of Edinburgh, Session 1870 - 71 . 
409 
until crystals formed in the liquid. These were necessarily thin, 
since their thickness was limited by the interval between the 
glasses. Of course the central part of each crystal, except the 
smallest ones, was hounded by parallel planes, but the extremities 
were bevilled at various angles, forming so many little prisms, the 
smallest of them floating in the liquid. When a distant candle 
was viewed through these glasses, having the little prisms inter¬ 
posed, a great number of spectra became visible, caused by the 
inclined edges. Most of these w r ere no doubt very imperfect, but 
by trying the glass at various points, some very distinct spectra 
were met with, and these could with some trouble be isolated by 
covering the glass with a card pierced with a pin-hole. It was 
then seen that each prism (or oblique edge of crystal) produced two 
spectra oppositely polarised and widely separated. One of these 
spectra was normal; there was nothing particular about it. The 
colours of the other were very anomalous, and, after many experi¬ 
ments, I came to the conclusion that they could only be explained 
by the supposition that the spectrum, after proceeding for a certain 
distance, stopped short and returned upon itself. 
No accurate measurements, however, w r ere made, because it 
always happened that, after the lapse of a minute or two, the 
crystals dissolved in the surrounding liquid, owing to the warmth 
of the hand or eye. The presence of the liquid, however, was 
necessary to give the crystals the requisite transparency, and, 
moreover, the liquid virtually diminishes the angle of the prism 
floating in it, which otherwise would be too great to give a good 
result. I never published this experiment, because I found it 
delicate and capricious, and I w r as reluctant to publish any facts 
that might be difficult for others to verify. But I have several 
times described it to Sir D. Brewster in conversation, and he always 
said that he thought it very important, at the same time suggesting 
that there might perhaps he some fallacy. This was because he 
doubted the possibility of a spectrum being partially inverted or 
returning on itself. But this doubt seems now to be wholly 
removed by Christiansen’s experiment, in which there seem to be 
two inversions in the spectrum, and therefore I no longer hesitate 
to state the grounds on which I concluded long ago that this 
phenomenon was possible. 
