When, however, it falls upon a grid placed some distance above the bottom of the water vessel, the various convection currents set up as parts of the liquid become warm, and the mechanical agitations produced by the upward current of gas rinse the spent lime from the carbide, and entirely prevent overheating, unless the lumps are excessively large in size. If the carbide charged into a hand-fed generator is in very large lumps there is always a possibility that overheating may occur in the centre of the masses, due to the baking of the exterior, even if the generator is fitted with a reaction grid. Manifestly, when carbide in lumps of reasonable size is dropped into excess of water which is not merely a thick viscid cream of lime, the temperature cannot possibly exceed the boiling-point--_i.e._, 100° C.--provided always the natural convection currents of the water are properly made use of. The defect which is, or rather which may be, characteristic of a hand-fed carbide-to-water generator is a deficiency of gas yield due to solubility. At atmospheric temperatures and pressure 10 volumes of water dissolve 11 volumes of acetylene, and were the whole of the water in a large generator run to waste often, a sensible loss of gas would ensue. If the carbide falls nearly to the bottom of the water column, the rising gas is forced to bubble through practically the whole of the liquid, so that every opportunity is given it to dissolve in the manner indicated till the liquid is completely saturated. The loss, however, is not nearly so serious as is sometimes alleged, because (1) the water becomes heated and so loses much of its solvent power; and (2) the generator is worked intermittently, with sufficiently long intervals to allow the spent lime to settle into a thick cream, and only that thick cream is run off, which represents but a small proportion of the total water present. Moreover, a hand-fed carbide-to-water generator will work satisfactorily with only half a gallon [Footnote: The United States National Board of Fire Underwriters stipulates for the presence of 1 (American) gallon of water for every 1 lb. of carbide before such an apparatus is "permitted." This quantity of liquid might retain nearly 4 per cent. of the total acetylene evolved. Even this is an exaggeration; for neither her, nor in the corresponding figure given in the text, is any allowance made for the diminution in solvent power of the water as it becomes heated by the reaction.] of liquid present for every 1 lb. of carbide decomposed, and were all this water run off and a fresh quantity admitted before each fresh introduction of carbide, the loss of acetylene by dissolution could not exceed 2 per cent. of the total make, assuming the carbide to be capable of yielding 5 cubic feet of gas per lb. Admitting, however, that some loss of gas does occur in this manner, the defect is partly, if not wholly, neutralised by the concomitant advantages of the system: (1)