26 SCIENCE PROGRESS 



times of passage of light along equal distances, the one parallel 

 and the other perpendicular to the drift of the ether. A 



difference of these times of the order of (^J is to be expected ; 



but, although this is so extremely small, the advantages which 

 arise from the possibility of now making the light travelling by 

 the different paths eventually recombine bring it within the 

 limits of delicate measurement. 



The actual method is the following : A beam of light is 

 split into two parts by incidence on a plate of glass placed at 

 45 to its direction. Part is reflected at right angles to its 

 original direction, and the remainder is transmitted with direction 

 unchanged. By means of two mirrors placed at equal distances 

 from the glass plate and normal to the respective beams, the 

 latter can be returned to the same point, and, being in a con- 

 dition to interfere regularly, will form a series of interference 

 bands. The positions of these will depend upon the times 

 taken by the light in traversing the two paths. If, from any 

 'Cause, a relative change of time occurs, a shift of the bands is 

 to be expected. Suppose now an ether-drift of magnitude n in 

 the direction of the original undivided beam. The transmitted 

 portion now travels toward the mirror with a velocity V + u, 

 V being the velocity of light in stationary ether. After re- 

 flection, however, it will have a velocity (relative to the earth, 

 •of course) V — u. If D is the distance between the unsilvered 

 glass and the mirror, the total time taken will be 



D D 



V + u ' V - u 



With regard to the other part of the beam, it might at first 

 be supposed that the velocity of transmission perpendicular to 

 the drift is unaffected, and that the time taken would be simply 



-y. This would lead to a difference of times equal to 



D D 2D 2D 



+ 



w 



V + u ^ V - u V V V 2 



u 



if powers of y higher than the second be neglected. However, 



the velocity perpendicular to the drift is not unaffected by it ; 

 and when this effect is duly allowed for, the above difference is 

 reduced by one-half. Now suppose that by rotating the whole 

 apparatus, the ether-drift is caused to be along the path of the 



