EXPERIMENTAL STUDY OF HEAT LEAKAGE. 759 



be insignificant. In such an experiment, the temperature difference 

 existing between the bulb of the thermometer and the binding-posts 

 at the head is not, as a rule, in excess of 5° C, whereas, in the above 

 test, this difference must have been at least 200° C. This would 

 signify nominally an error of 0°.013 in the measured temperature 

 difference ; but in the above-described test, only one of the four bind- 

 ing-posts was heated, so that the ordinary^ function of the compensat- 

 ing circuit was wholly inoperative. This is of course not the case with 

 the Joule-Thomson experiment, where we may reasonably expect 

 the nominal difference of 0°.013 to be eliminated to within 10 per cent, 

 of itself at most. 



To test the effect of inequalities in the temperature of the leads con- 

 necting the thermometers with the bridge, large differences of temper- 

 ature in the coil-circuit and compensating-circuit leads were artifi- 

 cially created. Several tests of this general nature showed that no 

 error from this cause need be feared with the leads as arranged under 

 normal operating conditions. The leads used were of No. 12 solid 

 copper wire, about 22 feet long; resistance of a pair, about 0.037 ohm. 



e. Contact resistances. 



With unplated copper binding posts on the thermometers, it was 

 found necessary to keep them and the lugs attached to them carefully 

 polished, and to set up tightly on the binding post nuts, in order to be 

 sure that the net contact resistance might not exceed 0.0444 ohm 

 (0°.0022 C), the limit of sensitivity of the bridge. With silver-plated 

 binding posts, much less care was necessary, and the comparatively 

 small amount of work needed to put on a thin plating of silver is 

 amply repaid by the abolition of the continual necessity of fussing 

 over contacts and by the additional security felt regarding experi- 

 mental results. 



/. Variations in lead-resistance. 



It is necessary in work of this sort to keep track constantly of the 

 resistances of the leads from the apparatus to the bridge, particularly 

 if all or portions of these are of stranded wire, as is sometimes neces- 

 sary for the sake of flexibility. Mere bending of a stranded lead will 

 sometimes produce a noticeable change in its resistance. An unno- 

 ticed parting of a strand will frequently cause a very appreciable 



