jEther u Drift '' and Rotary Polarization. 395 



only twice through the trough. Although the liquid appears 

 perfectly clear and colourless, the absorption through this 

 length was very marked and the intensity much reduced. 

 That a large part of the light was due to internal scattering, 

 was made evident by the bright shaft of light, seen in the 

 liquid between the polarizer and first mirror, showing- along- 

 its length the various orders of colour corresponding to the 

 reflexion of plane-polarized light in successive azimuths from 

 small particles. 



The light, however, was sufficient to make successive 

 settings to 1/120°, and the quality of the rotatory compensa- 

 tion of the liquid indicated that greater lengths were practi- 

 cable. Such observations were taken on different days 

 between 11 a.m. and 1 p.m. and at 4.30 p.m., the entire appa- 

 ratus having been previously dismounted and cleaned; as 

 with all the precautions the surfaces became fogged and the 

 quality of the definition reduced. 



In the final test I used better mountings for my mirrors, 

 and examined the light after it had passed four times through 

 the liquid. The adjustments in this case were much more 

 tedious, but I finally succeeded in getting a field with a suf- 

 ficient intensity and uniformity to attain a high degree of 

 sensibility. Under conditions which seemed to be freest 

 from residual disturbances, I could observe no change on 

 reversing the trough. With an unusually brilliant sun, I was 

 able to detect a change in the plane of polarization, by rota- 

 ting P, of less than 0°*01, or one tenth of a division of my 

 scale = 1/120°. I also made several successive settings in the 

 two positions, but did not note them down, as I treated them 

 as preliminary to more extended systematic observations at 

 another time. I am not therefore able to reproduce them de- 

 finitely, but the differences between the means of the averages 

 of individual sets of four or five, which I calculated mentally 

 as the positions of the index were read off by my assistant, 

 were perhaps one half of this. The observations upon which 

 I based my conclusions were taken on May 10 th at the noon 

 hours 11-1 and at 4,30 P.M., which was as late as I could 

 obtain the sun. I did note occasionally, while observing 

 with the trough at rest, a gradual change in the field amount- 

 ing in one or two instances to 0°'05, apparently due to in- 

 ternal motions of the liquid and relative changes of temperature. 

 Observations at reversal were taken, however, when no gradual 

 change could be detected. Slight changes were also noted 

 due to a shift in the reflected beam from the heliostat. This 

 disappeared when the proper adjustments were made to bring 



