T. E. Lawrence’s Undergraduate Dissertation – Page 104

so typically Byzantine there can be no possible hesitation in dating it before the arrival of the Crusaders : very little of its walls remain. Going further East, the site of Turbessel (Tell Bashar ...) is entirely laid waste. There had been once a little castle on a mound, but only a few stones of a square tower are left. At B... (Biradjik ...) on the Euphrates is a fortress, which would be typically Byzantine if the Arabs under Malek es Zahir had not rebuilt the high towers looking south- ward : and the huge castle of Edessa (Urfa ...) is also Byzantine with Arab additions. The rock moat, over 500 feet long, with an average depth of 40 feet, and a width of 30 feet is too huge a work for the Latins ever to have undertaken in their insecure tenure of the place : besides we have town walls tolerably perfect, and unquestionably Byzantine, and the Crusaders appear to have maintained a semi-Byzantine administration during the few years they held the province. It is important, having regard to other castles in Northern Syria, to notice the pier of rock left standing in the moat when the rest was excavated : its purpose was to support the centre of a timber bridge, and to make it high enough it was necessary to cap it with masonry : the only large wood to be found in Edessa to-day is the poplar-tree, and a long beam of this is quite untrustworthy, even in the far less trying strain of the roof of a native hut. In Europe a draw-bridge