Occasionally, the method by which we brought our conjecture to the test had to be applied to another kind of difficulty, and a choice had to be made between two alternative tracks, each clear and each with something in its favour; in one or two short and rare examples, the Road very plainly went one way, when by every analogy and experience of its general course it should have gone another. But, take the problem as a whole, whether applied to the commonest example, that of forming an hypothesis and then finding whether it could be sustained, or to the most exceptional (as where we found the Road making straight for a point in a river where there was no ford), the general method was always the same: to consider the geographical and geological conditions, the analogy of existing trails of the same sort, the characters we had found in the Road itself throughout its length as our research advanced, and to apply these to the parts where we were in doubt. The number of 'habits,' if I may so call them, which the Road betrayed, was much larger than would at first have appeared probable, and that we could discover them was in the main the cause of our success, as the reader will see when I come to the relation of the various steps by which we reconstituted, as I hope, the whole of the trail. The principal characteristics or 'habits' are as follows:-- I. The Road never turns a sharp corner save under such necessity as is presented by a precipitous rock or a sudden bend in a river. In this it does precisely what all savage trails do to-day, in the absence of cultivated land. When you have a vague open space to walk through at your choice you have no reason for not going straight on. We found but one apparent exception to this rule in the whole distance: that at the entry to Puttenham; and this exception is due to a recent piece of cultivation.[4] This alignment, however, is not rigid--nothing primitive ever is--the Road was never an absolutely straight line, as a Roman road will be, but it was always direct. II. The Road always keeps to the southern slope, where it clings to the hills, and to the northern bank (_i.e._ the southward slope) of a stream. The reason of this is obvious; the slope which looks south is dry. There are but four exceptions to this rule between Winchester and Canterbury, and none of the four are so much as a mile in length.[5] III. The road does not climb higher than it needs to.