57 



system, wliicli ought to have different temperatures in order that 

 the wliole may get the most probable division of temperature over 

 its parts (trillion tables, and upon each of them million balks). If it 

 is advanced against this that an inequality of this kind must 

 continually appear in precisely the same cubic centimetres, so that 

 our two portions of gas may still equalize their temperature, it must 

 not be forgotten that this demands that at the same moment another 

 arbitrary pair of cubic centimetres would be obliged to change 

 temperature in just the opposite direction. 



Further it must be remembered that in the case when the subdi- 

 vision is continued as far as the single molecules we do actually 

 take up the latter standpoint: the momentary kinetic energy accorded 

 to each separate molecule is in itself not the most probable; over 

 a sufficiently large number of molecules, however, the velocities are 

 divided in such a manner that we can only talk of the most probable 

 distribution for the whole of these molecules (quadrillion tables with 

 one ball on each, or, what comes to the same, one table with 

 quadrillion balls). 



