136 M. A. J. Angstrom on a New Method of 



error, depending on the so-called personal equation in different 

 observers, can of course be removed from the results in a 

 similar manner. 



§4. 



After these observations I give in Tables I. and II. the ob- 

 servations, or rather the mean values calculated from them. 

 Each observation is the mean of observations during 2 to 5 pe- 

 riods, in which the readings of the thermometers made before 

 the periods were complete and had assumed a constant character 

 were always excluded from the calculation of the mean *. 



Although in general the coefficients of the members which 

 contain the double and threefold angle are so small that there 

 can be no hope of obtaining from them a reliable value of k, 

 they yet furnish an interesting confirmation of the theory, 

 and hence deserve to be adduced. 



If the respective coefficients in No. 1 and No. 2 are divided 

 by one another, the quotients obtained multiplied, and the 

 square roQt extracted, we get 



31-745. 35-203 _ 1 , fin46 _ / 

 13010.23-885 •" 



4 



which values are quite independent of the values of the scales of 

 the thermometers used. 



If, further, the respective angular measures are subtracted from 

 one another, we get 



A/3. A/3\ A/3". 



25 3-5 



37 16-1 



42 25 



24 34-0 



36 1-0 



41 44 



24 48-7 36 38-5 42 5 



Meanwhile from formula (4) we have 



A/3= Vl.A/3'= VJ.A/3"; 

 and if the values of/,/', A/3', A£" are introduced, 

 i i 



/= 1-6046, /^=l-5994, f"^= 1-5201, 



A/3 = 24° 48'-7, s/\. A/3'=25° 55', ^jA/3"=24° 19, 

 an agreement as close as can be wished. 



* M. Thalen had the goodness to help me in these observations j his 

 name is indicated by Th in the Tables. 



