348 PROCEEDINGS OF THE AMERICAN ACADEMY. 



to this levelling ; but even a careful observer may have the habit of set- 

 ting the bubble a little to one side of its proper position, always to the 

 right, let us say ; and against error from such a habit the device of turn- 

 ing the plate through 180° in the course of the observations is no safe- 

 guard. If the observer could have moved about so as to be half the time 

 on the east and half the time on the west of the apparatus, his trick of 

 the eye could have been eliminated from the average conditions govern- 

 ing the experiments ; but this was not the case. Some of the levellings 

 were made by my assistant and some by myself, and accordingly it be- 

 came necessary to study the " personal equation " of each of us in the 

 use of the levels. This I did July 29, 1902, when it appeared that both 

 had the habit of setting the bubble too far to the right, which would 

 carry the plumb-line to the south of the release position of the balls and 

 therefore tend to cause an apparent northerly deviation of the balls in 

 falling. This effect, according to the observations of July 29, was prob- 

 ably somewhat more than 0.001 cm. It is not likely to have been as 

 large as 0.002 cm. 



As to faults of levelling at the receiving apparatus, the changes of posi- 

 tion of this apparatus, and of the observer with respect to it, were such 

 as possibly to eliminate error due to such faults from the final result ; but 

 whether it was perfectly eliminated, I have some doubt, as this levelling 

 was less careful than that at the sending apparatus. There were two 

 places for faults, one in levelling the ring which supported the receiving 

 pan, the other in levelling the tripod which held the tip of brass above 

 the imbedded ball. 



The latter operation was performed for every ball, and as the observer 

 was about half of the time south, and half of the time north, of the ap- 

 paratus during the levelling, his errors in this particular should very 

 nearly eliminate each other. The levelling of the supporting ring was not 

 so often or so systematically done. On July 24, almost at the end of my 

 main series of observations, I found the north side of the ring to be 

 a little higher than the south side, the inclination being, perhaps, 5'. 

 This would make the north side of the one inch ball fall throuirh about 

 0.0036 cm. greater thickness of tallow than the south side. The effect 

 of such a dissymmetry is rather difficult to estimate, but it seems unlikely 

 that it would push the ball as much as 0.001 cm. toward the south. 



The suggestion having been made that the copper wire used for the 

 plumb-line might be sufficiently magnetic to suffer a slight deflection from 

 the vertical toward the direction of the magnetic dip, I made such experi- 

 ments as seemed likely to meet this suggestion. Per this purpose I 



