I. ECLIPSES IN ANTIQUITY

Eclipses, both lunar and solar, were phenomena of fear and wonder in the ancient world.Footnote ⁵ The cause of eclipses was first understood in Greece during the sixth and fifth centuries b.c.e., possibly with the speculations of Thales or, failing that, certainly by the time of Anaxagoras.Footnote ⁶ Interestingly, however, the art of eclipse prediction in Antiquity developed sooner than physical understanding, and it took place not in Greece but in Mesopotamia. The Babylonian astronomer-priests, who, outside China, were the most assiduous observers of eclipses during the first millennium b.c.e., never seem to have developed any spatial astronomical theory and relied on manipulating numerical relationships to carry out their work.Footnote ⁷ We do not know how, or even if, they thought of eclipses in physical terms, whether in respect of solar eclipses (in which the Moon lies between the Earth and the Sun) or the conceptually more difficult and less obvious case of lunar eclipses (in which the Earth lies between the Sun and the Moon).Footnote ⁸ In cuneiform texts, now known as Astronomical Diaries, various celestial phenomena including eclipses were recorded. These go back to 652 b.c.e., although the earliest survive only in later copies from the Seleucid era.Footnote ⁹ Ptolemy states that eclipses were recorded from at least the start of the reign of the Assyrian king Nabonassar in 747 b.c.e.Footnote ¹⁰

From these diaries and related astronomical texts, including reports sent to Assyrian kings, almanacs, ephemerides (documents giving positional information on celestial objects), and eclipse tables, the progress of Babylonian astronomy as a predictive science during the Greek Archaic, Classical, and Hellenistic periods can be roughly gauged.Footnote ¹¹ It is my contention that aspects of Babylonian astronomy help us to resolve the puzzle of the Eclipse of Xerxes.

It has been shown that, by about 550 b.c.e., through arithmetical tabulation and extraction of difference sequences, the Babylonians were able to predict lunar eclipses (visible everywhere on Earth) with an accuracy of 90% to within a timing error of around an hour.Footnote ¹² For the much more difficult case of predicting solar eclipses (only visible in a restricted area), the Babylonians had achieved accuracy to within about two hours from around 350 b.c.e., provided that the geographical locus of the eclipse was ignored. In practice, their predictions for the latter appeared to fail about 55% of the time, since these missing eclipses – those which were said to have ‘passed by’ – were only visible in locations far removed from the Middle East.Footnote ¹³ Without some kind of geometrical theory, determination of not only when but where a solar eclipse would be visible could make no further progress.

In Greece, following the explanations first put forward by philosophers such as Anaxagoras, eclipses were clearly understood in broad spatial terms in the later fifth century by an educated elite, but probably not by the bulk of the population. Thucydides, for example, writes of the partial solar eclipse of 3 August 431 b.c.e. as if the phenomenon were quite well understood, at least by him (see below). The later accounts by Cicero and Plutarch of the same eclipse contrast the matter-of-fact explanation given at the time by Pericles (derived, possibly, from Anaxagoras in person) with the fearful ignorance of the general populace.Footnote ¹⁴ During the following century, when Aristotle presented his arguments in favour of the sphericity of the Earth, the geometrical explanation of eclipses was clearly taken for granted and was, presumably, seen as a consequence of the models of the heavens then being put forward.

However, while the Greeks of the Classical era understood eclipses in qualitative spatial terms, they had no tradition of data tabulation and arithmetical analysis to rival that of the Babylonians. Consequently, without adequate empirical support, they had no means of predicting even lunar eclipses, and it was not until the time of Hipparchus in the second century b.c.e. that data and theory combined sufficiently to enable geographically relevant solar eclipse predictions to be made.Footnote ¹⁵

Improvement in the qualitative understanding of eclipses went hand in hand with specialization in the vocabulary. The earliest use of the word ἔκλειψις to mean ‘eclipse’ in its full astronomical sense occurs in Thucydides (1.23.3), where he remarks:

a statement which was not true in fact, but perhaps reflected popular opinion among a citizenry abnormally attuned to portents in time of war. In 2.28 we find the description of the aforementioned solar eclipse of 3 August 431 b.c.e.:

The descriptive accuracy of this passage suggests that the author understood the basic astronomical configuration and had probably witnessed the event. The ‘stars’ referred to were probably the planets Venus, Mercury, and Saturn and some of the brighter stars. This eclipse was partial in Athens in the late afternoon, achieving a magnitude of 88% at 15:58 UT.Footnote ¹⁷

Similarly, in 4.52.1 we have the report of the partial solar eclipse of 21 March 424 b.c.e., again with mention of the necessary lunar phase:

Finally, in 7.50.4 he records the total lunar eclipse of 28 August 413 b.c.e., which took place during the infamous Sicilian expedition (note, again, the reporting of the lunar phase):

These passages show that, by the end of the fifth century b.c.e., eclipses were quite well understood as natural phenomena by at least a few of the educated Greeks. They could also describe them using appropriate vocabulary, even if many still regarded them as ominous (the two interpretations are not necessarily mutually exclusive, of course).

Before Thucydides, though, ἔκλειψις and related words occur only in the general sense of ‘abandonment’ and are not necessarily used in descriptions of phenomena that clearly are eclipses. However, a transitional use may occur in Aristophanes, Clouds 582–6, where the leader of the eponymous chorus says:

This, of course, is not intended as a historical report but it does illustrate, more or less, the ‘correct’ use of the term.Footnote ¹⁸ From these considerations, therefore, it seems that the use of this word in its specialized sense must have developed during the second half of the fifth century b.c.e. and, if not initiated by a philosopher such as Anaxagoras, then perhaps by the astronomer Meton or one of his circle.Footnote ¹⁹ Such terminological specialization fits in well with the emergence of ‘academic disciplines’ during these decades.Footnote ²⁰

II. ANALYSIS OF HISTORICAL ECLIPSES

The accuracy of Babylonian eclipse prediction was not high by modern standards, being essentially a matter of determining the Saros cycle of 223 lunar months and sub-periods of 6 months (33 occurrences) and 5 months (5 occurrences) within it.Footnote ²¹ Knowledge of these period relations, especially the two shorter ones, goes back at least to the seventh century b.c.e.Footnote ²² Modern eclipse theory, of course, is far more accurate, but only in the last two centuries has it become sufficiently comprehensive in terms of date, time, and place for the analysis of historical eclipses to be viable.

Two early attempts to analyse the first Herodotean eclipse (1.74), the Eclipse of Thales (now widely, but not universally, accepted as that of 28 May 585 b.c.e.), were made by Baily in 1811 and Oltmanns in 1812, though neither in fact succeeded in its correct identification. In a paper delivered to the Royal Society in 1852, George Airy, the Astronomer Royal, summarized and criticized past attempts to analyse historical eclipses.Footnote ²³ Airy, who had studied Classics as well as mathematics at Cambridge, turned his attention to two of the three Herodotean eclipses (1.74 and 7.37).Footnote ²⁴ Since then, ephemerides have become more accurate, the orbital dynamics of the Moon better understood, and computational techniques vastly improved. These factors, together with better estimates for the Earth's rotational clock error (ΔT), mean that nowadays it is possible for any researcher to make use of accurate and comprehensive datasets.Footnote ²⁵

Given that eclipses were general objects of wonder, even alarm, in the ancient world, it is no surprise that they are mentioned in early Greek literature. The language is not precise, but there are characteristic references to the disappearance of the Sun or an unexplained darkening of the sky, together with suggestions of coolness or humidity. If the day grows dark, there are only three possible explanations: atmospheric obscuration (such as clouds or dust storms), nightfall, or an eclipse; and it is inconceivable that ancient people were unable to recognize the first two. The language used before the later Classical period is vague but is appropriate to the genre in which it occurs and fairly consistent.

The earliest eclipse description in Greek may be in the Odyssey (20.356–7), reflecting in some way perhaps the solar eclipse of 16 April 1178 b.c.e., which was total over Corfu around noon.Footnote ²⁶ Another early poetic description occurs in a fragment of Archilochus (IE ², vol. 1, fr. 122 + P. Oxy. 22.2313 fr. 1a).Footnote ²⁷ If this is an allusion to a recent solar eclipse it may be that of 6 April 648 b.c.e. or (perhaps more likely, in view of the time of day) that of 27 June 661 b.c.e.Footnote ²⁸ With Archilochus we are still in the realm of the prescientific, of course; moving into the Classical era, however, we have Pindar's description of a solar eclipse in Paean 9.1–5, which suggests, at a time of great intellectual advance, the still primitive sense of bafflement and powerlessness occasioned by such events.Footnote ²⁹

Obviously, poetic references to eclipses need not refer to actual historical events; they are speculative conjunctions of real (and possibly contemporary) astronomical phenomena with imagined circumstances. However, the situation with historiographical texts is wholly different and it is clearly appropriate to enquire whether eclipse reports help us to understand the chronology of events or to identify the motivations of the actors in the narrative.

III. CURRENT INTERPRETATIONS OF THE ECLIPSE OF XERXES

The Eclipse of Xerxes in Herodotus 7.37 takes place during the first phase of the final Persian advance on Greece. As previously mentioned, it involves difficulties of timing and interpretation. In contrast, the Eclipse of Cleombrotus (9.10), which occurs while the Spartans are making defensive preparations, seems unproblematic. In order to fix the chronology, I shall recapitulate the accepted sequence of events.

In the summer of 480 b.c.e. Xerxes advanced southwards through Greece, reaching Athens via Thermopylae in the middle of September (8.51.1).Footnote ³⁰ This timing is secured by references to the Carneia festival and to the Olympic Games (7.206.1–2, 8.26.2) which pin the battle of Thermopylae to late August or early September of that year.Footnote ³¹ Around this time, owing to the threat of invasion of the Peloponnese, the Spartans under the command of Cleombrotus began to construct a defensive wall across the Isthmus of Corinth (8.40.2, 8.71.1–2, 9.7.1).Footnote ³²

However, after Xerxes had captured and burnt Athens, but had lost the subsequent battle of Salamis towards the end of September, he decided to withdraw for the winter to Thessaly (8.113.1), from where a new assault would begin in the following spring, led by his general, Mardonius (8.107.1). At this point, Cleombrotus considered whether to attack the retreating Persians. In the event, this plan was not executed and we learn instead that he returned to Sparta (9.10.3):

There seems no reason not to accept this as a description of a solar eclipse.Footnote ³³ It coincides perfectly with the partial eclipse of 2 October 480 b.c.e. which reached a maximum of 61% at 12:02 UT in the vicinity of the Isthmus. Moreover, it fits well with the chronology of the war as established from other textual references.

This was not a prominent eclipse and, ordinarily, few people would have noticed it. However, the circumstances were special. Cleombrotus was performing a ‘border ritual’ and would have taken care to ensure that there were no unfavourable omens in the sky.Footnote ³⁴ The narrative leaves it open for us to interpret his reaction. Did he consider the eclipse a bad omen or did he use it as an excuse to return home at the end of the military season? Notwithstanding Spartan piety and courage, it is difficult to decide from the information given, but the Spartan interpretation of this eclipse remains a relevant issue.Footnote ³⁵

We might ask whether there were any other eclipses that could serve to mark the ritual of Cleombrotus should it be necessary to reconsider the date. The answer to this is emphatically negative. It is clear from the narrative that the eclipse must have occurred in the autumn, but for several years on either side of 480 b.c.e. there were no autumnal solar eclipses visible in Greece. Therefore, the text effectively nails this eclipse onto the timeline of the war and reinforces the accepted chronology.

It follows that the Eclipse of Xerxes must have occurred in the preceding spring, when Xerxes had set out from Sardis in western Anatolia in order to march into Greece via Thrace.Footnote ³⁶ The invasion had been four years in the planning (7.20). A canal had been dug through the Athos peninsula for the accompanying fleet, provision dumps had been established en route, and, most famously, the Hellespont bridges constructed from Abydos to Sestos in order to permit the crossing of the army from Asia into Europe. At 7.37 we learn that:

The passage is problematic, despite the relatively full description and the interesting use of the word ‘eclipse’ twice (ἐκλιπών/ἔκλειψιν), though not with its later astronomical meaning. The foremost difficulty is that the otherwise ideally timed solar eclipse of 9 April 480 b.c.e. was only visible in the South Pacific. This makes it impossible to accept the text at face value. Xerxes cannot have seen a solar eclipse at that time.Footnote ³⁷

Airy was perhaps the first to establish definitively the non-occurrence of the alleged solar eclipse. He noted that the text asserts that the Magi had said the Sun was viewed as a ‘prognosticator’ for the Greeks and the Moon for the Persians, but that this was contrary to what was known about Persian religion (worship of Ahuramazda, ‘Lord of Wisdom’) and its veneration for the Sun.Footnote ³⁸ Citing the example of the aforementioned lunar eclipse in Thucydides (7.50.4), where the Greeks seem terrified, he argued that the Eclipse of Xerxes was not solar but lunar. Herodotus, Airy implies, had misunderstood the eclipse report and the justification for it had become scrambled. According to the Magi, he suggested, it was really the Moon that was the predictor for the Greeks and the Sun for the Persians.

Unfortunately for Airy, there appears to have been no convenient lunar eclipse either in the spring of 480 b.c.e. However, one in the following year fitted perfectly: the total eclipse of 14 March 479 b.c.e. which peaked at 02:17 UT with an umbral magnitude of 150% and whose totality lasted for 95 minutes. Of course, to accept this meant violating the accepted chronology of the Persian Wars. Airy acknowledged the difficulty but left it unanswered.

However, Airy omitted the Eclipse of Cleombrotus, which we have already examined and which is securely anchored to 2 October 480 b.c.e. If we accept Airy's lunar eclipse for the Eclipse of Xerxes we need to bring forward the Eclipse of Cleombrotus by a year also. Yet, as has been stated, there is no other solar eclipse with which it can be identified. It seems, therefore, that we cannot extend the Persian War by a year for astronomical reasons, even if we could square the historical chronology, which itself seems unlikely. All major commentaries agree on this point. An alternative solution would be to take the annular solar eclipse of 17 February 478 b.c.e. as the Eclipse of Xerxes.Footnote ³⁹ This reached a magnitude of 95% at 10:17 UT at Sardis and must on that account have been quite striking. But by extending the war for two years it violates the chronology even more strongly than does Airy's lunar eclipse and must, therefore, be dismissed.Footnote ⁴⁰

There is, however, a conventional way out of the problem. In the previous year, Xerxes was at Susa, the western imperial capital, and had begun the preparation for the campaign against Greece. Now, in the fifth year of his plan (7.20.1) in the spring of 481 b.c.e., he set out with his army for Critalla in Cappadocia (7.26.1), the designated mustering point for the many national forces which comprised the army. The location of Critalla has not been determined, but it was probably near the first crossing of the river Halys (modern Kızılırmak) in eastern Anatolia, perhaps at Caesarea Mazaca (modern Kayseri).Footnote ⁴¹ On the Anatolian plateau, conditions are harsh in winter, so the march must have been planned to have been complete by the autumn in order that the army could overwinter in Sardis, where it would be joined by contingents from western Anatolia and thus be ready for the second leg of the operation in the following spring.

The decision to leave Susa represented an irrevocable step for Xerxes, not only the culmination of four years of planning but a decisive action taken after much vacillation involving the interpretation of dreams (7.8–19). He could not now decide to abandon the invasion, but he still had some leeway in respect of timing and would have wanted to ensure favourable omens for the expedition by employing the Magi to scan for various celestial phenomena.

On 19 April 481 b.c.e. there was a solar eclipse. It was only total over south and east Asia, but it was partial at Susa, though not at Sardis.Footnote ⁴² The maximum occurred at 03:23 UT, just after dawn. With a magnitude of just 19%, the eclipse would have passed unnoticed by most, but to those looking for portents the Sun would have looked distinctly odd, with a flattened side.

The suggestion is, therefore, that original accounts of an eclipse observed by professional skywatchers at Susa in the spring of 481 b.c.e. became transferred in the narrative to the events of spring 480 b.c.e. at Sardis. Whether this was conscious design by Herodotus or confusion in his sources it is difficult to say. Despite being a seasoned traveller to many parts of the Near East, including probably Babylon (1.178–99) and possibly Susa (6.119.2–3), he may have felt more comfortable with locating the story in Sardis, a place well known to him and more relevant to the westerly focus of his narrative.Footnote ⁴³

IV. AIRY REVISITED

The conventional explanation has merit, but it does not account for the confusion over the religious significance of the Sun and Moon. Airy's judgement that Herodotus' description ‘is, on the face of it, absurd’ has generally been upheld by later commentators.Footnote ⁴⁴ Why, moreover, would Xerxes need to be reminded of the fundamentals of his religion? He would surely know how important the Moon was as a portent, if that were really the case?Footnote ⁴⁵

Herodotus is clearly trying to justify Xerxes' reaction to a solar eclipse which otherwise would seem perverse. Only a little later, he describes the Persian worship of the Sun (7.54.1–3), when Xerxes begins the crossing of the Hellespont, as he does also (7.223.1) before the battle of Thermopylae. Yet at 1.131.2 he also mentions that the Persians sacrifice to the Moon, water, and winds as well as to the Sun. In view of this, we might perhaps conclude that Herodotus' knowledge of Persian religious practice was sketchy and liable to adaptation as he saw fit.Footnote ⁴⁶ Equally, however, it might be argued that we are given only glimpses of a religion (whose main features are taken for granted by the author) which viewed all natural phenomena as divine manifestations.

If we can accept that Herodotus may have believed, rightly or wrongly, in the greater importance of the Moon to the Persians than the Sun or, perhaps, have believed that his report would seem credible to his audience, then the way is clear for us to accept that there may be some merit in Airy's hypothesis of a lunar eclipse. The idea that the Sun was the more important celestial indicator for the Greeks was probably acceptable to his Greek audience and few of them, presumably, were knowledgeable enough about Persian religion to gainsay the statement regarding the Moon. So, from a Greek perspective the statement seems fine; but from a Persian religious perspective, it is nonsense.Footnote ⁴⁷

However, since the Persians considered the Sun more important than the Moon, it might have seemed reasonable to them to think that the Greeks accorded greater importance to the Moon. This is the reversal of the Herodotean statement, but it functions in the same way as a cultural demarcator between the two peoples and as a symbol of otherness. Accordingly, a lunar eclipse might well have been regarded as favourable by Xerxes, since he would have assumed it was unfavourable for the Greeks. On this basis, Airy's instinct was right, but what can be salvaged from the astronomical facts?

It turns out that Airy ignored an earlier lunar eclipse than the one he proposed, namely that of 25 March 480 b.c.e. He must have been aware of it, but probably discounted it because his calculations showed that it was penumbral and, therefore, barely perceptible. In a penumbral eclipse, the Moon's disc still receives some direct light from the Sun, but the reflected light is typically reduced by 80–90%. However, even this reduction causes virtually no apparent diminution in brightness to the human eye.Footnote ⁴⁸ Keen-sighted and experienced observers though they may have been, the Magi could not have observed this eclipse had it been purely penumbral. Airy was right to pass it by, given the data and results that he had.

However, modern calculation shows that this eclipse was not penumbral. Even a small correction to the Greenwich observational data of 1750 to 1830 upon which Airy had relied would have sufficed to show the true state of affairs.Footnote ⁴⁹ The lunar eclipse of 25 March 480 b.c.e. was actually partial. In a partial lunar eclipse, a part of the Moon enters into the Earth's umbra where it receives no direct light from the Sun. Owing to the great differences in light levels between the penumbral and umbral parts of the Earth's shadow (about 500 times) and the fact that the latter forms a convex curve upon the Moon's surface (Aristotle's observation: Cael. 297b 24–31), partial lunar eclipses are unmistakeable.Footnote ⁵⁰

In the case of this lunar eclipse, the extent of umbral shading was comparatively small (magnitude 7.4%), with the maximum occurring around 23:13 UT. Yet even this small amount would have caused a pronounced distortion to the circular form of the full Moon, with the Earth's umbra biting fuzzily into the Moon's disc.Footnote ⁵¹ The event lasted for 66 minutes and all observers could have seen that there was something amiss. A preliminary suggested sequence of events is thus as follows:

1. 1) a partial lunar eclipse occurs on 25 March 480 b.c.e.;

2. 2) the Magi at Sardis deem it to be a bad omen for the Greeks, explaining (whether in good faith or bad) to an ignorant Xerxes that the Greeks set great store by the Moon;

3. 3) the story of an eclipse witnessed by Xerxes at Sardis enters the popular narrative of the war, together with the remarks of the Magi;

4. 4) Herodotus picks up a garbled account some decades later and, in attempting to make sense of it, writes it up as a solar eclipse, the only type he records;

5. 5) in order to make it credible to his Greek audience that a solar eclipse could be interpreted as a good omen for Xerxes, Herodotus assumes that the actual advice of the Magi (‘the Moon is a predictor for the Greeks, but the Sun for us’) has been misinterpreted and flips it around.

Yet there is a further twist to the story which needs examination. This is the tale of Pythius, the Lydian, who first appears as a potential financial contributor to the Persian campaign (7.27–9) and then reappears in the narrative, just as Xerxes is setting out, in order to ask if his eldest son can be released from military service (7.38). Hitherto, he has not only wanted to support Xerxes, but has been, we may assume, confident of Persian success. Now, however, he has been frightened by the eclipse and is less sanguine about the success of the invasion.

We first meet Pythius at Celaenae (modern Dinar) in Phrygia, whither he has, presumably, travelled in order to meet Xerxes and, as a member of the defunct Lydian royal family, to welcome him to Lydia with money for his campaign. He then, one assumes, travels back to Sardis with Xerxes as part of his wider retinue. It is difficult to see how he could have become attached to the story of the Eclipse of Xerxes had it originated in Susa, since he had no official role in the Persian hierarchy.

Since Pythius is a Lydian and the Lydians largely share the customs and outlook of the Greeks (1.94.1), we must think this through from the point of view of a Hellenized Lydian who has perforce thrown in his lot with the Persians.Footnote ⁵² By the time of Xerxes' invasion, the Lydians had been under Persian rule for over sixty years, but Pythius was of an age to have been brought up during the last years of Lydian independence and was on that account perhaps more likely to retain the intellectual outlook of his youth.Footnote ⁵³ His desire above all else is to ensure that not all of his sons are killed or enslaved, in view of which his commitment to the Persian cause would seem to be paramount. Why, then, does he not share Xerxes' view of the eclipse? It unlikely that he would take a solar eclipse as a personal prognosticator, since astrology, being mainly an invention of Hellenistic Egypt (though founded upon Babylonian techniques), was yet to get under way.Footnote ⁵⁴ If there had been a solar eclipse he would have assumed that this was bad news for the Greeks and, therefore, good news for him and the Persians. Nor can we have him disbelieving the authoritative pronouncement of the Magi regarding Persian beliefs. There seems to be no reason, on the assumption of a solar eclipse, why Pythius should think that it was a bad portent.

If, on the contrary, there had been a lunar eclipse, he would presumably not have been so easily persuaded by the Magi telling Xerxes that this was a bad omen for the Greeks, since he would be aware that the Greeks esteemed the Sun as much as the Persians.Footnote ⁵⁵ In other words, the story of Pythius' reaction seems better motivated by assuming a lunar eclipse than a solar eclipse. It also circumvents the difficulty of relocating the eclipse from Susa to Sardis, since it can hardly be supposed that the aged Pythius was at the Persian capital, although this in itself is perhaps not a serious objection.

Of course, as it turns out, Pythius is not lucky. Xerxes becomes enraged when he hears his request, pointing out the total commitment of his own family to the invasion, and orders that the eldest son be cut in half and the two parts of the corpse arranged so that the Persian army marches between them (7.39.3). Although this is presented as the barbaric act of a cruel and irrational tyrant, which undeniably it was, it may perhaps be viewed also as a punishment for challenging the official interpretation of the eclipse and as an attempt by Xerxes to garner further good fortune by offering a sacrifice along traditional lines.Footnote ⁵⁶

In fact, as Myres pointed out, the story of Pythius is essentially one which has been arranged around the fulcrum of the sojourn at Sardis (7.30–7).Footnote ⁵⁷ He offers Xerxes money for the campaign (7.27–9) and the quid pro quo is to be the release of his son from military service (7.38.43). In order to propel his narrative, Herodotus has bisected the story of Pythius as neatly as Xerxes bisected the eldest son so as to propel his passage into Greece. It is clear, therefore, that Herodotus has manipulated his sources to provide, as he sees it, a coherent narrative. For the author, the moral of the story is of great importance and the Eclipse of Xerxes is merely the first of three omens of increasing improbability which the great king chooses to ignore or misinterpret.Footnote ⁵⁸ We may suppose that this crescendo of arrogance and bad judgement would have appealed to Herodotus' listeners and readers. It seems that he knew that there was an eclipse which was ignored but, in default of sufficient source detail, devised a story that made sense.

V. THE PERSIAN PERSPECTIVE

So far, I have offered an alternative explanation, based on a lunar eclipse, for the problems surrounding the Eclipse of Xerxes. Its advantages are first, that it uses an eclipse which actually occurred at the right time (25 March 480 b.c.e.), as suggested by historical events; second, that it makes better sense of the remarks of the Magi; and third, that it better explains the story of Pythius. However, it is not wholly satisfactory. Against it is the text itself, which plainly states that

This clearly describes a solar eclipse. On the lunar hypothesis, therefore, we must accuse Herodotus of factual error. For this reason, it might seem preferable that, for the Eclipse of Xerxes, one should retain the idea of a solar eclipse transferred from either the one at Susa in April 481 b.c.e. or the one at Sardis in February 478 b.c.e. or a confusion of both. However, not only is the text unambiguously a description of a solar eclipse, but it also strongly suggests a total solar eclipse. Neither of the transferred solar eclipses were total: the preferred eclipse at Susa was barely perceptible and the unlikely one at Sardis, though of dramatic appearance at 95% magnitude, could not, with the Sun being quite high (38°) above the horizon, have literally turned day into night, if by that it is meant that stars became visible.Footnote ⁵⁹ Therefore, whatever the underlying astronomical event, Herodotus cannot escape the charge of distortion.

Since the facts, as they stand in the narrative, cannot be true, we must accept that either the story of a solar eclipse (or eclipses) has been detached from historical context, exaggerated, and reattached to the departure from Sardis in 480 b.c.e. or the nature of the eclipse itself has been misrepresented. A lunar eclipse has been considered, but is not compelling per se. However, another possibility exists.

Let us reconsider the incongruous words of the Magi (leaving aside the question of why they were inserted into the text and by whom). Was the Moon a predictor for the Persians and the Sun a predictor for the Greeks, as reported, or was it the other way round? This apparent binary choice may be fallacious: perhaps neither the Greeks nor the Persians thought of eclipses in that clear-cut way. All eclipses may have had ominous significance depending upon circumstances. After all, we only have Herodotus' word for the interpretation given and it has been established in this case that he is not necessarily reliable. Moreover, given the source, the viewpoint hitherto taken has been unavoidably Hellenic. However, it takes two to make a war, and it behoves us to consider what we can make of a possible Persian perspective.

Arguably, the text as we have it misunderstands the Persian attitude in a rather more fundamental way than previously considered, a suspicion bolstered by an almost contemporary passage in Aristophanes, Peace, 406–13, where Trygaeus remarks that, whereas Greeks worship gods, foreigners worship the Sun and Moon. This is a pronouncement put into the mouth of a simple farmer and perhaps not to be taken as authoritative but indicative of a mismatch between Greek and foreign belief systems, in so far as those in the Near East generally placed much greater emphasis on celestial phenomena. More specifically, we may cite Herodotus' own observation (1.131.1) that the Persians deny anthropomorphism in gods and do not set up statues to them as Greeks do, but (1.131.2):

Further to this, we may recall, as did Airy, the Greek fear of the lunar eclipse described in Thucydides 7.50.4 and then again the attitude of the Spartans to the solar eclipse in 9.10.3. All in all, we should conclude, perhaps, that neither Greek nor Persian attitudes to eclipses were resolvable along a simple Sun/Moon line of demarcation and that the reality was more complex. If we step back from the text and focus on what we really do know about the origins of ancient astronomy, we may be able to find a better approach.

Notwithstanding the famous eclipse ‘prediction’ of Thales, the Greeks in the fifth century had no innate tradition of predictive skywatching other than to forecast the seasons via the annual cycle of the stars.Footnote ⁶⁰ Nor, however, did the Persians, nor the Medes, nor, probably, the Magi (originally, for Herodotus, a Median tribe).Footnote ⁶¹ Whatever anyone knew about astronomy was derived ultimately from the Babylonian astronomer-priests, especially those working in the Neo-Assyrian (747–612 b.c.e.) and Neo-Babylonian (612–539 b.c.e.) periods, when the Astronomical Diaries were first compiled. Their skills and knowledge were disseminated widely into the surrounding cultures, who were keen to augment their own divinatory practices.Footnote ⁶² The domination of Babylon by the Assyrians in the early first millennium and then its subsequent capture by the Persians under Cyrus the Great in 539 b.c.e. witnessed a flowering of astronomical practice which continued into Seleucid times.Footnote ⁶³ By the start of the Persian era, however, Babylonian astronomy was already very ancient, having developed over the course of the previous 1,500 years its signature characteristics of planned observation, systematic recording, and omen interpretation. Its chief concern was the safety of the king and state.

The collection of 70 tablets known as the Enuma Anu Enlil (EAE) series, which dates from the eighth century, contains over 6,000 omen formulae used in astronomical reports sent to the Assyrian kings. Some of the descriptions refer back more than a thousand years.Footnote ⁶⁴ Celestial observations were part of a wider set of natural phenomena that were incorporated into divinatory schemata which consisted of explanatory rules of the form ‘If P then Q’.Footnote ⁶⁵ Typical examples read as follows:

Huge numbers of celestial phenomena were incorporated into such schemata and those that featured the Moon were arguably the most important, followed by the Sun.Footnote ⁶⁷ The general importance of celestial phenomena for the Babylonian astronomers in the Neo-Assyrian period derives from the fact that:

This precept underlies the divinatory approach to astronomy in the ancient Middle East, to the extent that areas and directions in the sky were often seen as corresponding to parts of the Earth, and events that occurred in the former presaged in some way events on the latter.Footnote ⁶⁹ There was no necessary causal connection; rather, the sky was viewed as a kind of text, written by the gods, indicating what might happen.Footnote ⁷⁰

The main aim of Babylonian astronomers was not to predict celestial events in the pursuit of objective science but to extract ominous meaning and not to be caught unawares by those events which did occur.Footnote ⁷¹ Almost all unusual events, if they were not predicted, were bad omens, especially eclipses, whereas predicted events, whether or not they occurred, were potentially benign. An eclipse that occurred ‘on schedule’ could have its deleterious power reduced or nullified by the performance of an apotropaic rite.Footnote ⁷² Curiously, the Babylonian/Persian king-substitution rite finds its way into Herodotus (7.15–18), although not in a celestial context.Footnote ⁷³ If, on the other hand, an eclipse failed to occur, then, apart from perhaps the wounded pride of the astronomers, no real harm could ensue. In fact, if the rite were performed before the failed event was due to occur (a strong impetus to prediction in itself) then the rite could be viewed as having been successful.Footnote ⁷⁴ Of course, the Babylonian astronomers did not want to be seen to be foolish or tempting fate and so they strove to improve the accuracy of their observations and predictions for all sorts of astronomical phenomena.Footnote ⁷⁵ However, there was a strong intrinsic bias towards making predictive claims on the shared understanding that they were declaring when events might occur, not when they would occur.Footnote ⁷⁶

It seems to be the case that solar eclipses were not included as predicted events in the extant texts until the fifth century b.c.e.Footnote ⁷⁷ This may be a sampling effect, owing to the relative paucity of recovered and published texts, but it may genuinely reflect the fact that Babylonian astronomers were more concerned, for calendrical purposes, with understanding the motion of the Moon (for which they reserved their most advanced techniques of tabulation and calculation) and the relatively tractable nature of lunar eclipses, owing to the five- and six-month cycles and the larger Saros cycle.Footnote ⁷⁸ Solar eclipses were simply too difficult to predict in the early fifth century b.c.e. and generally occurred too infrequently, although they were always regarded as significant and featured in the omen literature.Footnote ⁷⁹

Given the pervasive nature of Babylonian astronomical practice throughout the Middle East during the first millennium b.c.e., particularly as absorbed and promulgated by the Assyrians from the eighth century onwards, it seems likely that the Magi, whatever their own local traditions of skywatching and divination from the pre-Zoroastrian period (which we can scarcely guess at), would have assimilated this aspect of Babylonian culture when the opportunity occurred.Footnote ⁸⁰ Support for this comes from the fact that, although the names differ, the Old Persian months (like the Elamite and unlike the Avestan) correspond largely to the Babylonian scheme in respect of intercalations.Footnote ⁸¹ Regrettably, we know little about the Magi at this date and our primary source of information is again Herodotus. That they were regarded in later times as expert astrologers is underscored by their appearance in the New Testament (Matthew 2:1) but, as already suggested, this cannot have been their function in the early fifth century b.c.e. since astrology in the personal sense had yet to be invented.

However, the path from divination to astrology, from the public realm to the private, is not a difficult one provided that the necessary data and techniques have become widely available, as they had done during the Seleucid period. Since the Magi are credited by Herodotus with the art of divination in several places (1.107.1, 1.120.1, 1.128.2, 7.19.1), it seems likely that they fulfilled a similar role in fifth-century Persian society to the astronomer-priests in Assyria/Babylon during the preceding centuries. Of course, we do not have any documents to prove this. There is no royal library of Darius, for example, to match that of Assurbanipal in which many source texts of Babylonian astronomy were found.Footnote ⁸² Let us suppose, however, that such documents once existed and that the Magi learned astronomical techniques from their Babylonian peers. In this context, we can see another way of looking at the Eclipse of Xerxes.

The chief message that Xerxes took from his consultation with the Magi is that the eclipse was a good omen. This much seems guaranteed, almost literally, by the march of events. We may surmise that it was viewed as benign, not because it was a bad predictor for the Greeks (whether lunar or solar), but simply because it had been successfully predicted by the astronomers and that, consequently, there had been time to put in place an appropriate ritual. We do not know, of course, whether this was actually the case for the Eclipse of Xerxes, but we do know that, for lunar eclipses (but not for solar eclipses), the technology to predict them with sufficient reliability was already well established by the early fifth century b.c.e.

As far as I know, the lunar eclipse of 25 March 480 b.c.e. does not occur in the extant Babylonian records, but both earlier and later eclipses are tabulated. They show that the 90% success rate in predicting lunar eclipses can be broken down into 55% umbral (total and partial) and 35% penumbral, some 10% being outright failures.Footnote ⁸³ Since penumbral eclipses cannot usually be observed by the naked eye, the astronomers working in the Babylonian tradition must have considered (erroneously) their own failure rate to be relatively high, but within the context of celestial divination the apparent fact of non-occurrence was, as has been argued, not a poor result.

If we shift our focus to consider eclipses to be a matter of predictive success or failure for the Persians then we can understand why the Magi would have been pleased, perhaps even jubilant, to have caught, by a whisker, the partial lunar eclipse of 25 March 480 b.c.e., which they may have predicted using traditional Babylonian techniques – a feeling, we may surmise, that was communicated to Xerxes in the form of an omen. Of course, we do not know how the eclipse may have been interpreted in detail nor what apotropaic ritual may have been put in place. It is not necessarily the case that the Magi would have followed original Babylonian omina or rituals. They may well have devised their own, or at least modified Babylonian originals, since many of the latter referred to political situations which were no longer extant.

By contrast, most Greeks of the fifth century b.c.e., notwithstanding the Ionian philosophical tradition, probably had little interest in, or knowledge of, the predictive schemes emanating from the Middle East. The brief success of Thales a century earlier was not and (if the analysis by Couprie is correctFootnote ⁸⁴) could not be repeated. They knew that the barbarians studied the sky, but not to what end. What mattered to the Greeks was the fact of the eclipse itself, not whether it had been successfully foretold.Footnote ⁸⁵ When it occurred, therefore, it was likely to be interpreted as a bad omen, just as those observed by Cleombrotus and Nicias had been. On this basis, we can ascribe the alarm of Pythius at seeing the eclipse to his general Hellenic outlook and the fact that he was not part of Xerxes' inner circle, rather than trying to work out whether he would have been more or less concerned depending upon whether it was lunar or solar.

In addition, by attributing to the Magi general Babylonian techniques, we can rescue their reported remarks (‘the Sun is a predictor for the Greeks, but the Moon for the Persians’) since, although the Sun was arguably of greater significance within Persian religion, the Moon was far more important for the Babylonian astronomical tradition within which, I contend, they themselves were working.Footnote ⁸⁶ Moreover, we need not suppose that Xerxes himself, while necessarily familiar with the tenets of Zoroastrianism, would have known much about Babylonian astronomy. In this context, his initial misgivings are well motivated. On this interpretation, therefore, Herodotus has somehow misreported the facts of the eclipse, but he has correctly, albeit imperfectly and too briefly, characterized the attitude of the Magi.