Thursday, 28 June 2018

The not-so-cryptic oracle of Delphi

The most famous story about an ancient oracle comes from Herodotus’ tale of Croesus, king of Lydia. First, he tested various oracles to see which ones were reliable. He narrowed his choices down to two, then asked both of them whether he should invade Persia. The answer was:

They foretold to Croesus that if he campaigned against the Persians, he would destroy a great empire.
Herodotus 1.53

Croesus decides that yes, he will go to war against the Persians. And he does indeed end up destroying a great empire ... his own. The Persian king wins, he takes Croesus prisoner, and Lydia (in what is now western Turkey) becomes part of the Persian empire.

Michelangelo, the Delphic Sibyl (Sistine Chapel, 1509)

Herodotus’ Delphic Oracle is enigmatic and ambiguous. He depicts the Oracle with its mythological hat on: when mythical figures like Oedipus or Xuthus consult the Oracle, they come away with responses that make no sense at the time and can only be interpreted in hindsight.

That sits very nicely with stories that the Pythia, the priestess who spoke for the god, was intoxicated or drugged, so as to create a kind of artificially induced enthousiasma or ‘divine inspiration’. These stories come from Plutarch, six hundred years later than Herodotus, but they’re such a neat match that the modern popular image treats them as one and the same.

Herodotus’ Oracle often speaks in hexameter verse. That’s the same rhythm as epic poetry, but it’s very different in terms of poetic style:

Pallas cannot appease Olympian Zeus,
even if she beseeches him with many words and dense cunning.
But I shall tell you another word, and bind it with adamant:
for though everything will be ravaged that Cecrops’ border
holds inside, and so will the glens of holy Cithaeron,
wide-browed Zeus grants a wooden wall to Tritogeneia
which alone will be unconquerable, an aid to you and your children.
Herodotus 7.141

According to Herodotus, this reponse supposedly means: 1. the gods are grumpy at Athens (represented by the goddess Athena: ‘Pallas’ and ‘Tritogeneia’ are her titles); 2. Athens is going to be sacked by an enemy; 3. a ‘wooden wall’ will save the day, and that’s the Athenian navy.

And, wonder of wonders, this turns out to be an accurate prediction of the naval battle at Salamis in 480 BCE, where the Athenian fleet crushed the Persians.

Some important things to observe here.

  1. The poem Herodotus quotes wasn’t a real prediction. We should take it as given that virtually all stories about foretelling the future are written in hindsight. The prophecies we hear about were either composed after the events they supposedly predicted, or were vague enough that they could be reinterpreted in light of actual events. This is not a prediction, and it isn’t an authentic oracular pronouncement.
  2. Real-life oracular responses were not normally predictions, but instructions or, alternatively, statements about the will of the gods or what is morally right.
  3. Real-life responses were not typically obscure.
  4. Real-life responses were never in verse, at least until the Roman era.
  5. Real-life responses were not ambiguous, but straightforward and transparent: yes or no, this or that. They did not have ‘if’ or ‘when’ conditions. The response that Croesus supposedly got — ‘if you make war on the Persians’ — is exactly opposite to the kinds of instructions that we hear about in records kept at the time the responses were actually given.

Our best evidence for actual oracular responses comes from official inscriptions about occasions when a government consulted the Oracle. We don’t have anything approaching a complete record — even in antiquity these inscriptions were only made occasionally, and they cover governmental decisions rather than historical events. Herodotus couldn’t base his history on that kind of evidence. Still, we have a fair number of them. And they are all, almost without exception, drastically different from the picture that Herodotus paints.

How to test an oracle (source: Oglaf, ‘Double blind’; NB: site is not safe for work)

Here’s the most detailed account we have of a real procedure for consulting the Oracle. It comes from an Athenian inscription dating to 352-351 BCE.

[It is decided] that the Secretary of the Council should write on two sheets of tin, equal and alike: on one,

whether it is more beneficial and better for the People of Athens that the [Archon] Basileus lease out those parts of the Sacred Meadow that are currently under cultivation, the parts outside the boundaries, to pay for the building of the portico and the repair of the sanctuary of the Two Goddesses;

on the other sheet of tin,

or it is more beneficial and better for the People of Athens that the parts of the Sacred Meadow now under cultivation, the parts outside the boundaries, be left untilled for the Two Goddesses.

[There follows an elaborate set of procedures to put the tin sheets randomly into two jugs, one gold and one silver, without anyone being able to find out which sheet is in which jug.]

... the People shall choose three men, one from the Council and two from all Athenians, to go to Delphi and ask the god which of the inscriptions the Athenians should act on concerning the Sacred Meadow: the one from the gold jug, or the one from the silver.

Inscriptiones graecae ii3 1.292 = ii2 204, lines 23-47
(trans. Bowden 2005: 88-89, adjusted; Greek)

If you read the full inscription, you’ll see just how carefully the procedure is designed to make sure that there’s no human influence over the Oracle’s decision.

Inscriptions are the most reliable evidence, but we do get some seemingly trustworthy reports from literary sources too. It’s just that they have to be taken with a grain of salt. For example, Thucydides reports how Sparta established a colony in 426 BCE at Heracleia, a few kilometres to the west of Thermopylae. Step one was:

So first they inquired of the god at Delphi, and he gave the command ...
Thucydides 3.92.5

This is recent history, not a record of something that happened centuries earlier, so there’s a decent likelihood that this really happened as described. Thucydides doesn’t give us any details, but it’s a simple question with a simple answer, just like in the inscription quoted above.

Conversely, on another occasion when the Spartans consulted Delphi, the Oracle sounds more willing to make speeches:

... and they sent to Delphi and consulted the god as to whether it would be better if they made war. And he ordained, as it is said, that if they fought with all their might they would have victory, and he said that he himself would assist, whether asked or unasked.
Thucydides 1.118.3

That’s a lot more than a yes-or-no answer. It’s much more like the responses Herodotus and Pausanias tell us about, even if it is in prose. (An encyclopaedia from nearly 1400 years later claims to give the exact wording of the Oracle’s response: Suda α.899 ἄκλητον. It’s still prose.)

Now, the Spartans may actually have consulted the Oracle. The Oracle may have actually said yes. But unlike the situation with founding Heracleia, there are some solid reasons to be sceptical of the details:

  1. It’s cast as a foretelling, not an instruction. This is atypical for genuine oracular responses, but typical for Herodotean-literary-mythological ones.
  2. It’s effectively got an ‘if’ clause — you will win if you fight as hard as you can. (There’s no ‘if’ in the Greek: the use of a participle makes the prophecy’s conditional nature implicit.) Again, atypical for real responses, very typical for mythological ones.
  3. The timing is vague. Thucydides doesn’t put it into a linear chronology. Here’s how he frames it: ‘Not many years afterwards, there took place the events I described earlier at Corcyra and Potidaia ... All these events took place in a period of roughly fifty years ... In this period the Athenians consolidated their authority ... At this point, the Spartans regarded the situation as no longer tolerable’ and so they consulted the Oracle (Thuc. 1.118.1-2). Now, the Potidaian affair that he refers to was in 432 BCE, and the war began in 431, so you could argue that the consultation was in between. But this chapter is emphatically not a timeline of events between 432 and 431 BCE. It sounds much more like an anecdote that has been slotted into an appropriate free space.
  4. Thucydides distances himself from the story, inserting the phrase ‘as it is said’. This could be because he wasn’t totally satisfied about the story himself. It’s also possible that he’s reporting a secondary version of the oracle, designed specifically for public dissemination, and not necessarily what the priestess said on the day of the consultation.

This last point gets us to a core part of the problem. Even if we set aside mythological oracular responses, there’s a sharp distinction between the responses you get if you go to Delphi and go through the process of consulting the Pythia, and the responses that get published in literary texts. And, sometimes at least, the two had nothing to do with each other.

The biggest source of non-Oracular oracles in classical Greece was a group of poets known as ‘oracle collectors’ or chrēsmologoi. Most of them aren’t well-known names: it’s just possible you might have heard of Musaeus and Epimenides, but even trained classical scholars usually won’t know their way around Bacis, Onomacritus, Amphilytus, Abaris, Lysistratus, Lycus, and Euclus. (A few more appear in a list drawn up by Clement of Alexandria, Stromateis 1.21.)

Their poetry survives only in quotations found in other authors. The best represented is Bacis. Herodotus quotes three pieces of Bacis’ poetry, Pausanias quotes four, and there are another two or three of doubtful authorship. Here’s a sample, which Pausanias specifically assigns to Bacis:

But one day, when a Tithorean man pours libations
and offerings of prayer on the earth to Amphion and Zethus;
when Taurus is warmed by the might of the glorious sun;
at that time, beware of disaster for the city, no small one.
For the fruit of the harvest wastes away in it
when people divide the earth and bring it to Phocus’ grave.
Pausanias 9.17.5

It’s pretty obvious which kind of Delphic response this corresponds to, isn’t it?

  • Obscure: check. (What’s all this about ‘the harvest’?)
  • Ambiguous: check. (Which city? Where is Phocus’ grave?)
  • In verse: check.
  • Starts with a condition: check. (An indefinite ‘when’ clause with ὁπόταν + subjunctive.)

There’s a pretty good argument to be made that the Herodotean stereotype of the enigmatic Oracle is actually based on the poetry of oracle collectors and other similar concoctions.

Oracle collectors had a mixed reputation. Bacis gets brutally satirised in some of Aristophanes’ plays and by the Roman-era essayist Lucian; but Herodotus was a fan. In the Peloponnesian War, when the Athenians heard of the failure of the Sicilian expedition, they blamed the oracle collectors for misleading them (the Sicilian expedition: Thuc. 8.1.1); but when they lost the battle of Aegospotami, they took that as confirming prophecies made by a Sibyl and Musaeus (Paus. 10.9.11).

Oracle collectors aren’t the only possible source for oracles that are more than just yes-no answers. We don’t know what relationship, if any, existed between oracle collectors and the institutional oracles like that of Delphi. But we have so many responses attributed to the Oracles of Delphi, Dodona, Didyma, and other places, which are an exact match for the style of what we see in the oracle collectors, that it’s hard to avoid thinking there was at least some cross-influence between the Oracles and the oracle collectors.

Remains of the 3rd century BCE temple of Apollo at Didyma

In particular, there is some (sparse) evidence of individuals attached to the institutional Oracles who may have acted as a kind of publishing wing. Two inscriptions from Didyma refer to a building there called the Oracle Writing Office (chrēsmographion: McCabe, Didyma 107, 108 = Did. Inschr. 31, 32), present in the temple complex from around 300 BCE onwards. This sounds awfully like some officials were taking spoken oracular responses and turning them into something more literary — something closer to what we see in the oracle collectors.

Several inscriptions refer to an official called the hypochrēstēs. This person’s role isn’t made explicit, but the prefix hypo- regularly means someone who does work interpreting something: for example, hypokritēs meant ‘interpreter’ before it meant ‘hypocrite’, and the verb hypokrinomai means ‘to interpret a dream’ in Homer. So hypochrēstēs ought to mean Oracle Interpreter.

We don’t have evidence of an Oracle Writing Office at other oracular sites — as far as I can find out — but we do have someone called a hypophētēs at the Oracle of Dodona (Homer, Iliad 16.235-236). If a hypochrēstēs is an oracle (chrēst-) interpreter (hypo-), a hypophētēs ought to be an interpreter (hypo-) of the god’s speech (phēt-). And that’s exactly how an ancient commentary explains it (scholion A on Iliad 16.235).

Now, there’s room for debate over the meanings of these words. The foremost 20th century scholar on the institutional Oracles of ancient Greece, Joseph Fontenrose, thinks the Oracle Interpreter at Didyma was an attendant who was present at consultations. Personally I find it more tempting to link the Oracle Interpreter to the Oracle Writing Office.

Note. See Fontenrose 1988: 78-85 on the process of consulting the Oracle of Didyma and possible roles of the Oracle Interpreter; 1988: 43 on the Oracle Writing Office.

Be that as it may, there appears to be a distinction between a priestess giving an oral response, and an Oracle Writing Office disseminating a written version. It’s not unreasonable to imagine a comparable division of labour at other institutional Oracles, like Delphi and Lebadeia and Dodona.

The point is that if we’ve got an oracular response that is cryptic or ambiguous; if the response is in verse; if it uses conditionals, metaphors, or animal imagery; in other words, if it isn’t a straightforward yes-or-no answer — then it’s not likely to be an authentic response. But it could come from an oracle collector like Bacis or Euclus. Or it could be produced by a publicity office, like the Oracle Writing Office of Didyma.

Remember the ‘wooden wall’ prophecy I quoted near the start? Herodotus says that came from the Delphic Oracle. It appears in a few other places too. In one, the Historiae written by the 12th century Byzantine scholar Ioannes Tzetzes, it’s attributed to the oracle collector Bacis (Hist. 9.796-805). Now, Tzetzes is very very late, and he’s probably not right ... but it’s still a good example of how the different potential sources for oracular declaration could end up obscuring their real origins.

A consultation of the Delphic Oracle as depicted in the video game Titan Quest (2006), complete with the priestess suspended over chasms oozing psychoactive gases

What of the stories of chasms with psychoactive gases, inspiring the priestess with a divine ecstasy? Well, these stories are late too. By the Roman era, things had changed a lot at Delphi. To take one metric: from the 6th century BCE to the 1st century CE, according to Joseph Fontenrose, we hear of exactly one authentic historical response that was given in verse. From the 2nd and 3rd centuries CE, we hear of six. And even with the one earlier response that he classifies as authentic, I wouldn’t be at all confident that it’s the Pythia’s own wording.

Note. For stories of psychoactive gases, see especially Diodorus 16.26.2-5; Plutarch On the decline of oracles 432c-438b; also Pliny NH 2.208; ps.-Aristotle On the universe 395b; Pausanias 10.5.7; Strabo 9.3.5; ps.-Longinus On the sublime 13.2; etc. Note that these sources refer to gas emissions at other sites too: the H₂CO₃ and H₂S emissions at Ampsanctus and the CO₂ at Hierapolis (Pliny) are real, the gases at Lebadeia (ps.-Aristotle) are not. See Fontenrose 1978: 197-203 for more sources and discussion.

On the seven surviving authentic oracular responses given in verse see Fontenrose 1978: 186-195. The only pre-2nd century CE one, H28 in Fontenrose’s catalogue, dates to the 300s BCE (quoted in Dem. Against Meidias 52). Even there, we are almost certainly looking at the product of an Oracle Writing Office or somesuch, rather than of the Pythia herself.

To judge from the description of the Oracle that we find in Plutarch, it appears that the Oracle had changed the way it worked in order to match the Herodotean stereotype more closely. Life imitated art.

The notion of real chasms and real psychoactive gases still enjoys wide circulation, thanks to the efforts of geologists like Jelle de Boer and Luigi Piccardi. But it’s a solution looking for a problem. The problem is how to explain the obscurity of the Oracle’s utterances. But that problem didn’t exist for the first 800 years of the Oracle’s recorded history.


  • Bowden, Hugh 2005. Classical Athens and the Delphic Oracle. Divination and Democracy. Cambridge: Cambridge University Press.
  • Fontenrose, Joseph 1978. The Delphic Oracle. Its Responses and Operations with a Catalogue of Responses. Berkeley, etc.: University of California Press.
  • Fontenrose, Joseph 1988. Didyma. Apollo’s Oracle, Cult, and Companions. Berkeley, etc.: University of California Press.


  1. Onomakritos is familiar to readers of Mary Renault's The Praise Singer, along with his exposure by Lasos. The relevant excerpts are here.

    1. I know your comment is a week old now (sorry, been busy) but I too always think of that novel when I mention or think of Onomacritus! I do feel sorry for Renault's version of him, but on the bright side, he's one of the few chresmologoi to have a decent modern edition, in Bernabé's edition of epic fragments.

  2. All the contrivance with the gold and silver pots sounds to me like he was forcing a binary decision because he knew if he asked the Pythia directly she WOULDN'T give a simple yes or no answer.