CONTINUOUS ELECTRIC CALORIMETRY. 
119 
very closely with the value 13°'3, calculated for a similar case in § 32, hut the heat- 
loss is less than the limit for a small conductor in a large tube. On the other hand, 
the observed heat-loss in line (4) is greater than the normal value ’0700 instead of 
being less, as theory requires. It is possible that there may have been some other 
source of error in this experiment due to shifting the thermometer in the outflow- 
tube or similar causes. 
When the first apparatus was set up with the vacuum-jacket in June, 1897, only 
three strands of '006 / ' wire were employed, as it was necessary that the resistance of 
the heating conductor should be of the same order as that of the manganin current- 
standard. The latter had been made of 1 ohm resistance to suit the mercury 
experiment, and there was not time to make another, as the experiment had to be 
tried before the meeting of the British Association at Toronto. In order partially to 
compensate for this defect, I took pains to exaggerate the irregularity of the wire by 
bending it into a zigzag before pulling it through the tube, but the superheating of 
the wire (4° at 14 watts) proved to be somewhat excessive. In spite of this the 
readings were extremely steady, and could be taken easily to 2 or 3 parts in 100,000. 
The accuracy of this preliminary series of experiments was seriously impaired by the 
bad fitting of the thermometers in the inflow- and outflow-tubes, but the results 
showed that the method was capable, under suitable conditions, of attaining a very 
high order of precision on account of the great steadiness of the readings, which was 
much-more perfect than anyone could have anticipated without actual trial. 
Shortly after the British Association meeting, another manganin current-standard 
of 1 ohm resistance was made, and placed in parallel with the first. The platinum 
conductor, was composed of 6 strands of •006" wire, which gave a much better 
distribution of the heat. After numerous preliminary difficulties of temperature 
regulation, of leakage, and of bad fitting and connections had been overcome, the 
results, though-often very consistent, still showed occasional discrepancies, especially 
after refitting. I was inclined to attribute these difficulties at the time either to bad 
fitting of the thermometer in the outflow-tube as previously explained, § 26, or to 
variable contact between the conductor and the walls of the glass tube, or possibly to 
bad solder joins on the manganin current-standards, one of which had actually broken 
away on one occasion. 
With the view of preventing the uncertainty of contact with the walls of the glass 
tube, I thought at one time of trying conductor in the form of a twisted strip, 
which I had employed some years previously in experiments on the viscosity of 
liquids by the Wheatstone-bridge method, for which I required a continuously 
variable resistance analogous to a slide-wire in the electrical method. 1 found that a 
round wire sliding in a tube would not do for this purpose owing to the impossibility 
of centering. A twisted strip, sliding in a tube which it closely fitted, evaded this 
difficulty, but led to further anomalies, which proved, on investigation by the colour- 
band method, to be due to the fact that the motion of the liquid became turbulent if 
