2i6 ALBERT JOHANNSEN AND E. A. STEPHENSON 



great majority of types Notwithstanding the opinion of Williams, the 



Rosiwal method of estimation is capable of giving extremely accurate results. 

 The writer .... has found the chemical composition calculated from the 

 results of the Rosiwal measurements to be strikingly accordant with that 

 obtained by ordinary chemical analysis. 



In view of Williams' apparently careful study and rejection of 

 the Rosiwal and other microscopic methods, and of the increasing 

 use made of volumetric determinations of rock constituents by 

 other petrographers, it was thought desirable to repeat the experi- 

 ments with all possible precautions against errors in both the Rosi- 

 wal and the heavy solution methods. Williams' study is apparently 

 the only one in which there was a comparison of the percentages 

 determined by the microscopic measurement method and an 

 actual separation of the minerals; all others are comparisons 

 between the former and recalculated chemical analyses, with all 

 the possible errors in allotting constituents due to the uncertain 

 composition of the dark constituents and the possible soda content 

 of the orthoclase. 



The material chosen for the experiment by the present writers 

 was the Butte ''granite." It was selected because it had been chem- 

 ically analyzed, because there was available ample material from the 

 original locality, and because there were on hand six thin sections 

 and the two determinations by the Rosiwal method made by others 

 as mentioned above, and five determinations by students on one 

 slide (Table III, column I). Furthermore, the material is of 

 fairly uniform grain and composition in widely separated localities, 

 thus giving a fairer test than would a rock differing possibly in 

 different sections and thus probably different from the material 

 analyzed. 



The six different sections were measured by the Rosiwal method 

 from one to five times each, fourteen determinations being made in 

 all. The average of each slide, reduced to weight percentages, is 

 shown below. In Table III the first six columns represent the 

 six sections; the number in parentheses after each indicates the 

 number of times it was measured. The seventh and eighth columns 

 are the determinations taken from Cross, Iddings, Pirssons, and 

 Washington and quoted above. The variations between the read- 



