CLASSIFICATION OF IGNEOUS ROCKS 277 



is too scanty for us to dogmatise, but in this case also it is 

 impossible to believe that the units selected for the quantitative 

 classification correspond in any but the most imperfect manner 

 to those actually present in molten magmas, which probably 

 vary in molecular composition within wide limits under different 

 conditions of temperature and of pressure. 



But even if the elements of differentiation were sufficiently 

 permanent to be taken as units of classification, there would 

 be no justification for the arbitrary division of group from 

 group by lines drawn at arithmetical intervals, which can 

 correspond to nothing that has occurred in the evolution and 

 differentiation of igneous rocks. 



These purely arbitrary boundaries would in many cases 

 separate a common and simple rock-type into two or more 

 portions, with different names and positions in the classification. 

 Indeed, there would be few rock masses sufficiently uniform to 

 be included under one name, for a slight increase in the propor- 

 tion of one mineral to another would frequently be sufficient to 

 transfer the rock from one order, or even class, to another. 



The authors defend the position they have taken up by the 

 contention that rocks pass into each other by imperceptible 

 gradations, and claim that there are no natural groups that 

 could be adopted in classification. 



This view cannot, I believe, be supported. Rocks do not 

 differentiate alike in all directions. There are certain associa- 

 tions of oxides which increase or decrease together, so that in 

 the extremes at least definite types must become prominent ; and 

 while it is true that intermediate links occur which bridge over 

 the interval between rocks of extreme types, they are not, there is 

 every reason to believe, distributed with anything like uniformity. 



Prof. Iddings has constructed diagrams in which each rock 

 is represented by a point whose horizontal co-ordinate is 

 proportional to the percentage of silica and whose vertical co- 

 ordinate to the ratio of the total alkali molecules to those of 

 silica. In these diagrams he has included all the satisfactory 

 analyses available so as to indicate the frequency of different 

 types as far as concerns the amounts of silica and the alkalies. 

 A brief examination of these diagrams shows that there 

 are definite regions where the dots representing analyses are 

 closer together than in the intermediate areas, and we have 

 no right to assume, with Prof. Iddings, that this is due to the 



