VOLCANIC ISLANDS 327 



sitic, peridotitic, and "alkaline" magmas. Three other classes include 

 transitions between basaltic rock and each of the other three chemical 

 types or groups. An eighth class comprises quartz-bearing rocks, and a 

 ninth includes a few species not certainly referable to any of the other 

 classes. 



So far as the records go, the relative importance of these classes is 

 suggested by Table III, which shows the number of islands where each 

 rock species has been reported. 



Andesites and their genetic Associates 



In spite of the obvious difficulties, due to faulty nomenclature and 

 general lack of chemical analyses, it is clear that typical olivine-bearing 

 plagioclase basalt and as typical pyroxene andesite have been most often 

 discovered, and it is safe to hold that the two species mentioned are by 

 long odds the commonest lavas in Oceanica. The two occur side by side 

 in at least 17 of the islands listed. They are connected by transitional 

 varieties, including many olivine-free basalts, gabbros, diabases, etcetera, 

 as well as the species named in Class 2 of Table III. All these relations 

 are very often repeated on the continents, where, moreover, pyroxene 

 andesite and plagioclase basalt are associated, not only in eruptions of the 

 Kecent Period, but also in nearly all the great rock systems since at least 

 the late Precambrian. 



When a given kind of rock association persists, both in time and space, 

 it is reasonable to suspect syngenesis for the species involved. Elsewhere 

 the writer has elaborated the thesis that pyroxene andesite and certain 

 ultra-basic rocks may be complementary differentiates from plagioclase 

 basalt. 2 More recently Cross has concluded that the andesites and other 

 igneous rocks of the Hawaiian Islands have been derived from a general 

 basaltic magma by a process of pure differentiation. 3 The experiments 

 of Bowen have fully shown the possibility that the differentiation is based 

 on the gravitative separation of crystals, though the other possibility — 

 that the splitting is due to the gravitative separation of non-consolute 

 liquids chemically similar to the phenocrysts, can not be excluded. 4 In 

 view of this uncertainty, the present writer prefers to use the expression 

 "gravitative differentiation," rather than "fractional crystallization" or 

 "liquation," to indicate the method of splitting. 



In the writer's opinion, most of the rock species named in Classes 2, 3, 

 4, and 5 of Table III are best explained as formed by the more or less 



2 Jour. Geology, vol. 16, 1908, p. 401 ; Memoir 38, Geoi. Survey Canada, 1912, p. 782 ; 

 Igneous rocks and their origin, New York, 1914, p. 375. 



3 W. Cross : Professional Paper 88, U. S. Geol. Survey, 1915, p. 87. 



* N. L. Bowen : Amer. Jour. Science, vol. 39, 1915, p. 175, and vol. 40, 1915, p. 184. 



