certainty the general direction of the streams that flowed through this tract of country anterior to the glaciation. In addition to the two main branches of the river a large stream rising near the head of Lake Sumner followed the course of the Lake Taylor Valley, parallel with the South Branch; this entered the North Branch about half-way between Lake Sumner and the junction of the two main branches. A small tributary entered this valley on the north side, rising near the head of Lake Taylor and following the course of Lake Sheppard. Another small stream rose near Lake Katrine and joined the North Branch below the outlet of Lake Sumner. In pre-glacial times the ridges dividing these valleys would be more or less entire, though they might have saddles at their heads. It is impossible to reconstruct such features exactly, but the description just given affords a fairly accurate view of the stream conditions which obtained in this tract of country before it was modified by glaciation. Whatever was the prime cause which promoted glacier extension, it is reasonable to assume that it was gradual in its incidence. Snow would slowly accumulate, glaciers would be formed at higher altitudes and slowly extend down the valleys. Thus the heads of the small valleys would probably be filled with corrie glaciers, while the glaciers of the first order would be extending down the main valleys. These would help to lower the divides in the way suggested by Matthes.* F. E. Matthes, Glaciation of the Big Horn Mountains, U.S. Geol. Surv. 21st Ann. Rep., 1899–1900. As the ice-flow increased in volume the main streams would be filled, and in time overflows would take place over the lowest part of the divides, which would be lowered at the same time by active ice abrasion. It is significant that the greatest amount of lowering has taken place near the head of Lake Sumner. This would be due to the marked overflow of ice from the main Hurunui Valley, no doubt due to the narrowing of the cross-section of the valley at Lake Sumner, which caused the ice to overcrowd into the headwaters of the neighbouring streams, as it has done in several of the valleys of the Canterbury rivers. The full force of this would be felt at the head of the Lake Taylor Valley, and thus its divide has been completely reduced. The headwaters of the intermediate tributary valleys were also invaded and the saddles at their heads reduced. Thus a clear passage for the ice was opened down these valleys past the site of the Lakes Station in the direction of the south-easterly reach of the North Branch below Lake Sumner, while the main stream of ice followed down the valley now occupied by this lake. In addition to the overflow toward Lake Taylor a powerful stream passed over into the tributary which runs into the South Branch from the north. The saddle at the head of this stream was thus reduced, but not so much as its neighbour, which was more in the line where the ice-stream would impinge on the valley-wall. If, however, glacier action had continued this saddle would have been reduced and the mountain ridge to the north of the South Branch would have been completely isolated. It is possible that ice also overflowed into the valley of this stream near the Lakes Station, and as at the height of the glaciation the country in its vicinity would have the form of an intermontane basin, and would be an efficient gathering-ground, overflows from it took place along several lines from the front of the ice-sheet in the direction of the valley of the North Branch. These would produce the breaks in the valley-wall between the South Branch and the country to the north which occur immediately up-stream from the junction of the two branches.