TRANSACTIONS OF SECTION A. 547 



circle, remained approximately constant, but after each overhauling it suddenly 

 altered, increasing by 12' on the first occasion, by 23' on the second, and by 20' 

 on the third. Mr. Chambers states that he ' can give no satisfactory account of 

 this behaviour of the instrument,' but suggests that 'the needle gradually hollows 

 out a depression in the agate plates on vfhich it rolls, and that this characteristic 

 of the dip circle ' has not before been discovered owin? to the reluctance of 

 magnetic observers to interfere vyith the adjustments of instruments v?hich are 

 apparently working well. 



I do not think that this explanation will suffice. Dr. Thorpe and I employed a 

 new dip circle in the earliest part of our survey work, which has remained in accord 

 with Kew for ten years. During that time the dip has been measured some 700 times 

 with it. This corresponds, I believe, to more than the amount of work done with the 

 circle at Colaba in six years, which in turn is longer than some of the intervals in 

 which the Colaba instruments gave results erroneous to the extent of 20'. I feel, 

 therefore, quite sure that the difficulties which have been experienced at Bombay 

 are not due to any ' characteristic [defect] of the dip circle.' But, whatever the 

 cause may have been, surely the lesson is that, if such things can happen in so well- 

 known an institution, it is desirable that we should take the moderate pains 

 required to assure ourselves whether smaller — but, possibly, not unimportant — 

 errors are gradually atiecting the results at any of our observatories. 



This brings me to my next point, namely, that if we are to draw conclusions 

 from the minor differences between measurements of secular or diurnal change made 

 in the observatories, it is not only necessary that we should know whether the 

 instruments are strictly comparable and constant, but the observations must be 

 reduced by precisely the same methods. 



In 188G the late Mr. Whipple drew the attention of the British Association to 

 the fact that there was a systematic difi'erence between the diurnal ranges of 

 declination at Greenwich and Kew. His results were based on the three years 

 1870-72. In 1890 two of my students, Messrs. Robson and S. W. J. Smith, ex- 

 tended the comparison to three more recent years (1883-6-7), and obtained results 

 in complete accord with those of Mr. Whipple. 



It is well known that the average daily oscillation of the magnet is affected by 

 the magnetic weather. Sabine showed that magnetic storms do not merely buffet 

 the needle now in this direction and now in that — they affect its average behaviour, 

 so that the mean swing east and west is different according as we deduce it only 

 from days of magnetic calm or include tho.se of storm. 



Mr. Whipple reduced the Kew observations by two methods,' one of which 

 depended on the calmest days only, while the other included those which were 

 moderately disturbed. Neither agreed exactly with the method in use at Green- 

 wich, but the difference between the results deduced from them was so small when 

 compared with the difference between either and that obtained at Greenwich, that 

 it seemed possible that the diurnal variations, even at these closely neighbouring 

 places, might differ appreciably. The question whether this is so has now been 

 answered. In 1890, at the request of the Kew Committee, the Astronomer Royal 

 undertook to select early in each year five quiet days in each of the preceding 

 twelve months. It was also agreed that, whether they adopted other methods or 

 not, the chief English magnetic observatories should determine the diurnal varia- 

 tions from these days alone. The Greenwich^ and Kew observations for 1890 have 

 therefore been worked up in exactly the same way, with the result that the dis- 

 crepancy, which had persisted for twenty years, has entirely disappeared, and 

 that the two diurnal ranges at the two observatories are in as close accord as 

 could bo expected. 



If, therefore, we may judge from a single year, the cause of the difference lay in 

 the choice of days. Greenwich will in future gi%'e us two diurnal variations — one 

 obtained from the most quiet days only, the other from all days except those 

 of violent storm, and in these we shall have most valuable data for studying the 

 mean effect of distuibances on the diurnal variation. 



' Sabine's and Wild's. 



- The Greenwich observations for subsequent years have not yet been published. 



N N 2 



