354 
DR. P. E. SHAW ON THE NEWTONIAN CONSTANT OF 
is only a l/l6 chance that this accordance is mere coincidence. Yet the evidence is 
slender in view of the uncertainties of such delicate measurements which may involve 
spurious effects of this order of magnitude. 
In the case of the results in Boys’ research we have data quite definite, though small, 
as to temperature, and they are outside the range of experimental error. On the whole 
we are led to expect that in a full gravitative experiment where both M and m are 
involved we should find a plus temperature coefficient of the order l/lOOO for 1° C. 
After §§ 3 to 5 (pp. 350 and 351) we see that this result need not be at variance 
with the numerical figures obtained by Poynting and Phillips. 
In this subject there are three classes of work, the results from which should at 
present be kept separate, viz. :—(1) Change in temperature of both M and m ( indirectly 
by Boys, Baily, yon Sterneck and other pit experiments) ; (2) change in 
temperature of M only ( directly by the present research, indirectly by Mendenhall) ; 
(3) change in temperature of m only ( directly by Poynting and Phillips). 
There is only one class of experiment with which the present experiments can be 
brought into direct comparison, viz., the Schehallien type, and for this class we have 
no assigned temperatures. 
III. Methods Employed. 
1. There are left two possible lines of research with a full gravitative apparatus 
on the influence of temperature on attraction. 
(l) Changing temperature for M and m and the intervening medium ; (2) changing 
temperature of M alone. 
I commenced with method (l), but abandoned it after much labour, the difficulties 
appearing insuperable. A decisive result obtained for the method (2) should go far 
to settle the whole issue for the temperature range involved. 
As to the actual type of apparatus used there are three standard forms (a) the 
torsion balance ; ( b ) the weight balance ; (c) the pendulum. 
The torsion balance was chosen as combining great sensitiveness with accuracy. 
Under favourable conditions the results agree by this method to l/l0,000 so that it 
affords unparalleled refinement; but there are two great troubles attending this 
extreme delicacy, viz. :— 
(l) The law, for maximum load on the fibre short of breaking, is 
sensitiveness cc l/(diam. of fibre) 2 , 
since carrying power varies as (diam.) 2 , and stiffness to torsion varies as l/(diam.) 4 . 
Thus for attainment of the best results one employs a small factor of safety in 
loading the fibre and breakages easily occur. Quartz fibres (as used chiefly here) 
often stand a load for many hours and then break. Hence all fibres before use must 
be subjected to careful load-time tests. Even after taking this trouble a small shock 
during installation may cause breakage. 
