366 



SCIENCE 



[N. S. Vol. XXXIV. No. 873 



their relative abundance, it is possible to 

 calculate approximately the composition 

 of the parent-magma of a province. Noting 

 that nearly identical assemblages of rocks 

 sometimes occur in widely separate prov- 

 inces and at different geological periods, 

 we have some reason for expecting that the 

 provincial parent-magmas may ultimately 

 be reduced to a limited number of types. 

 Whether these types will be sufficiently 

 definite to serve as a basis of classification 

 it is too early to say. 



For the sake of argument, I have taken 

 chemical composition as the criterion. It 

 is certain, however, that a rock-magma 

 consists, not of free oxides, but mainly of 

 silicate-compounds, and the variation pro- 

 duced by magmatic differentiation is a 

 variation in the relative proportions of 

 such compounds. The characteristics com- 

 mon to a set of cognate rock-types will, 

 therefore, be more properly expressed in 

 mineralogical than in chemical terms. If, 

 to fix ideas, we take as representative of a 

 province its principal plutonic series, we 

 shall often find that some particular min- 

 eral or some special association of minerals 

 stands out as a distinctive feature. For 

 instance, in the charnoekite-norite series of 

 southern India the characteristic ferro- 

 magnesian mineral is hypersthene; in the 

 granite-gabbro series of the British Tertiary 

 it is augite ; and in the granite-diorite series, 

 which predominates among the "newer 

 granites" of the Scottish Highlands, horn- 

 blende and biotite. These three sets of 

 rocks, all of calcic facies, are easily distin- 

 guishable in isolated specimens. 



Each such rock-series embraces types 

 ranging from acid to ultrabasic. This 

 variation is ascribed to a later differentia- 

 tion of the parent-magma of the province, 

 and, therefore, in an arrangement based 

 on genetic principles, it will find expres- 

 sion, not in the main divisions of the 



scheme, but in the subdivisions. Here is an 

 essential difference between an ideal petro- 

 genetic classification and the petrograph- 

 ical systems which are, or have been, in 

 use. If we are content to limit our study 

 of igneous rods to specimens in a museum, 

 the distinction of acid, neutral, basic and 

 ultrabasic may seem to be of first impor- 

 tance. It has, in fact, been employed for 

 the primary divisions in some formal 

 schemes, e. g., in that put forward by 

 Lowinson-Lessing. In a less crude system, 

 like that of Rosenbusch, this element dis- 

 appears, but the underlying idea still re- 

 mains. There is a division into fam- 

 ilies, such as the granite-family and the 

 gabbro-family, but the term, in so far as it 

 implies blood relationship, is a misnomer. 

 The augite-granite of Mull is evidently 

 more closely related to its associated gab- 

 bro than it is, say, to the biotite-granite of 

 Peterhead or the hypersthene-granite of 

 Madras. 



The differentiation which evolves a 

 varied series of plutonic rocks from a com- 

 mon parent-magma is clearly not of the 

 same kind as that which gave rise to the 

 parent-magma itself. It appears that the 

 external mechanical element is here a less 

 important factor, and the variation set up 

 is, therefore, more closely in accordance 

 with the uninterrupted course of crystal- 

 lization. This is clearly indicated when we 

 compare the order of intrusion of the sev- 

 eral rocks of the series with the order of 

 crystallization of their constituent miner- 

 als. The history of the series is in a sense 

 epitomized in the history of each individ- 

 ual type, corresponding in both cases with 

 continued fall of temperature and progres- 

 sive change in the composition of the resid- 

 ual magma. In a large number of rocks, 

 more particularly those of complex consti- 

 tution, the order of crystallization follows 

 Rosenbusch 's empirical law of decreasing 



