98 J. P. IDDINGS 



cation appears to be more general than was at first suspected ; 

 and by means of both laws a crude approximation to the com- 

 position of the feldspathic constituents in any rock may be 

 obtained. 



With regard to the occurrence of the minerals of the second 

 group, it is a general law that they are most abundant in pro- 

 portion to the diminution of silica and of the feldspathic miner- 

 als, so that an expression of the variations in these will also 

 express inverse variations in the ferromagnesian minerals. To a 

 very considerable extent the occurrence of the orthosilicate, 

 olivine, and of the metasilicate, hypersthene, depends on the 

 available silica in the magma, the former occurring in rocks 

 with lower silica. 



From these considerations it is evident that a comparison of 

 the chemical composition of rocks may be undertaken so as to 

 brinsf out the relations between the alkalis and the silica, and 

 between these components and the feldspathic constituents, and 

 that indirectly some notion may be had of the relations of the 

 ferromagnesian constituents. Since two coordinates or variables 

 only can be employed in a plane diagram or table, and a third 

 involves a representation in three dimensions, or by some other 

 device, it is necessary to reduce the factors to at least three. 

 Those selected for the present investigation are, first, the silica 

 on account of its important role in conditioning the character of 

 the alkali-feldspathic constituents; second, the molecular ratio 

 between the alkalis, soda, and potash, taken together, and the 



... (Na»0 + K„0\ . . ... 



silica I qvq — — )» f° r this corresponds to the relative pro- 

 portions of orthosilicate, metasilicate, and polysilicate feld- 

 spar and quartz. The amount of silica in each case was made 

 the abscissa and the alkali-silica ratio the ordinate, and since 

 these quantities are not of the same kind, it is not necessary 

 that the scale of parts should be the same for both. For con- 

 venience the amounts of silica were plotted directly from the 

 percentages in each analysis. By this means it is possible to 

 investigate the distribution of all the rock analyses studied with 



