630 REPORT—1899. 
Peabody, in the preface to his well-known ‘Tables of the Properties of 
Saturated Steam’ (1896), as the result of a careful discussion of Rowland’s and 
Regnault’s experiments, adopts Rowland’s values from 0° to 40°, and expresses 
his results in terms of the mean specific heat between 15° and 20°. He finds 
that Regnault’s experiments may be sufficiently represented in terms of this 
unit by assuming the specific heat to be constant and equal to 1:008 between the 
limits 45° and 155°, and constant and equal to 1:046 between the limits 155° and 
200°C. This assumption would make the mean specific heat between 0° and 100° 
have the value 1:0044 in terms of the specific heat at 17°5, or the value 1:0056 in 
terms of the specific heat at 20°, assuming Rowland’s coefficient of diminution. 
The general effect of these changes is to make the tables agree tairly well 
throughout with Regnault’s experiments, but the method can only be justified on 
the ground of expediency, and can hardly be regarded as a satisfactory reconcilia- 
tion of conflicting evidence on account of the assumed discontinuities in the 
specific heat. ; 
Shaw (‘B. A. Report, 1896, p. 162) gives a similar reduction of Regnault’s 
experiments by means of Rowland’s original table, but tabulates only the total 
heat in joules at each point between 100° and 180°C. His reduction shows a 
similar flattening of the curve between 100° and 150°, as compared with 
Regnault’s formula. This may be a physical fact, but might also be explained by 
supposing that the earlier experiments at 108° to 120° were about 0:4 per cent. too 
high. Shaw’s reduction, expressed in terms of a thermal unit at 20° C., is given 
for comparison in the table on p. 631. 
Quite recently a direct determination of the mean specific heat in terms of 
mechanical units has been made by Reynolds and Moorby (‘ Phil. Trans,’ A. 1897) 
on a large scale with Reynolds's break and a steam-engine, Their result expressed 
in absolute units is 4°1832 joules, and is entitled to very great weight on account 
of the minute accuracy of the measurements, and the full discussion of possible 
sources of error. It exceeds the value found by Rowland at 20°C. by only one 
part in two thousand, but is no less than 1:20 per cent. smaller than the mean 
value found by Dieterici—a discrepancy far too large to be explained by any 
uncertainty in the values of the electrical units. 
Unless the mean value found by Reynolds and Moorby is summarily rejected, 
it is clear that the minima of specific heat at 20° and 30° indicated by the work 
of Bartoli and Stracciati and of Rowland respectively, must be due to some con- 
stant source of error inherent in their methods, and that all formule hitherto 
proposed for the mode of variation of the specific heat between 0° and 100° must 
be abandoned. 
It is possible, however, to deduce a more satisfactory comparison of the results 
of Rowland with those of Reynolds and Moorby by means of the present series 
of experiments, on account of their greater range, and the close agreement of the 
individual observations. Neglecting for the present the rapid change of the 
specific heat in the immediate neighbourhood of 0° C., it may be observed that all 
tlie authors’ observations between 10° and 60° (with the exception of one at 55°) 
are represented within one part in 5,000 (ze. within the limits of agreement of 
the observations with different calorimeters at any one point) in terms of the 
minimum value s,,40° at C., by the simple formula— 
8,= 84 (1+ 0°0000045 (¢-40)7) . . . PEC) 
which gives for the mean specific heat between 0° and ¢° the fornula— 
8‘) = 84. (1:0072 —0-00018 ¢ + 000000150 #?). 
If this formula could be assumed to hold beyond these limits over the whole 
range 0° to 100°, the ratio of the mean specific heat between 0° and 100° to the 
specific heat at 20° would be 1:0024. Assuming Rowland’s 4181 joules at 20°, 
this ratio would give the value 4:191 joules for the mean specific heat, a result 
which is still in excess of Reynolds and Moorby’s 4:183 joules, but is not so 
hopelessly beyond the range of possible errors of experiment as that given by 
