150
T. G. BONNEY ON COLTJMNAE, FISSILF.
roidal structure, though on a minute scale ; and here examination of the rocks, both macroscopic and microscopic, proves that the structure is in no way connected with previously formed cubes, or capable of being explained by simple exfoliation during decompo- sition. But the following instances of its occurrence will, I think, show that no theory of mere decomposition is adequate to explain the facts.
The annexed diagram (fig. 11) is taken from a large spheroid ex-
Fig. 10. — Spheroidal structure in Agglomerate (Burntisland).
Fig. 11. — Sjpheroid in Basalt near Murat.
hibited in a mass of basalt by the roadside on the way from Murat to the bridge over the Alagnon (Auvergue). A glance at it will, I think, show that the curious prolongation of the outer layer is impossible to explain on any principle of weathering. I may add that this occurrence of an isolated spheroid or small group of sphe- roids in a mass which generally does not show the structure is by no means rare.
But I can produce yet stronger cases. A few kilometres from Le Puy on the lloute de Brioude, close to where a road turns off to Polignac, is a mass of columnar basalt rather decomposed, part of which exhibits veiy well the spheroidal structure. Here spheroids may be seen, one above another, enclosed three or four at a time in a columnar shell without any dividing cross joints, so that they are just like Dutch cheeses packed in hexagonal cases (the interstices being filled up). The lid of the box has more or less fallen away, and exposed the contained spheroids (See fig. 12).
Again, a mass of trachyte, intrusive in trachytic ash, by the road- side on the flank of the Puy Gros (near the Roche Sanadoire men- tioned above) showed structures bearing on this as well as other points of interest. The surface in contact with underlying tuff exhibited a. platy structure for about three inches; then came a layer rudely spheroidal ; above that were some irregularly vertical joints ; two of these curved in opposite directions so as to enclose a
