Monday 18 January 2016

Colours in Homer #2: the wine-dark sea

Homer regarded wine, the sea, and sheep as all being the same colour, which is 'red'.
-- QI, series 2 episode 1 'Blue' (Oct. 2004)
As we saw last time, it's easy to get mixed up over colour terms in another language. Let's repeat an important point: different languages divide up the spectrum differently. As a result, colour terms often do not have one-to-one equivalents in different languages.

We can't conduct experiments to see exactly which labels a live ancient Greek would attach to which shades, so we have to rely on surviving texts and how they use colour terms. The colours I listed last time come out roughly as follows:

'Kyaneos': porpoise and barn swallow (source:
  • kyaneos -- conventional translation 'blue'. Used of: porpoises, a swallow's feathers, black hair. So actual range = medium grey-blue ranging towards black; not vivid.
  • melas -- conventional translation 'black'. Used of: soil, blood, wine, sacrificial rams. So actual range = black, but also ranging to deep browns and reds.
  • porphyreos (in Homer porphyrios) -- conventional translation 'purple'. Used of: flowers (of unspecified species) and dyes; also blood, blushing cheeks, rainbows. So actual range = vivid colours of any luminosity throughout the indigo-lavender-violet part of the palette (and perhaps beyond). Porphyreos doesn't have a one-to-one equivalence to any English colour-term.
  • chlōros -- conventional translation 'green' or 'pale'. Used of: the complexions of frightened people, leaves, live plants; also figuratively to mean 'unripe'. So actually = colours around chartreuse: low-saturation, high-luminosity colours with a high yellow component (i.e. not pink).
  • glaukos -- conventional translation 'grey'. Used of: (dull green) olives, grapes, vine leaves; (blue) eye colour, the sea, a clear sky. Ancient technical descriptions consistently treat it as a light shade of kyaneos; see the Plato passage below. So actual range= light blue to dull green, with the latter also understood as a greenish-grey. Like porphyreos, this term doesn't have a one-to-one equivalence to any English colour-term.
  • leukos -- conventional translation 'white'. Used of: snow, marble, an old person's hair, unusually fair skin. Pretty much = English 'white'.
  • erythros -- conventional translation 'red'. Used of: wine, ruddy complexions, blushes, clay, egg yolk. Egg yolk is more prototypically ōchros for the Greeks (conventionally translated 'yellow'); but yolk colour is affected by a hen's diet, and hens that are given scraps often produce dark orange yolks. So 'red', but also used for neighbouring colours figuratively, as also in English ('red robin', 'red-head').
Ancient Greek colour terms: estimated palette
Estimated palette ranges
for the colours listed above
One important piece of testimony is an account by Plato of how vision supposedly works (Timaeus 68b-c, 4th century BCE). The theory of colour perception is obviously very inaccurate. However, it has historical value for the fact that it gives colour terms for several combinations of pigments:
And again, when the intermediate kind of fire reaches the liquid of the eyes and is mixed with it, it doesn't gleam: rather, because of how the fire's light in blended through the moisture, it emits the colour of blood, and we call it erythros. But when lampros ('bright') is mixed with erythros and leukos, we call it xanthos ('light brown, fawn')... And erythros mixed with melas and leukos produces alourgos (lit. 'sea-worked'; apparently close to Tyrian purple in Xenophon Cyr. 8.3.3); but it is orphninos ('dark, dim'?) if these (pigments) are more burnt and more melas is added. And pyrrhos ('red-orange, russet') is made from xanthos and phaios ('grey'); phaios is made from leukos and melas; and ōchros ('yellow') is made from leukos mixed with xanthos. When leukos combined with lampros is poured into deep melas, it produces the colour kyaneos. And kyaneos mixed with leukos gives glaukos, and pyrrhos mixed with melas gives prasios ('light green' or 'khaki').
The sky. After all this, by the way, it turns out that the question asked in QI, 'What colour was the sky in ancient Greece?', actually has the answer: either glaukos 'light blue' or lampros 'bright', depending on context. Neither is very strongly supported in ancient testimony, but ancient Greek texts just don't talk about the colour of the sky much. We find clear references to blue sky as glaukos in the 1st century CE writer Lucius Annaeus Cornutus (Compendium 10.20 -- he also likens it to olive-tree foliage, confirming that that is also part of the glaukos area of the colour palette) and in the 3rd century CE writer Philostratus (Life of Apollonius 2.5). Lampros, 'bright', is less specifically about colour than the other terms: more properly it refers to the aithēr, the intense light that supposedly lies beyond the misty aēr 'air, atmosphere' in ancient Greek cosmology. Lampros is regularly used of the sun and stars, and more metaphorically of things like 'clear' voices, 'brilliant' diction, and 'splendid' horses. Yet Plato, above, treats it as a pigment; and a handful of passages associate it with the sky itself, like Suda χ.326 and Iamblichus De mysteriis 2.7 ('earthly things display an earthly and blacker [melanteron] fire, heavenly things a brighter [lamproteron] fire'). Cornutus, too, calls the sky pyrōdēs 'fiery in appearance' (Compendium 2.10).

There's no doubt that lampros would have been the most obvious term to describe a bright clear Mediterranean day. But someone who wanted to talk specifically about its colour would call it glaukos, as in Cornutus and Philostratus. (An overcast day would presumably be 'grey', phaios.)

Now, we still need to deal with the QI claim that Homer refers to red wine, red sea, and red sheep.

Red wine
Red wine (source:
Sheep. This one is just a misunderstanding: Homer doesn't call sheep 'red' anywhere. It's probably based on a few passages where lambs or rams are called melas 'black'. Melas can also be used of blood, so that's probably what caused the confusion. For the record, Homeric sheep are prototypically white: they're argennos in five places (literally 'silver-clothed'), and argyphos in two others (literally 'silver-woven'). (And by the way, as it happens red sheep do actually exist -- but they probably aren't relevant.)

Wine. The adjectives used for wine in early Greek epic poetry are:
  • aithops: conventional translation 'shining, gleaming'; also used of burning torches, smoke (glittering with sparks), and the mythical Aithiopes (whose skin was burnt by the sun). Used for wine 12× in the Iliad, 10× Odyssey, 2× Hesiodic Works and Days.
  • erythros: 'red'. 7× Odyssey, 1× Homeric Hymn to Demeter.
  • melas: 'black, deep red'. 3× Odyssey.
So wine gleams, but its colour is prototypically red. There, that was easy!

The sea. The business of the 'red sea' is tied up with the colour of wine. The 'red sea' claim on QI is almost certainly based on a frequent formulaic phrase in the Homeric Odyssey, epi oinopa ponton, conventionally translated as 'on the wine-dark sea'. Based on the discussion of wine above, 'wine-dark sea' sounds like it ought to mean 'red sea'.

This famous phrase lacks any totally certain explanation. Stephanie West's 1988-90 commentary on Odyssey books 1-4 takes an agnostic position, simply calling it 'puzzling'. In 1983, R. Rutherford-Dyer argued, based on his own observations, that it referred to the colour of the Aegean Sea at particular times of day ('Homer's wine-dark sea', Greece & Rome 30: 125-8; subscription required). A few years earlier, in 1978, C. H. Gordon argued that 'wine-dark sea' was a traditional East Mediterranean poetic phrase also supposedly found in Hebrew taršiš -- according to Gordon, 'a qaṭlîl formation of the denominative root trš derived from tîrôš, "wine," paralleling ḥaklîl, and meaning "wine-red" or "wine-dark"' ('The wine-dark sea', Journal of Near Eastern Studies 37: 51-2; subscription required).

Personally I'm not taken with the romantic suggestion of colours at sunset. Yes, the Aegean can take striking colours at that hour, but an explanation that relies on a specific time of day needs something to justify interpreting it with that specificity. I'm even less impressed by Gordon's argument: it relies on oinops meaning 'wine-dark'. The trouble is, that's not a firm foundation. Strictly literally, the phrase straightforwardly means 'wine-faced sea', from οἶνος 'wine' + ὄψ 'face'.

[Corrigendum, 31 March: a 1995 article by Beekes shows that this is wrong. See comments. 'Wine' + 'face' would be οἰνο- + ωπ-; οἰνο- + οπ- is actually 'wine-looking', that is, 'looking like wine'. I've silently corrected the translations below, where relevant.]

The Aegean Sea at sunset
Sunset on the Aegean Sea (source:
The phrase 'wine-dark sea' didn't find its popularity in English until the first English prose translations of Homer in the 1880s-90s (Iliad; Odyssey). Andrew Lang contributed to both translations, so perhaps he bears the greatest responsibility. I'd guess Lang felt 'wine-dark' had a better ring than 'wine-looking' -- and of course it absolutely does sound better.

That interpretation has a long-standing history prior to Lang, though. 'Wine-dark' appears as a secondary translation for oinops in the Liddell and Scott Greek Lexicon. The older Greek-German dictionaries on which Liddell & Scott was based, those of Passow and Schneider, refer to darkness ('Dunkelheit') and dark red ('dunkelroth'). These translations probably go back ultimately to the early Byzantine lexicographer Hesychius. His dictionary contains these entries [edit: and two more on the previous page]:
οἰνωπόν· πορφύρεον, μέλανα. ...
οἰνώψ· οἶνοψ. μέλαν.

wine-faced [noun]: porphyreos, melas. ...
wine-faced [adjective]: 'wine-looking', melas.
So, why does Hesychius gloss 'wine-faced' as 'black'? Probably because water in early epic, especially drinking water, is often melas: three rivers have melas water (Iliad 2.625; 21.202; Odyssey 6.91); ships take on stores of melas water (Od. 4.359); the sea nymph Calypso disappears into a melas wave (Od. 5.353); the sea around Charybdis is melas (Od. 12.104); a freshwater spring has melas water (Od. 13.409), and in five places there is a κρήνη μελάνυδρος 'melas-watered spring'. A mostly lost epic, the Cypria, referred to the 'barren melas water (of the sea)' (fr. 9.6 ed. Bernabé).

And how does that get us to wine? The answer can only be that Homer calls wine melas in three places. Wine is melas; water is melas; therefore (goes the argument) 'wine-looking' is 'dark'.

But as we saw above, Homer doesn't just call wine melas: he calls it erythros more often, and aithops more often still. So Hesychius' gloss doesn't look well-founded either. West, commenting on the phrase in Od. 1.183, is also unimpressed: 'the conventional rendering "wine-dark",' she writes, 'does not inspire confidence, [but] it is more convincing than alternative suggestions.'

Can we do any better? Maybe. I suggest that an explanation for the 'wine-looking sea' can be conjectured based on the Homeric system of metrical formulas. My suspicion is that epi oinopa ponton was designed to be a substitute for an unattested phrase, *ep' aithopa ponton 'on the bright-looking sea', but with a different rhythm. Hypothetically, an epic poet could choose which formula to use depending on the rhythmical context. *Ep' aithopa ponton was the original version (since it actually makes sense), but either it died out in the pre-Homeric phase of the epic tradition, or it was never permissible in strict hexameter; as a result only the oinopa variant is attested.

The hypothesised *aithopa/oinopa pair follows a pattern of rhythmic variants like the following --
polyēraton eidos echousa
epēraton eidos echousa

'(a woman) possessing lovely beauty'
-- where the formula is versatile because a syllable can be added or removed without a second thought. A pair of this kind would be absolutely typical for the epic tradition.

Technical discussion: [Note: a few changes and corrections have been made to this discussion, 20/1/16.]
Epi oinopa ponton has the extra syllable because the preposition epi ('on') gets abbreviated before aithopa ('bright-looking') but not before oinopa ('wine-looking'). In phrases like epi + aithopa, ancient Greek regularly contracted a short vowel at the end of one word when followed by a vowel in the next word. However, that couldn't happen with epi + oinopa, because when the formula was devised, oinopa had an extra consonant, w, which had dropped out of the epic dialect by the time the Iliad got written down. Originally the phrase was *epi w(o)īnopa (*w(o)īn- 'wine' comes from the same root that produced Latin vinum, originally pronounced wīn-). Again, it is absolutely typical for Homeric formulas to respect this lost consonant when deciding whether to contract a vowel or not. The phenomenon is known as 'observing digamma': digamma was an archaic letter used to write the w sound in some parts of Greece. Aithops never had a w sound at the beginning (it comes from the verb aith- 'gleam, shine') so there would be nothing to prevent contraction of epi + aithopa → *ep' aithopa.

Why do we never see *ep' aithopa ponton? If this conjecture has a weak point, this is it, because there is no convincing explanation. The formula epi oinopa ponton appears after a fourth-foot caesura: *ep' aithopa ponton in the same position would violate Hermann's bridge. Put simply, what that means is that it's nearly impossible in Homeric metre. Given that, where could it have come from?
  1. Could the formula have developed in a prehistoric form of the hexameter and then died out once Hermann's bridge developed? No: we have to presume that Hermann's bridge was an emergent feature of the forces that created the hexameter in the first place, not an epiphenomenon.
  2. Could the aithopa variant have appeared in a proto-hexameter where it was possible to have lines with fewer prolongations, that is to say with five feet rather than six? In such a line it would follow a tritotrochaic caesura, the most typical position for formula-pairs of this type. No, this can't be right either. If so, we'd expect to see evidence of other formulas where the same thing happened, with more violations of Hermann's bridge.
  3. Could the formula have developed in an Aeolic metre with dactylic prolongation but without the same constraints as hexameter, and subsequently epic poets borrowed the oinopa variant but not the aithopa variant? Again no: (a) what would be a poet's motivation to avoid the aithopa variant? Again, we have to presume Hermann's bridge is an emergent feature of the formulaic system, not a prescriptive rule. (b) We have no attestations of aithops/aithopa in extant lyric poetry of any kind: prior to 400 BCE, the only other uses of the word in verse are in Euripides' tragedies (3×; perhaps also a doubtful occurrence in Thespis fr. 4. Two are in melic/lyric passages, Suppl. 1019 and Bacchae 594; the third is the only one that refers to wine, Eur. fr. 896, a just-so rationale for why wine is called aithops).
  4. Perhaps *ep' aithopa ponton was never actually used in verse, precisely because it was metrically impossible, and oinopa was devised specifically to compensate for that (at a time when the w of *woinopa was still pronounced). This is the only possibility that is at all likely.
Now that I've undermined my own conjecture in the 'technical discussion' above, I'd better also lay out a summary of the reasoning in support of it:
  1. Aithops and wine are each closely associated with the other in early epic poetry:
    1. aithops is the most typical adjective for wine (24×; cf. erythros 8×, melas 3×);
    2. wine is the substance most typically described as aithops (24×; cf. 'bronze' 12×, 'smoke' 1×, 'hunger' 1×).
  2. Aithopa and oinopa are closely parallel forms.
  3. Oinopa ponton is almost impenetrable on its own, but makes perfect sense if it is understood as a substitute for a similar-sounding word meaning 'gleaming, shining'.
I believe that's plenty to make it a sensible conjecture, at least on a par with Hesychius', Gordon's, and Rutherford-Dyer's explanations. But it does involve a secondary conjecture (or rather speculation) about why we never see the aithopa variant. The upshot is that we still don't have a rock-solid explanation of why Homer's sea is 'wine-looking' -- but there are some good possibilities that we can suspect.


  1. What do you make of Aristotle in Athenaeus 9 (394A): "ἡ μὲν οὖν οἰνάς, φησὶν ὁ Ἀριστοτέλης, μείζων ἐστὶ τῆς περιστερᾶς, χρῶμα δ᾽ ἔχει οἰνωπόν..." ("So the rock dove, Aristotle says, is bigger than the peristera, and has an oinopon colour").
    That seems to push Hesychios' gloss back to the fourth century BC (Seems! I can only find Aristotle himself talking about size, not colour!).
    Beekes claims that the ending -οπ- in Aithiopes is pre-Greek and compares "Dryopes." Could oinops similarly be pre-Greek and the connection with wine be a coincidence?

    1. I apologise: I missed this comment until now, as I hadn't looked at the page that notified me of comments awaiting moderation. Oops.

      The Aristotle snippet certainly makes it clear that οἰνωπόν is a colour term. But what Hesychius is emphasising is that οἰνωπόν is *dark*. I don't see that in the Aristotle. (Unless there's a variety of rock dove that's darker than the others?) If anything I'd look at a pigeon's shimmering neck feathers and think that that's what οἰνωπόν is about - shimmering.

      Beekes' argument is very interesting and attractive. I didn't know the article he cites there, in Glotta 1995-96 ( It has a lengthy discussion of -op-/-ōp- roots at pp. 18-25; οἶνοψ is at pp. 22-5. He seems to have changed his mind between that article and the Etymological Dictionary. In the Glotta article he regards the old nominative forms as οἶνοψ and αἴθιοψ, with -οπ- related directly to the verbal root (as in ὄπωπα) and meaning "-looking, -seeming", not anything to do with faces. In the Etymological Dictionary he is happy to stick with this for οἶνοψ (see under ὄπωπα), but for Αἰθίοπες he treats it as pre-Greek for some reason. I wonder if that's an unintentional slip.

      There are a few other interesting points from the article:

      (1) Beekes comments on the description of the sky as πυρώδης that we find in Cornutus (p. 17); but it's the folk-derivation αἰθήρ < αἴθω that he's criticising, not the adjective itself.

      (2) At page 20 he mentions another entry in Hesychius, νέρωπα· λαμπρόν. When you put that alongside οἰνώψ· οἶνοψ. μέλαν ...I wonder.

      (3) Beekes' analysis implies that the name of Aesop most certainly means "burning-faced". αἰθ + ι + οπ- ("-looking") gives αἴθιοπ-, but αἰθ + ι + ωπ- ("-faced") would become assibilated like πρόσωπον and give αἴσωπ-. (Beekes explains the ι as a Caland i-stem: if you need an explanation of what that is, as I did, then see e.g.

  2. Just got here via a Twitter link; if a random comment on a year-old post is okay, I have some thoughts.

    Why couldn't ἐφ᾽ αἴθοπα πόντον have developed as a formula for use in a different part of the line? E.g. a line-beginning formula with a break at the regular "masculine" caesura:

    οὔδ᾽ ὅ γ᾽ ἐφ᾽ αἴθοπα πόντον ἔβη μετὰ τοῖς ἑτάροισιν

    (an invented example)

    1. I beg your pardon for neglecting this comment over the last week: busy busy.

      Your conjecture is a nice one, and sounds plausible, GIVEN my idea -- it's just that my idea is really just a speculation. If we can find some evidence to support your conjecture, it'd also support the speculation (and I'd like that).

      In the year I've had to think about this, I'd not thought about any potential evidence for αἴθοπα being used like this, and had pretty much given up on it.

      You've prompted me to take a look at the formulaic characteristics of αἴθοπα. Like οἴνοπα, αἴθοπα is strongly associated with a 5th foot position in extant hexameter (21x Il, 10x Od, 3x W&D, 1x Aspis), usually followed by οἶνον (24x; 11x with "bronze"). Less popular are 4th foot (αἴθοπα οἶνον in that position 2x Od, αἴθοπα πινέμεν οἶνον 1x W&D) and 2nd foot (αἴθοπα οἶνον in that position 1x Il, 2x Od).

      Perhaps the most telling instances are Iliad 1.462 and Od. 3.459 ἐπὶ δ' αἴθοπα οἶνον (at line end), with the desired ἐπί. These are the only occasions where we get a preposition before αἴθοπ- (W&D 592 has ἐπὶ δ' αἴθοπα πινέμεν οἶνον, but the ἐπί is adverbial). But they can cut either way. They give us some support for ἐπὶ δ' αἴθοπα, but not for ἐπ' αἴθοπα.

      I don't think mobility in the line is enough: support for ἐπ' αἴθοπα is essential for the idea to work. A pity, but there it is. If you can think of anything to support it, I'd love to know!

  3. I wonder if, corrigendum notwithstanding, it really is 'wine-faced' and means something like 'drunken sea', with connotations of heavy swells and stormy seas. Nothing to do with colour. Of course, this doesn't explain why it's used with oxen.

  4. The big Liddell does give as a definition of οἰνάς "a wild pigeon of the colour οἰνωπός, the rock-dove, Columba livia." Yet the Latin name for the stock-dove, columba oenas, does of course pay homage to the Greek word. Aristotle (History of Animals V) says that the οἰνάς is larger than the περιστερᾶ, or rock-dove.(Noted as much in the first anonymous comment here.) Such a distinction continues into modern Greek. But what is common is that they are not birds of one colour, but change from light grey body to purple chest to blue (cyan aka aqua) neck to darker grey head. It seems that this patterning, and the shimmering effect it might give in flight (I can't say I've ever really noticed pigeons this close, but then I'm not into divination by birds) would accord well with the changing colour of the sea in different depths, light conditions, etc. But this pigeon, whichever one is meant, is hardly of "dark" plumage.
    So maybe this is a wild pigeon chase? Maybe we are better off with Passow over Schneider, as Passow defines "oinop-" in terms of the restless seas and then ("also" [therefore]) says that this can be compared to the "rothspielende Dunkelheit des Wogenschlages." Of course, as the article points out, "wine-dark" is (for a classicist) of recent coinage, really only coming in with L & S (to my knowledge and-failing-memory, neither Chapman nor Pope much used the term, preferring motion-adjectives instead).
    I'm even less sure about this last point, but need wine be red/dark? I take the point that Homer etc use "red" and "black" but this adjectival use seems to specify, ie this wine, as opposed to another, rather than point to a general condition. A counter-argument might be that the colour term is a pleonasm, used in the "epic formula." But as we are aware, Greek wine comes in a variety of hues, from Retsina to Chian, so "wine-appearing" may not be specific.

  5. This is a fascinating blog!
    I wonder what wine looked like in a Greek drinking cup (like this ) --I would think quite a bit deeper and more opaque than in glass.