32 L. B. Fletcher — A Determination of the B. A. Unit 



closely with the general mean. Experiments made on Nov. 

 29 and Dec. 6 were rejected on account of leakage of 3 and 6 

 grams respectively. The results, however, are 4220 and 4227, 

 falling within the limits of the series. The duration of the 

 experiment was less than one-fifth of the interval between the 

 two weighings of the calorimeter, and probably the loss during 

 the experiment was tbe same fraction of the whole loss. 

 Furthermore, the leakage during the radiation experiment 

 would affect the radiation coefficient in such a manner as to 

 approximately compensate for the effect of leakage during the 

 main experiment. For these reasons I have retained the 

 experiments of Jan. 26 and Feb. 16, which showed a leakage 

 of 1 gr. and 1*5 gr. respectively, but have given the results half 

 the weight of the others. The remaining experiments are 

 satisfactory in this respect, the loss being a few tenths of a 

 gram, due principally to the removal of the thermometer. 



The result of the experiment is J=42, 039,000 O, where O is 

 the value of one-tenth of the 10 ohm coil in earth-quadrants 

 per second. Reduced to the new value for the constant of the 

 sine galvanometer, it becomes J = 42,055,000 O. 



I have also calculated the experiment from the formula 



c~Ht 

 J= — — , where R is the resistance of the calorimeter coil as 

 a 



measured in the ordinary manner, corrected to the mean tem- 

 perature of the water and further corrected for superheating. 

 I estimated the superheating from observations of the main and 

 derived currents when the strength of the former was varied. 



c'W 



The expression should give the true resistance of the coil 



at any instant. When the currents are smaller the superheat- 



c'R' 



ing is less and the comparison of the value of for a zero 



current, obtained by graphical extrapolation, with its value for 

 the full current, should give the superheating. The method is 

 not very accurate, as the observations with the smaller currents 

 are uncertain. I find the increase of resistance due to super- 

 heating to be about 1 part in 700, corresponding to a difference 

 of temperature of 2° C. When this correction is applied the 

 second method of calculation gives J= 42,156,000 O. 



This result depends upon the square of the main current, 

 and as the temperature of the coil changed 6° or 8° during the 

 experiment, its mean resistance is somewhat uncertain. Hence 

 this result has not the weight of the former. 



The discovery of this discrepancy has greatly delayed the 

 publication of this paper. It may be due to conduction to 

 the water, which was guarded against by varnishing the wire 

 and using distilled water, but was not proved not to exist. 



