In other places the bed of the glacier is rough and torn. The rocks have been disrupted and their fragments have been carried away,--a process known as _plucking_. Moving under immense pressure the ice shatters the rock, breaks off projections, presses into crevices and wedges the rocks apart, dislodges the blocks into which the rock is divided by joints and bedding planes, and freezing fast to the fragments drags them on. In this work the freezing and thawing of subglacial waters in any cracks and crevices of the rock no doubt play an important part. Plucking occurs especially where the bed rock is weak because of close jointing. The product of plucking is bowlders, while the product of abrasion is fine rock flour and sand. Is the ground moraine of Figure 87 due chiefly to abrasion or to plucking? [Illustration: Fig. 111. Roches Moutonnés, Bronx Park, New York] =Roches moutonnées and rounded hills.= The prominences left between the hollows due to plucking are commonly ground down and rounded on the stoss side,--the side from which the ice advances,--and sometimes on the opposite, the lee side, as well. In this way the bed rock often comes to have a billowy surface known as roches moutonnées (sheep rocks). Hills overridden by an ice sheet often have similarly rounded contours on the stoss side, while on the lee side they may be craggy, either because of plucking or because here they have been less worn from their initial profile (Fig. 112). =The direction of glacier movement.= The direction of the flow of vanished glaciers and ice sheets is recorded both in the differences just mentioned in the profiles of overridden hills and also in the minute details of the glacier trail. Flint nodules or other small prominences in the bed rock are found more worn on the stoss than on the lee side, where indeed they may have a low cone of rock protected by them from abrasion. Cavities, on the other hand, have their edges worn on the lee side and left sharp upon the stoss. Surfaces worn and torn in the ways which we have mentioned are said to be glaciated. But it must not be supposed that a glacier everywhere glaciates its bed. Although in places it acts as a rasp or as a pick, in others, and especially where its pressure is least, as near the terminus, it moves over its bed in the manner of a sled. Instances are known where glaciers have advanced over deposits of sand and gravel without disturbing them to any notable degree. Like a river, a glacier does not everywhere erode. In places it leaves its bed undisturbed and in places aggrades it by deposits of the ground moraine. [Illustration: Fig. 112. A Glaciated Hill, Norway. Sharp Weathered Mountain Peaks in the Distance]