APPENDIX. 
515 
On ship-board, j 
five different vessels ; and the general results, so tar as they are neces- 
sary to the present explanation, were these. 
1st. At or near the binnacle, the north point of the compass was 
attracted forward in all the ships ; but the quantity of error produced, on 
one side when the head was East, and on the other when W est, varied from 
6i° to 0° 21'. 
2nd. When the compass was placed in other parts of the different 
ships, the attraction was sometimes forward and sometimes aft ; but always 
aft from the forecastle. The error at some of the stations was greater than 
at the binnacle, and at others less. 
3rd. The errors were least when the ship’s head was at, or near to 
North or South, and greatest at, or near to East or West ; and as the head 
was made to deviate from the points of least error towards the greatest, 
the increase of error was found to be in proportion to the sines of the angles 
of deviation. 
This last was the particular subject of my anxiety ; and being then 
satisfied that the law before deduced from analogy was certain, I employed 
it to find a standard correction for all my observations in the Investigator. 
For this purpose a selection of them was made where the ship s head was 
in the most opposite points and furthest from the meridian, and where the 
true variation could be ascertained within a small quantity ; the diffeience 
between the observed and true variations gave the errors, and when the 
head had not been at East or West, they were proportioned to eight 
points or radius by the sines of the angles. These observations were col- 
lected into tables, one for the north, and another for the southern mag- 
netic hemisphere, and classed according to the dips of the needle , and 
the error for eight points at each dip being reduced to parts of that dip, a 
medium of the whole was taken, and considered to be the standard radius 
applicable to all situations. The two tables arc here inserted for the satis- 
faction of naval and philosophical readers ; and no further explanation of 
them seems requisite, than that when the errors of observation were to the 
right they are marked -}-, and — when to the left . 
