628 REPORT—1899, 
value of the specific heat could not be determined by his method with the same 
degree of precision at the extremities of the range as in the middle, and all the 
probable errors of the method would be greatly increased as the temperature of 
the calorimeter was raised so far above its surroundings. In particular, the 
corrections and changes of zero of the mercury thermometers, and the rate of 
external loss of heat, would be excessive at the higher points. In the authors’ 
method, on the contrary, there are no thermometric difficulties of this nature, 
owing to the direct employment of platinum thermometers, and the external heat 
loss increases very little as the temperature is raised, because the external water- 
jacket is always at the same temperature as the inflowing water current, so that 
the mean excess of temperature is always nearly the same. 
Another indication that the temperature of minimum specific heat should be 
not far below the middle of the range is afforded by the experiments of Regnault, 
and more recently by those of Reynolds and Moorby, on the mean specific heat of 
water between 0° and 100°C. Their results by entirely different methods agree 
in showing that the mean specific heat over the whole range does not greatly 
exceed the value at 20° C. 
There is apparently revealed for the first time by the authors’ experiments a 
very rapid increase in the specific heat as the freezing point is approached. The 
point at 4° C. on the curve CB represents the mean specific heat over the range 
0° to 8°. The rapid increase of the curvature as this point is reached is probably 
due to an exceptionally high value in the immediate neighbourhood of 0° C. The 
probability of this result was foreseen by Rowland on theoretical grounds, but 
his original curve, which is accurately a straight line from 5° to 20°, does not 
show the effect. The authors propose to investigate this point more closely by 
taking smaller ranges of temperature, such as 0°-2°, and 0°-4° C., from which the 
actual form of the curve may be deduced. 
With reference to the possibility of obtaining an independent verification of 
the accuracy of the electrical units, and, in particular, of the absolute value of the 
E.M.F. of the Clark cell, it is interesting to compare the absolute values of the 
specific heat deduced by the electrical methods with that of Rowland by the 
mechanical method. For this purpose the authors’ results, and those of Griffiths 
(G.), and of Schuster and Gannon (S.), have been reduced to joules on the assump- 
tion that the absolute value of the E.M.F. of the Clark cell at 15° C. is 
1:4342 volts, as found by Glazebrook and Skinner, assuming Lord Rayleigh’s 
value of the electrochemical equivalent of silver, and taking the international 
ohm as correct. It has been pointed out that the results of Griffiths would be 
brought into harmony with those of Rowland by supposing that the true E.M.F. 
of the Clark cells employed was about 2 millivolts lower, or one part in 700, The 
authors’ results, however, lie about midway between those of Rowland and 
Griffiths, and would require a correction of only 1 millivolt if the whole of the 
difference were to be debited to the Clark cell. It is not at all likely that the 
E.M.F. of the cells employed by the authors can have exceeded the B.O.T. 
standard by so much as 1 millivolt, or that their resistance standards can have 
been incorrect by so much as one part in 700. It is most likely that both the 
Clark cells and the resistance standards employed by the authors agreed with 
those employed by Griffiths to within one or two parts in 10,000, and that the 
difference of the results is mainly to be attributed to the radical difference in the 
methods of calorimetry. 
These and similar questions relating to the absolute values of the standards 
employed do not affect the accuracy of the relative results as regards the variation 
of the specific heat of water with temperature. The relative results are regarde4 
by the authors as being probably as accurate as their present apparatus is capable 
of affording. By far the most important consideration affecting the form of the 
curve in this respect is the particular thermometric scale to which the results are 
reduced. If, for instance, the results were expressed, as originally obtained, in 
terms of the platinum scale of temperature, which differs from the absolute scale 
by only 0°38 C. at 50° C., where the divergence is a maximum, the curve would 
be that represented by the dotted line PPP in the figure, The curve CB is 
