Upham's observations on the shore lines of Lake Agassiz have fully demonstrated that in northern Minnesota and Dakota, and in Manitoba farther north, the land was depressed, not elevated, at the time of the overflow of river Warren from Lake Agassiz; and hence, there as in the St. Lawrence Valley, the cause of overflow must be looked for in the retreating front of the Pleistocene ice sheet. In the Bulletin of the Geological Society of America (vol. ii, 1891, pp. 243-276) Upham has described a number of glacial lakes associated with large river channels, north of our boundary in Canada. The channels are now deserted by the great streams that carved them, and are occupied only by smaller streams, which are frequently "laked" by the alluvial deposits brought in by lateral tributaries, as will be referred to again further on.
In 1885 Gilbert traced out the shore lines of the expanded waters of Lake Ontario, afterward named Lake Iroquois by Spencer, and showed that they converged to the southeast, and at Rome, N. Y., an outlet was found through what is now called the valley of the Mohawk. Only a brief mention of the attractive problem offered by this locality has yet been published. Spencer has called attention to the probable former discharge of Lake Huron and Georgian Bay across the province of Ontario by way of the river Trent (Proceedings of American Association, xxxvii, 1888, p. 198); and Gilbert has suggested that at an earlier stage there was another outlet farther north, by way of Lake Nipissing and the Ottawa River the account of this being found in his excellent History of the Niagara River, published by the Commissioners of the State Reservation at Niagara, in their sixth annual report. The reading of this history will greatly increase the pleasure of an intelligent visit to the great cataract. It was of the outlet by way of the Ottawa River that Wright gave an account in the New York Nation for September 22, 1892.
The abandoned channel of overflow of the ancient Lake Bonneville at Red Rock Pass in northern Utah, and the "old river bed" leading from Sevier Desert to Great Salt Lake, well known from Gilbert's monograph, are analogous to the old channels here considered, although the overflows there were not produced by glacial barriers.
All these abandoned channels have certain features in common. At their upper end, where they trench across a divide of greater or less distinctness, they open out upon lacustrine plains of greater or less extent and distinctness, whose converging shore lines may be traced to the point of discharge. The breadth of the abandoned channel is relatively constant throughout a great part of its length; from which we may infer that the volume of water received from the lake at its head was large in comparison with that received from the tributaries lower down in its course. None
