232 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



TABLE III. 



Between these observations a considerable length of the wires was 

 necessarily clipped off. Reduced to a common temperature of 445°, 

 the maximum difference is but G microvolts in 5290, or 0.11 per cent, 

 while the average deviation of a single observation is but 0.02 micro- 

 volts, or 0.04 per cent, and of the mean but 0.02/V3 = 0.012 micro- 

 volts, or 0.024 per cent. At higher temperatures the discrejjancy was 

 even smaller. 



The validity of the interpolation formulae has been ah-eady dis- 

 cussed. A statement of the extreme error which may have been intro- 

 duced into the results by this source should however be added. This 

 is believed to be for aluminum less than ± 2°, for silver less than ±2°, 

 for copper less than ±0.5°, and for platinum less than ± 10°. 



Comparison with the temperatures computed by the Avenarius 

 equation show errors by the latter to be about 1.5 times as great for 

 water and naphthaliii, and of the same signs. It is therefore much 

 less reliable, especially for the platinum temperature, and no weight 

 is attached to its results. 



Melting Points hy various Authorities. — A collection of these is 

 given in Table IV. Except in the case of the Barus data, the results 

 are set down directly as given by their authors. A further discussion 

 of these with reference to the purity of the metals used, and the char- 

 acteristic errors of the methods employed, would doubtless prove 

 instructive, and might partly remove or account for some of the 

 apparent discrepancies, — a task which will perhaps be undertaken 

 later. 



