36 Prof. Everett on Normal Piling as 
The same assumptions, as to the grains, which give these 
results, give also the correct velocity of light, and the correct 
value of the gravitation constant. 
Prof. Reynolds nowhere hints at the existence or possibility 
of closest piling other than the normal kind. I do not 
think I am going too far in saying that he identifies a 
tendency to closest packing with a tendency to that particular 
kind of closest packing which he names “normal piling.” 
It is necessary to face the question, whether this ignoring 
of other modes of closest piling vitiates the arguments by 
which the theory is supported. 
It is legitimate, in framing a hypothesis as to the structure 
of the universe, to assume the particular structure which best 
answers the purpose of yielding an explanation ; and where 
there is a choice between several hypotheses, the simplest is the 
best, provided it proves sufficient. Of all the possible modes 
of closest piling, normal piling is comparably the simplest. 
The theory does not expressly assert that the grains, coming 
together fortuitously, have settled down into. this simple 
arrangement ; but nevertheless the question does occur to 
un inquiring mind whether fortuitous clashing might be 
expected to lead, in the long run, to such a result. It is 
near akin to the question whether equal prains of shot, 
shaken together in a, bag, will arrange themselves in normal 
piling. Prof. Rey nolds gives experiments showing that shot so 
treated do arrange themselves in closest packing, “but adduces 
no evidence as to the particular kind of closest packing that 
they assume. 
Looking at the matter from a theoretical point of view, 
we have, on the one hand, the fact that, in the casual laying 
of 3 successive tiers in closest order, it is an even chance 
whether the piling is normal. In the casual oyine of 4 tiers, 
the chance is only 1 in 4; for 5 tiers it is 1 in 8, and for 
2 1m tiers ib 1s 1 in’ 2”, 
On the other hand, we have the fact that a normally piled 
cluster has 6 sets of parallel lines of balls, and these will 
form efficient battering-rams in collisions with other clusters 
which can only have 3. 
Tf, instead of collision, we suppose a statical push, as in a 
football maul, there is still the same advantage. Balls 
ranged in parallel lines, pushing end on, will be better able 
to retain their formation than balls with only oblique support 
behind them. 
Again, as has been pointed out by Mr. Barlow * in a 
different connexion, if we suppose that, in the first instance, 
* W. Barlow, “A Mechanical Cause of eeu ppncity , &c.” Sei. Proc. 
Roy. Dublin Soe, vol. viii. pt. vi. (1897) p. 535, 
