I9I2.] CLARKE— SOME GEOCHEMICAL STATISTICS. 215 



even the areas exposed, we shall find ourselves relying in great 

 part upon arbitrary assumptions, a procedure fully as unphilo- 

 sophical as that which it would supplant. Estimates of that and 

 similar kinds have been made, most recently by Loewinson-Lessing, 

 whose figures give essentially the same result as that obtained by the 

 method he has criticized. The method by volumes is doubtless ideal, 

 but impracticable ; and the true, philosophical procedure is to do 

 the best we can with the available data. It is highly probable that 

 the rocks of minor importance will balance one another, the per- 

 sihcic and subsilicic varieties occurring in something like equal pro- 

 portions. This supposition is sustained by the groups of average 

 analyses which will presently be given. If we trust to individual 

 judgments, different observers will reach widely different conclu- 

 sions. Loewinson-Lessing supposes that the average rock may be 

 about the mean of an average granite and an average basalt; Daly^ 

 argues in favor of a fundamental basaltic magma ; Mennell, whose 

 experience has been gained in a granitic region, regards granite as 

 the dominant rock with all else of minor importance. Mennell 

 makes a strong argument in favor of his contention ; but there is 

 weighty evidence against it. If we study recent lavas, that is, the 

 rocks which issue from unknown depths far below the surface, we 

 shall see that rhyolite, the effusive equivalent of granite, is much 

 rarer than andesite or basalt. The Deccan trap, the Columbia 

 River ibasalt, the andesites of South America, the lavas of Iceland 

 and the Hawaiian Islands are good illustrations of this statement. 

 Moreover, the river waters which originate in areas of crystalline 

 rocks contain almost invariably an excess of lime over soda, which 

 would hardly be the case were granite predominant. Much so- 

 called granite is really either quartz diorite or quartz monzonite, 

 rocks which are probably far more abundant than has been com- 

 monly supposed. 



In order to test the method of averaging analyses we may now 

 compare the averages so far obtained by different computers, and 

 then pass on to averages of rocks from distinct and widely sepa- 

 rated areas. In these averages only the more important constitu- 



^Bull. U. S. Geol. Survey, No. 209, 1903, p. no. 



