OF ARTS AND SCIENCES. 21 



the observations of Prof. Cooke with the spectroscope, and of Prof. 

 Tjndall on aqueous vapor in a state of formation. 



Observations were next made to test the results found above, for the 

 light reflected and transmitted by several parallel surfaces of glass. To 

 check the results, which are given in Table XIII., two, and in some 

 cases three, independent methods were employed. For convenience of 

 reference, the series are numbered, as in the observations on sky polari- 

 zation. The general method employed to measure the polarization of the 

 reflected ray was to lay one or more sheets of glass on a piece of black 

 velvet, and render them horizontal with a spirit-level. The polarimeter 

 was then mounted a short distance from them, carefully levelled and 

 then turned down, so that the light should be reflected from their sur- 

 faces. Its angle of depression would then equal the complement of the 

 angle of incidence. The line of junction of the two images was then 

 rendered vertical, and the polarization measured in the usual way. The 

 polarization of the sky, if clear, would introduce a large error into the 

 results, and care was therefore taken to make these observations only 

 on cloudy days. Although it is commonly stated that no traces of 

 polarization can be detected in the heavens when completely covered 

 with clouds, yet it was found to be slightly polarized in a vertical jilane 

 at such times, the effect being most marked near the horizon, and prob- 

 ably due to reflection from terrestrial objects. To obtain a single sur- 

 face of glass, a piece was blackened on one side in the flame of a candle, 

 in the expectation that the oil in the lamp-black, having nearly the 

 same index of refraction as the surface to which it adhei'ed, would pre- 

 vent all specular reflection from it. But the results obtained did not 

 agree with those computed for a single surface, and a close examination 

 showed that double reflections were given of objects in front, as with a 

 common plate of glass ; in fact, that the lamp-black acted merely like 

 the velvet in the other cases. The two series, Nos. 20 a and 21 a, are 

 therefore placed with the observations on two surfaces, with which they 

 agree very well. A piece of colored glass was next used, which gave 

 the results in the column headed 17 a. Series 22 «, 27 a, 35 a, and 

 36 a were obtained in the same manner, using 1, 4, and 10 pieces of 

 plate glass, laid on one another so as to form a pile. 



The other measurements of the polarization of the reflected beam 

 were obtained by quite a different method. A large Babinet's gonio- 

 meter, or optical circle, was employed, the slit being removed and 

 replaced by a Nicol's prism, which was free to turn around its axis, the 

 angle of rotation being measured by a graduated circle and index. In 

 the eye-piece of the observing telescope, a Nicol's prism was placed, and 



