I ought to say that my own knowledge of Pontus is drawn from journeys in 1891 and 1899 along the following routes—(a) Sivas, Zara, Enderes, Purk (Nicopolis), Lycus valley, Niksar, Tokat, Amasia, Samsun (cf. Supplementary Papers of the Royal Geographical Society, vol. iii); (b) Samsun, Herek, Niksar, Tokat, Turkhal, back to Niksar, Sonisa, Ladik, Khavsa, Vezir Keupru, Halys bridge, back to Khavsa, Marsovan, Chorum. The books which I have found most useful after Strabo are for geography Murray's excellent Guide to Asia Minor, and for history M. Théodore Reinach's admirable study Mithridate Eupator.

^{page 53 note 1} Samsun. One would naturally interpret the name as Ἄμισον (cf. Stambul, Isnik, Ismir, etc.), but certain of the coins seem to attest a genuine ancient form with initial sibilant.

^{page 53 note 2} Mr. J. G. C Anderson tells me that Amasia is 71½ miles by road from Samsun, and estimates the total to Zela at 99½. No Pontic road from north to south could be much more, whereas the road from Niksar to the Halys, only a fraction of the total length of the kingdom, is 105 miles.

^{page 53 note 3} Turkish Charshembey Ova, Tash Ova, and Kaz Ova.

^{page 53 note 4} Cf. Xenophon, Anab. v.

^{page 54 note 1} The valley of the upper Halys, originally Cappadocian, was subject to Mithridates Eupator. It was linked to Pontus through Zela at the one end and through Nicopolis at the other. The road Zela, Verisa, Sebasteia, Camisa, Zara, Nicopolis, marks the line of this province. Dr. Carrington of the American College at Marsovan tells me the curious fact that on this line and southwards from it all the Armenians speak Turkish, whereas north of it all speak Armenian.

^{page 55 note 1} This road was included in Pontus Galaticus (Ptolemy, , Geogr. v. 6Google Scholar, 3), probably in order that the Galatian province might have its own free access to the sea. By it Lucullus entered the Phanaroea. One purpose of the fortress of Eupatoria-Magnopolis was to guard it.

^{page 55 note 2} There are Roman milestones in the Phanaroea belonging to the Neo-Caesarea Amasia road at Niksar, Fidi, and Chalgara. Fidi=Ptolemy's Πίδα, which appears also on the Peutinger Table.

^{page 55 note 3} See Mr.Yorke's, V. W. discussion of this road in the Geographical Journal, vol. viii, pp, 467–8.Google Scholar

^{page 55 note 4} Strabo, 555.

^{page 56 note 1} Appian, , Mithr. 18–9Google Scholar, Memnon 31, Strabo, 562.

^{page 56 note 2} Appin, , Mithr. 64–6Google Scholar, Memnon 36.

^{page 56 note 3} Appian and Plutarch are vague. Memnon (37) says ὴπείγετο διὰ τῆς Τιμωνιτίδος Παφλα ὴπείγετο διὰ τῆς Τιμωνιτίδος Παφλα Βιθυνίαν ἀφικνεῖται But according to Strabo (562) Timonitis bordered immediately on Bithynia. Possibly, as Reinach suggests, Mithridates' army marched in two columns. One would go by Osmanjik and Tossia, the other by Chorum and Changra. On the date see Reinach, , Mithridate, p. 321Google Scholar, note.

^{page 57 note 1} The accounts transmitted to us of these operations (Appian, , Mithr. 78–9Google Scholar, Plutarch, , Luc. 14, Memnon43–5)Google Scholar are meagre and obscure. The route followed by Lucullus is defined only as having lain ‘through the mountains,’ and no relation is recognised between it and his attacks on Themiscyra and Eupatoria. We are left in the dark as to the fate of Themiscyra, which must have been taken before Lucullus could proceed The capture of Eupatoria is falsely involved with the siege of Amisus. If there really was a second Eupatoria, a suburb of Amisus (which I doubt), it might explain this misapprehension and the contradiction between Memnon, who tells how Eupatoria was carried by an unexpected assault, and Appian, who implies that it surrendered to the Romans, (Mithr. 115)Google Scholar. But I am inclined to believe that the root of the confusion may have been that the story of the siege of Amisus was reserved to the date of the capture of the city, and with it the attack on Eupatoria (of. Memnon's order), so that the latter was divorced from the march of Lucullus and falsely connected with Amisus, to the dislocation of the whole plan of campaign. Memnon's account of the storming of Eupatoria looks like a reduplication of the capture of Amisus, whereas M. Reinach's ingenious combination of the surrender of the fortress with the treason of Phoenix is extremely plausible (Reinach, , Mithridate, p. 337)Google Scholar.

^{page 57 note 2} It is true that Plutarch, (Luc. 15)Google Scholar speaks of Lucullus having got through a ‘pass’ and occupied a position ‘overhanging Cabira’ but these expressions seem to me to be not unnatural exaggerations. The passage between the river and the hills, or even the whole valley at this point, may reasonably be called a pass. Lucullus had got over the crest of the ridge and overlooked the plain of Niksar. It must be remembered that he cannot be thrust too far eastwards, for he has to communicate with Cappadocia, and the road from Cabira to Comana must have been in Mithridates' hands (Appian, , Mithr. 82.Google Scholar). The ridge was the nearest and most obvious point for Lucullus to seize, and in every way fits the rest of the narrative

^{page 58 note 1} The only detailed accounts are Appian, , Mithr. 79–82Google Scholar, and Plutarch, , Luc. 15–7Google Scholar. Both evidently draw on the same souree.

^{page 58 note 2} Strabo 547 clearly indicates the position of Gazioura. Ramsay, , Hist. Geogr. 326–8Google Scholar, proves from Basil that Ibora must also be placed about Turkhal. No doubt Ibora and Ioura are the same word, and Gaz is preserved in the Turkish Kaz Ova.

^{page 58 note 3} Can Dadasa be Dazya near Turkhal?

^{page 59 note 1} Strabo 560, cf. 557 and 559.

^{page 59 note 2} On the whole campaign see Dio Cassius, xxxvi. 11–17, Plutarch, , Luc. 35Google Scholar, Appian, , Mithr. 88–90Google Scholar. Appian sends Mithridates back from Zela into Armenia Minor, but that is improbable, for Lucullus would hardly have crossed the Halys before Sivas, and was not many days journey from Zela when the battle was fought.

^{page 59 note 3} Dio Cassius, xxxvi. 46–50, Appian, , Mithr. 97–101Google Scholar, Plutarch, , Pomp. 32Google Scholar, Strabo, 555.

^{page 59 note 4} Bell. Alex. 34–40, Dio Cassius, xlii, 45–6.

^{page 59 note 5} Bell. Alex. 66–77, Dio Cassius, xlii, 46–8, Appian, , Mithr. 120Google Scholar, Bell. Civ. ii. 91, Plutarch, , Caes. 50Google Scholar, Suetonius, , Jul. 35Google Scholar, 37.

^{page 60 note 1} The precise position of Καινὸν has not been determined in spite of Strabo's detailed description (556), but it probably stood near the Lycus.

^{page 60 note 2} Strabo 560 The text seems to be corrupt although the sense is plain. Besides the correction Νεαπολῖτιν for Μεγαλόπολιν I fancy we ought to write παρὰ for κατὰ (cf. Meineke, Vind. Strab.), put a comma at κώμην and read ἀποδείξας [αὐ]τὴν κατοικίαν Κατοικία is almost a technical term for a military colony.

^{page 60 note 3} Neo-Claudiopolis is known from coins. Ptolemy, V 4, 6, identifies it with Andrapa, ἀποδείξας [αὐ]τὴν κατοικίαν Κατοικία Andrapa was certainly the later name of the place, and the only name in Byzantine times. There is no positive proof that it was the earlier, but all probability favours the assumption. Our inscription has been published by Mr.Anderson, in the Journ. of Hell. Stud. xx., p. 152Google Scholar.

^{page 61 note 1} Strabo's words (561), ἐκεῖνος Πομπήιος μὲν οὖν οὔτω διέταξε τὴν Φαζημωνῖτιν οἰ δ ὔς τερον βασιλεῦσι καὶ ταύτην ἔνειμαν imply that Neapolis was included. His account of Pompey's distribution of Paphlagonia (541), some of which was given to the House of Pylaemenes and some retained, implies that Pompeiopolis was included. Both these cities were again included in the province of Pontns before Strabo wrote (544—καὶ μέχρι δεῦρο τοῖς Ρωμαίοις ἡ Ποντικὴ ἐπαρχία ἀφώρισται cf. 561–2, where ἠ ἑκτὸς Αλυος χώρα τῆς Ποντικῆς ἐπαρχίας includes Pompeiopolis) and their era, 6–5 B.C., indicates the date of their re-admission. Cf. J.H.S. xx. pp. 152–3, 155, 160–1. The four great coast-towns, Heraclea, Amastris, Sinope, and Amisus, were certainly included (M. Th. Reinach exaggerates when he says, p. 400 note 7, that Heraclea was not restored before the time of Caesar. See Memnon 60). The other five of Pompey's eleven πολιτεῖαι may be open to doubt. The simplest hypothesis is that they were Pompey's other five foundations, Magnopolis, Diospolis, Nicopolis, Megalopolis, and Zela. If Nicopolis must be surrendered to Deiotarus, Amasia might take its place, or, as a last resort if Amasia cannot be admitted, Tieum. Until the problem of the Pontic κοινὰ is cleared up, they had better be kept out of the question. Strabo could not speak as he does on p. 541, if Pompey's province had been no bigger than the later Pontus; and to whom were all the districts afterwards known as Pontus Galaticus and Pontus Polemoniacus assigned? It must be remembered that Pompey's arrangements were upset fifteen years later when Pharnaees overran the country, and that Caesar, Antony, and Augustus made their own distributions.

^{page 61 note 2} Sueton, . Vesp. 8Google Scholar; cf. Tac., Hist. ii. 81Google Scholar, Joseph, . B. J. vii, 1, 3Google Scholar.

^{page 61 note 3} Josephus (l.c.) says that Legio XII. Fulminata was stationed at Melitene at the end of the Jewish War. Commagene, annexed A.D. 72, not without fighting, cannot have been left without a garrison. The camp at Samosata must date from the occupation, and possibly Legio XVI. Flavia Firma was posted there from the first. But Commagene was reckoned to the Syrian command, whereas Suetonius and Tacitus speak of legions (plural) ‘added to Cappadocia.’ We should expect, therefore, to find that Vespasian also established the legionary camp at Satala. A milestone at Sherif, Melik (C.I.L. iii. 306Google Scholar, probably on the road from Satala to Melitene), and an inscription near Tiflis (published with C.I.L. iii. 6052), both dated A.D. 75, show that his attention was, in fact, directed to the defences and lines of communication in the north-east, where the Alani were a pressing menace (cf. Joseph, . B.J. vii. 7Google Scholar, 4). On the other hand, we have no evidence of any legion having ever been stationed at Satala except Legio XV. Apollinaris, which can not be proved to have been established in Cappadocia until the last years of Hadrian's reign (Arrian, , Acies c. Alanos, 5Google Scholar). Indeed, the evidence seems at first sight to tell strongly against an earlier arrival. Legio XV. Apollinaris was on the Save in southern Pannonia at the death of Augustus, and took part in the Parthian and Jewish wars (Tac., Ann. i. 23, xv. 25Google Scholar. Hist. v. 1. Josephus, , B.J. iii. 4Google Scholar, 2). After the latter it was sent back to Pannonia (Joseph, . B.J. vii. 5Google Scholar, 3). Now the camp at Carnuntum would seem to have been built by Vespasian in A.D. 73 (C.I.L. iii. 11194–6), and the monuments of Legio XV. Apollinaris are so numerous there that they postulate a long sojourn (cf. Mommsen's, note, C.I.L. iii. p. 550)Google Scholar. Hirschfeld, however, has ingeniously argued from the nomenclature of the inscriptions that the camp at Carnuntum must date from before the reign of Claudius, probably from that of Tiberius, and that Vespasian merely restored or rebuilt it (Arch.-Ep. Mitt. aus Oesterr. v. pp. 216–9). If Hirschfeld's argument may be accepted, there can be no difficulty in supposing that Legio XV. Apollinaris, having built the new camp at Carnuntum, was sent on to build new fortifications at Satala and Harmozica. The antecedent probability that Vespasian planted a legion at Satala strengthens Hirschfeld's argument. I am not aware that the problem has ever been approached from the Cappadocian side.

In any case it can be no accident that the three camps are all mentioned in connection with Trajan. He struck the Euphrates at Samosata, marched up to Satala, and thence into Armenia, and is said to have conferred political status on Melitene (Dio Cass. lxviii. 18–19, cf. lxxi. 2, Procop, . De Aed. iii. 4)Google Scholar. Nothing can be built on Dio's remark that Trajan occupied Samosata without fighting. He says the same of the whole march to Satala.

If Tacitus, ' words (Hist. ii. 6Google Scholar) Cappadocia Pontusque et quidquid custrorum Armeniis prætenditur may be pressed, there were camps along the frontier before there were legions. That is probable enough.

^{page 62 note 1} The best detailed account of the Upper Euphrates frontier is Mr.Yorke's, V. W. paper, ‘A journey in the valley of the Upper Euphrates,’ in the Geographical Journal, vol. viii. 1896Google Scholar. Cf. also J.H.S. xviii. 1898.

^{page 63 note 1} These first three stones may perhaps belong to other reads. I refer the stone at Sherif, Melik (C.I.L. iii. 306Google Scholar) to the road from Satala to Melitene, and that at Aksheher or Ashkhar near Nicopolis, (C.I.L. iii. 6057Google Scholar, R. G. S. Suppl. Pap. iii. p. 727) to the road from Nicopolis to Melitene. See Mr.Yorke's, argument in Geogr. Journ. viii. pp. 467–8Google Scholar. Mr. Yorke, by an oversight, ascribes this stone to Trajan instead of Hadrian.

^{page 64 note 1} Even if this stone were nearly in situ, it could not affect my argument, for it is of Constantinian date and the boundaries were then different. Mr. Anderson reports another late stone near Zela.

^{page 64 note 2} An exact parallel occurs on the road from NeoCæsarea to Comana. Milestenes do not occur until the valley of the Iris (Pontus Galaticus) is reached.

^{page 64 note 3} Can αἰ πηγαὶ τοῦ ποταμοῦ in Ptolemy refer to the source of the Sepetli Su? His Φαναγορία seems clearly equivalent to Strabo's Φανάροια

^{page 65 note 1} E.g. at Καινὴ πόλις (Artaxata), C.I.L. iii. 6052. Cf. 6741–2.

^{page 65 note 2} Procopins, , De Aed. iii. 4Google Scholar.