Assuming these figures to be correct, the following statement is approximately true. On oxidising an alloy of 10 grams of copper and 10 grams of lead, and pouring off the slag when 3 grams of lead have gone into it, there will be a loss of (owing to the slag carrying it off) about 0.2 gram of copper. On repeating the operation, the next 3 grams of lead will carry with them about 0.5 gram of copper; and on again repeating, 3 grams of lead will remove 0.8 gram of copper. Finally, the last gram of lead will carry with it 0.3 gram of copper, and there will be left a button of copper weighing 8.3 grams. The slag will have carried off altogether 1.7 gram of copper, which is 17 per cent. of the metal originally present. With the more perfect exposure to the air, and quicker removal of the slag, which results from heating on a cupel, the loss would be heavier. Karsten got by actual experiment on cupelling copper and lead in equal proportions, a loss of 21.25 per cent. Going back to the example: if the slag were collected and fused with a suitable reducing agent so as to convert, say, half of it into metal, that half would contain nearly the whole of the copper (such a reduction is called "cleaning the slag"). On reoxidising this metal, another button of copper is formed which, added to the first, would reduce the loss from 17 per cent. to, say, 7 or 8 per cent. And it is conceivable that by a series of similar operations, almost the whole of the 10 grams of copper originally taken might be recovered. In practice the problem is (as far as the copper is concerned) not how to save, but how most easily to remove it; and since the removal of this metal is quicker from an alloy containing not too much lead, it is evident that two or three operations with small quantities of lead will be more effectual than a single treatment with a larger quantity. With those metals (tin, antimony, &c.) which pass quickly into the slag, the contrary is true; hence with these it is necessary to have enough lead present, so that the slag formed at the outset shall contain enough oxide of lead to make it fluid. As silver is so much less easily oxidised than copper, we should reasonably expect that the proportion of silver carried off in the oxide of lead would be considerably less than that of the copper indicated in the above example. Indeed, there are one or two facts which tend to encourage the hope that the operation may be conducted without any loss. If a piece of pure silver foil is exposed on a cupel to air at the usual temperature of cupellation, it undergoes very little change; it does not even fuse; it loses nothing in weight, and does not oxidise. In fact, even if oxide of silver were formed under these conditions, it could not continue to exist, for it is decomposed into silver and oxygen at a temperature considerably below redness. On the other hand, oxide of silver is not reduced to metal by heat alone, when