OF IRON, COPPER, AND GERMAN SILVER. 541 



unequally heated stem. In connection with this, Prof. Tait calculated that for 

 a temperature of 250° C. " the utmost error that can be introduced in the 

 indications of the thermometers used is somewhere about 10° C. That is to 

 say, the highest temperatures were read, at the most, 10° lower than they 

 would have been if the whole thermometer had been exposed to the same 

 temperature. This correction of 10° at 250° diminishes at lower temperatures 

 and increases at higher, nearly as the square of the excess of the temperature 

 above the freezing point " (Trans. R. S. E., 1878). 



Wishing, if possible, to obtain some experimental verification of this com- 

 putation, I had two thermometers constructed, with stem, bulb, capacity, bore, 

 and length of degree similar to each other and to those used in determining 

 the temperature towards the hotter end of the long bars. They were tested in 

 the following manner. One was placed vertically, with bulb only immersed, in 

 a bath of melted paraffin wax ; the other, placed horizontally, was wholly 

 immersed in the same bath. The paraffin was maintained at a steady 

 temperature throughout. After their indications had become stationary, the 

 thermometers were read, and the excess of the reading of the wholly immersed, 

 over that of the partially immersed, thermometer noted. From several ex- 

 periments of this kind, it was calculated that the error in question amounts, 

 at 250° C. to 9° 5, an exceedingly close verification of the previously calculated 

 estimate. 



In the reduction of the experimental readings, and in the deduction of 

 conductivity this error has not been taken into account, chiefly because the 

 thermometer used in the experiment on the cooling of the short bar was 

 almost exactly the same in construction as those used in the long bar, and 

 that therefore the difference in this respect between the two experiments must 

 be small. In Appendix II., however, I have endeavoured to correct for this 

 source of error. 



Statical Experiment.— -The bar to be experimented upon had one end 

 inserted into an iron crucible containing melted solder, which is heated by a 

 powerful Bunsen burner. Heat conducted along the bar raises the temperature 

 of each portion of it, and this is carried on until the flow of heat across each 

 section has become steady, a state indicated by the steadiness of the thermo- 

 metric indications along the whole length of the bar. This condition was 

 attained after (about) 7 hours in the case of iron, 6 in German silver, and 

 after 5 hours in copper. This steadiness of state is then maintained for some 

 time, generally at least an hour, after which the distribution of temperature 

 along the bar, as well as the temperature of the unheated short bar in the 

 vicinity, is carefully determined. 



Initially, therefore, the success of this part of the experimental work 

 depends upon the maintenance of this particular distribution of temperature, 



