6*4 PEOCEEDIJS'GS OF THE GEOLOGICAL SOCIETY. 



It will be seen at once that these examples do not very greatly 

 affect the main argument, because there is still an enormous area of 

 ocean left in which the only islands are volcanic. But still the 

 instances mentioned are worthy of notice, because they show that 

 there are exceptions to the general rule that no palaeozoic or meso- 

 zoic rocks occur on oceanic islands, and the additions made to our 

 knowledge in this respect of late years render it probable that other 

 instances remain to be discovered. Of course much depends on the 

 definition of an oceanic island ; but the accuracy of the term can 

 scarcely be contested in the case of South Georgia. As it is impos- 

 sible for denudation to hollow out the sea-bottom beyond a few 

 fathoms below the surface, the isolation must in all such cases be 

 due to depression. 



There are three other facts that should be remembered with 

 reference to the point under discussion. The first is that our 

 acquaintance with the geology of many oceanic islands is by no 

 means sufficiently complete to justify our being confident that no 

 sedimentary rocks of old date exist. The second is that the rocks 

 of an island may be entirely volcanic, and yet the island may be a 

 remnant of a continental mass. It must not be forgotten that 

 typically volcanic rocks in some continental areas, as in the Western 

 United States of North America, jS'orth-eastern Africa, and the 

 Peninsula of India, form vast horizontal or nearly horizontal sheets, 

 and completely cover the surface over areas the diameters of which 

 are measured by hundreds of miles. Such rocks may be of consider- 

 able antiquity, and they are typically continental, being all, so far as is 

 known, subaerial. It is, I think, far from clear that some oceanic 

 volcanic islands, such as St. Paul's Rocks, Fernando Xoronha, and 

 Kerguelen, are not composed of volcanic formations of the continental 

 type ; and rocks of this class are well developed in some ancient 

 continental islands, for instance, New Caledonia. At the same time, 

 this only proves that such islands were formerly of considerable 

 extent, not that they were attached to continents. 



The third fact is even more important. This, which has been 

 noticed by Prof. Bonney in one of his notes to the recent edition 

 of Darwin's ' Coral Beefs ' (p. 326), is that the occurrence of vol- 

 canic islands does not prove that the area in which they occur is 

 not a sunken continent. If Africa south of the Atlas subsided 2000 

 fathoms, what would remain above water? So far as our present 

 knowledge goes, the remaining islands would consist of four volcanic 

 peaks, Camaroons, Kenia, Kilimanjaro, and Stanley's last discovery, 



