the reef within consisting, as before noted, of loose masses of coral bedded in with shells and sand. A portion of this crust is photographed on Plate XXXIII; the upper surface (Fig. 72) with its _sections_ of contained shells has already been referred to. It is nearly smooth and very hard. The under side of the same fragment is shewn in the next figure and is seen to consist of an irregular mass of shells and coral branches lightly cemented to the crust, from between which the sand, which has not been consolidated, has fallen away. The formation of beach sandstone is practically the same process of cementation, by alternate solution and deposition of lime, taking place in a mass of shell and coral sand instead of larger fragments, the rock following exactly the curve of the sandbank, of which it is obviously a part which has been consolidated _in situ_. _Plate XXXV_ [Illustration] [Illustration: Figs. 76 and 77. Coral Cliffs, Zanzibar] Fig. 76. Chuaka Bay. Note undermining of fallen fragments „ 77. Bawi Island. Rock masses supported by narrow stalks Coral reefs are classified into three sets according to their relation with other land[55]. I. _Fringing reefs_, which, as the name implies, border the land, are continuous with it, and the seaward edge of which can be reached by wading. II. _Barrier reefs_, which run parallel to the coast but separated from it by deep water navigable for coasting vessels larger than canoes. III. _Atolls_, ring- or crescent-shaped reefs having no obvious relation to any land and typically found far out in the ocean, from the great depths of which they rise with steep slopes to, at most, a few feet above high tide level. Fringing reefs we have already dealt with; the two agents described—growth and abrasion of coral—will account for all of them. Barriers and atolls are more puzzling. Why should the barrier form its line parallel to the coast, though at a distance from it, and the very existence of atolls is one of the most striking phenomena of Nature. The problem is complicated by the fact that ordinary reef corals die out at a depth of 50 fathoms or so. Now 50 fathoms is a mere nothing compared to the depths from which the Pacific atolls rise, and is only a quarter the depth often found within a few hundred yards of the Red Sea reefs. How then to account for the building of reefs in deep water?