20 PROCEEDINGS OF THE AMERICAN ACADEMY 



polarization would then be at once given by Table II., making i equal 

 to one-half the solar distance. The curve thus obtained is given in 

 Fig. 5, at A. The polarization according to this should be complete 

 at 90° from the sun, while in reality it is only about 70 per cent. 

 If, however, we multiply the ordinates of ciu-ve A by this fraction, we 

 obtain curve B, which agrees almost precisely with the curve given in 

 column 2 of Table XII. Its ordinates are given in column 3, and the 

 differences in column 4. From the latter it will be seen that the em- 

 pirical curve gives results somewhat too great for solar distances less 

 than 60°, and too small for greater distances ; but the deviation is so 

 small, compared with the accidental errors, that we are scarcely justified 

 in drawing any conclusions from them. The agreement of all the 

 observations in the neighborhood of 120° from the sun is remarkable, 

 and not easily explained. The observations of series 10 for distances 

 less than 110° give results decidedly below that given by theory. A 

 possible explanation is the reflection of the sun on the sea to the east 

 of Boston, a source of error not present in the earlier observations 

 which were made inland. It will be noticed that no account is here 

 taken of the points of no polarization, or neutral points of the sky ; but 

 the polarization is very slight for some distance from them, and hence 

 is not easily measured. They must be regarded as due to some second- 

 ary distui-bing cause, as refracted light, which alters the general polar- 

 ization of the sky but little. 



When the polarimeter is directed towards a polished colored plane 

 surface, the two images assume different tints. One, which contains 

 the light polarized in the plane of incidence, or B, is composed mainly 

 of the light reflected specularly, and is therefore white like the source 

 of light. The image A contains but little of the light reflected specu- 

 larly, consisting principally of the rays emitted by the body, and hence 

 partaking of its color. The idea at once suggested itself, that testing 

 the light of the sky in this way might give a clue to the cause of its 

 color. The experiment was tried several times, with negative results, 

 the two images appearing of precisely the same blue tint. But on the 

 evening of July loth, near sunset, when measuring the polarization of 

 a point near the northern horizon, where the blue color was compara- 

 tively pale, a marked difference in the two images was observable. 

 The image B was found to be of a yellowish brown, ^ of a grayish 

 blue or violet tint. This observation has since been frequently re- 

 peated, and can, in fact, be made almost any clear evening near 

 sunset. Evidently we may conclude from these colors that the true 

 color of the sky particles is blue, a view quite in accordance with 



