This conclusion was later verified by sudden subsidences of the liquid lava until the cliffs bordering the liquid were fifty feet high. The eastern grottos turned into cascades, with the liquid pouring down a well. The liquid lake had become a river pouring over a ledge of its own bottom, across from the western source wells to the eastern sinkholes. These latter were fountaining grottos when the lakes were full, but they exhibited internal rectangular upright sinkholes when the lake level was down. This was verified repeatedly, and the phenomena of source wells at the west and cascading sinkholes at the east were confirmed and photographed. It thus became evident that the lava lakes were nothing more than convectional lava flows over pasty solidified substance of their own bottom sediment. Convection means rising foam, loss of gas, and sinking gas-free heavier liquid. In other words, the bench magma capped with overflows on the marginal platforms was a paste, cooled from the top and bottom and sides and making the saucer of streaming liquid. It was this paste which constituted the swelling heart of the bench magma. The fountaining of gas bubbles escaping from solution robbed the lava of heat and caused it partially to solidify, always at a depth of about fifty feet. Thus there were necessarily three substances: The deep lava fizzing with self-heating gases (later proved to be inflammable hydrogen, carbon monoxide, sulfur, and inert nitrogen and argon), the streaming foam into which the deep lava expanded, and the semi-solidified refuse of the foam created at the bottoms and banks of the liquid lava when it cooled from bright yellow heat (about 1150° Centigrade) to a dark-red heat (about 900° Centigrade). The streaming across the bottom from west to east meant that during six months of rising lava, some 600 feet in the last half of 1916, the lava column was a cylinder of semicooled lava, maintained by upward pressure of the deep lava bubbling up in the western crack between the cylinder and Halemaumau wall. Meanwhile, at all times, the lakes were nothing more than streams of foam fifty feet deep and skinned over on top, congealing on their bottoms and shores and cascading down sinkholes in the eastern wall cracks of the cylinder. A convectional circulation was what maintained the rising, foaming, heating, and cooling and the changes in density of the liquid as it lost its gas. Thus the entire fountaining phenomenon of the lava lakes was due to the self-heating of what is known as exothermic reaction of gas escaping from solution in molten basalt. Much of this is actually the burning of hydrogen in air, creating a convectional circulation wherever the deep lava can find an outlet.