sub-series. It is plain that the source of the hydrogen ought to be varied. This was done by passing the hydrogen over palladium, and getting the hydrogen from the hydride formed by heating it.
Finally silver oxide was substituted for mercuric oxide, as it decomposes at a more convenient temperature.
The fracture of the measuring vessel rendered a new apparatus necessary, and one or two improvements were now introduced, the chief being the use of a constant volume, so that pressures only required to be measured for the large volumes, and the use of a narrower tube for measuring the residues.
One interesting point about the fracture of the measuring vessel is that, while surrounded by water and full of mercury, it broke in exactly the same way that a spare one which was kept ready for fear of accidents broke at the same time. The one in use was found fractured on a Wednesday morning, and the spare one had been handled and put on a shelf on the Monday afternoon just preceding. On going on the Wednesday to take it down to replace the other which had broken, it was found to have broken in such a similar manner that the pieces had to be fitted to each other to find out to which apparatus they belonged. The spare one had never been used, and not even calibrated.
In Series II. the oxygen was always got from oxide of silver, and the hydrogen from sodium and water or from palladium hydride, the hydrogen for which was furnished by sodium and water, with the exception of the first four experiments in Series IIa., when the hydrogen charge in the palladium was what was left in it from Series If., the hydrogen for which was obtained by the electrolysis of dilute hydrochloric acid. The 93 grms. of palladium absorbed usually somewhat over 6 litres of hydrogen, enabling twelve experiments to be done consecutively from one charge.
In Series IIa. the only two experiments which perhaps ought to be rejected are Nos. XII. and XIII., which were the first two performed after the apparatus had stood unused for over a year. Why these results should be so low I can offer no suggestion, but they differ notably from those of the same charge both before and after them.
Series IIb.
The results in this series vary remarkably, two of the ratios being very high and the mean of the whole being high, but with a large probable error. The oxygen from the mixture of silver oxide and barium sulphate seemed to be very pure.
Series IIc.
The hydrogen from the palladium hydride at first behaved in a peculiar way, and some air seemed somehow to have got in with the oxygen, and in two of the experi-
