The formation or coagulation of ice-islands has not, to my knowledge, been thoroughly investigated.  Some have supposed them to be formed by the freezing of the water at the mouths of large rivers, or great cataracts, where they accumulate till they are broken off by their own weight.  My observations will not allow me to acquiesce in this opinion; because we never found any of the ice which we took up incorporated with earth, or any of its produce, as I think it must have been, had it been coagulated in land-waters.  It is a doubt with me, whether there be any rivers in these countries.  It is certain, that we saw not a river, or stream of water, on all the coast of Georgia, nor on any of the southern lands.  Nor did we ever see a stream of water run from any of the ice-islands.  How are we then to suppose that there are large rivers?  The valleys are covered, many fathoms deep, with everlasting snow; and, at the sea, they terminate in icy cliffs of vast height.  It is here where the ice-islands are formed; not from streams of water, but from consolidated snow and sleet, which is almost continually falling or drifting down from the mountains, especially in the winter, when the frost must be intense.  During that season, the ice-cliffs must so accumulate as to fill up all the bays, be they ever so large.  This is a fact which cannot be doubted, as we have seen it so in summer.  These cliffs accumulate by continual falls of snow, and what drifts from the mountains, till they are no longer able to support their own weight; and then large pieces break off, which we call ice-islands.  Such as have a flat even surface, must be of the ice formed in the bays, and before the flat vallies; the others, which have a tapering unequal surface, must be formed on, or under, the side of a coast composed of pointed rocks and precipices, or some such uneven surface.  For we cannot suppose that snow alone, as it falls, can form, on a plain surface, such as the sea, such a variety of high peaks and hills, as we saw on many of the ice-isles.  It is certainly more reasonable to believe that they are formed on a coast whose surface is something similar to theirs.  I have observed that all the ice-islands of any extent, and before they begin to break to pieces, are terminated by perpendicular cliffs of clear ice or frozen snow, always on one or more sides, but most generally all round.  Many, and those of the largest size, which had a hilly and spiral surface, shewed a perpendicular cliff, or side, from the summit of the highest peak down to its base.  This to me was a convincing proof, that these, as well as the flat isles, must have broken off from substances like themselves, that is, from some large tract of ice.