Appendix talk:Old Irish pronunciation

Representing the tense sonorants

I'm using /ʟ ʟʲ ɴ ɴʲ ʀ ʀʲ/ to represent the tense sonorants, in line with the traditional phonetic transcription of Irish but I'm worried about it. Although the page explains that those symbols are not being used with their usual IPA meanings, people who see things like /ɴax/ in an entry may not click through to read the explanation and, if they know IPA, will be left with the impression that Old Irish had a velar lateral (not just velarized but actually velar) and a uvular nasal and trill. I'm considering using the "strong articulation" diacritic ◌͈ from the Extensions to the International Phonetic Alphabet instead (e.g. /n͈ax/ instead of /ɴax/ and /ˈbun͈ʲe/ instead of /buɴʲe/); that's what Korean phonology does for the "tense" consonants of that language. What do others think? Shall we stick to the traditional letters, even though they misuse the IPA, or shall we use diacritics more in line with the IPA that will be unfamiliar to people who have studied Old Irish? —Aɴɢʀ (talk) 01:24, 9 August 2014 (UTC)

I prefer using proper IPA. We can either mark tense sonorants with a strong diacritic, or mark the other set with a lax diacritic (I think there is one?). —CodeCat 01:33, 9 August 2014 (UTC)

Yes, the symbol ◌͉ is the corresponding Extended IPA symbol for "weak articulation", but I'd rather use the "strong articulation" diacritic because people who are accustomed to seeing /ʟ/ vs. /l/ or /L/ vs. /l/ will be less thrown off by /l͈/ vs. /l/ than by /l/ vs. /l͉/. —Aɴɢʀ (talk) 02:05, 9 August 2014 (UTC)

I don't think marking lax articulation rather than strong is a good idea, because

1. the weaker articulation was the more common form (medial, final and lenited, as opposed to geminated and initial),
2. the weaker articulation seems to be considered, then and now, the baseline from which the strong articulation differs, in whatever period of Irish, and
3. a century of Old Irish scholarship has marked it this way, and we would be introducing a gratuitous difference with the scholarship for negligible benefit.

Personally, I prefer the /ʟ ɴ ʀ/ forms, but maybe enabling the IPA template to put up a note warning that these symbols are used outside the IPA definitions, because reasons. But then, I tend to conservatism. (I think the prime to mark palatalisation is neater than /ʲ/, but no-one asked me.) --Catsidhe ^{(verba, facta)} 03:38, 9 August 2014 (UTC)

I sympathize with that, and when I first started working on Irish phonology topics at Wikipedia, I used the traditional symbols /L N R ʹ/ too, but eventually I came to accept (and this applies to Wiktionary as well) that we're not a closed system here. If we were just a dictionary of Old Irish, we could use whatever transcription system we liked as long as we defined it properly somewhere. But we're not; we're a dictionary of all languages, and we use IPA in as uniform a way as possible to maintain some degree of consistency. What I'd really like to do is commit myself to a precise articulation for these sounds and use /l̪ˠ l̠ʲ lˠ lʲ/ for /ʟ ʟʲ l lʲ/, /n̪ˠ n̠ʲ nˠ nʲ/ for /ɴ ɴʲ n nʲ/, and /rˠ rʲ ɾˠ ɾʲ/ for /ʀ ʀʲ r rʲ/, but in fact we don't really know that those transcriptions are correct. That's what I like about using ◌͈ for the tense sounds: its definition is so vague ("strong articulation") that we can make the distinction between the fortis and lenis sounds without committing ourselves to precise articulatory details. —Aɴɢʀ (talk) 17:31, 9 August 2014 (UTC)

I'm going to implement it now. —Aɴɢʀ (talk) 13:19, 11 August 2014 (UTC)

Unstressed short vowels

According to Stifter’s Sengoídelc (§3.3.10, pp. 21–22), de Vries’s A Stundent’s Companion to Old Irish Grammar (pp. 2–3), and McCone’s A First Old Irish Grammar… (§ I.B.6, p. 16), in unstressed closed syllables only two short vowel qualities existed: /ə/ (spelt a, ai, i, e, o depending on neighbouring consonants) and /u/ (spelt (i)u, possibly o). Unstressed vowels kept their non-reduced quality only in absolute auslaut (when no consonant followed).

Shouldn’t this be represented by this appendix? It doesn’t currently acknowledge Old Irish /ə/ at all (which, at least to me, was confusing when first compared Wiktionary transcriptions with the mentioned books).

Old Irish entries consistently suggest non-reduced vowel qualities, eg. in carpat /ˈkarbad/ (instead of /ˈkarbəd/), beirid [ˈbʲerʲiðʲ] (instead of /ˈbʲerʲəðʲ/), etc. I’ve ‘corrected’ this in peccad, from /pʲekað/ to /pʲekəð/ (change) before I noticed that OIr. entries consistently don’t use /ə/ – should I revert this? // Silmeth ^@talk 10:59, 19 July 2021 (UTC)

Also the linked Wikipedia section on OIr. phonology mentions this (It is usually thought that there were only two allowed phonemes: /ə/ (written a, ai, e or i depending on the quality of surrounding consonants) and /u/ (written u or o).) ⁊ gives such transcriptions as /klaðʲəv/ for claideb or /arɡəd/ for arggat, so it’d be nice to be consistent with that too, I guess. // Silmeth ^@talk 11:10, 19 July 2021 (UTC)

+ ping @Mahagaja and @Mellohi! (since I see a similar change to mine done in lebor by them). // Silmeth ^@talk 12:53, 19 July 2021 (UTC)

@Silmethule: That represents a level of underlying representation that is more abstract than is useful in a dictionary. Compare languages with final devoicing like German, for example: at an abstract level of representation, Rad is underlyingly /ʁaːd/ because /d/ is present in inflected forms like Rades and Räder; but for a dictionary it's more useful to show the sounds that actually surface, i.e. /ʁaːt/. The highly consistent ways in which unstressed vowels between two consonants were spelled (between two nonpalatalized consonants as a, between a palatalized and a nonpalatalized consonant as e, between a nonpalatalized and palatalized consonant as (a)i, and between two palatalized consonants as i) strongly suggests that these vowels were actually pronounced differently. If they had all been pronounced [ə], there would have been a lot more instability in spelling – which is exactly what happened in Middle Irish, when unstressed vowels really did merge to /ə/. The hypothesis that all unstressed vowels between two consonants belong to the single phoneme /ə/ comes from Kim McCone, who is such a huge name in Old Irish studies that it's difficult for anyone to contradict him and be taken seriously. (David Stifter, author of Sengoídelc, was McCone's student and follows him in all things. McCone is well known for brooking no contradiction: as far as he's concerned, he is the representative of the Truth™ and anyone who disagrees with him is wrong and foolish.) But from the point of view of theoretical phonology, the hypothesis makes little sense. Yes, the quality of an unstressed vowel between two consonants is predictable, but because the quality of unstressed vowels is unpredictable in word-final position (/a/, /e/, /i/, /o/, and /u/ can all appear contrastively), it's already clear that five distinct vowel phonemes have to exist in unstressed syllables. From the phonological point of view, it's unnecessary and undesirable to posit a sixth unstressed vowel phoneme /ə/ that appears only between two consonants – especially when it's clear that the surface pronunciation was not [ə] but was rather one of these five vowel phonemes that have to be posited anyway. At any rate, the compromise I've been aiming for at Wiktionary is to use square brackets rather than slashes and show the sounds that were actually pronounced ([a e i o u] plus [ɘ ɨ] after nonpalatalized consonants without directly implying that those sounds were distinct phonemes (although I personally believe they were). —Mahāgaja · talk 12:58, 19 July 2021 (UTC)

@Mahagaja: Perhaps it’s worth to explicitly note the departure from McCone&Stifter’s transcription either here or in Wiktionary:About Old Irish (or in both places) then. Since it’s the system readers are probably most likely to be familiar with if they have any previous contact with Old Irish and it may cause confusion (certainly confused me, and I guess not only me, judging by the mentioned lebor edit). (On a side note, in the table, for initial fortis slender /l͈ʲ/, shouldn’t it be lebor instead of lebhor?) // Silmeth ^@talk 15:11, 19 July 2021 (UTC)

Yes, that's probably a good idea. I fixed lebor in the table. —Mahāgaja · talk 15:15, 19 July 2021 (UTC)

@Silmethule, Mahagaja: A few clarifications on Stifter and McCone's positions. — Ceso femmuin mbolgaig mbung, mellohi! (投稿) 19:34, 19 July 2021 (UTC)

• Stifter divides Old Irish prosody-vowel categories into the following:
  • Pretonic (/a i u~o/)
  • Stressed internal (/a(ː) e(ː) i(ː) o(ː) u(ː)/)
  • Posttonic non-final (/ə u/ and long vowels)
  • Unstressed final (short vowels only)
  • Stressed final (long vowels only)
• Mahagaja: I find it ironic that you keep complaining about McCone's system, while phonetically your system is barely different than McCone's (which you had admitted a while ago); the main difference is that his exact allophones are slightly different than yours. Stifter follows McCone in believing that /ə/ has allophones alternating phonetically based on the environment, but he is hesitant to assigning phonetic values to the allophones. Amusingly enough, in a 2020 presentation Stifter transcribed all schwas as [ə] — exactly the opposite of what McCone does.

I will propose a settlement between myself and Mahagaja. I will not delete Mahagaja's phonetic transcriptions or force schwas into them. In turn, Mahagaja, I would like you to not delete any more phonemic transcriptions, especially those that include schwas. — Ceso femmuin mbolgaig mbung, mellohi! (投稿) 19:34, 19 July 2021 (UTC)

@Mellohi!: I don't see what's so ironic; my only objection to McCone's system is the positing of a phoneme /ə/, which is both uneconomical and insufficient (as it fails to account for instances where o appears between two nonpalatalized consonants, like lebor). I won't remove phonemic transcriptions that include schwa, but I will add "phonetic" (actually, surface-phonemic) transcriptions in square brackets to them. I may, however, change existing /.../ transcriptions that do not contain a schwa to [...] transcriptions for consistency's sake. —Mahāgaja · talk 19:47, 19 July 2021 (UTC)