. Consequently he concludes that upon 

 purely theoretical grounds a chemical classification would be the 

 most desirable. But from the present condition of our knowl- 

 edge this would be impracticable. 



Moreover, he observes, that while the chemical analysis of 

 an unaltered rock furnishes us with the proportions in which the 

 elements existed in the molten magma, it is very probable that 

 rocks of like chemical composition may have been derived from 

 magmas consisting of quite different silicates (that is, possessing 

 different molecular constitutions). And this he suggests is one 

 of the reasons why eruptive rocks with corresponding analyses 

 can exhibit quite different mineralogical compositions. In a 

 foot note he observes that another cause for this phenomenon may 

 lie in differences in the process of solidification, which may 

 affect the rearrangement of the compounds originally in the 

 magma. To this extent the mineralogical composition is 

 dependent on the genesis of eruptive rocks, which he considers 

 essentially the same as their geological mode of occurrence. 



Since we are not in a position to infer the original molecular 

 constitution of a magma from its chemical analysis, he does not 

 consider a chemical basis of classification applicable. Neverthe- 

 less he states that a comparison of the chemical composition of 

 the andesitic lavas with their mineralogical composition shows 

 that certain differences of chemical composition go hand in 

 hand with others of mineral development, and with these are 

 also connected modifications of structure. 



The mineralogical features are the most pronounced, and 

 are therefore selected as a basis of classification. The first sub- 

 division is based on the presence or absence of quartz, and the 

 groups become andesites and dacites. They are not, however, 

 distinctly separated from one another, being connected by gradual 

 transitions. But this grouping, as a purely mineralogical one, 

 fails, as he himself points out, in cases where quartz has not 

 crystallized, as in certain dacitic glasses. 



In the further subdivision of these groups the ferromagnesian 

 minerals are employed as distinguishing characteristics, and the 



