Examples of the converse case are also common, an instance of the recurrence 

 of special conditions at the same hour on succeeding days having already been given 

 iii Section 1. A better example is, however, furnished by the observations of May Kith 

 to May 21st. Between these dates, May 17th and 18th were overcast, while no 

 observations were made during the afternoon of the 20th. 



On the Ifith, a bright aurora was specially reported at 17.20, this time licit u 

 between the regular hours (17 and 18) of observation. Aurora was again specially 

 reported at 17.30 on the 19th and also at the same time on the 21st, no aurora bein- 

 seen on this occasion at the adjacent regular hours of observation. 



This tendency to recurrence at the same time on successive days is clearly an 

 indication of a pronounced daily variation in the sequence of auroral conditions at 

 Cape Evans, which is more clearly brought out in the following sections. That tin- 

 tendency to recurrence of conditions extends also to the form assumed by the amora. 

 might be inferred from the remarks in Section 1 regarding the general course of the 

 aurora during the day. 



It would be of some importance to ascertain if auroree show any tendency to recur 

 after a period of about 27 days, corresponding to the period of rotation of the sun. 

 For the purpose of this investigation, days have been divided into five classes days 

 on which auroral displays have been very brilliant, brilliant, fair, poor, or lacking in 

 display entirely, these classes being given numbers 4, 3, 2, 1 and respectively. 



The results of the analysis must be considered inconclusive, possibly owing to 

 the fewness of the observations. Taking all the days (n) allotted figures over 2, we 

 obtained the following mean auroral character numbers on days between (n-|-24) 

 and (<+32). 



Mean auroral character for days () = 3-17. 



(4-24) (n+25) (w+26) (+27) (w+28) ( 

 1-79 1-62 2-16 2-41 2-38 2-15 1-55 1-47 1-30 



From these numbers (shown also in fig. 2), it will be seen that there is almost as 

 much evidence for a 28-day period as for a period of 27 days a result which is by 

 no means surprising when we consider the effect on auroral observations of the presence 

 of moonlight.' The result is, however, sufficiently definite to merit further investigation. 



The rounded form of the curve is certainly to be expected if there is a direct 

 connection between the recurrence of aurora and the recurrence of sunspots after one 

 revolution of the sun, in view of the fact that the tune of revolution in the sun varies 

 with latitude. In addition, the rounded form must owe its shape in some degree to 

 the fact that disturbed auroral conditions show a tendency to continue sometimes for 

 several days at a time. How far this may be due to sunspot conditions cannot be 

 estimated. 



No attempt was made to obtain evidence of an annual variation in the number of 

 aurorse seen, in view of the impossibility of estimating the effect of the twilight arch 



