3 i2 POPULAR SCIENCE MONTHLY. 



mm., or 0.00001 inch. In view of the newness of the subject, the 

 richness of the field, and the fact that the more active great telescopes 

 are now nearly all applied to this work, I append a list of the improve- 

 ments which have contributed most powerfully to recent progress : 



1. A realization of the fact that a spectrograph is an instrument 

 complete in itself. The telescope to which it is attached serves only to 

 collect the light and to deliver it properly upon the slit. 



2. The development of a method of reduction which permits the 

 use of all good stellar lines, irrespective of whether they correspond to, 

 or lie between, the comparison lines. 



3. The use of a longer collimator, permitting a wider slit, and re- 

 quiring larger prisms, with greater resolving power. 



4. The use of simple prisms, of better glass, with better optical 

 surfaces. 



5. Care in collimating, to insure that the star light and comparison 

 light traverse identically the same part of the collimator lens. 



6. The adoption of a compact and rigid form of spectrograph 

 mounting, designed in accordance with good engineering practise. 



7. The elimination of flexure effects by supporting the spectrograph, 

 in connection with the telescope, in accordance with engineering prin- 

 ciples. The conventional spectrograph had been supported entirely at 

 its extreme upper end; the instrument projected out into space, unsup- 

 ported, boldly inviting flexure under the varying component of gravity. 



8. The use of a constant temperature case around the instrument. 



9. Precautions taken to eliminate many sources of error from the 

 measures of the spectrograms. 



Up to December, 1900 — the last month of the departing century — 

 the speeds of 325 stars had been determined with the Mills Spectro- 

 graph in the northern two thirds of the sky. Omitting several stars 

 whose lines could not be measured accurately, and some thirty spectro- 

 graph^ and visual binaries for whose centers of mass the velocities 

 were still unknown, 280 stars remained available for deducing the 

 relative motion of our solar system. The observational data were 

 distributed symmetrically in right ascension, and the result for this 

 coordinate of the apex agreed with Newcomb's proper-motion result 

 within a small fraction of a minute of arc. The data were extremely 

 unsymmetrical in declination, as there were few observations between 

 — 15° and — 30° declination, and none whatever south of — 30°. 

 The solution placed the apex 15° south of Newcomb's position. The 

 deduced speed, 20 km. per second, is no doubt close to its true value. 



There is a question whether the direction of the solar motion 

 can be determined more accurately from proper motions or from 

 radial velocities, an equal number of stars being available in the two 

 cases; but as to the speed, no doubt of the very marked superiority of 

 the spectrograph^ method can exist. This, however, is but incidental, 



