140 
PROFESSOR HUGH L. CALLEXDAR OX 
It might, however, be claimed that either or both of the absolute determinations 
above cited were more accurate than this, and that the error lay with our comparison. 
It is a nice question whether the mean specific heat, which is practically indejiendent 
of the scale of the thermometer employed, should be chosen as the standard of 
reference, rather than the specific heat at a particular temperature such as 20° C. on 
the scale of a particular thermometer. My own opinion is that the latter unit is the 
more practical, and the most accurately reproducible. It is not very difficult in a 
thermochemical experiment to measure a quantity of heat to 1 in 1000 with a 
suitable rise of temperature and a good thermometer with known scale errors. But 
it is quite impossible to realize the mean thermal unit to this order of accuracy without 
the most elaborate apparatus on a scale approaching that adopted by Reynolds and 
Moorby. 
The fundamental difficulty in the determination of the mean specific heat between 
0° and 100° C. lies in the great range of temperature to be covered, and the 
consequent risk of excessive or uncertain loss of heat. Reynolds and Moorby 
endeavoured to meet this objection by working on a very large scale, and succeeded 
in reducing the heat-loss to 5 per cent., even without the use of lagging. But in 
working on this scale they encountered peculiar difficulties, which were not overcome 
without great pains and ingenuity, and which must in any case have materially 
affected the order of accuracy attainable by their method. Owing to the employment 
of a steam-engine as motor, it was difficult to secure a high degree of steadiness in 
the conditions of running, and the outflow-thermometer could not be read more 
closely than a tenth of a degree owing to its incessant oscillations. The largest 
variation recorded in the two trials of which full details are given, was 4’9° F. in 
two minutes on the outflow temperature, and four or five revolutions per minute in 
300 on the speed. The greater part of these variations being accidental would 
disappear in the mean, but it is probable that there would be systematic errors of the 
same order as the limit of accuracy of the instantaneous readings. In using so large 
a quantity of water, it was impracticable to deprive it entirely of air, which caused 
considerable trouble, owing to loss of heat by generation of steam in the air bubbles. 
Again it was naturally impossible to enclose the brake in a vacuum-jacket, or shield 
it from draughts and extraneous disturbances by means of an isothermal enclosure. 
The greatest uncertainty appears to have arisen from damp in the lagging, which 
caused the rejection of a number of trials.. The correction for conduction along the 
4-inch shaft was of much smaller magnitude, though probably more uncertain on 
account of the impossibility of determining the actual gradient in the shaft itself. 
The extreme limits of variation of the recorded results were from 776'63 to 779’46 
foot-pounds, but considering that the main sources of error above mentioned were 
accidental, it is probable that the accuracy of the mean would be of a higher order 
than might be inferred from the separate experiments. 
