The northeastern corner of Massachusetts and the southeastern corner of New Hampshire are largely covered with these peculiar-shaped glacial deposits, while they are numerous as far west as Fitchburg, in Massachusetts, and Ware, N. H., and in the northeastern part of Connecticut. A little later, also, we shall refer to an interesting line of them in central New York. Elsewhere in America, except in a portion of Wisconsin, they rarely occur in such fine development as in New England. In Europe they are best developed in portions of Ireland. One's first impression in examining an exposed section of a drumlin would lead him to think that the mass was entirely unstratified; but closer examination shows that there is a coarse stratification, but evidently not produced by water-action. The accumulation has probably taken place gradually by successive deposits underneath the glacier itself. Professor William M. Davis has suggested a plausible explanation which we will briefly state. [Illustration: Fig. 28.--Drumlins in the vicinity of Boston (Davis).] The frequency with which drumlins are found to rest upon a mass of projecting rock, the general co-ordination of the direction of their axes with the direction of the scratches upon the underlying rock, and the abundance of scratched stones in them, all support the theory that drumlins are formed underneath the ice-sheet, somewhat in the way that islands and bars of silt are formed in the delta of a great river. The movement of ice seems to have been concentrated in pretty definite lines, often determined by the contour of the bottom, leaving a slacker movement in intervening areas, which were evidently protected in some cases by projecting masses of rock. In these areas of slower movement there was naturally an accumulation at the same time that there was vigorous erosion in the lines of more rapid movement. There was doubtless a continual transfer of material from the end of the drumlin which abutted against the moving mass of ice to the lower end, as there is in the formation of an island in a river. If time enough had elapsed, the whole accumulation would have been levelled by the glacier and spread over the broader area where the more rapid lines of movement became confluent, and where the differential motion was less marked. Drumlins are thus characteristic of areas in the glaciated region whose floor was originally only moderately irregular, and where there was an excessive amount of ground-moraine to be transported, and where the movement did not continue indefinitely. It has been suggested, also, that some of the long belts of territory in New England and central New York covered by drumlins may represent old terminal moraines which were subsequently surmounted by a readvance of the ice, and partially wrought over into their present shape.