Though generally rather compact, becoming indeed dense, almost vitreous rocks in some sheets, the columnar basalts are often more or less cellular throughout, and highly slaggy along their upper and under surfaces. In some cases, as in that of a prismatic sheet which overlies the rough scoriaceous lava of Camas Tharbernish, in the island of Canna, the rows of vesicles are disposed in lines parallel to the under surface of the sheet (Fig. 259.) As already remarked with regard to the massive, rudely-jointed sheets, many of the most perfectly columnar rocks of the plateaux are not superficial lavas, but intrusive sills, bosses or dykes. Conspicuous examples of such sills are displayed on the coast of Trotternish in Skye, and of the bosses and dykes at the eastern end of Canna. To these further reference will be made in the sequel. It is not always possible to be certain that columnar sheets which appear to be regularly intercalated among the undoubted lavas of the volcanic series may not be really intrusive. In some instances, indeed, we can demonstrate that they are so, when after continuing perfectly parallel with the lavas above and below them, they eventually break across them. One of the most remarkable examples of this feature is supplied by the great sill of the south-west of Stromö, in the Faroe Islands, of which I shall give some account in Chapter xlii. (Figs. 312, 328, 329). 3rd. Slaggy or amygdaloidal lavas without any regular jointed structure, but often with roughly scoriform upper and under layers, and tending to decay into brown earthy debris. Some of the upper surfaces of such sheets among the Tertiary basalt-plateaux must have resembled the so-called "Aa" of the Sandwich Islands. A striking example of the structure may be noticed at Camas Tharbernish, on the north coast of the Island of Canna. There the hummocks on the upper surface of a slaggy basalt measure about 15 feet in breadth, and rise about three feet above the hollows between them, like a succession of waves (see Fig. 259). The steam-holes are disposed in a general direction parallel to the strike of the hummocks. Great variety obtains in the size and shape of the vesicles. Huge cavities a foot or more in diameter may occasionally be found, and from such extremes every gradation may be traced down to minute pore-like vacuoles that can hardly be made out even with a strong lens. In regard to the deformation of the vesicles, it is a familiar general rule that they have been drawn out in the direction of the flow of the original lava. Occasionally they have become straight, narrow, sometimes bifurcating pipes, several inches long, and only an eighth of an inch or so in diameter.[222] A number of such pipes, parallel to each other, resembles a row of worm-burrows (see Fig. 2).