PROFESSOR HUGH I. CALLENI>AR OX 
120 
the pitch of the twist was too steep. This change in the character of the flow was a 
serious defect in the viscosity experiment, but was exactly what was required for the 
calorimeter. As the calorimeters were already made, and I could not choose the 
flow-tube to fit the strip, I procured a special pair of rolls about the end of 1897 for 
rolling a wire to fit the flow-tube as closely as possible. But after carefully rolling a 
wire to fit one of the calorimeters at one end, 1 found that it would not go through 
the tube, on account of want of uniformity of diameter, and in particular on account 
of a join near one end which unduly constricted the bore. I accordingly abandoned 
the attempt at the time, as T did not feel at all certain of the necessity of making 
any change in the form of the conductor. But the twisted strip was subsequently 
employed by Dr. Barnes at my suggestion with great success in two other calori¬ 
meters, the tubes of which were probably more uniform. 
The method independently devised by Dr. Barnes for eliminating the effect of the 
stream-lines, and at the same time preventing all contact between the conductor and 
the glass, was to wind a rubber cord round the conductor so as nearly to fit the tube, 
after a similar manner to that employed in fitting the copper sleeve of the outflow 
thermometer. This is undoubtedly a very simple and effective method, and unlikely 
to get out of order. It is quite possible that some of the discrepancies in the earlier 
experiments may have arisen from the accidental disarrangement of the stranded 
conductor, which might be pulled tight along the side or middle of the tube in 
refitting the apparatus, either of which contingencies would lead to serious errors. 
There were many other sources of error and difficulties in the earlier experiments 
w hich might have accounted for the effects observed, but Dr. Barnes’ opinion on this 
point is entitled to the greatest weight, as lie was personally responsible for the 
greater part of the fitting-up of the apparatus. With proper care I have no doubt 
that it would be possible to obtain as accurate results with the stranded conductor as 
by any other method, but I should be inclined to prefer the rubber-spiral method as 
being safer and more certain. 
A possible objection to the rubber-spiral is that, since the wire is held central and 
considerably superheated, the temperature of the surface of the glass must necessarily 
be less than the mean of the flow, in spite of the mixing of the stream-lines. This 
would not matter if the difference of temperature were independent of the flow. Since 
it is impossible to calculate what the effect would be in the case of turbulent flow, the 
question can be answered only by trial. The experiments with the twisted strip, 
which was partly in contact with the glass and gave a higher heat-loss, were under¬ 
taken, at my suggestion, with the object of testing this important point. The 
extremely close agreement of the results with those obtained with the rubber-spiral 
at 80° C. are sufficient proof that the error, if any, must be extremely small. It is 
unlikely that it could amount to more than two or three parts in 10,000 at 30°, 
and there is no reason to think that this type of error should increase largely at 
higher temperatures, where the viscosity of water is much smaller and the conductivity 
probably much larger. 
