215] V. B. Mathews 17 



From the foregoing it would appear that mature workers 

 even in establishing their types believe it allowable to include 

 rocks whose dominant constituents show departures ranging 

 from 10% to 30% from their mean values. 



The use of average analyses in the description of rock 

 groups may or may not prove more serviceable. "While there 

 are more possible variations in the kinds and proportions of 

 minerals the graphs may show no wider variations than those 

 noted in discussing individual types. Thus the diagram 

 (Fig. 7) showing pyroxenites, exclusive of the websterites 

 (Fig. 8) although representing several kinds and many ex- 

 amples of pyroxenites, is not much more confusing or variable 

 than that for the websterites by themselves. The peak for 

 lime is more marked but the predominance of the magnesia 

 and lime with the subordination of the irons and alumina are 

 nearly as clear. The more general diagram carries aberrant 

 types like the pyroxenite from Rosetown, N. Y. (high alumnia 

 and low magnesia) and the magnetite pyroxenite from Cen- 

 tral City, Colo, (high in irons) . The former is a poor analy- 

 sis while the latter is recognized as aberrant and their inclu- 

 sion in any general average analysis is doubtful. 



Similar graphs of gabbro, dacite and camptonite show 

 fairly well defined figures which indicate that the impressions 

 gained from average analyses while incomplete may not be 

 incorrect in the major essentials. Average analyses cannot be 

 expected to bring out minor " clusterings " or many of the 

 relationships in constituents which are disclosed by the simple 

 serial diagrams here employed. 



The same is true of several of the systems of projection now 

 employed in petrography. These, moreover, often require ex- 

 perience and maturity beyond that of the average student of 

 systematic petrography for their complete appreciation. 



