284 



PROCEEDINGS OF THE AMERICAN ACADEMY. 



log L = n log (t c — t) + log A. 



That is, if log L is plotted against log (t c — t) one should get a straight 

 line. This turns out to be remarkably near the truth. Thiesen origi- 

 nally suggested n = 1/3 ; Henning 32 showed that his observations 

 below 100° could be represented by putting n = 0.31249 and A = 

 94.21 ; a careful plot, a year ago, of the values available before the 

 appearance of Henning's work above 100°, but including the values in 

 Table I. in this paper, led to n = 0.3150 and A = 92.93. The work 

 has been carefully repeated this fall. Including Henning's new work 

 and the values in this paper, 37 values of L are available. They were 



TABLE III. 



plotted logarithmically dn a large scale, and the slope of the line that 

 best represented them was determined graphically by stretching a thread 

 among the points. This was done several times by each of two 

 different people, their results being closely accordant. The average 

 of their values of n was then used to compute A arithmetically. The 

 result is exactly the same as that of a year ago, namely, 



L = 92.93 (365 - 0°- 3150 



The average of the numerical values of the differences between the 37 

 observed values of L and the numbers computed by means of the 

 above formula is one fourteenth of one per cent, which is less than the 

 probable accuracy of the measurements. It is true that there is some 

 evidence of regularity among the deviations as the above table shows. 



3 2 Wied. Ann., 1906, 21, 870. 



