Whether these processes furnish an adequate explanation of the changes or not, the observed fact is that there are all gradations from snowflakes and pellets into granular névé, and thence into glacier granules (_Gletscherkörner_), varying in size up to that of filberts and walnuts, and even beyond. In coherence, these aggregations may vary from the early slightly coherent granular stage, where the grains are small and spheroidal, to the ice stage, where the cohesion has become strong through the interlocking growths of the large granules. Even when the mass has become seemingly solid ice, sufficient space is usually left between the granules to give the dispersive reflection to light which imparts to glacier ice its distinctive whitish color. =The arrangement of the crystal axes.=—The most radical difference between glacier ice and ice formed directly from water is in the arrangement (orientation) of the crystals. In the ice formed on undisturbed water, the bases of the crystals are at the surface and their principal axes are vertical, as shown by Mügge.[134] As they grow, the crystal prisms extend downwards. This gives a columnar or prismatic structure to the ice, well seen when it is “honeycombed” by partial melting. In the glacier, on the other hand, the crystals, starting from snowflakes, have their axes turned in various directions according to the accidents of their fall; and as the snow develops into ice, the principal axes of the crystals continue to lie in all directions. Hence glacier ice, unlike pond ice, cannot usually be split along definite planes, except where cleavage planes are subsequently developed by extraneous agencies. [Illustration: +Fig.+ 291_a_. +Fig.+ 291_b_. +Fig.+ 291_c_. Figures to illustrate the method of deformation of ice crystals.] While the crystals of a glacier usually have their principal axes in various directions, there appears to be a tendency for them to approach parallelism in certain positions, especially in the basal parts of a glacier near its terminus. Observations on this point are not so full and critical as could be desired, but it is probable that the parallel orientation is partly general, and due to the vertical pressure of the ice, and partly special and local, and connected with the shearing planes and foliation.