and Indian Oceans. The islets are first formed some way back either on the
projecting points of the reef, especially if its form be angular, or on the
sides of the main entrances into the lagoon--that is in both cases, on
points where the breakers can act during gales of wind in somewhat
different directions, so that the matter thrown up from one side may
accumulate against that before thrown up from another. In Lutke's chart of
the Caroline atolls, we see many instances of the former case; and the
occurrence of islets, as if placed for beacons, on the points where there
is a gateway or breach through the reef, has been noticed by several
authors. There are some atoll-formed reefs, rising to the surface of the
sea and partly dry at low water, on which from some cause islets have never
been formed; and there are others on which they have been formed, but have
subsequently been worn away. In atolls of small dimensions the islets
frequently become united into a single horse-shoe or ring-formed strip; but
Diego Garcia, although an atoll of considerable size, being thirteen miles
and a half in length, has its lagoon entirely surrounded, except at the
northern end, by a belt of land, on an average a third of a mile in width.
To show how small the total area of the annular reef and the land is in
islands of this class, I may quote a remark from the voyage of Lutke,
namely, that if the forty-three rings, or atolls, in the Caroline
Archipelago, were put one within another, and over a steeple in the centre
of St. Petersburg, the whole world would not cover that city and its
suburbs.

The form of the bottom off Keeling atoll, which gradually slopes to about
twenty fathoms at the distance of between one and two hundred yards from
the edge of the reef, and then plunges at an angle of 45 deg into
unfathomable depths, is exactly the same (The form of the bottom round the
Marshall atolls in the Northern Pacific is probably similar: Kotzebue
("First Voyage," volume ii., page 16) says: "We had at a small distance