GRAVITATION AS AFFECTED BY TEMPERATURE. 
383 
The process in the three differs in important respects and the conditions differ as 
to («) vacuum pressure; ( b) sensitiveness; and (c) lagging. 
£ 
£ 
c 
liJ 
o 
2 
< 
CC 
20 ° 40° 60 c 80° 100° 120° 140° 160“ 180° 200° 220° 240° 
T £ M P. ( Centigrade). 
Fig. 14. Graph for Table III. The first three crosses show rising temperatures, the nest three are for 
temperatures high but variable, and the two circles are for falling temperatures. 
Out of a total of many scores of experiments eighteen are summarised in Table IV. 
A great number of experiments have failed because the conditions were unsuitable ; 
for instance, the apparatus may be changed in the attempt to gain accuracy 
or to apply some test—but this change produces, sometimes, results which are 
non-cyclic. Or, a new vacuum pressure may be tried, which is unsuitable, the 
torsion system getting out of control through convection or tremors. In a few cases 
small negative temperature effects have been found, but in every case the results have 
been non-cyclic ; for instance, a negative effect when temperature rises may be 
followed by a positive effect when it falls again. The net result for both rise and fall 
is found invariably to be a positive effect. But it is non-cyclic and is considered 
worthless. These negative effects are due to strain and change in position of the 
parts of the apparatus and have always been removed by proper care. 
In Table IV. the high value +I'8xl0~ 5 and the low value 0'5xl0 _o occur and 
are included in finding the arithmetic mean, which might be considered to be 
vitiated by their presence. But it is noteworthy that the longest and best 
experiments give results about +l'2x 10“ 5 which agree with the mean result and so 
tend to confirm it. 
3 f 2 
