lOi An Account of the Repealuig Circle, 



observations for finding the attraction of a mountain with a 

 small instrument of the construction R ; and obtained a de- 

 flection of the level 'equal to two seconds ; and, although his 

 telescope could not have been more than 15 inches long, fi'om 

 this experiment brought out a density of the earth nearly co- 

 inciding with the Schehallien experiment, and with the more 

 recent one which Cavendish obtained by direct attraction. 

 Yet has that same astronomer, from later experiments, found 

 ihojixed error of the same or a similar instrument (made by 

 a foreign artist by no means unknown to fame) to amount to 

 from five to ten seconds. Should this error, however, which 

 is called fixed, turn out to be at all fortuitous, it is possible 

 that, taking five for its amount (to say nothing about ten), a 

 result might have been obtained of an equal quantity contrary 

 to attraction. But on the other hand, were the error of such 

 an instrument absolutely fixed, although it gave the altitudes 

 incorrectly, yet might it give the dilierences correctly; and 

 consequently the above dv/arfish experiment might be pei- 

 mitted to stand on its own little base. It is a circumstance 

 little suspected, and not very easy to explain, how an instru- 

 ment, that is capable of reversion, can have errors that are 

 not correct in the double result. But although this, like the 

 above statement, and die remarks I have made upon it, may 

 appear digressive, yet will it not be altogether useless to in- 

 quire into the nature of two of the most obvious sources of 

 error, to which this instrument in particular, and others in a 

 less degree are subject. 



It is not perhaps so well known as it ought to be, either to 

 observers or artists, that the air-bulb of the spirit level changes 

 its position with a difference of temperature. This, it is pro- 

 bable, is wholly occasioned by the glass tube (I mean from its 

 expansion) being larger at one end than at the other: for I 

 have observed that as the temperature increases, the bubble 

 alwaj'S deviates towards the larger end, and as it diminishes, 

 the deviation takes place the contrary way. The error occa- 

 sioned by this cause, whatever it may amount to at the end of 

 a series of repetitions, will be divided, and affect- the result by 

 no more than the mean tleviation belonging to a pair of obser- 

 vations. This kind of error is fixed, as far as a single course 

 of repetition is concerned ; but, when the change of tempera- 

 tui*e is reversed, an error equally fixed will affect observations 

 in the opposite diriiction. 



Another and still more fatal source of error to which R is 

 peculiarly liable, arises from the resistance of the centre-work 

 to the action of the tangent-screw. This will be more or less, 

 according to the care anil judgement that have been employed 



in 



\ 



