stellar clusters 



49 



desirable. If I have not performed such a discussiou it is because the agreement 

 with the americau observations, regarding tlie first regression hue, is so great tliat 

 a note-worthy correction to the equation of the second regression hue is scarcely 

 not to be expected. 



In table 15 (p. 21) Schwarzschild gives the mean values of his colour indices 

 corresponding to the different spectral types in H. 5U. From these values we find 

 that the color-indices in Göttingen are to be augmented by -|- 0.26 for being 

 reduced to the international scale. Having added this constant I compared his num- 

 bers with the formula (2), written in the form 



c = + 0.44 S. 



The result of this comparison is found in the diagram fig. 4. 



C 



Fig. i. Comparison between table 10 and Yerkes (•), Harvard ( + ) and Göttingen (O)- 



Excepting the vahies s= + 1-8 (= -F8) and s = — 1.0 {BO), the agreement is 

 satisfactory. As to BO the explanation may, without doubt, be sought in the lack 

 of homogenity of this subclass which I have found in my researches on the distri- 

 bution of the jB-stars (Meddel. Ser. II, N:o 14), in conjunction with the small num- 

 ber of stars observed (iV"=4). 



As to FS we observe that the large jump in the colour indices between the 

 spectral-classes FS and GO is found also in the Yerkes observations (but scarcely 

 in the Harvard observations). Tlie explanation of this discontinuity seems to be a 

 certain ambiguity concerning the definition of the spectral class GO. Parkhukst 

 has called attention to the discrepancy between the definition given in Harvard 

 Circ. 172 and H. 28. Another illustration is found in the diagram on p. 19 of the 



