^* I express my gratitude to the Managing Committee of the British School of Archaeology at Athens for appointing me Visiting Fellow and to the Managers of the Leverhulme Research Fellowships for a grant which enabled me to work in Greece during the summer of 1953; to Professor Homer Thompson and Mr. C. W. J. Eliot, who accompanied me on a visit to Salamis and discussed the problems; to Captain Pringle, R.N., and to Mr. Vasilis Deleyannis, seaman at the Greek Yacht Club, who informed me of local conditions; and to Professor Sir Frank Adcock and Professor A. W. Gomme for meir comments on this article. I have not given references to all the works on this subject. I append a list of those to which I shall refer under the name of the author alone, unless he has written more than one work:

Beloch, K. J., Klio 9 (1909) 477Google Scholar; 11.431; 13.128; Arch. Ephem. 1910 383; Griechische Geschichte 2². 2. 107.

Dähnhardt, O., Scholia in Aeschyli Persas (1894).Google Scholar

Goodwin, W. W., Papers of the American School of Classical Studies at Athens 1 (1885) 237.Google Scholar

Grégoire, H., Les Études classiques 4 (1935) 519.Google Scholar

Grundy, G. B., The Great Persian War (1901).Google Scholar

Judeich, W., Klio 12 (1912) 129.CrossRefGoogle Scholar

Keil, J., Hermes 73 (1938) 329.Google Scholar

Kromayer-Veith, , Antike Schlachtfelder 4 (1924–1928) 64 by Keil, W..Google Scholar

Labarbe, J., BCH 76 (1952) 384.CrossRefGoogle Scholar

Legrand, Ph. E., Revue des études anciennes 38 (1936) 55.Google Scholar

Lolling, H. G., Aufsätze E. Curtius gewidmet (1884).Google Scholar

Milchhöfer, A., Text zu Curtius und Kaupert, Karten von Attika, II and VII (1883–1895).Google Scholar

Munro, J. A. R., JHS 22 (1902) 321CrossRefGoogle Scholar; Cambridge Ancient History 4 (1926) 304.

Myres, J. L., Herodotus father of history (1953).Google Scholar

Positano, L. M., Demetrii Triclinii in Aeschyli Persas Scholia (1948).Google Scholar

Rediades, P., plan in P-W. RE s. Salamis col. 1830 with references.

Rodgers, W. L., Greek and Roman Naval Warfare (1937).Google Scholar

Tarn, W. W., JHS 28 (1908) 202.CrossRefGoogle Scholar

The maps and the geographical data are derived from the Greek Staff Maps 1/20,000 of the island; 1/100,000 of Greece; and for civil divisions on the recent map 1/100,000 of the Greek Statistical Division of the Staff; Admiralty Chart 894; and the Mediterranean Pilot 4 (1941) 107 f.

¹ Thuc. 2.19.2; Schol. Dem. 24.129; Aristodemus I 2 (FHG V), regarding Mt. Aegaleos as an extension of Mt. Parnes, placed Xerxes' throne

² Hdt. probably refers here to the town which is a much more accurate point of departure for the description than the island of Salamis; so, too, Diodorus 11.18.3,

³ The sanctuary of Artemis was probably on the hill of Kamateró, where a decree of her thiasotae was found (Milchhöfer 7.27; IG 2.620). The trophy, with the cult of Zeus Tropaios, was on a peninsula (Schol. A. Pers. 303, ), which may be identified with Cape Kamateró (see Appendix). Milchhöfer and others prefer Cape Varvári and suggest that the cape had two names, Cynosoura and Tropaea. The Polyandreion, which is known from the inscription first published in Ephem. Archaeologike 1884, p. 169, was the tomb of the Greek dead; its site is not known, but it was probably not far from the trophy. There is no doubt that Kamateró is the site of classical Salamis town; when Strabo mentions an early Salamis town which faced south towards Aegina (9.1.9, C 393), he is presumably speaking of the pre-classical or ‘Homeric’ town.

⁴ For καὶ ἄλλη meaning ‘there is another Atalante’, compare Steph. Byz. s.v. Oropos.

⁵ Mr. V. Deleyannis, mentioned in n. 1. Cf. Goodwin 257 ‘the only passage over which it would not have seemed insane even for Xerxes to attempt to build a causeway to Salamis is from this point of Aegaleos (a little north-west of the ferry), over the shoal above mentioned, to the island of St. George and thence to Salamis’; despite this, Munro, , CAH 4, 305Google Scholar, thought that Xerxes intended to build a causeway to the tip of Cape Varvári.

⁶ Leake, , Die Demen von Attika 159Google Scholar, mentions two possible sites for the quarry, one on the peninsula of Skaramangá, where there is an artificial mound of excavated soil, and the other on the coast east of the reef of Arpedhóni, where he speaks of a ‘small ancient quarry’. Leake evidently preferred the former site for Strabo's quarry. Lolling, p. 6, mentions a quarry near Pérama and places it on the hill just north-west of Pérama on his map; Lolling claims it is the only quarry between Pérama and Skaramangá, and denies Leake's two sites without justification. It is clear that a suitable site is available whether Cape Amphiale is equated with Cape Filatoúri or with the bend of the Channel by Pérama which significantly is not named as a cape on modern Greek maps, and the precise dating of a quarry is too difficult to permit one to say that one and not another of the three sites was quarried in Strabo's day. As regards Heracleum, Lolling recognises that for the purposes of the mole, Heracleum must be at or near Pérama, but he wishes also to locate a Heracleum at Keratsfni Bay; his argument that the former term is used ‘in a wider sense’ and the latter ‘in a narrower sense’ would be acceptable only if Heracleum was a deme and not a shrine.

⁷ Strabo 322 gives the length of his stade. Lolling, p. 7, would amend to and Müller, C. (FGH V, p. 1)Google Scholar to

⁸ Goodwin, p. 241, reckoned 600 yards of navigable water, allowing for the shoals on each side of the fairway.

⁹ See Godwin, H., in The New Phytologist 46.1.29 f. (1945), especially p. 65Google Scholar, to whose kindness I owe this information.

¹⁰ Scylax, writing probably c. 338–335 B.C. (GGM, p. XLIV), describes Ambracia as up the river and continues (ch. 33). I visited the site of Ambracos and found that the footings of the circuit-wall are only a foot or two above the present sea-level and that the enclosed area is all under shallow water. When the sea is rough the walls are sea-washed. It may be assumed that in the fourth century B.C. most of the enclosed area was dry land and the walls were not sea-washed. If follows, then, that since the fourth century B.C. the sea-level has risen by some 4 or 5 feet at least. I noticed evidence of a similar rise in sea-level at Actium and at Treporti on the north shore of the Gulf of Valona, and the sea has risen since antiquity to invade the cave which was excavated by Miss Benton in Ithaca. Mr. R. M. Cook tells me that in the sea-fed lake east of Perachora there are the foundations of an ancient house which are now in sea-water.

¹¹ The Greek seaman, Mr. Deleyannis, considered that the ship-sheds at the Peiraeus and other foundations indicated a rise of one metre or more in the level of the sea. Mr. E. J. A. Kenny tells me that the slip-way at Sunium seems to extend farther into the sea than is now necessary, but he has no precise measurements. See note 36 below for the walls at Salamis, and Milchhöfer 7.29 for the presence of walls under the sea on the west side of die Bay of Ambeláki.

¹² That there has been since antiquity no large rise of sea-level (e.g. by some 20 feet), is clear enough from the position of ancient harbours. In the case of the Salamis Channel such a rise would mean that in antiquity the present islands of Nera, Kyrádhes, Áyios Yeóryios, and the rock off Pérama would be joined to the adjacent coasts, and Strabo's account would be impossible to explain at all.

¹³ The reefs known as Skróphes (off Selínia) and Arpedhóni (in the northern arm of the Channel) would remain almost the same in size, and neither would be joined to the adjacent coast.

¹⁴ In the short passage, which I have quoted, Strabo uses καί, meaning ‘also’, four times although it is never strictly necessary in logic. This is so in the phrase where it would be enough to say ‘it is like Psyttalia’ without adding ‘also’ to suggest it is rocky and uninhabited. This peculiarity of Strabo's diction is an argument against rather than in favour of emending, as, for instance, Beloch did in supplying after καὶ τοῦτο the words (Klio 11.432). Milchhöfer 2.46 and Honigmann (P-W. RE s. Korydallos) believe tfiat Aegaleos was the early name and Corydallus the late name for the same mountain, but it should be noted that Diodorus 4.59.5 probably borrowed it from an early source and Theophrastus used it (Athen. 9.390). In the modern divisions of this area (see sheet 81, 1: 100,000 Greek Staff Map, ) Mt. Skaramangá is in Megaris and Mt. Aegaleos in Attica.

¹⁵ A further clue to the position of Phoron Limen is afforded by Bekker, Anecdota 315Google Scholar (a reference which I owe to the kindness of Mr. C. W. J. Eliot): The boundary between Attica and Salamis probably followed the line of the Channel; in the modern civil divisions the line of demarcation between the Eparchia of Megaris and that of Attica runs west and north of Áyios Yeóryios and eastwards through the beach of Amphiale, Salamis island belonging to Megaris. It should be noted that Lipsokoutáli is exposed to rough seas, is steep-to, and has only one poor landing-place on its north-west side (Milchhöfer 7.29); it is unsuitable for ships to lie to, whether at anchor or not It cannot, therefore, have been an eyesore of the Peiraeus in the sense which we attach to the word, and it is also likely to have lain within the waters controlled by the Peiraeus authorities in the fourth century B.C.

¹⁶ Strabo often uses πλησίου in a very loose and general sense, e.g., of the Rivers Acheron and Thyamis, which are some 30 miles apart (7.7.5, C 324).

¹⁷ The bracket in the Oxford Text may be unnecessary, since Steph. Byz. s.v. Psyttaleia quotes Herodotus' words, The same ambiguity is inherent in most of Herodotus' mentions of ‘Salamis’ (8.74.1; 76. 1; 78; 82. 2; 86; 89. 1; 90. 4; 95–7).

¹⁸ The occurrence of the phrase is peculiar in the context of the battle; for it is not used in the many other mentions of Athenians. As their descent is stressed, it is probable that they are Athenian cleruchs at Salamis and were normally known as ‘Salaminioi’. A parallel case is supplied by the 4,000 cleruchs at Chalcis. They are referred to as ‘Athenians’ in contrast to the Eretrians (Hdt. 6.100), but generally, it seems, as ‘Chalcidians’. Thus before the battle at Artemisium Athens gave twenty of her ships to the ‘Chalcidians’ who manned them, the normal full complement being 4,000 men (Hdt. 8.1.2); these ‘Chalcidians’ fought also at Salamis (Hdt. 8.46.2). Now it is most unlikely that so many native Chalcidians were so friendly or loyal to Athens that they would have fought at Salamis, or indeed that Athens would have entrusted twenty of her ships to the native Chalcidians. This consideration and the similarity in number of the cleruchs and of the men required to man twenty ships make it clear that these ‘Chalcidians’ were the Athenian cleruchs of Chalcis. In short, then, Herodotus normally named cleruchs not by the city of origin but by the city of residence. In the fourth century we have an interesting parallel in Athenodorus, who was described as an ‘Imbrian’ and also ‘by descent a citizen’ of Athens (Plu. Phoc. 18; Dem. 23.12). There is not space here to develop the argument, but it casts some light upon the terminology of Thucydides in 7.57.2–3. The Chalcidians named on the serpent-column and in Paus. 5.23 may refer, then, to the cleruchs at Chalcis, and they may have provided the 400 hoplites who fought at Plataea beside contingents from Euboea (Hdt. 9.28.5 and 9.31.4). Bury's suggestion in CR 10, 416 f., that these hoplites on Salamis were ‘by far the greater part’ of the Athenian hoplite army, does not account for the peculiar use of the phrase and sets an impossibly high figure for the Athenian population. Athens manned 180 triremes with crews which totalled some 36,000 men drawn from (Hdt. 8.44.1 and Plu. Them. 10, excluding slaves); these, says Bury, were ‘a minority’ of the whole, but he does not proceed to the deduction that Athens must have had an adult male population of 80,000 or more to justify his statement. The unanimous tradition that the Athenians ‘embarked on the ships’ (Thuc. 1.18.2) should not be called in question.

¹⁹ To suppose Diodorus to say ‘they descended together from the (island of) Salamis’ is to postulate an unnecessary vagueness.

²⁰ This phrase (misquoted by Beloch, Klio 8.480) is a periphrasis for ‘Salamis’.

²¹ Aristodemus 1.4, may refer to Salamis town or Salamis island; so too Steph. Byz., indicating that at some time the island was inhabited, although this was not so in the days of Herodotus and of Strabo. Pliny, , in NH 4.62Google Scholar, ‘Salamis, ante earn Psyttalia’, is giving a list of islands, and thus refers to Salamis the island.

²² The view advocated by Munro, , CAH 4, 305Google Scholar, Grégoire, pp. 519f., and Myres, p. 274, that Cape Cynosoura is to be located in the bay of Marathon and Ceos is to be identified with the island off Sunium, is entirely inconsistent with the narrative of Herodotus; it would have been a long and purposeless voyage of up to 70 miles for a part of the Persian fleet to undertake during the hours of darkness until midnight (Hdt. 8.76.1), and it would not have returned in time for the action.

²³ In view of the ancient evidence I do not attach any importance to the occurrence of modern place-names which are similar to ancient place-names. The survival of ancient place-names except in the case of large islands is rare, and their survival at the original site is rarer still; Salamis as the name of a town, Marathon and Corinth are obvious examples of a place-name surviving but migrating to another site. Klephtikolimani, which Curtius, E., Altertum u. Gegenwart 2.96Google Scholar, says was a current name for Keratsíni Bay, has the same meaning as Phoron Limen, but both arise from smuggling, and a smuggler's choice varies with political and social conditions. As Keratsini Bay is not far from the Peiraeus entrance, it is most unlikely that it was outside the control of the Peiraeus customs authorities in the days of Demosthenes. The name Talantónisi may represent the ancient Atalante, but in its modern and meaningless form it may have been displaced from its original site by the modern and meaningful name Lipsokoutáli, ‘the defective soup-ladle’. We may be warned by the modern name Selinia; on the face of it Σελήνια may be a survival of the ancient name Σιληνίαι (A. Pers. 303), but its present location is one of the few which do not fit Aeschylus' description. Its beach is sandy and not oruqAos, and the corpse of Artembares could not have been washed there because the fighting was nowhere near modern Selinia and the wind was westerly when the drift began (Hdt. 8.96.2). See note 85 below.

²⁴ J. Keil, for instance, p. 334 n. 1, attributes Xerxes' length of sight to ‘poetic exaggeration’; but this is not the type of poetic exaggeration to which Attic Tragedy is prone, and Hdt. 8.86 fin., 88.2, 89.2, 90.3–4, which Keil does not mention, cannot be explained away as poetic exaggeration. If Xerxes' throne is moved to within human sight of the fighting on Psyttalia–Lipsokoutáli, it has to be taken to the hills beside the Peiraeus, which dissociates it from Mt. Aegaleos.

²⁵ The verb διαχοῦν is translated in L-S-J ⁹ ‘to complete’ a mole, but Strabo (9.1.13, C 395), does not use it in this sense.

²⁶ By assuming that the accounts of Aeschylus and Herodotus are both comprehensive, most scholars detect inconsistencies between them. They also dismiss the evidence of Diodorus as ‘built up by the reflection, inference, rationalism and conjecture’ of Ephorus (e.g., Munro, , JHS 22, 239Google Scholar); but this is to overlook the possibility, and I think the probability, that Ephorus drew on sources independent in some respects of Aeschylus and Herodotus. The same may be said of Plutarch. A less severe attitude towards Aeschylus, Herodotus, Diodorus, and Plutarch was taken by Goodwin 245 f.

²⁷ Klio 8, 484 f.

²⁸ JHS 28, 226; the only support for his view is in Aristodemus 1.4, a passage clearly dueto an error of understanding and not mentioned by Tarn.

²⁹ Études classiques 4, 519 f.; see the answer to Grégoire by Legrand, pp. 55 f.

³⁰ Löschcke, , Jahrb. f. Philol. 115, 25Google Scholar; Beloch, , Klio 13, 130Google Scholar; Wilhelm, , SB Akad. Wien 211.1.28Google Scholar; Jacoby, , Philologus 86, 369Google Scholar; J. Keil, p. 340.

³¹ E.g., by J. Keil, p. 335.

³² J. Keil, for instance, does not consider the evidence for the Greek position which is contained in Persae 386–98.

³³ For instance, Tarn, p. 226, held that the Greeks out numbered the Persians, which is a bold reversal of the evidence. A healthy reaction against such methods has begun; for in the most recent discussion by Labarbe, pp. 421 f., Herodotus' figures for the Greek fleet are accepted.

³⁴ The interpretation which I give of A. Pers. 337–44 is in accordance with the normal meaning of χωρίς and of μέν and δέ.

^34a On the disputed passage in Thuc. 1.74.1, (or with one codex ) several views are possible. Gomme, op. cit. 1.234, suggests reading or which would dispense with the rather meaningless γε and is closely paralleled by the phrase in Hdt. 8.44: This would mean that the Athenians, providing 200 ships at Salamis, claimed to have provided a little less than two-thirds of 300, i.e., less than 200, which is correct if they did in fact provide 180 (Hdt. 8.44); on the other hand, if they claimed to have provided rather less than two-thirds of 400 ships, this being a round figure for 380, the claim is an extravagant one, even if the twenty ships of the Chalcidian cleruchs are added. Labarbe, p. 419, keeps the text and reads τριακοσίας, interpreting the passage to mean that Athens supplied up to 300 ships (in fact, some 270 on his calculation) and the total Greek fleet was some 450 ships; he then refers these figures to the beginning of the war. The objection to his view is less in the figures than in the context; for Thucydides appears to be speaking of the decisive Battle of Salamis. A third view is to keep the text, but assume that the Athenian claim is made with reference not to triremes but to warships of all kinds; on this interpretation the Greeks had a total of more than 600 ships and Athens supplied up to 400 ships, and the former figure is in accordance with Ctesias' statement that the Greeks had 700 ships at Salamis. But, whatever is done with this passage in Thucydides, the evidence of Aeschylus and Herodotus is more authoritative.

³⁵ Labarbe p. 425 n. 5, holds that the word ναῦς in the epitome of Ctesias must mean a trireme, but this technical meaning may not be necessary in an abbreviated version.

³⁶ Myres, p. 281, allows 7 yards, which seems to be too tight, if one envisages the launching of a heavy ship by manual methods. For the hull amidships was 5 yards wide (see p. 50 below for the dimensions of the trireme).

³⁷ The jagged rocks of Cynosoura promontory, which are apt to cut a bather's feet, begin at the knob of rock on the south side of the mouth of Ambeláki Bay; they render the station of the Greek fleet proposed by Munro (CAH 4, map 9) completely impossible.

³⁸ It is unlikely that there has been any marked silting up of the Bay of Ambeláki since antiquity, because the area draining into it is so small. Moreover, the remains of Hellenic walls (shown in Lolling, p. 9 and map facing p. 10) are now close to the sea and almost at sea-level. If the walls were built to enclose a habitable area, then the coast-line was more distant in antiquity than today and the sea-level was correspondingly lower (see n. 10 and 11 above for indications of a change in sea-level).

^38a This is consistent with Herodotus' usage of πορος for the narrows of a strait, e.g., 7.176.1 and 183.3 between Sciathos and the mainland and 7.34 from Abydus to the European coast. It is interesting that Aeschines 3.158, in speaking of the ferrymen who plied to Salamis described the water they crossed as ὁ πόρος.

³⁹ Hdt. 8.70.1 and 76.1 is more compendious than Aeschylus, but he clearly indicates an interval between the approach of nightfall and the operations at midnight, during which the Persian fleet was making its preparations not in the open sea but at its base.

⁴⁰ The same manoeuvre is described in more grandiose terms by Plutarch Arist. 8: and Them. 12

⁴¹ The inference to be drawn from this statement is that the visibility was such that the Greeks did not see the part of the Persian fleet which was closing the eastern exit of the Salamis Channel. On the other hand, they apparently saw some Persian ships to the south of Ceos, a sight which led them to suppose the Persians to be in the same position as on the pre ceding afternoon. Otherwise it is difficult to understand how they believed the Persians to be in this position during the night. This raises the question of moonlight on this night. Busolt, , GG 2 ², 702Google Scholar, Beloch, , GG 2 ², 2, 47Google Scholar and Myres, p. 270, state that ‘the moon rose at midnight or after’, but they do not explain why Xerxes waited for moonlight before he put his fleet out in an advanced position, hoping to escape observation. Goodwin, p. 242, refers to the two dates given by Plutarch for the battle, namely ‘16th Munychion’ and ‘about 20th Boedromion’, and to his statement that the Greeks dedicated the 16th Munychion to Artemis: (de Glor. Ath. 7 and Camillus 19). As Munychion corresponds roughly to April and Boedromion to September, the mention of Munychion in Plutarch is clearly an error, and we may conclude either that the passage is to be rejected in its entirety or that it is an error only of Munychion for Boedromion (the same error occurs in Plu. Lys. 15, where he puts the Battle of Salamis and the intervention of Lysander in 404 B.C. on 16th Munychion instead of in Boedromion, for which date see Fuks, A., The Ancestral Constitution, p. 70Google Scholar). If the latter alternative is adopted, we are left with two dates in Boedromion, the 16th and ‘about the 20th’. Moreover, the 16th of Boedromion is the date given in Polyaenus 3.11.2. The moon was full on the 16th and was waning on the 20th; in either case the moon rose early in the night and set late. As the moon rose in the east it would be riding in the south about midnight. During a moonlit night at sea one benefits by the reflection of the moon's rays off the water to see a fleet between oneself and the moon. In order to deprive the Greeks at Salamis of this advantage, Xerxes waited until midnight to bring his fleet forward; and he moved his force of Persians on to Psyttalia when the moonlight was not to his disadvantage. In this calculation there are several imponderable factors: we do not know whether Plutarch's statement is dependable in its confused form and whether the night before the battle was cloudy, hazy, or clear. If I am correct in believing that a south wind was blowing out at sea, the probability is that the sky was cloudy or hazy during the night.

⁴² Herodotus names as ‘west’ any direction which is west of north (cf. 7.36.2 and 7.176.3).

⁴³ I translate the full passage as follows: 76.1—‘On the one hand, they landed many of the Persians on the islet Psyttalia which lies between Salamis and the mainland. On the other hand, when midnight came, they were moving out at sea, curving their western wing on the one hand towards Salamis and being already in formation on the other hand off Ceos and Cynosoura, and they occupied the entire strait with their ships as far as Munychia.’ 76.2—‘They were moving the fleet out at sea with the intention of preventing the Greeks from escaping and of ensuring that they should be cut off at Salamis and pay the penalty for their actions at Artemisium. They were landing some of the Persians on the islet Psyttalia with the intention that, when a battle developed and men and wreckage were cast up mainly in that area (for the island lay in the narrows where the battle was likely to take place), the Persians would aid their friends and destroy their enemies.’ 76.3—‘They were execut ing these movements in silence, so that they should not be observed by their opponents.’ While the aorist tense of ἀνάγειν or ἀνάγεσθαι with or without τὰς νέας ‘to put out to sea’ from the shore (e.g., 3.41.2 ἀναγαγεῖνν, 7.100.3 ἀναγαγόντες, 3.138.1 ἀναχθεντες), the imperfect is used of ships moving out at sea (e.g. 8.70 and in the participles 6.12.1 8.83.2 ). The use of κέρας shows that Herodotus is speaking of a continuous formation and not of a detached squadron, and the μέν and δέ refer to the two wings or extremities of this continuous formation, which extended in depth towards Munychia. The imperfect tenses in 8.76.1 mark the continuous movement of the fleet (cf. A. Pers. 382, ). The purpose of the movements is clearly stated in 8.76.2, and the need for silence was due to the proximity of the enemy. I find no difficulty in the Greek of this passage, which was a stumbling block to Goodwin (pp. 251 and 261–2) and others in their plan of the whole engagement.

⁴⁴ I noticed the remains of a tower at the spot marked X on map 2, which does not figure on Milchhöfer's map, op. cit. Bl. XXI. The masonry is probably of late-fifth- or early-fourth-century date, and the occasion for its construction may have been the Corinthian War, when a Spartan fleet lay at Aegina; it would serve as an observation post and as a blockhouse against a landing of troops designed to raid Salamis town. Greek observers on the night before the battle may have been posted on this part of the peninsula rather than on the tip, which is less high.

⁴⁵ The question, whether the oracle was a vaticinium post eventum or not, cannot be resolved; but it should be noted that the collection of Musaeus' oracles from which it came (Hdt. 8.96.2) was in general of earlier origin, that the Greeks them selves were aware of one case of tampering with oracles (Hdt. 7.6.3), and that Herodotus himself was convinced that this oracle was ante eventum. But the question is not relevant to the bearing of the oracle on the position of the Persian fleet.

⁴⁶ The word γεφυρόω may mean also ‘to dam up’ or ‘make into a causeway’ in the sense that a pontoon-bridge forms a causeway.

⁴⁷ Beloch, , Klio 8, 477–8Google Scholar and others assume that ‘the shore of Artemis’ is the Acte of Peiraeus, since there was a sanctuary of Artemis at Munychia (Paus. 1.1.4). They believe, then, that the oracle refers to the midnight position of the Persian fleet when its formation extended from off Cape Cynosoura to Munychia. This interpretation, however, does not accord with Herodotus' understanding of the oracle; for he quotes it not to explain his words at 8.76.1 but to point forward to the battle itself (8.96.2). Beloch's argument that one makes a bridge to connect two points which are opposite to one another is only one aspect of the truth; for in Greek, as in English, one can also speak of bridging a river or a strait (Hdt. 4.88; 118).

⁴⁸ So long as the Persian fleet threatened the eastern entry into the Salamis Channel, the quickest crossing from Aegina to the Greek headquarters at Salamis town was the direct one to the western side of the waist of the island, whence it is a short walk or ride to Kamateró. This route must have been much in use while Aegina was held by the main Aeginetan fleet. Aristides will have guessed that the Persian ships off the western bay moving northwards intended to close the exit from the Bay of Eleusis. This guess was confirmed at dawn by the Aeginetan ship, bearing the images of the Aeacidae, which had come by sea from Aegina (Hdt. 8.83); the only possible route for her was through the Bay of Eleusis. Bury's, suggestion in CR 10, 418Google Scholar, that Aristides in fact travelled on the Aeginetan vessel which brought the images of the Aeacidae, is contrary to the statements of Herodotus and Plutarch; it entails the consequence that the Greeks were informed first by the Tenian trireme, that Aristides arrived when the Greeks were already embarking at dawn (Hdt. 8.83.2) and that the whole story of Aristides entering the conference of commanders (Hdt. 8.79–81 and Plu. Arist. 8, especially ) has to be discarded as nonsense. Bury's suggestion encouraged Grundy, p. 391 n., to accuse Herodotus of having mistimed the arrival of the Aeginetan ship and Beloch, , GG 2 ², 2, 121Google Scholar to dismiss Herodotus' account of Aristides' arrival as unhistorical. The choice between them and the ancient authorities is an obvious one. For they, having scrapped Herodotus and Plutarch, have no evidence for their view, but Herodotus no doubt spoke to combatants at Salamis and used contemporary accounts.

⁴⁹ For the timing see below, p. 51.

⁵⁰ This essential point, that the Greek fleet was at first out of sight of the Persian fleet and then came swiftly into sight, is not met by many reconstructions of the battle (e.g., J. Keil, plan on p. 333 and other plans in Kromayer-Veith, map facing p. 106).

⁵¹ Whereas Aeschylus mentions the disappearance and the reappearance of the Greek fleet from the Persian point of view, Herodotus omits this step in his narrative of the Greek fleet's preparations. He simply says that the whole Greek fleet was already at sea when the Persians bore down upon them (Hdt. 8.83 fin.). In the same way, whereas Aeschylus, Persae 374–9Google Scholar describes the preparations of the Persians on the preceding evening, Herodotus omits this step and proceeds to midnight, when the Persian fleet was already at sea in formation (Hdt. 8.76.1). In these respects the two accounts happen to be complementary. They do not contradict one another.

⁵² See below, p. 49, for a discussion of the Corinthians' task.

⁵³ See Map 1. This precise description of the Greek position is not taken precisely by many scholars, e.g., by J. Keil, p. 335 n. 2, who says it means only that the Greek fleet lay between Attica and Salamis.

⁵⁴ My informant is Mr. Deleyannis; see note 1.

⁵⁵ The opposing ships were, I take it, facing one another more or less bow on, like sparring boxers, before the swell came up. There was sufficient space between the opposing ships for a ship to make a racing start and gain enough momentum to use its ram.

⁵⁶ Plutarch's account is not to be confused with the story of the dust-cloud in the Thriasian plain (Hdt. 8.65), which was seen by the exiled Athenian, Dicaeus, and by the exiled Spartan king, Demaratus, a day or two before the battle. Many supernatural sights were associated with this campaign.

⁵⁷ I take it that the oarsmen's vision was limited to what they could see through the oar-ports and that those on deck would be facing in the direction of the enemy. On my disposition of the Athenian ships some of the oarsmen on the bow side saw Eleusis. That the westward part of the lines of battle were within sight of Eleusis when the engagement began is almost certain from Herodotus' phraseology alone. At 8.76.1, when he describes the position of the Persian fleet at midnight, he mentions simply ‘the west wing’ at 8.85.1, when he describes the opening of the engagement, he uses the fuller phrase In this phrase ‘pointing in the direction of Eleusis’ is no idle addition. It means that at midnight the Persian right wing was not pointing in that direction, but next morning it was. So, too, the Persian left wing was pointing towards Peiraeus and was in sight of Peiraeus. Both phrases become nonsensical if the scene of the battle is supposed to be outside the eastern exit of the channel, as in the plans of Rediades, Munro, Myres, etc.

⁵⁸ The retarding of the Persian left wing, which was held by the Ionian squadron, may have lent colour to the view that the Ionians fought backwardly; for in the action they did well (Hdt. 8.85.1).

⁵⁹ For the Persian fleet now thronged the Channel from Attica to Cynosoura and to the shore of Artemis by Salamis (cf. p. 45 above).

⁶⁰ Goodwin, pp. 257–8, failed to see the object of the Greeks in backing water, and, despite the statement of Herodotus, assumed not a withdrawal but a further advance which carried the Greek fleet out of sightof Eleusis.

⁶¹ The importance of such a position is clear from the action off Artemisium (Hdt. 8.16).

⁶² The best appreciation of Themistocles' genius is given by Thucydides, who saw that Themistocles was responsible for the battle being fought and that this saved the situation (1.74.1). In his strategy Themistocles showed the qualities for which Thucydides praised him (1.138.3): the finest judgement in immediate situations which admitted of least deliberation, the most accurate insight into the future development of events, and the ability to improvise the necessary measures.

⁶³ The Phoenician captains had a short climb to Xerxes' throne, where they were promptly decapitated (Hdt. 8.90.1–3).

⁶⁴ Or on the reef of rocks off Pérama to which Timotheus, , Persae, 97 f.Google Scholar may refer: So, too, A. Pers. 421

⁶⁵ Honigmann, in P-W. RE 11, 1, col. 1077Google Scholar, discusses the difficult problem of the identification of Κωλιὰς ἄκρα. Strabo (9.1.21, C 398) places the temple of Aphrodite Colias not far from Cape Sunium near Anaphlystus (now probably Anávyso) and adds that this was the spot to whichthe last wreckage of the Persian fleet after the battle was cast up by the waves I take it that Strabo means that the wreckage extended along the coast as far as Cape Colias and that a heavy sea running on to the coast, which had been raised by a southerly wind in the Aegean Sea, washed the wreckage up on that stretch of coast. Honigmann rejects Strabo's placing of the temple of Aphrodite Colias and prefers to put Cape Colias at the Cape of Ayios Yeóryios just east of Phalerum Bay (following Milchhöfer, 2, 2). Although it makes little difference to my argument, I think that Strabo's evidence is too strong to be rejected in this manner and that Cape Colias should beplaced at Cape Anávyso.

⁶⁶ The figure of seventy ships is obtained by subtracting Aeschylus' figure of 310 Greek ships at the opening of the action from Herodotus' total of 380 ships at the concentration of the fleet at Salarais. The Corinthians had forty ships, and their colonists at Ambracia and Leucas had ten ships, which probably acted with the Corinthians. Among the remaining twenty ships it is probable that six Spartan ships under Eteonicus and some Aeginetan ships were included (Lycurg. c. Leocr. 70; for the number of Spartan ships see Hdt. 8.43 and Isoc. Panath. 50). This point is excellently made by Labarbe, pp. 434 f. I differ from him in my view that the Corinthian force did not engage the Egyptian squadron but returned to take part in the main action; the words of Hdt. 8.94.4, which Labarbe seems to take in a temporal sense, are to my mind qualitative and mean that the Corinthians considered themselves to be among the most distinguished fighters.

⁶⁷ Cf. Bürchner, and Kock, in P-W., RE, 1, 2Google Scholar, col. 1829, and 3, 1, col. 534. If the island Nerá was in antiquity a cape of Salamis, it is a possible site for the temple.

⁶⁸ Their absence has often been noted. Herodotus refers to the Egyptians in two speeches only, where the King's subjects are enumerated as a worthless lot (8.68.7; 8.100.4).

⁶⁹ In summer the normal wind of the early morning breathes off shore into the Bay of Eleusis (so also Grundy, p. 398 note), and comes down the Salamis Channel as a north-westerly or westerly breeze, raising no sea, but adverse to anyone sailing up the Channel. On this day in September the weather was not normal; for the Corinthians hoisted sail to proceed up the Channel, and, as a trireme under sail was aided only by a following wind, it is clear that the wind was southerly at dawn. About 9 a.m. (on my approximate reckoning) an offshore wind was breathing offshore into the Bay of Eleusis and carried some mist from the Eleusinian plain towards the Channel (Plu. Them. 15), where it settled suddenly. It is probable that the weather was fluky in the Channel, because a southerly wind had been, and perhaps still was, blowing hard out at sea. Such conditions were to be expected in September when the Battle of Salamis was fought. Compare the proverb of the Middle Ages quoted by Rodgers, p. 85, ‘in the Mediterranean there were only four good ports for a fleet, namely June, July, August and Port Mahon’, and note the storms which befell the Persian fleet in August.

⁷⁰ My informant is Mr. Deleyannis (see n. 1). The only local wind which creates a sea in the Salamis Channel is the strong north-westerly wind, common in winter, which makes the passage from Pérama to Kamateró difficult. In the ancient evidence about the winds on the day of the battle the only problem is contained in the words of Plutarch (Them. 14) that Themistocles waited for to bring up the strong wind from the sea and the swell. He may refer to the fact that the wind normally is strong about 9 a.m., but it does not apply necessarily to a southerly wind of the type which brings up the swell from the open sea.

⁷¹ For the position of Cape Colias see n. 65. I obtained this information about wind and drift from Mr. Deleyannis. The Mediterranean Pilot, 4, III, comments on the strength of the Maistro.

⁷² Grundy, p. 399, says the Greek ships carried thirty-six marines, which seems to rest on no ancient evidence.

⁷³ Rodgers, p. 55, misled perhaps by modern analogies, believes that the early trireme was less large than the later model, but he forgets the change in tactics and its effect on ship construction in the second half of the fifth century B.C.

⁷⁴ Rodgers, p. 30, puts the longest oar for one oarsman at 14 feet and demonstrates that almost one-third of the oar must be inboard of the thole-pin. Assuming that no outrigger was used, since it would be vulnerable in boarding and does not appear in vase-paintings, etc., of triremes, I conclude the width of a trireme under oar was 35 feet amidships. Some scholars reckon 45 feet, e.g., Myres, p. 281.

⁷⁵ Grundy, p. 396, allows 60 feet per ship, that is an interval between ship and ship of only 25 feet, which seems to be too small for the ships to be able to manoeuvre freely.

⁷⁶ Rodgers, pp. 89–90, allows 25 feet and 45 feet of open water in line and in column respectively for the Persian fleet during its advance.

⁷⁷ Rodgers, p. 87, ‘certainly the speed was not over 2 miles an hour’.

⁷⁸ Rodgers, p. 53, gives examples of trireme speeds. The night was calm; for the main Persian fleet was under oar during the night.

⁷⁹ See note 48 above. I am assuming that he went on a ship as fast as a trireme; if he went on a slower vessel, he left before the Aeginetan ship and still arrived well before dawn.

⁸⁰ See note 41 above.

⁸¹ See note 41 above.

⁸² Xerxes' men, who held the area by the Heracleum, had no doubt observed during the day that Psyttalia was not occupied.

⁸³ Xerxes probably had a throne prepared for him on Mt. Kérata near these narrows, so that he could watch a battle there; this makes sense of Acestodorus' statement which Plutarch, , Them. 13Google Scholar, records without explanation. There are no grounds for supposing, as Rediades does, that a spur of Mt. Aegaleos was called Kérata and that Acestodorus transferred it to the borders of Attica and Megaris.

⁸⁴ The difficulty of such a landing is clear from Thucydides 4.10.5. The same difficulty caused Datis and Artaphernes to land their army at Marathon, where they were unopposed, and not to attempt a landing at Phalerum Bay after the battle when they were opposed. The alternative policy of blockade, which was suggested in the speech attributed to Artemisia (Hdt. 8.68), would have been difficult to execute during the bad weather of autumn; for the coast-line of Salamis is a long one, and the example of Sphacteria is Illuminating.

⁸⁵ In this article I have not discussed the identification of the for which Lolling, p. 8, and Milchhöfer, 7.28, give the references, as they shed little light on its position and do not affect the location of the battle.

⁸⁶ The only point not yet mentioned which may be regarded as inconsistent is the statement in Diodorus 11.17.3 that the Ionian commanders sent a Samian swimmer to inform the Greeks of Xerxes' plans and of their own intention to desert. This statement was probably derived through Ephorus from an early source. For Herodotus remarks that only a few of the Ionians fought backwardly (8.85.1), and he discounts the charge made by the Phoenicians that the Ionians betrayed the fleet (8.90.1). In these passages Herodotus was probably refuting such a claim as appears in the narrative of Diodorus. Whether an individual Samian swam across during this eventful night, there is no means of determining. The mention of a Tenedian instead of a Tenian trireme (Plu. Them. 12) is presumably a slip by him or by his copyist.

⁸⁷ To catalogue the errors attributed by modern scholars to the ancient writers, and Herodotus in particular, would serve little purpose, and a few examples may suffice. Grundy further believes that Herodotus forgot ‘the enormity of the losses he represents the Persians to have suffered in the two storms’ (p. 374), that Herodotus made a mistake as to the time and means of Aristides' arrival (p. 390), and that Herodotus and Diodorus both erred in not placing the Athenians and the Aeginetans next to one another in the battle (p. 400).

Munro, , JHS 22, 326–7Google Scholar believes that Herodotus ‘holds a brief to vindicate the veracity of Bacis and tries to force the situation at Salamis into conformity with his oracle’, ‘Herodotus' conception is hopelessly irreconcilable with the descriptions of Aeschylus’, ‘Herodotus would seem to have antedated the start from Phalerum to the afternoon and postdated the envelopment to midnight’, ‘Herodotus has simply adapted or misinterpreted his information to suit his preconceived idea of the battle’, and both Aeschylus and Herodotus were mistaken in supposing Psyttalia to have been occupied for the purpose of assisting in the naval battle. In CAH 4, 304 f. and map 9, Munro adds to this list a spoken or tacit rejection of Herodotus' and Diodorus' descriptions of the Greek order of battle by putting the Athenians on the right wing pointing towards Salamis and not Eleusis and the Aeginetans, etc., on the left wing facing south, so that the Greek seamen would have had to swim through the press of ships to reach Salamis (Hdt. 8.89.1) and the Phoenicians would have been pushed through the Aeginetan ships to reach land near Munro's site for Xerxes' throne (Hdt. 8.90; D.S. 11.9.2). To attribute to Herodotus ‘a childish misconception of the battle’ in this wholesale manner is justifiable if one regards Herodotus as aman of childish intelligence in these matters and if one assumes that the conception of the battle in Herodotus' pages comes entirely from Herodotus; but it is most unlikely that Herodotus failed to consult the earlier accounts of the battle and to learn from participants in the battle, and these cannot all have been childishly misconceived.