PRESSURE ON RESISTANCE OF METALS. 577 



of resistance due to heating. It should be emphasized that this 

 avoids a source of error present in the work of Lisell ^ and Beckman.^ 

 They allowed the current to flow continuously through the wire. 

 An}^ change in the thermal conductivity of the oil under pressure 

 would change the heating effect and so partly mask the change of 

 resistance due to pressure. If thermal conductivity of a liquid 

 decreases under pressure, as seems plausible, this source of error would 

 make their pressure coefficient too low; which as a matter of fact is 

 the direction of discrepancy between most of their results and mine. 

 The magnitude of the error from this source varies with the material 

 and the size of the wire. Lisell states that for lead his error from this 

 source is probably not as much as 1%, but for zinc he is willing to 

 admit the probability of an error as large as 10%. Beckman, working 

 with Lisell's apparatus, makes no mention of this source of error, and 

 would seem not to have sufficiently guarded against it. 



Temperature control was one of the most troublesome difficulties 

 of the preliminary work. The effect of temperature on resistance is 

 very large compared with that of pressure; for some substances 1° C 

 may make a change in resistance 3000 or 4000 times as large as 1 kg. 

 pressure. The upper thermostat in which was the manganin measur- 

 ing coil gave no trouble; it was sufficient to keep the temperature of 

 this within 1°. For the lower thermostat a bath of water violently 

 stirred was used, with a very sensitive regulator consisting of a spiral 

 of thin walled copper tubing filled with ethyl benzoate and connected 

 to a regulating mercury column in a glass capillary by means of a steel 

 intermediary part to which the copper was soldered and in which the 

 glass was directly sealed without cement. By means of a Beckmann 

 thermometer and adjustment by hand of the mercury contact if 

 necessary, temperature could be kept constant within two or three 

 thousandths of a degree during a run. Error from slow drift was 

 avoided in those cases where the temperature coefficient was high 

 compared with the pressure coefficient by using as the comparison 

 coil another coil of the same substance and resistance as the coil 

 subjected to pressure, placed in a glass tube in the bath in close 

 proximity to the cylinder. For those substances with larger pressure 

 coefficient this precaution was not necessary, but a comparison coil 

 of manganin at room temperature was sufficiently good. Of course 

 the temperature of 0.0° was obtained in an ice bath, and no tempera- 

 ture trouble was ever found here. Difficulty of temperature control 

 probably accounts for the fact that most previous measurements 

 have been made only at 0.0°. It was not possible to run the thermo- 



