232 PROCEEDINGS OF THE AMERICAN ACADEMY. 



During the course of the experiments the steel cylinders containing 

 the mercury resistance were placed in thermostats by which the tem- 

 perature was usually kept constant within 0.01° during a day's work. 

 Such constancy of temperature as this was not necessary, differences 

 of temperature in the mercury of 0.06° being just perceptible on the 

 bridge wire. Most of this work was carried out at temperatures of 

 about 25°, which was high enough above room temperatures to insure 

 the satisfactory performance of the thermostat. The temperature of 

 the bath was read by a small Goetze thermometer graduated to tenths 

 of a degree and calibrated at the temperature of the bath against a 

 standard Tonnelot thermometer. 



Before making the final calibration against the absolute gauge, 

 many preliminary experiments carried out with varying success showed 

 the necessity of observing rather carefully certain apparently insignifi- 

 cant matters of detail. 



These preliminary tests were made by comparing together a number 

 of pairs of mercury resistances, there being for this purpose two steel 

 pressure cylinders to contain the resistances, two thermostats, and, 

 as has already been mentioned, two bridge wires, either of which 

 could be connected to the extension and balancing coils. The pro- 

 cedure in comparing two mercury resistances was : read resistance 

 No. 1 on slide wire No. 1 ; throw in slide wire No. 2 and measure 

 resistance No. 2; interchange the extension coils with the eight point 

 switch and measure resistance No. 2 again; and finally throw in slide 

 wire No. 1 and measure resistance No. 1 again. If these readings 

 were made at equal intervals of time, as they usually were, the average 

 of the two determinations of each resistance gives the value at the same 

 instance of time. In this way the effects of slight changes of pressure 

 due to dissipation of heat of compression and elastic after effects are 

 eliminated. There was no leak. The pressure was roughly measured 

 with the Bourdon gauge described above. These preliminary tests 

 are competent to decide the question of the reproducibility of the 

 mercury resistance gauge. The question of entire freedom from 

 hysteresis, however, cannot be settled merely by a comparison of two 

 gauges, for complete agreement would indicate only that hysteresis 

 in the glass envelope was the same in either gauge. Entire freedom 

 from hysteresis, within the limits of error, can be shown only by a 

 comparison with the absolute gauge. 



The results first obtained in the comparison of the two gauges were 

 irregular beyond possibility of experimental error, discrepancies of 

 1 per cent being not uncommon. This was found to be due principally 

 to three causes: minute impurities in the mercury, the effect of which 



