usually placed in the apparatus as shown by the dotted lines in the sketch. It does not seem that there is any particular reason for the employment of more than one grid, provided the size of the carbide decomposed is suited to the generator, and provided the mesh of the grid is suited to the size of the carbide. A great improvement, however, is made if the grid is carried on a horizontal spindle in such a way that it can be rocked periodically in order to assist in freeing the lumps of carbide from the adhering particles of lime. As an alternative to the movable grid, or even as an adjunct thereto, an agitator scraping the conical sides of the generator may be fitted which also assists in ensuring a reasonably complete absence of undecomposed carbide from the sludge drawn off at intervals. A further point deserves attention. If constructed in the ideal manner shown in Fig. 6 removal of some of the sludge in the generator would cause the level of the liquid to descend and, by carelessness, the level might fall below the point _a_ at the base of the shoot. In these circumstances, if gas were unable to return from the holder, a pressure below that of the atmosphere would be established in the gas space of the generator and air would be drawn in through the shoot. This air might well prove a source of danger when generation was started again. Any one of three plans may be adopted to prevent the introduction of air. A free path may be left on the gas-main passing from the generator to the holder so that gas may be free to return and so to maintain the usual positive pressure in the decomposing vessel; the sludge may be withdrawn into some vessel so small in capacity that the shoot cannot accidentally become unsealed; or the waterspace of the generator may be connected with a water-tank containing a ball-valve attached to a constant service of water be that liquid runs in as quickly as sludge is removed, and the level remains always at the same height. The first plan is only a palliative and has two defects. In the first place, the omission of any non-return valve between, the generator and the next item in the train of apparatus is objectionable of itself; in the second place, should a very careless attendant withdraw too much liquid, the shoot might become unsealed and the whole contents of the holder be passed into the air of the building containing the apparatus through the open mouth of the shoot. The second plan is perfectly sound, but has the practical defect of increasing the labour of cleaning the generator. The third plan is obviously the best. It can indeed be adopted where no real constant service of water is at hand by connecting the generator to a water reservoir of relatively large size and by making the latter of comparatively large transverse area, in proportion to its depth; so that the escape of even a largo volume of water from the reservoir may not involve a large reduction in the level at which it stands there.