The decomposition which with perfect chemical accuracy has been stated to occur quantitatively between 36 parts by weight, of water and 64 parts of calcium carbide scarcely ever takes place in so simple a fashion in an actual generator. Owing to the heat developed when carbide is in excess, about half the water is converted into vapour; and so the reaction proceeds in two stages: half the water added reacting with the carbide as a liquid, the other half, in a state of vapour, afterwards reacting similarly, [Footnote: This secondary reaction is manifestly only another variety of the phenomenon known as "after-generation" (cf. _ante_). After-generation is possible between calcium carbide and mechanically damp slaked lime, between carbide and damp gas, or between carbide and calcium hydroxide, as opportunity shall serve. In all cases the carbide must be in excess.] or hardly reacting at all, as the case may be. Suppose a vessel, A B, somewhat cylindrical in shape, is charged with carbide, and that water is admitted at the end called A. Suppose now (1) that the exit for gas is at the opposite end, B. As the lumps near A are attacked by half the liquid introduced, while the other half is changed into steam, a current, of acetylene and water vapour travels over the charge lying between the decomposing spot and the end B. During its passage the second half of the water, as vapour, reacts with the excess of carbide, the first make of acetylene being dried, and more gas being produced. Thus a second quantity of heat is developed, equal by theory to that previously evolved; but a second elevation in temperature, far more serious, and far less under control, than the former also occurs; and this is easily sufficient to determine some of those undesirable effects already described. Digressing for a moment, it may be admitted that the desiccation of the acetylene produced in this manner is beneficial, even necessary; but the advantages of drying the gas at this period of its treatment are outweighed by the concomitant disadvantages and by the later inevitable remoistening thereof. Suppose now (2) that both the water inlet and the gas exit of the carbide cylinder are at the same end, A. Again half the added water, as liquid, reacts with the carbide it first encounters, but the hot stream of damp gas is not permitted to travel over the rest of the lumps extending towards B: it is forced to return upon its steps, leaving B practically untouched. The gas accordingly escapes from the cylinder at A still loaded with water vapour, and for a given weight of water introduced much less acetylene is evolved than in the former case. The gas, too, needs drying somewhere else in the plant; but these defects are preferable to the apparent superiority of the first process because overheating is, or can be, more thoroughly guarded against.