Module talk:la-pronunc

Archive for 2014–2020.

Discuss ɑ/ä, voiced /ex-/
The generic reference will be [https://acda-publications.s3.us-east-2.amazonaws.com/De'athDec16.pdf De'Ath L. (2016). The Latin Problem] because the name's dope xD The ɑ is mentioned by most manuals including the Liber Usualis; some other manuals (Copeman H. (1990), Singing in Latin) insist on the "palatal" or Italian/Spanish vowel, but they might be describing speech instead of sung performance. In Russian singing for example, the same opening applies especially to /a/ and /o/ to the effect that I don't know which one's being sung at all. Since people need the Roman Ecclesiastical primarily for sung performance, it makes sense that we cater to that, especially seeing the current Italian pronunciation of Latin is at variance with official prescriptions. If we want to include it, it should wait till after we have the German, Polish and French traditional pronunciations (nobody mention the English one!).

Anyway, the same goes for the voiced /s/. The Italians voice it, but that's not what the singing practice and prescription aims at. Even if you listen to some chants, fully voiced /s/ is characteristic of German and French, not of authentic Roman, and even the Americans seem to go lightly on it. Probably this should be treated in the earlier discussion here (Intervocalic s).

The CL ä is because wiktionary seems to stick that symbol wherever it can, so why not have it in Latin, where it seems to have been dead center, but phonologically back, and earlier (in Plautine) probably phonetically back too. I also want to add [wɑ] and [jæ] to account for the variations vocāre~vacāre (which is earlier) and jānua~jēnua (+jēnuārius) (which is later).

I've tried the best I could to make the prefix /ex-/ be voiced also when a following voiced consonant results in the syllabfication /eks.l/, but the result seems hacky. I don't really understand what I'm doing so if anyone can think of a neat way to write this, go right ahead. Brutal Russian (talk) 20:13, 27 March 2021 (UTC)
 * I recall father being the typical example for Latin long a and Ancient Greek long ᾱ in the the textbooks I used, and I'd always thought that this was simply because father is the closest thing to a central a in most English dialects. (Though nowadays the cat vowel is closer in quality at least in quite a few dialects of British English, including "modern RP".) De'Ath L. (2016) seems to be saying a is back in the Roman pronunciation, but is inconsistent in writing it as  for the most part in Tables 5 and 7 and as  in Table 6, and as far as I can see doesn't explicitly say it is back or why it would be back when the official ecclesiastical pronunciation was based on the Italian. The Liber Usualis looks like weak evidence to me, as the only two reference words are father and cat. Cat was often fronted and raised in old-fashioned RP, even to, and would often be similar in American English, so it would have been unsuitable as an approximation of even a central a. For either central or back a, the English reference word would have to be father, and if they cared about the backness, they'd have to mention some other language like Italian. So the evidence for a back a doesn't look very clear to me. Unfortunately I haven't listened closely to the quality of a in choral music. — Eru·tuon 07:21, 5 April 2021 (UTC)
 * It doesn't look particularly clear to me either from reading the materials, but I know that the most open vowel is the back ; this might be more in the domain of vocal technique than phonetics to be honest, something I know close to nothing on; what I do know is that the /a/ needs to be as open as humanly possible. For a while I was a bit obsessed about figuring out what kinds of /A/ (the archiphoneme) there were in Italy; so far I've figured out that it's a peculiar, charming back type in Veneto; Neapolitan too has it so open that it even resists the vowel deletion that operates there (the same exception is observable historically in most of Gallo-Romance); it retreats further back the louder and more emotional the utterance xD Central varieties on the other hand have it - well, central! - or even front, with raising metaphony in and around Umbria. Lombard (among the most awesome-sounding Romance languages) has both: front when short, back when long, like in Parisian French 50 just years ago (still in Quebecois), which squares nicely with basic articulatory physiology; albeit Hungarian for instance has /ɒ~ɔ/ vs /a:/, as evidently did proto-Slavic; pre-Attic raised /a:/ instead.
 * Back to external sources. I had always thought it was a device for English speakers too; but the Liber Usualis (at the end) talks about the importance of the "rich, open, warm sounds of the vowels A and U", and it's echoed by whatever book this is from. Alderson R., Alderson A. (2020). A New Handbook for Singers and Teachers refers the same doctrine. For reference, here's what I consider a canonical - in Latvian (it's consistent there, mildly centralised when unstressed). Admittedly English speakers on both sides of the Atlantic might be prone to turn that into a rounded vowel. But the vowels in Catalan/Dutch/French here seem rather to match the short /æ/ of General American accents, and Spanish is even centralised, as usual. That's definiteny not the Roman choral vowel; as a guide to singers/conductors, specifying the maximally open vowel seems to me like the right thing to do. Brutal Russian (talk) 10:26, 5 April 2021 (UTC)

/z/ in classical
I confess I haven't found an short yet illuminating treatment of the matter with which I would agree, but I've found a passage in Velius Longus (50.8 sqq) that I think is key (here's fresh critical edition that adopts some different readings). Allen seems to misinterpret it, partly due to the incertainty of the reading (which is as usual), partly following Sturtevant and Kent. Basically this part:  The "d" reads as "p" in one codex, which confuses one of the three Greek letters (Ψ) for another (Ζ), but the former of these is not being discussed yet, and so is the false reading.

—There's seemingly no two ways to read the statement that Z starts with a D; further Longus treats the problem of whether it consists of DS or SD; the latter he rejects. Despite all this, Allen still ends up wrongly concluding that it doesn't in fact contain a D at all but is a basic /z/, which would make the whole previous discussion make no sense. He's being mislead by the final remarks about the geminate ZZ, which Longus says is in fact different from a single one in a way that Χ and Ψ aren't: ZZ starts and ends in the same sound, which he earlier refers to as "semiuocalis". This is his way of equating it to stop-liquid clusters: a single Z to him syllabifies as the onset in its entirety. This then was the native Latin pronunciation of the sound - as a true affricate, hence my current transcription in the module.

—The double ZZ passage is easiest to understand if we assume that the first affricate lost the stop part when in gemination, and optionally the second affricate as well, giving /z(d)z/; if we assume this, the reading  "if there's some additional sound when I pronounce this letter" can be preferred to the confusing   "if someone joins in" - even though admittedly the next sentence has  ; but this would in fact explain the scribe's confusion.'

—Now the fact that in Attic Greek the letter was always what Longus calls ZZ intervocalically, this would very nicely explain the confusion as to which sound it starts and ends with, since intervocalically it started in the opposite sound /zdz/ from the absolute position /dz/. It's quite possible that Attic and the Doric of Magna Graecia did in fact have different pronunciations of this letter. Touratier here concludes that the letter already existed in Old Latin with the Etruscan value of /ts/; I think this value never truly disappeared from the region as much as the languages where it was employed were seen and heard less and less at Rome. It does survive into Italian, after all.

—There's also a phonological/etymological reason it couldn't have been the voiced equivalent of /s/: Romance shows no such outcome anywhere; a century or so after Longus is writing, Z pops up to spell the outcome of /dj/, and after another century of /tj/. Nuorese Sardinian even has /θ/ where other varieties have affricates (faθo < facio, puθu < puteum); some Logudorese varieties have /t/ in its place. This at any rate is my evidence that the marginal sounds notated with Latin Z were specifically dental, as opposed to the retracted /s/ which continued to be subject to rhotacism if and when voiced, which in good old Sardinian ended up being completely regular before voiced consonants. This by Martianus Capella is also very suggestive, and also shows it was a known letter with its own sound in pre-Plautine Latin:. Those who couldn't pronounce the dental, or didn't want to affect a Hellenising pronunciation, substituted a /s/ for it. Here I think there's a need for automatically-generated variants with an /s/ in place of /z/ since they were obviously current. As a reverse, particularly in Spain but also elsewhere, there are some words with initial /ts~dz/ in place of /s/, and there's some Greek borrowings among these IIRC.

—Feel free to leave your thoughts or further references. Brutal Russian (talk) 22:51, 2 April 2021 (UTC)


 * Certainly an interesting discussion can be had here about ⟨z⟩, but does any scholar support the assignment to Classical Latin of the realizations that you have implemented here? What do you mean by "it does survive into Italian"- were any words with  inherited from Classical Latin into Italian with /dz/? The Nicodene (talk) 22:34, 17 July 2021 (UTC)
 * Duly noted that neither of the above questions have been answered in the past three and a half years.
 * Since someone mentioned this on Discord today, I have decided to look further into the matter. The following will indeed be 'illuminating', but not in the way that you'd imagined.
 * You have misinterpreted the Latin in multiple places. First and foremost you have incorrectly concluded that Velius Longus agrees that z begins with a d sound when even the sentence that you've directly quoted makes it clear that this is rather the view of another man:
 * Verrio Flacco placet mutas esse, quoniam a mutis incipiant, una a c, altera a d
 * (In case it still isn't clear why: quoniam + subjunctive.)
 * Two lines later Velius begins a new sentence with mihi videtur. It should have been clear, both from that phrase and the contrary arguments that immediately follow, that this is the moment when he switches from reporting others' views on the matter to presenting his own.
 * Second, Velius uses a system for indicating sounds that you've neglected to consider. It takes no virtuouso of Greek phonology to realise that when he insists that ⟨z⟩ does not equate to sd, nor ⟨ζ⟩ to σ+δ, the pronunciation he is arguing against is [zd]. A few lines later we learn that ⟨ζ⟩, in intervocalic position any rate, does not sound like δσ; that is, it does not sound like [dz]. Next, and most importantly of all, we learn that ⟨ζζ⟩ sounds like σσ, and here I will insist on pointing out that the only remotely plausible interpretation of this is [zz]. Were Velius actually trying to transcribe *[zdz], he would obviously have written sds or σδσ according to his own system.
 * Third, Velius insists that ⟨z⟩ is not a 'two-part letter' (si quis secundum naturam vult excutere hanc litteram inveniet duplicem non esse). As can be seen in the last sentence of the section, a 'two-part letter' is a letter that begins with one sound and ends in another, like ⟨ψ⟩ [ps] or ⟨ξ⟩ [ks]. That means ⟨z⟩ would in fact be a 'two-part letter', contrary to what Velius says, if it sounded like either [zd] or [dz]. Ho-hum.
 * Fourth, and most damningly of all, Velius states that ⟨z⟩ maintains one sound from beginning to end (si quis supervenerit me dicente sonum huius litterae inveniet eundem tenorem a quo coeperit). In case it isn't clear: the use of tenor unambiguously implies a single sustained sound like [zː], not a sequence that jumps back and forth like *[zdz].
 * You, not Allen, are the one who has misinterpreted Velius Longus. Nicodene (talk) 00:59, 20 February 2024 (UTC)

Pronunciation of dein(de)
Cser 2016:  Why this anomaly, when it could be easily repaired with the syllabification /de.in/? In my opinion it's because there simply wasn't any /i/, but the two vowels coalesced into a short [e] that was neither /e/ [ɛ] nor /i/ [ɪ], and for this reason as well as due to clear etymology, continued to be spelled EI; they probably did so at a stage when /ei/ was still a dipthong, which then shortened before a sonorant with the same resulting [e] as in. The forms where there was an actual /i/ are the ones where it makes a syllable; these were then not alternative syllabifications, but recomposed anew. Cf. also all the other -postfixed forms which Priscian says were all deaccented. Cf. finally dende, which reflects /ẹ~ĭ/ with no hint of any anomalous glides; but since it was deaccented, it couldn't have been the /ĭ/. The difference from the current transcription is minor, but it might contribute to preventing some children from damaging their tender articulation organs. Also it doesn't require changing the modulea as there are still a few words with genuine /ei/, I'm just posting it here because this is the place. Brutal Russian (talk) 02:50, 4 April 2021 (UTC)

Proto-Romance transcription
The Vulgar Latin pronunciation that you're currently changing was transcribing a variety of Campanian Latin (bay of Naples), which was suggested and at least not opposed here. The aim was to replace a clearly ahistorical mash-up pronunciation in the diglossic tradition with a good approximation of a Classical-period variant. It's not clear to me what you're doing, and it seems like you're trying to return to that earlier state. If you wish to add a Proto-Romance phonemic continuation, this needs to be done in a separate transcription, and this will contain no long vowels, for instance; but giving a single phonetic transcription of Proto-Romance is misguided not only because Proto-Roman is a phonemic reconstruction with no certainty about phonetic reality, but also there potentially existed any number of concurrent PRmc. pronunciations and nothing like the single prestigious urban standard of the Classical period.

I would like to ask you not to replace existing pronunciations, and discuss your proposed changes with others. Converting Classical phonemes into PRmc. should be easy enough - it's adding a new transcription that seems non-obvious to me. Perhaps we can find someone who can help out; in the meantime I'd like you to undo your changes - remember you can experiment with a sandbox copy. Brutal Russian (talk) 07:57, 26 May 2021 (UTC)


 * You seem to be under the impression that this is supposed to be a purely phonemic transcription. That is not the case; many, perhaps even most, of the thoroughly cited phenomena described here and here are in fact allophonic.
 * The presence of long vowels in the phonemic transcriptions is not deliberate on my part, but rather the result of my being forced to work with the existing transcription scheme in this module, after my request to start a separate one was rejected. It can be solved by either substituting the phonemes, e.g. /i ɪ e ɛ a ɔ o ʊ u/ for the vowels, or simply hiding the phonemic transcriptions for Late/'Vulgar' pronunciations. The latter would have the added bonus of forestalling wasteful debates about what should count as a phoneme in that transcription system, e.g. whether intervocalic /b/ should count as /β/.
 * "no certainty about phonetic reality" also describes our understanding of Classical pronunciation. We work with the data we have to come to what conclusions we can.
 * You claim that the Classical period had a single prestigious urban standard but that the later period did not. Leaving aside any other details for the moment, I would like you to cite this supposed difference.
 * J.N. Adams talks about 'Campanian Latin' a handful of times, but I'm not seeing any mention of it having, for instance, a bilabial fricative already (let alone turning word-initial /b/ into a bilabial fricative, which is what your transcription system did). The only phonetic feature that he mentions in relation to it, at least the only one I have been able to find by looking through his references to 'Campanian [Latin]', is the monophthongization of /ae̯/.
 * For the moment I am even refraining, purposefully, from adding some of the more 'exotic', even if well-sourced, allophonic features such as /ll/ [ɭɭ] to the Late/'Vulgar' transcriptions, focusing rather on the features that are widely agreed-upon in modern scholarship. No similar caution seems to have existed on your part, as shown e.g. by your applying an essentially Spanish system of alveolar taps to Classical Latin, apparently the result of a hunch on your part, as you provided no sources when asked about it. The Nicodene (talk) 16:25, 26 May 2021 (UTC)
 * I'm really quite perplexed about this. Proto-Romance reconstruction does not deal with allophony, which is physical realisation: it reconstructs abstract phonemic structure. Reconstructed Proto-Romance is not claimed to be a language of a specific time or place, and it cannot in principle have one pronunciation even based on this. Allophony belongs to a given variety of late Latin, all of which had different realisations and different allophonies. I have to ask you again to explain what exactly you're trying to transcribe, to which I will return presently.
 * It should be easy to remove the length, Ecclesiastical lacks it. The 9 vowels you list are clearly what we want to see in the final result. By the way, could you link to your request for starting a new module?
 * I don't see how a debate about an intervocalic /b/ can be possible, because I've never seen anyone suggest that the Latin /b/ and /v/ merged into anything other than /β/. In any case, since we're transcribing Proto-Romance, which is a phonemic/structural/segmental/etc. reconstruction, if such a debate existed, it would have to be decided on.
 * Ok, now this is the most confusing part. Tell me what of the following you want clarification on: Classical Latin is the form of Latin language recognized as a literary standard by writers of the late Roman Republic and early Roman Empire (straight from wikipedia), it was developed on the basis of the speech of the city of Rome, which was considered, what we would call prestigious, and in opposition to the non-standard/non-presitigious varieties. Like most modern standard languages that aren't English, it had an associated standard pronunciation. All attempts to describe the pronunciation of (Classical) Latin from Alcuin through Erasmus and down to today have been trying to reconstruct this speech variety, and any other pronunciations that might have existed have been described in terms of their differences from that Classical standard; and the best-described has been traditionally considered an entirely different language, now generally recognised as a dialect - Faliscan.
 * Nothing of the sort applies to Late Latin, and I don't understand why you want a citation for this. The political situation in late/post-5th century Roman empire should explain why no such pronunciation existed. The Late Latin literary tradition was a continuation of the Classical tradition, but I would be wary of the idea that the uniformity of written Late Latin means there existed a uniform school-taught pronunciation. In any case, this school pronunciation is precisely not what we're transcribing.
 * The DERom puts the initial break-up of its reconstructed Proto-Romance at the late 2nd-early 3d century AD, when Sardinian split from the rest. This corresponds to the stage Banniard (2013) calls Classical Latin, followed by Late Spoken Latin phases 1 and 2, and then (!) Proto-Romance (8th and 9th centuries). Now this needs to be stressed: only the first of these stages, the pre-break-up one, can logically represent a common, standard and universal pronunciation. The unsettling truth is that this pronunciation is in fact what we currently transcribe as Classical Latin, and so DERom's ultimate stage of reconstruction ends up being a parallel language to Classical Latin (not even Late Latin). This paradox is exactly why I don't see a way to marry their Proto-Romance to our Latin. The historical reconstruction method they're using is unable to reconstruct the phonological system of Classical Latin as we know it existed, for example the length and unstressed vowel oppositions. We can compare, but we cannot equate.
 * Campanian Latin is spread around the literature for Vulgar Latin. It's found everywhere that uses Pompeian evidence. The go-to reference for this is Väänänen 1966 and 1981, but obviously there's mountains of literature, most of Adams' monographs.
 * I've had a thorough discussion of the current transcription of /r/ and provided plenty of evidence, it's unfair to call anything I write a hunch. I explained that the system is not that of Spanish. The other party of the discussion, namely Urszag, seemed to be satisfied by the reasoning and evidence adduced. If you're simply after a nominal reference, then I didn't bother giving any because I don't believe in offering nominal references in lieu of sound self-conducted research, but you can have one here. Brutal Russian (talk) 20:39, 26 May 2021 (UTC)
 * The DERom very much does deal with allophony. The citations on the Proto-Romance page point to discussions about the realization of intervocalic /b d g/, of /g/ before /n/, of /ll/, of /i/ and /u/ before vowels, of the effects of [j] (from /i/ before vowels) on preceding consonants, of the prosthesis before word-initial pre-consonantal /s/, and so forth. That is all very much allophony, and that is what I am doing here as well.
 * I'm sure the length can easily be removed, I just do not know how to do it. As for the request, I made it several months ago on the Information Desk page. I don't remember when exactly.
 * Thereotically someone could argue that intervocalic [-β-] counts as an allophone of /b/, as it is in complementary distribution with [b]. Similarly, one could argue the phonemic status of /j/. These are just meant as examples of the sort of potential debates about phonemes that can result, even when people agree about the actual realizations.
 * Prestigious pronunciations very much still existed in the Late Latin period, right through to the end of the Western Empire in 476, exactly as you mention. J.N. Adams (2008: chapter III § 1.2) quotes numerous late authors apologizing for their rustic habits and praising the speech of the capital.
 * That isn't to say that the transcription system I am making aims at a cultivated late pronunciation, necessarily, although with some tweaks it certainly could. It is rather to say that the situation in the 1st century b.c. and that in the late Roman world were not fundamentally different. If anything, one would expect less, rather than more, regional divergence in Latin pronunciation after centuries of centralized rule by Rome (Wright 2002: 27–28). That there existed a relative degree of uniformity is supported by the fact that it has proven exceedingly difficult to find tangible evidence of regional divergence in the period in question, other than vague comments by various writers. Adams dedicated an entire book to this topic and only managed to establish two phonological differences by the sixth century, in the process tearing down misconceptions of older scholarship about the supposed 'distinctness' of the Latin spoken in Iberia or Britain.
 * The DERom does, as you say, pin its reconstruction at about 200 AD, though Dworkin puts it centuries later. (This is incidentally why I made the disclaimer earlier that I do not agree, necessarily, with all the details of what they say.) As you have probably been able to gather, I support a date later than 200, but in the end, it does not really matter for our purposes. What is important is that the comparative method gives us a synchronic sketch of what spoken Latin was like at some point well after the 1st century b.c., and that is what is important in the end. The label of the new transcription system can simply be set to something like "Late, informal" - which is what I was planning to make it anyway.
 * There is no paradox. If what the DERom reconstructs, pronunciation-wise, truly does reflect the situation as early as AD 200, then it would simply mean that that is how Latin was pronounced at the time, not that there were 'parallel languages'. Of course I would rather put the date at ca. 300-350, since it is about then that we begin to see solid evidence for the collapse of the CL vowel quantity system, among other considerations. But again, disagreements on the precise date have no bearing on the validity of the phonological reconstruction, since it is synchronic.
 * As for Banniard (who is, incidentally, not a linguist but rather a historian) I can't quite bring myself to believe that he would claim that a unitary Proto-Romance existed as late as the ninth century AD, considering that the Oaths of Strasbourg were written in 841 in a language that already shows numerous innovations that clearly distinguish it from, e.g., Central Italian. If he really does speak of a protoroman in that period, most likely he is simply using the term to mean 'pre-literary Romance', rather than 'pre-divergence Romance'.
 * Adams barely even mentions the term 'Campanian Latin' in his works, and only attributes to it the single phonological feature that I mentioned. A Google search for that exact phrase reveals a mere 323 results, often simply repetitions of each other, and many of them surrounded with quotation marks. Far from 'mountains of literature', that amounts to an obscure and often simply ad-hoc term. I gather from your comment that you are using it to refer to the spoken Latin attested in Pompeiian graffiti.
 * If you want a 'Campanian Latin' transcription system, you need to provide sources regarding the various features you assigned to it. For instance you gave short /i/ a realization as [e] in all positions, despite Pompeiian graffiti only showing that phenomenon with any consistency in final unstressed syllables, generally in verb forms (Adams 2013: 58-61). The distribution of [b] and [β] was rather adventurous and appears to claim a complete 'proto-Neapolitan' betacism; that too will need to be sourced, if it was intentional. Likewise the nasalization of vowels before all syllable-coda nasals.
 * Lastly, your 'Campanian Latin' would be inappropriate for lemmas reconstructed from Romance via the comparative method.
 * As for CL /r/, it's good to see that there is a source that examines the Latin evidence and argues for it being an alveolar tap. The Nicodene (talk) 00:43, 27 May 2021 (UTC)
 * While I'm just watching your very interesting discussion, I just wanted to mention I definitely favour the label "Late, informal" over "Proto-Romance". I have a complaint about the use of "Proto-" in this term for what is really a reconstruction of common sort of informal pronunciation of the Late Latin period, including allophony. And I like "informal" because Late Latin grammarians still talked about pronouncing final -m in oratory before a vowel, giving some odd recommendations even, all due to rhetorical concerns (as Anna Zago's paper "Mytacism in Latin grammarians", 2018, DOI: 10.1515/joll-2018-0002, shows), not to mention Augustine of Hippo's mention of people caring whether people (well, qui illa sonorum vetera placita teneat aut doceat, at any rate) correctly remember where to pronounce h in words (Confessiones 1.18.29). Yet "Proto-Romance" doesn't have /h/ or a realization of final -m. (On similar lines, I also wonder about the differing use of "Pre-" in e.g. "Pre-French", that is an earlier unattested stage of French, versus Beekes' "Pre-Greek", which is not an early unattested stage of Greek, but rather one or more substrate languages... Sorry for going off-topic.) --Ser be être 是 talk/stalk 17:56, 27 May 2021 (UTC)
 * As for CL /r/, it's good to see that there is a source that examines the Latin evidence and argues for it being an alveolar tap. The Nicodene (talk) 00:43, 27 May 2021 (UTC)
 * While I'm just watching your very interesting discussion, I just wanted to mention I definitely favour the label "Late, informal" over "Proto-Romance". I have a complaint about the use of "Proto-" in this term for what is really a reconstruction of common sort of informal pronunciation of the Late Latin period, including allophony. And I like "informal" because Late Latin grammarians still talked about pronouncing final -m in oratory before a vowel, giving some odd recommendations even, all due to rhetorical concerns (as Anna Zago's paper "Mytacism in Latin grammarians", 2018, DOI: 10.1515/joll-2018-0002, shows), not to mention Augustine of Hippo's mention of people caring whether people (well, qui illa sonorum vetera placita teneat aut doceat, at any rate) correctly remember where to pronounce h in words (Confessiones 1.18.29). Yet "Proto-Romance" doesn't have /h/ or a realization of final -m. (On similar lines, I also wonder about the differing use of "Pre-" in e.g. "Pre-French", that is an earlier unattested stage of French, versus Beekes' "Pre-Greek", which is not an early unattested stage of Greek, but rather one or more substrate languages... Sorry for going off-topic.) --Ser be être 是 talk/stalk 17:56, 27 May 2021 (UTC)

Reconstructed 'Vulgar' terms
@Urszag "Undo revision 62608313 by Ser be etre shi (talk) With this edit, the module was only displaying the "Vulgar Latin" transcription on all pages".

The change was deliberate, since those lemmas are unattested and only reconstructed by comparing the Romance languages. These are not terms that are used in Modern Latin, so adding an Ecclesiastical or modern Classicizing pronunciation doesn't make sense. Since they are not attested in Classical times either, adding a reconstructed Classical pronunciation is anachronistic. The Nicodene (talk) 16:30, 26 May 2021 (UTC)
 * @User:Ser be etre shi @User:The Nicodene it was displaying only the vulgar transcription on all Wiktionary pages, both for reconstructed and non-reconstructed entries. I assume that was not deliberate.--Urszag (talk) 19:20, 26 May 2021 (UTC)
 * Ah, that's what it did? Yeah that was completely unintentional. The Nicodene (talk) 19:26, 26 May 2021 (UTC)
 * I'm very mystified by this report, because I didn't see this effect yesterday. Namely, non-reconstructed entries showed Classical+Ecclesiastical pronunciations, as normal. And if I redo my changes in the module and test pages with "Show preview", I still see Classical+Ecclesiastical pronunciations just fine. What word(s) did you try that showed you bad output? I tried e.g. os and palatium and they were fine. I haven't reverted your revert because I have since realized I should probably discuss this change before applying it.--Ser be être 是 talk/stalk 19:49, 26 May 2021 (UTC)
 * Removing Ecclesiastical from reconstruction pages looks like a sound idea, but I'm in favour of leaving Classical as well as the former Campanian Latin, those reconstructed entries aren't tied to any specific period - which should be considered together with the fact that late date of attestation does not mean that the word had not existed before (an oft-cited example is ). I believe it's beneficial to display the full range of ancient Latin pronunciations for pre-Medieval pages at least, to illustrate the language's development. On the other hand, the reconstructed Proto-Romance that we're seem to be in the process of adding as a third ancient Latin pronunciation will probably need to be excluded from the stages of the language that postdate the existance of Proto-Romance (Medieval and New Latin). Brutal Russian (talk) 20:50, 26 May 2021 (UTC)
 * That a word reconstructed only from Romance could have, in theory, existed in Cicero's day doesn't mean that it really did, nor does it mean that we should be so bold as to claim that it did by giving it a pronunciation from the 1st century b.c. in its Wiktionary entry.
 * As for excluding the late/'vulgar' pronunciation from words that date from post-Carolingian times: yes, that we can agree on. The Nicodene (talk) 00:57, 27 May 2021 (UTC)
 * Maybe it is only certain pages (ones that have Vulgar pronunciations allowed?). I saw it on saeculum and irremeabilis.--Urszag (talk) 00:55, 27 May 2021 (UTC)
 * Ah yes, now I see the problem. I'll wait until the ongoing discussion winds down to make changes, but will account for them.--Ser be être 是 talk/stalk 18:04, 27 May 2021 (UTC)

To-do list for the Proto-Romance section
If someone could help with any of these, I would be infinitely grateful.

- Shorten long vowels in syllables that are closed by the allophonic gemination of a palatalized consonant.

- Caveola, asciola, etc. should have stressed lax vowels, not tense ones (in those two examples it should be [ɔ], not [o]).

- /ks/ should reduce to /s/ before or after another consonant as well as word-finally.

- Disable Classical and Ecclesiastical pronunciations for reconstructed Proto-Romance lemmas.

- Potentially change the name from 'Vulgar' to something like 'Late, informal'.

- Figure out why [gʲ] is not yielding [ɟ], even though [dʲ] is.

- Figure out a way to manually set stress position. Exvŏlo, for instance, should be stressed on the second syllable, not the first, because stress generally moved away from verb prefixes in Late Latin / Proto-Romance. (This can be seen in every Romance derivative of exvolo.) The Nicodene (talk) 05:53, 20 June 2021 (UTC)


 * - Figure out a way to add (allophonically) the vowel [ɪ] before words beginning with /s/ plus another consonant. E.g. scribere > [ɪsˈkɾiːβeɾe]. The Nicodene (talk) 05:59, 20 June 2021 (UTC)
 * /m/ should only be deleted at the end of word-final unstressed syllables, not at the end of any syllable. The Nicodene (talk) 21:44, 3 July 2021 (UTC)

Stress
At the moment the Classical module transcribes stress as phonemic. Isn't it rather allophonic (dependent on vowel length, syllable count, and syllable structure)? Perhaps there is a counter-example or two, like illic, where the 'anomalous' stress location is, I presume, simply the result of a loss of the final syllable /ke/ (archaic illicce). I would be surprised if there is any more than a handful of cases like that, though. The Nicodene (talk) 03:59, 10 July 2021 (UTC)
 * There are exceptions to the regular stress rule, as you note, so the position of stress is not predictable solely from the segmental phonology of a word, despite being regular in the vast majority of cases. If you consider minimal pairs to hold special significance for the question of phonemic contrast, there is even a possible minimal pair for stress in Latin: nosˈtrās (nominative singular, genitive nostrātis) vs. ˈnostrās (feminine accusative plural of noster). (I say "possible" because I haven't reviewed the actual evidence for the form nosˈtrās with ultimate accent, just read about its alleged existence).
 * It is possible to analyze away this surface contrast. But regardless of whether theorists would categorize Classical Latin stress as allophonic or phonemic, I think the simplest way to present the information in a dictionary is to show stress in both the phonemic and phonetic transcription. I see no advantage to stripping it out of the phonemic transcription. Wiktionary also shows the position of stress in our phonemic transcriptions for Swahili, another language where the position of stress is almost always predictable (in that case, on the penult), but may be unpredictable in a handful of exceptional words (e.g. barabara). --Urszag (talk) 20:10, 10 July 2021 (UTC)
 * Fair enough. Nostrās (-ātis) is an interesting case, thank you. I didn't know that it was stressed on the second syllable. The Nicodene (talk) 08:32, 11 July 2021 (UTC)

Vulgar Latin
As can be seen in previous discussions on this topic, many found fault with the Vulgar Latin sub-module.

The problem, essentially, is that the term 'Vulgar Latin' is too vague: per Lloyd (1979) it has more than a dozen varying definitions, and the times span it covers (ca. 200 B.C.E.–600 C.E., sometimes even later) is massive. Pronunciation was changing throughout that period, needless to say, as in any living language–so how could Wiktionary give one transcription to cover it all?

Everyone seems to agree that more precise alternatives would be preferable.

At the moment I have implemented in its place a Proto-Romance sub-module, but that is meant for reconstructed entries. (I should probably rename it from 'vul' to 'proto', or similar.)

For attested Latin words, we can have a range of different pronunciations, to be implemented as additional sub-modules, e.g. Plautine, circa 200 B.C.E., and Pompeiian, late first century C.E. (I could even try my hand at implementing a pronunciation based on Robert Politzer's study of eighth-century documents from Lombardy.)

I propose removing the remaining vul=1 codes on Wiktionary and then, once the Pompeiian or Plautine sub-modules are ready, enabling them for chronologically-appropriate Latin words with pomp=1 or plaut=1. The Nicodene (talk) 11:43, 11 July 2021 (UTC)

w as u̯
I have seen no bibliography that shows w was pronounced as u̯. This should be fixed, since it was pronounced as w --PastelKos (talk) 13:48, 13 August 2021 (UTC)
 * There is scarely a difference between [w] and [u̯]: they are more notational variants than meaningful alternatives, especially when transcribing an ancient language that has no phonetic recordings available. I agree that [w] is preferable; I was going to post about it as part of a more detailed discussion of the treatment of glides and diphthongs (I would also use [j] rather than [i̯]).--Urszag (talk) 18:42, 13 August 2021 (UTC)
 * As a passerby I wish to offer input. I have tried to think in the way that the greatest majority of people will think who see these reconstructions. In my opinion it is aesthetically advantageous to use [u̯], as [w] is popularly associated with Germanic languages, and using it creates this ever so slightly unsatisfactory situation whereby Latin "v" is enforced as being actually a different, and moreover "Germanic", letter thrust into the precious mediterranean Latin, and as if not the "v" that it seems to represent naturally by itself (and which it is denied, and made a charade), while [u̯] in my mind dodges this entirely, and presents a nice and organic look that is untarnished by German philologists, English professors, and other enemies. Consider that one is like asserting that "vacca" is not pronounced "vacca" but instead "wacca" (or indeed "wakka"), and the other is merely a suggestion that "vacca" is "uacca" with a consonant--it can be literally spelled "uacca", for that matter. Somehow the hard "c" is less of an issue. This is tangential, but I have even thought that Latin entries might benefit from the elimination of the distinction between "v" and "u", since it is not characteristic of most of Latin's existence, and it looks decently arbitrary to its "classical" conception (why not spell "Avgustus", "qvoqve", "svasit" if these all represent /w/?), for all it can disambiguate a couple of words. Draco argenteus (talk) 01:03, 24 September 2021 (UTC)
 * There is no rule in phonetic transcription that [w] must be reserved for the transcription of Germanic languages. It's fairly commonly used in transcriptions of Italian and Spanish, including on this wiki: compare Italian /ˈswɔ.no/ and Spanish  /ˈkweɾpo/, [ˈkweɾ.po]. I think [w] is clearer and looks visually cleaner than [u̯]. It's true that the spelling distinction between the letters V and U that Wiktionary (as well as many other texts) follows as a convention is somewhat arbitrary, and I'm not particularly attached to it; however, revising the standard spelling of Latin words in Wiktionary would be a much bigger project that is not in scope for this module.--Urszag (talk) 07:06, 6 October 2021 (UTC)
 * Privately, I have had to witness indignation from common folk at the presence of w instead of u̯ in the transcription for Spanish "muerte" at Wiktionary, but it looked like in that case the person also believed that w was somehow different from u̯, and that w existed only in English. I wished to make this comment earlier, but I support now [w] and [j], although I propose the phonemic /u̯/ /i̯/ as better used over /w/ and /j/ for making the reconstruction go a longer way among the less educated (as a passerby). Since "u as a consonant" (or, admittedly, sometimes the Aeolic digamma) and "i as a consonant" is apparently all you will ever find among Roman grammarians, there is a sense of accord with that. Draco argenteus (talk) 10:33, 20 October 2021 (UTC)

/ˈstr/ > [ˈs̠(t̪)ɾ]
For some reason, the module is claiming that Classical /t/ was variably deleted in this specific environment, as can be seen in the pronunciations generated for stringo or extraneus. Considering the fact that every last Romance derivative retains /t/, I can only assume that this was unintentional. The Nicodene (talk) 19:55, 4 September 2021 (UTC)
 * This is caused by line 139, {"st([.ˈ]?)([^aeɛiɪoɔuʊyʏe̯u̯])", "s(t)%1%2"}, which makes /st/ into [s(t)] whenever it is followed by anything other than a vowel or semivowel (within or between syllables). I agree that the display of /str/ as [ˈs̠(t̪)ɾ] is incorrect. I'm guessing it's unintended and the line is meant to show optional elision of /t/ in words like postpono. While t-deletion in that context seems plausible, I'm not actually sure we need to show it, as I'm not aware of any other texts on Latin pronunciation that mention it. The thesis "Aspects of the Phonology and Morphology of Classical Latin", by András Cser (2016), which among other things attempts to summarize the phonological behavior of Latin prefixes, says that post- "is rare in Classical Latin and shows no phonological interference with the stem. The only possible exception to this is possidere ‘possess’, which is believed to go back to post+sedere ’sit’ (de Vaan 2008:552)."


 * Aside from /str/, there's also /stl/, which occurs rarely, but does occur in the prefixed word postliminium and at the start of a few exceptional forms like stlata. My first vote would be to remove the line entirely; my second vote would be to edit it to make the syllable division after /t/ required rather than optional: {"st([.ˈ])([^aeɛiɪoɔuʊyʏe̯u̯])", "s(t)%1%2"}), which should avoid producing [ˈs̠(t̪)ɾ].


 * User:Erutuon, what's the long-term plan for implementing edits on this module? In the time since it's been locked, the edit-warring participants don't seem to have come to any resolution between them: there aren't many other users active here either. I assume few moderators want to have to deal with evaluating edit requests. There are other changes aside from this that I would like to propose; what is the best way to accomplish this?--Urszag (talk) 08:02, 5 September 2021 (UTC)
 * For post, at least, there is a bit of evidence of elision, now that you mention it. Grandgent (Introduction to Vulgar Latin, p. 121) says that final /st/ "may have lost its t before consonants–as post illum but pos' me, est amatus but es' portatus: Lat.Spr. 473. Cf. pos, posquam in R. 470. According to Velius Longus, Cicero favored posmeridianus; Marius Victorinus preferred posquam: S. 368. Both st and s are represented in Romance languages."
 * András Cser (2016) actually mentions this in a footnote at the bottom of page 168, despite his earlier comment about post- "show[ing] no phonological interference with the stem". Perhaps he considers the change too minor and sporadic (at least in writing) to count as a significant feature. Compare that with, say, the transformation of ex- to ē- before voiced consonants, as mentioned on the page that you cited, which is consistently reflected in spelling and has consequences on the phonemic level. A phonemic represention of, say, [ˈeːdɛrɛ] as /eksdere/ would be a bit dubious, while a representation of [ˈpɔskʷã] as /postkʷam/ is rather less so, especially if we allow for the possibility of spelling-pronunciations restoring t in 'cultivated' speech.
 * Be that as it may, both of the solutions that you have proposed sound good to me. The Nicodene (talk) 05:55, 6 September 2021 (UTC)
 * I see! Thank you for the references. Now that I've seen them, I'd switch my preference to keeping the line but making it only apply when a syllable break or stress mark follows /st/.--Urszag (talk) 07:38, 6 September 2021 (UTC)

Please re-enable Vulgar Latin/Proto-Romance reconstruction function
This was one of the best features of this module. VL/PR is one of the most tragically-overlooked linguistic stages that had profound effect on the world. Even if it's nigh-purely speculatory (so is Classical even, to a far lesser extent), this should be at least demarcated in the module instead of disabling it entirely. The disclaimer would please both the purists and serve the Wikipedian spirit of pro scientia. Personally, I believe most use would be to display a reconstruction of PR circa late 400-circa 600s AD, when the Western Empire collapsed but quite a while before Charlemagne artificially restores high learning, as well as a good midpoint between Classical and true Medieval Latin in a murky yet profound epoch. Restituet! Sigehelmus (talk) 05:18, 7 September 2021 (UTC)


 * The dust is still settling, I suppose, from the last time it was discussed. Perhaps bring it up in the Beer Parlour in a week or two? The Nicodene (talk) 22:58, 9 September 2021 (UTC)
 * Why was it never fixed:((--Sigehelmus (talk) 01:00, 20 April 2022 (UTC)

Transcription of Latin's rhotic consonant
I'm dissatisfied with the current module's transcription of /r/ as [ɾ] and of /rr/ as [ɾː]. The module comments say:

"-- Pultrová (2013) argues for Latin /r/ being an alveolar tap. -- Lloyd (1987: 81) agrees: “The /r/ was doubtlessly an alveolar flap." -- Allen (1978: 33) expresses doubt: “By the classical period there is no reason to think that the sound had not strengthened to the trill described by later writers.” -- No source has been provided for a word-initial trilled [r] versus tapped [ɾ] in other positions in Classical Latin."

Pultrová 2013 is a self-acknowledged outlier. I haven't yet read Lloyd 1987, but a contrary view is expressed in Wireback 2015, a source that goes into quite a lot of detail on the question of the probable allophones of /r/ in Latin and in the intermediate variants between Latin and Spanish, and which does in fact support the view that word-medial /r/ was more frequently tapped than word-initial /r/. Most descriptions of Classical Latin phonetics say that the primary phonetic realization and underlying phonemic representation of the rhotic consonant was an alveolar trill, which is uncontroversially transcribed as [r] in the IPA (Wireback 2015: 305, citing Catford 2001: 179; Allen 1978: 32; Pultrová 2013: 22). Trills can have different durations and numbers of contacts; however, any kind of trill would be appropriately expressed in IPA with the letter [r].

A minority of sources such as Pultrová 2013 do suggest that the primary phonetic realization or underlying phonemic representation of the Latin rhotic consonant was a tap/flap (i.e. /ɾ/) rather than a trill. However, because a tap consists of a single brief contact, a phonetically geminate tap is articulatorily implausible (Colina 2010: 78). In Spanish, a phonemically geminate tap /ɾɾ/ is sometimes hypothesized to be the underlying representation of a phonetic trill [r] (Recasens and Pallarès 1999: 146). A similar analysis could potentially be applied to Latin, with [rː] as the surface realization of /ɾɾ/. But I have not seen any source that clearly describes the geminate rhotic in Latin as phonetically consisting of a long tap [ɾː] (Pultrová seems to neglect to discuss the phonetic realization of the geminate); I strongly disagree with our use of this transcription, as it seems highly unlikely to me in comparison to [rː].

Aside from trills and taps/flaps, some sources mention other possible allophones of the rhotic consonant, such as a fricative realization (Allen 1978: 33, Wireback 2015: 305).

A single-contact (monovibrant) trill is phonetically similar to a tap or flap, but distinguished by the means in which it is produced (Wireback 2015: 301). If singleton /r/ had both tap and trill allophones, the conditions in which it was realized as a short trill versus a tap/flap seem to be not fully known and not exhaustively described in any one source, but in general appear to be as follows based on what I've read:

Contexts where a tap/flap realization of singleton /r/ was most likely:
 * in intervocalic position (preceded and followed by a vowel) in the middle of a word. The intervocalic /r/ in the word  is derived from /d/, which likely resulted from a phonetic change of a stop [d] to a tap [ɾ], with the latter reinterpreted as the phoneme /r/. The development of /d/ to /r/ in this word supports the reconstruction of [ɾ] as a common allophone of Latin intervocalic /r/ in the word-medial intervocalic context. (Wireback 2015: 306)
 * at the end of a word when the following word started with a vowel
 * as the final element of a syllable-initial consonant cluster, such as /pr/, /br/ or /tr/

Contexts where a trill realization of singleton /r/ was most likely:
 * in absolute utterance-initial position (Wireback 2015: 300)
 * at the start of a root after a consonant-final prefix (ab-, ad-, con-, in-, ob-, sub-). The prefixes ad-, con-, in-, and sub- have variant forms where the final consonant of the prefix was fully assimilated to the following root-initial /r/, producing a long trill [rː], as in,  ,  , and  . Heterosyllabic clusters ending in the trill [r], such as [b.r] or [n.r], are the most likely source of the [rː] found in these forms.
 * at the start of a word when the preceding word ends in a consonant (forming a heterosyllabic consonant cluster ending in /r/). It is possible that, as in prefixed words, some some such clusters could be assimilated to geminate [rː] (e.g. in prepositional phrases starting with ad per Wireback 2015: 307)

Contexts that seem fairly unclear:
 * in absolute utterance-final position
 * before another consonant sound (forming the first element of a heterosyllabic consonant cluster), either in word-medial position, or in word-final position when the following word started with a consonant. A few cases of syllable-final /r/ are derived from dissimilation of /n/ before the nasal /m/, as in and possibly, and some, mostly archaic forms exist with syllable-final /r/ derived from /d/ in ad-, such as arfuisse and arvorsum (Allen 1978, Pultrová 2013: 27), although the shift back to ad- spellings for most words could indicate that syllable-final /r/ became less similar phonetically to /d/. Per Wireback 2015: 321, the realization of syllable-final /r/ in Classical Latin may have been variable in a similar manner to the variable realization of pre-consonantal rhotics as taps or trills in Ibero-Romance.
 * at the start of a word when the preceding word ends in a vowel (possibly optionally lenited to [ɾ] per Wireback 2015: 300, 304, but possibly the frequency of [r] was greater compared to the word-medial intervocalic context per Wireback 2015: 303-304, 306), ([r~ɾ] per Wireback 2015: 315), "equally plausible" that word-initially a trilled [r] was more frequent even in post-vocalic sandhi contexts (Wireback 2015: 316)
 * in a morpheme-internal intervocalic consonant cluster that is a valid syllable-initial consonant cluster, but is optionally treated as heterosyllabic for the purpose of syllable weight, such as /br/ in when scanned with a heavy second syllable. As discussed above, heterosyllabic clusters involving originally word-initial /r/ appear to have been trilled, but originally word-medial clusters ending in /r/ don't show variant forms with assimilated /r.r/ in Latin.

Due to the apparently complicated nature of the allophony and lack of complete clarity, and the long-established convention of using [r] in IPA as a broad transcription of a tap/flap (no reverse convention to my knowledge exists of using [ɾ] as a transcription for a brief or monovibrant trill); I propose using the transcription /r/ in all contexts and adding a summary of the situation to Appendix:Latin pronunciation, which is linked from every transcription as the key. Here's my attempt; suggestions for improvement are welcome:

Allophones of /r/
"Latin has one rhotic consonant phoneme, transcribed as /r/ in the phonemic transcription used on Wiktionary, that may have had multiple phonetic realizations (allophones).

The realization of /r/ in Classical Latin was likely predominantly an alveolar trill [r] in at least the following two contexts: after a pause (in utterance-initial position), or when it was geminated (doubled) as /r.r/ [rː]. Most sources describe [r] as the primary pronunciation of Latin /r/.

A alveolar tap (or flap) allophone [ɾ] may have existed for singleton (non-geminate) /r/. A minority of sources describe [ɾ] as the primary pronunciation of Latin /r/.

There is disagreement about the overall frequency of tap versus trill realizations of /r/, but they may have varied based on on the position of word and morpheme boundaries, or on other factors such as speech style, rate, or prosody. Here is a summary of plausible tendencies to the distribution of different allophones of /r/:


 * Between vowels in the middle of a word, a tap may have been favored as the pronunciation of singleton /r/. At the start of a word, a trill allophone of /r/ was possibly favored even when the preceding word ended in a vowel.


 * After a consonant, a tap allophone was likely favored except for when /r/ was the first sound in a word, or had originally been the first sound in a word before the addition of a prefix. In some cases, the final consonant of a prefix could merge with a following originally word-initial /r/ to form the geminate /r.r/. The same assimilation of a final consonant to form geminate /r.r/ might also have been able to occur in some cases between word-initial /r/ and the final consonant of a preceding word, for example, in phrases such as "ad regem". Regardless of whether it was assimilated, a preceding word-final or prefix-final consonant was always pronounced in a separate syllable from the (originally) word-initial /r/, as in /sub.rogō/,  /ad.rogō/: this syllabification is shown by the heavy scansion in poetry of the first syllables of words and phrases of this form.


 * Before a consonant, there may have been variation between a trill and tap allophone, as in modern Spanish.

Because of the uncertainty in the distribution of the trill and tap realizations of /r/, and because the IPA recognizes [r] as a valid broad transcription of [ɾ], the phonetic transcriptions in Wiktionary entries do not differentiate between the possible allophones of Latin /r/, and use only the transcription [r]. Aside from trills and taps, other allophones such as fricatives may existed as possible pronunciations of the phoneme /r/."

However, I can certainly see the argument for the transcription that we previously had where intervocalic and post-consonantal /r/ was transcribed narrowly as [ɾ]. I would also support returning to that transcription, which I prefer over the current one.--Urszag (talk) 02:11, 17 September 2021 (UTC)


 * I agree that [ɾː] is dubious. It appears to be an oversight rather than an intended feature.


 * From what I have gathered, Wireback (2015) supports a tap realization, at least variably, of Latin /r/ in intervocalic and preconsonantal position (leaving aside, for the moment, the matter of sandhi contexts), not in postconsonantal position.


 * As mentioned by the same author, both Penny (2002: 94) and Lloyd (1987: 81) support a general tap realization, although neither provides a detailed argument to that effect. Pultrová (2013) presents a better case, but one that I, at least, find less convincing than Wireback's refreshingly detailed paper.


 * I have no objections to adopting the latter's model. A simplified one, with a uniform /r/ [r] in all environments, also seems acceptable.


 * The Nicodene (talk) 08:27, 18 September 2021 (UTC)
 * I didn't see anything in Wireaback (2015) about the pronunciation of Latin /r/ in complex onsets, and I couldn't find any other source discussing this particular topic, unfortunately. I think it's likely that a tap was frequent in this context based on the subsequent developments (the common presence of taps in complex onsets in Romance languages, the potential of metathesis between /r/ in complex onsets and in coda, or dissimilation or metathesis with /l/) and the fact that other modern languages with allophony between a trill and tap often seem to show high frequency of the tap after another consonant (Stolarski (2015) regarding Polish and Egurtzegi1 and Carignan (2020) regarding Basque, the latter contradicting descriptions of Basque allophony such as Bradley (2001) that say that the tap only occurs word-medially).


 * Since we both agree that uniform [r] is an acceptable transcription, I'm going to request for an admin to edit the module to use that.--Urszag (talk) 21:05, 19 September 2021 (UTC)


 * Per Canepari (2007: 131–132), the Italian singleton /r/ has the following distribution of allophones:
 * [r] in stressed syllables, whether before or after the nucleus.
 * [ɾ] in unstressed or half-stressed syllables.
 * Accordingly, he gives riprendere as [ɾiˈprɛːndeɾe] (ibid.).
 * Not all sources agree, necessarily, with Canepari, but I am not aware of any that posit a uniform /Cr/ [Cɾ].
 * Sardinian does show /Cr/ [Cɾ]. It presents, however, another complication: coda /r/ is realized as a trill before various consonants; cf. /arˈtista/ [arˈtiˑsˑta] (Mura & Virdis: 69, 128).
 * What the various sources and lines of evidence do seem to agree upon is a tap realization in intervocalic position. (Stress might still matter here, however; cf. Italian naturali [ˌnatuˈra:li], per Canepari 2007: 137). I think including that in the module, at least, would be fair. We could email Wireback and ask for his view on Latin /Cr/, etc. The Nicodene (talk) 23:48, 21 September 2021 (UTC)

Apparent logic error in transcription of /l/
Setting aside the disputes about how to transcribe l exilis and two-way vs. three-way contrasts, the current code incorrectly leaves out velarization entirely on word-medial /l/ before a consonant, e.g in (currently transcribed [ˈälɡä]). I’m pretty sure this is a logic error, since it’s undisputed that /l/ in this context was velarized (in fact, very likely to a greater extent than before any vowel).

The simplest solution would be, as in the previous version of the module, to start by defining dark [ɫ̪] as the general/“elsewhere” value of /l/, and then specify the contexts where it is lightened.

Specific fix:
 * Add {"l", “ɫ̪"} as the new line 160 (after current line 159 and directly before the current line 160)
 * Remove lines 168 {"l([eɛaoɔuʊ])", “ɫ̪%1"} and 169 {"l$", “ɫ̪"} which will be made redundant.

Can you confirm whether this is a logic error, since you're the one who made that edit. Another small question: did you intend to change the transcription of words like to [ˈɛs̠ːiː]? This contradicts Cser (2016), who specifically cites it as an example of a word with [ts] (4.3, page 73).--Urszag (talk) 02:14, 21 September 2021 (UTC)


 * Yes, that was unintended. Coda /l/ should be velarized.
 * The code in question is line 138, which I did not write. I imagine it was intended to apply to words like adsum.
 * The Nicodene (talk) 02:22, 21 September 2021 (UTC)
 * Thanks. Line 138 was changed from {"[d]([.ˈ]?)s", "s%1s"} to {"[dt]([.ˈ]?)s", "s%1s"} in your edit on 02:40, 22 July 2021‎; it previously applied only to prefixed words such as adsum and now applies also to the compound etsi.--Urszag (talk) 02:31, 21 September 2021 (UTC)
 * Ah. I don't remember why I did that. That can be removed. The Nicodene (talk) 02:34, 21 September 2021 (UTC)

Potential model for Ecclesiastical pronunciation
Canepari (2018: 14) comments that "Church Latin or International Latin (Italic, i.e.), is similar to Italian Academic Latin, and in fact it should be pronounced exactly like it." Earlier (p. 9), he provides a precise transcription of a passage in Italian Academic Latin. Translating his transcription into standard IPA, and simplifying the prosodic markers, it reads as follows:

ˈɔːlim ˌin̪teɾˈsɛ | ˈäːkwilo e̞tˈsɔlː | ˈuːte̞ɾ ˈfɔrːʦjoɾ ˈɛsːset | ʧeɾˈtäːbänt | ˌkumviäˈtɔˑɾem ˈkwɛn̪ːdäm | ˈpɛˑnulä äˈmik̟ːtum pɾoʧeˈdɛn̪ːtem ˌkon̪speksˈɛːɾun̪t ˈätːkwe ipˈsɔːɾum foɾˈʦjɔːɾem eɡˌzistiˈmän̪ːdum ˈɛsːse ˌkon̪sen̪ˈsɛːɾun̪t | ˈkwi efˈfiːʧeɾe̞t | ˌutviˈäːtoɾ ˈilːle̞ ˈpɛːnuläm deˈpɔːneɾe̞t || ˈäːkwilo ˈäˑu̯tem ˌveemen̪ˈtisːsime̞ ˈfuːɾeɾe̞ ˈʧɛːpit | ˈsɛdː kwo̞fo̞ɾˈʦjɔːɾes ˈfläːtus ˌemitˈtɛːbät | e̞oˈärːʦjus viˈäːtoɾ ˌse̞ʧiɾˈkumːdäbät ˈpɛːnulä | ˈtänːdem ˈviˑɾibus ˌdestiˈtuːtus pɾoˈpɔːzitum ˈsuˑum oˈmiːzit | ˈtumː ˈsɔlː ʧɛːlum kläˈrissimä ˈluːʧe̞ illusˈträːvit | ˈmɔkːs ˈvɛːɾo | viˈäːtoɾ käˈlɔˑɾe ˈvikːtus | ˈpɛːnuläm ˈɛɡːzwit || ˈiːtäkwe ˈäːkwilo ˈkwämvis iɱˈviːte̞ | koɱˈfɛsːsus ˈɛsːt ˈsɔːlem ˌe̞ssefoɾˈʦjɔːɾem ||| ˈtiˑbi ˈpläːkwit ˈfäːbulä | liˈbɛtne ˌe̞ämɾeˈpɛːteɾe ||

For comparison, here is the text in normal spelling:

Olim inter se Aquilo et Sol uter fortior esse certabant, cum viatorem quendam paenula amictum procedentem conspexerunt; atque ipsorum fortiorem existimandum esse consenserunt, qui efficeret ut viator ille paenulam deponeret. Aquilo autem vehementissime furere coepit; sed, quo fortiores flatus emittebat, eo artius se circumdabat paenula; tandem, viribus destitutus, propositum suum omisit. Tum Sol caelum clarissime luce illustravit; mox vero viator, calore victus, paenulam exuit. Itaque Aquilo, quamvis invite, confessus est solem esse fortiorem. Tibi placuit fabula? Libetne eam repetere?

Perhaps this could serve our purposes.

The Nicodene (talk) 06:20, 27 September 2021 (UTC)


 * It could be useful. I'd prefer not to use it as the sole source for a transcriptional feature. Canepari's transcriptions are not conventional, and he publishes them along with extensive descriptions of what they're intended to indicate. We obviously can't do that kind of presentation, and even though you mentioned converting it to IPA, I still find the transcription above a bit too complicated in places. In particular, the length markers on vowels (distinguishing long vowels in äːkwilo, pɾoˈpɔːzitum, ˈpɛːnuläm from half-long vowels in viäˈtɔˑɾem, ˈviˑɾibus, käˈlɔˑɾe) and coda consonants (distinguishing sːs with both gemination and a length marker on the coda portion of the geminate in ˈɛsːset, ˌveemen̪ˈtisːsime̞ versus shorter? geminate in kläˈrissimä, ˌe̞sse) seem excessively detailed and I'd guess subject to prosodic variation, which we can't account for in a standard dictionary entry that gives the isolated pronunciation of a single word. I'd prefer to simply mark vowels as long or short, and to not mark lengthening of coda consonants. I'm also not in favor of us transcribing three levels of allophonic vowel height, although maybe we could use a two-way distinction in transcription such as stressed [ɛ(ː)ɔ(ː)] vs. unstressed [eo] or stressed [ɛ(ː)ɔ(ː)] vs. unstressed [ɛ̝ɔ̝] to indicate that unstressed /e/ and /o/ are likely to be closer than the stressed vowels. Something Canepari shows that I've suspected should maybe be changed here on Wiktionary is non-assimilation of coda /m/ to the place of a following consonant in Ecclesiastical pronunciation.--Urszag (talk) 06:55, 6 October 2021 (UTC)
 * Agreed that several of the finer distinctions are best left unspecified, at least unless/until other sources can be found to support them.
 * Unstressed [eo] for [ɛɔ], on the other hand, would be a clear improvement and quite uncontroversial. I think that could be done now. The Nicodene (talk) 07:24, 15 October 2021 (UTC)'
 * I think that unfortunately, it is not uncontroversial to change the transcription of unstressed /eo/ to [eo] (versus stressed [ɛɔ]).

Prescriptive resources such as singers’ guides often give only a single pronunciation for E O, and many of these make it clear, either by explicit statement or by choice of examples, that this is meant to apply in all positions, including both stressed and unstressed syllables:
 * SUNG ECCLESIASTICAL LATIN (ROMAN) PRONUNCIATION GUIDE (KU CONCERT CHOIR): "e = eh as in [English] met", examples "te, video"
 * PRONOUNCING CHURCH LATIN: A QUICK REFERENCE, Ideals and Styles in the Western Musical Tradition, Douglass Seaton: "e is pronounced as in egg: ante (ahn-teh)"
 * Ecclesiastical Latin Pronunciation, CanticaNOVA Publications: æ = "eh", "pronounced exactly like Latin E sound", "example æterna"
 * https://en.wikibooks.org/wiki/Ecclesiastical_Latin/Vowels
 * http://lyricchorale.org/wp-content/uploads/Latin.pdf
 * https://learnchurchlatin.com/2019/05/20/ecclesiastical-latin-pronunciation-guide/

Since singing is a significant portion of contemporary use of what is described as “Ecclesiastical” Latin, and there does not appear to be a definite consensus among singers/directors on what vowel qualities to use for /e o/ in sung Latin, I’m unsure whether we ought to indicate one particular system of allophony that may not even be more common than the invariant [ɛ ɔ] system.

There may even be other, alternative systems of allophony: the thesis "A Musician’s Guide to Latin Diction in Nineteenth and Twentieth Century Choral Repertoire", by Sean Taylor (2013), (which seems dubious in its transcriptions in various places), contains a quotation (page 19) saying the following bizarre system is use: "In Singer’s Liturgical Latin, Hines says that “many American conductors and singers advocate that /e/ is to be used exclusively for written E except when it is followed by T, ST, X, or R.” In those cases, /ɛ/ is used." (citing Robert S Hines' Singers’ Liturgical Latin)

Canepari is obviously a much more phonetically sophisticated source than the references above, but Canepari is still only a single person, and it's not entirely clear the degree to which he is engaging in prescription (as opposed to description) himself. The wording "is similar to Italian Academic Latin, and in fact it should be pronounced exactly like it" raises the question of whether the transcription with [e] and [o] is based on direct experience of actual Ecclesiastical Latin usage, or based on his knowledge of the allophonic rules of normative Italian pronunciation, plus a belief that it is logical and therefore preferable to apply the same rules when pronouncing Latin. It is difficult to find detailed and high-quality descriptive sources on Ecclesiastical Latin. Do you know of any source other than Canepari that supports the system with conditional use of [ɛ ɔ] in stressed syllables and [e o] in unstressed syllables in Latin?--Urszag (talk) 20:21, 15 October 2021 (UTC)


 * The module currently encodes for numerous Italianizing features that are not specified in the Anglophone singing-manuals. We have also changed [ɑ] to [ä], contrary to what the same manuals generally prescribe, in order to reflect the speech of Italian clergy.
 * It seems that we need to decide what sort of Ecclesiastical we want to represent. If we opt for choral Anglo-Ecclesiastical, then Italian features not mentioned in the manuals should be removed, such as the stressed open syllable lengthening rule. If we choose Italo-Ecclesiastical, then Italian phonotactics have to be considered, including the unstressed [ɛ ɔ] > [e o] rule.
 * If we opt for the latter, I will look for other sources describing its pronunciation. I do not know of any offhand. The Nicodene (talk) 03:46, 16 October 2021 (UTC)
 * I agree that it seems likely that the Italian vowel-lengthening rule would go together with the Italian unstressed-mid-vowel-raising rule, but judging from GianWiki's confusing comments on the la-IPA discussion page, that isn't necessarily completely obvious to native Italian speakers. I'm trying to see now what I can find written in Italian about Latin pronunciation. I've never heard anyone refer to a distinct "Anglo-Ecclesiastical" pronunciation tradition and I don't believe the guides linked above intend to position themselves in that space, as opposed to guides to the Roman/Ecclesiastical tradition for English speakers. I would consider Canepari a better source than all of the English sources linked there put together; however, I still would prefer to have a second source before making changes.
 * I also think it's best to figure out all of the other changes needed and revise the test cases accordingly before making/requesting module edits. So in the meantime, I wanted to ask what your view is on the transcription of vowels, hiatus and diphthongs for Ecclesiastical Latin. Canepari notably transcribes semivocalization of certain cases of unstressed i and u (ˈfɔrːʦjoɾ, ˈɛɡːzwit, ˈpläːkwit) although also hiatus in some positions (viˈäːtoɾ). My understanding is that in practice, an Italian speaker would usually use a semivowel in the positions Canepari indicates in normal speech, but that a syllabic variant would be an option (in Gregorian chant, unstressed i and u correspond to a separate note from the following vowel). This is a case where I would actually see a potential advantage to a transcription making use of [i̯ u̯] rather than [j w], since the phonetic glides in Ecclesiastical Latin seem to represent allophones of the vowel phonemes /i/ and /u/. I found an Italian blog post ("UNA BREVE STORIA DELLA PRONUNCIA ECCLESIASTICA DEL LATINO") that mentions that gliding in stressed-unstressed vowel sequences is also commonly found ("*Dovrebbe pronunciarsi in due sillabe: De-us, De-um. Nella pratica si sente pronunciare con un dittongo, [dɛus] e [dɛum], per quanto in Wiktionary si riporti soltanto con iato, addirittura come ['de:us].")
 * There's also the matter of Italian's inherently geminate consonants [t͡s d͡z ʃ ɲ], which are usually described as always geminate between vowels in Italian, but which we currently transcribe as single for [t͡s ʃ].--Urszag (talk) 07:06, 16 October 2021 (UTC)
 * By 'Anglo-Ecclesiastical' I mean the pronunciation that Anglophones would settle on when strictly following the official manuals written for them. Although it is an imitation—in broad strokes—of the Italian pronunciation, and not intended to be distinct from the latter, it is inevitably different in practice.
 * An example would be the prescription of a uniform pronunciation for ⟨e⟩ and ⟨o⟩ in all environments. That presumably originates in the identical advice that L2 learners of Italian are often given (Lorinczi 1996: 20–21). A single pronunciation for ⟨e⟩ and ⟨o⟩ is common in regional Italian accents, so it is regarded as relatively acceptable, or at least still 'Italian' in a sense. For Anglophones in particular, it would make sense to prescribe [ɛ ɔ], since the nearest alternative would be the jarringly un-Italian* [eɪ oʊ~əʊ].
 * * un-Standard-Italian, that is. Regional languages may have the sounds in question.
 * As for the 'autogeminating' consonants, these can be expected to remain so in Italo-Ecclesiastical, as with other phonotactic rules. No example is to be found in the paragraph trancribed above, but Canepari provides [ˈɔʦːʦjum], [ˈpiʃːʃe̞m], [ˈliɲːɲum] (op. cit.: 13).
 * Viator with [i] might be due to the first morpheme also existing as an independent word with a stressed /i/, or it might be due to the morpheme boundary. I'll see if I can get in touch with the author and ask if this is the reason. Incidentally, he transcribes the Italian viaggiatore with [i ] in another work.
 * Edit: Well, that did not take long. It seems that the cause is indeed morphemic. I continue to be surprised by the willingness of academics to reply to random emails.
 * It is unfortunately proving difficult to find detailed non-Caneparian sources. The Nicodene (talk) 21:21, 16 October 2021 (UTC)
 * I see, thank you for checking. I would support changing the transcription of unstressed /e o/, and I've edited the sandbox testcases. I've also added gemination of [t͡s ʃ] and assimilation of /nf/ > [ɱf]. Based on the response you received from Canepari, I feel like it would be more trouble than it's worth to attempt to transcribe semivowel allophones of /i/ and /u/, since the module does not normally have access to the morphological structure of a word, which would be needed to determine when it is appropriate. Could you check and see if there's anything else that should be added to the testcases?--Urszag (talk) 00:02, 17 October 2021 (UTC)
 * Perhaps the input for that word can be changed to vi-ātor, and the module can be programmed to subtract the hyphen at certain points (so that CL ɪ > i applies, but EL i > j does not). A morpheme boundary indicator might prove useful in other cases as well.
 * Looking at the test cases, I see that CL nasalized /a/ has an unfortunate diacritic conflict (ä̃). The centralization diacritic may not really be necessary in general; Wiktionary's transcriptions of Spanish, for instance, opt for a simple [a], and the practice is common elsewhere as well. (Mahāgaja's comment comes to mind.) On the other hand, it is not a bad thing to be precise, and perhaps another solution can be found for ä̃, or it could be tolerated as-is.
 * Nothing comes to mind regarding adding more testcases. Incidentally, I am in favour of your earlier proposal for syllable-initial [j w] in the CL transcriptions. The Nicodene (talk) 05:45, 17 October 2021 (UTC)
 * Thanks for reviewing. The biggest problem I see in the way of differential transcription of CiV and CjV is the difficulty of doing it consistently. I also question whether, even if we can do the work to mark the distinction consistently, any simple system will accurately describe the actual usage that can be expected to be produced by Italian speakers. Considering just one specific case, like viator, it would be fairly simple to set up a system to display a special pronunciation with hiatus. But are we really going to be able to go through all words and correctly mark them for morpheme boundaries? In some cases, whether a relevant morpheme boundary is present may not even be obvious: trivium is also a derivative of via, but is that etymology relevant, or should it get pronounced like a simple word? (Canepari's Italian dictionary marks the pronunciation of Italian trivio as ˈtrivjo, -io). Or, given that modern Italian is pronounced with a glide (Canepari gives the pronunciation only as pjeˈta), is an Italian speaker actually going to never produce a glide for unstressed i in Latin forms like pietātem? I suspect that in at least some cases (and possibly in many cases), there will only be tendencies towards the use of a syllabic or non-syllabic pronunciation rather than strictly followed rules.
 * Thanks for reviewing. The biggest problem I see in the way of differential transcription of CiV and CjV is the difficulty of doing it consistently. I also question whether, even if we can do the work to mark the distinction consistently, any simple system will accurately describe the actual usage that can be expected to be produced by Italian speakers. Considering just one specific case, like viator, it would be fairly simple to set up a system to display a special pronunciation with hiatus. But are we really going to be able to go through all words and correctly mark them for morpheme boundaries? In some cases, whether a relevant morpheme boundary is present may not even be obvious: trivium is also a derivative of via, but is that etymology relevant, or should it get pronounced like a simple word? (Canepari's Italian dictionary marks the pronunciation of Italian trivio as ˈtrivjo, -io). Or, given that modern Italian is pronounced with a glide (Canepari gives the pronunciation only as pjeˈta), is an Italian speaker actually going to never produce a glide for unstressed i in Latin forms like pietātem? I suspect that in at least some cases (and possibly in many cases), there will only be tendencies towards the use of a syllabic or non-syllabic pronunciation rather than strictly followed rules.


 * The diacritic stacking with ä̃ seems tolerable to me. I would not mind dropping the centralization diacritic, but Stick Daze and Brutal Russian indicated a preference for using it.


 * I hoped that more people would comment on this page or on the la-IPA page about [j w]. Currently, you and I prefer [j w], but Brutal Russian evidently prefers [i̯ u̯]. Further up on this page, PastelKos prefers [w], but Draco argenteus prefers [u̯]. I was waiting to get a larger majority in favor, but the discussion appears to have died down.--Urszag (talk) 07:15, 17 October 2021 (UTC)

In the course of preparing a revision that adds unstressed [e o], intrinsic geminate [t͡s ʃ], as well as a few other small changes, I found that I don't know what transcription the module is supposed to generate for the phonetic Ecclesiastical Latin pronunciation of words with stressed falling diphthongs like nauta, auctus, causticus, neu. Currently, it generates a diphthong with a long initial component in stressed open syllables (such as [ˈnäːu̯t̪ä]) but short in closed syllables ([ˈäu̯kt̪us]). I'm not sure this is correct. In terms of rules, that could conceivably be correct if diphthongs count as one unitary vowel for the purpose of vowel lengthening, but with phonetic lengthening applied only to the first element. However, alternative rules that also seem a priori plausible are a) diphthongs behave like /ˈVC/ sequences and so are not transcribed with lengthening ([ˈnäu̯t̪ä]: a previous talk page discussion implied that this is expected behavior for this module), or b) diphthongs behave like /ˈVV/ sequences, and so lengthening applies even when there is a coda consonant ([ˈäːu̯kt̪us]) because the coda consonant doesn't come directly after the stressed vowel element of the diphthong. The only example in Canepari's passage is ˈäˑu̯tem, but I haven't looked at what Canepari says about Italian.

Italian has pretty much no words with falling diphthongs in closed syllables: the only examples I can find, which are rare, are words with "au" before an s-cluster (s-cluster syllabification in Italian is disputed). The only stressed example I found so far is. To me, the pronunciation on Forvo sounds like the [a] element is fairly long.--Urszag (talk) 19:09, 19 October 2021 (UTC)


 * /ˈau/ seems to behave like any other /ˈVV/ sequence, at least with regards to vowel lengthening. So autem [ˈäˑutem] as also suum [ˈsuˑum] (1), like the Italian mai, ciao, boa, causa [ˈmäˑi, ˈtʃäˑi, ˈboˑä, ˈkäˑuzä] (2 3). (The half-lengthening is also noted by Martin Kramer in The Phonology of Italian, page 53.) Canepari never marks the latter element of a falling diphthong as non-syllabic, as you can see, and this time I have not either.
 * Some other Italian words with /ausC/ on Forvo are fausto, olocausto, austro, esauste, and auspice. A general lengthening of the a may be heard, as in the recording that you provided. This makes sense if /au/ here is really /a.u/; the first vowel then is in an open syllable. Kramer (p. 136) discusses cases like these, commenting that '...the preferred realization breaks up the offending sequence /aus/ into two syllables.' The Nicodene (talk) 00:32, 23 October 2021 (UTC)
 * Those additional examples are useful, as is the reference to Krämer. He seems to be saying on page 136 only that the specific word has a realization with hiatus (and without lengthening): [balaˈustɾa] (preferred over [baˈlawstɾa]). He doesn't list equivalent pronunciations with hiatus for Austria, Australia, austro (instead, he appeals to the fact that they are not "ordinary" nouns). Unfortunately, he also doesn't provide any further support for the use of ˑ in this context, as he transcribes [ˈawstɾo] and [ˈawstɾja].--Urszag (talk) 02:03, 23 October 2021 (UTC)
 * He renders pausa and causa without a half-long diacritic on page 88, despite earlier transcribing them with the diacritic on page 53. This seems to be a case of transcriptions varying in precision, depending on what the author wants to focus on at the moment.
 * He does give ˈawstɾja and ˈawstɾo. Following the comment below, this would be the less common (non-'preferred') pronunciation. His comment about /aus/ being split into two syllables would imply hiatus. At least, I do not see another way to interpret it.
 * Perhaps best to email him for clarification? The Nicodene (talk) 03:30, 23 October 2021 (UTC)
 * Sure, I'll email him to ask about this. I'm reading the first clause in that sentence as specifying a condition: "In the only form listed here that is an ordinary noun, the preferred realization breaks up the offending sequence /aus/ into two syllables." Given that he explicitly lists [balaˈustɾa] (before [baˈlawstra]) in (17) b., I don't see why he would not list [aˈustɾja], [ausˈtɾaːlja] [aˈustɾo] in (17) a. if he had intended to say that they all existed. Urszag (talk) 03:42, 23 October 2021 (UTC)
 * I read his comment as implying [ˈa.ustɾo] (ignoring vowel length), with stress on the /a/.
 * He does imply that this phenomenon does not apply to the other two words. My mistake.
 * The realizations of /au/ in Austria and austro sound similar to me, incidentally.
 * If you do email him, perhaps you could also ask about the syllabification in the words listed earlier? (Fausto, etc.)
 * The Nicodene (talk) 04:13, 23 October 2021 (UTC)
 * On second thought, I am starting to think it more likely that the split that Kramer was describing is /au.s/ rather than /a.us/, despite his phrasing ('split into two syllables', not *'split across two syllables'). Was that your interpretation? The Nicodene (talk) 06:24, 25 October 2021 (UTC)
 * For balaustra, there is a pronunciation where au is split into two syllables, which is clearly indicated by the position of the stress on the penultimate syllable of the word: [balaˈustɾa]. I received a reply to the email I sent to Krämer. He says that he recommends collecting more data and performing measurements on the length of the vowels. He is not aware of the austr- words showing any variation of the type that he describes for balaustra, and he said that Austria doesn’t sound to him like it is syllabified as ‘a.us.trja. For further reading, he suggested Marco Bertinetto’s article “On the undecidable syllabification of /sC/ clusters in Italian: Converging experimental evidence” (2004). As the title hints, that article doesn't support much optimism about the answerability of questions like "is Austria syllabified as Aus.tria or Au.stria?"--Urszag (talk) 16:37, 30 October 2021 (UTC)
 * Thank you for confirming.
 * The syllabification of /-sC-/ does seem to be a vexata quaestio in Italian studies.
 * For /ˈau/ in open syllables, I think we can go with [äˑu̯], following Kramer and Canepari (and our own perception of lengthening in the recordings). The module's current [ˈäːu̯] may also work, if you do not want to specify different levels of lengthening. Nicodene (talk) 00:48, 10 November 2021 (UTC)

I've emailed Krämer about these topics and will give an update when I receive a response. Meanwhile, another topic that I would like to get resolved now is the problem of how to transcribe and syllabify "quu" and other "uu" sequences in Italian/Ecclesiastical Latin (something Brutal Russian brought up in May 2020). Italian, per Krämer, has no *[wu] (page 54). (In fact, it has no [u.u] either, acording to "Durations of Italian Diphthongs and Vowel Clusters", Pier Luigi Salza, 1988, Language and Speech, Vol. 31, Part 2, page 100.) I've found a passage that confirms that Italians confronted with this graphical sequence naturally pronounce it like "cuu", presumably as [kuu]: "la labiovelare sorda qu (ku̯) non pone problemi di pronunzia agli Italiani (quattuor come quattro, qui come qui), tranne che davanti a u: equus, sequuntur, etc. Noi pronunziamo ecuus, secuuntur, con due u, ma a torto." (§11 "Il gruppo quu", page 62, Propedeutica al Latino Universitario, Alfonso Traina & Giorgio Bernardi Perini). Currently, the module generates 	/ˈe.kwus/ [ˈɛːkwus]. And in contrast, the module generates a long vowel [uː] in /abˈpa.tru.us/ [äbˈpäːt̪ruːs], something that I'm also not sure it should be doing.--Urszag (talk) 02:24, 25 October 2021 (UTC)


 * Agree regarding equus. [u(.)u] would be best, if the module can be made to output that rather than [uː]. The Nicodene (talk) 06:16, 25 October 2021 (UTC)

Absence of voicing assimilation in Italian/Ecclesiastical Latin
There are sources that say that Italian does not have productive assimilation of voicedness or voicelessness in general between obstruents (except the special case of [s]~[z]). Although clusters of obstruents with different voicing do not appear in native Italian vocabulary, they occur in loaned or borrowed words, including words taken from Latin, so I assume the same lack of voicing assimilation would also apply to "Ecclesiastical Latin". "How can Italian phonology lack voice assimilation?", by Bálint Huszthy (2019) provides the Latinate examples obcordate, subtropicale and subcultura; the last is transcribed as [subkulˈtuːɾa] (46).

Based on this, I assume voiced [b] [d] occur in Italian pronunciation in all words containing a prefix spelled with b or d, despite the evidence we have from Ancient Roman grammarians that describes /b/ assimilating to [p] in bt, bs clusters in Classical pronunciation (and so presumably also in other clusters ending in a voiceless obstruent such as bc, although I'm not sure whether any ancient source explicitly states the general condition for devoicing). I'd assume the absence of assimilation in Italian pronunciation also applies to word-final bs as in urbs, although I couldn't find an exactly analogous example in the linked source.--Urszag (talk) 01:26, 23 October 2021 (UTC)

Classical /s/
The three sources cited in support of a retracted CL /s/ all describe an apico-alveolar realization. (As do others.) Accordingly we would have [s̪] rather than the less-precise [s̠]. If a broad transcription is desired (due to uncertainties), [s] would do. The Nicodene (talk) 06:48, 25 October 2021 (UTC)
 * I just tried to look into this and was unable to understand the phonetic details involved in the notation of these sibilants. I'm not convinced yet the transcription needs to be changed. You indicate that [s̠] is "less precise" than [s̪], and then indicate that you don't object to a broad transcription as [s]: what then is wrong with a broad-ish transcription as [s̠]? Is it accurate to say that [s̪] falls within the range of [s̠]? I do see that the sources that are currently listed as sources describe the sound as "apico-alveolar", and at least one uses the transcription [s̠]. However, the Wikipedia article "Voiceless alveolar fricative" seems to indicate there is uncertainty or disagreement about which exact articulatory elements characterize "intermediate"-sounding sibilants:
 * "There is no general agreement about what actual feature distinguishes these sounds. Spanish phoneticians normally describe the difference as apical (for the northern Iberian sound) vs. laminal (for the more common sound), but Ladefoged and Maddieson[6] claim that English /s/ can be pronounced apically, which is evidently not the same as the apical sibilant of Iberian Spanish and Basque. Also, Adams[7] asserts that many dialects of Modern Greek have a laminal sibilant with a sound quality similar to the "apico-alveolar" sibilant of northern Iberia. Some authors have instead suggested that the difference lies in tongue shape. Adams[7] describes the northern Iberian sibilant as "retracted". Ladefoged and Maddieson[6] appear to characterize the more common hissing variant as grooved, and some phoneticians (such as J. Catford) have characterized it as sulcal". (Voiceless alveolar fricative: Comparison_between English and Spanish)
 * As a practical matter, I doubt non-expert users of this site will be familiar enough with either of these diacritics to be mislead by the use of [s̠] vs. [s̪] or vice versa; and I would guess expert users who are familiar with the theoretical distinction would also be aware of the notational variability that apparently exists in practice.--Urszag (talk) 09:04, 29 October 2021 (UTC)
 * A broad transcription would serve to encompass realizations other than [s̺]. The only other ones that have been proposed for Latin /s/, as far as I am aware, are [s] and [s̪], neither of them retracted. Since ⟨s̠⟩ cannot stand for them, a broad transcription for our purposes would rather be [s].
 * If we commit to a narrower transcription (while rejecting [s] and [s̪]), then [s̠] would be misleading, as it suggests a range of retracted realizations, when the only one that has been proposed by scholars, as far as I am aware, is [s̺].
 * Here are various other sources' comments on Latin /s/:
 * “Whether the coronals [t d s n l r] were dental or alveolar is difficult to establish with certainty…” (Cser 2020: 13).
 * “…The northern dialects [of Portuguese] have kept their Latin apico-alveolar fricatives, as is also the case in northern Castilian..." (Mateus & d'Andrade 2000: 2).
 * "...it is likely that the Latin s of many areas was apico-alveolar…” (Penny 2002: 94).
 * Cser, András. 2020. The phonology of Classical Latin. Transactions of the Philological Society 118. 1–218.
 * Mateus, Maria Helena & d'Andrade, Ernesto. 2000. The phonology of Portuguese. Oxford University Press.
 * Penny, Ralph. 2002. A history of the Spanish language. Cambridge University Press.
 * Nicodene (talk) 00:34, 10 November 2021 (UTC)
 * Penny, Ralph. 2002. A history of the Spanish language. Cambridge University Press.
 * Nicodene (talk) 00:34, 10 November 2021 (UTC)
 * Nicodene (talk) 00:34, 10 November 2021 (UTC)

Is there a point?
@Mahagaja, @Urszag, @Erutuon: The more I think about it, the more I question the need to have detailed phonetic transcriptions for a dead language, particularly for an online dictionary. Latin is the only one for which Wiktionary has attempted such a venture—very few living languages even get narrow transcriptions—and a sea of ink has been spilt here about the finer details, few of which have been worked out to any satisfaction. (Final /-Vm/ is only one of many unopened cans of worms.) It is telling that no scholarly source has attempted a comprehensive phonetic sketch of Classical Latin, and I fear that we may be on a fool's errand here.

What do you think? Might it be best to switch back to phonemic transcriptions, leaving discussion of phonetic intricacies for the Wiki page? Nicodene (talk) 11:34, 19 December 2021 (UTC)
 * Apologies for my delayed responses to your posts. I no longer find it worthwhile on a personal level to dedicate attention to this project, so I have no opinion either way about removing phonetic transcriptions from Latin Wiktionary entries.--Urszag (talk) 14:27, 19 December 2021 (UTC)

⟨uv⟩ narrow transcription bug
There is a bug in the current implementation where the string ⟨uv⟩ produces [uː̯] (that is: a 'u', a 'ː' and then a '̯'). An example can be seen here. I'll try to fix it myself and add a comment if I've succeeded. --LonleyGhost (talk) 14:27, 4 February 2023 (UTC)


 * Ok, since the module is edit-locked I cant fix this. Someone higher up the food chain will have to do this. --LonleyGhost (talk) 14:36, 4 February 2023 (UTC)