1902 - 3 .] Mr Manley on Electrical Conductivity of Sea- Water. 359 
on the whole an absence of that uniformity which obtains in 
column A, and this even after the application of the troublesome 
corrections due for the changes both in the temperature and 
pressure of the air. The irregularity is probably caused by 
variations in the amount of moisture condensed upon the exterior 
surface of the pyknometer. When a proper counterpoise is used, 
such variations are of no consequence, for both the pyknometer 
and counterpoise are equally affected. Attention may also be 
drawn to the fact that, whilst the extreme difference in the values 
given under A amounts to *0009 grm., that for those under B is 
equal to *0022 grm. In the former we have a definite suggestion 
of loss through evaporation, but in the latter such probable loss is 
not even hinted at, except in the most uncertain manner. 
Judging then from these results, it may, I think, be safely 
asserted that when it is desired to obtain a value for the relative 
density of a water which shall be as nearly as possible correct to 
the fifth decimal place, the use of a counterpoise for automatically 
eliminating certain incidental errors is absolutely essential. It is 
important to note that the smaller the differences in the ‘total 
salinities’ of any waters under examination, the more necessary 
does it become to introduce the use of a suitable counterpoise. 
Instead of a specially constructed counterpoise, use may be 
made of a second and exactly similar pyknometer, which for 
convenience should be from 0T to 0*2 grm. lighter than the 
pyknometer proper. After the weight has been suitably adjusted, 
the two open ends should of course be sealed up before the blow- 
pipe flame. Such a counterpoise possesses the further advantage 
of having in all probability been made from the same kind of 
glass as that which constitutes the pyknometer itself ; hence the 
conditions for the condensation of moisture upon both will be 
equally favourable. Several such counterpoises were actually 
tried, and the results obtained with them fully justified this con- 
clusion. 
Experimental Results. 
The results obtained, according to the methods which have been 
described at some length, are here grouped together in Table II. 
Under p 2 4 and — are given the respective values in true ohms 
r24 
