In correcting the reading of a barometer to the freezing-point, it is customary to ascertain the temperature of the instrument by observing a thermometer attached to the brass frame, but it is doubtful how far this attached thermometer indicates the temperature of all parts of the barometer under the conditions in which most observations are made. In a Fortin barometer the level of the mercury in the cistern is adjusted to a fixed point at each observation, and we have only to consider the expansion with heat of the mercury and of the brass scale.