Our course lay somewhat west of due north, over a bare and sterile country. Small hamlets, surrounded by a narrow belt of cultivation, were seen at rather wide intervals; and, save a few olive trees near the houses, we did not pass a single tree during the day. _Artemisia Herba-alba_, and _Chenopodiaceæ_ of the _Salsola_ tribe, were the prevailing plants, indicating the presence of gypsum and of soluble salts in the soil. About three o’clock we approached a large village, with a massive square _kasbah_, and soon after, following a dry watercourse, entered the singular defile which, as well as the village, is known by the name Aïn Tarsil. It has evidently been formed by erosion from the limestone strata which dip slightly towards the south. The surrounding country here shows a hilly undulating surface, unbroken by any marked inequalities; but the stream, though dry in ordinary weather, has cut a trench from two to three miles in length, and from thirty to fifty feet in width, between steep walls of rock about equal in height to the width of the trench. M. Balansa, the only traveller who is known to have passed through this defile, must have been more hurried than we were, as he does not speak of the numerous rare and local plants which grow upon the rocks, most of them, indeed, the same that we had found on the rocks about Seksaoua. But he could not avoid being struck by the singular excavations in the rock, evidently used at some remote period for human habitation, which extend at intervals along both sides of the defile. In some cases there may have been a natural recess in the rock, afterwards artificially enlarged; but the majority appear to be altogether the work of human hands; and in most of them, where the entrance had become difficult owing to the breaking away of pieces of rock from the edge, this was afterwards made good by building up a bit of loose wall of irregular blocks of stone. The height of the entrance does not exceed four feet, and is often less. The most singular point about these dwellings is the fact that they are all near the top of the cliff, where the rock is nearly vertical, in positions that cannot now be reached without a ladder, or other artificial assistance. It might be suggested that since these prehistoric dwellings were abandoned, the work of erosion has deepened the trench, and thus increased the difficulty of access; but unless we suppose that during the same period the climatal conditions have been profoundly modified, this seems a highly improbable explanation. As far as we could afterwards judge, the watercourse running through the defile receives the drainage of only a small tract of hill country, and the marks of water action do not extend to the rocks on either side. It is impossible to see these remains without being reminded of the notions current in antiquity as to troglodytes who dwelt in the neighbourhood of the Atlas mountains, and who could run faster than horses;[1] but until the dwellings