EXPERIMENTAL STUDY OF HEAT LEAKAGE. 755 



thermometer ends of the leads which connect the thermometers to the 

 bridge, in order to reduce contact resistance to as low a jBgure as 

 possible. One thermometer, with flat leads similar to those of the 

 Dickinson-Mueller thermometer, enclosed in a flat tube, was con- 

 structed, but difficulty in secm-ing adequate insulation resistance was 

 encountered, and the type was- abandoned. The bulb and stem, 

 which in certain thermometers were separated by a layer of porcelain, 

 are of steel, the stem being \ inch in diameter and having a wall thick- 

 ness of about 0.03 inch. The stem is secured to the head by means 

 of a porcelain joint. The bulb is flat, about | inch wide, and with a 

 space of about 3^ inch between the two flat walls inside. The 

 length of a thermometer from the bottom of the bulb to the upper 

 porcelain joint is about 9^ inches. 



Certain experiments with these thermometers, having for their 

 object the investigation of the magnitude of various errors in the 

 measured difference of temperature on the two sides of the plug, will 

 now be briefly described. 



b. Effect of conduction of heat down the stein of the thermometer. 



The porcelain joints which have been mentioned were employed 

 for the purpose of minimizing any error which might arise from this 

 source. As nothing was actually known regarding the possible 

 magnitude of this error, it was considered desirable to make some 

 experimental tests on it. To this end, a thermometer having two 

 porcelain joints, one between the bulb and the stem and one between 

 the stem and the head, was used. Two sets of clamps, made of copper 

 and soft iron, were constructed: one set of clamps could be applied 

 to the upper joint so that it short-circuited the heat-insulation afforded 

 by the porcelain; the other set could similarly be used to short- 

 circuit thermally the lower joint. Both sets were carefully made to 

 •fit the joint concerned closely and to give good thermal contact with 

 the metal on each side of the joint. Comparative readings of the 

 thermometer when immersed in the vapor of boiling naphthalin, 

 exactly as for calibration at this point, were made, with the following 

 different arrangements: (i) neither joint short-circuited; (ii) both 

 joints short-circuited; (iii) upper joint free, lower short-cii'cuited; 

 (iv) lower joint free, upper short-circuited. Taking (i) as standard, 

 case (ii) lowered the apparent boiling point by 0°.01 C; case (iii), by 



