CONVERTIBILITY OF FREE HEAT. 85 



atmospheric pressure ; and this being the only work done 

 by the escaping air would be the equivalent of the only heat 

 abstracted from the receiver and the pipe, other than that 

 which passed to and from the surrounding water. Three 

 sets of experiments, with this contrivance, gave values for the 

 equivalent of 820, 814, and 760, with a mean 798, which is 

 only 3*3 per cent above the final value, although the actual 

 differences of temperature measured did not exceed o*i 

 degree Fah. Neither are the discrepancies, which do not 

 amount to 5 per cent., nor yet the slight excess of the mean, 

 to be attributed solely to errors of thermometric measure- 

 ment ; for it was necessary to stir the water to equalize the 

 temperature before reading this temperature, and the work 

 done in this stirring exercised such a sensible effect on the 

 temperature that this effect had to be determined and 

 deducted ; affording a source of error. 



These determinations, by the first and last series of 

 ■experiments, of the relations between the heat and work 

 involved in the expansion and contraction of air would not, 

 in themselves, have afforded conclusive evidence of the 

 convertibility of heat into work like that deduced from 

 Joule's magnetic or friction experiments ; for these experi- 

 ments still left it possible to suppose that the heat evolved 

 or absorbed was merely transformed from latent to sensible 

 and vice versa from sensible to latent, in consequence of 

 the changes in the density of the gas. That Joule was 

 fully aware of this was shown by his undertaking the 

 second series. In this the air underwent exactly the same 

 final change as in his third series, but without doing any 

 external work, in which case if there was any change from 

 sensible heat to latent it would still have appeared. It was 

 only on finding in this second series that there was no 



