( 94 ) 



For the simple means of the 4 winter and the 3 summer rates 

 and of the temperatures belonging to them I find : 



26°.28 + 0.77 



20 .61 — .65 



hence : Variation per degree -j- s . 25 l ) 



By means of this coefficient of temperature I have reduced all the 

 rates to 20° ; these reduced rates are given in the table above in 

 the column: D. Rate 20° C. The simple mean value of these 

 reduced rates is — s . 87, from which the real mean reduced rate 

 — s . 83 differs only little. By forming the differences between the 

 reduced rates and their mean I found for the mean error of a daily 

 rate, disregarding the different lengths of the intervals between the 

 time determinations : 



M.E. ± (K225 



a very satisfactory result, especially in consideration of the fact that 

 for the whole period of more than 3 years we have adopted a 

 constant rate depending on the temperature only. 



For the chronometer of Hohwü the results are somewhat less 

 favourable. One sees at a glance a distinct variation with time which 

 from 1904 seems to continue in the same sense. 



In order to derive the coefficient of temperature I have compared 

 each summer rate with the mean of the two neighbouring winter 

 rates and thus found : 



Rate summer — winter -\- s . 78 



+ .03 

 — .05 



A regular influence of the temperature does not appear from these 

 data and the greater value of the first difference must be ascribed 

 to an irregular variation in the beginning. Therefore I have accepted 

 for the coefficient of temperature s . 00 and, in order to investigate 

 the variation which is independent of the temperature I have formed 

 mean rates for periods of about three months. They follow here 

 together with the corresponding values for the chronometer of Hewitt 

 reduced to 20°. 



l ) It was impossible to determine also a quadratic term on account of the small 

 differences of temperature. For the years 1901 — '02 the temperature coefficient 

 was found to be -f 0*.18. 



