452 PROCEEDINGS OF THE AMERICAN ACADEMY. 



The wave-number differences of these pairs are approximately constant ; 

 they are : 96, 97.5, 97.7, '06. o, 98.4, ami 80. The last pair, being on 

 the verge of invisibility, is very inaccurately measured. These differences 

 are much less than for the usual series, being about 0.177 as much. 

 The line of gi'eater wave-length in the pairs has slightly greater intensity. 

 All the lines in these series are so exceedingly diffuse that their positions 

 cannot be measured with much accuracy ; it did not seem worth whiie 

 on this account to attempt to fit a formula to them. From an inspection 

 of their positions, it seems likely that they are not so directly connected 

 with the first and second subordinate series as in the case of Leuaid's 

 series in the sodium spectrum. They appear to run together to a 

 common end, which lies somewhat nearer the red than the end of the other 

 series. It seems most unlikely from their appearance that they can be 

 due to any impurity, as in that case they might f:xirly be expected to be 

 sharp when faint ; they are, however, always diffuse. These series are 

 faint, and of approximately equal intensity. 



It does not appear to have been previously noticed that the heavier 

 line of each pair in the first subordinate series is accompanied by 

 a satellite (lines all observed by Ramage) on the red side, forming a 

 " secondary " series, using Rydberg's term. The writer has calculated 

 the wave-number differences of the pairs in tlie caesium spectrum, using 

 Lehmann's data for the two extreme pairs and his own for most of the 

 rest. The results are as follows : 



553.2 



It will be noticed that the values for the first subordinate series are 

 lower than the others, and have an evident drift, with the exception of 

 the first, which is possibly in error. As the first member of each pair is 

 accompanied by a satellite (none have yet been observed for the first 

 and last lines), the wave-number differences were calculated between 

 the satellite and the line of shorter wave-length. The result is as 

 follows : 



