INSTRUMENTS AND RECORDS. 77 



during these months must have been much larger than I had supposed, and I decided that some means 

 must be devised for eliminating its effects. The absolute sign of the Vertical-Force changes during 1903 

 was, as already explained, in doubt, but there was no ambiguity either in 1902 or 1903 as to the direction 

 in which rise of temperature deflected the Vertical-Force trace. It was thus apparent that if one could 

 determine the change of Vertical-Force ordinate answering to a rise of 1 in temperature, one could from 

 the trace of the metallic thermometer eliminate the effects of temperature, whatever their sign might 

 prove to be. 



Thanks to the large irregular changes of the Antarctic temperature, the task proved simpler than 

 anticipated. The method adopted was one which I have found to work successfully in several cases. The 

 principle is that if one can get a number of instances in which there is a large change of temperature in 

 the course of 24 hours, a coefficient calculated by assigning the 24-hour apparent change of Vertical Force 

 to temperature alone will not be much in error. There are, of course, individual occasions when the true 

 values of Vertical Force at the same hour on two successive days differ considerably ; but still, if one is 

 dealing with 20 or 30 days on which no specially large magnetic disturbance has occurred, the effects of 

 natural magnetic changes will in most instances be very nearly eliminated. In most months there were 

 fairly copious temperature data from the trace given by the thermometer inside the Vertical-Force box, 

 standardised by reference to the readings of the mercury thermometer. Numerical values having been 

 obtained for the teriiperature coefficient, hourly measurements were made of the temperature trace, and 

 corrections were thus obtained to the diurnal inequalities already calculated for December, 1902, and 

 February, 1903. Considerably to my surprise, the result was not merely a large reduction of the range in 

 each case, but a complete inversion of the inequality for February. 



9. A difference of sign between the true inequalities of Vertical Force for December and February 

 appearing highly improbable, I came to the conclusion that the hypothesis that the magnet had been 

 changed end for end on December 31, 1902, must be wrong, and that there must in reality have been 

 a change of sign in the temperature coefficient. To obtain further light on the subject, two direct 

 determinations were made at Kew of the temperature coefficient, the soft-iron bar being on one occasion in 

 the position it occupied during 1902, on the other occasion in the position it occupied during 1903. The 

 coefficients obtained differed notably in size, but they agreed in sign. At first sight this was rather 

 staggering, but surmising the true explanation of the phenomenon, I had a further determination made 

 under conditions as similar as possible to those at Winter Quarters. With the aid of a number of bar 

 magnets it proved possible to produce at the position of the Vertical-Force magnet a fairly uniform vertical 

 field of about - '72, the natural field at Kew being about -I- '44, and now, much to my relief, the expected 

 difference in sign appeared. With the soft-iron bar below the Vertical-Force magnet, as in 1902, the trace 

 went down the sheet as temperature was raised ; with the soft iron on the top of the magnet box the reverse 

 happened. The experiments were made with both rising and falling temperatures, the readings being 

 taken by Mr. T. W. BAKER and Mr. G. W. WALKER quite independently of me, and the results appeared 

 quite decisive in favour of the view that the Vertical-Force magnet was not altered in position in December, 

 1902 movement up the sheet meaning increase of force throughout but that the temperature coefficient 

 was negative in 1902, and positive in 1903. This view was thus finally accepted. If any additional 

 evidence in its favour is thought necessary, it will be found in the general consistency of the results for 

 the diurnal inequalities in Table XVIII. In several individual months the corrections from temperature 

 to the mean diurnal inequality exert but a trifling effect on its nature, and it is out of the question that 

 the diurnal inequalities in corresponding months of 1902 and 1903 should be the antitheses of one 

 another. 



