440 
MR. E. H. GRIFFITHS OH THE VALUE OF 
sufficient to fill the calorimeter to about the top of the tube up which the water is 
thrown, and it is possible that the conditions would, at this depth, change slightly. 
In any case, the work done by the stirrer alters but very little as the height of 
water changes, jarovided that 130 cub. centims. is regarded as the minimum quantity.* 
The examples given above show that the limit of error introduced by appl}dng this 
method of reduction to variations far exceeding any occurring during our experiments 
is less than 1 per cent., and as the heat generated by the stirrer during the majority 
of our 1892 experiments varied from about to ^ of the heat due to the current, it 
is evident that any error must be less than xo-q-o of whole supply. Bearing in 
mind the small variations which actually occurred, we may safely regard the limits of 
error as a small fraction of this tvoo- 
As will be seen later on, we performed an experiment in which the current was so 
diminished that the stirring supply was half that due to the current (see J 9), and 
another in which the current was so increased that the proportions were as 1 :20 
(see J 34). Any inaccuracy in the determination of the stirring supply would have 
caused considerable divergence in the corrected results; but the numbers obtained are 
themselves a proof of the accuracy of the method adopted. (See note, p. 406.) 
It appeared possible that changes in the viscosity of water at different temperatures 
would have an influence on the results. 
This may be the case, although our experiments do not lead to that conclusion 
(see p. 448). If so, however, it would not affect us, for the total heat lost or gained 
by the calorimeter at different temperatures (independently of the current supply) 
was directly determined for each weight of water, and therefore although our values 
of the radiation coefficient, &c., might be modified, the expression for the total loss or 
gain would remain unchanged. 
Want of time alone prevented our investigating this matter more fully, but the 
consideration that neither the value of J, nor the changes in the specific heat of 
water would be affected thereby, justified its postponement to another season. 
No use was made of the numbers obtained in Table XVIII. ; our only object being 
to establish the law tr® = K. This was fortunate, for on afterwards examining the 
water used for this series we had reason to believe that it had become slightly soiled. 
We found that the lead ring used in the calorimeter lid had been insufficiently gilded, 
and also we had not allowed the gutta-percha varnish, which we used as a “ size ” 
for the gold leaf, to harden sufficiently before introducing the water. 
The results of nearly ninety additional experiments of a similar kind indicate that 
the irregularities of the numbers in the above table are unusual, and that the actual 
values of obtained from them, are unreliable, but that the conclusion drawn 
(viz., that under the same circumstances tr^ is a constant) is correct. 
Very small changes in the adjustment caused considerable changes in K, but its 
* This minimum qxiantitj, 130 cub. ccntims., is deduced from the rcsxdts of latei’ experiments. 
