378 



PROCEEDINGS OP THE AMERICAN ACADEMY. 



is satisfactory to obtain practically the same numerical value of the 

 Thomson Effect and with one exception the same decided evidence of 

 the diminution of the Thomson Effect with a rise of temperature. 



Accuracy of the Resistance Measurements. 



On page 371 it was pointed out that the decrease in the value of the 

 Thomson Effect with rise of temperature could not be explained by 

 even large errors in the assumed value of the thermal conductivity or 

 in the form of the " h curve." Omitting these two sources of error, the 

 balance of the work depends on tlie accuracy of the resistance measure- 

 ments. The accurate measurement of such low resistances as 1 cm. 

 lengths of no. 18 copper wire, presents considerable difficulty. How- 

 ever, an examination of the values obtained for the same sections of 

 the bar at different times and under different temperature conditions 

 will show that some success was attained. 



These figures show a very satisfactory accuracy, the agreement in one 

 case being as close as 1 part in 7,000, the worst agreement being 1 

 part in 700. 



Comparison of the Results with those of Batelli. 



Owing to the fact that Batelli made no experiments with copper, 

 no direct comparison with his results can be made. He obtained for 

 iron a value of .00001215 at a temperature of 108° and his values in- 

 creased very rapidly with the temperature. This result is roughly 

 twice as great as the value found above for copper. From the slope 



