A lava-flow is sometimes divided into large rudely spheroidal masses, which fit into one another, and which show signs of more rapid cooling on their surfaces. These were particularly observed on the mountains near Mont Genèvre by Cole and Gregory[64], who compared the forms to "pillows or soft cushions pressed upon and against one another." It was suggested that these forms were produced by the seething of viscid lavas, masses being heaved up and falling over, and the outer layers having time to cool in a glassy state before they were deformed by contact with others. This _pillow-structure_ has been widely recognised, and J. J. H. Teall has remarked how often "pillow-lavas" are associated with radiolarian cherts. He regarded them, therefore, as of submarine origin. Sir A. Geikie[65], moreover, stated that the spheroidal sack-like structure was produced by the flow of such lavas into water or watery silt. This acute suggestion has now been verified by Tempest Anderson[66], who has observed in Samoa the chilling of the lobes of lava, as they are thrust off into the sea and washed over by the waves. H. Dewey and J. S. Flett[67] have pointed out that pillow-structure commonly occurs in lavas in which there has been a conversion of lime soda felspars into albite, a change frequent in a series of rocks which they call the "spilitic suite." The importation of soda is attributed to vapours entering soon after the consolidation of the rock, and it is urged that any excess of sodium silicate must have escaped into the sea-water in which the pillow-lavas were produced. Hence radiolaria will flourish in the neighbourhood (presuming that a decomposition of the silicate can be brought about), and their remains will in time form flint in the hollows of the lavas. The paper quoted contains numerous references to previous work, and is a suggestive example of how petrographic study may go hand in hand with the appreciation of rocks from a natural history point of view. It is only characteristic of the subject of petrology that G. Steinmann[68] has with equal ingenuity explained the relations between radiolaria and spilitic lavas by reminding us that gravity-determinations show an excess of basic material under the oceans and of lighter material, rich in silica, under continental land. Hence, when deep-sea deposits are crumpled by earth-movements, basic types of rock, graduating even into serpentine, become associated with radiolarian chert, partly as extruded lavas, but usually as intrusive sheets injected at the epoch of mountain-building.