( 260 ) 



the errors appeared to satisfy fairly well the set of equations. Where 

 it was necessary I took away the last dilFerences by adding 



^.% 



to the errors. 



Table III contains the results of the compensation which are 

 necessary to compute the clock correction at any instant between 

 the observations. Let this instant be q -\- T, the epoch of the im- 

 mediately preceding clock correction being q; if moreover during 

 the interval from q to q~\- T: 

 the mean atmospheric pressure, expressed in mms. of mercury 



of 0° C. is 760 + Br, 

 the mean temperature in centigrades 10^ C -|- d-x, 



the mean difference in temperature in centigrades Vt, 



then we can compute the clock correction S(j-\-t according to the 

 following formula: 



5,4.r=5,-/, + Y^(G,4-15,4^r-26,4^T+318F7' + C,T+^„T'). 



The values of S, f, G, C, E, occurring in this formula can be 

 derived from table III. 



The 5"' column shows the mean rates for each interval between two 

 successive determinations of tiie clock correction after the reduction and 

 the compensation. From this we may judge of the constancy of the rate 

 of the clock. It must be remarked that the small yearly inequality 

 occurring in these values is very probably due to a little inaccuracy 

 in the coefficient of temperature obtained as described above. 



The last column of table III shows the quantities 2, the diffe- 

 rences between the successive values of E. They give us a simple 

 test for the computation of the compensation, because the}'' must be 

 equal to the errors ƒ multiplied by 20, or ƒ, = 2,^ : 20. The adopted 

 series of errors satisfies tolerably well this relation, if we admit 

 small diiferences, which in thousandth parts of a second of time do 

 not exceed the intervals m. or n expressed in days, and iience give 

 rise to a difference less than 0-*,00I in the mean daily rates. 



