^{page 23 note 1} See Ramsay, , Cities and Bish. of Phrygia, vol. i. p. xivGoogle Scholar. n., and the Excursus at the end of this paper.

^{page 23 note 2} Larissa and Herpa must have been near each other. Herpa (Strabo, pp. 537, 539) or Herpha (p. 663) was on the road from Caesareia to Melitene at the point where it crossed the Karmalas, (see Hist. Geog. pp. 289, 272–3)Google Scholar. Larissa cannot be located with certainty, but lay on the direct road to Melitene, not far east of Arasaxa, (H. G. pp. 272Google Scholar–3, and campaign of 1069 infra). It was given, along with Komana, Tzamandos, etc., to the Armenian prince Gagik in 1064.

^{page 23 note 3} By Const. Porph., De Them. p. 32Google Scholar and De Adm. Imp. p. 228; in 976 (Kedr. ii. p. 423) and 1068 (Mich. Att. pp. 121–2, Skylit. 678), see infra. The Armenian name is Dzamentav or Dzamentou, Arab. Samandou (St. Martin, , Mém. sur ľArménie, I. p. 191Google Scholar).

^{page 23 note 4} The quotation is from MS. notes of Prof. Ramsay's, to whose unfailing kindness I owe far more than can be actually specified in the preparation of this paper.

^{page 23 note 5} For similar cases see pp. 279 n., 280 n.

^{page 23 note 6} Τζαμαντὸς in Mich. Att. 121.

^{page 23 note 7} It is quite likely that Herpa is the older name of the town, which was renamed Ariarathia after one of the Cappadocian kings, Ariarathes (ca. 350–36 B.C.): Herpa is not mentioned after beginning of first century B.C.

^{page 24 note 1} See Class. Review, April 1896, p. 140 and 137, and Map accompanying this paper.

^{page 24 note 2} Cl. Rev. l.c. p. 137. The comparison of the Arabic Taranda (Bilãdhuri, p. 186, ed. M. de Goeje), Armenian Daranda, Syriac Turanda (St. Martin l.c. p. 190), makes assurance doubly sure. Taranta was a place of importance, especially in the first half of the eighth century, when the district of Melitene was in Saracen hands. It is mentioned by Theoph. p. 312 De Boor (see infra); in 701 A.D. it was beseiged in vain by Abd Allah (Ἀβδελᾶς), p. 372. Some further information about the fortress is given by Bilādhuri, for all references to whom I am indebted to the kindness of Mr. Le Strange who has taken the trouble to translate for me his chap. on Mesopotamian Fortresses in the ‘Book of the Conquests’ (ca. 869 A.D.). When Malatia was taken by the Greeks ca. 693, the Moslems settled at Taranda after it had been captured in 702. It lies three marches distant from Malatiyya, deep in the Greek country; and was held by a garrison of about 2,000 troops from the Mesopotamian army, during the summer only (Bilādh. p. 185). When Omar II. became Caliph (end of 717 A.D.), ‘he removed the Moslem population from Taranda, for he feared for their safety, the enemy being so near; the people however objected, and when they were perforce removed, they would leave nothing for the enemy, breaking even their oil and vinegar jars. The Caliph transported them to Malatiyya, leaving Taranda to ruin’ (p. 186). In the following century it was a Paulician stronghold and surrendered to Basil I. in 872 A.D. (Theoph. Cont. 267).

^{page 25 note 1} Only the most important evidence can be given here. Ptolemy (v. 13, 19) places including amongst its towns and Arsamosata (below). In Byzantine and Arabic times it clearly denotes the district indicated above. It is always connected with the κλεισοῦρα Romanopolis-Palu: before Romanus I. it was attached to Melitene (, Const. De Adm. Imp. p. 226), and was assigned by him along with Kamacha, Keltzine (Acilisene), etc. to the newly-eonstituted Theme of Mesopotamia (p. 227). Nicephorus, , De Velit. Bell. p. 250Google Scholar (quoted at end of Pt. I.), is quite explicit: the trans-Euphrates passes into Saracen territory are those crossing ‘the (Tauros) mountains which separate Chanzit from the enemy's country as far as Romanopolis.’ All the passages fsom Syriac and Arabic writers collected by Gelzer, , Geo. Cyp. p. 178Google Scholar f., confirm this localization: e.g. Faustus Byzant. v. 16 gives as conquered in succession Arzanene, Sophanene, Ingilene (about Egil), and Anzitene, preserving the geographical order from southeast to north-west. The Arabic writers all agree in placing the fortress Hanzit close to the Euphrates between Malatya and Sumaysāt (Samosata), on a tributary of the Euphrates, says Ibn Serapion, which ‘passes the city of Hanzit and the province thereof’ and then falls into the Euphrates (ed. Le Strange, p. 54, cf. n. on p. 49). The Euphrates dépasse la ville de Hanzyt, puis tourne vers ľouest, arrive à Sumaysat… (Khordadbeh, Ibn, Trans. p. 177)Google Scholar. Space forbids further quotations. The fort then, should apparently be looked for near the Euphrates, west of Kizin: the position of the ‘province’ Anzitene-Hanzit seems clear.

^{page 25 note 2} Ibn Serapion's description is confirmed by Ibn Khordadbeh and Yakut (cf. Le Strange, l.c. p. 57).

^{page 25 note 3} This is a reproduction of the Armenian form Ašmušat, as ΧανζΙτ is of Handzith (Prof.Gelzer, on Geo. Cyp. p. 172Google Scholar). The Theme of which Arsamosata was the central fortress apparently extended north of Murad Tchai (Arsanias).

^{page 26 note 1} See the description in Theoph. p. 313.

^{page 27 note 1} Measured apparently from about the Euphrates, though this is not precisely stated.

^{page 27 note 2} Tzamandos was situated on the hill above the modern Azizie, which occupies the lower slopes.

^{page 27 note 3} Cf. H. G. p. 271.

^{page 27 note 4} ‘One too many,’ Hogarth l.c. p. 51.

^{page 27 note 5} With Arga it would be possible to identify the Paulician fortress Argaous, which occurs in the marches of Basil I. in 872 (Ἀργαούθ, Theoph. Cont. p. 270) and of Romanus IV. in 1068 (Ἀργαοῦ, Ioan. Skylit. 670). Kedr. II. p. 154 tells us that Argaous and Amara were the first cities founded by the Paulicians with the aid of the Emir of Melitene; and that when their numbers increased Tephrike was added. The first city would be in, or close to, Saracen territory: but as the sect grew in strength and became to a certain extent an independent state, they would have to find sites for their new cities outside Saracen territory, i.e. further north. Now Amara (see infra iv. (2) b) is north of Argaous and Tephrike is north of Amara. Argaous therefore might be Arga. But it is far more probable that it should be identified with modern Argovan, about twenty-five miles almost due north of Malatia. This suits the line of both marches (infra) far better and is in itself a more natural position for the first Paulician city.

^{page 28 note 1} For the importance of Arabissos, see H. G. pp. 277, 311, 280, etc. From Tanadaris—Tanir there is also a direct road to Sebasteia (iv, (2) b infra).

^{page 29 note 1} Cl. Review, l.c. pp. 138 f.

^{page 29 note 2} Cf. pp. 271, 281, 291.

^{page 29 note 3} See H. G. pp. 386, 451.

^{page 31 note 1} See Cl. Rev. l.c. p. 138 f.

^{page 31 note 2} ‘Palas = Αἰπολίους, ί being pronounced as y, as in modern Greek.’—Prof. Ramsay in MS. notes.

^{page 32 note 1} This might possibly be the ‘Aragines in Pharakn’ of the Armen. Notitia. Is it possible that Pharakn (=‘the sheepfold’) is the pass βούκου λίθος near the Euphrates (Kedr. ii. 421, supra, p. 24)? Prof. Ramsay, however, think that Pharakn=Everek at base of Mt. Argaios.

^{page 33 note 1} Perhaps taken as a fem. sing., but ordinarily in Bonn ed. is clearly wrong.

^{page 33 note 2} From this passage alone it would be natural to connect Kaësoun with modern Khesun in Kommagene, south of Besne: but see above II. (2) and campaign of 877 infra. The Sis passes should strictly be included under but Niceph. is evidently thinking of the passes leading from the Anti-Tauros region generally across Tauros to Germaniceia and Adata. The Sis passes ought to come under those leading from the Theme Lykandos into Cilicia.

^{page 34 note 1} See Class. Review, l.c. p. 140 note, and H. G. p. 311. If the text of Theoph. is right, the description is obviously very bad. It may be answered that he meant to say ‘on his way over Tauros (Amanos, cf. Mich. Att. 120, Skyl. 677) he reached Germaniceia, and passing Adana came to the Saros.’ If so, the text requires much alteration, for the proper order is Germaniceia—Amanos—Saros—Adana; and it is most improbable that he would cross (1) Amanos, (2) Tauros by Cilician Gates, and thence by a most difficult route come round to Sivas.

^{page 34 note 2} The date is 877, not 880, for Sima, who submitted to Basil, was killed by Tulun of Egypt in 878 (Weil, , Gesch. der Khal. ii 473Google Scholar n.).

^{page 34 note 3} Kedr. calls it Kameia.

^{page 35 note 1} Phalakron may be Frakhtin (Ferakhtin) on the western pass to Sis; the -tin is the Arabic word Din widely adopted in Turkish [W. M. R.]. Psilo-, Xylo-, etc. are all Graecized forms.

^{page 35 note 2} Cont. and Kedr. do not precisely say so; after enumerating all the forts they say vaguely knowing only that the forts were in this district somewhere. The Saracen army would not take to flight, nor would Simas, ‘who held the Tauros passes’ (infra), submit before Basil had reached the Karmalas, as their language would imply.

It might be suggested that the curious name Onopniktes is a popular word expressing the difficulty of fording the river

^{page 35 note 3} Cont. says probably to be changed to

^{page 35 note 4} Simas was not ‘son of Tael’ but his surname was Tawil, i.e. ‘the tall’ (Weil, l.c. ii. p. 473 n.).

^{page 35 note 5} Endelechone — Andala may perhaps be Enderessi on the western pass.

^{page 35 note 6} Which shows that this was not the ordinary route to Germaniceia.

^{page 36 note 1} This means that the war begun the year before in the south-west and north was being carried on at the same time as Basil's expedition: next year (878) Abdallah, Emir of Tarsus, was decisively defeated at Podandos.

^{page 36 note 2} (Skyl. 670): the order of the words does not prove that Argaous is north of Tephrike: he has just said . The site assigned to Argaous (supra) at Argovan suits this passage well.

^{page 37 note 1} Δολιχὴ becomes Dolouk and then again in Greek Τελούχ ! It is mentioned both as πόλις and as θέμα in Kedr. ii. 494.

^{page 37 note 2} ‘Azaz is about twenty miles north by west of Aleppo,’ Finlay, i. 472 quoting from Col. Chesney. It is called “Αζάζιον, two days' march from Berroia (Aleppo),” in Kedr. ii. 492.

^{page 37 note 3} Modern Kutma, nearer Antioch. The description of the march is very accurate.

^{page 38 note 1} The crossing of Murad Tchai is not mentioned, but must be assumed.

^{page 39 note 1} Keramon can hardly he connected with τὸ Κεραμίσιον on the Zarnuk, the most easterly tributary of Tokhma Su (Melas) flowing past Melitene (Theoph. Cont. 268).

^{page 39 note 2} The Turks therefore took their favourite route by Caesareia [I. (1)] and thence to Iconium [II. (3) b].

^{page 39 note 3} I have followed the accounts of the Latin writers in Migne's Patrol. Lat. vols. clvi. (Guibert) and clv. (Rob. Mon., Tudebodius, etc.).

^{page 40 note 1} The name is given by Baldric.

^{page 40 note 2} ‘In qua erat maxima ubertas atque stipata omnibus bonis quae nobis erant necessaria,’ Tudeb. l.c.: so Guibert, etc.

^{page 40 note 3} Von Moltke, quoted by Hogarth, in Mod. and Anc. Roads, etc. p. 20Google Scholar, describes the road from Marash to Geuksun as difficult. This may have been the route he took.

^{page 41 note 1} Discussed in Class. Rev., l.c. pp. 136 ff., and only summarised here in tbe briefest possible manner.

^{page 42 note 1} i.e. if his authority mentioned Kilikia.

^{page 43 note 1} If Hdt.'s authority was an official document, should we not expect it to be based upon this division?

^{page 44 note 1} It ought to be etc.

^{page 44 note 2} Cf. the importance attached to the fortification of the west bank of the Euphrates in Amm. Marcell. xviii. 7 (supra iii.).