OP ARTS AND SCIENCES. 277 



Passing now to the second part of the table, we find a wholly different 

 condition of things. Every star has a companion near it at an average 

 distance of only 1'.7, or these stars are twenty-six times as thickly 

 placed as in the rest of the heavens, since 8.7'^ : 1.7*^ = 26 : 1. This 

 effect is partly due to the surrounding square degree, which contains 

 nearly double the average number of stars. Only a small part of this 

 effect may, however, be thus explained. We may, therefore, infer 

 that there is a physical connection between these variables and their 

 companions, or that they are at nearly the same distance from the 

 Sun, and not optically double. The singular character of tiiese stars 

 renders them interesting objects for the measurement of parallax. 

 This is especially the case with those of very short period, since from 

 the rapidity of the changes we might infer that they were really 

 small, and therefore near. Now an observer .would be very likely 

 to select the companions as points to measure from, since their 

 distances are much greater than that separating the components of 

 most stars which are binary, or are supposed to be physically con- 

 nected. A measurable parallax might thus escape detection. 



The stars of longer period occupy an intermediate jjosition as re- 

 gards the distances of the components, and the number of stars in 

 the square degree. 



If the direction of the components depended wholly on chance, we 

 should find that they would differ, on the average, from that given by 

 any theory, by about 45°. It therefore seems scarcely probable that, 

 in each of the five cases, a chance distribution would give the angle 

 less than 45°, for the stars of short period. The uncertainties in the 

 measurements would in general increase the discrepancies, so that it is 

 to be expected that a more accurate determination would diminish the 

 mean value, although it would doubtless alter the separate results by 

 many degrees. The position of the components of the stars of the 

 fifth class has not been determined, as it seems very improbable that 

 they have any physical connection. The stars of long period, with 

 one exception, give results which do not agree at all with theory. 

 Some more precise test of the class to which these variables should 

 be assigned, is therefore much needed. They are distinguished from 

 many of the stars of the second class only by the length of their period, 

 no other known variables having a period less than that of R Vul- 

 jjeculce, or 137 days. Stars of Class 11. have banded spectra, and are 

 of a red color. This suggested a test dependent upon observations 

 already made. The last column shows what stars have been regarded 

 as red^ and may, therefore, in some cases belong to Class II. The only 



