Talk:Proto-Romance language

Nominative Plural of Feminine Nouns (Class I)

Why is *kápras reconstructed for both nominative and accusative plural? Shouldn't it be *kápre for nominative plural (and *kápras only for accusative plural)? Making *kápras the nominative ignores e.g. Italian, which has "capre" as plural of "capra". The -e ending in Italian is a reflex of Latin's -ae, which is the nominative plural suffix in this declension. — Preceding unsigned comment added by Brunoczim (talk • contribs) 00:24, 17 December 2021 (UTC)

@Brunoczim The Italian feminine plural /-e/ derives from an earlier /-as/. The reasoning for this, other than comparison with other Romance languages, is that feminine nouns with a velar consonant stem do not palatalize in the plural; cf. the classic example amica 'girlfriend' > amiche 'girlfriends' (not *amice /aˈmitʃe/). That suggests that, during the Late Latin palatalization of velars before front vowels, the nominative plural feminine ending was /-as/ still, hence without a front vowel. Later, [as] > [ai̯] > [e]. For the sound changes, compare Latin portās 'you (sg.) carry' > Old Italian porte (Maiden 1995: chapter 2, §12).

• Maiden, Martin. 1995. A linguistic history of Italian. London: Routledge.

Nicodene (talk) 07:52, 17 December 2021 (UTC)

Confused on some points

I was surprised by these bullets about the vowel system:

• In unstressed position, there was no distinction between /ɛ, ɔ/ on the one hand and /e, o/ on the other.
• In intertonic (unstressed word-internal) position, there was no distinction between /i, u/ and /ɪ, ʊ/ respectively.

In particular, I'm confused about what the second bullet is saying. Since Latin long /iː/, /uː/ in the penultimate syllable attracted stress, there shouldn't be any inherited examples of words with original /i, u/ in penultimate syllables: is the second bullet point simply pointing out that historical gap in post-tonic syllables? Inherited /i, u/ and /ɪ, ʊ/ do not appear to have generally merged in unstressed syllables between the first syllable and the stressed syllable, based on examples such as French engloutir, nourrisson, Spanish antigüedad, amistad.

The following bullet points are contradictory about whether intervocalic /b/ does or does not exist:

• /b d ɡ/ may have been realized as fricatives or approximants between vowels or after /r/ or /l/.
• /b/ did not occur in intervocalic position, having previously spirantized to /β/.

Is the reference to /b/ between vowels in the first meant to apply to word-initial /b/ in sandhi contexts (in which case I'd say "intervocalic position" in the second should be revised to read "word-internal intervocalic position")?

If palatalized consonant phonemes such as /rʲ/, /mʲ/ are being reconstructed, shouldn’t they go in the consonant table? Or if it’s possible to analyze these as phonemically being clusters /rj/ /mj/ (which is an analysis I think I have seen), then it might make more sense to add mention of that possible analysis and use square brackets [rʲ] [mʲ].--Urszag (talk) 07:20, 12 February 2022 (UTC)

Hi @Urszag.

The first bullet-point is, to the best of my knowledge, uncontroversial. No Romance language reflects, in atonic position, an original distinction between /ɛ/ and /e/, nor between /ɔ/ and /o/.

————

Due to Latin stress-rules, as you've noted, the second bullet-point is only relevant for pre-tonic, not post-tonic position. The cited source simply states the rule, without elaboration, but it points to Lausberg 1970 (§§292–296), where a fair amount of examples are provided, at least for the front vowels. Here are some of them:

• Latin vestīmentum, suspīciōnem > Old French vestement, sospeçon (/ə/)
• Latin cīvitātem, *venīre-hábeō > Italian città, verrò (∅)
• Latin vestīmentum, ōscitāre, rādīcīna > Romanian veșmânt, ușta, rădăcină (∅, /ə/)

The overall pattern is that /i /and /ɪ/ have a shared outcome: a 'weakened' sound that may reduce to zero, depending on the environment- even in Italian.

What your example nourrisson shows is that this weakening may subsequently be reversed by analogy with related forms with stressed /i/, such as nourrir. Note that Old French originally had norreçon (/ə/).

All of the preceding examples, as well as the others that I have not copied here (dormitorium, latrocinium, catenionem...), are four-syllable proparoxytones. (Here one has to account for /i/ > /j/ in hiatus and *venīre-habeō, despite the highly etymological spelling, standing for *veniráyo.) The prototypical example that Lausberg provides for this type of word is 'càntatóre', where he uses the grave to indicate secondary stress. Neither *amīcitātem nor antīquitātem fit this type, hence they are not subject to the same rule.

In retrospect, I should have provided a better description than 'intertonic'.

————

For possible lenis allophones of /b d ɡ/, there are two relevant environments: intervocalic and post-liquid-consonant. The bullet-point is not meant to imply the existence of /b/ in both environments- that is an unfortunate byproduct of the concise phrasing. I had tried to think of a decent unambiguous way to explain it, but nothing came to mind. Here is the sort of mess that I kept recreating:

• /b d ɡ/ may have had lenis (fricative or approximant) realizations after a liquid consonant and, for /d/ and /ɡ/, in intervocalic position as well.

The possibility of sandhi /-b-/ lenition may be worth noting, now that you mention it. I'll have to add another citation since, if memory serves, the current ones do not cover it. I believe Lloyd (1987) discusses the matter somewhere.

————

It is possible to analyze the palatalized consonants as phonemically /Cj/. Then, however, one has to add a rule stating that, whenever /j/ is found after a consonant, it palatalizes and is 'absorbed' into that consonant and hence has no independent realization. I find that the representation /Cʲ/ (per e.g. Lausberg, who consistently uses palatalization diacritics) simplifies matters and helps prevent the misreading of a palatalized consonant as a sequence of [C] + [j].

Following this approach, the palatalized consonants should indeed go on the consonant table. If I could find a way to fit them in elegantly, that is, without making a hopeless mess of the table, I would. From memory, palatalized variants would need to be added for all of the labials (except, I believe, /f/*), all of the coronals, and the 'plain' velars /k ɡ/.

* /fʲ/ is not found in any reconstructed lemma that I'm aware of, and very few Classical or Late Latin words come to mind that have the sequences required to produce /fʲ/. Only specific conjugations of fiō- which do not appear to have been carried over into Romance- and suffiō, a verb that didn't survive anywhere.

Nicodene (talk) 12:11, 12 February 2022 (UTC)