The ice of the interior cavity always retains a certain quantity of water adhering to its surface, which may be supposed to belong to the result of the experiment; but as, at the beginning of each experiment, the ice is already saturated with as much water as it can contain, if any of the water produced by the caloric should remain attached to the ice, it is evident, that very nearly an equal quantity of what adhered to it before the experiment must have run down into the vessel F in its stead; for the inner surface of the ice in the middle cavity is very little changed during the experiment. By any contrivance that could be devised, we could not prevent the access of the external air into the interior cavity when the atmosphere was 9 deg. or 10 deg. (52 deg. or 54 deg.) above zero. The air confined in the cavity being in that case specifically heavier than the external air, escapes downwards through the pipe x y, Fig. 3, and is replaced by the warmer external air, which, giving out its caloric to the ice, becomes heavier, and sinks in its turn; thus a current of air is formed through the machine, which is the more rapid in proportion as the external air exceeds the internal in temperature. This current of warm air must melt a part of the ice, and injure the accuracy of the experiment: We may, in a great degree, guard against this source of error by keeping the stop-cock u continually shut; but it is better to operate only when the temperature of the external air does not exceed 3 deg., or at most 4 deg., (39 deg. to 41 deg.); for we have observed, that, in this case, the melting of the interior ice by the atmospheric air is perfectly insensible; so that we may answer for the accuracy of our experiments upon the specific heat of bodies to a fortieth part. We have caused make two of the above described machines; one, which is intended for such experiments as do not require the interior air to be renewed, is precisely formed according to the description here given; the other, which answers for experiments upon combustion, respiration, &c. in which fresh quantities of air are indispensibly necessary, differs from the former in having two small tubes in the two lids, by which a current of atmospheric air may be blown into the interior cavity of the machine.