On Gravimeters. 



531 



becoming aware that there was something wrong. I traced this at 

 last distinctly to a want of yer-martence of the first adjustment. I had 

 spent from first to last not less than ten hours on these adjustments 

 alone, which may give some idea of their uncertainty. 



It was now necessary to ascertain what was the source of the 

 instability. It must be clearly understood that there is not any part 

 of the whole design which is of greater importance than the attain- 

 ment of permanence in this part. If, after disturbance, the lower 

 weight does not return to the same position, relatively to the upper, 

 with absolute exactness, I see no chance of obtaining anything which 

 can be called a result. It is literally a sine qua non. 



The instability might be due to one of two causes — I see no third 

 alternative. Either the holding was insecure ; or the wire was 

 strained beyond what its elasticity would bear. I tried various plans 

 to test this. At first it seemed clear that the holding was in fault. 

 Then I fancied that the wire was strained. I mention this vacillation 

 purposely, because my final conclusion, which condemns the holding 

 after all, though more hopeful, might be wrong, and as it challenges 

 the " special manner " noticed in Mr. Broun's description, it will be 

 best not to be too certain. The followiug test, however, seems con- 

 clusive. 



I prepared two needles of deal, about 2 inches long and as thin and 

 light as seemed necessary. These were split at one end, and thrust 

 upon the taut wire, so as to stand out from it horizontally. One was 

 placed close to the upper holding, but free from contact ; the other 

 about 0*3 inch lower. The upper weight was clamped, and the lower 

 then turned through two entire revolutions. [I did not scruple to 

 overdo it, having ascertained from collateral experiments with other 

 wire that a wire will bear being turned twenty or thirty times to 

 every inch, without any other ill effect than a permanent twist.] 

 The result was that the upper index turned through a few (8 or 10) 

 degrees, and the lower through 70° or 80° — the latter being sensibly 

 in due proportion to its distance : as to the former, I could not say 

 exactly where the holding point might be. Now, the test would be in 

 the positions to which they would return. If the wire was strained, 

 the lower index would not return to conjunction with the upper : if 

 the holding had failed, the two would not return to the starting point. 

 The event proved the latter alternative. The lower index returned to 

 conjunction with the upper ; but both failed to return to the original 

 position. I repeated the experiment, giving the weight only a single 

 turn. The result was the same, in less degree : the holding had again 

 failed, i.e., had allowed the wire to turn in its socket still further. 

 The screws were all firm, on trial. 



The above test is so easy, and useful, that a description of it needs 

 no apology. Still, it is rather by way of proof that I give it ; for a 



