78 ' THE ROYAL SOCIETY OF CANADA 



The first column shows the current in amperes passing through 

 the heating coil. In the second column are given the values of a 

 determined from the corrected observations. For purposes of com- 

 parison the column headed a^ gives the values of the temperature 

 co-efficient obtained from the uncorrected observations. Values of 

 n from the corrected observations are given in the fourth column. 

 In order to show the range of temperature involved, the temperature 

 of the upper plate, obtained by extrapolation, is shown in the fifth 

 column. 



The only other determination of a for temperatures higher than 

 100°C. is that of Stafford.^ His results may be expressed approximate- 

 ly by the empirical formula 



K = Ko (14- -00 15(9 +-0000057^2) 



This would give mean values of a over the range covered in the 

 above experiments from -0030 to -0042. 



It will be seen from the values of « given in Tables I, II and III, 

 that they are considerably higher than those which Stafford's formula 

 gives. In general they increase with increasing range of temperature. 



The values of n are more concordant among themselves and de- 

 crease with increasing temperature range. 



The values of a and n depend on the corrections for radiation 

 applied to the original curves, the corrections in some cases materially 

 altering the original values. The probable errors involved in the 

 assumptions made in evaluating the corrections are large. The 

 largest error is probably in the value of the "effective diameter" of 

 the wires of the thermo-element. In the process of fusing the wires a 

 comparatively large bead of metal was unavoidably formed at the 

 junction. It was assumed that the effective diameter of the wires was 

 that of the larger wire, v\z. 1-6 mils. This value is very unlikely to 

 be too large, but may be considerably too small. The shape factor S 

 is only slightly affected by changes in the diameter of the wire, but the 

 heat gained by radiation, Wr, is proportional to the diameter. An 

 increase in the effective diameter of the wire would thus require 

 larger corrections for radiation, which would further decrease the 

 values of a. For instance, if the radiation corrections for the last 

 series of observations for blackened glass (Table III) had been doubled, 

 the value of a would have been reduced from -00500 to -00357. 



In view, then, of the uncertainties involved in the determination 

 of the radiation corrections, the values of a given above may be con- 

 sidered only as upper limits. From their variation with the tempera- 

 ture range it seems evident that a simple formula of the type 



1 Stafford, O. J., Zeits. fur Phys. Chem., 77, p. 67, 1911. 



