There is, however, a peculiarity about the development of the vesicular structure in this sill which I have not observed anywhere else. If we examine the southern side of the crag near its eastern end we observe that the successive bands of vesicles are arranged in the same direction as the surface of contact with the underlying tuffs, precisely as they are ranged in dykes parallel to the bounding walls. So far the structure is quite normal. But, moving a few yards westwards, we find that the bands begin to curve, and, instead of following the contact surface, strike it first obliquely and then at right angles, until we have the structure shown in Fig. 191. The bands here vary from less than an inch to more than a foot in breadth, and where broadest assume a slaggy texture. I sought in vain for any evidence of subsequent disturbance such as might have truncated these parallel rows of vesicles and pushed the rock bodily over the tuffs. The perfect parallelism of the bands with the surface of the tuff at the east end, and the absence of all trace of a thrust-plane at the base of the sill, seem to show that, though the rows of vesicles were undoubtedly at first arranged parallel to the surfaces between which the intrusion took place, the mass, before completely consolidating and coming to rest, was ruptured, and a portion of it was driven onwards at right angles to its previous line of movement. A consideration of the singularly slag-like structure of the injected masses in the tuffs and agglomerates leads to the conclusion that though what we now see of these rocks did not actually flow out at the sea-bottom in streams of lava, it was intruded so close to the surface that the imprisoned vapours had opportunity to expand, as in superficial outflows.[59] This inference is in accord with that derived from an examination of the necks, wherein we find evidence of the probable survival of parts of the actual craters and volcanic cones. [Footnote 59: As illustrative of the occurrence of the vesicular structure in superficial intrusions, I may again cite the dyke which cuts the ash of the outer crater-wall of the Puy de Pariou in Auvergne. The andesite of this dyke is in places as vesicular as the lava-stream with which it was doubtless connected, but the vesicles have been flattened and drawn out parallel to the walls of the dyke. In this instance it is quite certain that there could never have been any great depth of detrital material above the fissure into which the material of the dyke was injected (see vol. i. p. 66).]