REPORT ON THE COMPOSITION OE OCEAN- WATER. 
85 
only tlie constant terms need be retained in the computation, which leads to the value — 
16-7^17 — 1 028 ‘01 , 
whence, by a series of corrections, 
4815-56= 1027‘202 by hydrometer. 
By analysis y = 20*345, whence by tables (if we deduct 0‘04 from the tabular value for the 
vacuum correction), 
4 S 15 . sg = 1027-287, 
or hydrometer result + 0‘085 ; but this latter result is liable to a correction of — 0‘042 for 
the capillarity (see p. 83), which brings it down to 1027 ‘160, whence error in corrected 
hydrometer reading = — 0 ‘ 1 2 7. 
By a specially made plunger experiment (plunger displacement specially determined, 
weighings reduced to the vacuum from actually observed air-density) I found 
4 S 19 -i 5 = 1026-43. 
The hydrometer result, when reduced to 19° "15, is 
4 S 19 . 15 = 1026-315 
(capillarity not allowed for), hence error in tmcorrected hydrometer reading = — 0*12 ; in 
corrected = — 0"16. We see the capillarity correction alters the result in the wrong direc- 
tion, and consequently had better be omitted. The number 0*00845, which in my final 
equation figures as the water-value of 1 0 of the hydrometer, is the result of a compromise 
between those two series. The water series by itself gave 0*00805, which coincides with 
what I obtained before by a more direct determination, in observations stretching over 
a wider range of the scale. By, so to speak, forcing this value (i.e., 0*00805 x 1*028) 
on the sea-water curve, slightly altering the constants, and calculating on this basis, 
I find — 
4 S 19 . 15 = 1026*349, 
which is only 0*08 less than the plunger result. Adopting this method, the hydrometer 
value is too low: if we neglect the capillarity correction by 0*08; if we take in the 
capillarity correction, by 0*12. 
There is no need of our searching for the cause of this error ; it is fully explained by 
the inherent uncertainty in the hydrometer’s position of apparent equilibrium. On com- 
paring the several readings obtained in series I. for identical over -weights and reduced to 
the same temperature, I often found them to differ from the mean by as much as 
