160 Dr. C. Cliree : Auroral and 



variations from the two periods also differ, and in the same 

 direction as before. From May to September the relative 

 frequency is decidedly less, and at midwinter considerably 

 greater, for the later period than for the earlier, and the 

 equinoctial maxima are but indistinctly shown in the later 

 period. As in the case of Scandinavia, northern stations mav 

 have contributed more to the means for 1827-59 than to 

 those for 1761-93. It is clear, however, from the substantial 

 frequencies in the summer months, that even in the period 

 1827-59 data from temperate latitudes must largely have 

 prevailed. Thus the closer approach to the Arctic or single 

 maximum type is difficult to wholly explain, unless we admit 

 that it is partly a real phenomenon, representing a real 

 difference in the distribution of auroras throughout the average 

 year of the two periods 1761-93 and 1827-59. As regards 

 the intermediate period 1791-1826, we see that its mean 

 annual variation differed from that of the neiohbourino- 

 periods in the same direction as it did in the case of Scandi- 

 navia . The fall in the summer frequencies is very pronounced , 

 even as compared to the period 1827-59 alone. Of the 

 equinoctial maxima, that in March is much enhanced, but 

 that in October seems to have vanished, unless it is represented 

 by the maximum now shown in December, 



The average year from the two periods 1761-93 and 

 1827-59 combined shows fully five times the auroral frequency 

 of the average year of the period 1794-1826. This is inter- 

 mediate between the ratios deduced for the Scandinavian 

 districts IY. and V. 



§ 10. Table V. shows the annual variation of auroral 

 frequency in several pairs of groups of years characterized 

 respectively by many and by few sun-spots. The first line 

 gives results for the whole of Scandinavia from 30 years, 

 made up of the 3 years of largest sun-spot frequency from 

 each of the ten 11-year cycles covered by Tromholt's table ; 

 the second line gives the results for the corresponding 

 30 years of few sun-spots. Any gradual change in the nature 

 of the observations should affect the two sets of results nearly 

 equally. The third and fourth lines give corresponding 

 data calculated from Lovering's general catalogue, also for 

 30 years of many and 30 years of few sun-spots. 



Of the ten 11-year cycles employed in the two cases nine 

 are identical ; the Lovering data cover one cycle prior to the 

 common nine, the Scandinavian one later. This was un- 

 fortunately necessary owing to the dates when the Lovering 

 table ended and the complete Scandinavian table commenced. 

 The two sets of data combined represent eleven consecutive 



