29 & 
on Magnetic Attraction. 
tuation s C n, ought to deviate by the same angle towards the 
ball, in the whole of its revolution round the line sCn. Hence 
it would follow, if the hypothesis were correct, that the angle of 
deviation of the horizontal needle, when referred to the angulai 
deviation of the line sCn, should give the same angle during 
the whole revolution of the ball round the line sCn. Thus, 
let N'MS' be in thep lane of 
the table, N'C S' being paral- 
lel to the meridian line; C the 
centre of the compass; sCn 
the direction of the dip; iEQ 
a circle perpendicular to sCn , 
and passing through the cen- 
tre of the needle ; l B a any 
other circle perpendicular to 
the line sCn^ and in which 
the ball is supposed to be car- 
ried round that line. The circle iEQ, that in which we have 
found the deviation to be nothing, being perpendicular to 
our magnetic axis, may be termed the magnetic equator; 
M l will then be the latitude of the ball, and II s JR the com- 
plement of its longitude, reckoning from the intersection of 
iEQ with the horizontal plane N'MS'. Suppose that S' CM, 
or the arc S' M is the angular deviation of the horizontal needle, 
when the ball is at the point B ; then the deviation of the par- 
ticles, in the line sCn, will be in the plane of the circle s B n 9 
and s<r may be considered the measure of that deviation, as 
causing the deviation S' M ; s <r ought, according to our theory, 
to be the same wherever the ball may be in the circle l B a. 
I resolved, therefore, to observe the deviations of the hori- 
zontal needle, caused by the ball when in different situations in 
the circle l B a, and, reducing the arcs S' M to arcs s <r, see how 
near they coincided with each other. As, however, the nature 
of the apparatus could not admit of the ball being carried round 
the compass, the compass was carried round the ball, in such a 
manner that the ball was always at the same perpendicular dis- 
tance from the same point in the line sC n. 
In the first set of experiments, I took CB 14 inches, and C p 
6 inches, and having observed the deviations at every 10° of 
longitude, I found the mean value of s <r computed from these 
