480 



PROCEEDINGS OF THE NATIONAL MUSEUM. 



VOL. 37. 



The accompanying figure represents graphically the relationships 

 brought out by columns I and III of the preceding table. 



A few words may be added in regard to the method here used of 

 comparing the composition of a described rock with composite analy- 

 ses. Usually it is customary, in quoting an analysis, to include for 

 comparison several analyses of related rocks. The analyses selected 

 are apt to be those which correspond most closely with the rock 

 described. Occasionally, however, a dissimilar analysis may be 

 inserted, or a series of analyses ranging from a close approximation 

 to an extreme of the same type; but in any case the chief feature 

 brought out is whether the individual occurrence is like or unlike 

 other individual occurrences, and unless one has a sufficient work- 

 ing knowledge of analyses to visualize an average the exact quanti- 

 tative relation of a member of a type to that type does not completely 



CaO 



MqO 



510,-1::;^;^ 



;;=>5lOL 



A 1,0, 



Graphic representation of the constituents or the olivine-diabase described (solid lines) 



AND THE AVERAGE NORMAL DIABASE (DASH LINES). 



appear. Now that Washington's admirable collections of rock analy- 

 ses are available, and in these the reliable analyses are grouped, it is a 

 matter of little difficulty to determine, by averaging, the composition 

 of any rock type desired, and in doing this to note the extreme 

 values, both low and high, for each component. A column thus 

 obtained, including the average and extreme values for each compo- 

 nent, serves well in determining the significance of each constituent, 

 and the analysis as a whole, of any rock under discussion. The use 

 of extreme values must be made with caution, however, as such values 

 are more likely than normal values to be the result of some error in 

 analysis. 



Certain differences between a composite olivine-diabase and a 

 composite normal diabase may be observetl in columns II and III, 

 but a discussion of these does not come within the province of the 

 present paper. 



