358 ON THE ERRORS OF CHRONOMETERS, ARISING 



north. Its previous rate was in mean time ; and it was set by, 

 and compared with, a chronometer, whose rate was very near- 

 ly the same. After an interval of 2 hours 18 minutes, it was 

 found to have lost 13 seconds, being at the rate of 135 seconds 

 per day j the XII o'clock mark being now turned towards the 

 south, it gained 12 seconds in 2 hours 12 minutes, being at 

 the rate of 131 seconds per day. The position being again 

 reversed, it began to lose as in the first instance ; but at a 

 rate somewhat less ; and on being once more turned round, 

 with its XII o'clock mark towards the south, it gained 56 

 seconds in the course of the night, being at the rate of 156 se- 

 conds daily. This acceleration above its former rate was pro- 

 bably owing to the diminution of temperature. The regula- 

 rity of a change of rate, from losing to gaining, with each re- 

 version of position in the watch, every thing else being the 

 same, was a sufficient proof that some part of the watch was 

 magnetic, and that that part was acted upon by the influence 

 of the magnet. 



Three means have occurred to me, of either lessening, or 

 altogether obviating, the anomalous action of the magnetism 

 of the balance, viz. 



1. To employ a substance in the construction of the balance 



without magnetical properties. 



2. To free the balance of any magnetism accidentally ac- 



quired. 



3. To prevent the unequal action of the magnetic influence, 



by giving to the chronometer a fixed position as re- 

 gards the magnetic meridian. 



The first method, should there be no practical objection, 

 would go far towards the removal of all magnetic influence. 



The 



