THE CLASSIFICATION OF IGNEOUS ROCKS. 479 



(26) in a paper dealing with the volcanic ejectamenta of 

 Krakatoa. These belong to the hypersthene-andesite 

 group, and are comparable with the ancient and modern 

 lavas of Santorin, the rock of Buffalo Peaks in Colorado, 

 and the freshest of the Cheviot lavas. All these consist of 

 the same minerals in roughly the same relative proportions, 

 with residual glass of approximately the same composition 

 in all. Nevertheless these rocks range in total chemical 

 composition from an acid type with jo per cent, of silica to 

 a basic type with 52 per cent. This variation is connected 

 with the relative proportion of the acid glass to the total 

 minerals (on the whole basic), this proportion being 9 : 1 

 and 1 : 9 in the extreme types mentioned. In the author's 

 opinion these and other facts point to the conclusion that 

 " after the partial separation of a magma into crystals and a 

 colloid residue, the two may be separated by a process of 

 liquation, and subsequently be mingled again in varying 

 proportions". In a later paper on the intermediate lavas 

 of the Western Islands of Scotland the same writer applies 

 this idea to the associated rocks belonging to one volcanic 

 centre, that of Ardnamurchan, and expresses the opinion 

 that "there is no direct and necessary relation between the 

 porphyritic crystals of a volcanic rock and the magma by 

 which they are enveloped ". Without endorsing the rather 

 extreme conclusion we may note that we have in this sugges- 

 tion a method of differentiation in rock-magmas demanding 

 only mechanical means, and possibly capable of producing 

 important results. 



Apart from the question just raised, if we know the 

 component minerals of a holocrystalline rock, and approx- 

 imately the relative proportions in which they are present, 

 we have a sufficient index of the place of the rock in any 

 scheme founded upon total chemical composition. It is 

 interesting to note that the converse does not hold ; a bulk- 

 analysis of a rock would not always enable us to predict its 

 mineralogical constitution. For example, Iddings (33) has 

 pointed out that some of the lamprophyres are identical in 

 chemical composition with certain leucite-bearing and other 

 lavas, although the two sets of rocks differ widely as regards 



