26 



CLARKE— STATISTICAL "METHOD 



[April 18 



The final average differs from that of the igneous rocks alone, only 

 within the limits of uncertainty due to experimental errors, and the 

 assumptions made as to the relative proportions of the sedimentaries. 

 If the work were ideally exact, the last column of figures should 

 differ from the first symmetrically, being higher in water and carbon 

 dioxide and lower in all other constituents. Lime and potash, how- 

 ever, show small gains, which are abnormal, and indicative to some 

 extent of the errors above mentioned. It is possible that excessive 

 weight has been assigned to the limestones, but on that theme it is 

 hardly worth while to speculate. The values chosen for the sedi- 

 ments are approximations only, and nothing more can be claimed 

 for them. They seem to be near the truth ; as near as we can ap- 

 proach with data which are necessarily imperfect, and so they may 

 be allowed to stand without further emendation. 



Now, with the help of this new average, we are in a position to 

 compute the relative abundance of the chemical elements in all 

 known terrestrial matter. For this purpose, the column is restated 

 in elementary form, with an arbitrary allowance of 0.5 per cent, 

 for all of the elements not specifically included in it. As for the 

 atmosphere, it is represented in the final result by 0.02 per cent, 

 of nitrogen, which is a little too low. The mean composition of the 

 lithosphere, the ocean and the atmosphere, then, is as follows : 



93 Per Cent, 



Lithosphere. 



Oxygen 47.07 



Silicon 28.06 



Aluminum 7.90 



Iron 4.43 



Calcium 3.44 



Magnesium 2.40 



'Sodium 2.43 



Potassium 2.45 



Hydrogen .22 



Titanium .40 



Carbon .20 



Chlorine .07 



Bromine 



Phosphorus .11 



Sulphur .11 



Barivim .09 



Manganese .07 



Strontium .03 



Nitrogen 



Fluorine .02 



All other elements. .50 



