404 On the present Mode of finding 



time ; and to ascertain the degree of dependence that can 

 be placed upon it, during the intermediate part of that time : 

 or, to find whether the quantity of error increases uniformly 

 and regularly in proportion as the time increases. It would 

 seem, that in all the trials hitherto made of timekeepers in 

 fixed observations, this object has only been partially pur- 

 sued ; whilst another of equal importance has been alto* 

 gether neglected : that is, no method has been adopted, for 

 finding the unavoidable alteration in the rate, produced by 

 the different changes of temperature, to which most long 

 voyages are liable : but on the contrary, the rate has been 

 ascertained, merely for that temperature which happened to 

 occur whilst it continued under trial, without endeavouring 

 to find, whether any change would take place in the rate, 

 if a material alteration should be produced in the weather, 

 from heat to cold, or from cold to heat. 



Can it reasonably be expected, in machines like these, that 

 a rate found in a temperature of 30° of Fahrenheit, without 

 any greater variation than lo 3 en either side, will be ade- 

 quate to compute forward, and find the error of a watch 

 that is afterwards to be kept going in a temperature of 100° 

 or higher, wherein the expansion of the metals is so diffe- 

 rent from that of the former? It has also frequently hap- 

 pened, that the rate has been obtained in the coldest part of 

 winter, when metals are most contracted, and applied to 

 the hottest part of summer when they are most dilated, 

 without any correction for the unavoidable defect of exact 

 compensation ; or that the rate has been found in a very 

 cold latitude, to apply to the going of the watch in a very 

 warm one ; and vice versa. 



\ Perhaps it may be asked here, Then of what use is your 

 compensation ? To which I reply, The value of this com- 

 pensation in the balance is not depreciated because the ut- 

 most degree of perfection cannot be attained in adjusting it, 

 anv more than the value of the chronometer because it does 

 tiot keep exaetly with mean time, or any more than the in- 

 direct method of finding the true from the mean anomaly, 

 because it is done by means of an approximation. 



When chronometers have 5een sent on trial for twelve 



month* 



