62 2 HENRY S. WASHINGTON 



That this identity of the two, based on mineralogical grounds, 

 is correct, is substantiated by a chemical analysis of one of 

 Derby's specimens made by myself. For this purpose an 

 apparently medium specimen was chosen, composed largely of 

 a violet-brown augite, with some magnetite (more than in the 

 Arkansas rock) and only a little nephelite (less than in the 

 other). No biotite was present, and only traces of apatite. 

 This analysis, given in III of the table, is most remarkably close 

 to that of the Magnet Cove jacupirangite in all respects, except 

 the iron oxides. Indeed the figures for silica, magnesia, lime, soda, 

 water, titanic acid and manganese are close enough to belong to 

 duplicate analyses of the same specimen, and those for alumina 

 and potash do not differ greatly. The higher iron oxides are 

 of course connected with the more abundant magnetite, but, 

 apart from this, the mineralogical composition is closely similar. 



The closest known analogue of these rocks is probably the 



pyroxenite of Brandberget, an analysis of which is given in IV 



above. The only noteworthy differences are in SiOg and Fe^Og. 



That of the former apparently conditioned the formation of 



nephelite in the Magnet Cove and Brazil rocks and plagioclase 



at Brandberget, while the higher ferric oxide of II and III is to 



be connected partly with the more abundant magnetite in the 



former. 



Henry S. Washington. 



\^To be continued. \ 



