Escape of Gases from Atmospheres. 697 
The cases in which Maxwell’s approximate law may 
legitimately be employed can be pointed out. Whenever an 
approximate law presents itself in an exponential form with 
negative index, the approximation holds good as an approxi- 
mation over that small part of the range where the exponential 
function acquires large values, but can no longer be depended 
upon as an appr oximation in regard to the parts of the range 
where the exponential function is small. Maxwell makes a 
legitimate use of his law when, through its instrumentality, 
he discovered his remarkable explanation of viscosity and 
diffusion, and investigated the laws of those phenomena. In 
reference to these, what happens in the case of velocities 
which are infrequent is of small account; but the appli- 
cation made by Prof. Bryan and Mr. Cook is to the rare 
events which occur within that part of the range where the 
approximation breaks down and where, in consequence, the 
exponential Jaw is misleading. It is this ove ersight to which 
I think it likely that we are mainly to refer numerical 
results which are found to clash with events that have taken 
place or that are taking place upon the moon and the earth. 
The inquiry in which I engaged in the sixties of the last 
century led also to the detection of other defects in the 
premisses made use of by those who have trusted in the 
deductive method. One of these concerns the ambiguities 
whick surround the use of the term “temperature.” Tem- 
perature is not one physical measurement but two groups 
of physical measurements, essentially different according as 
we test equality of temperature by there being no transfer 
of heat by conduction when two bodies are brought into 
contact, or by radiation when they are made to stand apart. 
This establishes a division of temperatures into two principal 
groups, and these groups require further subdivision. The 
temperature of a body determined in these two different ways 
may be called its conduction temperature and its radiation 
temperature; of each of which there are several varieties. 
There are accordingly many different kinds of temperature. 
In the case of gases, conduction (including convection) is 
mainly concerned with the translational speeds of the mole- 
cules, while radiation in the first instance affects only the 
internal events going on within the molecules. In most 
laboratory experiments (carried on as they must. be at the. 
bottom of our atmosphere) the partition of energy between 
the internal events of each molecule and its translational 
movements takes place so frequently—probably several times 
every second in a gas at standard temperature and pressure— 
that’ the distinction even between the two main kinds of 
