530 MR JOHN ALLAN BROUN ON THE HORIZONTAL FORCE 
unifilar magnet deflected by two bar magnets, so as to make a large angle with 
the meridian. As the suspension thread is of silk cocoon fibre, and the possible 
error from the smallest change in its structure from continued tension consider- 
able, the directive force being small, the confirmation of epochs is in this case 
the more marked. 
3d, Cape of Good Hope, 1842-45, uncorrected for secular change. 
Maxima in December and June. 
Minima in September and April. 
Our examination of the monthly means will have shown how impossible it is in 
this case to make any correction for the secular change: the coincidence in the 
epochs derived from these means seems purely accidental. 
4th, St Helena, 1842-45, uncorrected for secular change. 
Maximum in J une, 
Minimum in April. 
The same difficulty occurs in this case for the secular correction as for the Cape, 
though to a less extent: the existence of the minimum in September and the 
maximum in December is barely indicated by a diminution of the rate of decrease 
of force after July, as will be seen in the dotted curve below that for St Helena 
(Plate XXIV.), which shows the change from one monthly value to the next fol- 
lowing. No satisfactory result can be obtained for the annual law where the 
secular variation changes its rate and sign during the period considered. 
5th, Singapore. 
Maxima in January and June. 
Minima in April and September. 
6th, Trevandrum. 
Maxima in December and June. 
Minima in April and August. 
40. The maximum in June and minimum in September are less marked at 
Singapore and Trevandrum than at Makerstoun, Hobarton, and Munich, or than is 
indicated generally at the Cape and St Helena. Though this may be due, to some 
extent, to instrumental causes, it is probably partly a consequence of the less 
effect of disturbances on the mean force near the equator. 
41, The double epochs may be considered to exist in all the series in spite of 
the various errors or movements that cannot easily be eliminated.* That the 
variations follow the same law everywhere, independently of the secular change, 
may be best shown by an example. The year 1845 exhibits considerable differ- 
ence in the character of the monthly variations (see Plate X XITI.); let us then 
consider the differences of the values of each month with that following it, in 
the year 1845; these differences are projected in Plate XXIII. It will be seen 
* The means of the Toronto Observations, July 1846 to June 1848, corrected for secular 
change at the rate of — 1°33 per mensem, are projected in Plate XXIV. 

