Vol. 70.] COMPOSITION OF ROCKALLITE. 299 



Classification. 



In the current classifications rockallite is generally considered as 



a highly sodic variety of the alkali-granites, although nothing is 

 known of the structural geology of the islet, except that the 

 rockallite overlies an apparently bedded or foliated rock of unknown 

 character. In the 4th edition of his ' Mikroskopische Physiographie ' 

 Rosenbusch described it under the granites (vol. ii, pt. 1, 1907, 

 p. 79), but is inclined (ibid. p. Gil) to consider it as a grorudite 

 (a?girine-quartz-tinguaite), belonging to the dyke-rocks, with which 

 also Prof. Judd compares it. This is an apt example of the fallacy 

 of basing the classification of a rock on the particular portion of 

 the lithosphere where it happened to consolidate, rather than only 

 on characters inherent in the rock itself. 



According to the Quantitative Classification, the norms of the 

 two analyses figure out as follows : — ■ 



I. II. 



Q 37-80 30-18 



Or 1-12 



Ab 24-10 25-15 



Z 1-83 



Ac 30-95 37-88 



Di 1-70 2-64 



Hy 5-08 



Mt 0-23 



II 0-61 



Ap 0-17 



They both, therefore, fall in the subrang rockallose, with 

 the general symbol III. 8. 1. <3. Makins's specimen however, is 

 transitional towards the dosalane class, and the domalkalic order, 

 with the exact symbol (II) III. (2) 3. 1. 5, and mine more nearly 

 central, having the symbol "III. 3. 1. 5. It is worthy of note, as 

 indicating the highly exceptional character of this rock, that these 

 two analyses of rockallite are the only representatives of the 

 subrang rockallose (III. 3. 1. 5) among about eight thousand 

 analyses now collected by me. 



Affinities of Rockallite. 



Both Judd and Rosenbusch remark on a similarity with the 

 grorudites of Brogger. They resemble each other in so far as the 

 qualitative mineral composition is concerned, but the grorudites 

 contain less pyroxene and more felspar, especially considerable 

 orthoclase (as shown by their analyses), and Judd finally considered 

 them as quite distinct from his rock. Similar distinctions also 

 hold good for the pantellerites, with which Judd tentatively 

 compares rockallite on the basis of Foerstner's analyses, but which 

 we now know to be incorrect in many particulars. 1 In recent 

 years there has been an extension of our knowledge of highly sodic 



1 Cf. H. S. Washington, Journ. Geol... Chicago, vol. xxi (1913) pp. 702, 

 707. and vol. xxii (1914) p. 19. 



