31 



2. Some other sources of uncertainty remain to be mentioned. Under Antarctic conditions, with the 

 instrumental outfit supplied, it did not prove possible to take astronomical nbscrvatiniiu <>f mitiicient 

 accuracy to determine chronometer rates from day to day, with the high precision desirable for pendulum 

 observations. From time to time, at intervals varying from 8 days to 4 months, stellar and solar observa- 

 tions were taken by Lieutenant ARMITAGE with a theodolite, and from these he deduced the error of the 

 chronometer " A," which served as the standard to which the others were referred. These observations 

 were carefully made, but between the dates of two successive observations the rate of the standard A had 

 to be assumed uniform. 



During the pendulum observations, Mr. BERNACCHI compared the chronometer, Kulllicrg 0711, used 

 in the pendulum observations, with A, through the intermediary of a chronometer watch ; this was 

 compared with A in the ship, and with 6711 in the observational hut. The rates accepted for 6711 thus 

 depend on the accuracy of the comparisons with it and with A of the chronometer watch used as inter- 

 mediary, on the steadiness of A, and ultimately on the accuracy of Lieutenant AHMITA<;K'S observations. 



To reduce the uncertainties of the comparison with the intermediary watch, Mr. BKIINACCIII'S usual 

 practice in the later observations was to compare it with 6711 after an exact 24-hour interval by the 

 watch, taking seven successive readings of the watch at 10-second intervals, and estimating the 

 corresponding times on 6711 to O'l second. Different watches were employed during the different sets 

 of observations. That employed in September, 1903, had the steadiest rate, and the uncertainties as 

 to the rate of 6711 were then probably least. 



As to the accuracy of Lieutenant ARMITAGE'S observations, one can form an opinion only from the 

 greater or less apparent regularity in the results. For some time after its arrival at \Vinter Quarters, 

 chronometer A seems to have gained slightly. It then began to lose, and continued to do so during the 

 remainder of the time. The losing rates deduced from Lieutenant ARMITAGE'S observations varied ;i- 

 follows : 



. Bate of A losing. 



May 13, 1902, to August 7,1902 .... 0'60 seconds 

 August 7 October 26 . . . . 0'90 

 October 26 February 4, 1903 .... 0-75 

 February 4, 1903, to February 21 .... 1-35 

 February 21 March 1 . . . . 1-12 

 March 1 July 6 .. ... 1*17 



July 6 October 8 .... 1-12 



The apparent irregularity in the rate about February, 1903, is suggestive of some uncertainty in 

 the observation on February 4. The rates actually assumed as applicable to A during the pendulum 

 observations at Winter Quarters were : 



July to August, 1 902 - 6 seconds 



February, 1903 -1-1 



September - 1 1 



These assumed rates are hardly likely to be affected by any large errors. At the same time it is 

 impossible to feel absolutely certain that an error as large as 5 second may not have existed, especially 

 in the result assumed for February, 1903. 



The rates finally deduced for the pendulum Chronometer 6711 during the observations at Winter 

 Quarters were : 



July to August, 1902 -0-8 seconds 



February, 1903 -T2 



September , -1'3 



Chronometer 6711 when at Kew, both before and after the Expedition, possessed a very steady rate, so 

 that the uniformity in the above results is at least in harmony with its general character. 



