262 CARNEGIE INSTITUTION OF WASHINGTON. 



The assumption that the increase in color is a consequence of absolute 

 luminosity nmst be considered in connection with the spectroscopic 

 results discussed elsewhere. These show that stars of high luminosity 

 (mostly of great size) are redder than those absolutely fainter, but that 

 the effect is a function of spectral type and probably inappreciable for 

 first-type stars. The color indices, however, would indicate that the 

 absolutely faint stars are the redder. To reconcile these results we must 

 suppose, first, that among the faint stars there are few or none of early 

 type, for, as just stated, the color of these is apparently independent of 

 their luminosity and should be the same for all magnitudes; second, that 

 the relatively high color of the stars of small luminosity is only indirectly 

 a luminosity effect and probably is to be explained on the basis of spec- 

 trum. The spectroscopic results make it unlikelj^ that abnormally high 

 color for any given spectrum is associated with low luminosity, while 

 there is a very definite relation between small luminosity and advanced 

 spectrum. This, however, merely throws us back upon the previous 

 explanation, which attributes the increase in the minimum color index 

 to a very special distribution of spectral types. Upon the color effect 

 thus produced is superimposed that due to luminosity in the sense 

 revealed by the spectroscopic observations. The completeness with 

 which the earlier spectral types would have to be excluded from the 

 faint stars in order to make such an explanation tenable is indicated 

 by the fact that for several hundred stars fainter than the eleventh 

 magnitude no single value of the color index falls appreciably below 

 the sharply defined limit fixed by the smaller values. 



The decision between such an explanation and that of the third possi- 

 bility — space absorption — is at present a question of probabilities. 

 Some light might be thrown upon the matter by a precise determination 

 of the behavior with increasing magnitude of the mean or the maximum 

 color index, but the data at present available are insufficient for this 

 purpose. 



Finally, it should be remarked that a phenomenon pointed out by 

 Hertzsprung must also be involved, but it seems unlikely that this 

 affords a complete explanation. 



Considered by themselves, the color results for N. G. C. 1647 afford a 

 valuable control upon the relation of the magnitude scales to each other. 

 The colors of a large number of stars in this cluster have also been 

 determined by Professor Hertzsprung by means of a large objective 

 grating. The transformation of his effective wave-lengths into color 

 indices gives results agreeing verj^ closely wdth those found directly 

 from the photographic and photovisual magnitudes. Fifty-two stars 

 between magnitudes 11 and 14.5 have, according to the effective wave- 

 lengths, a mean index of -|-0.46 mag. That found from the magnitudes 

 is 4-0.47 mag. The agreement is closer than was anticipated, for the 



