read by bringing the
zero-point of the graduated scale to the level of the surface of the lower
limb by means of a screw, and reading off the height at once from the
surface of the upper limb. This barometer requires no correction for errors
of capillarity or capacity. Since, however, impurities are contracted by
the mercury in the lower limb, which is usually in open contact with the
air, the satisfactory working of the instrument comes soon to be seriously
interfered with.

[Illustration: FIG. 2. Cistern Barometer.]

Fig. 2 shows the _Cistern Barometer_ in its essential and simplest form.
This barometer is subject to two kinds of error, the one arising from
capillarity, and the other from changes in the level of the surface of the
cistern as the mercury rises and falls in the tube, the latter being
technically called the _error of capacity_. If a glass tube of small bore
be plunged into a vessel containing mercury, it will be observed that the
level of the mercury in the tube is not in the line of that of the mercury
in the vessel, but somewhat below it, and that the surface is convex. The
capillary depression is inversely proportional to the diameter of the tube.
In standard barometers, the tube is about an inch in diameter, and the
error due to capillarity is less than .001 of an inch. Since capillarity
depresses the height of the column, cistern barometers require an addition
to be made to the observed height, in order to give the true pressure, the
amount depending, of course, on the diameter of the tube.

The error of capacity arises in this way. The height of the barometer is
the perpendicular distance between the surface of the mercury in the
cistern and the upper surface of the mercurial column. Now, when the
barometer falls from 30 to 29 inches, an inch of mercury must flow out of
the tube and pass into the cistern, thus raising the cistern level; and, on
the other hand, when the barometer rises, mercury must flow out of the
cistern into the tube, thus lowering the level of the mercury in the
cistern. Since the scales of barometers are usually engraved on their brass
cases, which are fixed (and, consequently, the zero-point from which the
scale is graduated is also fixed), it follows that, from the incessant
changes in the level of the cistern, the readings would be sometimes too
high and sometimes too low, if no provision were made against this source
of error.

[Sidenote: Fortin's Barometer.]

A simpl