ON THE SPECIFIC HEAT OF WATER. 201 



lengthy without proceeding to the ultimate goiil of a comparison of the specific heat 

 curves of solutions with that of water, and the present paper deals only with the 

 results of our apparatus as applied to water. 



It is unnecessary to go into the history of previous researches, since this was 

 carefully and compendiously reviewed by CALLENDAR and BARNES,* and we are not 

 aware that any further work of importance has been done since. The Callendar- 

 Barnes method was based on the electrical heating of a stream of water flowing 

 through a small tulie, the flow-tul>e being surrounded by a vacuum vessel, and the 

 temperatures of inflow and outflow Ix-ing measured by platinum thermometry. 



One of the difficulties with which CALLENDAR and BARNES had to grapple arose 

 from the uncertainty in the electrical units and, in particular, in the value of the 

 Clark cell. BARNES' last correction on this score (April, 1909) involved an alteration 

 of about 1 part in 1000 on the 1902 results. At the present time the international 

 electrical units, in terms of which our results are given, are so closely ascertained 

 that any uncertainty in their value does not affect the third place of decimals in the 

 value of J. But other difficulties remain of a serious character, one of which in 

 particular does not appear heretofore to have received sufficient attention. Measure- 

 ments of the electrical energy developed in a heater usually depend in one way or 

 another upon the value in actual use of the resistance of a standard resistance 

 carrying the heavy heating current (in our case a current of 5 amperes). The 

 resistance of a standard carrying only the small current required for bridge measure- 

 ments may easily be determined with an accuracy much greater than 1 in 10,000 at 

 the various temperatures of the bath in which it is immersed. But the resistance of 

 the standard at the same temperatures of the bath when a heavy current is passing 

 may be quite different. It has long been known that a hysteresis effect may be 

 produced by the paasage of a heavy current, that is to say, a change in the resistance 

 which persists for a time and may afterwards disappear with a well-annealed 

 resistance. We have found reason to Ixslieve that, in addition to this, there is 

 another effect which we may call a thermoid effect, that is to say, a change in 

 resistance which is of the same kind as would be produced by a considerable heating 

 of the resistance, the greater part of which change exists only ichilst the current is 

 passing, the hysteresis effect being merely the small residual effect which persists 

 when the current ceases. We cannot find that the possibility of such an effect 

 has been guarded against in previous electrical determinations of the mechanical 

 equivalent of heat. It might differ according to the particular alloy used for the 

 standard resistance and the current density used in each experiment, and we suggest 

 that it may help to account for discrepancies such as those to which we have called 

 attention. 



In order to meet this difficulty we atandoned the use of a standard resistance made 



* See CALI.KNDAR, "On Continuous Electrical Calorimctry," 'Phil. Trans.,' A, vol. 199, p. 55, 1902; 



UAUNES, "On the Capacity for Heat of Water, &c.," *ime vol., p. 149. 

 VOL. ccxi. A. 2 D 



