463 



tion, no absolute measures have been attempted. All 

 moveable magnetic substances were removed from the vici- 

 nity of the needle. A vrindow near the needle was carefully 

 closed by a shutter of wood, and by two curtains fastened 

 to the window frame, and with an interval between them. 

 There is no fire in the room ; and a double door is between 

 the observing room and an adjoining one where there is a 

 fire. 



*' The instrument is one of Gauss's dechnation magne- 

 tometers, made by Meyerstein of Gottingen. The arrange- 

 ment of it agrees exactly with that described in the Resultate, 

 which has been followed as nearly as possible. The reading 

 telescope is supported upon a small wooden shelf fastened 

 to one of the side walls of the room ; the scale is attached 

 to a wooden frame before the shelf. One of the smallest 

 divisions of the scale, which is divided by estimation in the 

 observations to tenths, is 25''-975 nearly in value. The 

 zero of the scale did not vary sensibly in position during the 

 observations. 



" The observations were made every eight seconds, an 

 assistant striking two seconds before the time of each obser- 

 vation. The ticks of the half-seconds chronometer being 

 distinctly audible, the observation was made at the fourth 

 beat after the signal given by the assistant, and thus the 

 time was independent of the minute accuracy of the signal. 

 Checks were adopted to prevent or detect large errors in 

 giving the signal. The interval of eight seconds is very 

 nearly one-third of the time of oscillation of the magnet bar. 



" The time was observed by a chronometer beating half 

 seconds. This was compared before and after each set of 

 magnetic observations with one, and after the morning of the 

 I4th with two chronometers. One of these was carried from 

 the Girard College to the city after each set of observations ; 

 but the others remained during the night at the Girard 

 College, and were removed to the city after the morning 



