The mode of construction is illustrated by Figs. 21 and 22. It consists in hermetically sealing a glass tube about three feet long and filling it with mercury. The finger is placed over the open end of the tube, which is then inverted and placed in a cistern of mercury and the finger withdrawn. The left-hand figure shows the result; the mercury is seen to fall some three or four inches, leaving an empty space at the top of the tube, which is called the “Torricellian vacuum.” The mercury is prevented from falling lower than is shown, by the external pressure of the atmosphere on the cistern. The _weight_ of this column, therefore, represents the _weight_ or pressure of a corresponding column of air many miles in height; and so close is the relation between the column of mercury and the external air that the _height_ of the former changes with the slightest variation in the _weight_ of the latter, and the instrument thus becomes a measure of the weight of the air, from which property its name is derived, the Greek words _baros_ and _metron_ signifying respectively “weight” and “measure.” When the mercury in the barometer tube falls, that in the cistern rises in corresponding proportion, and _vice versa_, so that there is an ever-varying relation between the _level_ of the mercury in the tube and the mercury in the cistern, which affects the accuracy of the readings. In M. Fortin’s cistern this difficulty is obviated by the use of a glass, with flexible leather bottom and a brass adjusting screw, as shown in the cut. Through the top of the cistern is inserted a small ivory point, the lower end of which corresponds with the zero of the scale; and, to secure uniformity, the level of the mercury in the cistern should be adjusted by the screw at each observation, until the ivory point _appears_ to touch its own reflection on the surface. The reading is then taken. [Illustration: 23. Fortin’s Cistern. Scale about 1/6.] In making barometric observations for comparison with others, it is necessary that all should be reduced to the common temperature of 32° F., and for this purpose tables have been calculated which will be found to save much time. Tables also for reducing observations of the barometer to sea level, an operation equally indispensable with the other corrections to make the readings intercomparable, have been published by direction of the Meteorological Committee.