MARYLAND GEOLOGICAL SURVEY 439

tions were favorable. Only in two or three cases the stations had
to be decided upon—if observations were to be made at all—under
conditions not the very best. In these cases, however, an additional
station was occupied in the same county when possible, or somewhere
in the vicinity. The descriptions and locations of the stations are
given in another chapter.

METHODS OF OBSERVATION.

The highest refinement possible at any one station was not sought.
It was believed, for example, that it is far more valuable to obtain
two declinations to within a few minutes at two stations some distance
from each other than to observe one declination at one station to the
nearest minute. It is absolute folly, as far as the matter of distribu-
tion of declination is concerned, to occupy a station two or three days
with the view of determining the declination to the nearest minute
and then not observe the decimation again to within 25 to 50 miles
or more of this first station. Even in undisturbed regions the error
made by a linear interpolation between the values at two distant
stations is generally much greater than the error of the station obser-
vation. In disturbed regions the extremely refined methods, when
pursued at the expense of limiting the number of stations, are entirely
out of place. In the establishment of secular variation stations the
utmost refinement should of course be employed.

At the same time it was the aim to arrange the observations at two
distant stations in such a way that, while they individually might be
in error by several minutes, they would not necessarily both be in
error in the same direction, so that in making an adjustment, graph-
ical or otherwise, of all the observations, these station errors would
be in the nature of " accidental errors, " i. e., some would be plus and
others minus. If it had been possible to carry out this scheme per-
fectly, the isogonics drawn with a free hand, for example, as based
on observations defective in this way, might be just as accurate, or
nearly so, as those based on observations made with the utmost refine-
ment at an equal number of stations, and would be more accurate than
those based on refined observations at the number of stations which