MOUNT WILSON OBSERVATORY. 



247 



is necessary to multiply the spectroscopic parallaxes in order to secure 

 the best possible agreement with the trigonometric results. Due 

 account was taken of the fact that the errors in the spectroscopic paral- 

 laxes are dependent upon the size of the parallaxes themselves. After 

 considerable investigation it was decided to base the reduction upon 

 the four excellent series of trigonometric parallaxes made by the pho- 

 tographic method at the Allegheny, McCormick, Mount Wilson, and 

 Yerkes Observatories. A total of 657 stars of types F, G, K, and M 

 was used in the computation. 



The systematic corrections to the approximate absolute magnitudes 

 have been assumed to consist of two parts — systematic errors and 

 errors which are accidental and distributed according to the usual 

 frequency law. As a consequence of this assumption the most proba- 

 ble parallax does not correspond exactly with the most probable abso- 

 lute magnitude. After the determination of the systematic errors for 

 groups of stars with but small variations in line-intensity and spectral 

 type, the condition of a continuous change of absolute magnitude with 

 these variables is found to hold for all stars except the giants and dwarfs 

 of the later K and M types. Since there appear to be almost no ob- 

 jects of intermediate magnitude among stars of these types, the discon- 

 tinuity is of little consequence. 



The reduction tables for the Cepheid variables and other stars of 

 similar spectral type have been based upon a consideration of proper 

 motions alone, since the measured parallaxes are too small to furnish 

 reliable data. The absolute magnitudes as derived from line-intensities 

 for the Cepheid variables show a marked correlation with the proper 

 motions of these stars. 



A comparison of the values of the trigonometric and the spectroscopic 

 parallaxes for the stars used in this reduction shows the following re- 

 sults, the differences being taken directly and without regard to weights : 



The Spectra of Certain Nov.^ and of the Variable Star T Pyxidis. 



The discovery of several stars with the variations of light characteris- 

 tic of Novae, made by Harvard observers from a comparison of photo- 

 graphs, has added a number of interesting objects of this character to 

 those already known. Two of these stars have been observed with the 

 100-inch reflector in addition to Nova Aquilse of 1918 and the variable 



