260 //. H. Robinson — Chemical Analyses of Igneous Rocks. 



and not over 10 per cent of water are used. The high per- 

 centage of water, is unavoidable because the olivine-bearing 

 rocks, which contain large amounts of this element, are par- 

 ticularly susceptible to alteration. The results are : 



Quality of work Number of anals. Ave. summation 



Superior.. 29 100-14 



Inferior 10 100-60 



The analyses in the superior group have the same average 

 summation as the general average ; presumably the common 

 plus error in the determination of magnesia is sufficiently small 

 to be counterbalanced. On the contrary, the inferior group 

 has a high summation. It is 0-47 per cent above the general 

 average, which is a significant amount and must largely repre- 

 sent the plus error in the determination of magnesia. In 

 general it is no doubt true, except for the best work, that mag- 

 nesia has been determined in excess of the real amount by the 

 older analytic methods. Gooch and Austin* have shown the con- 

 ditions which cause this error, as well as how it may be avoided. 

 On the basis of their special determinations an error of 1 in 40 

 may be expected. The inferior analyses just given show a 

 smaller average error than this, caused probably by counterbal- 

 ancing errors in the determination of other constituents. It 

 is possible, of course, that the plus error in the determination 

 of MgO itself might raise the general average summation 

 above 100 per cent. The average igneous rock of the earth's 

 crust contains about 4 per cent of M^O and on the proportion 

 of 1 :40 the plus error would be - 10 per cent, However, so 

 large an average error is not allowable, because in many rocks 

 MgO undoubtedly has been correctly determined. Yet with 

 this qualification a small part of the constant plus error in the 

 general average summation might well be due to this cause. 



An unbalanced error might be expected in those analyses in 

 which iron has been determined wholly as ferric or as ferrous 

 oxide. Such a procedure should affect the summation pro- 

 vided the oxide not determined is present in any considerable 

 amount. Theoretically, if an analysis shows a total of 15 per 

 cent of ferric oxide, for example, and this includes 5 per cent 

 of ferrous oxide, the summation will be +0"5 per cent in 

 error. If 15 per cent of ferrous oxide is determined and this 

 includes 5 per cent of ferric oxide, then the summation will be 

 — - 6 per cent in error. To test these errors in actual summa- 

 tions analyses have been selected in which the iron, wholly as 

 one oxide or the other, is greater than 10 per cent and in which 

 magnesia is less than 15 per cent. Magnesia can not be set at 



* Gooch, F. A., and Austin, M., The Constitution of the Ammonium Mag- 

 nesium Phosphate of Analysis, this Journal, 4th ser., vol. vii, p. 187, 1899. 



