for the Chronometers at Greervwich. \ 35 



nometer: and in adopting this method, the limits might be re- 

 duced to 4 seconds for the first premium, ^\ for the second, 

 and 5\ for the third. Why the error from the rate has al- 

 ways been doubled, I am at a loss to conceive. And now I ask, 

 what would have been the result of the trial as exemplified 

 by Caleb Mainspring, in which he makes his own common 

 watch to beat his master's chronometer ? According to my 

 proposal it would have been as follows : 



My Master's Chronometer. ^^ 



Greatest mean monthly rate 6*70 



Least ditto 5'26 



Difference r^* 



Greatest extreme variation"^ 



in April = l"-20 ^Mean= 1-20 



in May = 1 '20 J 



Trial number 2-64 



My own Common Watch. ^ 



Greatest mean monthly rate 6*80 



Least ditto 6-80 



Difference 0-00 



Greatest extreme variation ~) 



in April = 3"-80 > Mean = 3-80 



in May = 3 -Soj 



Trial number 3*80 

 And so would have been the reverse of what he made it by 

 the present method of ascertaining the trial number. I have 

 now suggested what I consider to be the best mode of deciding 

 the quality of a chronometer. With respect to the proposal 

 of J. L. T., it appeared to me tLat it would not be so practi- 

 cable, nor perhaps so good a test, as the one I have made ; as, 

 on examining his system, an idea entered my mind, that if, 

 instead of taking the mean rate of the whole time of the trial 

 as a standard rate for deducting each day's difference, the 

 mean rate of the first month was taken as that given rate, the 

 result would not be a bad test. But after trying the result 

 upon two chronometers that had equally varied from their 

 rates, and had also gone with the same regularity, I found 

 that the result of the trial would be in favour of that chrono- 

 meter wiiicli went nearest to the first month's rate during the 

 first part of the trial, though it varied at the latter end : but 

 by taking the mean of the whole time as the given rate, the 

 result was in favour of that which varied from its rate at the 

 commencement of the trial, but which made no greater error 

 afterwards. As I consider those (wo chronometers ecjual in 



riualitv, 



