150 VAPOR NUCLEI AND IONS. 



tude. There being but two observations, as a rule, for the day, the nature 

 of the fluctuations is not referable to periods; and indeed there is as 

 liable to be a rise as a fall of values during the middle hours of the day. 

 Towards the end of the month and in the beginning of September there 

 is an absence of agreement in the march of positive and negative ioni- 

 zations. Frequently the variation of one ionization is apt to lag behind 

 the other. 



After the 7 th of September the positive and negative variations 

 tend to take the same sign again, but the agreement in the course of a 

 month is less marked than before. 



In October the earlier observations are as a rule in the same phase until 

 October 10, where the first of a series of anomalies occurs, to be specially 

 considered later (section 94). While the data throughout the remaining 

 part of October are regular, there are similarly displaced variations 

 towards the end, which run quite into the next month. 



During November similarity of variation of the positive and negative 

 ionization may still be recognized, but in December the divergence of 

 data is so marked that it is not possible to coordinate them; and the same 

 discrepancy shows itself in January, both as regards the signs of vari- 

 ations and their absolute values. One may note, moreover, that the 

 positive curve (the observations for which were first taken) is more 

 irregular in its march and fluctuates between relatively enormous values. 

 Though there is some agreement in phase between January 18 and 25, 

 the anomalies increase again at the close of the month. 



The attempt was therefore made in February (section 94) to account 

 for and remove these discrepancies, and though this was but partially 

 successful, the positive and negative results during the remainder of the 

 season again return to an unmistakable agreement in character. There 

 is, moreover, a curious parallelism between the general march of the 

 nucleation curves and the ionization after February 15 as far as March. 

 In the latter month the positive and negative ionizations, though at first 

 fluctuating and uncertain, are finally in very close agreement. 



From what has been stated it appears that the positive results during 

 December and January are liable to be untrustworthy. The negative 

 results, which were taken after the positive, show less irregular fluctua- 

 tion and are in a measure acceptable throughout the eight months of 

 observation. 



94. Errors of measurement. The abnormal data during the occurrence 

 of cold weather, and as a rule in December and January, show that some 

 grave error must here have crept into the results. As every part of the 

 condenser and appurtenances functioned faultlessly, this error is liable 



