22 CLARKE— STATISTICAL METHOD [April i8 



means of composite analyses. That is, instead of averaging analyses, 

 average mixtures of many rocks were prepared,^ and these were 

 analyzed once for all. The results appear in the preceding table. 



A. Composite analyses of 78 shales ; or more strictly, the av- 

 erage of two smaller composites, properly weighted. 



B. Composite analysis of 253 sandstones. 



C. Composite analysis of 345 limestones. 



In attempting to compare these analyses with the average com- 

 position of the igneous rocks, we must remember that they do not 

 represent definite substances, but mixtures shading into one another. 

 The average limestone contains some clay and sand, the average 

 shale contains some calcium carbonate. Furthermore, they do not 

 cover all of the products derived from the decomposition of the 

 primitive rock, for the great masses of sediments on the bottom of 

 the ocean are left out of account. The analyses of the latter are too 

 few to give conclusive averages, but the data published in the Chal- 

 lenger Reports"^ indicate a difference between them and the terri- 

 genous deposits. The "red clay," for example, which covers 51,- 

 500,000 miles of the ocean floor, at its greatest depths, is much richer 

 in iron than the average shale. In twenty-one analyses of this sedi- 

 ment a mean of 13.61 per cent, of ferric oxide was found. The 

 thickness of this deposit is quite unknown. There are also meta- 

 morphic rocks to be considered, such as amphibolites and serpen- 

 tines ; although their quantities are presumably too small to seriously 

 modify the final averages. They might, however, help to explain a 

 deficiency of magnesium which appears in the sedimentary analyses. 

 Partly on account of these considerations, and partly because the 

 sedimentary rocks contain water and carbon dioxide which have 

 been added to the original igneous material, we cannot recombine 

 the composite analyses so as to exactly reproduce the composition 

 of the primitive matter. To do this it would be necessary also to 

 allow for the oceanic salts, which represent, in part at least, losses 

 from the land; but that factor in the problem is perhaps the least 



' These mixtures were prepared by G. W. Stose, under the direction of G. K. 

 Gilbert. The analyses were made by H. N, Stokes in the laboratory of the U. S. 

 Geological Survey. See Bull. 228, p. 20. 



2 " Volume on Deep Sea Deposits," p. 198. 



