Module talk:sa-pronunc

Layout change

@Aryamanarora , just curious, why does the classical sanskrit pronuciation now appear twice? 2405:204:9387:50DE:0:0:84F:70A0 02:03, 15 June 2017 (UTC)

@2405:204:9387:50DE:0:0:84F:70A0: I tried to make the template match visually with {{zh-pron}}, {{bo-pron}}, {{ko-IPA}} etc. because they've been around for a while and so are pretty standardized, and they're good for maintaining IPA for many dialects. So if we ever decide to add day modern liturgical Sanskrit pronunciations or different romanizations, or even pre-Vedic reconstructed pronunciation we can do it easily. Do tell me if it looks bad though, I can revert it. —Aryaman ^{(मुझसे बात करो)} 02:06, 15 June 2017 (UTC)

@Aryamanarora: oh i see. No, it doesn't look bad and if it'll help document an Old Indo-Aryan dialect parallel to Vedic, it's certainly something that should be there. 2405:204:9387:50DE:0:0:84F:70A0 02:12, 15 June 2017 (UTC)

@Aryamanarora If you would like to provide descriptions for the (multiple) liturgical pronunciation systems, and the "Old Indo-Aryan dialect parallel to Vedic", I can implement those. —JohnC5 16:14, 15 June 2017 (UTC)

@JohnC5: Mainly, modern liturgical Sanskrit does not have a syllabic r (instead there's /ɾV/) the final ः is sometimes dropped. All the diphthongs are merged into monophthongs except in Uttar Pradesh and South India where they remain diphthongs. As for vedic, is there not also the svarita accent that needs to be handled by the module? We could also try to reconstruct where hiatus occurs as a result of the laryngeal in Proto-Indo-Iranian. —Aryaman ^{(मुझसे बात करो)} 17:14, 15 June 2017 (UTC)

@Aryamanarora: Ok, could you list the particular liturgical accents you'd like to see and what their respective realizations of ṛ and the diphthongs are? As for the svarita, it was not phonemic at the time of the composition of the Vedas (there was a phonetic falling tone). The independent svarita occurs the composition and the transcription. Which are we representing here: the theoretical composition stage or the later transcription stage? Also, do you know of a thorough write up of the svarita? —JohnC5 18:08, 15 June 2017 (UTC)

@JohnC5: For ऋ, ऐ, औ, अ:

• (North India) IPA^(key): /ɾɪ/, /ɛː/, /ɔː/, /ə/
• (Maharastra) IPA^(key): /ɾu/, /ɛː/, /ɔː/, /ə/
• unchanged for all, except अ is /ɑ/

There are some more differences but I haven't been able to compile a complete list. As for Bengali Vedic chanting, I have no idea what it sounds like off the top of my head. As for svarita, I have not been able to find much but [1] seems promising. —Aryaman ^{(मुझसे बात करो)} 19:56, 15 June 2017 (UTC)

@Aryamanarora: I will look into this. And any information you can provide is much appreciated. Two questions: the liturgical pronunciations are based off of the Classical ones, right? And should I account for the independent svarita or not in Vedic? —JohnC5 05:15, 16 June 2017 (UTC)

@JohnC5: The liturgical pronunciations are based on Vedic Sanskrit, they're most used in Vedic chanting. I think svarita is a result of Sandi, not an independent occurrence, so maybe it isn't necessary? —Aryaman ^{(मुझसे बात करो)} 13:02, 16 June 2017 (UTC)

@Aryamanarora: Ok, I've added the sannatara-udātta-svarita (low-rising-falling) functionality to Vedic as seen here using this code. Feel free to mess around with them. I will add the liturgical dialects, but I was wondering whether you had any sources for these pronunciations you're giving (no offense). I would like to make accent labels for them in Module:accent_qualifier/data, but I'd like to know a bit more about them before doing so. Also, there is an independent svarita (not following an udātta which came from an elision or coalescence of a udātta-bearing vowel). The independent svarita did not appear in Vedic, but does in liturgical recitation, so I'll add another parameter for it. —JohnC5 04:23, 17 June 2017 (UTC)

@JohnC5: That's amazing! Unfortunately, the liturgical pronunciation is based off of my own experience, there's almost no easily accessible literature about dialectical differences in Vedic chanting. We should hold off on that I think until we have something more reliable than a Quora answer. —Aryaman ^{(मुझसे बात करो)} 13:50, 18 June 2017 (UTC)

@Aryamanarora: That Quora answer is fairly regionalist and unscientific for my taste. I would agree that we should find a better source. —JohnC5 22:17, 18 June 2017 (UTC)

@Aryamanarora: So, I've engaged Madhav Deshpande and Peter Scharf in conversation. They have recommended the work of Wayne Howard for modern recitation practice. If you're up to it, I might recommend reading Chapter 5 of this paper. I'm thinking about breaking up the phonetic representation of Vedic by scholarly tradition:

1. Śākalasaṁhitā of the Ṛgveda
2. Vājasaneyisaṁhitā of the Śuklayajurveda
3. Taittirīyasaṁhitā of the Kṛṣṇayajurveda
4. Śaunakīyasaṁhitā of the Atharvaveda
5. Maitrāyaṇīsaṁhitā of the Kṛṣṇayajurveda, Ṛgveda khilāni, and Kashmiri mss. of #2
6. Kāṭhakasaṁhitā of the Kṛṣṇayajurveda
7. Paippalādasaṁhitā of the Atharvaveda
8. Sāmavedasaṁhitā in the Kauthuma śakhā
9. Śatapathabrāhmaṇa

It's gonna take some more research before I can get the liturgical stuff started. —JohnC5 17:08, 19 June 2017 (UTC)

@JohnC5: I've started reading it, and wow, I did not know about the depth of Sanskrit phonological studies. And yeah, it would be better to detail different scholarly traditions instead of regional variation since there's a wealth of information about them. The liturgical stuff can wait. —Aryaman ^{(मुझसे बात करो)} 17:22, 19 June 2017 (UTC)

Collapsible

@Aryamanarora, JohnC5 I object to it being collapsible. I think it looks bad, is unnecessary, and is at odds with other templates of its kind. --Victar (talk) 02:25, 5 September 2017 (UTC)

@Victar: Excusing {{grc-IPA}} and every template used for Chinese. I could perhaps be conviced to remove the functionality. —JohnC5 06:09, 5 September 2017 (UTC)

@JohnC5: {{grc-IPA}} actually looks a lot better -- {{sa-IPA}} looks like garbage with that bounding box. How many forms are being displayed in {{sa-IPA}} on average? If it's just two in most/all cases, that it hardly seems merited, and more akin to {{la-IPA}}. --Victar (talk) 06:41, 5 September 2017 (UTC)

@Victar: Oh, I had forgotten that Aryamanarora added the bounding box. I kinda like it more with the box, to be honest. The intent will be for more reconstructed and recitational pronunciations, but I'm working on Sanskrit declension stuff at the moment. No offense, but I'll probably leave it is for now. —JohnC5 08:20, 5 September 2017 (UTC)

@Victar: I did it because I was hoping we could more dialects eventually (esp. liturgical Sanskrit of today). The following languages use similar boxes: Korean, Japanese, Tibetan, Chinese, and I guess Ancient Greek. —Aryaman ^{(मुझसे बात करो)} 12:48, 5 September 2017 (UTC)

@Aryamanarora: You did what, make it collapsible or add the bounding box? Have you seen how it looks for Greek? It looks much better than this jarring and out of place blue box. --Victar (talk) 14:55, 5 September 2017 (UTC)

@Victar: I added the bounding box. Also "blue"? It's a light gray on my screen... IMHO, it looks nice, but if you really don't like it it can be removed. —Aryaman ^{(मुझसे बात करो)} 14:57, 5 September 2017 (UTC)

@JohnC5: Let's deal with what we have now, not what might be in months or years. I don't think we don't have enough variants to warrant it. It can always be added back if needed. --Victar (talk) 15:00, 5 September 2017 (UTC)

@Victar: At the moment, both Aryamanarora and I think it looks nice, so... —JohnC5 18:36, 5 September 2017 (UTC)

@JohnC5: "Looks nice" has nothing to do with it being collapsible and Aryamanarora was actually referring to the bounding box. So, at the moment, it's you, who created it as collapsible, and me who wants it removed, and despite claiming to be open to changing it, you are completely unwilling. There are only two pronunciations, Vedic and Classical. Collapsing them is unnecessary as its reduction is space is negligible and it hides them away, requiring readers to click and see them. It's a lose lose and makes zero sense. --Victar (talk) 19:35, 5 September 2017 (UTC)

@Victar: Ask more people. I said I'd be open to changing it, and I still am, but when I said that, I was misremembering what it current appearance. It turns out that I'm more in favor of keeping it as is. Regardless, you discover a module I've been working on for months and say that something I built looks "bad" and "like garbage", and then you want me to change my mind? It may well look like garbage, but at the moment you're trying to attract flies with vinegar. I appreciate your making a suggestion about how to improve the layout, but please do not insult my and others' work then expect me to agree immediately with you. As I said, I can be convinced, but in this case, I will need more opinions on aesthetics than yours. I do not wish to speak so tersely, and I did appreciate the compliment you gave the module in your original post, but I hope you can see my perspective: I like the layout as it is and do not see an immediate need to change it based on the current line of argumentation. If, however, you can show me that more people dislike the aesthetic than like it, then I will follow the will of the group. —JohnC5 19:58, 5 September 2017 (UTC)

@JohnC5: I think two issues are being confused here.

1. The bounding box is what I was saying looked bad. That was added by Aryamanarora and not original to the module you created. Aryamanarora said he was OK with it being reverted and did so. My apologies to Aryamanarora for my vinegary language. I should have been more delicate.

2. The issue which I'm bring up to you is the collapsing function, which is far less an issue of aesthetics and much more an issue of functionality. In that, I've only spoke practically but if my arguments don't speak to you, I can bring it up to a wider audience. --Victar (talk) 20:16, 5 September 2017 (UTC)

@Victar: Your clarification is appreciated, but I think you should bring up both questions to a wider audience. After an adjustment period, I too have come to like the bounding box. In particular, I think it draws the user's eye to the fact that the box is collapsible in a way that {{grc-IPA}} sometimes doesn't (we often get remarks from people that they only just noticed the collapsing functionality of {{grc-IPA}}). My current feeling is that, if something is collapsible, the experience is improved by a bounding box to draw the user's eye to the fact more information is included. If we remove the collapsing functionality, I certainly think the bounding box should go with it. —JohnC5 20:28, 5 September 2017 (UTC)

Having used and seen it in use for a while, I think it looks much worse now (especially since the positioning of the dropdown is dependent on the screen size). I think we should retain the bounding box if we have the dropdown. —Aryaman ^{(मुझसे बात करो)} 22:45, 13 September 2017 (UTC)

I concur. I'm not entirely sure why Victar decided to change it without full agreement. @Aryamanarora, you would agree that it should be changed back? —JohnC5 23:29, 13 September 2017 (UTC)

Again, @JohnC5, you're making untrue accusations due to not fulling reading what has been said. I not not change it; Aryaman did, following my complaint.

If neither of you find it appealing as is, I recommend revisiting removing the dropdown functionality all together. It makes absolutely no practical sense with just two entries.

--Victar (talk) 23:44, 13 September 2017 (UTC)

@Victar: sorry for the false accusation; I misread which changes to the module belonged to whom. That's my bad. As I said, if we remove the collapsibilty, we should remove the box. Otherwise it should be kept. For now, remove both the box and the collapsibility if you'd like. —JohnC5 00:15, 14 September 2017 (UTC)

(edit conflict) Yes I removed it, mostly because it's really not that important. However, the placing of the dropdown on my screen is kind of annoying now though, it's in the middle of the page. Another reason for the dropdown is to standardize the treatment of Asian languages, which all use dropdowns (even a small template like {{ko-IPA}}). Some things that could be added quite easily: alternate transliteration systems (available at MOD:sa-utilities), some sort of marking of the pitch accent to make it more visible, and maybe accent sandhi? I do agree we could remove the dropdown for now. —Aryaman ^{(मुझसे बात करो)} 00:18, 14 September 2017 (UTC)

TBH I liked the dropdown... —Aryaman ^{(मुझसे बात करो)} 22:25, 15 November 2017 (UTC)

I don't want to get into this again, but it unnecessarily hid data. --Victar (talk) 02:06, 16 November 2017 (UTC)

Please do something about {{sa-desc}}, it looks like garbage!! --2A02:2788:A4:F44:B0CD:80E3:189A:8FF2 14:06, 16 November 2017 (UTC)

स्वधा (svadhā)

So this can have the accent on the first or second syllable. Is there a way to show this using only one invocation of the template? @JohnC5 —Aryaman^{A (मुझसे बात करें • योगदान)} 17:08, 17 December 2017 (UTC)

@AryamanA: Not yet. Unfortunately, I really need to refactor this, and in particular, I want this to use IAST input and have better tonal realizations. —*i̯óh₁nC^[5] 05:16, 18 December 2017 (UTC)

र (ra)

@AryamanA Was the र really a retroflex flap (ɽ) in Sanskrit? Would it not be an alveolar flap (ɾ) instead? -- CueIn (talk) 11:33, 6 February 2018 (UTC)

@JohnC5 Are you sure about this? —Aryaman^{A (मुझसे बात करें • योगदान)} 11:56, 6 February 2018 (UTC)

Most modern literature uses a retroflex flap based on the descriptions of the grammarians and its widespread production of retroflexion. Some authors (Peter Scharf, for example) prefer the retroflex approximate, /ɻ/. @CueIn On what basis would you construct the alveolar tap instead? —*i̯óh₁nC^[5] 06:40, 7 February 2018 (UTC)

@JohnC5: It probably seems strange to him and me because pretty much all New Indo-Aryan languages have an alveolar tap. Hindi does also have the retroflex flap but it's written like ड़ (ṛa), a totally different letter. Also, I find clusters like in अर्ध (ardha) strange if the "r" is retroflex and the "dh" is dental. —Aryaman^{A (मुझसे बात करें • योगदान)} 15:10, 7 February 2018 (UTC)

@JohnC5: Sorry for the late reply. Like @AryamanA said, most modern IA languages use the alveolar tap and I expected Sanskrit's r to be the same (modern Sanskrit chantings all use the ɾ and not ɽ). The retroflex flap sounded more like a stop to me. I did not know of the linguistic literature though. Also, next to dental stops doesn't the retroflex flap sound odd? -- CueIn (talk) 01:48, 14 February 2018 (UTC)

@AryamanA, CueIn: As with many of these things (palatals, rhotics, clusters, tones), there seem to be multiple competing opinions on the matter. I'm steadily researching this, but don't expect progress soon. —*i̯óh₁nC^[5] 02:01, 14 February 2018 (UTC)

@JohnC5: No worries, the progress has been great so far :) —Aryaman^{A (मुझसे बात करें • योगदान)} 02:02, 14 February 2018 (UTC)

@JohnC5: It's ok, I thought it was a genuine mistake. -- CueIn (talk) 02:12, 14 February 2018 (UTC)

@AryamanA, JohnC5 So I was reading w:Grammar of the Vedic language where it is given that ancient treatises on Vedic phonology say that r was originally alveolar which later became retroflex and त (ta) and द (da) were originally /t/ and /d/ and later became completely dental (/t̪/ and /d̪/). What do you guys make of this? -- CueIn (talk) 05:41, 4 March 2018 (UTC)

I would also like to add that if OIA -ta- and -da- are represented as /t/ and /d/, it will also explain better the corruption of OIA -tya- and -dya- into MIA -cca and -jja respectively (cf Ved. mṛtyu/ MIA maccu or miccu, Ved. adya/MIA ajja. etc). Basically, palatal "ca" and "ja" are more likely to emerge if the t and d were articulated from the root of the teeth as opposed to a purely dental t and d. -- CueIn (talk) 05:51, 4 March 2018 (UTC)

Didnt the grammaticians consider it to be a murdhanya repha (harsh/grating retroflex) sound? so it cant be an approximant or flap and there is r̥̄ which cant be a [ɽː] so a retroflex trill [r̠] (considering other retroflexes were also post alveolar)? AleksiB 1945 (talk) 17:08, 6 November 2022 (UTC)

Anusvara

@AryamanA, Erutuon, Victar, JohnC5 At पुंस् (puṃs) the module is producing ⟨ṃ⟩, which is not a valid IPA character. Can this be fixed? —Mahāgaja · talk 16:16, 4 February 2019 (UTC)

@Mahagaja: I've fixed it with the modern solution [ɴ]. —*i̯óh₁n̥C^[5] 20:38, 4 February 2019 (UTC)

@JohnC5: Thanks, John. The module is also producing superscript vowel letters ᵄ, ᵅ, ᵉ, ᵒ, ⁱ, ᵘ, which also aren't valid IPA. Could we use the breve instead, which is valid IPA for marking an ultrashort vowel, thus [ɐ̆, ɑ̆, ĕ, ŏ, ĭ, ŭ], instead? —Mahāgaja · talk 17:01, 13 February 2019 (UTC)

@Mahagaja: I've updated it. Could you find an example of this functionality being used? —*i̯óh₁n̥C^[5] 22:48, 15 February 2019 (UTC)

@JohnC5: Any term ending with visarga, उषाः (uṣāḥ) for example. It works! —Mahāgaja · talk 06:55, 16 February 2019 (UTC)

@Mahagaja: Yeah, I know. I just didn't feel like digging around for one. Thanks! —*i̯óh₁n̥C^[5] 09:03, 16 February 2019 (UTC)

Visarga

Since visarga isnt phonemic, shouldnt it be represented as /-s, -ɾ/ and [h] / [hV]? AleksiB 1945 (talk) 11:05, 7 November 2022 (UTC)

Classical stress ?

@Mahagaja Hi, the module produces a stress accent for Classical Sanskrit, but what is the source for this stress pattern, apart from an unsourced mention on the Sanskrit IPA help page? On the talk page there, it has been claimed that this pattern is just a convention invented by Western scholars. Exarchus (talk) 10:42, 1 January 2024 (UTC)

I don't know. I thought that was the stress pattern used by modern-day Indian Hindus when pronouncing Sanskrit. It may well be an invented convention, considering how many centuries has existed only as a liturgical/literary language and not a naturally spoken one, but if so, I bet it was invented by Indian scholars, not Western ones – sort of like pronouncing ऋ as /rɪ/ or अः आः इः etc. as /əhə ɑːhɑ ɪhɪ/ etc. At any rate w:Sanskrit grammar § Accent says "The [Vedic pitch] system of accent disappeared completely at some point during the classical stage. It was still alive in Pāṇini's time and even after Patañjali. The author of the Kāśikā commentary (c. 700 CE) declares its use optional, and it might have disappeared from popular speech in the early centuries of the Common Era." —Mahāgaja · talk 11:11, 1 January 2024 (UTC)

I doubt it has been invented by Indian scholars, as Hindi (as far as I know) does not have a stress accent. Exarchus (talk) 11:32, 1 January 2024 (UTC)

OK, it does have a stress accent per 'Hindustani phonology'. But the rule stated there doesn't seem to correspond to the rule for Classical Sanskrit. For example: kɪ.ˈdʱə(r) / sə.ˈmɪ.t(ɪ) / ˈreːz.ɡaː.ri(ː) Exarchus (talk) 11:40, 1 January 2024 (UTC)

This paper says: "it is the general custom among Western scholars to pronounce [Sanskrit words] according to the rules of Latin accentuation. This is because the accented syllable is, in the majority of words, unknown" Exarchus (talk) 11:26, 1 January 2024 (UTC)

Interesting here is https://www.academia.edu/download/99417659/Sandell_WeCIEC_33_Handout_V2.pdf

The rule as stated might actually be mostly correct (I'm still reading) Exarchus (talk) 12:09, 1 January 2024 (UTC)

My understanding (which may be incorrect) of the convention for Classical Sanskrit is that it's a little different from Latin. In Latin, stress cannot fall earlier than the antepenultimate no matter what, but in Sanskrit it can go as far left as possible in a word with all light syllables. So in Latin, a five-syllable word with all light syllables is stressed σ σ ˈσ σ σ while in Sanskrit it's stressed ˈσ σ σ σ σ. —Mahāgaja · talk 12:45, 1 January 2024 (UTC)

Summarising the above article: the rules for stress in Classical Sanskrit seem to go back to Bühler (1883). They can be confirmed by looking at where syncopes occurred in Gujarati.

There are two points where Bühler's rules differ from the rule as stated currently on Wikipedia:

- A disyllabic word with a light syllable + final syllable with long vowel (followed or not by a consonant) gets stress on the last syllable. This particular rule hasn't been independently confirmed (or denied).

- A word with 5 light syllables is: ˌσ σ 'σ σ σ (although 'σ σ ˌσ σ σ can't technically be excluded) Exarchus (talk) 12:52, 1 January 2024 (UTC)

A word where this is relevant is महा: is it /ˈmɐ.ɦɑː/ as currently stated, or /mɐ.ˈɦɑː/? It's possible that morphology plays a role in such cases, an example given by Bühler is "जगौ", so a reduplicated form and a stress on the final syllable seems more logical to me there than in the case of महा, where it would contrast with महत [ˈmɐ.ɦɐt̪]. I'm obviously just speculating. Exarchus (talk) 13:18, 1 January 2024 (UTC)

I do think the stress currently given for अंसफलक /ɐ̃ˈs̪ɐ.pʰɐ.l̪ɐ.kɐ/ has to be wrong. Should be rather /ˌɐ̃.s̪ɐ.ˈpʰɐ.l̪ɐ.kɐ/, also taking into account the morphology. Exarchus (talk) 13:40, 1 January 2024 (UTC)

The module is also wrong about अच्छभल्ल, stress should be on penultimate. Here it's just badly programmed. Exarchus (talk) 14:05, 1 January 2024 (UTC)

I think I see the problem: 'r' and 'l' are defined as short vowels, I can see why, but now the syllable 'bʱɐl̪' is interpreted as ending in a vowel... Exarchus (talk) 14:23, 1 January 2024 (UTC)

@AryamanA, Erutuon, Victar, JohnC5 maybe one of you can add something here? Incidentally, I'm not convinced that stress placement in modern Hindu-Urdu is relevant anyway. What were the stress placement rules of the Middle Indo-Aryan Prakrits and Pali? —Mahāgaja · talk 14:34, 1 January 2024 (UTC)

The (Latin-like) stress placement given by Bühler was apparently based on comtemporary "Hindu" pronunciation. So one might think that this might not be relevant for Sanskrit from two millenia ago, but the syncope data from Gujarati suggests that this pronunciation was at least the case for the ancestor of Gujarati.

I tried to fix the module error interpreting [l̪] as a vowel, but it doesn't seem to work. And 'l' can be omitted anyway as it's never a vowel at syllable coda. Exarchus (talk) 14:47, 1 January 2024 (UTC)

I did another attempt to fix the bug, I think it's okay now (although the code looks weird, but I added an explanatory note)

edit: I figured out the workaround doesn't work for syllabic r ... ... Now it's okay Exarchus (talk) 15:44, 1 January 2024 (UTC)

@Exarchus, Mahagaja As y'all know, the data for the stress in Classical Sanskrit is notoriously fraught and measly. The major data sources are Bühler and certain syncopes that occurred in Middle Indic. I heard Ryan Sandell's talk and (as with most of his work) found it very convincing even though the evidence is so scarce. It is stunning the extent to which the "Latin stress + one syllable" (i.e. within the last 4 syllables, stress the rightmost non-final heavy syllable else the preantepenult) has penetrated modern Western practice despite the scanty basis for it. I just naïvely implemented the Classical pronunciation when I made this module but (as with so much else in the Sanskrit modules), I intended to update it later to be more accurate. (I would also liked to have given various Vedic pronunciations according to the different Prātiśākhyas, but I have never found the time to come back to this.) —*i̯óh₁n̥C^[5] 21:35, 1 January 2024 (UTC)

@JohnC5 I think it would be useful to be able to overrule the 'automatic' Classical accent in the case of compounds (e.g. आनापानस्मृति). Bühler states the rule as such: "In compounds, unless the first member be a monosyllabic word, each part generally retains its own accent" and something else wouldn't make sense to me given the huge Sanskrit compounds that can be constructed. Exarchus (talk) 14:40, 2 January 2024 (UTC)

Doesn't Vedic have a rule that finite verbs in main clauses have no pitch accent, or something like that? (I don't remember the details and am too lazy to look it up.) Does that apply to Classical stress accent as well? —Mahāgaja · talk 21:52, 2 January 2024 (UTC)

@Exarchus That seems like a very reasonable idea, though how exactly to do it? Two methods present themselves:

1. It would be nice to mark compound boundaries and calculate the stress for each independently, but, as ever, sandhi and Devanagiri impose impediments on marking morpheme boundaries. I should have been clever and allowed IAST inputs to mark accents (e.g., thus अश्व could be {{sa-IPA|áśva}} instead of {{sa-IPA|a=1}}). With such a system we could do compounds with {{sa-IPA|átithi-pati}}, but this also has problems. We would have to de-sandhi many compounds (e.g., कारागार as {{sa-IPA|kārā-āgārá}}), and that would bring in the complexities of module:sa-utilities. There would also sometimes be some ambiguity about where the compound boundaries should be (though a good example does not immediately spring to mind). Finally, this doesn't solve the problem of specifying which member is the head of the compound if such a problem is solvable in the first place. Can the Vedic accent be a reliable indicator of the morphological head? What about compounds without accentual attestations? Should there be a head parameter |h= to identify which morpheme is the head?
2. Alternatively we could make parameters |p= (index of syllable with primary stress) and |s= (a list indices of syllables with secondary stress) similar to the |a= parameter, with which one could specify the syllables to receive stress. Thus कारागार might be realized with {{sa-IPA|p=3|s=1}} as /ˌkɑː.ɾɑːˈɡɑː.ɾɐ/. This would require a fair amount of knowledge on the user's part.

The greater question would be the distribution of primary and secondary stresses (especially in longer compounds). This seems like it would leave much to speculation since we have so little information. Are we fine with that? Or should we mark them all as primary (with a note)?

@Mahagaja You are correct about that Vedic verbal accentuation rule, but so far as I am aware nothing like that is attested in Classical Sanskrit. —*i̯óh₁n̥C^[5] 07:33, 3 January 2024 (UTC)

@JohnC5 My thought was that automatically programming this would be complicated to implement, so I was simply thinking of doing it manually. I don't think this requires deep knowledge for the user, as the etymology is mentioned in most lemmas.

What Bühler says about compounds is that the "Haupttheil" receives the primary stress. Ryan Sandell says about this: "Whether one should understand that primary stress falls on the rightmost member of any compound – endocentric or exocentric – is unclear. If it applies only to endocentric compounds, then one would presume that the general rules for non-compound stress apply to exocentric compounds."

So at least in tatpuruṣa's (the most 'normal' type), I'd assume primary stress on the rightmost member. Marking multiple primary stresses feels very unsatisfactory. Exarchus (talk) 08:59, 3 January 2024 (UTC)

About your question "Can the Vedic accent be a reliable indicator of the morphological head?", I think the answer is no, as I happen to have read here about a nine-syllable word with (Vedic) stress on the first syllable. Exarchus (talk) 11:20, 3 January 2024 (UTC)

@Exarchus Ok, so what would be the desired input method for primary and secondary stress? |p= and |s= or something else? This seems like the easiest to implement, but it's rather ugly, so I am interested to other ideas. —*i̯óh₁n̥C^[5] 21:19, 3 January 2024 (UTC)

That'd be the same as for the Vedic accent, so I'm fine with that. Exarchus (talk) 21:37, 3 January 2024 (UTC)

👀 --{{victar|talk}} 09:28, 3 January 2024 (UTC)

@Mahagaja: Stress rules for Pali seem to be unknown! Geiger remarks that the 'Sanskritic stress' seems to have affected some changes of vowel quantity, but "Nothing has been handed down to us about the nature of the Pali accent". --RichardW57m (talk) 15:13, 14 March 2024 (UTC)

Notes

pronunciation of ऋ

@Pulimaiyi @Svartava So what is the evidence it was a trilled [r̩] ? If it was, then the Wikipedia page needs to be edited.

Whitney gives, for what it's worth: "The vowel ऋ ṛ is simply a smooth or untrilled r-sound, assuming a vocalic office in syllable-making"

Adding @Getsnoopy @RichardW57m Exarchus (talk) 13:37, 14 March 2024 (UTC)

I'm confused as to what our IPA is meant to be showing for the various forms of Sanskrit. Modern pronunciation? Earliest classical? We've got a problem with the palatals - there's evidence both for palatal stops and for affricates. The Wikipedia article contradicts itself.

As to the pronunciation of <ṛ>, I think we have a problem. While the modern Indian pronunciation might be a flap or tap, how natural is that as a syllabic resonant? Plausible values for a syllabic resonant are trills and frictionless continuants, and it may have been retroflex as some do describe it. No wonder Wikipedia is currently showing it as "ɽɪ", which I think no-one believes of Panini. (Notifying AryamanA, Bhagadatta, Svartava, JohnC5, Kutchkutch, Getsnoopy, Rishabhbhat, Dragonoid76, RichardW57, Exarchus): --RichardW57m (talk) 16:19, 14 March 2024 (UTC)

I'm confused as to what "Modern pronunciation" is. There are no extant speakers of it, so the latest would be the Classical pronunciation, which is a retroflex approximant (as evidenced in Sandhi rules, as well as the explicit classification of it as a mūrdhanya [retroflex] vowel in Sanskrit grammar). Pulimaiyi and Svartava reverted my changes to the module, but I've yet to see any evidence corroborating their claim. Getsnoopy (talk) 04:22, 15 March 2024 (UTC)

@Getsnoopy: My understanding is that there are people whose living is underpinned by a claim to be pronouncing it properly. Whether they individually can converse in it may be another matter, but I have no doubt that some people can. --RichardW57 (talk) 08:57, 15 March 2024 (UTC)

@Getsnoopy @RichardW57m There are in any case supposed to be modern speakers of Sanskrit, but how they pronounce ऋ is not relevant to "Classical Sanskrit", just as the "modern Italianate Ecclesiastical" pronunciation of Latin does not determine the Classical one. One could think of adding an extra "Modern Sanskrit" pronunciation but I don't think there is much consistency there.

My personal guess is that the trill [r] is simply easier to pronounce for modern speakers as it can also occur in Hindi. Exarchus (talk) 10:46, 15 March 2024 (UTC)

Wikipedia: "The 1991 Indian census reported 49,736 fluent speakers of Sanskrit." Exarchus (talk) 10:56, 15 March 2024 (UTC)

Also Wikipedia: "However, despite attempts at revival, there are no first-language speakers of Sanskrit in India." Getsnoopy (talk) 22:03, 15 March 2024 (UTC)

@Exarchus According to this article on Slovak, the Slovak vowel <r> is usually realised as a flap, but the long vowel <ŕ> is realised as a trill. Note, though, that the paper, which is used to illustrate the IPA, writes the sounds as [r̩] and [r̩ː]. Note that this article on Slovak uses the general principle of using 'r' if there is only one rhotic, which I think is a sound principle. In other words, using 'r' does not imply the sound is a trill. --RichardW57m (talk) 13:58, 15 March 2024 (UTC)

Using 'r' can definitely be construed as implying a trill when in a pronunciation section. Exarchus (talk) 15:38, 15 March 2024 (UTC)

I think you mean 'misconstrued'. Especially as the pronunciations are shown between slashes, not in square brackets. The key (w:Sanskrit_grammar#Phonology) should show what we mean, but it doesn't really. The only sane interpretation of the 'key' is that for vowels, 'r' denotes a retroflex approximant. The author of the IPA module shows a touching faith in Wikipedia. It looks as though we need to create Appendix:Sanskrit pronunciation and then add Sanskrit to langs_with_infopages in Module:IPA/data so that the appendix serves as the key. --16:29, 15 March 2024 (UTC) RichardW57m (talk) 16:29, 15 March 2024 (UTC)

But using the incorrect symbol and then pointing readers to another document that they might or might now encounter to clarify "what we mean" is not only circuitous, but defeats the purpose of IPA in the first place. Even when distinguishing between phonemic and phonetic transcriptions, the trill never exists in Sanskrit, so using /r/ makes no sense whatsoever. Getsnoopy (talk) 22:14, 15 March 2024 (UTC)

@Getsnoopy I’m certain the trill is used. Even today, the trill is the prescribed pronunciation despite regional pronunciations getting more common. And whatsoever it is not the pronunciation you add, without a doubt. Svartava (talk) 16:21, 16 March 2024 (UTC)

Sanskrit never had a trill. What evidence do you have that it is/was a trill?

Because the evidence to the contrary is overwhelming. The ऋ is classified as a vowel, which per se disqualifies any sounds which involve the tongue touching any part of the vocal tract, trill included (because vowels are defined in the Sanskrit grammar as sounds that are made without touching the vocal tract—asparśa sounds—but are ones that could be produced "on their own"; hence, svara sounds.

Moreover, all Sandhi rules point to stops being converted their retroflex counterparts when a ऋ is present. Furthermore, there are different vowel lengths for the sounds, with the short one being ऋ (r̥) while the long one being ॠ (r̥̄). This is impossible to do for stops (as the name implies) or for trills.

And again, I'm not sure what you mean by "despite regional pronunciations getting more common"; Sanskrit is not spoken natively by anyone, so there wouldn't be any regional pronunciations. Even if there were, they'd be incorrect because it's those people using their native language's phonetics to try to speak Sanskrit, which is akin to a Spanish person trying to pronounce Latin using Spanish phonetics. Getsnoopy (talk) 01:25, 17 March 2024 (UTC)

@Getsnoopy: It is dishonest or stupid to say, "Sanskrit is not spoken natively by anyone, so there wouldn't be any regional pronunciations". Have you never heard of the old English pronunciation of Latin (which mostly survives in Latin law terms)? Indeed, you seem to be aware of one mechanism for regional pronunciations to arise.

Trills are perfectly capable of being varied in length - which is why the long rhotic vowel of Slovak is a trill although non-syllabically and as a short vowel it is generally a tap. --RichardW57 (talk) 11:46, 17 March 2024 (UTC)

No, it isn't either. An attempt at pronouncing the words in a target language using someone's native language phonetics and actually speaking the target language natively are two very different things. If words are borrowed into one's native language from the target language, and thus, those words have to be phonetically assimilated, that's one thing. But it's ridiculous to suggest that such a borrowing or attempt at pronouncing the target language changes the phonetics of the target language itself so as to have "regional pronunciations".

And it seems like you misunderstood what I'm referring to. The length of a trill is akin to repeatedly pronouncing a stop sound: it requires multiple plosive articulations and touches of the tongue to a place of articulation. It's akin to pronouncing "ttttttt"; it's a bunch of stop–starts. This isn't considered "length"; at least, not in Sanskrit grammar. Length is only when there's a smooth flow of air without having the tongue touch any part of the mouth. So under this definition, a trill is not capable of being varied in length, as it's essentially a bunch of taps strung together.

But all of that is neither here nor there; the point is that a trill involves touching the tongue to a part of the mouth, which disqualifies it as a vowel in Sanskrit grammar. Getsnoopy (talk) 07:03, 18 March 2024 (UTC)

@Getsnoopy It is simply wrong that a trill is a bunch of taps, the articulation is different, though they may sound the same.

But then I don't think there is any hint in ancient Sanskrit grammarians that ऋ is a trill. Exarchus (talk) 16:02, 19 March 2024 (UTC)

@Exarchus, Getsnoopy: Actually, there's a word for it, or at least for र there is - barbaratā - footnotes 12 and 13 on p54 of Allen's book. It's not a pronunciation that's approved of, contra @Svartava. --17:09, 19 March 2024 (UTC) RichardW57m (talk) 17:09, 19 March 2024 (UTC)

@RichardW57m Would 'barbaratā' refer to Dravidian languages?

w:Proto-Dravidian says "The singular alveolar plosive *ṯ developed into an alveolar trill /r/ in many of the South and South Central languages" Exarchus (talk) 20:07, 19 March 2024 (UTC)

@Exarchus: It refers to a pronunciation of Sanskrit. As to who used such a pronunciation, I don't know. --RichardW57 (talk) 22:21, 19 March 2024 (UTC)

I know that it's not the same as a bunch of taps; my point is that it's essentially the same (hence, "akin") in the view of Sanskrit grammar, which is what matters here. Getsnoopy (talk) 01:56, 20 March 2024 (UTC)

@Getsnoopy:: Your claim that "all Sandhi rules point to stops being converted [to] their retroflex counterparts when a ऋ is present" makes no sense. The associated conversion rules are for n > ṇ and s > ṣ, a nasal and a fricative respectively. Moreover, the latter is part of the RUKI effects, so it is by no means clear that it is retroflex assimilation. For the former, I think an alveolar position could have a similar effect, and remember that the literary Prakrits lost the distinction between ungeminated dental and retroflex nasals, and also between dental and retroflex sibilants. --RichardW57 (talk) 21:14, 17 March 2024 (UTC)

Those rules are there precisely because of the retroflex characteristics of the ऋ / र. While the latter is due to RUKI, the whole point of Sandhi is to facilitate pronunciation, and the rules/effects of it are such because that's what ends up facilitating pronunciation the most.

While it's speculative to say that an alveolar position could have a similar effect, it makes no sense that it would in the case of n > ṇ because the n is already the alveolar sound, so the tongue would have to travel farther than where it already is/was positioned in order to articulate the retroflex equivalent, which defeats the fundamental purpose of Sandhi. Getsnoopy (talk) 07:34, 18 March 2024 (UTC)

@Getsnoopy: It looks as though we going to have to start quoting the grammarians rather than interpretations, which makes persuasive arguments more difficult. I recently came across a paper on the Sanskrit ban on /rr/ bemoaning that separate ancient authorities could be cited for dental, alveolar and retroflex pronunciations of the rhotics. I've also noticed a report that one grammarian classifies rhotics with w:jihvāmūlīya, which makes me suspect that a velar pronunciation was also known!

I have seen the later merger of dental and retroflex nasals and sibilants blamed on a confusion between Indo-Iranian contrast between retroflexes and not very distinct dentals and the 3-way Dravidian dental ~ alveolar ~ retroflex contrast. Alveolar pronunciation being the key to nati was beginning to make sense to me because of the curious fact that retroflex nasals and plosives actually block it. Rhotics and sibilants tend to behave differently to plosives and nasals, but apparently one of the differences is that the tongue generally uncurls after oral and nasal retroflex stops, which is offered as an explanation of the behaviour of the rule. Source for nati mechanism: https://dash.harvard.edu/bitstream/handle/1/34945788/ryan_nati.pdf?sequence=5&isAllowed=y --RichardW57m (talk) 11:24, 18 March 2024 (UTC)

A commonly quoted source is:

Allen, W. Sidney. 1951. Some prosodic aspects of retroflexion and aspiration in Sanskrit. Bulletin of the School of Oriental and African Studies 13.939–946.

That's largely hidden behind paywalls.

However, there is some good material in his book "Phonetics in Ancient India" (https://ia800102.us.archive.org/30/items/in.gov.ignca.7855/7855_text.pdf)

Most of the discussion of the rhotics is on pp53-6 of the book. The assertion that ṛ contains r is reported on p62. --RichardW57m (talk) 13:11, 18 March 2024 (UTC)

@Getsnoopy: But using 'r' for a language's only rhotic is a widespread practice, and is not normally regarded as wrong - compare that paper on Slovak. --RichardW57 (talk) 22:55, 16 March 2024 (UTC)

It is not a widespread practice, but is only considered to be allowed in exceptional cases where there's a specific reason to do so. In English's case, for example, it's because the language has many native accents which all realize the (only) rhotic in considerably different ways, so it's a sort of "symbol of last resort".

This is not only not true for Sanskrit (or for any Indic language for that matter), but the ऋ is not the only rhotic in the language; there's also र. It's simply confusing and contradictory for using IPA—let alone incorrect—to use the wrong symbol for the sound when the sound represented by that symbol doesn't occur in the language at all. I can't comment on why it's done in Slovak though. Getsnoopy (talk) 01:31, 17 March 2024 (UTC)

@Getsnoopy: The ancient grammarians state that ऋ contained र - which seems to leads some to assert that ऋ is [ərə], which may be a modern pronunciation. I'm surprised that you insist that ऋ is retroflex but are happy with र being given as an alveolar tap or flap [ɾ]. Note that ऋ and र are equally effective in the nati rule that converts 'n' to 'ṇ'. --RichardW57 (talk) 21:32, 17 March 2024 (UTC)

You seem to have it backwards: the traditional Sanskrit grammar defines all semivowels as the transition (which would be defined as a diphthong in modern linguistics) from whatever the corresponding vowel is to the अ vowel:

• य = इ → अ
• र = ऋ → अ
• ल = ऌ → अ
• व = उ → अ

So it's not that ऋ contains र, but that र contains ऋ. And I never said I'm happy with र being given as an alveolar tap/flap; Classical Sanskrit has it as a retroflex flap, so that would be the correct pronunciation. And it's not me "insisting" on anything; it's literally defined as such in all literature, and corroborated by evidence.

Even if we were to say that the retroflex sound became fronted over the years, the point I'm making is that the internal grammar and relationships remain. This would mean if र is now in the alveolar place of articulation, so is ऋ, and the manners of articulation stay the same: flap and approximant, respectively. A trill simply doesn't exist anywhere.

Note that ऋ and र are equally effective in the nati rule that converts 'n' to 'ṇ'. Indeed, because र and ऋ have basically the same function when it comes to Sandhi (because like I said above, र contains ऋ). Getsnoopy (talk) 06:48, 18 March 2024 (UTC)

@RichardW57m It is in any case not common practice in the Wiktionary pronunciation sections. English uses /ɹ/. French uses /ʁ/. Bengal uses /ɾ/ or /ɹ/ depending on region.

I'm not sure the Slovak example makes your point, because Slovak /r/ can obviously be realised as a trill, it's just that [ɾ] is a more common allophone in practice, basically because it takes less effort and is auditorily similar (like in Dutch where the 'rolled r' is used).

About Svartava's argument that "the trill is the prescribed pronunciation despite regional pronunciations getting more common", well, this obviously reminds me of Erasmian pronunciation of Greek, which is/was also prescribed. Exarchus (talk) 21:20, 17 March 2024 (UTC)

@Svartava: What is your source for the prescription of a trill? --RichardW57 (talk) 22:55, 16 March 2024 (UTC)

The most plausible claim I've seen is that the ancient grammarians described the rhotics as approximants, possibly retroflex. Reportedly only Panini described them as retroflex. --RichardW57 (talk) 22:55, 16 March 2024 (UTC)