The basalt-plateau of Canna resembles in all essential particulars those of the other Western Isles. Its base is everywhere concealed under the sea, but from the fragments of Torridon Sandstone in its agglomerates we may infer that it probably rests on that formation, like the volcanic outliers in Rum. It is formed of successive sheets of different basalts including the usual banded, amygdaloidal and columnar forms. Some of them towards the west are specially marked by the great abundance and large size of their porphyritic felspars. The magnetic properties of the basalts at the east end of the island have long been known, and have given rise to various modern myths regarding their influence on the compasses of passing vessels. [Illustration: Fig. 268.--Section of the cliffs below Compass Hill, Isle of Canna.] But it is in its conglomerates, tuffs and agglomerates and the light they cast on some aspects of the volcanic period, elsewhere hardly recorded, that the geology of Canna possesses a special importance. To these, therefore, we may at once turn. The conglomerates are best developed at the eastern end of the island, where the cliffs present the structure represented in Fig. 268. At the base, and passing under the level of the sea, lies the agglomerate (_a_) of a vent which will be described in Chapter xli., together with other eruptive orifices of the various plateaux (p. 288). This rock has a somewhat uneven upper surface which rises in places about 150 feet above high tide-mark. Here and there it shades off upward into the conglomerate that overlies it; water-worn pebbles appear among its contents, and rude traces of bedding begin to show themselves, until, within the course of a few feet, we pass upward into an undoubted conglomerate. Elsewhere, however, and particularly along the precipices west of Compass Hill, the two deposits are more distinctly marked off from each other. The agglomerate has there a hummocky, irregular upper surface, as if it had been thrown down in heaps. The hollows between these protuberances have been filled up with conglomerate and sandstone, forming the base of the thick overlying deposit. It is thus clear that the loose materials of the vent were directly exposed at the surface when the conglomerate was accumulated, and, indeed, that these materials served to supply some of the detritus of which the conglomerate consists. The absence of any trace of a cone and crater at the vent may perhaps be explicable on the supposition that their incoherent material was washed down by the currents that swept along and deposited the conglomerate.