ome heat is lost when the
calorimeter is raised above the temperature of its enclosure, and
before the final temperature is reached. This can be roughly estimated
by observing the rate of change of temperature before and after the
experiment, and assuming that the loss of heat is directly
proportional to the duration of the experiment and to the average
excess of temperature. It can be minimized by making the mixing as
rapid as possible, and by using a large calorimeter, so that the
excess of temperature is always small. The latter method was generally
adopted by J.P. Joule, but the rise of temperature is then difficult
to measure with accuracy, since it is necessarily reduced in nearly
the same proportion as the correction. There is, however, the
advantage that the correction is rendered much less uncertain by this
procedure, since the assumption that the loss of heat is proportional
to the temperature-excess is only true for small differences of
temperature. Rumford proposed to eliminate this correction by starting
with the initial temperature of the calorimeter as much below that of
its enclosure as the final temperature was expected to be above the
same limit. This method has been very generally recommended, but it is
really bad, because, although it diminishes the absolute magnitude of
the correction, it greatly increases the uncertainty of it and
therefore the probable error of the result. The coefficient of heating
of a calorimeter when it is below the temperature of its surroundings
is seldom, if ever, the same as the coefficient of cooling at the
higher temperature, since the convection currents, which do most of
the heating or cooling, are rarely symmetrical in the two cases, and
moreover, the duration of the two stages is seldom the same. In any
case, it is desirable to diminish the loss of heat as much as possible
by polishing the exterior of the calorimeter to diminish radiation,
and by suspending it by non-conducting supports, inside a polished
case, to protect it from draughts. It is also very important to keep
the surrounding conditions as constant as possible throughout the
experiment. This may be secured by using a large water-bath to
surround the apparatus, but in experiments of long duration it is
necessary to use an accurate temperature regulator. The method of
lagging the calorimeter with cotton-wool or other non-conductors,
whi