252 CARNEGIE INSTITUTION OF WASHINGTON. 



system. Tliis method of clnssification has been of great value in its 

 bearing on discussions of radial velocity and investigations connected 

 with stellar magnitudes. All of the available spectra of stars with 

 types between F and M have now been classified in accordance with this 

 system. 



Spectkal Determinations of Stellar Parallax, 



It is well known that if the intrinsic brightness or the absolute magni- 

 tude of a star can be determined, its parallax may readily be derived 

 by means of the relationship connecting apparent magnitude with 

 distance and absolute magnitude. An investigation by Mr. Kohlschiit- 

 ter about two years ago showed clearly that certain Unes in the spectra 

 of the stars of the more advanced types vary in intensity with the 

 intrinsic brightness of the stars in which they appear. The possibility 

 of applying these results to the determination of absolute magnitudes 

 was considered at that time. 



The rapid accumulation of stellar-spectrum photographs, particu- 

 larly for stars with measured parallaxes, has made it possible during 

 the past year to investigate this question more thoroughly. A con- 

 siderable number of stars of different spectral types was selected for 

 which reliable parallax measures were available, and their absolute 

 magnitudes were calculated with the aid of these values. The spectra 

 of these stars were then investigated, and estimates were made of the 

 intensities of the lines which appeared to be subject to variation with 

 absolute magnitude. These estimates were then plotted, and equa- 

 tions were derived by means of least-squares solutions giving the 

 relationship between hne intensity and absolute magnitude. These 

 equations may be used to detennine the absolute magnitude of a 

 star when the intensities of the selected lines are known. From the 

 absolute magnitude the parallax is readily calculated. 



An application of these results to 124 stars of measured parallax may 

 be summarized as follows: 



1. The average deviation of the spectroscopic parallaxes from the 

 measured values is 0''025. 



2. The only discrepancy exceeding OTl is in the case of a star with a 

 parallax measured by a single observer. 



3. The average value found for 25 stars with negative measured 

 parallaxes is +0''03, the largest value being -}-0''08. The spectro- 

 scopic method gives no negative parallaxes. 



4. The spectroscopic method appears to give its most accurate 

 results for stars of types K and M. It is of httle value for stars with 

 types earUer than F5. 



5. An increase in the number of measured parallaxes for very 

 luminous stars should add materially to the accuracy of the spectro- 

 scopic method. 



