291 



tion must remain equal to many pounds at its place of action ; 

 all the parts, therefore, which intervene between the pivots of 

 the instrument and the point of resistance of the largest screw 

 must be under strain equal to the moment of this friction. 

 But, as is well known, any slight tremor will relieve a part of 

 the friction, and the circle must take a new position. Ac- 

 cordingly, I find that if (for example) I make the two images 

 coincide by moving the reflected from right to left before a 

 train, when that has passed it will be found to have started 

 from two to four seconds towards the left, and vice versa. 

 This I think the most dangerous of all these disturbances ; for 

 if a train passes after the star is bisected, and while the obser- 

 ver is reading the microscope, the whole or a part of this jump 

 will be changed to the star's place, and the direction of the 

 error will depend on that in which the screw was last turned. 

 I have, therefore, been obliged to establish it as a standing 

 rule, that all such cases are to be noticed in the journal, and 

 the observations rejected. 



" 3. The adjustments appear to be slightly affected ; of this, 

 however, I cannot speak with confidence, except in respect of 

 the horizontality of the circle's axis, and its meridian position. 

 The transit instrument has always been kept closely adjusted, 

 and, therefore, cannot show these deviations; and the index 

 correction of the circle is by its construction scarcely liable to 

 change but by extreme violence. The circle's axis is levelled, 

 by first correcting collimation in right ascension with two col- 

 limating telescopes, and then making the images of the central 

 wire coincide ; so that it can be at any time easily examined. It 

 is kept in position by two check-nuts, on a strong screw, and 

 formerly would remain many months before it erred to the 

 amount of five seconds, which it was not allowed to pass. At 

 present as great a change will sometimes occur in four days, 

 and that abruptly. The disturbance in azimuth is less in 

 amount, and I think independent of the other. I have not 

 detected any effect on the clock. 

 vol. v. 2d 



