4 Palmer — Pressure Coefficient of Mercury Resistance. 



ner that a spring contact, which served as one terminal of the 

 galvanometer, could be readily placed on any point of the 

 wire and its position accurately determined by a large microm- 

 eter head divided into one hundred equal parts. This arrange- 

 ment presents a great advantage over the ordinary form of 

 drum bridge since the only friction connections are in the gal- 

 vanometer circuit, where the worst effect they can produce is 

 small variations in sensitiveness, and not, as they are usually 

 placed, at the terminals of the wire, where changes in their 

 resistance produce the maximum effect on the result. Current 

 was supplied to the bridge by a single Daniell's cell and was 

 so regulated by a small rheostat in series with the battery that 

 its intensity was never sufficient to appreciably alter the tem- 

 perature of the resistances in circuit. The atfainment of bal- 

 ance was judged by a very sensitive and dead beat D' Arson val 

 galvanometer of the horizontal magnet type and its indica- 

 tions were observed by the telescope and scale method. 



When the Queen resistance box was bought, some years ago, 

 it was accompanied by a certificate from Professor Anthony to 

 the effect that its readings were correct to one-fiftieth of one 

 per cent at 17*5° C> and that its temperature coefficient was 

 •00023. The coils used in the present investigation have 

 nevertheless been very carefully calibrated and the values thus 

 found used in all the calculations, for, though many of them 

 came quite up to the guarantee, several showed deviations 

 somewhat larger than the probable errors of observation. The 

 resistance of the bridge wire was determined in the following 

 manner. Let the reading of the bridge micrometer, when 

 balance has been obtained, with two nearly equal resistances R 

 and R' in circuit be x, and when R and R' are interchanged x' . 

 Then if z and z' are the corresponding readings when R has 

 been increased by a known increment oZK it is easy to prove 

 that 



_ dR 



~~ (x— z) + (z'— x') 



where r is the resistance of a length of the wire correspond- 

 ing to one division of the micrometer. About one hundred 

 determinations of this quantity, involving various lengths and 

 different portions of the wire, gave the mean value 

 r = -000898 ohms 



the greatest difference between a single observation and the 

 mean being less than 3xl0" 6 ohms. These measurements 

 also showed that the error of a single setting of the micrometer 

 was about one-tenth of one division and hence that the mean 

 error of a single determination of a resistance, due to this 

 cause alone, was less than *0001 ohms. Throughout the entire 



