CHAP. XIII.] OBSERVATIONS FOR ERROR 299 



employed in any string of meridian distances should be the 

 same, the instruments and watches the same, the temperature 

 and time of observing the same, as far as possible. Also, 

 supposing temperature to be the same, that a rate will be 

 probably more correct if obtained by combining single altitudes 

 of different days, both either a.m. or p.m., than by taking 

 equal altitudes one day and single altitudes the other. 



If clouds prevent observation at 'precisely the same altitude, 

 after transit, the mean of Error obtained by absolute sights, 

 a.m. and p.m., at nearly the same altitudes, will be almost as 

 good as equal altitude sights. 



When the observers are good, the greatest error is frequently Com- 

 introduced in comparing the hack-watches to be used for watches, 

 taking time with the clu-onometers, and great pains should 

 therefore be taken with tliis operation. 



The watches used for taking time must be compared before, 

 and after, both forenoon and afternoon sights, with the standard 

 and another clu*onometer ; and at noon all the clironometers 

 should be compared with the standard, and the hack-watches 

 with the same two chronometers. 



In the case of stars, when return cannot be made to compare, 

 the proper comparison for middle time must be deduced by 

 interpolation from the comparisons before leaving and on 

 retm-ning. 



Before saying more about comparing, we must remark that Defects in 

 the seconds hands of pocket-chronometers are rarely placed q?^*^®*" 

 symmetrically in the centre of the dial on which the seconds meters. 

 are marked. 



These watches beat five times in 2 seconds, commencing 

 with the even minute. The beat of the watch should therefore 

 coincide exactly with every even second ; the first beat from 

 the minute being 0-4 second, the second 0-8, the third 1-2, the 

 fourth 1-6, the fifth at 2-0, and so on throughout the whole 

 60 seconds. 



But it will be found, in nearly every watch, that the hand 

 does not fall over the even second on some parts of the dial, 

 although it may on others, and each watch must be examined 

 by counting the beats from the even minute, to ascertain how 

 the hand falls in different parts of the dial, or the time-taker 

 will be at a loss to know what is the exact decimal which his 



