The author regrets that the circumstances in which this excavation was conducted make it impossible to provide, in addition to the important material presented, all the technical information which is taken for granted from current excavations.

² It is impossible to be consistent in using either the ancient or the Arab names of towns and cities in Cyrenaica because neither system is internally consistent and the Italian occupation of Libya also had its impact upon standard usage. In the Roman period the port of Cyrene is known as Apollonia, but in the Byzantine town list (the Synecdemos of Hierocles dated 527–528 A.D.) it appears with the name Sozusa, the old name having been discarded for its connection with a pagan cult. Today the town is called either Susa or Apollonia; archaeological reports, however, usually retain the name Apollonia. Throughout this report we shall follow the practice set by most twentieth century authors for both choice of name and its spelling: Apollonia rather than Sozusa or Susa; yet the present Derna rather than ancient Darnis; Latrun rather than Erythrum or even El Atrun.

³ All of this coast lies along an ancient fault line recently established with some degree of accuracy by oil company geologists. We know Cyrene suffered a major earthquake in 365 A.D. The tidal wave from the quake is said to have swept seagoing vessels on to the rooftops of houses as far away as Alexandria. Some time after the Byzantine period (but just when is uncertain) a good portion of the harbour at Apollonia ‘fell’ into the sea. This same medieval disaster seems to be reflected at both Teuchira and Ptolemais. It is quite possible that the coastal area of Latrun was somewhat altered at the same time and that the present shoreline configuration is the result of such late fault activity.

Harrison, R. M. discusses briefly the geography of the Bay of Marsa Hilal in his article ‘A Sixth Century Church at Ras el Hilal in Cyrenaica’, PBSR, xxxii (1964), 1 ff.Google ScholarJones, G. D. B. and Little, J. H. in ‘Coastal Settlement in Cyrenaica’, JRS, lxi (1971), 64–79Google Scholar, offer a description and interpretation of the economic function of the ancient port cities of Cyrenaica, including Latrun.

⁴ Beechey, F. W. and Beechey, H. W., Proceedings of the Expedition to Explore the Northern Coast of Africa from Tripoly Eastward (London, 1828), 478Google Scholar.

⁵ Ptolemy, , Geography, IVGoogle Scholar.

⁶ The Stadiasmus ap. Müller, G.Geographi Graeci Minores (Paris, 1882)Google Scholar.

⁷ Migne, , Patrologia Graeca, LXVIGoogle Scholar; this translation ap. FitzGerald, A, The Letters of Synesius of Cyrene and The Essays and Hymns of Synesius, 2 vols. (London, 1930)Google Scholar. For a longer biliography relating to Synesius cf. Kraeling, C., Ptolemais, City of the Libyan Pentapolis (Chicago, 1962), 23Google Scholar, n. 112.

⁸ Migne, , Patrologia Latina, XXIIGoogle Scholar.

⁹ Cf. Honigman, E., ‘The Original Lists of the Members of the Council of Nicaea, the Robber-Synod and the Council of Chalcedon’, Byzantion xvi (1942–1943), 20 ff.Google Scholar As regards the bishops of Erythrum, Pius Bonifacius, Gams, Series Episcoporum Ecclesiae Catholicae (Graz, 1957)Google Scholar, lists the following: ‘Orion, ex. saec. 4—Sabbatius—Paulus, temp. Theophili—451, Theophilus’. This Theophilus of 451 obviously contradicts the Honigman compilation for Chalcedon. But then, Gams ignores completely Gemellinos. Although the entire clergy of Cyrenaica was nominally subject to the See of Alexandria from the time of the Council of Nicaea onward, the majority of Cyrenaican bishops must have bowed to the temporal power of the dux and the spiritual power of his presumed ally in effecting policy, the resident metropolitan bishop of the capital.

¹⁰ Cf. Goodchild, R. and Reynolds, J., ‘Some Military Inscriptions from Cyrenaica’, PBSR, xxx (1962), 45–6Google Scholar, where an inscription from Ain Mara is discussed.

¹¹ Goodchild, R., ‘Mapping Roman Libya’, Geographical Journal, cxviii (1952), 150Google Scholar or Libyan Studies: Select Papers of the Late R. G. Goodchild, ed. Reynolds, J. (London, 1976), 145 ff.Google Scholar

¹² The 1960 excavation was in cooperation with the Cyrenaican Department of Antiquities which provided heavy equipment (including a Decauville system) and the payroll for the local labour. My direction of the operation was undertaken for New York University and made possible by a grant from the Charles and Rosanna Bachelor Memorial Foundation. Funds for the second campaign of 1962, also in cooperation with the Cyrenaican Department of Antiquities and for New York University, were provided by New York University directly. Travel to the site in 1965 was afforded by a grant from the Research Council of Rutgers, the State University. Expenses on the site were ‘out of pocket’. None of the three campaigns could have been accomplished without the encouragement and friendly help of Richard Goodchild. His passing leaves a great gap in Cyrenaican archaeology.

¹³ Other possible tombs have since been discovered adjacent to the modern Moslem cemetery some distance south of the two promontories. In one of these underground chambers or caves were found four portrait heads carved in limestone.

¹⁴ Two different building periods and alignments were revealed; however, the evidence was insufficient to establish whether the second building phase totally eliminated the walls of the first. The much decayed remains of a water channel cut through the building on a northeast-southwest angle and both phases of construction seemed to accommodate it.

¹⁵ There were no columns in front of the apse haunches to carry a triumphal arch screen, a feature found at Basilica B, both excavated churches at Cyrene, Apollonia's three churches, and several rural Cyrenaican churches.

¹⁶ Fragments of two reliquary boxes were found in the antechamber and baptistery room itself. Undoubtedly these represent the robbed boxes from the converted font (the measure of the two boxes is correct: the fragments of the more complete box establish its size as 44 × 34 × 40 cms.; that of the other cannot be calculated exactly but it would be even smaller).

¹⁷ I do not know a single parallel to this arrangement of flanking rooms and antechambers throughout the whole of the Mediterranean. One has either fully developed rooms flanking an inscribed apse or well defined rooms flanking the chancel should the apse be exposed, but never both kinds of rooms in series. Latrun's antechambers, which appear in Basilica B as well as A, could be derived from a tripartite or even less differentiated transept form. The careful adjustment of the chancel enclosure to these spaces reinforces this suggestion. Certainly they are withdrawing rooms for the clergy during the mass and probably functioned additionally as standard prothesis—diaconicum rooms.

¹⁸ Against my own earlier view in Arch. Anz., (1962) 433–4Google Scholar. Composite Ionic volute—impost block capitals generally are reserved for gallery support systems, yet St. John at Ephesos (if correctly restored!) is a marked exception. Cf. Forschungen in Ephesos, IV, 19 ff. and 157 ff.Google Scholar; also Kautzsch, R., Kapitellstudien (Berlin and Leipzig, 1936), 90 ff. and 176 ff.Google Scholar, and Krautheimer, R., Early Christian and Byzantine Architecture (Baltimore, 1965)Google Scholar pl. 84A. When building with spolia no ’rule’ would pertain; nor would it should the quarry supply of fittings be short.

¹⁹ On the other hand, no sets of voussoirs came to light for either Basilica A or B. Does this curious fact mean these arches were systematically robbed from the ruins of the churches shortly after their destruction for subsequent building operations in the area?

²⁰ Neither H. Demetrios nor the Acheiropoeitos churches in Salonica help us since the apse vault of H. Demetrios, a basilica which always has had a clerestory, springs from the line between the nave arcade and gallery zones, while that of the Acheiropoeitos, a basilica which now lacks a clerestory, springs from a line approximating the gallery capitals. Restoration sections of both these churches have the apse vault springing from a point just above the tops of the gallery parapets. Cf. Orlandos, A., Basilike (Athens, 1952)Google Scholar.

²¹ The two column bases in situ of the right side aisle arcade have cuttings to hold screens; the marble footing for the screen between them is still preserved. The respond on the west of this same arcade has a short slot to receive a screen (possibly late?), but none of the other responds does. Therefore, passage from nave to aisles is assured for at least three points.

²² The upper levels of the two rooms flanking the apse might similarly have comprised low projecting elements separately roofed. But here, as well, the corner masonry is too thin to allow real towers.

²³ The chancel enclosure for Basilica B, if not also A, was similar to that Orlandos reconstructs for the cruciform church at Thasos. Cf. Hoddinott, R. F., Early Byzantine Churches in Macedonia and Southern Serbia (London, 1963) 180CrossRefGoogle Scholar, fig. 91. Except for Basilica B, chancel enclosures with lintels are unknown in Cyrenaica.

²⁴ Another fragment from an unidentifiable feature is a strange capital cut from local stone and found in the room flanking the apse on the north. It has a square abacus on a cylindrical drum. The drum is stuccoed and has a red painted lozenge pattern with dots in the centres. The date of this capital, too, seems late.

²⁵ The peristyle of the palaestra of the Gymnasium at Salamis in Cyprus has pavings with these same patterns, also worked as carpets and executed in local stone. The Gymnasium was converted into a public bath some time in the fourth or fifth century and its pavings date from this late phase. Cf. Nicolaou, K., Ancient Monuments of Cyprus (Nicosia, 1968), 24–5Google Scholar.

²⁶ Because Basilica A was laid out on a small rise of rock and earth, it was sometimes necessary to cut into the bedrock for the footings of the walls. As a result, the plinth of the exterior walls is often only a single course in height. All of the exterior walls perhaps rest on solid bedrock.

²⁷ As has been stated in the discussion of Basilica A's elevation the exact height of the springing of the apse half dome cannot be established.

²⁸ The span between antechambers is too great to easily accommodate a barrel vault over the chancel area although the walls themselves are substantial enough. Examples of barrel-vaulted chancels are known in Cyrenaica; however, they pertain to types quite different from that of Latrun's basilicas.

²⁹ A new door may have been cut in the wall shared by the nave and the narthex and the original door blocked completely. At any rate, the central portion of the narthex now became a closed room.

³⁰ It is most unlikely Erythrum was ever a walled town, not even in the late sixth century when it was called ‘polis’ (if, indeed, the Ain Mara inscription does refer to Erythrum). Cf. above and n. 10.

³¹ The size and upward extension of the blocks forming the jambs of the lateral doors suggest this configuration. But the evidence is by no means conclusive.

³² It is interesting to speculate if the contents of these tombs might have supplied the relics for the boxes buried in the converted baptismal font of Basilica A (see above).

³³ The style and patterns of these carved cover slabs are unique among the decorative elements of Basilica B. Their date could be any time from the construction of Basilica B to its destruction, since no exact time can be assigned to the creation of this tomb chamber (nor for that matter, the other two). Because the material of the carved slabs is sandstone, it can be assumed they were locally executed.

³⁴ After their investigation, the south aisle tomb chambers were refilled with earth and their cover slabs replaced. As a result these two tombs are not visible today.

³⁵ This cistern was not investigated due to the need to destroy part of the nave pavement for access.

³⁶ It was the eroded top of the southern marble jamb and a fragment of the lintel which offered the only surface evidence for the existence of an ancient building prior to the excavation of 1962.

³⁷ Nonetheless, the talus is somewhat later than the construction of the building since the plaster surface of the north wall continues behind it.

³⁸ Could this be simply a difference in the treatment of the same design by individual carvers, or does it, indeed, reflect two designs? Masons' marks (recorded in Appendix II) are not helpful in deciding the issue. Exact parallels for both types can be found at various Mediterranean sites; sometimes both types appear at one site.

³⁹ Bases and shafts were doweled; so, too, shafts and capitals. But curiously, impost blocks were simply set upon the capitals. More extraordinary, however, was the practice of providing a cushion of lead between bases and shafts. The upper surfaces of the bases were actually channeled to allow any excess of lead to run off after the join had been made. At least half a dozen of these lead disks were found complete in the course of the excavation. The same procedure may have been used during the construction of Basilica A. Since here only small fragments of lead survived, any such interpretation of these pieces was impossible at the time of A's excavation.

⁴⁰ Cf. above and n. 19.

⁴¹ The string course created a physical as well as visual separation. The gallery floor beams must have been fixed immediately below the undersides of the string course blocks (the blocks themselves do not have beam cuts) thereby allowing the upper surface of the floor makeup to be flush with the upper surface of the blocks. This arrangement can be inferred from the setting of the parapet screens as indicated by the full vertical channelling of each side of the plinths. The screens would have been held by these channels and footed on the upper surface of the string course. Not only did the string course look like the division between zones, but it actually was this division since it corresponded to the position and thickness of the gallery floor. The floor makeup was mortared rubble, really a crude kind of concrete, and remnants of this material overlayed the aisle floors. The surviving gallery floors at Ras el-Hilal are similar in composition.

⁴² A groove drawn on the string course blocks marks the position of the plinths. The set-back of the gallery supports perhaps dictates a thinner masonry for the upper reaches of the nave wall. Confirming this supposition is the 63 × 63 cm. dimension of the top surface of the gallery capitals.

⁴³ These capitals were not doweled although the column shafts have dowel cuts. Actually, the recessed undersurface of the capitals assured a stable join without dowels.

⁴⁴ These are the columns which today have been restored (see above).

⁴⁵ The screen designs and dimensions are recorded in Appendix II.

⁴⁶ Certain churches of northeastern Syria such as Qalb Louzeh are representative. It should be noted that the type is never associated with an atrium.

⁴⁷ Herein may lie the importance of the Latrun churches. They provide an elevational prototype, a gallery basilica, which is admirably suited for use as a castrum. Apparently such a role was not paramount for either Basilica A or B at the time of their construction and, therefore, considerations of security would not have influenced the choice of church type. But these considerations may have prejudiced certain alterations to the Greek Mainland/Aegean Coastland norm like the elimination of atria, etc., and even aisle windows. These changes already suggest a true castrum and the form demonstrates the effectiveness of a gallery as an enclosed rampart.

⁴⁸ The addition of the talus on the north, as has been said, essentially a measure to shore up the north wall, might have interfered with exterior staircases on that side of the basilica. When exclusively defensive needs increased, any exterior wooden staircases could have been internalised in the aisles, whatever inconvenience this might have caused.

⁴⁹ Mathews, T. F., The Early Churches of Constantinople: Architecture and Liturgy (Pennsylvania State University, 1971)Google Scholar, especially 129.

⁵⁰ The base slab has four roughened areas near its corners which mark the position of the supporting colonnettes.

⁵¹ These larger columns and separate capitals do not seem to have formed a ciborium over the altar since there are no cuttings or dowels in the chancel pavement (preserved complete) to hold such a feature.

⁵² Goodchild claimed a wooden synthronos for the apse of the Central Church at Apollonia on the basis of very real cuttings in the apse wall. But these cuttings more likely stem from squatter occupation.

⁵³ For well-recorded examples of apse burials see Duval, N., Les Eglises africaines à deux absides, I and II (Paris, 1971–1973)Google Scholar. It is important to note that most apses containing burials are rigorously screened from their preceding chancel spaces and that they represent one of two apses within the church structure. Unscholarly musing might interpret Latrun's two basilicas (should Basilica B have had apse burials) as a single, double-apsed unit in terms of functional arrangement. Such an argument would account for the opposed orientation of the two basilicas and link them to the double-apsed church type found at Cyrene and several rural locations in Cyrenaica.

⁵⁴ Certain Greek and Macedonian churches have crypts associated with the apse (i.e. the Cruciform Church at Thasos; the Basilica of Bishop Philip at Stobi) or under the chancel space (H. Demetrios at Salonica; the Extra Muros Basilica of Philippi). These churches are of particular interest because they all include fittings which match some of those found at Latrun. A few Cyrenaican rural churches also have crypts. Because of the possibility that Basilica B might have had an altar tomb or miniature crypt, it was decided to lift the base slab of the altar and probe underneath. Nothing at all was found, but the action caused a rumour to spread that ‘treasure’ was buried deep below. As a consequence unknown persons later conducted a clandestine excavation which broke into several pieces the altar base slab. To curb further destruction, no other soundings in the vicinity of the apse and chancel were undertaken.

⁵⁵ As has been stated above, this might have provided the footing for a wooden staircase to an upper level room.

⁵⁶ Were there tombs beneath the floor which were looted and left open some time before the surrounding superstructure fell in?

⁵⁷ Because of the existence of the cisterns it is hard to imagine there were more tombs in the narthex, even though the cisterns are sufficiently below the destroyed floor to provide space.

⁵⁸ Cf. Appendix I and II.

⁵⁹ The Cyrenaican Department of Antiquities would not authorise the lifting of floor slabs or the endangering of walls to investigate their trenches.

⁶⁰ The full catalogue of all fittings is found in Widrig, dissertation ch. II.

⁶¹ Krautheimer, on the basis of photographs of the fittings, argues from a belief that the Latrun churches are Justinianic. At the same time he states that these very fittings are already old-fashioned. Krautheimer, op. cit., 190–1 and n. 19.

⁶² Those changes which are effected at Latrun perhaps place Basilica A and B in a more or less datable position in the whole evolutionary chain of Cyrenaican church architecture (cf. Widrig, dissertation, ch. II). One change is the envelopment of the chancel by antechambers to the rooms flanking the apse. The antechambers must have been used liturgically like transept wings. In all Cyrenaica, only the East Church at Apollonia is a transept basilica; it is, as well, the best candidate for the earliest of the excavated Cyrenaican churches. Cf. n. 17. Another change at Latrun is the elimination of all adjoining units such as atria or contiguous chapels in favour of a rigorously closed plan and solid external elevation. In this regard, the usual side aisle and apse windows may be abandoned even though this would transform the character of the interior space. The churches take on a castrum-like appearance, if not the actual function, suitable to their surroundings and hence could provide the model for later Cyrenaican fortress churches of the countryside and undefended cities. Galleries seem to be a particular characteristic of these true castra. Cf. n. 47. Extensive vaulting along with a gallery elevation appears to be an introduction of the Justinianic period. Cf. Goodchild, R., ‘The Roman and Byzantine Limes in Cyrenaica’, JRS, xliii (1963)Google Scholar or Libyan Studies, 195 ff., and Krautheimer, op. cit., 186 ff.

⁶³ Cf. Harrison, op. cit.; more inclusive, Jones and Little, op. cit.

⁶⁴ Concerning the shift of capital city to Apollonia cf. Goodchild, R., Cyrene and Apollonia: An Historical Guide (London, 1959), 80Google Scholar and ‘The Roman and Byzantine Limes in Cyrenaica’, op. cit., 74, n. 36; also ‘A Byzantine Palace at Apollonia’, Antiquity, xxxiv (1960), 247Google Scholar. Kraeling, (Ptolemais, 27)Google Scholar also concerns himself with the same problem but defers to the opinion of Goodchild rather than accept a post Justinianic date offered by Romanelli, P. in La Cirenaica romana (96 a.C.–642 d. C.) (Verbania, 1943), 172–3Google Scholar.

⁶⁵ Since there were no Arab coins, not even clipped Byzantine coins, found during the excavation of either of the churches, it would seem that their destruction came at the time of the second Arab invasion of 644 A.D. Possibly the Arabs destroyed the churches, but it may have been the Monophysite Christians or Berbers. The columns of Basilica A were ‘chopped’ and then pulled out in a manner characteristic of several of the Cyrenaican city churches; those of Basilica B were pulled out on the ground-floor level without much mutilation. As noted, no sets of voussoirs were found among the ruins of either basilica, suggesting that these were retrieved and re-used in subsequent, but modest, building operations in the vicinity. Since there were no charred roof timbers, neither church suffered the torch, perhaps in order that the timbers, too, could be salvaged along with the voussoirs. The evidence does not indicate that Basilica A, the fortress, sustained a long siege before its destruction. It was hastily abandoned or else surrendered to the enemy without a fight, most likely the former. It could be argued that the haste of departure afforded insufficient opportunity to transfer even the reliquaries outside the town, but only time to hide them away superficially within the sunken baptismal font of Basilica A. The presence of almost all the church furniture of Basilica B indicates its immediate razing rather than a period of disuse and decay with only gradual appropriation of materials and eventual toppling of the superstructure.

⁶⁶ Procopius in his Buildings, VI cites the projects undertaken in North Africa after the reconquest. Erythrum is not mentioned for Cyrenaica.

⁶⁷ Pairs of churches are often found in rural Cyrenaica. Originally I believed that one was likely Orthodox while the other would be Monophysite. So strong were the arguments in favour of this hypothesis that Goodchild, R. adopted it in his important article ‘Byzantines, Berbers and Arabs in Seventh Century Libya’, Antiquity, xli (June 1967)Google Scholar, especially 120 or Libyan Studies, 255 ff. Further reflection reveals the falseness of such an assumption. Instead, what continued excavation seems to prove is that one church of a pair is somehow tied to burial or commemoration of the dead.

⁶⁸ Cf. n. 53. This kind of reasoning would require that Latrun's churches be early rather than late within their established time range; so, too, does acceptance of the suggestion that they are a model for the typical castrum-type Cyrenaican church.