A. S'arlcer — British Igneous Rocks. 221 



necessary, I have halved the figure, and taken the titanic acid as 

 0-69 per cent. The difference between this and 0-12 is to be 

 subtracted from the silica and alumina, and half has been taken 

 from each. The same reasoning does not apply, at least in the same 

 degree, to manganese and phosphoric acid. Manganous oxide was 

 estimated in 103 analyses and recorded as traces in 51. The mean 

 for these 154 rocks is 0*40, and this figure may be adopted in 

 the general average without fear of much error. The difference 

 between 0-40 and 0-15 is to be subtracted from the magnesia. 

 Phosphoric anhydride was estimated in 35 analyses only and 

 recorded as traces in 20. The mean for these 55 rocks is 0*25, 

 and this figure may be adopted in the general mean, the difference 

 between this and 0-02 being accordingly subtracted from the 

 alumina. The result as thus corrected is seen in column II. In 

 the paper already cited Clarke gives the mean analysis of 880 

 miscellaneous rocks, American and others; but I quote for com- 

 parison, instead of this, a mean analysis of American rocks 

 exclusively, given in a later work by the same author.' This, 

 which is calculated from 680 complete analyses, supplemented 

 by a number of partial analyses, is given in column III. 



There are still certain points to be observed in comparing the 

 two columns. In 157 of our 397 analyses ferrous oxide was not 

 estimated. The figure for this constituent is therefore too low 

 and that for ferric oxide correspondingly too high. Again, the 

 water has not always been estimated in the same way. In many 

 of the analyses only the loss on ignition has been determined, 

 and by taking this as water 1 have no doubt increased unduly 

 the figure for that constituent. Carbonic acid, chlorine, fl^uorine, 

 and sulphur, when separately estimated in the analyses, have been 

 discarded. These, of which only the first is of importance, are 

 therefore partly included, together with ' loss ' and errors, in the 

 0-29 per cent, required to make up the total to 100. This defect 

 should really be somewhat greater, since a certain amount of 

 ferrous oxide has been reckoned as ferric. 



The British mean is seen to have a rather lower silica-percentage 

 than the American, though it agrees closely enough with the figure 

 (58-59) in Clarke's old mean for rocks from various quarters of the 

 globe. The difference is perhaps not a significant one. As regards 

 titanic acid and alumina, the two columns show a fair correspondence. 

 The British rocks are evidently richer in iron than the American. 

 The proportions of the two oxides are, as has been remarked, not 

 correct, and it is not possible to rectify them ; but converting the 

 whole to ferric oxide, we get 8*01 for column II, as compared with. 

 6-37 for column III. The relatively high percentage of manganese in 

 the British rocks is certainly of significance, and must be regarded 

 as a characteristic of the whole assemblage. The precise figure is 

 not to be insisted upon, since it has been reached in a rather 

 arbitrai-y fashion ; but even in column I, where all rocks in which 

 manganese had not been recorded were assumed to contain none, the 



1 Clarke & Hillebrand, " Analyses of Eocks " : Bull. 148, U.S.G.S., 1897. 



