Wiktionary talk:About Proto-Italic

Does the change of word-internal /θr/ to /fr/ predate Proto-Italic?

@Catonif, Kwékwlos, Mellohi!, Mahagaja: Wikipedia mentions: "*θ(e)r, *ð(e)r > *f(e)r, *β(e)r in all but Venetic. Compare Venetic louder-obos to Latin līber, Faliscan loifir-ta, Oscan lúvfreis." Since the grouping of Venetic is unclear, either convention seems conceivable, so it would be nice for this page to explicitly describe what to use. The change is currently not shown in *ruðros.

This also seems related to the reconstruction of the following sequences:

• -rð- /rθ/ to -rβ- /rf/, which we show in *werβom, *farβā.

• -zr- /sr/ to -βr- /fr/: if this developed via θr, as sometimes supposed, then it must have developed βr at the same time as *ruβros. Carl Darling Buck says in his Grammar of Oscan and Umbrian (1904) that "118. A change of sr to fr, whence in Latin initial fr, medial br (fūnebris from *fūnes-ris), belongs doubtless to the Italic period, and in Oscan-Umbrian we should expect fr in all positions (as, from bh, O.-U. f = L. f, b). A probable example is O. tefúrúm 'burnt-offering,' U. tefru-to 'ex rogo', tefra 'carnes cremandas', from *tesro-, *teps-ro: L. tepor, Skt. tápas, etc." (page 78)

• -ðl- /θl/ to -βl- /fl/

• -uð- and -ðu- to -uβ- and -βu-

Urszag (talk) 06:06, 22 June 2023 (UTC)

Hi @Urszag! Yes, *-θr-, -θl-, -sr-, etc. are -fr-, -fl-, -fr- in already Italic (presumably PIE gʷʰr > fr as well, cf. febris, ebrius, though we have no example of it outside Latin), except in the Venetic you mentioned and possibly shakily a Sicel origin of λῑ́τρᾱ (lī́trā), though that's even more controversial than the Venetic evidence. Both de Vaan and Sihler reconstruct with a *θ, though most likely as they are Italo-Veneticists, while we are not (itc-pro isn't set as the parent language of xve), and IMO, shouldn't be. Another pan-Italic change is *-dy-, -gy- > -yy- (not reflected in *odjom, *agjō) and of course the change of other post-consonantal *-y- to a syllabic *-i- (despite what this WT essay explicitly seems to deny). What is not pan-Italic at all is having *β ð ɣ: our standard and the PIt Wikipedia page, which says patent nonsense like "Osco-Umbrian f (probably voiced)", are based on one paper, which does not reflect current consensus among linguists. What is even less pan-Italic is ɣʷ, since gʷʰ already becomes w between vowels. What is also likely pan-Italic is -ens for what I otherwise see reconstructed as -ents here on the project. Sorry for rambling, the point is, yes, we can now change *ruðros to something like *ruβros, but it will still be as wrong, until of course it isn't *rufros (or in my opinion *rufr̥, though there is some doubious epigraphic evidence to contradict this, and it's not what sources say anyways... just OR :) ...), we need to change our convention wholesale. My point in deleting so many entries was to have less entries to move when changing the convention. Catonif (talk) 09:07, 22 June 2023 (UTC)

I see. I agree it will be easier to discuss when the number of entries is reduced. While I don't know if there is a consensus for voiced medial fricatives as the Italic reflexes of PIE voiced aspirates, that idea is not just based on one paper: Stuart-Smith 2004 (Phonetics and Philology: Sound Change in Italic, reviewed favorably by James Clackson) is a notable advocate of that view and describes it as a position that goes back to Rix (1957) (in opposition to the model, which Stuart-Smith sees as exemplified by Ascoli (1868), where voiced fricatives developed secondarily from originally voiceless medial fricatives). Regarding gʷʰ, Stuart-Smith says "Umbrian is the only Sabellic language to show any evidence for the treatment of *gʷʰ, but even this is limited to two rather uncertain forms which may be from the same PIE root *wegʷh- 'vow' [...] vufru (e.g. lib 21) may be semantically equivalent to Lat. uotīum 'votive', and if so, may be from *H₁wogʷhro- (Untermann 2000: 863); vufetes (IIa 31) is often taken to be a past participle from *wogʷheto-, 'offered', to the same root [...] This limited evidence suggests that <f> was the regular Umbrian reflex of *-gʷh- between vowels and before *r." (page 106)--Urszag (talk) 09:43, 22 June 2023 (UTC)

That 𐌅𐌖𐌚𐌄𐌕𐌄𐌔 (vufetes) is interesting, we'd best keep it at *-hʷ- then¹⁾. I was basing myself on de Vaan reconstructing PIt. sneiwe/o- "to snow" and on 𐌅𐌖𐌅𐌜𐌉𐌔 (vuvçis) : Vovicius, though checking sources that would the 𐌅𐌖𐌅- may stand for /wō-/, with the vowel lengthened by the simplification of /fš/, like in uouse (/⁠wōše⁽ⁿ⁾⁠/). About voiced fricatives, I'm surprised to see how widespread the idea is, yet it still seems a Latin-biased overcomplication. Of course, even if they did exist, they'd be undeniably in complementary distribution, so using the voiced symbols would be uninformative at best and disinformative at probable. IIRC there's good evidence for intervocalic *-s- in particular being [z] in PIt. already (though I can't make a thorough check ATM), yet given the unphonemicity of it, I still don't think it should be transcribed as *z. Catonif (talk) 11:19, 22 June 2023 (UTC)

¹⁾ Well, *-f- when before r, theoretically at least, since there wouldn't be PIt. reconstructions (i.e. with more than one descendants) containing the cluster. Catonif (talk) 11:57, 22 June 2023 (UTC)

Wiktionary:About Proto-Italic/sandbox

@Urszag, and @Cicognac, Nicodene, Fay Freak due to the recent discussion, ongoing back and forths and disagreements have convinced me to finally attempt to formalise here on WT:AITC the "more than one descendants/derived term" rule for both entries and unlinked reconstructions (painfully the hardest one to convince the community of) and at the same time ditch the Hartmann–Rix-ist voiced affricates (as explained in the stub, we theoretically would not be accepting nor denying either theory anyways), as well as amending some I consider minor graphical irregularities. There is likely much important informations missing, and some could be incorrect, so I thought I'd first discuss it here before going to the BP asking for the procedure to legitimise the policy. Perhaps with more time and effort one should present the different opinions of scholars with the needed footnotes, although the scope of the policy is not really being a descriptive insight into the modern literature but rather perscriptive guidelines for Wiktionary editors. Catonif (talk) 12:07, 29 March 2024 (UTC)

Yup, I noticed that the guidelines are old and have pitfalls here and there, even if Proto-Italic (PIt) is an important, useful and gorgeous proto-language as well as the bridge between Old Latin (OL) and Proto-Indo-European (PIE). Putting forward an updated and complete standardization could help users and patrollers in the future to differentiate between valid edits and invalid/non-standard edits.

• My first proposal is to banish *ɸ and use *f instead for reasons of practicality. The first symbol/letter can give a better understanding of the pronunciation, but on the other hand is difficult to read for those who don't know IPA or have a superficial knowledge of PIt. Secondly, it's harder and longer to type. Of course, this is just my opinion. If the majority opts for *ɸ, I'm fine with that and I'll provide Wiktionary with redirects.
• The second issue at stake is to banish earlier reconstruction which differentiate both early PIt *θ, *x (labialized and non-labialized) > late PIt *ɸ. This differentiation is super-useful as a bridge between late PIt and PIE, but I noticed that this is so controversial (I still can't understand the explanation) that it is better to overlook it (or am I plain wrong?). My suggestion is to keep it only if the reconstruction with *θ, *x is sourced; then, in order to avoid unwelcomed generalization, I suggest to add a sentence like "...but this theory is controversial because <reasons> and should not be overextended". This is a nice compromise IMHO.
• To conclude, I noticed in many pages that the early version of the suffix of the 1° person singular contains an intervocalic glide *-j-. If we label this mutation as 'unsourced', we can erase it everytime it is unsourced/unsupported. By the way, I use *j to indicate a diphthong instead of *i since *j is a semi-vowel and *i is a vowel. I had to correct a lot of *-io with *-jo in verbs since *io was a diphthong. But I noticed that closed diphthongs contain *-i instead of *-j ("-ai, -oi" should be "-aj, -oj", no?). But this is a minor issue.

Whatever we decide, we can check as well the etymologies under the PIE roots to see if their formatting lines up with our brand-new standard. I decided I won't touch anything until a standard is established and approved. Maybe @Kwékwlos, Mellohi!, Mahagaja, Agamemenon can offer nice insights in this discussion. Cicognac (talk) 12:31, 29 March 2024 (UTC)

@Cicognac, yes, those proposals are endorsed in the stub for the new policy I'm proposing. Sorry, forgot to make that clear in the post's text rather than in just the title. About the nature of *y after consonants (e.g. in verbs) I am conflicted, I'd say they must have already syllabised into *i by the PIt stage, maybe Urszag can confirm. Catonif (talk) 13:01, 29 March 2024 (UTC)

I did some minor edits in your sandbox, I'll add three other proposals:

1. personally, I use the schwa/neutral vowel to point out the loss of the sonorant, such as in L 'frūgmentum' < very similar in OL < PIt *frūgməntom < PIE *bʰruHg-mn̥-to-m. I would use schwa.
2. I don't know if the PIE 'y' is readable in PIt, what's more is that 'j' seems to me the natural counterpart of 'w'. Then, I don't want other users to mix up this sound with rounded /y/ if they try to pronounce it and, at the same time, if they don't know the contemporary PIE orthographic standard. Hence, *io, ai, oi > *jo, aj, oj. E.g., *oinos < *ojnos, *aiðilis < *ajðilis. But this is a rather minor issue, I wait for feedbacks.
3. I agree to banish u̯, i̯, χ and þ. The first 2 letters are used in Proto-Balto-Slavic and Proto-Baltic (their are ignored in Proto-Slavic), the third one seems to be a non-standard variation of *x (perhaps coming from Greek or from the uvular theory in PIE which is still not widely accepted, i.e. *g < *G and palato-velar *gˈ < *g?) while the fourth one is used in Old English and Proto-Germanic if I'm not wrong. I would never use ẽ to point out the loss of the sonorant since it's unclear and seems that the sound is nasalized. Cicognac (talk) 13:05, 29 March 2024 (UTC)

Another proposal, then I'll sum up everything I proposed until now while waiting for the others (no need to rush): to allow pre-proto-italic (pre-PIt) reconstructions only if sourced, the same I propose for θ, x in early PIt, again with some kind of memento so that nobody will overextend such changes, e.g., "...but this theory is controversial because <reasons> and should not be overextended".

To sum up:

1. use *f for *[ɸ]
2. allow early PIt vs late PIt only if the reconstruction is sourced and add a memento to avoid overextensions; hence, restrict the use of *x, *θ
3. allow pre-PIt only if the reconstruction is sourced and add a memento to avoid overextensions
4. use schwa/neutral vowel to point out the loss of sonorants
5. use *j for *[j] and *[i̯] (Catonif proposes *y, which has pros and cons)
6. I agree with all the other proposals in the sandbox, e.g., avoiding to reconstruct PIt only using Old/Classical Latin and PIE (I understand that this could give more authority to PIt as a proto-language), banishing u̯, i̯, χ and þ
7. tiny doubt: if I find in a source (e.g., De Vaan) a PIt reconstruction that doesn't show the other Italic languages, should I add it and quote the source?

See you later. Cicognac (talk) 13:50, 29 March 2024 (UTC)

@Catonif I don't know enough to confirm the statement on the current main page that "a distinction between -CjV-/-CwV- and -CiV-/-CuV- [...] should be indicated in Proto-Italic". This seems to allude to something like w:Sievers's law. De Vaan doesn't go over it in the introduction to his dictionary, but has a brief mention in his entry on orior, which says "Meiser interprets this in such a way that adorītur has a regular Sievers variant *-ye- > *-ije- after two syllables, whereas oritur has retained *-ye-. Yet even *-je- would be expected to yield -ī-, as Schrijver 2003 objects." Sihler 1995 argues that it is unlikely that Latin and Germanic share this rule as a common inheritance from PIE (§178); confusingly to me, though, Sihler nevertheless goes on to also refer to the Latin version as Sievers' law, e.g. writing that rēgius "reflects *reH₁ǵ-iyo- with the expected Sievers alternant *-iyo-" (§194a). Sihler also writes that the durative verb suffix *-ye-/-yo- "does not show the effects of Sievers law" (§456), but the cited example is Sanskrit (médyantu) so I can't tell what scope this statement is supposed to have. Fortson mentions "The counting of *(C)VR and sequences of two syllables as a heavy sequence for the purposes of “Sievers’ Law”, e.g. *or-i̯e- > *or-ii̯e- > Lat. or-ī- ‘rise’, *her-i̯e- > *her-ii̯e- > U. her-ī- ‘wish’, *sepel-i̯e- > *sepel-ii̯e- > sepel-ī- ‘bury’. I place quotations around this phenomenon because it is not clear whether we are dealing in Italic with Sievers’ Law proper or a similar, independent phenomenon; cf. Weiss (2009: 40) and now Byrd (2015: 188 with n. 21)" (page 837). Weiss gives the law as follows: "In the last syllable of a word after a syllable containing a long vowel or diphthong, or closed by a sonorant consonant, a glide developed a preceding anaptyctic vowel which was the syllabic version of itself" (p. 39); Weiss also notes though that "The Latin evidence for the operation of Sievers' Law is very scant due to the interference of later sound changes, especially anaptyxis between most consonants and a following i̯" (page 40). The verbs I see in Category:Proto-Italic_verbs where this would be relevant are as follows (some don't have any non-Latin descendants listed and should maybe be removed for that reason anyway): *dormjō (De Vaan explicitly reconstructs this with PIt. dormj-), *folkjō, *mentjōr, *sankjō, *sentjō. In the extended formulation mentioned by Fortson that includes -VR-, we would also have *-ij- in *herjō, *morjōr, *orjōr, *parjō, *saljō.

Other comments: I agree with transcribing the second element of "diphthongs" as glide consonants (Latin vowel weakening seems easier to explain if this analysis is used), and *y and *w seem good to me for this (although I would also be fine with *j for the palatal glide). I agree with using *f rather than *ɸ and *h rather than *x or *χ. I'm not sure about getting rid of *β, *ð, *ɣ, *z; it feels particularly strange to me to use *θ medially if we are also using *f word-initially, since by the point that change had occurred is there any context where it would even have been realized as voiceless [θ]?--Urszag (talk) 00:21, 30 March 2024 (UTC)

What about my proposals at point 2, 3 and 7? Cicognac (talk) 19:59, 31 March 2024 (UTC)

I agree that if intermediate pre-Proto-Italic reconstructions are included, they should be sourced. I'm not sure what "add a memento to avoid overextensions" means: if it means to include that reminder on the About Proto-Italic page, I agree. I don't think we should necessarily include reconstructions made by De Vaan that are based solely on Latin and no other Italic language (if we do include it, certainly cite the dictionary as the source).--Urszag (talk) 06:12, 1 April 2024 (UTC)

Yes, something to let users know that some features must not be extended to all pages, e.g. a redirect to the PIt guidelines. I would add reconstruction from earlier PIt if sourced and avoid overextensions to all PIt lemmas: thus, I would recognize the earlier version of PIt but I would only accept sourced reconstruction. That's why I disagree with Catonif. My solution is a compromise.

I would also add reconstructions by De Vaan but point out what you said, e.g. '...but De Vaan doesn't quote any descendand in other Italic languages, thus his reconstruction is based solely on Classical/Old Latin'. Again, this sounds like a nice compromise to me that could avoid further problems in the future. Cicognac (talk) 08:16, 2 April 2024 (UTC)

I'm sorry for the delay, I'll come back as soon as I'll be able with a thorough answer to everything. Catonif (talk) 13:14, 4 April 2024 (UTC)

@Urszag, Cicognac. Again sorry for the delay, how have you been doing?

Justification for the one-desc rule. [T]iny doubt: [...] should I add it and quote the source? In short, no. This is the one biggest change of the policy. We should not "decant", as you put it, de Vaan into here.

1. What is Proto-Italic? It is a hypothetical lect ancestor of all Italic languages, the very concept of PIt makes sense only if achieved by comparative evidence of the Italic languages. In Latin-only terms a PIt reconstruction would be nothing more that:
   1. If a PIE term can be reconstructed and in the same shape of its Latin descendant (i.e. the phonetic shifts are regular), I don't think there are more that a couple dozen of cases like this, it would be nothing more than a phonetic in-between. But the the word is in constant change, why specifically reconstruct the phonetics of the term in that specific point of time? Once again, that is the moment in which Italic languages split. Something rather irrelevant for a term not attested elsewhere than Latin. Note this situation is somewhat unusual in the first place, since there is most often some irregularity in the evolutionary process. Note I could back down to a compromise where the reconstructions that fall into this category could be accepted (of course, unlinked) in descendant sections (and not in etymologies), however I'd much rather not accept the terms falling into the category of the next point.
   2. If a PIE term cannot be reconstructed (i.e. not in other IE branches) or the Latin descendants has evolved irregularly, conditions often met, the PIt term could not exist in the first case and have many possible different forms in the other.
2. Even if it's sourced? Yes, note our beloved Michiel de Vaan, PBUH, did not get his PIt forms by divine revelation. Every choice needs to be explicitely backed up by the reasons it was taken, and we should not turn Wiktionary into an oracular presentation of whatever source. Note how de Vaan reconstructs PIt *abiē̆ts (p. 20f.) for seemingly no reason and we Wiktionary had followed him as dogs asking no questions, before the page for it finally got deleted.

Given the existence of the family is challenged in the first place, in the words of Urszag, [h]ere on Wiktionary, we reconstruct Proto-Italic in accordance with the mainstream view, but I think including a bunch of forms based only on Latin makes it harder to use our entries to see how much evidence there is for the family itself.

Justification for voicelessness of fricatives. From my perspective I don't see any reason do even doubt on which view is the valid one, I assume it is because of my ignorance, so please correct me. To me Hartmann's, later Rix's, theory was born as Italic data is ovewhelmingly from Latin, and hence a Latin-centric view (where sister languages are seen as "dialects") would naturally emerge. I'll sum up how I understood the two views. Following Ascoli (employed by Sihler and de Vaan):

• bʰ dʰ gʰ gʷʰ
• Devoicing to pʰ tʰ kʰ kʷʰ as in Hellenic.
• Fricativisation into f θ h hʷ as in Koine, or Grimm's Law.
• Much becomes f.
• Split from O-U, and later Faliscan (in both the system remains largely intact, though understandably θ > f, as in Russian, Cockney or Gheg).
• Voicing intervocalically and before liquids, a most understandable shift, innumerous examples, Western Romance just to cite one

Whereas the Hartmann-ists' view (in grey are the claims in common):

• bʰ dʰ gʰ gʷʰ
• Spirantisation into β ð ɣ ɣʷ without devoicing, may look straightforward at first, but does this actually happens in any IE language? Actually, does this development (fricativisation of aspirate voiced plosives but not of simple voiced plosives, assuming the traditional analysis and not the ejective one) occur in any language at all?
• Devoicing only word initially. I'm sure there are examples of this kind of thing per se, but it's definitely not as common, and why would it affect only fricatives?
• Much becomes β.
• Split from O-U and Faliscan, both (independently!) re-devoicing only fricatives this time also word-internally (but why?), and then θ > f.
• The Latin system stays intact, proving that Latin is the purest language and Romans are superior to the uncultured Sabellians who have distorted their speech. /s

So is there any context where it would even have been realized as voiceless [θ]? Later sensibly developping into O-U f and L d, IMO yes, very much. In any case, even if I am strongly convinced by Ascoli's view, I recognise that the discussion in the literature is still ongoing so for fairness (since Wiktionary should reflect modern literature) our reconstruction scheme should ideally be a compromise between the two, and by keeping all symbols voiceless, since even if following Hartmann one cannot deny that there would not be any phonemic distinction, one can leave wondering about phonetic realisation to the opinion of the reader. One can just assume that we are writing [ð] as *θ if that is what they like to believe.

Justification for using *y. De Vaan uses *j *w, but also uses *i̯ *u̯ for PIE. Sihler uses *y *w for PIE like us, and *y *w for PIt. Given how we handle PIE, this seems likes the most sensible choice. Poultney also uses *y *w. However this is not an important point.

I still need to properly read and understand the *iy and *n̥ situation before I can talk about them. Catonif (talk) 12:48, 14 April 2024 (UTC)

I remain unconvinced that Ascoli's model is clearly the best, but I don't have a particularly strong preference about whether we transcribe Proto-Italic fricative voicing, so the following is just for the sake of discussion.

My point about *dʰ was that its outcomes in Latin seem easiest to explain to me if its reflex was already voiced in word-medial position at the time when word-initial [θ] was fronted to [f]. I see no obvious motivation for why [θ] to [f] would occur word-initially but not word-medially; it seems more plausible for [θ] > [f] to have occurred unconditionally in the ancestor of Latin at a point where medial *dʰ had a phonetically voiced reflex (compare how some accents of English show [θ] > [f] as a general sound change but treat [ð] differently).

Stuart-Smith 2004 (Phonetics and Phonology: Sound Change in Italic) proposes that *bʰ dʰ gʰ gʷʰ had different phonetic evolutions in medial and initial position: word-medial *bʰ dʰ gʰ gʷʰ became [β ð ɣ ɣʷ] while word-initial *bʰ dʰ gʰ gʷʰ became [pʰ tʰ kʰ kʷʰ] and then [ɸ θ h hʷ]. Stuart-Smith argues that it isn’t clear that the the word-medial fricative reflexes found in other Italic languages were phonetically voiceless, citing in support of voiced fricative values both some primary evidence and some prior literature. Unfortunately the material is too spread out and complex for me to summarize Stuart-Smith's evidence now, but an example quote: "Meiser (1986: 29, 73f.) suggests that /f/ was voiced in a voiced context word-internally [in Umbrian]" (pages 107-108). The phonemic native spellings obviously tell us nothing about whether voiced allophones existed (any more than the use in Old English of the spellings -s- -f- -þ- between vowels). The main evidence I can think of for phonetic voicelessness is Latin and Romance forms with -f- that are explained as intra-Italic borrowings, such as rufus, scrōfa, farfecchia. But are these conclusive? I’m not sure how likely it is that Latin, at the point where it had no [v], would have adapted foreign [v] as [f] rather than as [b] or [w], but it doesn’t seem impossible: compare Arabic فيروس from virus.--Urszag (talk) 03:44, 15 April 2024 (UTC)

Yes, I am fine, I have been patiently waiting for you. I didn't do any new edit or rollback and I won't do any before we find an agreement on the official guidelines for PIt. Do you allow for any Pre-PIt/Pre-Italic reconstruction (as an intermediate stage between PIE and PIt) if sourced? Cicognac (talk) 15:06, 15 April 2024 (UTC)

Again sorry for the delay, these were (and soon will yet again be) some of my busiest moments thus far. @Cicognac, whether I allow pre-PIt reconstructions, personally most likely not, could you give an example of what you have in mind? @Urszag, this is an interesting and enjoyable discussion and I would continue it (I still have counterpoints) but I don't want to clog this too much. Catonif (talk) 17:42, 25 April 2024 (UTC)

Don't panic, we are still waiting. I'll remove just some stuff I know to be wrong within the week-end, then I won't touch anything else. Sorry, just pretend that I wrote nothing. Cicognac (talk) 17:36, 26 April 2024 (UTC)