Less than a hundred years later Mutawakkil (A.D. 847-861) built the mosque of Abû Dulaf, which is closely related in plan to Manṣûr’s mosque at Raqqah (Fig. 33).[422] There is the same enclosing wall of sun-dried brick garnished with [Illustration: FIG. 33. Mosque of Abû Dulaf.] rounded towers. The arcades are of burnt brick, but the central aisle of the ḥaram and the corresponding aisle of the north riwâq are more than a metre wider than the others (7·33 metres as against an average of 6·20 metres), and a transept 10·40 metres wide runs along the qiblah wall. Dr. Herzfeld informs me that he has by excavation ascertained the existence of a miḥrâb in the centre of the qiblah wall where I had placed a door. In one respect Abû Dulaf differs from all other mosques built with piers; the arcades of the south riwâq, instead of lying parallel to the qiblah wall, are placed at right angles to it. I do not think that this variation is of great importance, for the outer and inner arcades (that is to say, the arcade on the ṣaḥn and the arcade next to the qiblah wall) are placed parallel to the qiblah wall, and it is only between these two that the ḥaram arcades run at right angles. The divergence from the normal scheme is not therefore so great as would at first appear. The mosque is surrounded by an outer enclosure, or ziyâdah[423], within which stands the spiral minaret, to the north of the centre of the north wall. The piers and arches of the riwâqs must undoubtedly have carried a flat wooden roof; nevertheless in the façades of the ṣaḥn each pier is adorned with the blind niche which I believe to be derived from the tubular system of Mesopotamian vaulting (Plate 89, Fig. 2). This decoration is a direct link between the unvaulted mosque of Abû Dulaf and the vaulted palace of Ukhaiḍir, and the fact that it appears again in the mosque of Ibn Ṭulûn is to my mind an indubitable proof of the essential exactitude of the tradition which connects Ibn Ṭulûn’s building with Mesopotamian architecture (Plate 89, Fig. 1). Other structural evidence is not wanting. The position of the minaret in the northern ziyâdah (to say nothing of its spiral form) corresponds with the position of the minarets both at Abû Dulaf and at Sâmarrâ, and even if we leave on one side the much-disputed question of the origin of the stucco ornament in the Cairo mosque, there is another feature of its decoration which points directly to Mesopotamia. The walls are crowned with a fantastic parapet, which probably goes back, in design at least, to the ninth century, and below the parapet, just above the level of the roof, runs a decorative band consisting of a recessed square pierced by a circular hole (Plate 91, Fig. 1). The same motive appears upon the walls of the Sâmarrâ mosque, with this difference, that it is placed below the level of the roof and not above it (Plate 91, Fig. 2). Instead of forming part of a light parapet, it forms part of the solid wall;