Vowel / Diphthong length

Greek Learning Greek
1

I came across a reference to the accentuation of -αι and -οι at the ends of words that said something like “while accented short, as diphthongs, these syllables were pronounced long.”

Smyth states the following: "The difference in the quantitative treatment of -αι and -οι depends on an original difference of accentuation that may have vanished in Greek. -αι and -οι, when short, were pronounced with a clipped, or simple, tone; when long, with a drawled or compound, tone.

So I’ve always tried to give final -αι and -οι a short value in Attic prose, but a long value in reading Homer. What is the metrical evidence for their length in Attic verse?

(Strangely accented -εως words like πόλεως are a product of transfer of quality from ηο → εω, so a completely different phenomenon.)

2

Final -αι and -οι before a vowel usually scan short in Homer, or at least those parts of Homer that I’ve read; I’m not sure why you would pronounce it long. I haven’t really read any Attic verse, but skimming through several pages of tragedy, every occurrence I found was followed by a consonant. If indeed hiatus was avoided with those finals, then that might say something about the Attic pronunciation of them, probably that they were almost monophthongs.

One of the nice things about my pronunciation system is that it gives those two diphthongs a base vowel ([æ] and [ø] respectively) different from any of the short vowels, so they can be pronounced as short vowels without causing any confusion.

Allen’s general recommendation for diphthongs is with a doubled (extra-heavy, I guess) glide, as [ajj] and the like. This gives you a straightforward way to describe a diphthong that shouldn’t scan long, namely [aj] and the like. If you accept that, you then have to ask what the standard way is to handle a glide after a short vowel that doesn’t make a long syllable. The answer is that the Attic/Ionic orthography did not normally write such glides. This could mean that those glides were not pronounced, or were reduced to the point where they were not distinctive, or it could mean that ancient writers chose not to write letters for some of the sounds they spoke if writing those letters would confuse readers into thinking that a short syllable was long. This is my view. I pronounce lots of glides where hiatus is written, always referring to the etymology to figure out which glide (or none) is most likely, and that makes the matter such that there is nothing exceptional in the sound system about final -αι and -οι being short vowel plus glide and therefore in a category that makes them shorter than diphthongs.

Another case that I put into the same category is forms of ποιέω/ποιῶ, which were frequently written without the iota: ποῶ. I interpret that as meaning that ποιῶ was the standard pronunciation, but the first syllable was short, and so a spelling was invented to express that fact. Eventually the spelling that properly expressed the vowel color became standard.

Incidentally, I’m beginning to think that it’s inefficient to think of short vowel phonemes and long vowel phonemes. It seems more accurate to describe the situation in terms of vowel phonemes that simply give vowel color, plus lengthening and shortening processes that can be lexical (prosodemes, if you wish). Thus a long vowel is a vowel plus a lengthening process; a long diphthong (such as ῃ) can be formed as a diphthong plus a lengthening process or a long vowel plus a glide. With this view, any vowel or diphthong can occur in short and long variants. The advantage of this approach is that prosodic patterns cut across vowels. The exceptional long accentuation of final -αι and -οι in optatives I interpret as meaning that there is a length prosodeme in the optative thematic vowel making in effect, two more long diphthong phonemes that are written as short diphthongs because they have different base vowels from ᾳ and ῳ.

3

hi, there are two points to separate out above:

  1. scansion of -οι and -αι in Attic verse when they are accented short - the question is, do they scan long or short? answer: they scan long. see e.g. line 16 of aeschylus’ agamemnon: the position guarantees that the syllable containing it is heavy, even though the word is accented as if the ending is short (and there is only one consonant after it and so the syllable is not heavy by being closed):

16 ὅταν δ᾽ ἀείδειν ἢ μινύρεσθαι δοκῶ,

  1. scansion of -οι and -αι in Attic prose when followed by a word beginning with a vowel: the question is, do they corrept and so scan short, or are they blended into the next word, or are they left long, or what? answer: no clear rule, but certainly in some clear cases correption does not occur. i discussed this back in 2009 here giving evidence for dionysius of halicarnassus’ scansion of Attic prose showing no correption: dionysius scans …ἤδη εἰρηκότων… as long: see the post from Thu May 07, 2009 12:32am here:

http://discourse.textkit.com/t/reduction-of-final-circumflex-accented-vowels-into-grave/8714/1

see in particular these bits of the post:

"…your first precept, however, is not right: correption does not always occur at vowel junctions between words. i gave in my first post above some e.g.s from ancient sources of blending or just hiatus (with no other change) occurring in these cases. in fact Martin West in his 1982 book on grk metrics shows that five different things can happen at a vowel junction. i have scanned the relevant section for you (remove spaces):

www . freewebs . com / mhninaeide / west1982-voweljunction . pdf

the next question is which of these five possibilities which West mentions occurs in prose. your first precept chooses correption. however dionysius of halicarnassus scanned some prose from thucydides and didn’t use correption. He took the sentence from thucydides 2.35.1:

“οἱ μὲν πολλοὶ τῶν ἐνθάδε ἤδη εἰρηκότων ἐπαινοῦσι τὸν προσθέντα τῶι νόμωι τὸν λόγον τόνδε, ὡς καλὸν ἐπὶ τοῖς ἐκ τῶν πολέμων θαπτομένοις ἀγορεύεσθαι αὐτόν.”

he scanned this as follows (once again I have noted in square brackets the metrical feet he is referring to):

“τρεῖς μὲν γὰρ οἱ τοῦ πρώτου προηγούμενοι κώλου σπονδεῖοι πόδες εἰσίν [i.e. three lots of long-long for “οἱ μὲν πολλοὶ τῶν ἐν-”],
ὁ δὲ τέταρτος ἀνάπαιστος [i.e. short-short-long for “-θάδε ἤ-”],
ὁ δὲ μετὰ τοῦτον αὖθις σπονδεῖος [i.e. long-long for “-δη εἰ-”: this shows that there is no correption of -δη before a vowel-initial word],
ἔπειτα κρητικός [i.e. long-short-long for “-ρηκότων”],
ἅπαντες ἀξιωματικοί.”

therefore you can’t assume that correption will apply at each vowel junction between words in attic prose. you need to look at it on a case by case basis and see if there is any good evidence for the choice you make.

to take the case you referred to above, τοῦ, there are attic prose texts which show blending of this into a following vowel-initial word, e.g. τἀνδρός for τοῦ ανδρός in e.g. Plato Laches 179e:

“Ἔδοξε δὴ χρῆναι αὐτούς τε ἐλθεῖν ἐπὶ θέαν τἀνδρὸς καὶ ὑμᾶς συμπαραλαβεῖν ἅμα μὲν συνθεατάς, ἅμα δὲ συμβούλους τε καὶ κοινωνούς, ἐὰν βούλησθε, περὶ τῆς τῶν ὑέων ἐπιμελείας.”

Also τοὐνόματος for τοῦ ὀνόματος in e.g. Demosthenes’ against Boeotus 1, section 21:

“Ἀκούετ’, ἄνδρες Ἀθηναῖοι, ὅτι ἐγὼ μέν εἰμ’ ἐπὶ τοὐνόματος τούτου πάντα τὸν χρόνον”

so blending τοῦ into a vowel-initial word would be one reasonable choice. if on the other hand you choose to read τοῦ as you see it, i.e. with no change even if a vowel-initial word follows (following generally dionysius’ scansions of vowel junctions in attic prose quoted above), note that Probert on accenting (2003) suggests that it is not likely that the circumflex in τοῦ was simply written by convention (see s277), i.e. you should pronounce it as a circumflex and not just as an atonic proclitic. the choice of correption in such a case in attic prose, however, goes against all the sources I have seen so far.

as a side point, correption in homer is quite interesting: a study by kelly in 1990 has shown that correption is twice as frequent in the speeches compared with the narrative, and occurs much more frequently at some positions in the line compared with others. i have a note on this study on page 37 of my old Iliad A notes (remove spaces):

www . freewebs . com / mhninaeide / IliadANotes . pdf"

cheers, chad

4

Ariphron: Final -αι and -οι before a vowel usually scan short in Homer…

Yes, but that’s epic correption. Most long vowels scan short before another vowel in Homer, so it doesn’t tell us much about a possible clipped version these at word end, as suggested by Smyth.

cb: 1. scansion of -οι and -αι in Attic verse when they are accented short - the question is, do they scan long or short? answer: they scan long.

This makes me wonder if the accentuation produced by the clipped versions of -αι and -οι wouldn’t be a later feature of the Greek language (post 200-B.C.) rather than an earlier. Chandler believed that most of our accentuation information is late (post 900-A.D.) rather than early, even if it is a (mostly?) highly conserved feature of the Greek language.

On correption, Allen had this to say, and I always assumed it was true:

This is a feature that perhaps goes back to Indo-European, since it is also attested in Vedic; it is commonest in Homer, and is therefore termed ‘correptio epica’ (more generally the principle is stated as ‘vocalis ante vocalem corripitur’); the rarer non-epic instances are in any case largely confined to dactylic/anapaestic rhythms, as e.g. Euripides. Hec., 123 ὄζω Ἀθηνῶν.

Also, interestingly – referring to general hiatus rather than epic correption – Allen suggests that more deliberate or official Attic speech (slower?) seemed to indicate more hiatus. Ie., the vowels were written out completely. He directs the reader to Threatte’s Grammar of Attic inscriptions, but I don’t have access to it.

That chapter also has a discussion of the different sorts of vowel juncture:

5

It seems we have three hypotheses floating around here, not counting the one proposed by Lucus Eques in the other thread:

  1. The traditional view, that these finals are diphthongs, normally making the syllables “long by nature” except in cases of “epic correption”, but that for some unknown reason they function as short vowels for purposes of accentuation;

  2. What has been my working hypothesis for quite a while, that the word-final position causes a reduction in unmarked diphthongs: in medial position they are long vowels evolving over time, making them “long by nature”, but when reduced they are best analyzed as short vowel plus semivowel functioning as a consonant, making them always short when followed by a vowel, and “long by position” (i.e. closed) when followed by a single consonant, and this reduction is a manifestation of a pattern that also made epic correption natural and that persisted from very ancient times until the loss of lexical quantity;

  3. An alternative idea that Joel is trying to evaluate, perhaps that the received accentuations are wrong for the classical period and that the words in question were re-accented in late classical times after these diphthongs had actually become short vowels [ɛ] and [y].

The traditional view is completely true as far as it goes, as it refers strictly to phenomena on the printed page and makes no assertions about how things were actually pronounced. The main test case for my view would occur when final -αι and -οι occur before a vowel (written hiatus) in verse where epic correption does not usually apply, that is to say long vowels occur in similar position and scan long. If these finals scan short by contrast, then it supports my view; if they scan long, then it opposes my view. Unfortunately, I am not aware of any genre where this test situation actually occurs.

Then there are the two distinct pronunciations described by Smyth. Honestly, I find what he writes here so vague that I’m not sure what to say. Is he saying there was a distinction in pronunciation at earlier times that had probably merged by classical times? And if so, which one would describe the later pronunciation? In any case, both pronunciations would scan long except in cases of epic correption (as Chad’s Aeschylus example illustrates), so there’s little or no metrical difference between them, just as there’s no metrical difference between αι and ᾳ.

It certainly seems more plausible that the received accentuations would reflect an earlier prosodic distinction that had been lost by classical times than that the accentuation of so many polysyllabic words would change in late classical times on account of this kind of rule.

Joel, does your earlier statement, that you give final -αι and -οι a long value in reading Homer, also apply to cases of “epic correption?” In that case when you read Homer, you prolong the vowels in certain syllables that scan short, which sounds strange to me.

6

Epic correption applies to all vowels in Homer, not just diphthongs. And of course it has to be read as shortening the long vowel, whatever that may be. Often there is no semi-vowel present:

ἡμένη ἐν βένθεσσιν ἁλὸς παρὰ πατρὶ γέροντι·

There are plenty of places where correption doesn’t happen and they can provide the counter-examples you asked for. Two lines from the Iliad where αι/οι is long in hiatus, and there is no possibility of a digamma:

ἢ ὁδὸν ἐλθέμεναι ἢ ἀνδράσιν ἶφι μάχεσθαι;

τῶν ἤτοι αὐτῶν τέρενα χρόα γῦπες ἔδονται,

Also, in synizesis, which you’d think might naturally be explained by semi-vowels, you don’t see the semi-vowel prolonging the previous vowel:

Πηληϊάδεω

Something else seems to be going on.

So while a diphthong in correption generally gets turned into a short-vowel semi-vowel pair (per Allen, and that’s how I try to pronounce it) it’s not really a general explanation for correption, nor does it seem to happen at word end normally.

EDIT:

Also – and maybe I should have put this first – if there is no hiatus for αι and οι, why are they elided in Attic? And why would ει be accented differently (or optatives, or certain infinitives, for that matter)? None of these are impossible questions, but they reduce the explanative power of word end semi-vowels.

EDIT 2:

Apparently, according to a footnote in Allen, Sommerstein’s Sound Pattern of Attic Greek claims “the elision of -αι and -οι was not a feature of careful Attic Speech.” The book was $1 on Amazon and I have just ordered it, so I might know more in a few weeks.

7

“Something else seems to be going on.”

A lot of things are going on underneath the text of the Homeric poems, which didn’t take final shape until the Hellenistic era but were the end-product of many layers of tranformations of traditional epic language.

ἢ ὁδὸν ἐλθέμεναι ἢ ἀνδράσιν ἶφι μάχεσθαι; – the hiatus occurs at the main caesura, a strong break in the verse. No wonder there’s no correption.

Πηληιαδεω Αχιληοσ – Underlying this Ionic formula is probably an Aeolic (or “Achaean”) form Πηληιαδα’ Αχιληοσ, for Πηληιαδαο, which was presumably transformed first to an ealier Ionic form Πηληιαδη’(ο), then by “quantitative metathesis” to Πηληιαδεω. I don’t even think we can be sure that the spelling -εω was ever pronounced as anything other than a monophthong in the “original” Iliad, whatever that was and whenever and however it was composed.

τῶν ἤτοι αὐτῶν τέρενα χρόα γῦπες ἔδονται, – I don’t have an explanation for this, except that the hexameter is a very difficult meter to write or compose in Greek, and perhaps the aoidos could be allowed licenses.

8

And notice the lack of correption following Πηληϊάδεω (or Ἀτρεί̈δεω for that matter) everywhere in the Iliad.

However I don’t want this to distract from the point that -εω couldn’t have been pronounced as -yω, because is would have affected the quantity of the preceding vowel. I guess it’s a monophthong, but that wouldn’t have to be the only possibility.

πόλεως in Attic, also a product of vowel-exchange, scans as short followed by a long in Euripides. (@Ariphron: ποιῶ scans as long-long in Euripides.)

My short pronunciation for αι or οι has been a glide, following Allen, just as Ariphron does. Much like the diphthong / semivowel at the end of the English word “fly.” You can hear both the diphthong and glide versions here: http://forvo.com/word/fly/#en

Word-internal, not following Allen – except insofar as he mentions that Greek was likely mora-based – I aim for something closer to Japanese mora-values: http://forvo.com/word/aoi/#ja (she’s speaking slowly here, but not too slowly).

I’ve been a bit anxious about this pronunciation for the reasons listed in the thread, but I will report back once I’ve read Sommerstein.

9

I don’t think there was any difference in length in the pronunciation final syllables of words like οἶκοι and οἴκοι. Syllable length in poetry and vowel length for the purpose of accentuation are two completely different things. -οι and -αι are treated just the same in poetry, they are long unless there’s correption, just like any other word-final long vowel or diphthong.

My guess is that accentuations like οἴκοι are a survival of an earlier stage of the language, where either the rules of accentuation were different or the words themselves were different or both. I mean a strange survival analogous to πόλεως, but earlier. The important point is that in the historical period it’s a phenomenon that pertains to accentuation, it doesn’t change the final diphthong in any way (although perhaps the final diphthongs in οἶκοι and οἴκοι were not originally the same – maybe we had something like οἴκοϊ originally. I really have no idea, this is just a guess to show the sort of thing I think it could be, but I’m sure it’s treated somewhere).

On a different note, there are many cases with -εω that can’t be explained by the sort of explanation given for Πηληιαδεω < Πηληιαδηο. E.g. Theogony l. 44: θεῶν γένος αἰδοῖον πρῶτον κλείουσιν ἀοιδῇ.

10

I don’t think this is a valid argument, compare the treatment of mute+liquid or mute+nasal, e.g. Ἀφροδίτη, where the first syllable is always short.