40 
Proceedings of the Royal Society 
this the density of three samples of water was determined twice 
with the hydrometer stem dry, and twice immediately after the 
the stem had been immersed. The results are shown in Table IX, 
Temperature unaltered, about 7°. 
Table II. 
Weights.; 
Water Sample Ho. 263, 
Hydrometer stem. 
Ho. 267, 
Hydrometer stem. 
Ho. 268, 
Hydrometer stem. 
Dry. 
Wet. Dif. 
Dry. 
Wet. 
Dif. 
Dry. 
Wet. 
Dif. 
1 
86+ 
84'0 2'5 
95-5 
94-0 
1'5 
99+ 
96 '5 
2-5 
1+7 
49' 0 
48'0 1-0 
62+ 
60'5 
1-5 
1+6 
... 
37'5 
37+ 
0-5 
Here the adhering film of water produced a sinking of the hydro- 
meter of from 0*5 to 2*5 millimetres, according to the length of ex- 
posed stem. This is an error which is practically negligible when 
the reading is less than 30, but rises to a serious amount when it is 
greater than 80. 
Taking 2*0 mm. as the amount of sinking produced by the film 
of water when the stem reading is over 80, and since the sinking 
produced by weight No. 7 (0*3312 grammes) is on an average 36*5 
0"3312 x 2 
(see page 38) the weight of the water is — ' i. 0 . f = O/QI 8 I 4 . 
o\)*D 
In water sample No. 250(1) the reading was 80 and the calculated 
density 1 '02734. Supposing the scale to have been wet, and adding 
0*01814 grms. to the weight of the hydrometer, the density comes 
out as 1*02746, an increase of 0*00012. 
Here there is a sufficient explanation of the discrepancy increas- 
ing with the value of the reading ; and since in each case the mean 
of the determination when the stem was almost all out of the water, 
and that when the stem was almost entirely immersed, and when 
consequently the error from the attached film was reduced to almost 
nothing, the uncertainty of a determination (from this cause) is re- 
duced by a half, and comes to be 0*00006, exactly 0*00002 more 
than that brought out by the discussion of the 204 discrepancies 
without consideration of their cause. The reason of this, assuming 
the above explanation to be correct, and considering that 0 00006 is 
