BETWEEN THE FREEZING AND BOILING-POINTS. 
167 
values -f- 16'60 and — 19 '45 respectively, which agree very closely with the values 
given by the old cells, but are smaller than the values given by the inverted cells. 
X 2 in the above table is seen to be lower than the mean of the new crystal cells 
by — '17. Under date March 14, in Table I., it is seen to be + T4 mv, above the 
mean of the old crystal. This would make the mean old crystal lower than mean 
new crystal by '31 mv. This excessively low value is influenced by ceils X 5 and X 2 
on the mean, which have apparently gone down since the earlier tests. If we reject 
these two cells from the mean, as being too low, the mean value of the four remaining 
cells is increased by TO mv. and the difference between mean old crystal and new 
crystal reduced to '21 mv. In Table II. it was seen that the inverted cells in 
January, 1898, were '17 mv. higher than the mean old crystal. If we reject 
cell X- from the mean of the old crystal cells, as being; too low, then the difference 
between the two sets of cells is reduced to T2 mv. Also in Table III. the mean of 
the old cells is seen to be '23 mv. lower than the mean of the new and old sealed 
cells. If we reject X 5 as before from the mean,, the difference is reduced to T 7 mv. 
If now we can assume that the mean inverted and mean sealed was the same as 
the mean new crystal (which unfortunately could not be verified by a direct 
comparison), then we see that the old cells have lowered in value since January, 
1898, by '06 mv., or 4 parts in 100,000. Another indication that the cells have all 
lowered somewhat in value is afforded by a comparison as early as 1896 with six 
sealed cells, including S L and S 5 of Table III. The mean value of the sealed cells 
was '08 mv. higher than the mean old crystal including cell X 5 , which was more 
nearly in agreement with the mean at that time. (See Table VII., page 151, loc. cit.) 
It is evident that for some reason the old crystal cells, even from the first, are lower 
than what may be taken as the true Clark-cell value, if we may assume that the 
mean old sealed, mean new sealed, mean inverted, and mean new crystal are all 
within a few hundredths of a millivolt of each other, and of the true Clark-cell 
value. 
In 1896 the old crystal cells were lower than the sealed cells by '08 mv. ; in 1898 
lower than the inverted cells by '12 mv., than the new sealed cells by T7 ; and in 
1900 lower than the new crystal cells by '21 mv. This indicates that the mean 
value of the old crystal cells is T4 mv. lower than the most probable value that we 
can assume, combining all our Clark cells, and this within the limits of error of 
perhaps '02 mv. 
We can now, from the table of comparisons, assign individual values for the two 
Clark cells which were used throughout in the present investigation. These cells 
were X 2 and X 10 of the old crystal cells. From January to December, 1898, X 2 was 
'08 mv. higher than the mean of the crystal cells, and.X 10 was at the same time 
'15 mv. higher. If we neglect cell X 5 from the mean, as being too low, then this 
gives for cells X 2 and X L0 in series the values —|- '03, + T0, or + T3 mv. above the 
mean. But as mean crystal, neglecting X 6 , is T4 mv. lower than what we have 
