466 N. L. BOWEN 
For the systematic presentation of the deduced order of crys- 
tallization, the rocks (sub-alkaline) are treated in the broad types 
of systematic petrography. ‘There is, of course, complete over- 
lapping of the types and a similar gradual transition in order of 
beginning of crystallization. For example, the granite by increase 
of lime-alkali feldspar relative to alkali feldspar passes to the 
quartz diorite, and in the quenched equivalents alkali feldspar 
becomes less and less important as a mineral of early consolida- 
tion and lime-alkali feldspar more important, the result being a 
like passage from rhyolite to dacite. Again, the granite passes 
into the syenite by gradual decrease in quartz, and in the quenched 
equivalents quartz is represented in less and less amount as a 
mineral of early consolidation giving the passage from rhyolite 
to trachyte. 
In short, there is abundant evidence throughout of the tendency 
of special richness in any constituent to cause the early beginning 
of crystallization of that constituent as theory would lead one 
to expect. In every case a plutonic type whose distinguishing 
characteristic is richness in a certain mineral, or pair of minerals, 
has for its equivalent effusive type a rock whose chief characteristic 
is the appearance of this constituent or pair of constituents as 
minerals of early consolidation. 
It is hard to conceive a reconciliation of these well-known 
facts with the statement of any approximately constant ‘‘order 
of crystallization”? on the basis of which accessories, ferromag- 
nesian minerals, and lime-alkali feldspar should always be the 
minerals of early crystallization. 
The broad generalization which is necessary when any type 
is represented by a single diagram is perhaps likely to lead to some 
misunderstanding. In the gabbro diagram, for example, the 
ferromagnesian minerals are represented as beginning first, but 
this must be regarded as true only for the more femic gabbros. 
In a gabbro approaching anorthosite, plagioclase undoubtedly 
begins to crystallize considerably before the ferromagnesian 
minerals. The difficulty is inherent in the breadth of the types. 
The chief desire has not been to establish definitely the course 
of crystallization for all rocks referable to a given type, but to 
