304 PROCEEDINGS OF SECTION A. 



cell, but found that the liquids diffused iuto one another too 

 quickly, and that it -would rsquire attention every few minutes, 

 and so turned uiy attention to Clark ceils. Professor Threlfall and 

 Mr. Pollock had pointed out (Phil. Mag. 1891), its suitability for 

 small constant currents as well as for constant e.m.f. on open 

 circuit. I set up three, following Lord Rayleigh's directions as 

 closely as possible, and found that they gave perfectly consistent 

 results when closed through resistances of not less than 6,000 

 ohms, the minimum being still lower in the case of one of them. 



To eliminate thermo-electric effects other than the one being 

 measured, I took a copper wire of the same resistance as the 

 two metals whose junction was being examined, and arranged 

 it as an alternative circuit, so that by closing the galvanometer- 

 key the deflection produced was that due to the unequally heated 

 portions of the circuit. Thougli this consisted only of brass and 

 copper, I do not remember ever closing circuit without getting a 

 deflection, so sensitive was the galvanometer. Both junctions of the 

 metals being examined when copper was one of them, as it generally 

 was, were immersed in an oil bath ; and when the junction consisted 

 of two other metals, say lead and tin, the junction of copper and 

 lead and copper and tin were kept in the same cold bath, with a 

 thermometer, the lead -tin junction being in the other, which was 

 heated. No precautions were taken to keep the cold bath at a 

 constant temperature, it being sufficient to know its temperature ; 

 it being possible, from the observations taken, to correct exactly 

 for the rise or fall of the temperature of the cold junction. 



The results obtained are, in a sense, unsatisfactory. Thus, 

 while the mean error of a set of observations might be '5° C, the 

 results would differ from another set, taken under exactly the 

 same circvmistances, by 3° or 4°. I am convinced that the 

 so-called thermo-electric "constants" are not constant, but vary 

 considerably with the least change in temper or condition of the 

 body, and sometimes appear to vary arbitrarily. In a paper read 

 before the Royal Society of Victoria, in the early part of this year, 

 I discussed several months' experiments on the heating of a single 

 metal, and showed that the ordinary thermo-electric phenomena are 

 swamped at a temperature of about a red heat by great and 

 arbitrary (apparently) e.m.f. s. generated in single metals them- 

 selves. In the case of half a dozen different metals one-third of 

 a volt was reached by heating them to about 1,000° C. In some 

 metals this effect on a small scale could be observed at com- 

 paratively low temperature, and would interfere with the ordinary 

 thermo-electric effect. To avoid it as far as possible I only went 

 up to 100° C. in my observations, but even then I sometimes 

 recognised, on a small scale, the irregularities with which I was 

 familiar from jjrevious work. In any set of observations, the 

 relation between the e.m.f. e and the excess of the hot junction 

 over the cold (assumed constant, or corrected for change) /, is 



