A network of passages situated along three-grain intersections enables water to percolate through temperate glacier ice. The deformability of the ice allows the passages to expand and contract in response to changes in pressure, and melting of the passage walls by heat generated by viscous dissipation and carried by above-freezing water causes the larger passages gradually to increase in size at the expense of the smaller ones. Thus, the behavior of the passages is primarily the result of three basic characteristics: (1) the capacity of the system continually adjusts, though not instantly, to fluctuations in the supply of melt water; (2) the direction of movement of the water is determined mainly by the ambient pressure in the ice, which in turn is governed primarily by the slope of the ice surface and secondarily by the local topography of the glacier bed; and, most important, (3) the network of passages tends in time to become arborescent, with a superglacial part much like an ordinary river system in a karst region, an englacial part comprised of tree-like systems of passages penetrating the ice from bed to surface, and a subglacial part consisting of tunnels in the ice carrying water and sediment along the glacier bed. These characteristics indicate that a sheet-like basal water layer under a glacier would normally be unstable, the stable form being tunnels; and they explain, among other things, why ice-marginal melt-water streams and lakes are so common, why eskers, which are generally considered to have formed in subglacial passages, trend in the general direction of ice flow with a tendency to follow valley floors and to cross divides at their lowest points, why they are typically discontinuous where they cross ridge crests, why they sometimes contain fragments from bedrock outcrops near the esker but not actually crossed by it, and why they seem to be formed mostly during the later stages of glaciation.
