TABLE XXXI. 



Giving for the probable error of a single difference 



for M ; , 5 == 5-8X1Q- 7 sec. 

 M /; -7 =2-9 

 21 M,, = 3-8 



The mean of these (4 2 X 10~ 7 sec.) is less than half the corresponding figure obtained 

 from an analysis of the results of series " A." 



CAPE EVANS, SERIES C. 



JULY 13TH TO 16TH, 1912. 

 Height above sea-level 10 feet. 



In view of the disappointing discordance of the results of Series A and B, not only 

 as regards the individual observations of swing, but also as regards the means of the 

 different series, it was decided that a revolutionary change must be made in the method 

 of observation. 



Apart from the large errors in time determination which were due almost wholly 

 to irremediable defects in the transit instrument, it was clear that low temperature 

 was the cause of a great part of the difficulty. It was obvious, in the first place, that 

 the low temperature in the ice cave had a very unfavourable effect on the action of 

 the coincidence apparatus. Possibly greater, was the uncertainty due to deposition 

 of hoar-frost on the agate planes and on the pendulums themselves. 



It was decided, therefore, to build a small hut of petrol cases covered with rubberoid 

 and canvas, which was to be artificially heated during the observations. By March 17th, 

 the hut of heavy cases, still containing petrol and weighing some 150 Ib. each, was 

 entirely built except for the roof, the cases being tightly nailed together. No oppor- 

 tunity to put on the canvas roof occurred and, on the 20th, a blizzard levelled the 



45 



