442 FIRST REPORT UPON MAGNETIC WORK

about the same hour angle as already stated, or to observe the latitude
by means of circum-meridian altitudes of the sun. The former
method would very often have necessitated the loss of an additional
station, and the second method involved additional observation and
computation and often considerable annoyance, as other work would
have to be abandoned for the sake of the latitude observation. The
latitude error is besides a temporary error which can be allowed for
when more correct positions have been obtained. I therefore decided
that generally I would not attempt to eliminate the latitude error at
any one station, but endeavor to arrange the work so that at some of
the stations the azimuth observations would be made in the morning
and some in the afternoon. Some of the errors to be ascribed to
defective latitudes would be plus and some would be minus. I be-
lieve I can safely trust that an isogonic line which, in a certain sense,
represents an adjustment of all the observations on either side of it,
will not be in error by reason of imperfect latitudes to an amount
greater than l', if as much as that. The error due to direct interpola-
tion between two stations will far exceed any error that may be due
to defective latitudes of the station. It was my endeavor, therefore,
to reduce the interpolation error rather than the latitude error. And
the only way to reduce the interpolation error is by increasing the
number of points of observation, or what amounts to the same thing,
by decreasing the distance between the stations.

We next come to the observing error of the purely magnetic part
of the determination of the magnetic declination. Here we must
remember in the first place that we are not observing a fixed object,
but one in ceaseless motion. All we can aim at is an average position
corresponding to a definite moment of time. At some other moment
the average position assumed by the needle will be a different one. If
the interval of time over which the observations are made is too long,
then will the arithmetical mean of the different positions assumed not
correspond to the arithmetical mean of the times of observation.
For an interval of about 15-30 minutes this will generally, however,
be the case, except of course in times of magnetic storms. This was
about the interval used in this work. A reading of position of needle