178 COBLENTZ: CONSTANTS OF SPECTRAL RADIATION 



used in this research the aforementioned correct factors for reduc- 

 tion to the optical scale are too large. ^ The writer chose the 

 latter conclusion and the results obtained at high and at low tem- 

 peratures are now in agreement. 



The systematic errors, which formerly caused the mean values 

 of the various sets of observations to differ by 0.2 to 0,5 per cent, 

 are now reduced to less than 0.2 per cent. They were caused by 

 the presence of a disproportionate number of high temperature 

 energy curves, in some sets of data, which contained the afore- 

 mentioned corrections for reduction to the optical temperature 

 scale. This modification of the correction for reduction of the ther- 

 moelectric temperatures to the optical scale has no effect upon the 

 data obtained in 1912 owing to the fact that intentionally, (to save 

 the thermocouple calibration ; the difficulty with the scale correction 

 was not forseen) no energy curves were observed at these high 

 temperatures. The former unexplained disagreement between 

 the sets of observations made in 1911 and those of 1912 is now 

 reduced to an excellent agreement. This was to be expected 

 for there was no apparent difference in the methods of operation 

 other than the addition of a number of high-temperature, energy 

 curves to the series of 1911, which required the optical scale 

 corrections. 



The series of 1911 is the most accurate as regards low humidity 

 while the series of 1912 excel in having the most improved methods 

 and the longest experience in the work. The results of the two 

 years work are given in Table I. Each value of c is the mean of 

 the number of isothermal energy curves (94 in all) given in 

 column 2, computed by equation (2) published in the previous 

 communication to this Journal. The mean value is 



c - 14456 ± 4 

 A = 2911 ± 1 



In the complete paper it will be shown that the values obtained by 

 Lummer and Pringsheim, if computed by the present methods 



^ In the complete paper it will be shown that, for these particular thermocouples, 

 the mean value correction factors for reducing the thermocouple temperatures to 

 the optical scale are too large by 5 to 7, depending, of course, upon the tempera- 

 ture above 1100°C. 



