758 TRUEBLOOD. 



The mean apparent rise in the temperature when the glass tube is 

 replaced by the steel tube is 0°.037 C; as will be seen, the difference 

 is quite uniform. Being in the direction opposite to that of the differ- 

 ence which could be caused by conduction along the tube, it was 

 attributed to radiation. To test this, a fm'ther measurement was 

 made with the inside of the glass tube blackened to resemble the inside 

 of the steel tube. The result was Rs-Ra = 0.06895 ohm = 3°.610 C, 

 which is practically the value obtained with the steel tube. In addi- 

 tion to verifying the suspicion that the difference was due to radia- 

 tion, this result substantiates the conclusions reached in the experi- 

 ments described under b regarding the effect of heat conduction along 

 the thermometer tube. The steel tube of No. 8 was of twice the 

 diameter of the stem of No. 4 and also somewhat thicker, ha\dng an 

 area of cross-section perhaps three times as large as that of No. 4. 



The surprising magnitude of this radiation error was sufficient 

 effectually to discourage the use of thermometers of the type of No. 8 

 in the work of the main research. If it is thought that it ought also 

 to arouse suspicion as to the existence of a comparable error in the 

 case of the thermometers actually used, it may be replied that the 

 coil of 8, wound in contact with quartz tubes of high reflecting power 

 and separated by a comparatively thick layer of air from the walls 

 of the containing tube, is necessarily far more susceptible to radiation 

 effects than the coil of a thermometer of the type of No. 4, wound on 

 a strip of mica and fitting snugly into its flat tube, from either side of 

 which it is separated by only a thin layer of mica. Moreover, such 

 radiation error as actually exists must be practically the same for 

 both high side and low side thermometers, and hence without effect 

 on the difference of temperature. 



d. Effect of heat conduction along the thermometer leads, and of 

 unequal temperature distribution in these leads. 



To test this, a wing of thin sheet copper was attached to one of the 

 binding posts of the coil circuit of thermometer No. 8 dining an ice 

 point calibration. The apparent elevation of the freezing point 

 observed when this wing was heated to redness to within one inch of 

 the binding post was 0.00102 ohm = 0°.53 C. The heat was sufficient 

 to oxidize the copper binding post very appreciably. ^Yhile the mag- 

 nitude of this difference seems disturbingly large at first glance, a 

 simple calculation shows that the actual effect in the case of a meas- 

 ured difference in temperature in a Joule-Thomson experiment would 



