1 82 
DR. H. T, BARNES ON THE CAPACITY FOR HEAT OF WATER 
would have produced an error less than '00002 ohm, or less than 2 parts in 50,000, 
whereas it is probable the actual difference in temperature did not make an error 
one-tenth of this amount. 
Sec. 3 c.—Measurement of Temperature. 
By far the most important factor that determines the character of the curve for 
the variation of the specific heat of water with temperature is the particular 
thermometric scale to which the results are referred. This was most forcibly brought 
out by Rowland in his memoir, and it was pointed out by him that without the 
greatest care in reducing his mercurial thermometers to the air scale, the value of the 
specific heat of water would have apparently remained constant in terms of the 
mercurial scale over the range of his experiments. The discovery of the rapid 
decrease in specific heat with increase of temperature from 0° to 30° C. was only 
made through this careful reduction. 
In the present series of experiments there were no thermometric difficulties such 
as are to be met with in the use of a mercurial standard owing to the use of platinum 
thermometers. In working to the 10,000th part of a degree Centigrade, such 
corrections as a change of zero, pressure on the bulb, capillary and stem corrections, 
are so large in the case of the mercurial standard, that for large intervals of 
temperature the readings are far from reliable. With the platinum thermometer we 
still have to deal with the question of a change of zero and a stem correction, but 
these are so small that with sufficient care they may be eliminated altogether. 
In speaking of these possible sources of error in connection with the measurement 
of temperature with the platinum thermometer, I am referring to a limit of accuracy 
seldom required in most determinations. The first source of error is already well 
known, and has often been subject of controversy over the reliability of the platinum 
thermometer, though chiefly, I am convinced, by those who are either prejudiced or 
who require more experience in this class of work. I have met with no difficulties of 
this nature that could not be attributed to my own carelessness, or could not be 
easily avoided with sufficient patience and care. In regard to the second source of 
error, I have never seen it referred to before in connection with this subject, and will 
therefore speak about it somewhat further on. We should, strictly speaking, include 
with the electrical measurements the subject of platinum thermometry. We shall, 
however, include it with the thermal constants and treat it entirely from that point 
of view. 
The measurement of temperature by the change in resistance of a platinum wire 
has been carefully studied by Professor Callendar, and his work is already too well 
known to make it necessary for me to dwell on the fundamental part of it. His 
introduction of the idea of a platinum temperature which depends on the term, 
