498 
MR. E. H. GRIFFITHS OH THE VALUE OF 
closer degree of accuracy by means of a microscope fitted with a micrometer scale, 
but this method is not applicable when large differences have to be read, nor does it 
appear to have been adopted by Rowland.) On p. 94, Rowland remarks that on 
repeating the observations upon an accurate scale, they “ agreed with the scale to 
within two or three hundredths of a millimetre, which was as near as I could read on 
such an object,” and 0‘03 millim. appears to have been equivalent to 0°’01 C. Also, 
the temperature of the manometer column was only ascertained by means of a 
thermometer placed alongside it, and the uncertainty thus introduced might nearly 
double the residting error. 
A study of Rowland’s Tables, XL to XV., leads to the conclusion that the dis¬ 
crepancy between individual observations, due to the above and other causes, in some 
cases amounted to as much as 0°‘03 to 0°'04 C. 
The following extract will, I think, show that Rowland did not rely upon his 
observations to i 0°‘01 C. In his concluding remarks he says : “I now believe that, 
with the improvement to the air-thermometer of an artificial atmosphere of constant 
pressure, we could be reasonably certain of obtaining the temperature at any point up 
to 50° C., within 0°‘01 C. from the mean of two or three observations. I believe that 
my own thermometers scarcely differ much more than that from the absolute scale at 
any point up to 40° C., but they represent the mean of eight observations. How¬ 
ever, there is an uncertainty of 0°'01 C. at the 20° C. point, owing to uncertainty in 
the value of m. But, taking m — •0001 5, I hardly think that the point is uncertain 
to more than that amount for the thermometers Nos. 6163, 6165, and 6166.” 
The causes of error above enumerated do not affect the platinum thermometer 
determinations to the same extent. It is true that the values of d (which corresponds 
to Rowland’s quantity, m), are dependent upon the boiling-points of water and 
sulphur, wdiich are again dependent on barometer readings, but, on the other hand, 
an error of as much as 0°'l C., at those points, would produce so little change in d as 
to leave the value of our range unaltered, although it would slightly affect the actual 
elevation; more especially, it should be noticed that any small changes of pressure 
which may take place during the comparison of the platinum thermometer with other 
thermometers do not affect the results. 
The above considerations appear to me to justify the conclusion that the value of 
Rowland’s temperature range, 14° to 25° C., may be in error by as much as 0°'011 C., 
and, if this is the case, the values of the C.G.S. unit of heat, and of the temperature 
coefficients of the specific heat of water, as deduced from his experiments, may be 
identical with ours. 
2. I have, since the publication of an abstract of this paper, received several com¬ 
munications criticising our statement (see p. 496, supra), that we differ from the 
results obtained by Joule, by “1 part in 350,” and my correspondents have pointed 
out that the ratio, 772‘55 to 778’99, is very different from that of 349 to 350. It 
would, therefore, appear necessary that I should indicate, more fully than I have 
