It is true that the extensive decomposition of the underlying rock, penetrating sometimes to a considerable depth, makes it often difficult to distinguish between it and the drift; and the problem is made still more puzzling by the fact that the surface of the drift, when baked by exposure to the hot sun, often assumes the appearance of decomposed rock, so that great care is required for a correct interpretation of the facts. A little practice, however, trains the eye to read these appearances aright; and I may say that I have learned to recognize everywhere the limit between the two formations. There is indeed one safe guide, namely, the undulating line, reminding one of _roches moutonnées_,[97] and marking the irregular surface of the rock on which the drift was accumulated; whatever modifications the one or the other may have undergone, this line seems never to disappear. Another deceptive feature, arising from the frequent disintegration of the rocks and from the brittle character of some of them, is the presence of loose fragments, which simulate erratic boulders, but are in fact only detached masses of the rock in place. A careful examination of their structure, however, will at once show the geologist whether they belong where they are found, or have been brought from a distance to their present resting-place. But, while the features to which I have alluded are unquestionably drift phenomena, they present in their wider extension, and especially in the northern part of Brazil, some phases of glacial action hitherto unobserved. Just as the investigation of the ice-period in the United States has shown us that ice-fields may move over open level plains, as well as along the slopes of mountain valleys, so does a study of the same class of facts in South America reveal new and unlooked-for features in the history of the ice-period. Some will say that the fact of the advance of ice-fields over an open country is by no means established, inasmuch as many geologists believe all the so-called glacial traces—viz. striæ, furrows, polish, etc., found in the United States—to have been made by floating icebergs at a time when the continent was submerged. To this I can only answer that, in the State of Maine, I have followed, compass in hand, the same set of furrows, running from north to south in one unvarying line, over a surface of one hundred and thirty miles, from the Katahdin Iron Range to the sea-shore.[98] These furrows follow all the inequalities of the country, ascending ranges of hills varying from twelve to fifteen hundred feet in height, and descending into the intervening valleys only two or three hundred feet above the sea, or sometimes even on a level with it. I take it to be impossible that a floating mass of ice should travel onward in one rectilinear direction, turning neither to the right nor to the left, for such a distance. Equally impossible would it be for a detached mass