The present average southward slope of the highest terraces west of the trap ridge from Northampton, Mass., to Farmington, Conn., forty-four miles, is seven inches per mile,[49] and Professor Dana is inclined to believe that this is approximately the slope at the time the terraces were built. The character of the deposits shows that the current which formed these deposits flowed south. The present river, flowing north, falls twenty feet between Farmington and Tariffville, or 1⅔ feet per mile. The reversal of the river was probably determined by two factors. Near the village of Farmington, the waters of 200 square miles of territory are poured into the valley by the upper Farmington and its tributary, the Pequabuck. During the terrace building stage the great mass of débris contributed by these streams was deposited where the steep gradient of the highlands was exchanged for the gentle slope of the lowland. The main north and south valley was thus choked by the débris of its tributaries and a long stretch of comparatively still water extended north from Farmington, in which nearly horizontal deposits were made. South of Farmington the terrace deposits are much coarser than to the north, and the face of the terraces is much greater. It is not impossible that, as the deposits between Farmington and the Massachusetts state line approached nearer and nearer to horizontality, the waters of the upper Farmington began to divide, part flowing north and part south, the northward flowing portion finding an outlet at the sag at Tariffville. If this was the case, the terraces between Farmington and Tariffville must have had a slight slope to the north. Their present southward slope could readily be accounted for by the re-elevation of the land after the disappearance of the ice. This explanation rests upon the ability of the upper Farmington and the Pequabuck to have completely dammed the southward flowing current and turned it northward by the great mass of their deposits. If this was not the case, and there may be some doubt on the matter, the subsidence which accompanied the later stages of the ice-retreat is the other factor in the problem. It is estimated that an average depression of 1.25 feet[50] through the Connecticut valley would restore it to an altitude approximating that at the close of glaciation. It seems highly probable that these terrace-deposits were built before the maximum depression was reached. If this was the case, the depression would be efficient in reversing the Farmington, and this factor would supplement the first. It is impossible at present to say to what extent these two factors enter into the problem. That they are not mutually exclusive is evident, and that they are together quantitatively competent seems certain. Among the several hypotheses which have been considered, this seems the most probable, and in the light of the present evidence the most rational.