¹ Three inscriptions were found in Temple 9 at Hatra (Sumer, XI, Pt. 1, p. 39Google Scholar). In the first text only the consular date A.D. 235 survives. The other two record dedications by Q. Petronius Quintianus, tribune of the Legio I Parthica, in command of the Cohors IX Maurorum, which bore the title Gordiana; these texts must accordingly be dated to the reign of Gordian III (A.D. 238–44). A milestone of Trajan was reported from the village of Karsi, on the north side of Jebel Sinjar, and published by Cagnat in 1927 (Syria, VIII, p. 53Google Scholar); and one of Severus Alexander was discovered 5 km. south-west of Sinjar in 1951 (Sumer, VIII, Pt. 2, p. 229Google Scholar; see p. 220, n. 20 below). There are a number of Greek inscriptions from Nineveh, and a few meaningless graffiti from Hatra, but none has any specific connection with the period of Roman occupation; the only considerable inscription from Nineveh is almost certainly at least two centuries earlier (Thompson, Campbell, Archaeologia, LXXIX, pp. 140–42Google Scholar).

² Both sites were visited by Sarre and Herzfeld, who published a sketch plan of Ain Sinu under the name of Ain al Shahid, now unknown locally (Archaeologische Reise im Euphrat und Tigris Gebiet, II, Fig. 283) They identified the site as Zagurae on the basis of the distances given in the Tabula Peutingeriana (Weltkarte des Castorius, ed. Miller, K., Segmentum XI, 5Google Scholar). No evidence of the ancient name was obtained during the recent excavations, but there is no reason to doubt this identification.

³ Geographical Journal, CXXII, Pt. 2, pp. 195–97, Figs. 3–5Google Scholar.

⁴ Geographical Journal, XCII, Pt. 1, p. 62 and XCV, Pt. 3, p. 428Google Scholar.

⁵ These underground channels (Arabic qanat) are constructed by linking the bottoms of a series of vertical shafts sunk from the surface of the ground; the line of a disused qanat can often be detected by the depressions which mark the blocked shafts. They are most widely used at the present day in Iran, where the systems are of high antiquity, but in their nature difficult to date precisely. They occur in Northern Iraq in the districts of Kirkuk and Erbil east of the Tigris, at Eski Mosul on the river, and in the Sinjar region, and are reported by Poidebard (La Trace de Rome dans le Desert de Syrie, Pl. CXLV) from Jebel Cembe just across the Syrian border, where they are said to be associated with structures of the Roman period. It has been suggested by ProfessorLaessøe, J. (J.C.S., V, pp. 21–32Google Scholar) that a somewhat obscure passage in Sargon's account of his Eighth Campaign, in Urartu in 714 B.C., refers to an Urartian system of irrigation by qanat, and that the technique of construction was copied by the Assyrians. Although there is no evidence that irrigation by this means was ever practised in Assyria, where dry farming was in any case the rule, it is significant, as Laessøe points out, that the water supplies of Nineveh, Erbil and Nimrud (Calah) were all carried at various points through tunnels excavated by the method characteristic of the qanat-bulders. Moreover all three tunnels are the work of Sargon's immediate successors; the water supplies of Nineveh and Erbil were organised by Sennacherib (Jacobsen, and Lloyd, , Sennacherib's Aqueduct at Jerwan, O.I.P. 24Google Scholar; Safar, , ‘Sennacherib's Project for supplying Erbil with water,’ Sumer, III, Pt. 1, pp. 23–25Google Scholar), and a recent survey has shown that Esarhaddon substituted a tunnel for the original open channel at the head of Aššur-naṣir-pal's canal bringing water from the Upper Zab to Nimrud (‘The Negoub Tunnel Inscription,’ L.A.R. II, pp. 278–79Google Scholar). At all events the technique was known in Northern Mesopotamia long before the Roman period, but we cannot be sure whether its adoption in the Sinjar was due to Assyrian or to Achaemenian precedents.

⁶ The khan at Al Khan, about 25 km. east of Beled Sinjar on the caravan track, has a remarkable sculptured gateway with reliefs depicting St. George slaying the dragon, and an inscription recording its erection by Badr-ad-Din Lulu, A.D. 1234–58 (Iraq, V, Pt. 2, pp. 150–51Google Scholar and Pl. XXIII, Figs. 9–11).

⁷ The use of mortar containing a high proportion of ash, which gives it a distinctive dark grey colour, seems to have been introduced in the Islamic period. It is typical of masonry of the Atabeg and later periods in the north, but does not occur in the Roman masonry here or at Sinjar, or in a church of the sixth century at Qasr Serij, about 60 km, north-west of Mosul.

⁸ The rows of barrack-rooms correspond approximately in size and arrangement with the barrack blocks in the northern half of the fort at Chesters on Hadrian's Wall (Bruce, Collingwood, Handbook to the Roman Wall, 10th edition, ed. Richmond, I. A., p. 90Google Scholar and plan, p. 83). At Chesters each block consisted of ten rooms, opening on to a portico, with a more complex suite of officers' quarters at one end, and each accommodated two squadrons of auxiliary cavalry. This suggests the possibility that of the eleven pairs of barrack rooms in each block at Ain Sinu, ten served as contubernia or messes for other ranks, while the end pair, although superficially identical, housed the officers of the unit. It must be admitted, however, that we know of no parallel for such an arrangement, and can find no plausible explanation for the pairs of intercommunicating rooms; it is possible that one room of each pair served as a stable, but we found no material evidence to support this, or any other, hypothesis. It is strange, too, to find the barrack rooms separated so widely from the smaller store-rooms or armouries which were obviously designed to complete the accommodation of each unit, but we must conclude that the intervening space was an important feature in the layout, perhaps intended for the picketing of horses or as an exercise ground. The question of the purpose for which the camp was built is discussed more fully below (p. 218).

⁹ During a visit to the local police post we observed an exactly similar mud-brick pedestal on the right of the guard-room doorway, with two posts set in its upper surface which formed the ends of an arms rack. We did not find any pedestals of this sort in the other excavated chambers, and it is possible that this pair of rooms near the west gate did in fact house the guard.

¹⁰ Corbels of almost identical design were found during the excavation of the north gate of Beled Sinjar in 1955, but the pattern is so simple that it would be dangerous to draw any chronological deduction from this similarity.

¹¹ The Mesopotamian origin of pitched brick vaulting has recently been discussed by Perkins, J. B. Ward, in The Great Palace of the Byzantine Emperors, Second Report, ed. Rice, D. Talbot, pp. 93–95Google Scholar. It was essentially a mud-brick technique; examples in this material date from the second millenium B.C. to the present day, and are found in the Parthian period at Seleucia, Aššur and Hatra. Hatra has also produced a mortared stone vault built on the same principle, but the use of baked brick in vaulting of this type appears to be an innovation due to Roman influence; there are two instances in buildings of the Roman period at Dura, which must be approximately contemporary with the castellum at Ain Sinu (Dura-Europos, Preliminary Report, Sixth Season, pp. 84 and 266).

¹² The flimsy construction and irregular layout of the buildings exposed in the comparatively small excavated area of the castellum are in striking contrast to the massive walls and regular plan of the barracks, and would not in themselves be out of place in any ancient or modern village in the countryside.

¹³ One of the sherds, AS 125, bore three letters, roughly incised and underlined, which appear to be the Greek CEII, perhaps the beginning of a personal name such as Septimius.

¹⁴ The coinage of the Resaina mint has been definitively studied by K. O. Castelin, Numismatic Notes and Monographs, No. 108, whose dating we have followed. The vexillum issues clearly indicate that this was the base of the III Parthica at least as early as Caracalla, and Castelin rightly remarks that it was probably stationed here to watch Rome's somewhat unreliable allies in Osrhoene as well as the eastern frontier; Osrhoene was in fact annexed under Caracalla. A city wall of Resaina has recently been pubhshed (McEwan and others, Soundings at Tell Fakhariyah, O.I.P. LXXIX, pp. 14–17, Pls. 10–12, 24–27), but its close resemblance to the enceintes of Diyarbekr and Sinjar must date it to the fourth century A.D.

¹⁵ Bellinger, A. R. (The Excavations at Dura-Europos, Final Report VI, The Coins, p. 208Google Scholar) has shown by comparison of the dies that these coins were actually struck at Nisibis for use at Singara. It is probable that in the Severan period at least a detachment of the Legio I Parthica was stationed at Singara, with perhaps anothet detachment at Nisibis, the capital of the province (Pauly-Wissowa, Real-Encyclopadie der Klassischen Altertumswissenschaft, Vol. XII, s.v. Legio); certainly the I Parthica was in Singara, together with the I Flavia, when it was captured by Shapur II in A.D. 359 ((Ammianus Marcellinus XX, 6, 8).

¹⁶ Poidebard, , La Trace de Rome, p. 144Google Scholar and Pl. CXXII; Syria, IX, p. 219Google Scholar.

¹⁷ Poidebard, , La Trace de Rome, p. 150Google Scholar and Pl. CXXXIX. The location of these two sites is shown on Poidebard's map of the Khabur region, Pl. CXL.

¹⁸ Ancient descriptions of parts of the desert south of Ain Sinu, in particular the Tharthar basin around Hatra, make it clear that it has changed little in 3000 years, and its inhabitants must have been almost entirely nomadic (in the ninth century B.C., Annals of Tukulti-Ninurta II, L.A.R.I, p. 128, para. 407; in the fourth century A.D., Ammianus Marcellinus XXV, 8, 6). We have no information about the northern plain, south of Nisibin, in the third century A.D., but Segal, J. B. (‘Mesopotamian Communities from Julian to the Rise of Islam,’ Proceedings of the British Academy, XLI, p. 119)Google Scholar has shown that its population was in part semi-nomadic in the Late Roman period.

¹⁹ XXV, 8, 7.

²⁰ The only known milestone of this period, with an inscription of Sevetus Alexander, was found in 1951 5 km. south-west of Sinjar (Sumer, VIII, Pt. 2, p. 229Google Scholar), obviously marking a route from the Middle Khabur; this is almost certainly a continuation of the road reported by Poidebard (La Trace de Rome, pp. 156–57) which led from Circesium through Fadghami in the direction of Sinjar, but a search on the ground revealed no physical trace of its course, and Poidebard states that it was an unpaved road. Certainly there is no evidence that the Romans built roads in Northern Iraq in the manner and on the scale of the great highways of the western provinces.

²¹ The Cohors I Augusta Parthorum, stationed in Mauretania, seems to have been in existence for at least a century before this date, and the other known eastern units come from Palestine, Syria or Osrhoene (Pauly-Wissowa, Real-Encyclopädie, Vol. III, s.v. Cohors).

²² VI, 7, 8.

²³ We are indebted to Professor Richmond for a very helpful discussion of this problem, and especially for the suggestion that the barracks may have been an initial training centre for recruits.

²⁴ The Roman site at Tell Hayal was situated in the ploughland about half a kilometre east of the tell and the modern village, where the debris of a massive stone wall, probably the exterior wall of the castellum, can be traced running from east to west for a distance of over a hundred metres. Among the piles of stones collected by ploughmen near the site was found, in 1956, a brick bearing the stamped inscription COH (ORS) VI. I (TVRAEORVM). The brick was rather roughly made (c. 26 by 24 by 3 cm.; stamp 12 by 3 cm.) with finger grooves entirely covering the stamped surface; the other side was plain. We are greatly indebted to the Directorate General of Antiquities, and to Sayid Kadhim al Jenabi to whom the find was first reported, for information about this discovery. The Cohors VI Ituraeorum was not previously known, although its existence had been inferred from the VII Ituraeorum, stationed in Egypt (Pauly-Wissowa, Real-Encyclopädie, Vol. III, s.v. Cohors).

²⁵ Tell Ibra was already known as a Parthian site of a somewhat earlier period, owing to the discovery there of an Aramaic building inscription dated to the year 116 B.C. (Sumer X, Pt. I, p. 145Google Scholar, Arabic text). The identification of the ancient names of sites along the frontier is in all cases based on topographical considerations and the distances given on the Tabula; see Geographical Journal, CXXII, Pt. 2, pp. 197–98Google Scholar.

²⁶ We wish to thank the Directorate-General of Antiquities, Baghdad, for their kind permission to examine the unpublished pottery and small finds from Hatra that are in store in the Iraq Museum.

²⁷ Baur, P. V. C., The Excavations at Dura-Europos, Final Report IV, The Lamps, 1947Google Scholar; no. 401. R. Campbell Thompson, “The Excavations on the Temple of Nabu at Nineveh,” Archaeologia 79, Pl. LV: 216 (VI, 6), 217.

²⁸ Dura, Lamps, no. 303. The moulded ornament on a fragment of a light buff lamp found in the north-east circular tower of the castellum at Ain Sinu is very like Dura no. 287, but with a laurel wreath rather than a rosette pattern on the rudimentary discus.

²⁹ Ibid., no. 408 and p. 84. The tendril decoration on the Dura lamp does, of course, suggest a somewhat later date than the Ain Sinu lamps, but the general similarity of the two types is so close as to suggest some relationship between the two. Waagé, , Antioch I, pp. 65 fGoogle Scholar. Waagé writes, “The presence of the handle, quite lacking in the Roman lamps of the second and third centuries at Antioch, is the surest proof that the type did not evolve there.”

³⁰ Waagé, , Antioch I, p. 63Google Scholar. For a general discussion of heart-shaped lamps see Walters, A. B., British Museum Catalogue of Lamps, pp. 167–181Google Scholar.

³¹ Dura Lamps, nos. 348–400 and p. 83. R. Campbell Thompson, op. cit. Pl. LV. Debevoise, N.C., Parthian Pottery from Seleucia on the Tigris, 1934, pp. 119–129CrossRefGoogle Scholar.

³² Dura, Lamps, no. 421.

³³ For an extremely useful discussion of the Dura green glazes, see N. Toll, The Excavations at Dura-Europos, Final Report IV, Pt. 1, The Green Glared Pottery, 1943, pp. 1–5Google Scholar, and pp. 81-95 (“Technological Notes on the Pottery” by Frederick R. Matson).

³⁴ Dura, Green Glared Pottery, type XI–F, fig. 28, cf. Ain Sinu 8, 14–17; Dura type XI–H, fig. 29, cf. Ain Sinu 5, 6, 10–13.

³⁵ Cf. ibid. fig. 20, D. E.

³⁶ Cf. ibid., fig. 10.

³⁷ One sherd from the neck of what appears to have been a similar type of jar was found at the north gate of the camp, and a single sherd from Hatra may be of this type.

³⁸ H. Goldman, Tarsus I, Pl. 164, B. The usual Ain Sinu ornament is applied more in the manner of Pl. 168, C.

³⁹ AS 124: Extant ht. 35·5 cm., rim diam. 13·5 cm. Very coarse flaky reddish clay; body decorated with four bands of irregular wavy comb incision. III, west round tower, inside outer wall, depth 40 cm.

A very early Parthian example of incised wavy comb ornament was found at Nimrud, Iraq XX, Pt. 2, Pl. XXI, 23.

⁴⁰ Tarsus I, Fig. 162: 797, and figs, 204–6. cf. also the ‘Roman Pergamene’ cauldron, fig. 160: 758. I am indebted to Mr. F. R. Matson for the information that the Ain Sinu cooking ware is identical with that found at Dura.