34 Art. 9. -T. Terada : 



that the secondaiy maximum of frequency at about 90' corresponds 

 to the most frequent waves observed during night hours. 



No conspicuous group of waves with about 10' period, 

 observed by Birkeland, could be confirmed in our records, though 

 such waves might have been detected if the amplitudes were 

 somewhat larger than 0'2 y. 



10. Fluctuation of frequency of regular waves in successive days. 

 The number of hours (not that of trains) disturbed by regular 

 waves was counted for each of the successive days and plotted in a 

 diagram with the days as abscissa. The results showed apparently 

 very irregular fluctuations, either if we take all waves with the 

 periods 30^-200^ or only 30^-70' waves. Though we must be 

 very cautious in deducing any periodicity from such material, the 

 results show nevertheless a remarkable tendency to suggest the 

 existence of a period, or rather of an "interval" of about 25-30 

 days. The most conspicuous examples are illustrated by the 

 interval from October, 1912, to April, 1913, especially when the 

 waves with the periods 30^-70^ alone are taken into account 

 (Fig. 10, A). Fig. 10, B and C were obtained by superposing the 

 successive series of 14 or 28 days^^ respectively. These graphs 

 seem to suggest the existence more of a 14 days period than a 28 

 days one, the amplitude amounting to the same order of magnitude 

 as the mean vahie. The last procedure will however, have a 

 definite meaning only when the phenomena are simply and purely 

 periodic, but not when the frequency maxima are repeated with 

 conspicuous amplitudes only a small number of times in a definite 

 period, and then replaced by anotlier of a similar kind, but setting 

 in more or less abruptly with quite accidental time relation to the 

 former. The general aspect of the frequency diagram here in 

 question strongly suggests that the nature of the phenomena 

 concerned is of the type above mentioned. In the above 

 example, we ma}^ probably assume the accidental superposition 

 of two independent serijs witli the longer period, say 28 days. 

 The lunar period is to be excluded, since the above fluctuation 



1) 28 days instead of 27 days was taken simply because it is an even number, no weight 

 being laid here on the exactitude of the period. 



