Wiktionary talk:About Ancient Greek/Archive 2

Redesigning the IPA tokens?

In the years since I wrote the Ancient Greek IPA templates, I have learnt much about efficient template design, and now I realize that I wasted a lot of unnecessary resources writing them, and also made them harder to maintain by spreading them over multiple small templates. I'm considering rewriting the templates with a different approach — instead of having separate templates for every little phoneme with per-template logic to handle Classical vs. Koine vs. Byzantine, I'm thinking of having individual templates for Classical vs. Koine vs. Byzantine and having per-template logic for token names. This seems to be the more orthodox design, being easier to maintain and gentler to embed. If I can get them redesigned, the theory is that the main IPA template syntax will not change — just the underlying mechanics will. So if I can get around to this, I wanted to let everyone know that it might happen. We shouldn't need hundreds of tiny templates to do what maybe a half dozen templates can do better. Compare the work I've already done with Marshallese templates. - Gilgamesh (talk) 19:15, 20 May 2012 (UTC)

Eh...I'll hold off for now. This is a little more than I'm willing to bite off and chew. I just thought it would be nice to go back and reform the clutter. - Gilgamesh (talk) 20:56, 20 May 2012 (UTC)

If you do get around to overhauling them, could you change two things that have always bugged me about how Ancient Greek is transcribed at Wiktionary? First, the stress mark (ˈ) is supposed to be before the stressed syllable, not the stressed vowel. Second, could you please remove the "downtack" marks from [e] and [o]? They're very distracting and don't add any useful information. In a language where [e] and [o] are not phonemically distinct from [ɛ] and [ɔ], it's sufficient to write [e] and [o] even if the vowels are lower than their cardinal values. It's just standard IPA practice not to use diacritics unless either (1) they're needed to distinguish phonemes, or (2) they're illustrating a subphonemic distinction that's under discussion. Neither of those is the case here. So could we please transcribe πόλις in Koine as [ˈpolis] rather than [pˈo̞lis]? Thanks. —Angr 21:41, 20 May 2012 (UTC)

I think I understand that. I'm considering a lot of things. Note however that if the stress mark is to be moved, a new token needs to be made, and every embedded template updated to reflect it. And until that work is done, the Koine and Byzantine rows will not indicate stress at all. In the interrim, maybe that will be clear enough as long as pitch accent is still indicated in the Classical row. - Gilgamesh (talk) 13:46, 21 May 2012 (UTC)

By the way, note that the downtack was not applied frivolously. In Koine, there is opposition between [e] and [ɛ], and so neither is marked. But there is no opposition between [o] and [ɔ] — the downtacked [o̞] was meant to reflect "neither quite [o] nor [ɔ], or maybe both at once". However, you are correct that a specific diacritic is not needed. Also, considering that omega became [u] in some Byzantine dialects while omicron remained [o], there might have been a real opposition afterall, but just wasn't deemed distinctive enough in most Koine dialects. I was wary of injecting too much OR into the transcription. - Gilgamesh (talk) 13:58, 21 May 2012 (UTC)

After some investigation, I remembered why the system wasn't designed to put Koine/Byzantine stress symbols before consonants instead of vowels — it significantly complicates the token logic. There are already a variety of tokens for consonant assimilations, especially geminate consonants and prenasalized consonants — the stress mark would have to be sandwiched between the IPA symbols, necessitating more tokens to compensate for it. Also, Greek hasn't historically broken syllables up between consonants, as it has focused mainly on the vowel nuclei. If only there were a stress accent symbol that could be attached to the vowel without having the appearance of cutting off the consonants before it. I'll have to research the IPA diacritics to see what's available. - Gilgamesh (talk) 19:30, 21 May 2012 (UTC)

Found it. At Stress (linguistics), it says under "Notation" that there is established precedent to mark stress accent with an acute mark, so that there doesn't need to be a decision about syllable boundaries. I suppose we could just do this. - Gilgamesh (talk) 19:35, 21 May 2012 (UTC)

By the way, I came to realize that our definitions for the time periods are rather fuzzy. "Classical", "Koine" and "Byzantine". The problem is...each of these was pluricentric.

• For "Classical", I've been assuming conservative Classical Attic, with perhaps some phonemic considerations for Ionic, etc. But there was also Doric and Aeolic. At the time, these were not separate languages—they were mutually intelligible dialects of what was recognized as the same language. Since the language is "Classical" and not specifically "Athenian", I wonder how much abstraction we should afford. That might mean υ being phonemically /u/ in all circumstances, as it remained firmly a back vowel in several of the dialects.
• For "Koine", I've been assuming something close to well-documented 1st century BC Egyptian Greek. This was considering that, in the Hellenistic world spread by Alexander the Great, the most important Hellenistic city was Alexandria, and even Athens itself had greatly diminished in importance. But there were still other dialects, including well-documented standards in the 4th century BC among the Roman educated population, and in Boeotia. These seem to have had pretty different vowel inventories. The Roman conventions seemed to preserve several classical archaicisms, while the Boeotian convention was one of the first to start monophthongizing its diphthongs and raising its vowels. Aspirated stops finally became fricatives towards the end of Koine, but this had not happened by 1st BC in Egypt.
• For "Byzantine", this has been rather tricky, because the Medieval Greek period spans twelve centuries, only the first three of which can still be considered Late Antiquity. I've been trying to be rather conservative, including details known since the end of Koine, with some of the earliest changes of the Byzantine period.
  • Thus far I haven't completely been using Byzantine consonant sandhi—only for ξ /ks/ and ψ /ps/, which in the very conservative sense should probably actually be /xs/ and /fs/, having changed to plosive-fricative because of Byzantine sandhi. The sandhi became a highly predictable rule of Byzantine-era phonotactics, reflected in common misspellings of words. This was only reversed during the 19th century Katharevousa pronunciation reforms, when the sandhi for archaicized words became more stigmatized, but ξ and ψ remained /ks/ and /ps/. (While Katharevousa itself is no longer official, its pronunciation reforms remain a marker of modern educated speech.) If used pronunciation tokens to change all κτ/χθ and πτ/φθ to /xt/ and /ft/, it might raise the eyebrows of editors applying the modern stigma to known ancient phonotactics. I'm considering, instead, using up-tacks and down-tacks to indicate tendencies for these clusters, such that κτ is /k̞t/, χθ is /xθ̝/, πτ is /p̞t/, φθ is /fθ̝/, etc. The up-tack on fricatives would mean a pronunciation traditionally fricative, but tended to become plosive as a matter of phonotactics. The down-tack on plosives would mean a pronunciation traditionally plosive, but tended to become fricative as a matter of phonotactics. Even classically-trained Byzantine scholars may have been aware of the ancient pronunciations, but may have struggled to actually make distinctions not part of their language, outside of writing. Classicist Byzantine usage made use of contemporary sounds and their associated vowel mergers, but to apply a strict distinction against consonant sandhi because the 19th century stigmatized it when there's no evidence Late Antiquity did, seems like original research from a retrospective modern position.
  • It is well established that the Byzantines retained plosive pronunciations of μβ γγ νδ as /(m)b (ŋ)ɡ (n)d/ throughout their existence even as the plain β γ δ had long already became fricatives /v ɣ ð/. It is also known that the 19th century reforms adopted a pronunciation that had never been natively used before, with μβ γγ νδ becoming /ɱv ŋɣ nð/ (though /ŋɡ/ for γγ was often adopted instead)—this was artificial then, but has become standard today. As for γκ μπ ντ, it is only known that they became voiced /(ŋ)ɡ (m)b (n)d/ sometime during the Byzantine period, and certainly by the time they had regular contact with the Turks, but at an unknown time before then. To mark this as a Late Antiquity uncertainty, I've been thinking of using voicing diacritics /ŋk̬ mp̬ nt̬/.

Any thoughts? I don't want to simply go off on a grand editing tangent without some discussion first. I know we try to avoid original research, and that seems like a hazy topic here—to apply modern pronunciation stigmas where there was no evidence they existed in the distant past, seems itself like original research to me. It doesn't seem acceptable to just remove Koine and Byzantine pronunciation rows for inconsistent or uncertain phonologies, particularly when various important "ancient" Greek words were coined no earlier than those times (in the classicist usage of the time). All these considerations seem to make it impossible to avoid any potential allegations of original research no matter what we decide to do. - Gilgamesh (talk) 17:14, 23 May 2012 (UTC)

I'd definitely advise against using diacritics in the way you're contemplating. /k̞/ would mean a sound that is neither /k/ nor /x/, but something in between, and likewise /k̬/ would mean a sound that is neither /k/ nor /ɡ/, but something in between. But you don't mean that; you mean a sound that is sometimes /k/ and sometimes /x/ depending on speaker and time period, and a sound that evolved from /k/ to /ɡ/. I wouldn't worry too much about the ancient dialects; they're often spelled differently anyway. Why indicate a Doric pronunciation for μύλη (for example) if the Doric word was actually μύλᾱ? —Angr 19:39, 23 May 2012 (UTC)

Generally, I think our current setup for {{grc-ipa-rows}} is pretty damned good, if perhaps a little over-conservative, especially in Koine. I'll try to address all of your suggestions, and then make a few of my own.

• For upsilon, I think we should actually go the other way and represent it with /y/, to better fit standard scholarly consensus. Certainly /u/ and /ʉ/ are well accepted theories, but they're theories of very old (older than Classical) Greek.
• I think that our Koine is already too stodgy as it is. 4th century is Classical, not Koine. Representing learned archaisms doesn't really seem appropriate. Specifically, I think that for Koine we should get rid of vowel length distinctions and get rid of word initial aspiration (rough breathing marks). Additionally, I think we should consider representing the voiced stops (β, γ, δ) as fricatives (/β/, /ɣ/, /ð/). I think if we do that, and leave the aspirate stops (θ, φ, χ) as is, it'll show a nice progression. Finally, I agree with Angr that we should drop the diacritics on omicron's and omega's representation.
• For Byzantine phonology, I'm not in the least bit worried about offending a few stodgy Greek scholars. /xt/ for κτ and /ks/ for ξ is totally reasonable. I don't think the diacritics (up-tacks and down-tacks) are a good idea.

Finally, I think using the acute mark over vowels to represent stress without determining syllable boundaries is an excellent idea, and I think we should implement it immediately. -Atelaes λάλει ἐμοί 00:10, 24 May 2012 (UTC)

Thank you for your responses, and it's good to have fresh opinions added to the mix. I have some notes to add.

• The reason I had been using /ʉ/ instead of /y/ for the Classical IPA is because upsilon was not a front vowel except in late Classical. It's telling that alpha was considered a front vowel while upsilon hadn't been. In using /ʉ/, it was more like saying /y/ is a valid allophone, sort of like how Swedish and Dutch prescribe front rounded vowels but their most common articulations are usually closer to central rounded—I know this is very true for Dutch. And, to be honest, I don't know exactly how to treat υι—it's often completely absent from vowel tables for historic Greek, though it is known to have existed (υἱύς, etc.). I'd always thought υι was a long monophthong of two coalesced vowels, as it was the only diphthong that had no long/short diphthong contrast (*ῠι/*ῡι), the way the others can be found (αι/ᾳ, ᾰυ/ᾱυ, ει/ῃ, ευ/ηυ, οι/ῳ, ου/ωυ). Until now, I've been using /ʉː/ for ῡ and /yː/ for υι—they had certainly merged by earliest Koine. If I was mistaken about any of this, please let me know.
• 1st BC Alexandrian Greek is well within the borders of Koine, and it was where the Septuagint was codified. That's what I'd already been using. We'd been using voiced plosives because they're plosive at the beginnings of words, with possible fricative allophones between vowels, not unlike modern Castilian Spanish. Aspirated consonants remained plosives at the time.
• 1st BC Alexandria still retained vowel length (ε, -αι = /ɛ/, αι = /ɛː/), though I understand it may have even then been a holdout in this respect. When it comes to vowels, it seems like the only meaningful difference between Alexandria and Popular Koine of the same time period, was preservation of vowel length. I sort of wish we could preserve the illustrative intermediate stages of phonetic change, where the early Koine monophthongs had long/short distinction, especially for such an influential Greek literary center as Alexandria.

Sorry, I was called away mid-response to help someone in their task, and I lost my train of thought. I'll also continue editing the new templates I've been designing—when they're done, they should be much easier to maintain and edit than the old 200+ templates which we'll no longer use. - Gilgamesh (talk) 16:56, 24 May 2012 (UTC)

I'm...done. Wow. I didn't expect to be done this fast. I haven't updated the documentation, but suffice it to say that the whole of Category:Ancient Greek IPA tokens is now obsolete. - Gilgamesh (talk) 21:17, 24 May 2012 (UTC)

That is an excellent redesign. Not only will it be easier to see at a glance and maintain, but it should also be easier on the system as well, not having to call as many separate templates. Nicely done. -Atelaes λάλει ἐμοί 00:56, 25 May 2012 (UTC)

Thank you very much. I am rather pleased with it. So what shall we do with Category:Ancient Greek IPA tokens? Much of the instructions can probably be salvaged, but the category's templates themselves have been completely abandoned. - Gilgamesh (talk) 04:43, 25 May 2012 (UTC)

If you're fairly sure that there's nothing left to be taken from them, I can delete them all, and then move the documentation to an appendix. Also, I should note that I do intend to respond to the above discussion, I'm still mulling it over a bit. -Atelaes λάλει ἐμοί 11:49, 25 May 2012 (UTC)

Alright, take your time. It's not like I'm entirely sure how to face revamping the documentation anyway. I was always much better with technical details than with trying to speak like a normal human being in layman's terms people can understand. - Gilgamesh (talk) 17:08, 25 May 2012 (UTC)

On tone marks and syllable breaks

Why does |o| generate "ō" with a long mark in Classical Greek? See κάπρος#Pronunciation for an example. —Angr 15:19, 28 May 2012 (UTC)

In the International Phonetic Alphabet, that doesn't represent length. It represents normal (middle) pitch, neither raised nor lowered. Classical Greek had a pitch accent. To demonstrate:

• /ő/ = extra-high pitch
• /ó/ = high pitch
• /ō/ = normal pitch
• /ò/ = low pitch
• /ȍ/ = extra-low pitch

I marked every normal vowel and /ː/ symbol with the macron when the pitch is normal. Exact location of pitch is critical in distinguishing the difference between polytonic acute and polytonic tilde accents—the difference in tone affects only one of the morae, so only a portion of a non-short vowel. For /aː/, acute accent is /āː́/, grave accent is /āː̀/, and tilde accent is /áː̄/. In modern browsers and IPA fonts, the diacritics align (and if necessary, stack) elegantly. Koine Greek and later forms lost their pitch accent, and instead developed a stress accent. In IPA, this can be marked /ˈo/ or, when pitch is not an issue, /ó/. We're using the latter so we don't have to always arbitrarily guess ˈsyl.la.ble ˈboun.da.ries in Ancient Greek IPA. For pitch accent, would you rather I use the tone contour symbols for every vowel? /a˧ː˧/, /a˧ː˥/, /a˧ː˩/ and /a˥ː˧/, etc. - Gilgamesh (talk) 22:12, 28 May 2012 (UTC)

I don't think the redesign with a large switch statement is an improvement, especially not when it comes to speed. From experience I know that switches become slower the more possibilities they contain. I remember when {{poscatboiler}} used a switch rather than separate templates, it would take around half a minute for some of the pages to load. For separate templates there is no such slowdown, looking up pages is much more scalable. —CodeCat 22:13, 28 May 2012 (UTC)

At least these templates have a fairly limited role in articles. Having 200+ templates for pronunciation tokens was an enormous impediment to maintenance — even for me, who wrote the 200+ templates. Having half a dozen or so templates is more manageable from a human editor perspective. - Gilgamesh (talk) 22:17, 28 May 2012 (UTC)

I've noticed no slow-down on page-load with the new format, so I suspect that, if there is any, it's irrelevant. It most certainly is easier to maintain. -Atelaes λάλει ἐμοί 00:48, 29 May 2012 (UTC)

I'm still not comfortable describing the Classical pronunciation of υ as /y/. It was not a front vowel until later in the Classical period, right? And it wasn't /u/, which ου would creep into later. I'm not really comfortable describing it as anything other than /ʉ/, which is neither front nor back. - Gilgamesh (talk) 06:39, 29 May 2012 (UTC)

I'm curious where you're getting the info that it wasn't a front vowel. Did some Classical grammarian state that? Every scholar I've ever read has described upsilon with /y/. -Atelaes λάλει ἐμοί 11:38, 29 May 2012 (UTC)

Well, Attic is associated with its particular flavor of front-vowel triggered mutations of... ...labialized consonants. Never mind, got my time periods a tad mixed up. - Gilgamesh (talk) 14:50, 29 May 2012 (UTC)

I would prefer it if what you're calling "normal pitch" were left unmarked, and not just because people more familiar with classical philology than with phonetics will interpret "ō" as "long o" rather than "short o with 'normal pitch'". The difference between pitch-accent languages and tone languages is that in pitch-accent languages only one syllable per word bears a distinctive pitch; the other syllables carry a default pitch. This is much clearer if we leave the syllables without pitch accent unmarked, as I believe we already do for other pitch-accent languages like Serbo-Croatian and Swedish. Also, I've never seen anyone put a tone mark on top of the long mark in IPA. The usual ways of showing a contour tone on a long vowel are (for example) [âː] and [áà]. You seem to be suggesting that the acute accent is a rising accent on long vowels and a flat high accent on short vowels, while the grave accent is a low-falling accent on long vowels and a flat low accent on short vowels. Are we sure about that? If short έ and ὲ are [é] and [è] respectively, couldn't ή and ὴ be [ɛ́ɛ́] ([ɛ́ː]) and [ɛ̀ɛ̀] ([ɛ̀ː])? Then we could transcribe ῆ as [ɛ́ɛ̀] or [ɛ̂ː] and leave unaccented syllables unmarked. That would make the transcriptions easier to read and less misleading. —Angr 21:10, 29 May 2012 (UTC)

I can remove the extra macrons as a matter of simplicity, but absolute mid pitch is what macrons are used for in IPA, no matter how counterintuitive that might seem. It's sort of like how IPA unexpected uses the diaeresis to mark a vowel as more centered. Circumflex and caron have specific meanings too. See Wikipedia:International Phonetic Alphabet#Diacritics and Wikipedia:International Phonetic Alphabet#Suprasegmentals. /ȁ à ā á a̋/ are absolute pitch distinctions relative to the entire IPA sample. /ǎ â/ (global rise and global fall respectively) are only relative to the previous pitch. We've been using /à ā á/ instead of /ǎ â/, because the unaccented vowel tokens are completely agnostic about relative nearby pitch, but using /ǎ â/ would not make this possible, such that πόλις = /pǒlîs/). But using absolute pitch marks, πόλις = /pólīs/ or, as you may prefer, /pólis/. And although it may be strange to put a diacritic over /ː/, it's not impossible if /ː/ is an elongation of the previous vowel, such that /aː/ is essentially equivalent to /a͡a/ as a diphthong of itself. But I opted not to use tie markers in the redesign because it makes pitch diacritics more complicated to place...ultimately I did away with the ties and just use /./ to separate immediately adjacent vowel syllables.

In the Ancient Greek pitch accent, as I understand it, the pitch change only happens on one vowel mora and only lasts for one vowel mora, and then immediately reverts for the rest of the vowel and for all remaining vowels. For acute or grave pitch, it only falls on the final vowel mora of the syllable, and none of the vowel morae before it in the same syllable. With tilde pitch, it only falls on the penultimate vowel mora of the syllable, and none before the penultimate and none on the final. For one mora vowels, there can only be /ā á à/. For vowels of two or more morae, the possibilities are /āā āá āà áā/. The combination /áà/ cannot work as you may expect, because IPA strictly defines /à/ as an absolute low, and not a relative drop (which is /â/). - Gilgamesh (talk) 00:25, 30 May 2012 (UTC)

You both make some good points. I suggest we limit ourselves to the macron as an indicator of normal pitch to long, accented vowels, so that we can show which mora is changed in pitch and which isn't. All vowels which have no accent would then be shown without any diacritics. Also, as excited as I initially was about the idea, I'm starting to wonder if the switch to acute accents as indicators of stress accent was a good idea. At least with the previous system, the fact that something was fundamentally different about Classical vs. Koine/Byzantine accent was clear. Now it isn't. -Atelaes λάλει ἐμοί 00:53, 30 May 2012 (UTC)

It would make more sense to me for the pitch profiles to be more like the long accents in Lithuanian: one rises quickly to a peak in the first mora, then falls slowly to the normal pitch by the end of the second mora, while the other is a mirror image: it slowly rises over the first mora to a peak on the second mora, then drops rapidly back to normal. The greatest pitch change is on the mora with the peak, but the other mora is involved as well. Chuck Entz (talk) 05:27, 30 May 2012 (UTC)

As usual with the phonology of ancient languages, we know approximately what's going on phonologically, but we really don't know precisely how things sounded phonetically. For one thing, how sure are we that the grave accent really represents a low pitch, and not simply the absence of an high pitch on the final mora of a word? For another, it's dangerous to simply assume that "normal pitch" is mid pitch, which is what the macron stands for in IPA – we don't actually know what the "normal" pitch was in an absolute sense. An advantage to using orthographic gemination rather than the long mark to indicate long vowels is that we can then treat them the same way as diphthongs: we can transcribe ή and ῆ as [ɛɛ́] and [ɛ́ɛ] just as we transcribe αί and αῖ as [aí] and [ái]. —Angr 06:38, 30 May 2012 (UTC)

Alright. I'll address points one by one.
For Koine and Byzantine, should we go back to using /ˈ/ immediately before the vowels (/pˈolis/)? Otherwise we'd have to start guessing the syllable margins for every word, which...not only would take an enormous amount of editing work (thousands of entries now), but is also something not easy to objectively determine when the accent is not on the first syllable and falls on a consonant cluster.
I have no inherent problem with IPA orthographic gemination. One issue I can see is whether to extend this to the Koine row, where vowel length itself had started to weaken. 1st BC Alexandria retained it, but not everyone did, and it was certainly lost pretty much everywhere by the end of the Koine period. I'd be uncomfortable not marking it when it was retained in such an influential place as Alexandria.
And if vowels are orthographically geminated in our IPA, should consonants be too? It might look strange to have one and not the other.
If we decide on orthographic gemination, we can decrease the ambiguity by putting asyllabic diacritics under all but one of a long/diphthong's vowels (/ɛɛ̯/, /aj/); since the diacritic goes below, it cannot potentially interfere with pitch diacritics that go above (/ɛɛ̯́/, /ɛ́ɛ̯/). I say /aj/ instead of /ai̯/ because the IPA specifies cardinal approximant symbols /j ɰ w ʕ/ etc. as being absolutely equivalent to their corresponding semivowels /i̯ ɯ̯ u̯ ɑ̯/ etc., and this is also the convention I use in the Marshallese templates I wrote. - Gilgamesh (talk) 19:21, 30 May 2012 (UTC)

I looked at some of my experimental changes to the template (not saved), and some of these changes deeply anger my mental centers of logic and neatness. For instance, the token eoo (coalesced εω after a stressed syllable as in Βριάρεως) is looking like /eɔ̯ɔ̯/. Instant headache just looking at that. Are we at all certain how this highly unusual diphthong is pronounced? - Gilgamesh (talk) 20:36, 30 May 2012 (UTC)

Hmm... maybe I have to retract my claim above that all diphthongs in grc are falling. I hadn't thought about cases like this. I suspect /e̯ɔɔ/ is far more likely. If so, we can still dispense with the nonsyllabicity mark for falling diphthongs by convention and use it only on the comparatively rare rising diphthongs. —Angr 21:10, 30 May 2012 (UTC)

Are we certain either way that it's a rising or falling diphthong? All we know about it is that it was a way to squeeze the two vowels into one syllable to preserve Classical Greek meter. - Gilgamesh (talk) 21:21, 30 May 2012 (UTC)

Well, "two vowels squeezed into one syllable" is pretty much the definition of a diphthong, isn't it? Is it possible for a diphthong to be neither rising nor falling? —Angr 21:40, 30 May 2012 (UTC)

Interesting you ask that question. Smoothly-transitioning diphthongs exist in the Marshallese language, whose Wiktionary templates I wrote. It's a vertical vowel system with only four vowels, but 36 vowel allophones, of which 27 are diphthongs. Laboratory analysis has confirmed that they smoothly transition from one vowel to the other, since frontness and roundedness are not phonemic. So yes, it's possible for diphthongs to be neither raising nor falling. Whether this is true in Classical Attic, I can't be certain. - Gilgamesh (talk) 22:46, 30 May 2012 (UTC)

I would rather abandon the whole system and enter IPA manually (as is the usual practice for almost all other languages at Wiktionary) than go back to the very distracting transcriptions like /pˈolis/, which make it look like /p/ is a syllable on its own. Since Koine isn't showing pitch accents anyway, the issue is moot there and we can use the ː mark for long vowels. For consonants, I have a preference (orthographic gemination), but no strong opinion one way or the other. Since all diphthongs in Greek are falling, there is no ambiguity in transcriptions like /ai/ and we don't have to indicate nonsyllabicity--especially if we're marking vowel hiatus with "." anyway. The transcriptions are easier to read the fewer diacritics we use. —Angr 20:32, 30 May 2012 (UTC)

Abandon the system and use manual IPA...? About half a decade ago when I first started browsing the ancient Greek entries that were here, there were atrociously inaccurate pronunciation guides. I mean, not just a little "off", but hugely off-base. And pronunciations were haphazard across multiple entries. It was...an intolerable state of things. Neatness and consistency are demanded. - Gilgamesh (talk) 20:40, 30 May 2012 (UTC)

Finding syllable boundaries in Ancient Greek isn't much of a problem, since it kindly provides us with enormous amounts of poetry in which the distinction between consonant clusters that do and don't "make position" is made very clear. Any cluster that fails to make position can be treated as an onset cluster (e.g. πατρός [pa.trós]); all others are split up by the syllable boundary. It might be difficult to program the template to find those clusters and behave accordingly, but it isn't actually difficult for humans to know where the syllable boundaries go. But "instant headache just looking at that" is exactly how I react to things like /pˈolis/. —Angr 21:10, 30 May 2012 (UTC)

I've been thinking of an idea that I've been avoiding until now, because it would necessitate the editing of thousands of articles. A (') token, which would be invisible in the Classical row, but would appear as /ˈ/ in the Koine and Byzantine rows. In turn, the Classical pitch diacritics would completely disappear from Koine and Byzantine rows. This could also retire the many vowel accent tokens for one-syllable words, such as o(') returnint back to o' in the tokens. So πόλις, encoded (')|p|o'|l|i|s would appear as /pólis/ in Classical, but as /ˈpolis/ in Koine. Once again, this may necessitate the editing of thousands of articles, since it's the introduction of a fundamental new token. Also, the token doesn't have to be ('). We could use " or ' instead if that's simpler on the eyes, for "|p|o'|l|i|s or '|p|o'|l|i|s. - Gilgamesh (talk) 21:17, 30 May 2012 (UTC)

That sounds like it would have a good final result, but it would of course be an enormous amount of work adding the additional symbol to the existing words, and it's not really something we could get a bot to do, is it? —Angr 21:40, 30 May 2012 (UTC)

It's theoretically doable with an algorithm written from scratch that makes modifications to the embedded template arguments. But I've never even written a bot, much less petitioned to have one allowed to edit Wiktionary. So you are correct — the only reasonable option we have at this moment is manual labor. Such would also be required for a wholesale switch to manual IPA. In the interests of making the wiki more maintainable (not less), I'd be pressed to keep the templates. - Gilgamesh (talk) 22:46, 30 May 2012 (UTC)

The more I examine the Classic IPA system, the more I am comfortable with it. I mean, we can get rid of the macrons, but otherwise...it seems okay.
But if it actually came down to dismantling the template system and going manual...I do have experience with writing input-method-oriented JavaScript apps that generate converted text as you type. Here is an older example I wrote, for Korean and Hangeul. May or may not work with IE, though I've been getting better about that too. I'd once already thought of what an IME-oriented ancient Greek would be like. For example, Ἰξίων might be something like i x ii' oo n. Or, in an IME format that doesn't use spaces, perhaps ixii'oon or ixii'hoon (h between vowels cannot become a consonant, but may serve an effective syllable break). But really...I hope it never comes to us having to use such a beast...especially if we end up debating once again over the formatting that will have been manually edited into thousands of articles with IPA generated by such a program after the fact. - Gilgamesh (talk) 21:03, 30 May 2012 (UTC)

I just removed the macrons from the template, but otherwise changed nothing. - Gilgamesh (talk) 21:10, 30 May 2012 (UTC)

Thank you. —Angr 21:40, 30 May 2012 (UTC)

I have another idea. It's technically ad hoc, but then again, so is using acute marks for stress accent; the latter merely enjoys widespread legitimate ad hoc precedent. For now at least, we can use the symbol /a̍ e̍ i̍ o̍ u̍ y̍/ above a vowel. It resembles the /ˈ/ symbol, and by being directly above (instead of to the side), it marks the vowel as the syllable nucleus and doesn't concern itself with consonant syllable boundaries. It's a diacritic that the IPA currently doesn't use, and it also has the advantage of resembling the purely vertical stress diacritic in modern monotonic Greek. In a Koine row, it would look like /po̍lis/. - Gilgamesh (talk) 23:14, 30 May 2012 (UTC)

It just occurred to me that, since IPA diacritics can be optionally placed above or below a letter in case the letter has ascenders/descenders, that the diacritic I recommended could be seen as an alternative presentation of /z̩/ (the z is just for demonstration), which syllabifies a consonant. But when used for a vowel, there is no such need...I might consider this a reasonable ad hoc anyway. - Gilgamesh (talk) 23:23, 30 May 2012 (UTC)

Or, we could just use acute mark for non-pitch stress, like we had been doing. Atelaes, I know you aren't entirely comfortable with the distinction being unmarked between the rows, but do remember that the acute mark of polytonic Greek is widely used as the stress accent mark in modern monotonic Greek. Not all typefaces use a purely vertical mark. Tahoma uses a vertical mark:

πόλις

But Palatino Linotype uses an acute-slanted mark:

πόλις

It is true that the Unicode space for polytonic Greek supplements provides a second acute-slanted mark, here shown in Tahoma:

πόλις

But the Unicode standard considers the two shapes absolutely equivalent and interchangeable, with the supplements being a mere glyph variant of the default Greek counterparts. This is evidenced by how in the third example I had to use an HTML Unicode escape sequence to circumvent MediaWiki's automatic canonicalization of edit text. - Gilgamesh (talk) 23:37, 30 May 2012 (UTC)

Exact time periods

I renamed the vague "Classical", "Koine" and "Byzantine" more specifically to "5th BC Attic", "1st BC Egyptian" and "10th AD Byzantine", with links to Wikipedia articles that document their phonologies. I really wished for a Byzantine phonology from before 600 AD (the cutoff of Late Antiquity), but technically the entire Byzantine period from the Eastern Roman Empire to its fall in 1453 is collectively considered "Medieval Greek", with changes happening gradually instead of abruptly. It's one of the few ancient states that retained its infrastructure well into the Middle Ages, without having gone through the Dark Ages much of Europe went through. ...I'm not going to lose too much sleep over documenting 10th AD pronunciation practices considering they still had a strong tradition of archaicist (Atticist) writing — polytonic orthography flourished even though it had long since lost pronunciation distinctions. Now, if there are other well-documented phonologies associated with time periods and places, we could add additional rows. I occasionally hear clamoring for Late Koine. - Gilgamesh (talk) 06:09, 31 May 2012 (UTC)

There's certainly an interest in 1st Century AD Palestine, for obvious, non-linguistic reasons, though I don't know if we have anything meaningful to say about it. People will want to know what pronunciation is correct for the New Testament, so it can be read aloud with some authenticity. Of course, that leads us down the textual criticism rabbithole in order to try and nail down time and place... Chuck Entz (talk) 06:48, 31 May 2012 (UTC)

Yes, obviously. Wikipedia:Koine Greek phonology is not exact on that time period, but implies that the time between 1st BC and 4th AD was when the aspirated consonants θ φ χ were transitioning to fricatives. First φ, and then θ and χ. But evidence seems to vary (Jewish inscriptions vs. Pompeian inscriptions). It seems that transitions happened differently in different places, and in fact branched off in an entirely different direction in the Pontus region at the time. But the East Mediterranean (excluding Pontus) had a generally common pronunciation by the time of the Byzantine Empire. Anyway, I'm mainly going with the established milestones, which are pretty well documented. - Gilgamesh (talk) 07:06, 31 May 2012 (UTC)

Posessive Adjectives

It doesn't look like we have ἐμός, ἡμέτερος, σός, or ὑμέτερος. They look to me like possessive adjectives, but many grammars treat them as pronouns (to be used with nouns). Which POS header do we want to use? Chuck Entz (talk) 17:23, 28 May 2012 (UTC)

I'd call them determiners. —Angr 20:04, 28 May 2012 (UTC)

Determiner sounds right to me as well. Unfortunately, my lack of knowledge of Ancient Greek grammar prevents me from rigorously testing it with respect to w:Determiner_(linguistics)#Differences_from_pronouns, but it seems correct. -Atelaes λάλει ἐμοί 00:46, 29 May 2012 (UTC)

Sounds like enough of a consensus for me. It's only a few entries in a small category, so it would be very easy to change if this proves wrong. Chuck Entz (talk) 01:42, 29 May 2012 (UTC)

Done. I'm sure, as always, that I'm overlooking something: it seems like every time I nail down N details, N+1 is something any idiot would have taken care of in the first 5 minutes... Chuck Entz (talk) 03:46, 29 May 2012 (UTC)

Template:grc-ipa-rows; Inclusion of Standard Modern Greek pronounciation

I proposed to user Gilgamesh we include another row for a Standard Modern Greek pronounciation of Ancient Greek words. He did it experimentally and asked me to contribute here. So I propose we keep this change because:

• Many words are exactly the same morphologically throughout the history of Greek, i.e. in all the phases of the languages from the 5th century BC until now. However, as Wiktionary separates the entries between Ancient and Modern Greek according to whether they are written using the polytonic orthography or not, we cannot have unified entries for the moment.

• The language is the same albeit in different periods. As a result there are many instances where ancient words, now obsolete, can be heard in every day speech, for example by educated people. Such words may not exist in a Modern Greek dictionary but may be heard even today. A Greek speaker does not hesitate today to use ancient words in certain occasions of his speech to the extent that an foreign speaker would do.

• I do not mean that everyone uses freely Homeric words in his everyday speech (!) but when this happens the pronounciation is definite across Greece and Cyprus.

• In schools and universities in both Greece and Cyprus only one accent is used when reading Ancient, Koine and Byzantine Greek texts, that is the pronounciation of Standard Modern Greek. Moreover, such ancient words that have not been integrated in everyday speech or have not evolved through the centuries are written in one specified way which excludes many possible pronounciations. For example ⟨κτ⟩ in ancient words is always pronounced as /kt/ in Greece and Cyprus even though /kt/ has evolved into /xt/ in other words. But when /kt/ has evolved into /xt/ it is also usually written as ⟨χτ⟩. The way an ancient word is written does not allow for many different pronounciations.

• I find no reason why we include a 10th AD Byzantine pronounciation for Homeric words such as πολύτροπος which ceased to exist very early and were not used in either Koine or Byzantine Greek and we do not include a Standard Modern Greek pronounciation. Although such words have ceased to be used in everyday speech for millennia they are used for example when reciting the Odyssey. And since people are nowadays more educated than people in the 10th c. AD, I am sure the word πολύτροπος is now used more widely than during the medieval period.

• Finally, there are few differences between the 10th AD Byzantine pronounciation and the Modern Greek one. Perhaps the greatest difference is the pronounciation of ⟨οι⟩ and ⟨υ⟩ which both changed from /y/ to /i/ already in the 11th and 12th cc. AD.

Dimboukas (talk) 13:45, 2 June 2012 (UTC)

I am opposed to the inclusion of Modern Greek in {{grc-ipa-rows}} for the simple reason that, like it or not, Wiktionary treats Modern Greek as a different language from Ancient Greek. In many cases, the words are spelled differently (because of the loss of the grave and circumflex accents as well as the breathing marks), and even in the cases where the words are spelled the same, we have separate language entries for "Greek" (i.e. Modern Greek) and "Ancient Greek". The place for the Modern Greek pronunciation is under the Modern Greek header, not the Ancient Greek header. —Angr 14:03, 2 June 2012 (UTC)

Did people in the 10th century AD speak Ancient Greek? However we include a medieval pronounciation for Homeric words. The language of the 10th century is closer to the modern one that to the language of the 5th century BC. Moreover, the fact that Wiktionary treats Modern Greek as different language from Ancient Greek isn't such a huge obstacle to include another row. The Modern Greek language has used diacritics for longer than it hasn't. Dimboukas (talk) 14:19, 2 June 2012 (UTC)

Whether you want to call 10th-century Greek "Ancient Greek" is a separate issue. The ISO 639-3 code "grc" is defined as "Ancient Greek (to 1453)", while the ISO 639-1 code "el" is defined as "Modern Greek (1453-)", and that's what we follow on Wiktionary. We could have Modern Greek entries that use diacritics, but they would still be under a ==Greek== (i.e. Modern Greek) header, not under an ==Ancient Greek== header. The ==Ancient Greek== header shouldn't have any pronunciations under it that are later than 1453. Now if we want to change the row for "10th AD Byzantine" to "14th AD Byzantine", or add a separate row for "14th AD Byzantine" in addition to "10th AD Byzantine" I wouldn't oppose. But including a row for "21st AD standard" under the ==Ancient Greek== header would be like including Modern English pronunciations for Old English and Middle English words. —Angr 14:34, 2 June 2012 (UTC)

You are rendering the subject too technical. I understand what you want to say but an inclusion of the contemporary pronounciation wouldn't bother. However we could consider to add a 14th AD Byzantine row instead of a 21st Standard. After all this coincides with what I said that after the 12th century Greek pronounciation has not changed especially for ancient and literary words. Dimboukas (talk) 15:01, 2 June 2012 (UTC)

This is a familiar debate. Thought I'd test the waters again after Dimboukas spoke to be initially. If it comes to it, the 21st AD Standard row can be removed. - Gilgamesh (talk) 19:20, 2 June 2012 (UTC)

I've opposed this in the past, and I continue to oppose it, for two reasons. First, as Angr rightfully notes, modern Greek is considered a separate language on Wiktionary; there, and only there, does modern pronunciation belong. Secondly, I think that a pronunciation scheme based solely on spelling is only appropriate for a dead language, and absolutely inappropriate for a living language. -Atelaes λάλει ἐμοί 22:54, 2 June 2012 (UTC)

Hm, now that you say it, these are familiar points. I'm beginning to remember bits and pieces of why I removed it the first time, years ago. Pronunciation schemes based solely on spelling is a rather compelling argument...atticism was alive and well in the literary tradition of the Byzantine Empire, but the modern Greek approach uses a prescribed pronunciation. Even those philhellene Ottoman sultans who enjoyed Greek studies were probably not studying it as their own unbroken tradition, but as a foreign dead language of someone else's conquered empire (unless they happened to be islamized Greeks who still preserved atticism in a state where Arabic, Persian and Turkish had become more influential...which may not have been likely in context). - Gilgamesh (talk) 00:58, 3 June 2012 (UTC)

You know, 21st AD may not have been the best idea, but what about 15th AD? What information exists on the state of phonology just before the fall of Constantinople in 1453? If we can find references on that, we can have a row for that. I know that by that it was always becoming more diverse than the relatively united form it was in 10th AD (barring Pontic and Griko). The Byzantine Empire was shrinking rapidly, having only Constantinople and Mystras left at the end. But the phonology of Constantinople at that time would certainly be interesting. - Gilgamesh (talk) 23:58, 4 June 2012 (UTC)

This is a good idea since as Angr said the iso code grc is up to 1453. The Greek pronounciation may have become more diverse by then but for sure, as I have said before, [y] of ⟨υ⟩ and ⟨οι⟩ became [i] too.Dimboukas (talk) 18:48, 5 June 2012 (UTC)

Already done. The dialect of Constantinople is relatively well known at that time, because it's among the most similar to Modern Greek as it was spoken in the core of the Greek world before the advent of Katharevousa. Now my largest concern is vowels...the period had a lot of diphthongization of adjacent vowels. By the end of the 10th century, all /e/ before a vowel that wasn't /i/, became /i/. And in the 13th century, all /i/ (not after a consonant cluster ending with /r/) became /j/ and surrendered any accent it may have had to the next vowel. However, unaccented forms of the second declension endings -(i)ος and -(i)ον became /is/ and /in/. This was an active phonotactical process (reflected by atticist misspellings or by non-atticist written dialect) until reversed by 19th century Katharevousa-era prescribed pronunciations. By the end of the Byzantine period, these processes that make it relatively more impractical to represent pronunciation through tokens in continuity with Classical Attic. I think I know of some ways to adapt tokens to this...but the question is, do we really want to? - Gilgamesh (talk) 20:14, 5 June 2012 (UTC)

⟨ις/ης⟩ and ⟨ιν⟩ instead of ⟨ιος⟩ and ⟨ιον⟩ respectively, had appeared long before in Koine. For example we find Αντωνις and παιδιν instead of Αντωνιος and παιδιον already in the 1st century AD. Moreover, I don't think we have to adapt tokens to such rules. The entries are ancient Greek; let me use an example: we cannot find the ancient word μηλέα ([miˈlea], "apple tree") as μηλιά [miˈʎa]. What I mean is, that although the pronounciation of many words had changed, here we have to do with ancient words and μηλιά cannot be thought as an ancient word. Μηλέα, on the other hand, would have been pronounced as [miˈlea] when used in either trully Attic or Atticist texts. The same about ἀλήθεια ([aˈliθia] instead of [aˈliθça]) and many other words. Dimboukas (talk) 21:49, 5 June 2012 (UTC)

I can understand the is/in issue, but what I'm saying is that sound changes weren't necessarily a matter of "right" vs. "wrong"—phonotactics themselves changed, where the speaker naturally pronounces a certain combination of sounds a certain way, and individually spelt consonants and vowel sequences don't exist in a vacuum. The vowel interactions happened initially through diphthongization of adjacent vowels. /e.a/ became /e͡a/, then that was no longer stable and became /i͡a~ja/.

We see something similar in the development of English too, where Early Modern English had separate diphthongs for /ɪu ɛʊ/ (new, dew), but /ɛʊ/ became unstable and merged with /ɪu/, which itself later changed to /juː/. Ask modern speakers to try to pronounce all these words this way, even in historical context, and it feels phonotactically unnatural—as one of my friends lightheartedly said, "it makes my mouth angry". This all happened in English with the merger of /eː ei/ (pane, pain) as /eɪ/, and of /oː ou/ (toe, tow) as /oʊ/, and or /ɹ wɹ/ (right, wright) as /ɹʷ/, and the monophthongalization of /ɒʊ/ (law) to /ɔː/, and so forth. Changing accent through speech therapy is rarely done, except with learnt prestige accents like Received Pronunciation in the United Kingdom, which is itself not even as well-respected as certain natural regional accents like Yorkshire.

With a language's spoken sound changes also comes a change in phonotactics that pushes things along. It seems unlikely that medieval Greek speakers would have such fundamental pronunciation changes without some phonotactical ruleset making the previous pronunciations feel unnatural and uncomfortable. The Modern Standard Greek approach of reseparating adjacent phonemes in pronunciation started with Katharevousa. It was when new phonotactics were prescribed to replace natural ones, that continuity between written Attic and the way it was pronounced was truly broken.

There are plenty of other languages that retain very old phonetic writing systems that haven't changed, even when their pronunciation has. Written Tibetan is heavy and consonant clusters and otherwise does not mark any tones, but the spoken Lhasa dialect drastically simplifies most of the clusters and is tonal—and yet Tibetan continues to be written and adapt new words as needed using its ancient phonetic writing.

In the 19th century when Greece became independent again, Western scholars pointed out that Attic Greek was pronounced very differently in the distant past than it had come to be pronounced, and this was considered a foreign concept in Greece at the time—words were pronounced with traditional phonotactics people had picked up naturally from their parents. And from the end of the Byzantine period to before the pronunciation reforms of the 19th century, pronunciation sequences like /vm vn ŋɣ ŋk ɣm ɣn ŋx kp kt ɱv mp ɱf nð nθ ns nt sθ sf sx pt fθ fs xθ/ really didn't exist in most of the Greek world, as matter of what was phonotactically stable in speech. Though I admit I don't know enough about sequences like /ðm ðn θn nz fn xn/ etc. and what their status was. - Gilgamesh (talk) 00:47, 6 June 2012 (UTC)

I think these apply to words whose pronounciation changed naturally throughout the ages. However, in the 15th AD words such as πλάγχθη (Od.1.2.) would have been pronounced as [ˈplaŋxθi] and not as [ˈplaŋxti] or [ˈplaxti] for example. Such words have not undergone phonetic changes. Moreover, when a Greek speaker of 1450 came across words such as ἐννέα he wouldn't pronounce them automatically as [eˈɲa] but as [eˈnea]. Dimboukas (talk) 13:18, 6 June 2012 (UTC)

Nor would words, such as Εὐμενίς, be pronounced as [emeˈnis]. These sounds may have disappeard in every day speech but the words that contained them either changed too or disappeared. Moreover, if you look at the excerpt at w:Digenes Akritas (Escorial Digenes where the manuscript is of the 15th AD) you will see ευθύς. If ευθύς were pronounced as [eˈftis] it would be written as ευτύς too. But this last argument may disorientate again from my general idea, that εὐθύς of Thucydides wouldn't be pronounced as [eˈftis]. Dimboukas (talk) 00:51, 7 June 2012 (UTC)

I'm not saying people didn't understand that φ is /f/ and θ is /θ/, but that φθ as a sequence defaulted to /ft/ because of phonotactics, and there was no real effort to widely reverse this until the 19th century. Atticist spellings were largely frozen, and phonotactics are revealed by misspellings betraying the word's pronunciation at the time. Some sounds, within a given language region, are just uncomfortable to pronounce in certain positions and/or in certain combinations, even if people technically have the ability to do so in phonetic isolation. It takes actual active speech therapy to train yourself to pronounce sounds that fall outside of normal phonotactics. It seems unlikely that a common recitation of text would try so painstakingly to do so. Phonologies are made up of what is common, not what is specifically meticulously learnt and polished. - Gilgamesh (talk) 04:05, 7 June 2012 (UTC)

I know that there was some degree of written diglossia at the time, with a difference between texts written in the dialects, and texts written in atticist tradition. But the same processes that were causing isolated sounds to change (υ /y > i/ were also affecting the stable pronunciations of sounds in sequence (ξ /xs > ks/, ντ /nt > nd/, πτ /pt > ft/). It seems very illogical that some changes like /nt > nd/ would be kept (when it caused a merger with νδ at the time, which was itself only changed to /nð/ in the 19th century), but other atticist spellings would always be diligently recited as a stable string of the same phonemes they would be in isolation. If atticist spellings were pronounced as you say, there wouldn't be so much evidence of their underlying pronunciations from misspellings. I think people often tried to make a good effort at recitation, but would not routinely unnaturally articulate pronunciations that were already so thoroughly lost in common life. - Gilgamesh (talk) 04:26, 7 June 2012 (UTC)

Grave on articles in inflection templates

Templates like {{grc-decl-blank-M-full}} have acutes over the article where graves should normally be. Is this an oversight or am I missing something? --Vasiliy Faronov (talk) 21:21, 7 September 2012 (UTC)

We don't have entries using the grave accents because they're more of a co-variant with the acute used in contexts other than in isolation, which is the context that describes dictionary headwords. Chuck Entz (talk) 22:49, 7 September 2012 (UTC)

I’m not talking about headwords, I mean the templated definite articles that precede the inflected forms in the table. For example, in χείρ this gives τήν χεῖρα rather than the expected τὴν χεῖρα. Sorry I wasn’t clear. --Vasiliy Faronov (talk) 07:58, 8 September 2012 (UTC)

Yes, that's a mistake and should be corrected. I'd also support adding ὦ before the vocatives. —Angr 11:35, 8 September 2012 (UTC)

I fixed the accent in all the grc-decl-blank- templates. As for ὦ, it’s your call. --Vasiliy Faronov (talk) 12:00, 8 September 2012 (UTC)

Transliteration of iota subscripts

This should be addressed, even if it's to say "ignore them". Chuck Entz (talk) 05:06, 15 September 2012 (UTC)

Well, "ignore them" is our current de facto policy, right? Are there any counterexamples? (By the way, thanks for alphabetizing the digamma.) --Μετάknowledge^{discuss/deeds} 06:33, 15 September 2012 (UTC)

"It goes without saying" only works for those who have taken the time to look at comparable entries. If everyone knew the de facto policies, we wouldn't need an About page. I became aware of the issue via a revert by an IP not that long ago (I don't remember if it was adding or removing the "i" from the transliteration), so not everyone is on the same page on this, yet. Chuck Entz (talk) 08:30, 15 September 2012 (UTC)

I'm not saying that we should rely on "it goes without saying" — I think we should use current practice to determine policy. --Μετάknowledge^{discuss/deeds} 15:23, 15 September 2012 (UTC)

I never ignore them, but they're also never ambiguous in context. They can only appear under one of ᾳ ῃ ῳ, and I always transcribe them āi ēi ōi. For ᾱϊ ηϊ ωϊ (or what was later simply written ηι ωι), I transcribe āï ēï ōï, etc. See, they became written later as subscripts because the iota had become silent by the time Classical Greek became Koine Greek. But earlier in Classical Greek, they were written out, and formed part of very real long-diphthongs. So no, I would never omit them in romanization of Classical Greek. - Gilgamesh (talk) 00:55, 8 November 2012 (UTC)

Accents in transliteration

Is there any reason why the accent is not indicated in the transliteration? It seems to me that it is rather essential to the word, and definitely useful. —CodeCat 23:08, 8 October 2012 (UTC)

ῆ = ? Chuck Entz (talk) 04:40, 9 October 2012 (UTC)

ή = ḗ, ῆ = ế (?) ... I think polytonic Greek was not meant for transliteration with accents. If anyone really cares, they'll learn the script (it's pretty easy, even compared to other European alphabets). —Μετάknowledge^{discuss/deeds} 05:01, 9 October 2012 (UTC)

I'd transliterate ῆ as an e + ¯ + ^. So Metaknowledge got it right I think. In any case, what I would prefer is if the accents were at least allowed in transliteration, rather than not allowed by policy as this page currently describes. The argument that someone should learn the script if they want to know is really moot. After all, if people knew the script they wouldn't need transliterations; the idea of a transliteration is to explain the spelling to someone who doesn't know the script. To me, that means it should reflect all the significant details of the original spelling, which includes accents and (as mentioned above) iota subscripts. —CodeCat 12:36, 16 October 2012 (UTC)

Since ε and ο never take the circumflex anyway, we can transliterate ῆ and ῶ simply as ê and ô, saving the double-diacritics for ή, ώ, ὴ, and ὼ. If we want to transliterate iotas subscript, I would recommend using the underdot diacritic (ᾳ ῃ ῳ = ạ ẹ ọ). Again the macron is unnecessary since ε and ο never take the iota subscript. —Angr 15:49, 16 October 2012 (UTC)

Couldn't the iota be written with a following i? Or are there minimal pairs that are distinguished this way? —CodeCat 16:23, 16 October 2012 (UTC)

Well, there's a difference between αι and ᾳ. We could distinguish them as ai and āi, but I think ai and ạ is more intuitive. —Angr 16:34, 16 October 2012 (UTC)

Is it intuitive to people who can't read Greek though, and who have no understanding of what a iota subscript is or why it ought to be written below the letter? Look at it from someone who is rather clueless about Greek... they would treat something like ạ as just a funny looking a, since they do not understand the significance of the mark below it. On the other hand, āi is clearer. I realise that this also applies to other marks like macrons and circumflexes, so it's not a watertight argument. But in this case I think āi simply represents it more exactly than ạ, although ạ might resemble the Greek more graphically. —CodeCat 16:40, 16 October 2012 (UTC)

η=ē ή=ḗ ὴ=ḕ ῆ=ê ῃ=ēi ῄ=ēí ῂ=ēì ῇ=ēî, ηι/ηϊ=ēï, etc. It's pretty trivial. For other situations like ᾱ́=ā́, if absolutely necessary, combining diacritics can be used. - Gilgamesh (talk) 01:01, 8 November 2012 (UTC)

I agree with CodeCat about the accents, they are phonemically contrastive and completely essential. Compare words like νόμος vs. νομός: are we meant to write them both as nomos? How silly would the etymology entries on pages like ܢܡܘܣܐ look ("etymology 1: nomos, etymology 2: nomos--but they're different, I swear!")? --334a (talk) 08:22, 21 December 2012 (UTC)