¹ The main channel of the Çark Deresi (Çark Deresi I on the road signs) flows about 200 m west of the bridge, in a converging side valley.

² Texier, C., Description de l'Asie Mineure i (Paris 1839) 55–6Google Scholar and pl. IV; there is an enthusiastic but imprecise description by de Hell, Xavier Hommaire,Voyage en Turquie et en Perse (Paris 1854–1860) ii 277–80Google Scholar, and a rough sketch of the east apse at iv pl. XVI.2; a good photograph of the bridge is published by Dalman, K. O., Der Valens-Aquädukt in Konstantinopel (Bamberg 1933) Taf. 3, Abb. 10Google Scholar.

³ See Fig. 1; most of the measurements in this article are derived from Texier's description, roughly checked where possible by my own observations. The silting up of the bridge piers means that many measurements cannot be precise. My visits to the bridge have been made possible by generous assistance from Merton College, Oxford, the British Academy and the European Science Foundation.

⁴ See Plates VIc, VIIa.

⁵ See fig. 2 and plate VII a, b. Texier referred to eight arches, a mistake which has been handed down in subsequent descriptions, whereas the number of main arches was five, as proved both by size and by the number of buttressed piers. Theophanes 234.17, ed. C. de Boor (Leipzig 1883) correctly referred to five tremendous arches, and in antiquity the bridge was probably known as Πεντεγέφυρα. (corrupted to Ποντογέφυρα, i.e. pontoon bridge, at Pachymeres ii 330.20, ed. E. Bekker [Bonn 1835]), a name preserved in the modern village name Beşköprü (Five bridges).

⁶ Moore, F. G., ‘Three canal projects, Roman and Byzantine’, AJA liv (1950) 108–10Google Scholar, an account which is mainly based on that by von Diest, in Petermann's Geog. Mitt. Erg. xxvii, Heft cxxv (1898) 70Google Scholar. Pliny, Epist. x 61Google Scholar.

⁷ A.P. ix 641; see further n. 9 below. Cf. Plb. xxi 37 (also Livy xxxviii 18.7) who records that even near its source the Sangarius was very difficult to cross.

⁸ The average gradient of the Adapazari plain is about four feet drop per airline mile, according to Russell, R. J., ‘Alluvial morphology of Anatolian rivers’, Ann. Assoc. Am. Geog. xliv (1954) 364Google Scholar. Moore's hypothesis requires that the Sangarius should flow more than five miles back from Adapazan to Lake Sapanca, which would entail a rise in level of about 20 ft. It cannot be proved that the physical geography of the Adapazan area was exactly the same in the sixth century, since earthquakes might have affected the terrain. However, during the first century AD (during Pliny's governorship of Bithynia), and probably still in the fourth century (Amm. Marc. xxii 8.14), the Sangarius flowed into the Black Sea, as it does today, and it is difficult to see how even the most serious of tremors could have forced the Sangarius to flow back from north to south in the intervening period.

⁹ Procop., de Aed. v 3.8–11Google Scholar, ed. J. Haury, re-ed. G. Wirth (Leipzig 1964).

¹⁰ A.P. ix 641, also quoted by Zonaras iii 159; Constantine Porphyrogenitus, de Thematibus, ed. Bekker, E. (Bonn 1840) i 27Google Scholar, quotes the epigram without attribution, but notes that it was inscribed on the bridge. There is no trace of the epigram on the bridge; it was perhaps inscribed on the triumphal arch, like the inscription on the Tagus bridge at Alcantara.

¹¹ Ekphrasis S. Sophiae, ed. Friedländer, P. (Leipzig 1912) 928–33Google Scholar.

¹² Theoph. 234.15–18.

¹³ Loc. cit. (n. 10).

¹⁴ See Gazzola, P., Ponti Romani (Florence 1963) e.g. nos 28, 29Google Scholar; this book is certainly not a comprehensive account of Roman bridges and it contains little information on bridges in Turkey.

¹⁵ Rounded, Gazzola nos 41 (ponte Cestio in Rome), 161 (Merida in Spain); combination, no. 157 (Chaves in Portugal).

¹⁶ Gazzola nos 29 (angular), 41 (rounded), 91 (square).

¹⁷ Gazzola nos 40 (ponte Fabricio at Rome), 196 (Medjerda in Tunisia), 234 (Moselle bridge at Trier), 259 (Sabun bridge in Syria). These are the only close parallels in Gazzola for the Sangarius design.

¹⁸ I am very grateful to R. E. Franklin of the University of Oxford Department of Engineering Science for his professional advice on the problems of water flow and the design of bridge piers, and for the diagrams which are reproduced in Fig. 4. F. Sear's recent discussion of Roman bridges, Roman Architecture (London 1982) 42Google Scholar, accurately analyses the problem of river turbulence, but is marred by a confusion between upstream and downstream.

¹⁹ Procop. de Aed. v records repairs to several bridges in Asia Minor; at Adana, for example, Justinian had to carry out major repairs to the bridge piers which had been weakened by the force of the river Sarus (de Aed. v 5.8–12). The Romans did recognise that scouring of the foundations was a serious problem; counter-measures which they adopted were to pave the river bed around the pier foundations (e.g. at Dara) or to reinforce the foundations (e.g. the angular cut-water of a bridge at Pergamum is founded on a solid rectangular base).

²⁰ One detail for which I have no explanation is the presence in the south face of the bridge pier of a small rectangular opening (approx. 80 cm high by 40 cm wide) that is partially blocked by the south-east pier of the inner vault. Behind the opening there appeared to be a narrow tunnel, but it was not possible to establish where it led.

²¹ Cf. the massive breakwaters stretching upstream from the Merida bridge, or the promachōn mentioned by Procopius at Justinian's bridge over the Siberis (de Aed. v 4.3); on the latter, see the brief account by Anderson, J. G. C., ‘Exploration in Galatia cis Halym II’, JHS xix (1899) 65–7Google Scholar.

²² Theoph. 234.16–17. There are two possible routes for this diversion, either about 200 m to the west so that it would have flowed down the present main valley of the Çark Deresi, or else about two km to the east so that it could have flowed around the hill at the east end of the bridge, and then back across the flat site of Adapazan into the Çark valley. The former seems more probable.

²³ Paul Sil. Ekphrasis 928–33. For other allusions in Paul's poem to very recent events, see Whitby, Mary, ‘The occasion of Paul the Silentiary's Ekphrasis of S. Sophia’, CQ xxxv (1985)Google Scholar.

²⁴ For similar hyperbolic references to the peace, cf. Paul Sil. Ekphrasis 13–15, 138, 936, and Cameron, Averil, Corippus: In laudem Iustini Augusti Minoris (London 1976)Google Scholar comm. on preface 30 f. (p. 122).

²⁵ Agathias' epigram was probably composed before Paul's Ekphrasis, since the linguistic affinities between the epigram and ekphrasis are best interpreted as a deliberate compliment by Paul to his friend and fellow-poet.

²⁶ The starting point of Theophanes' annus mundi years is debated, see Ostrogorsky, G., ‘Die Chronologie des Theophanes im 7. und 8. Jahrhundert’, BNJ vii (1928/1929) 1–56Google Scholar and Grumel, V. ‘L'année du monde dans la Chronographie de Théophane’, EO xxxvii (1934) 396–408Google Scholar. It is probable that Theophanes was not consistent throughout his Chronographia and that he was influenced in part by the practice of the particular sources which he was using; for his account of the latter part of Justinian's reign, Theophanes seems to have begun his year in September.

²⁷ Procop., de Aed. v 3.10Google Scholar; Stein, E., Histoire du Bas-Empire ii (Paris 1949) 837Google Scholar.

²⁸ On the inceptive force of ἄρχομαι, see the Appendix below.

²⁹ Cf. Theoph. 232.27–233.3, where the completion of Justinian's restoration of S. Sophia is described in the notice which records the start of repairs.

³⁰ On this, see Whitby, Michael, ‘Theophanes' chronicle source for the reigns of Justin II, Tiberius and Maurice (565–602 A.D.)’, Byzantion liii (1983) 312–45Google Scholar.

³¹ This point was clearly stated by Mango, C., ‘The date of the Studius Basilica at Istanbul’, BMGS iv (1978) 119Google Scholar: ‘Since the dates given by Theophanes are sometimes right and sometimes wrong, it may be advisable to examine from this point of view a sample section of the Chronicle in the area that concerns us.’ Mango's examination showed that Theophanes' chronology for the 460s was not reliable, but because of the great variability in the quality of Theophanes' narrative one cannot conclude that his account of the 550s must be inaccurate simply because it is inaccurate for the 460s, 570s or 580s.

³² Ed. L. Dindorf (Bonn 1831).

³³ Krumbacher, K., Geschichte der byzantinischen Literatur (Munich 1897) 331Google Scholar; Moravcsik, G., Byzantitioturcica (Berlin 1958) i 329Google Scholar. The exact date of the termination of Malalas continuatus is uncertain, because the last folios of our sole MS of Malalas are lost. The Laterculus Imperatorum Romanorum Malalianus, a Latin chronicle which shows some affinities with Malalas, contains a list of emperors that terminates in the ninth year of justin II (AD 573/4), but this need only designate the date when the Latin translator or excerptor was reworking Malalas (Laterculus, ed. T. Mommsen, MGH Auct. Ant. xiii, Chron. Min. 3 [Berlin 1894] 426–37). In Theophanes, there is a marked difference between the relatively full account of the latter years of Justinian's reign, which was mostly derived from Malalas, and the less detailed and more imprecise account of Justin II's reign; hence the text of Malalas used by Theophanes most probably ended in AD 565.

³⁴ Moravcsik (n. 33) placed the break between the original Malalas and Malalas continuatus immediately before the account of the Nika riot (Mal. 473.4), but there is a more obvious break after Malalas' very long account of the riot (Mal. 477.3), where the narrative becomes much less comprehensive.

³⁵ Theoph. 174.1–186.5, 12 pages of Teubner text for the five years.

³⁶ Cf. the discussion by Bury, J. B., ‘The Nika Riot’, JHS xvii (1897) 101–4Google Scholar; also Stein (n. 27) ii 830, who noted that the Syriac chronicle of Ps.-Dionysius of Tel Mahre confirmed that Theoph. 229.10–14 came from a part of Malalas excised by the epitomator.

³⁷ Theoph. 177.33–178.5, an account of a harsh winter and of prayers for the termination of a series of earthquakes at Antioch, which directly continues Theophanes' report of the first earth tremor; this tremor was recorded by Mal. 442.18–443.7. Theoph. 181.10–11, a stage of Roman-Persian negotiations that would probably have been recorded between Malalas' other information on this diplomacy (453.12–14 and 454.11–15).

³⁸ Theoph. 181.24–31; cf. Artec. Cramer 112.19–27 (Cramer, J. A., Anecdota graeca e codd. manuscriptis bibliothecae regiae parisiensis ii [Oxford 1839]Google Scholar).

³⁹ On the Akta, see Cameron, Alan, Circus Factions (Oxford 1976) 318–33Google Scholar, and Whitby, Michael, ‘The Great Chronographer and Theophanes’, BMGS viii (1982/1983) 9–10Google Scholar.

⁴⁰ Theoph. 174.27–176.17 groups three separate reports of diplomacy in Malalas; Theophanes 178.27–179.14 amalgamates two different stages of a complex scries of Roman-Persian negotiations (Mal. 447.22–448.2 and 455.10–456.18), and as a result the second stage is antedated; Theoph. 179.15–27, an account of Roman-Arab warfare based on Mal. 434.19–435.17, is inserted one year late.

⁴¹ Theoph. 186.26–216.4 from Procop. Bell. iii–iv (Vandal expedition) and 219.19–222.8 from Bell. ii 20–1 (Persian campaign of AD 542).

⁴² Malalas covered these years in only five pages (477.13–482.18), whereas the first five years of Justinian's reign occupy 52 pages of text.

⁴³ Notices for which source passages are no longer extant can probably be assigned as follows: Theoph. 216.7–14 (Iberian embassy), 219.14–16 (Bulgarian captives) and 224.29–33 (earthquake at Constantinople) are likely to have originated in Malalas; Theoph. 216.23–24; 217.12; 222.9–19 (ecclesiastical information) in Anecdota Cramer. The summary of Vandal history at Theophanes 186.18–26 is paraphrased from Procop., Bell. iii 3Google Scholar.

⁴⁴ Thus, as the result of an incorrect indiction date in the Anecdota Cramer, Theophanes reported the same earthquake under two different years, at 222.25–30 based on Anec. Cramer 113.24–30 (but with the regnal year adjusted to fit the incorrect indiction), and at 229.5–10 based on Mal. 486.23–487.9 and correctly dated.

⁴⁵ Thus Theoph. 216.24–25; 217.4–12; 218.18–20 and 224.11–15 were derived respectively from Mal. 479.15–20; 479.7–12; 479.23–480.7 and 482.12–13.

⁴⁶ Theoph. 186.8–13 (AM 6025); 186.15–17 (AM 6026); 216.17–22 (AM 6028); 217.26–218.17 (AM 6031); and 218.31–219.14 (AM 6032) were derived from Mal. 441.8–12; 449.12–14; 436.17–437.2; 437.19–438.20 and 450.19–451.15, information which all belongs to the same indiction year.

⁴⁷ Theoph. 222.33–223.27 (Axumite–Homerite conflict) and 224.15–27 (Andreas the Italian entertainer), derived from Mal. 433.3 434.18 and 453.15–454.4.

⁴⁸ These twenty years occupy 16½ pages of text, 225.1–241.15.

⁴⁹ Malalas' own account of this period was fuller than for the previous section: his surviving account occupies 19½ pages, and some allowance has to be made for the defective text (on which, see Neumann, K. J., ‘Der Umfang der Chronik des Malalas in der Oxforder Handschrift’, Hermes xv [1880] 356–60Google Scholar). The extent and general accuracy of Theophanes' use of Malalas in this section can be seen in the lists of parallel passages in Rochow, I., ‘Malalas bei Theophanes’, Klio lxv (1983) 469–471Google Scholar. The detail of Malalas' account of these years, before it was epitomised, is revealed by a few palimpsest fragments published by Mai, A. (Spicilegium Romanum ii [Rome 1839] de fragmentis historicis tusculanis, fr. IV, pp. 22–8)Google Scholar and by Mal. frr. 48–51, in Const. Porph. Excerpta de Insidiis, ed. de Boor, C. (Berlin 1905) pp. 173–6Google Scholar.

⁵⁰ Anec. Cramer 113.8–9 provided Theophanes' incorrect date for Theodora's death (226.8–9; contrast Malalas 484.4–5 and Anec. Cramer 111.10–11), and his account of Justinian's heresy (Anec. Cramer 111.15–19; Theoph. 240.31–241.5); probably also Theophanes' summary of Pope Vigilius' actions in Constantinople (225.13–28; contrast Mal. 483.3–4; 484.11 –13; 485.4–7; 483.14–16) and the ecclesiastical information at 228.28–229.3; 230.30–33; 240.26–30 and 241.6–15. The account of an earthquake and storm (229.29–230.3) could have come from either Malalas or Anecdota Cramer.

⁵¹ Some regnal year dates in the Anecdota Cramer are one year late (113.8–9, 24–30).

⁵² E.g. Theoph. 232.27–233.3, the collapse of S. Sophia, which should have been placed in AM 6050, not 6051.

⁵³ Stein (n. 27) ii 837.

⁵⁴ Theoph. 234.6; this statement that Justinian remained at Selymbria until August accords with the account of his adventus at Constantinople on 11 August (Const. Porph. de Caerimoniis, ed.J. Reiske [Bonn 1829] 497.13–498.13). This provides useful confirmation for Theophanes' accuracy in this section.

⁵⁵ Theoph. 234.20.

⁵⁶ Theoph. 234.20–22; Justinian was believed to be in Thrace, probably supervising defensive constructions, but it would have been a simple matter for him to have sailed from Selymbria to Nicomedia to view the progress of the bridge.

⁵⁷ Theoph. 240.11–13; the remains of this substantial church, which Justinian had repaired, can be seen at the village of Yürme in the vicinity of Sivrihisar (for a brief description of the church, see Crowfoot, J. W., ‘Notes upon late Anatolian art’, ABSA iv [1897/1898] 86–92Google Scholar).

⁵⁸ Caesar, Gallic Wars iv 17–18Google Scholar records that the construction of a wooden trestle bridge across the Rhine in 55 BC took 10 days. Caesar's speed on this occasion was exceptional, and his wooden bridge would have been much simpler to construct than the masonry bridge over the Sangarius, but it illustrates what could be achieved with energy and determination.

⁵⁹ Ps.-Zachariah of Mytilene, Ecclesiastical History, trans Hamilton, F. J. and Brooks, E. W. (London 1899) vii 6Google Scholar.

⁶⁰ Dahn, F., Prokopius von Cäsarea (Berlin 1865) 38Google Scholar; Haury, J., Procopiana (Augsburg 1891) 27–8Google Scholar; Bury, J. B., A history of the later Roman Empire from the death of Theodosius I to the death of Justinian (London 1923) 428Google Scholar; Downey, G., ‘The composition of Procopius’ de Aedificiis', TAPA lxxviii (1947) 182–3Google Scholar.

⁶¹ Agathias, , History, ed. Keydell, R. (Berlin 1967) v 1–2Google Scholar.

⁶² Cod. Just. iii 129, ed. Schoell, R. and Kroll, W.⁶ (Berlin 1954) 647Google Scholar; this was issued in June 551 at the suggestion of Scrgius, bishop of Caesarea, who believed that it would encourage the Samaritans to behave better and would ensure their future tranquillity. This Novel merely lifted some of the restrictions that had been imposed on the Samaritans in the early years of Justinian's reign.

⁶³ The analogy with a faction riot is brought out in our main source for the incident, Mal. fr. 48 (Exc. de Insid. 173), which compares the alliance of Jews and Samaritans with the famous prasinobenetoi of the Nika riot. This incident was not sufficiently important to be recorded by Agathias in his History of the period, and it is only briefly noted in the Chronicle of Michael the Syrian (trans. J. B. Chabot [Paris 1899–1910] ii 262).

⁶⁴ Anec. 11.24–30.

⁶⁵ Haury loc. cit. (n. 60).

⁶⁶ Downey (n. 60) 176–81 claimed that differences between the two main manuscripts of de Aedificiis proved that one tradition (A) represented an early version of the work, which was subsequently revised, particularly in Bk i, to produce the other tradition (V). This argument is totally unfounded, since many of the minor differences between the manuscripts reflect the carelessness of A's scribe, and the few major differences were caused by A's abbreviation of the tradition represented by V. A's revision of i 1.22–6, the passage regarded by Downey as most significant, is ungrammatical and very clumsy, while the effects of abbreviation are evident at iv 3.15, where A carelessly preserves a reference forward to a list of Balkan forts (iv 4) which is actually omitted from A.

⁶⁷ The restoration of the walls of Chalcis in Syria is reported twice (de Aed. ii 11.1 and 8), as is the restoration of the Crimean city of Bosporus (iii 7.10, 12); there may be some repetitions in the lists of Balkan fortifications in Bk iv, although this is harder to prove since some of the forts may have had similar names. Procopius' arrangement of his material is occasionally awkward: the baths near Anchialus in Thrace belong in the second half of Bk iv, but are described at the end of the periplus of the Black Sea (iii 7.18–23); some of the monastic constructions included in the list at the end of Bk v (e.g. v 9.29, Antioch, and 9.31, Mesopotamia) should have been recorded in ii.

⁶⁸ De Aed. i 1.45–6; furthermore, the first dome had been more impressive than its replacement (Agathias, Hist. v 9.5Google Scholar).

⁶⁹ For an evocative description of the dome and its decoration, sec Paul Sil. Ekphrasis 489–511, accepting the readings of Ludwich, A., Textkritische Noten zu Paulus Silentiarius (Königsberg 1913) 7–9Google Scholar.

⁷⁰ De Aed. i 1.64–5; Theoph. 232.30–1 records the destruction of the ambo etc., whose original appearance is described (in exaggerated terms) in the ninth-century Narratio de Sancta Sophia, ed. Preger, T., Scriptores Originum Constantinopolitanarum (Leipzig 1901) 16, 17, 21Google Scholar. The reconstructed ambo was described by Sil., Paul, Ekphrasis Ambonis, ed. Friedländer, P. (Leipzig 1912)Google Scholar.

⁷¹ Narratio 28 records that the timber scaffolding was left in position for a year after the physical reconstruction of the dome had been completed, to ensure that the dome's masonry had set properly before the mosaic decorations were repaired by the decorators, who would have made use of the builders' scaffolding (cf. Harrison, R. M., ‘Anicia Juliana's church of St. Polyeuktos’, XVI Internationaler Byzantinisten-kongress Akten II.4, JOB xxxii (v) [1982] 436Google Scholar).

⁷² Paul Sil. Ekphrasis 177–85, 204–13, apologised profusely for rousing Justinian's grief by references to the disaster, references which could not be avoided by Paul since the collapse was the root cause for the rededication ceremony to which Paul's poem contributed; Paul excused himself from describing the process of reconstruction on the grounds that it was too technical (Ekphrasis 279–83).

⁷³ De Aed. i 1.67–8.

⁷⁴ See Crow, J., ‘Dara, a late Roman fortress in Mesopotamia’, Yayla iv (1981) 11–20Google Scholar, and Croke, B. and Crow, J., ‘Procopius and Dara’, JRS lxxiii (1983) 143–59Google Scholar, although the criticisms levelled at Procopius in these two articles are seriously exaggerated, as I hope to demonstrate in a forthcoming article.

⁷⁵ Rubin, B., ‘Prokopios von Kaisareia’, RE xxiii. 1 (1953) 573Google Scholar; Veh, O., Zur Geschichtsschreibung und Weltauffassung des Prokop von Caesarea iii (Bayreuth 1953) 15Google Scholar.

⁷⁶ Anec. 18.38, ὥς μοι ἐν τοι̑ς ἔμπροσθεν λόγοις γεγράψεται.

⁷⁷ Emending γεγράψεται to γέγραπται, as Dindorf in the Bonn Corpus text.

⁷⁸ Emending ἔμπροσθεν to ὄπισθεν, as Haury in the Teubner text.

⁷⁹ E.g. Anec. 1.3.

⁸⁰ This indeed was Haury's first suggestion (n. 60) 18, but he changed his opinion before he edited the Anecdota.

⁸¹ It is possible that Procopius continued to collect information about these events, information of the type that subsequently enabled Agathias, who was not a diligent historical researcher, to carry on Procopius' narrative.

⁸² Downey (n. 60) 176; CIL xi 1.288 and 294 record that S. Vitale and S. Apollinare in Classe were both built, decorated, and dedicated (in AD 547 and 549 respectively) by Julius Argentarius. See the discussion by Deichmann, F. W., Ravenna, Kommentar ii (Wiesbaden 1976) 3–33Google Scholar, particularly 15–21.

⁸³ Procopius does not mention the buildings recorded by Mal. 430.18–19, 435.18–20, and 445.8–9 (the suburb of Sykae and the baths of Dagistheus at Constantinople, and an aqueduct at Alexandria), or the church at Germia (cf. n. 57).

⁸⁴ Theoph. 233.8–11, 234.3–6; this date was accepted by Croke, B., ‘The date of the “Anastasian Long Wall” in Thrace’, GRBS xxiii (1982) 69 n. 39Google Scholar. On his return to Constantinople in AD 559, Justinian probably travelled from Selymbria by the old inland Via Egnatia, which passed round the heads of the two large inlets of Büyük and Küçük Çekmece, since he entered the capital by the Edirne Gate (Porph., Const.de Caer. 497.13–16Google Scholar); this choice of route would be explained if the coastal Via Egnatia was being reconstructed at the time.

⁸⁵ This identification cannot be proved, but was accepted by Cameron, Averil, ‘The Sceptic and the Shroud’, Continuity and Change in Sixth Century Byzantium v [Variorum 1982] 23–4 n. 46Google Scholar. Michael the Syrian ix 29 records Amazonius' construction of the church. The Syriac hymn is translated by Mango, C. in The Art of the Byzantine Empire 312–14.53 (Englewood Cliffs, N.J. 1972) 57–60Google Scholar.

⁸⁶ Verse 2 of the Syriac hymn reads ‘And now Amidonius, Asaph and Addai have built for Thee at Edessa this glorious Temple’; Mango suggested, in his n. ad loc., that Asaph and Addai were presumably the architects of the cathedral, but this Addai could equally well be Amazonius' predecessor as bishop.

⁸⁷ His presence is recorded in the official acts, Acta Conciliorum Oecumenicorum iv 1, ed Schwartz, E. and Straub, J. (Berlin 1971)Google Scholar. The Council's preliminary proceedings probably began in March, and the official proceedings concluded in June. Making allowance for travelling, which is unlikely to have been hurried, Amazonius would have been absent from Edessa for at least six, more probably eight, months in 553.

⁸⁸ Apart from the Sangarius bridge, the only works in de Aedificiis that can be securely dated after 545 are the reconstruction of Topirus in Thrace following its capture by the Slavs in 550 (de Aed. iv 11.14–17; Bell. vii 38.9–23), and of the walls of Chalcis in Syria (de Aed. ii 11.1, 8; W. K. Prentice, , Greek and Latin inscriptions, Publications of an American Archaeological Expedition to Syria iii [New York 1908] nos 305–6Google Scholar).

⁸⁹ E.g. Corippus, , In laudem Iustini Augusti minoris ii 331 ff.Google Scholar, with Averil Cameron's nn. ad loc. (op. cit. n. 24); see also Cameron, Introd. 6–7. The financial problems of Justinian's reign, to which Corippus alludes, were caused not by neglect but by Justinian's activities (the major construction projects around Constantinople in AD 558–562, and the substantial peace payment to Persia in 562).

⁹⁰ I am most grateful to Ewen Bowie and Mary Whitby for advice, criticism, and encouragement, to Prof. C. Mango for comments on part I of this article, and to Prof. Averil Cameron for comments on an earlier draft of parts II and III; Prof. Cameron does not agree with my arguments about the date of de Aedificiis.

⁹¹ Although there is a strong tendency for the obliteration of different verbal systems, and for a parallel development of a periphrastic perfect to replace the loss of the distinctive perfect tense, this periphrastic perfect was formed by εἶναι or ἔχειν plus participle (see Mitsakis, K., The language of Romanos the Melodist, Byzantinisches Archiv xi [Munich 1967] para. 272, and cf. 96, 98Google Scholar; cf. also Jannaris, A. N., Historical Greek grammar [London 1897] paras 1864–5Google Scholar, and Psaltes, S. B., Grammatik der byzantinischen Chroniken [Göttingen 1913] para. 345)Google Scholar.