extension of the line of this scarp, and may have some relation to it. It should be noted that no block of greywacke or associated rock from the back country was found on the southern end of the Surrey Hills and such blocks should be easily detected if they lay on the surface of the volcanics forming these hills, so it is reasonable to conclude that the ice did not reach as far as their lower slopes.

It would be a simple solution of the problem to regard the veneer of boulders on the face of the top terrace as having been dropped initially by a glacier retreating from a more advanced position and included subsequently in the material forming the top layers of the terrace, the larger blocks sinking gradually in the loosely compacted gravel as happens at the present time in front of many glaciers in New Zealand, and also in the case of concrete blocks washed from the protective works on the banks of gravel-bearing rivers. The scouring of the finer material either on the downstream or the upstream side of the block is largely responsible for its settlement and ultimate disappearance. If this is correct, one must explain why large blocks lie thickly on the treads of the lower terraces, and also why a veneer of coarse material masks a finer-grained alluvial basement. I can only suggest that the underlying gravels, as they form part of the lower strata composing the plains, have had an opportunity to become more compacted, and therefore more capable of supporting larger blocks on a water-worn surface; such a compacted surface will be less liable to scour. The upper layers of the top terrace have not had this chance of becoming consolidated, and the larger blocks have sunk into them when washed by water, only to appear on the edge of the terrace when the river has cut down to a lower base-level and has removed the finer material forming a matrix in which the larger blocks are included. Of course, if the glacial river be considered competent to move such blocks for a considerable distance from the ice-front and to concentrate them into definite accumulations, then it must be conceded that the terrace affords no positive proof of the former presence or proximity of ice, but the break near Montalto facing the plains has still to be explained, and its somewhat irregular line, with embayments facing upstream, cannot be accounted for by subsequent erosion by a river flowing along its front.

When considering this aspect of the matter, it must be remembered that at times glaciers do not form terminal moraines, and the conditions that determine their absence are not thoroughly understood. In this case a paucity of moraine may be due to a glacier issuing from a somewhat confined valley and deploying on to a plain where it would tend to spread, thin out, and develop an indefinite front, and the material it carried through the gorge would be scattered as a thin veneer over a wide area. The paucity of evidence in this case may be accounted for in some such way.

Another problem in this connection is the origin and subsequent history of the blocks that veneer the treads and risers of the lower terraces (Plates 31 and 32). They may have been derived from material included in the top terrace when its surface extended right across the present site of the river, the large blocks contained therein merely sinking to a lower and lower level as the stream eroded its bed deeper and deeper into the pre-existing cover. But in all probability

this does not represent the entire case. As noted earlier, the lower main terrace ends upstream in a moraine, which is at a much lower level than the uppermost terrace on the north side of the river, so the river must have cut down to the level of the second main terrace before the moraine was deposited, the vertical erosion taking place during an interglacial interval. It must be noted, however, that the cutting down of a terrace is often associated with glacier advance in areas outside those where glaciers operate, but this case is within the glacial area and the cutting down has been interglacial. The formation of terraces during a glacier maximum is frequently credited to the lowering of the sea-level and consequent stream base-level when the water has been taken from the ocean to provide the ice-cover of the land, but the terrace formation in the case of the Canterbury rivers in their course across the plains can hardly be attributed to eustatic changes, for the terraces are all lowest near the sea and rise steadily when followed inland. If they were due entirely to change in sea-level this should not be the case unless the change were accompanied by a tilting movement along an axis parallel to the western edge of the plains, and of this there is no evidence at present available, although the removal of the ice-cover and the consequent crustal adjustment has been mentioned as a possible explanation.

It is therefore probable that the river had cut down its channel to the level of the second main terrace during an interglacial interval and that there was a subsequent advance of the ice, certainly as far as the moraine inside the gorge, and perhaps much further. The accumulation of large angular masses on the surface and flanks of this terrace certainly suggests the proximity of ice. It seems hardly possible that they could not have been moved to their present positions from an ice-front near the mouth of the gorge by water alone. It is admitted that they have been subject to strong water action, but this could have taken place as the ice retreated from a more forward position. The size of the masses suggests that the ice passed the site of Mr. McIlwraith's farm, and there is also the accumulation of blocks up to seven feet in diameter just beyond Moorhouse's Road. Could a stream of the size of the Rangitata or perhaps a little bigger have moved and concentrated such blocks when it was entirely unconfined and free to wander over the surface of a plain with an inclination of about 50 feet per mile? It is possible, but I hardly think it probable. I therefore conclude that the glacier extended about six miles from the mouth of the gorge.

I should like to refer very briefly to the evidence as to the maximum extent furnished by the other glaciers that deployed on to the western margin of the Canterbury Plains during the Pleistocene period. This is most complete in the case of the Rakaia, since undoubted moraine, travelled blocks, boulder clay with scratched stones, and rounded hills composed of gravels, occur some six miles out on to the plains from the outlet of the gorge of the river. This glacier was larger than the others since it was fed by a greater length of the snow-clad main divide, and also the outlet of the valley was wider than the others and this allowed freer egress for the ice-stream.