318 CAPTAIN ELLIOT’S xMAGNETIC SURVEY OF THE INDIAN ARCHIPELAGO. 
being 2''7 per annum increase^ or 0'‘22 per mensem. This rate of 0'‘22 per mensem 
is assumed as the correction to be added to observations taken prior to January 1848 
to correct them to that epoch, and subtracted from observations taken subsequent to 
that date. This change at Madras is I believe strictly in accordance with the ob- 
servations of Captain Ludlow of the Madras Engineers, the talented director of the 
magnetic establishment at that station. As the south end of the needle dips at 
Singapore and the north end at Madras, and the inclination is increasing at both 
stations, there must be at some intermediate spot a point of contrary flexure. 
The method of observing on shore after the reception of the three new dip circles 
was never altered with those needles whose poles were changed ; the circle being always 
in the magnetic meridian ; two sets of readings were taken in each position of the 
needle, the usual eight positions being observed for the correction of instrumental 
error as follows : viz. needle direct and reversed for the non-coincidence of the mag- 
netic axis with the axis of form ; the circle east and the circle west for the correction 
of the zero of the vertical limb, and the same four positions with the poles changed 
to correct the error arising from the centre of gravity not coinciding exactly with the 
axle of the needle. 
With a view of ascertaining the value of the observations of the inclination, I drew 
up the following Table of the probable error of a single determination, and of the 
probable error of the mean value of the inclination at some of the principal points 
of the Eastern Archipelago. 
The first column contains the year and month ; the second, the name of the station ; 
the third, the number of needles ; the fourth, the number of dip circles observed ; 
the fifth, the number of observations ; the sixth, the mean value of the dip ; the 
seventh, the probable error of a single determination ; and the eighth, the probable 
error of the mean value. 
