5G4 
DR. A. SCOTT ON THE COMPOSITION OE WATER BY VOLUME. 
merits the residues were very large, so large that the results were absolutely valueless, 
the Don-absorbable residue amounting to many cubic centimetres in Experiments IX. 
and XL It is therefore advisable to neglect also Experiments YIII., X., and XII., 
the results of which are somewhat abnormal. No reason can be given for the 
abnormal behaviour of all five experiments, but being all consecutive there is no 
doubt that some common source of impurity infected all five. 
Series IId. 
The first experiment was made with the last lot of hydrogen from a charge of sodium ; 
fresh sodium was added, and the second experiment was made with the hydrogen at 
first given ofii Both these results are high. The next three agree wonderfully with 
one another and are, as it were, made with the purest hydrogen obtainable from 
sodium, any impurity whicb wouhi come off first being used in Experiment XXL, 
and any whicb tended to remain being left behind, and was rejected as not quite 
enough sodium remained to give hydrogen for another experiment. The gases were 
in the last three experiments also perfectly pure as far as residual gas was concerned. 
The silver oxide in this series and the next was the same sample and prepared from 
silver sulpiiate and potassium hydrate. 
Series He. 
This series as a whole is the best which has been done, the gases being throughout 
pure, with the exception of very small impurity in the first experiment. They agree 
remarkably with one another, whether the hydrogen or the oxygen was in excess. 
The whole twelve experiments were successful without exception, and were all that 
could be performed with one charging of the palladium. 
The values deduced from the various series are given below and together so that 
they may readily be compared. The equations employed in their reduction are the 
following well known ones :— 
1. E(n’ the probable error of the arithmetical mean 
ju’obable error = 
dz •G745 
Avhere n — nmnber of observations, 
6’ = sum of squares of the variations of the individual results from the mean. 
