THERMAL CONDUCTIVITY ON WATER. 
545 
by the sides of the vessel and by the convection currents thereby established, the fol¬ 
lowing arrangement was made. Within the cylindric tank in which the testing was 
carried on, a circular screen of thin cotton material was hung, which was everywhere 
6^- centims. from the outer walls, and which extended from top to bottom of the 
tank. Any convective action due to conduction of heat downwards by the walls of 
the vessel would go on outside the screen : and the enclosed column, 77 centims. in 
diameter, would be but little affected by these slight motions. The thermometers 
projected into the middle of the enclosed column through small holes cut in the 
screen. 
Experiments made in March and April, 1880, with aud without the screen, gave 
the following numbers for the apparent thermal conductivity of the water (in terms of 
an undetermined 
unit):— 
without screen 
•00137, 
•00135, 
•00130 ; 
with screen 
•00154, 
•00149, 
•00158. 
These numbers, though not in complete agreement, are yet quite of the same order. 
They seem to indicate that the screen does produce some effect, and I propose to carry 
out a much more complete series of experiments in order to determine exactly to what 
this is due. One of the greatest difficulties of the investigation is the determination 
of what is to be considered as the effective distance between the levels represented by 
the thermometers A and B in the diagram given in the foregoing paper; and it 
prevented me from reducing the results of these last experiments satisfactorily to 
absolute measure, though it does not at all affect their relative values. Any error in 
estimating this distance produces a proportionate error in the absolute value of the 
numerical result. I am therefore making preparations to eliminate as far as I can 
this source of error. I believe however that the numbers given in Table II. may be 
taken as approximations to the absolute conductivity of water, which has not been 
hitherto even roughly determined. 
