The fourth, with some interruption, led along the north downs to Canterbury, whence tracks would radiate to the ports of Kent. A fifth followed the Berkshire Downs and the Chilterns, and so led on to the Wash and earlier parts of the Norfolk coast, which have now apparently disappeared in erosion. The sixth line led with more difficulty (and has been more obliterated by later Roman work) directly westward to the mines of the Mendips, and to the borders of the Severn estuary. It could not take advantage of the chalk beyond Wiltshire, but it had fairly dry going along the ridge of the Mendips. The seventh, it must be presumed, though the traces are largely lost, used the height of the Cotswolds; but here the soil, being oolitic and not chalk, was much less favourable and the extension northward ceased earlier. This system, then, we regard as the earliest of all. ii The second system, as I have said, seems to have been connected with London, but here the later track of the Roman engineers and the continuous development of nearly twenty centuries has left us little to go on save conjecture. There are points in that conjecture, however, which are fairly certain. But there seems to have been, from the earliest time, communication between north and south on the lowest crossing of the Thames. Now, the lowest permanent crossing of the Thames, even before a bridge, was in the neighbourhood of London. The crossing of a river is determined by the hardness of the land upon either bank, as we have seen, more than by any other factor. The lower Thames everywhere had extensive marshes either upon one side or the other, and usually upon both. At Grays, Tilbury, Erith, etc., the hard ground approached right up to one bank, but was always countered by extensive marshes on the other, or by marsh behind gravel, forming a sort of island of hard land which could not be used for continuous travel.