NO. 6.] INTRODUCTION. CHRONOMETERS. LIII 



The increasing value of x for Htv with increasing temperature seems to 

 indicate that the temperature coefficient had a smaller numerical value on 

 board than ashore. It was next tried to form some means of the satellite 

 points, by which the number of intervals was increased from 3 to 7 ; and by 

 putting y = 0. 8 10 it was found that x could be made approximately con- 

 stant for all the intervals except one (1894 November 1895 March) where it 

 was sensibly (about 0. 8 4) smaller. No pendulum observations were made 

 during this period. The means of the satellite points being, however, rather 

 uncertain, these numbers are not reproduced here. 



As it was apparent from these trials that the second solar eclipse, whose 

 conditions were much less favorable than those of the first, introduced some 

 constraint if the satellite points of the preceding winter were not to be enti- 

 rely neglected, it was lastly tried to leave it out and to use the two remai- 

 ning equations for a direct determination of x and y, viz : 



1893 July 181894 April 5, x + 8.822 y = 0. 8 253 



1894 April 51896 Aug. 22, x + 11.202 y = 0. 4265 



which give x = 0. 8 390 and y = 0. 8 073 and would imply a correction of 

 -(- 12 s to the result of the second solar eclipse, corresponding to a somewhat 

 early observation of the second contact as compared with the first, parti- 

 cularly for Sverdrup's observation with the smaller instrument. 



By putting in round numbers 



y = -O."080 



and determining x from the mean temperature of the whole time (10.656 C) 

 and the mean rate ( 0. 8 3864) viz : 



x = +0. 8 466, 



only 3 seconds were sacrificed from the first solar eclipse, which brings the 

 result nearer to the mean of I 8t and 2 nd contact, estimated as corresponding. 



As these values seemed to be slightly more concordant with the satellite 

 points, the formula 



Daily rate = 0. 8 466 0. S 080 t 



was finally adopted. On this basis the first table of the "Results", containing 

 the error of chronometer Hohwu for every 10 th day, has been calculated. 



It may be noticed that the rate of Hw during the 11 series of pendulum 

 observations, with temperatures between +5 and +15 C, as calculated by 

 this formula, nowhere differs more than +0."1 from the values obtained in 



