The effect of the river as a barrier must, of course, have largely depended upon the level to which the waters rose in early times. It is exceedingly difficult to get any evidence upon this--first, because however far you go back in English history some sort of control seems always to have been imposed upon the river; and secondly, because the early overflows have left little permanent effect. As an example of the antiquity of the regulation of the Thames we have the embankment round the Isle of Dogs, which is Roman or pre-Roman in its origin, like the sea-wall of the Wash, which defends the Fenland; and at Ealing, Staines, Abingdon, and twenty other places we have sites probably pre-historic, and certainly at the beginnings of history, which could never have been inhabited if the neighbouring fields had not been drained or protected. The regularity of the stream has therefore been somewhat artificial throughout all the centuries of recorded history, and the banks have had ample time to acquire consistency. It is certain, of course, that works of planting, of draining, or of embankment, which required continuous energy, skill, and capital, decayed after the coming of the Saxon pirates, and were not undertaken again with full vigour until after the Norman Conquest. Even to-day the work is not quite complete, though every year sees its improvement: we are still unable to prevent regularly recurrent floods in the flats round Oxford and below the gorge of the Chilterns; but for the purpose of this argument the chief fact to be noted is that no serious interruption to the approach of the river seems to have existed in historic times. In pre-historic times many stretches of the river must have afforded great difficulties of approach. The mouths of the Ock, the Coln, the Kennet, the Mole, and the Wandle must each have been surrounded by a marsh; all the plain between Oxford and the Hinkseys must have been partially flooded, as must the upper reaches between Lechlade and Witham (on one side or the other of the stream as it winds from the southern to the northern rises of land), and as must also have been the long stretch of the right bank below Reading. The highest spring tides may have been felt as high up the stream as Staines, and both the character of the surface and the contour lines permit one to conjecture that the valley of the Wandle and several other inlets from the lower river were flooded. Yet it is remarkable that in this alluvium, more disturbed and dug than any other in Europe, little or nothing of human relics, of boats, or of piles has been discovered, and this absence of testimony also points to the remoteness of date from which we should reckon the human control of the river.