In estimating extensions in the unpolarised images no account was taken of light, which it seemed fair to attribute to diffusion from clouds. Extension was gauged along streamers, not in structureless luminosity. In many cases bands can be traced in the polarised images further than streamers in the unpolarised.
Many years ago Sir G. G. Stokes pointed out that the Savart polariscope might be used for the detection of faint extensions of the corona, and it may be that these photographs give evidence of the value of that suggestion.
Across the dark moon very faint bands are detectable only in the 17-second exposure. I am not yet prepared to say how far the discussion of the results may lead me to attribute the extension of the bands to the diffusion of polarised light which emanates from the corona by terrestrial clouds, but it is clear that the resultant polarisa- tion of such light must be, so far as the Savart bands are concerned, in a plane parallel to the equatorial extensions.
The value of the photographs, as a series, with related exposures is much interfered with by the variation in the effect produced by cloud, and in consequence it will hardly be possible to utilise them for finding at what distance from the sun there is the maximum relative polarisa- tion, even were the impressed extensions great enough. But as each plate had two exposures made on it simultaneously, one of the polarised image and one of the unpolarised, it is anticipated that they will serve well for the determination of the amount of polarisation in many regions of the corona.
The accompanying figure (fig. 1) indicates the contours of equal brightness in the two images, polarised (dotted line) and unpolarised (unbroken line) obtained in the second (the longest) exposure. It is reduced from a drawing made on a screen on which the two images were successively thrown by means of a lantern. The most marked streamers in the corona are indicated as well as the contours. No great precision has been aimed at in the diagram, its object being merely to give an idea of the sort of material obtained for the investigation. The phenomena are somewhat complicated; in con- sequence of the radial polarisation of the corona, the bands in the quadrants through which the central band passes form a system with white centre, those in the other quadrants form a system with dark centre ; the two systems meet along lines passing through the centre and inclined at 45 to the central band. Hence some care must be exercised in interpreting the appearances presented.
14. Atmospheric Polarisation.
The Savart polariscopes A, B, and C were fixed on the framework of the hut in such position that the sky in the neighbourhood of the
