338 'Prof. Tyndall un some Physical Properties of Ice. 



The interior walls of the segments were thickly covered with rich 

 liquid discs ; in some cases the vision plunged unimpeded into 

 the ice to a depth of several inches, while in others the prismatic 

 segments were dotted with discs to their very centres. Fig. 1 



Fig. 1. 



represents one of a number of cases of this kind which I had an 

 opportunity of examining. The network of lines are the inter- 

 sections of the bounding surfaces of the segments with the sur- 

 face of the slab of ice ; the short lines in each segment represent 

 the sections of the discs ; and it will be observed that, while in 

 each single segment their directions are alike, in no two seg- 

 ments is this the case. The flat rings denote that the discs at 

 these places were parallel to the surface of the ice slab, while in 

 all the other cases they were perpendicular to it. 



18. I placed this slab on a table which could be caused to 

 rotate, and bringing it into the path of a concentrated sunbeam, 

 caused the beam to travel all round it. The little flowers started 

 into existence with extreme suddenness and beauty. In all cases 

 the planes of the flowers ivere parallel to the planes of the discs. 



19. Hence the conclusion that the flowers are always parallel 

 to the surface of freezing, must not go without qualification. 

 There is no doubt that each of the fragments above referred to, 

 possesses an optic axis perpendicular to the planes of the flowers 

 developed in it ; the directions of these axes being therefore as 

 various as the planes aforesaid. How is this result brought 

 about ? Has the molecular structure of the ice been always what 

 the last experiments show it to be ? Sir David Brewster's ob- 

 servations are in opposition to this idea. Or can it be that the 

 ice has changed, by a rearrangement of the constituent particles 

 of each segment, this arrangement producing the surfaces of 

 discontinuity by which the segments are bounded ? At present we 



