[Illustration: FIG. 8. GROUND-PLAN OF A SMALL CORRIDOR HOUSE FROM FRILFORD, BERKSHIRE. (_From plan by Sir A.J. Evans._)] [Illustration: FIG. 9. COURTYARD HOUSE AT NORTHLEIGH, OXFORDSHIRE, EXCAVATED IN 1815-16. (Room 1, chief mosaic with hypocaust; rooms 8-18, mosaic floors; rooms 21-7 and 38-43, baths, &c. Recent excavations show that this plan represents the house in its third and latest stage. See p. 31.)] [Illustration: FIG. 10. DETAILED PLAN OF PART OF SILCHESTER. Showing the arrangement of the private houses and the Forum and Christian Church. (_From the plan issued by the Society of Antiquaries._) (See p. 31.)] The origin of these northern house-types has been much disputed. English writers tend to regard them as embodying a Celtic form of house; German archaeologists try to derive them from the 'Peristyle houses' built round colonnaded courts in Roman Africa and in the east. It may be admitted that the influence of this class of house has not infrequently affected builders in Roman Britain. But the differences between the British 'Courtyard house' and that of the south are very considerable. In particular, the amount of ground covered by the courts differs entirely in the two kinds of houses, while for the British houses of the plainer 'corridor' type the Mediterranean lands offer no analogies. We cannot find in them either _atrium_ or _impluvium_, _tablinum_ or peristyle, such as we find in Italy, and we must suppose them to be Roman modifications of really Celtic originals. This, however, no more implies that their occupants were mere Celts than the use of a bungalow in India proves the inhabitant to be a native Indian.[1] [Footnote 1: _Vict. Hist. Somerset_, i. 213-14. A few Romano-British houses at Silchester (_in insula_ xiv. (1), see _Archaeologia_, lv. 221) and at Caerwent (house No. 3, see _Arch._ lvii, plate 40) do bear some resemblance to the Mediterranean type, as I have observed in _Archaeol. Anzeiger_, 1902, p. 105. But they stand alone. Similarly, parallels may be drawn between Pompeian wall-paintings of houses and certain 'villa' remains in western Germany, as at Nennig; see Rostowzew, _Archaeol. Jahrbuch_, 1904, p. 103. But these again seem to me the exception.]