280 ANNUAL. REPORT SMITHSONIAN INSTITUTION, 1920. 



of Cuba, Jamaica, Porto Rico, and Haiti, we do not possess a single 

 analysis. 



Most of the countries of Europe are well represented, but for the 

 most part with not very complete analyses. North America is well 

 known, especially as to the rocks of the United States and southern 

 Canada. The analyses of both these countries are of exceptionally 

 high general quality. Parts of Australia, especially New South 

 Wales, Victoria, and Queensland, with New Zealand, are well repre- 

 sented, as is also British Guiana, and it should be said that the 

 analyses of Australia and British Guiana rocks are almost the only 

 ones that, as a whole, are comparable as to accuracy and complete- 

 ness with those of the United States, which holds a preeminent posi- 

 tion through the labors of the chemists of the United States Geologi- 

 cal Survey. 



A second disturbing factor, and one that has been often advanced 

 against the validity and representativeness of the estimates of the 

 average composition of rocks, is that the true relative amounts of 

 various rocks are not properly represented because of the selection 

 of material for analysis. It has frequently happened that the 

 petrographer has had analyzed rather the rare or most interesting 

 rock types than those which, though much more abundant in the 

 region described, are of more usual character. While this is often 

 to be expected and, from a special point of view, is almost justifiable, 

 yet it certainly may involve a serious disturbance in the estimation 

 of the composition of the crust as a whole. This is so, because the 

 most interesting types, often ipso facto, are much less abundant than 

 the common ones, so that, as regards the relative masses of the va- 

 rious kinds of rocks in any given region, they are disproportionately 

 represented. It is needless here to give examples, of which there are 

 very many ; it would lead us too far into the technicalities of petrog- 

 raphy. 



Although this objection is serious, and is entitled to consideration, 

 yet it would seem, on detailed examination, not to be of the over- 

 whelming character that is often attributed to it. For one thing, the 

 satellitic rocks of the dikes and other small bodies (which are most 

 prone to furnish "interesting" types), tend to be complementary to 

 each other, through processes of differentiation, and so, as Dr. 

 Clarke, says, " they tend to compensation, and so to approximate to 

 the true mean." Also, as in a number of examples from many locali- 

 ties that could be cited, only the main body or the most prominent 

 types have been analyzed, chiefly because of the labor or expense of 

 making chemical analyses of rocks. Again, as I have pointed out 

 elsewhere, the more " basic " rocks — that is, those that are lowest in 

 silica and highest in iron oxides, magnesia, and lime — are most 

 liable to alteration, so that many of their analyses would be ex- 



