Cad'inorian vs. Romance Thread

[BGMan]

I've sometimes wondered what Verdurian and its relatives would look like if they developed from Latin instead of Cad'inor, but with the same sound changes (i.e., sort of like Wenedyk for Polish) -- similarly, what the major Romance languages would look like if they developed from Cad'inor instead of Latin.

[Mornche Geddick]

You'd have to start out with Vulgar Latin, not Classical Latin.

So, anybody know where to find a good lexicon of VL?

[hwhatting]

So, anybody know where to find a good lexicon of VL?

This is a start.

[Kuchigakatai]

I've sometimes wondered what Verdurian and its relatives would look like if they developed from Latin instead of Cad'inor, but with the same sound changes (i.e., sort of like Wenedyk for Polish) -- similarly, what the major Romance languages would look like if they developed from Cad'inor instead of Latin.

If you like romlangs that sounds like a fairly interesting project actually, and a project that I think no one has carried out before. I like romlangs but I'm not familiar with Zompist's modern Central languages (Verdurian, Ismain, etc.), so I can't make sketches yet... I don't know if I'll get around it.

You'd have to start out with Vulgar Latin, not Classical Latin.

So, anybody know where to find a good lexicon of VL?

To pull this off well, you'd have to start with both CL and VL really, adopting continuous learned and semi-learned borrowings into the descendants just as Zompist's modern Central adopts words from Cadhinor.

(Also, I find the usual distinction between CL and VL plainly ridiculous, since CL and VL are just registers of the same language, Latin. Yes, in the Roman Republic and Empire, there was morphosyntax and vocabulary that rarely if ever got written down but survived in Romance (e.g. ecce/eccum hīc > Portuguese aqui, Old French ci, Italian qui, Romanian aci; or the have-perfects in the TAM system) and morphosyntax and vocabulary that became, or was always, so lofty it stopped being used entirely so it didn't survive in Romance (e.g. flēre 'to cry', īlia 'guts (poetic)', or syntax sandwiching in noun phrases). This is also true of tons of other languages, such as 18th century English, except that we don't say there was a Classical English and a Vulgar English at the time. The Wikipedia article says "Vulgar Latin had no official orthography of its own."--what does that even mean? Would an article on 18th century English point out that the colloquial ways of speaking English at the time didn't have "an official orthography of their own" (different from standard English!)?

A colloquial word is just a synonym of more formal words or expressions, and mutatis mutandis also colloquial morphosyntax, but somehow in Latin the colloquial registers are often thought of as a Different Language, even in the 1st century BC when there wasn't much of a distance between speech and writing. If you like to distinguish CL and VL, you might as well also distinguish Classical Spanish from Vulgar Spanish today, in the 21st century. Okay, I think I should stop myself from going into a much longer rant now...)

This is a start.

I get:

Accès refusé ! Pour vous ouvrir l'accès, veuillez contacter gilles.souvay@atilf.fr.

Maybe you can access it without a special permission because you're in Germany?

[hwhatting]

It works from my PC, but I get the same message when I try it from my mobile...

[hwhatting]

It seems that the problem is with the direct link. If you follow this link:
http://www.atilf.fr/spip.php?article146, click on "Publications", and then on "Consultation du dictionnaire" in the side bar, it should work. (The VL entries are under "Étymons protoromans".)

[Kuchigakatai]

It seems that the problem is with the direct link. If you follow this link:
http://www.atilf.fr/spip.php?article146, click on "Publications", and then on "Consultation du dictionnaire" in the side bar, it should work. (The VL entries are under "Étymons protoromans".)

Ah, I see, now I can get to it. Here's a good direct link:

Dictionnaire Etymologique Roman

Don't you love it when programmers make websites in a way that they're hard to bookmark? :)
(I sure f****** don't.)

[hwhatting]

Yep

[malloc]

(Also, I find the usual distinction between CL and VL plainly ridiculous, since CL and VL are just registers of the same language, Latin. Yes, in the Roman Republic and Empire, there was morphosyntax and vocabulary that rarely if ever got written down but survived in Romance (e.g. ecce/eccum hīc > Portuguese aqui, Old French ci, Italian qui, Romanian aci; or the have-perfects in the TAM system) and morphosyntax and vocabulary that became, or was always, so lofty it stopped being used entirely so it didn't survive in Romance (e.g. flēre 'to cry', īlia 'guts (poetic)', or syntax sandwiching in noun phrases). This is also true of tons of other languages, such as 18th century English, except that we don't say there was a Classical English and a Vulgar English at the time. The Wikipedia article says "Vulgar Latin had no official orthography of its own."--what does that even mean? Would an article on 18th century English point out that the colloquial ways of speaking English at the time didn't have "an official orthography of their own" (different from standard English!)?

From what I understand, though, vulgar Latin had considerably different morphology than classical Latin, especially among nouns which had lost most case distinctions. That seems like enough reason to treat them as distinct, particularly since it was the isolating declension of vulgar Latin rather than the fusional declension of the classical variety that fed into Romance languages.

[Mornche Geddick]

Perhaps an analogy might be made between Classical / Vulgar Latin and Literary / Colloquial Welsh today.

[Kuchigakatai]

From what I understand, though, vulgar Latin had considerably different morphology than classical Latin, especially among nouns which had lost most case distinctions. That seems like enough reason to treat them as distinct, particularly since it was the isolating declension of vulgar Latin rather than the fusional declension of the classical variety that fed into Romance languages.

Nah, it's trivial to show that case was well alive in classical times in the 1st centuries BC and AD. I don't know who you got the idea that "most" case distinctions had already died by the time of Classical Latin from, but whoever it was should've read more. For one, it's attested aplenty in graffiti at Pompeii from the 1st century AD, including graffiti with terrible spellings and outright vulgar (in the modern sense) content. For another one, it survives into Romance.

Some examples from Pompeii:

• secundus prime suae ubique isse salute. / rogo, domina, ut me ames.
  (In standard orthography: Secundus primae suae ubique esset salutem. Rogo, domina, ut me ames.)
  
  'Secundus writes this to his favourite, wherever she is. My mistress [with power over me], I ask you to love me.'
  (CIL I.10.7, at the "House and Office of Volusius Iuvencus", on the left of the door)

Although there are Old Latin inscriptions with the nominative singular -s missing (notably in the 3rd-century BC epitaph of Lucius Cornelius Scipio at the beginning), it survived in Old French and Old Occitan, so the -s in Secundus likely stands for a sound. The dative -ae in primae is spelled -e since at the time [aɛ̯] had merged into [ɛ], but the point is that this dative singular was still there in speech, surviving later on into modern Romanian.

• futebatur, inquam futuebatur, civium romanorum atractis pedibus cunus, inqua nule aliae veces erant nisissei dulcisime et pissimae.
  (In standard orthography: Futuebatur, inquam futuebatur, civium romanorum attractis pedibus cunnus, in qua nullae aliae voces erant nisi dulcissimae et piissimae.)
  
  'The c*nt of the Roman people got f*cked, f*cked I say, with the legs grabbed back [i.e. r*ped]. And that day there were no voices except sweet ones and respectful ones.'
  (political polemic as a parody of orators; CIL IV.1261, at the "House of the Tragic Poet", on the outside wall)

This one is interesting because it shows the nominative plural -ae was perfectly alive, misspelled as -e (nule, dulcisime). This ending is believed to either not have survived anywhere, or to have phonologically merged in Italian/Romanian with accusative -as (> [ai] > [e]).

• Filius salax, qud tu mulierorum difutuisti?
  (In standard orthography/morphology: Filius salax, quot tu mulierum defutuisti?)
  
  'My lustful son, how many women have you f*cked?'
  (CIL IX.8.3, at the "House of the Centenary", in the atrium)

Here, the genitive plural of mulier 'woman', mulierum, has been analogized with the 2nd declension ending -orum. It may seem surprising as the 2nd declension is associated with masculine and neuter nouns, but this is exactly the outcome we see in modern Romanian where the genitive plural ending survives as -or in the feminine article suffix and feminine demonstrative/indefinite pronouns. Note how the nominative filius is used as a vocative here, as opposed to standard fili, suggesting an early loss of the vocative at Pompeii.

Perhaps an analogy might be made between Classical / Vulgar Latin and Literary / Colloquial Welsh today.

I don't know anything about Welsh to comment on this.

[kodé]

My recollection from back when I knew a lot more about Latin and Romance was that the change from a 5 vowel inventory with contrastive length to a 7 vowel inventory without a length contrast plus the loss of word-final <m> really blew up the case system, leading to much greater syncretism than in Classical Latin. So, for instance, vanilla 1st and 2nd declension nouns that looked like this in Classical Latin:

a	ae	us	ī
ae	ārum	ī	ōrum
ae	īs	ō	īs
am	ās	um	ōs
ā	īs	ō	īs

became this:

a	ɛ	os	i
ɛ	aro	i	oro
ɛ	is	o	is
a	as	o	os
a	is	o	is

Especially in the singular, that’s a ton of syncretism! Though the sound change wouldn’t have touched the dative and ablative plural, which I guessed just disappeared for non-sound-change-related reasons.

That was followed by the increasing usage of <ad> + accusative for the dative and <de> + accusative for the genitive. But FWIW, the distinction between nominative and accusative lasted for a long, long time; I know Old French retained that case distinction for many nouns.

[hwhatting]

And as I think Ser has pointed out somewhere, Romanian has a merged Genitive and Dative in its case system, so the minimum case system we must assume for Proto-Romance is Nom., Acc., Gen.-Dat.

[BGMan]

Latin cases appear to be more unstable than, say, Slavic ones -- I suspect because Latin's cases resemble each other too much, whereas the Slavic ones are more distinct. But since Verdurian has cases, I think the Romance mirror would too, and probably not change too much from Latin. So the accusative singular would remain unchanged from Latin. BUT -- Verdurian doesn't tend to have final s, except maybe if it's a stressed syllable. So Romance Verdurian would be more like Italian or Romanian in that regard. So masculine nominative plural could be -i but the accusative plural be -î, from -os > -oi ?

I imagine Romance Verdurian could derive its definite article from ipse like Sardinian did.

The most interesting question, IMO, would be the sound changes from Latin to Romance Verdurian. I mean, some obvious ones would be lactem to laže. But how to make ď distinct from d? Maybe have ď evolve intervocalically as it did in Spanish, but have d arise from intervocalic nt or t?

Verb conjugating. Portar in the present tense: porto, portai, porta, portam, portad, portan? (Can you end a word with -d? Or would it be portač? Although I would expect that in Xurnese Romance...)

[Emily]

The most interesting question, IMO, would be the sound changes from Latin to Romance Verdurian. I mean, some obvious ones would be lactem to laže. But how to make ď distinct from d? Maybe have ď evolve intervocalically as it did in Spanish, but have d arise from intervocalic nt or t?

i'm working on a long-term large project where i'm switching proto-germanic and latin (classical, which i know is cheating but the phonology matches up a lot better) and deriving the germanic and romance languages from the "opposite" parent languages: so english, german, and norwegian will all descend from latin, and french, spanish, and romanian will all descend from proto-germanic. after spending some time trying to figure out how to develop the english (and scots, and icelandic) sounds /θ/ and /ð/ from latin, which didn't have them, in order to make the language as much like irl english as i could, i ultimately decided that it was better to leave these sounds out of english, as any sound change i came up with to create them would be historically unmotivated—there was no change from PG that created /θ/, it was simply retained from PG straight into modern english, and /ð/ (in most instances) is simply a phonemicized allophone of earlier /θ/, so a new rule to create these sounds out of latin would violate the core idea, which was seeing what would happen if latin followed the general course that irl proto-germanic did.* this is a roundabout way of saying i don't think a romance verdurian would (or should) have /ð/ unfortunately

*of course, proto-germanic did have /θ/, which means i could retain it in the new "romance" daughters; i eventually settled on keeping it in romanian and sardinian and losing it in the rest of the romance languages, since much like english and icelandic these languages are somewhat geographically isolated from the rest of the family and tend to be weirdos in comparison to the rest of them. but this isn't helpful for the proposed romance verdurian!

[Tropylium]

But how to make ď distinct from d? Maybe have ď evolve intervocalically as it did in Spanish, but have d arise from intervocalic nt or t?

Start one step earlier with *ð < Proto-Italic *θ still separate from *d? (You could use this same approach to also get e.g. /z/ for free.)

[BGMan]

Finally got around to working on this, if anyone is curious. Here's the result... pardon the lost formatting:

**********
# Verduroman

*A Romance language derived by feeding Classical Latin through the sound
changes that carried Caďinor into Verdurian.*

Verdurian is the daughter of Caďinor, the imperial language of Almea's
Caďinorian empire, and Mark Rosenfelder documented the roughly one hundred
regular changes that separate them. Caďinor occupies the same position
relative to Verdurian that Latin occupies relative to the Romance languages:
an imperial literary standard with some fifteen centuries of wear behind its
daughter. Verduroman is what comes out when Latin is substituted for Caďinor at
the head of that chain and the changes are allowed to run.

---

## Introduction

### Premise

This is a derivation, not a design. The vocabulary is Latin; the phonological
history is Verdurian's. Nothing has been invented for effect: where the
language looks strange, it is because an attested Latin word met a documented
Verdurian sound change, and where a choice had to be made, the choice is
recorded in this document rather than buried in the output.

A number of decisions were forced by the fact that Latin and Caďinor are not
the same language — some because Latin supplies phonemes Caďinor lacked,
others because it supplies contrasts Caďinor's rules were never written to
preserve. Each is set out below with its justification. Everything else
follows mechanically.

### Sources

* Latin input: Classical forms with vowel length marked, cited in the
nominative singular for nouns and the present infinitive for verbs.
* Sound changes: Rosenfelder's Caďinor-to-Verdurian list, applied unmodified
except where noted.
* Pre-processing: a program assigns Latin stress, forms glides in hiatus, and
collapses the vowel system before the list runs. Stress cannot be expressed
as a rewrite rule — it requires syllable counting and weight evaluation —
which is why this stage is a program rather than more rules.
* A second rule file carries Early Modern Verduroman to Modern, following the
Modern Verdurian changes.

Second-declension nouns and adjectives are fed in with the archaic Latin
ending **-os** (*mūros*, *lupos*, *bonos*) rather than the Classical -us, and
so are all other endings in short -us: the third-declension neuters
(*corpos*, *tempos*, *genos*, *pectos*, *vetos*) and the dative-ablative plural
(*manibos*, *patribos*, *diēbos*).

This is not a convenience. Latin's -us continues PIE *-os in every one of those
endings and is the exact cognate of Caďinor -OS; only the fourth declension's
nominative singular continues *-us and takes the -US route. Classical Latin
conflates them by a change of ŏ to ŭ before final s around 200 BC. Latin's own
morphology preserves the evidence: the r of *corporis* and *generis* is the
rhotacized s of a stem in *-os-, which is why *corpus* and *genus* decline the
way they do. The archaic spelling restores a distinction the rule file depends
on and makes the correspondence literal rather than stipulated.

Nouns are cited in the nominative singular. This is Caďinor practice and it
suits a language that keeps its cases, but it should be read with the
third-declension stem alternations in mind: *nox* and *noctem* give **nöč** and
**nožem**, and the citation form is not always the informative one.

### Periodisation

Three stages, following Verdurian's own.

Rosenfelder identifies a break roughly halfway through his change list, at the
word-final v → f change: everything before it corresponds to Caďinor → Old
Verdurian, everything after to Old Verdurian → Early Modern. The same break
applies here, and it is useful, because Old Verduroman is the stage at which
the Latin declensions are still legible.

A third stage follows the Modern Verdurian of two centuries later. Its changes
are almost entirely unreflected in the spelling, so Modern Verduroman is a
layer of pronunciation over unchanged written forms and requires no separate
orthography. The changes that bear hardest on Verduroman are the reduction of
ny to y — which undoes the gn palatalisation, while leaving ly untouched — the
nasalisation of final n and m, which affects every Latin accusative and
neuter, and the loss of post-syllabic r.

### Dialects

Where a change admits more than one plausible outcome, the variants are
recorded as dialectal rather than adjudicated. This is a bookkeeping device
first; no culture, geography or history is proposed for this language.

One such split has emerged and is substantial enough to name. The syncopated
third-conjugation infinitive leaves a syllabic r — *bibere* → *bebr*. The
**Avele** variety keeps the cluster, which is what one would expect of a
variety spoken near Ismaîn, where such codas are unremarkable. The **standard**
vocalises it to **î**: *bebî*, *vivî*, *caðî*. The two then diverge further at
the Modern stage, where post-syllabic r is lost: the standard's infinitives are
already safe, while Avele's would lose the r outright and fall together with
the bare stem.

---

## Phonology

### Input inventory

Latin is fed in with its full consonant inventory plus the Greek graphemes,
which do substantial work here. The Greek stratum is not decorative: because
Latin lacks /x/, /θ/ and /ð/ entirely, several Verdurian sounds would be
unreachable without it.

| Latin | Treated as | Notes |
|---|---|---|
| ph | f | merges with native f |
| ch | x | h alone; ř only in the chr cluster, see below |
| th | Þ | sent to ð between vowels and to t elsewhere; survives as neither |
| rh | r | Greek rho; no longer a source of ř, see below |
| qu | k | fills the second velar slot |
| x | cs | Latin's x is /ks/; the rule file reserves the letter for /x/ |
| y | ü | Greek upsilon; the main lexical source of ü |
| z | double consonant | makes position for stress |
| h | — | invisible to weight and to every rule, but restored in the spelling |

### Stress

Latin stress is carried through unchanged: penultimate if the penult is
heavy, antepenultimate otherwise, with a stop-plus-liquid cluster counting as
a single onset and therefore leaving the penult light. Stress is assigned on
the Classical form, before the vowel merger, because penult weight depends on
the vowel length that the merger removes. It is then carried as a diacritic
through every subsequent stage; a good half-dozen of Rosenfelder's rules key
on it.

### Vowel development

The Classical ten-vowel system collapses to the Vulgar Latin seven, then to
five, as in Spanish:

| Classical | Vulgar Latin | Stressed outcome | Unstressed |
|---|---|---|---|
| ī | i | í | i |
| ĭ, ē | e | é | e |
| ĕ, ae | ɛ | ié | e |
| ă, ā | a | á | a |
| ŏ | ɔ | ué | o |
| ŭ, ō | o | ó | o |
| ū | u | ú | u |

The diphthongs are not decoration; they are the mechanism by which vowel
length survives. Rosenfelder's list already converts ié to **ë** and ué to
**ö**, so Latin short e and short o both leave a distinct trace in the modern
language, in clean contrast to ē and ĭ giving plain e and to ŭ and ō giving
plain o. Quantity therefore remains visible in the output without a single
quantity-sensitive rule.

The list also supplies **ä** from stressed au, which Verduroman reaches
directly from Latin (*cauda* → *cäða*, *auris* → *är*). Unstressed au gives
plain a. Verduroman acquires further au later, from l before a stop, and in
reborrowings.

**Length shields.** Caďinor has no phonemic vowel length, so its reduction
rules were written on the assumption that every vowel is short. Latin's long
vowels are therefore exempt from them, and the exemption is not a patch but the
governing principle of this whole section — the same fact surfaces four times:

* An unstressed long vowel is not syncopated. *Amābāmus* gives **amabám**, not
*ambám*; the long ā of the stem survives, as it does throughout Romance.
* A final s survives after a long vowel long enough to matter. This carries the
plural through the whole system, and when the s is later lost it leaves its
vowel behind as the marker: *rosās* → **rösî**, *mūrōs* → **murî**, *patrēs*
→ **pači**. Without the exemption the plural would have gone with the s.
Spanish keeps the s straightforwardly; Old Italian turned it into i.
* It separates the second person singular from the third. Latin *amās* and
*amat* differ only in length, and without the exemption both would reduce to
the same shape — leaving the far more frequent third person the longer of the
two, which is the wrong way round. The two end up **amî** and **ama**.
* And it routes the endings. See below.

**Two further exceptions**, protecting contrasts Caďinor's rules were not
written to handle:

* Short u is held back in a final *-us*, so that the fourth declension takes
the -US route to **-o** while the second declension's archaic *-os* takes the
-OS route to zero. Caďinor's contrast is between two short-vowel endings, and
the merger would otherwise collapse them.
* Long vowels before a final s are exempt from both routes, for the same
reason: the rules encode a distinction Caďinor made only among short vowels.
This is what preserves the second-declension accusative plural in *-ōs* and
the fourth declension's *-ūs*. See the nominal declension below.
* Short u in hiatus is exempt, being the second element of a diphthong rather
than a nucleus. This is what makes the ui sequence, and therefore one source
of otherwise scarce ü, reachable at all.

A further pre-processing change belongs here rather than in the rule file: an
unstressed front vowel in hiatus becomes a glide, as in Vulgar Latin, so
*palea* gives *palja* and *vīnea* *vinja*. Without it *palea* would fail to
palatalize while *folium* did.

**Blocking before a yod.** Diphthongization is blocked where a yod follows,
as in Spanish: *noctem* gives *noche* and not *\*nueche*, *oculum* gives *ojo*
and not *\*huejo*, *folia* gives *hoja*. The blocking environments are -ct-,
-cl-, -gl-, the -c'l- and -g'l- clusters that syncope will later create, and
-li- before a vowel. It applies to both open-mid vowels, which is why Spanish
has *lecho* and *espejo* rather than *\*liecho* and *\*espiejo*.

One consequence is a stem alternation across the paradigm. The nominative
*nox* has no yod and diphthongizes; the oblique *noctem* has one and does not.
Spanish never had to resolve this because it continued the accusative and let
the nominative die, but Verduroman keeps its cases and therefore keeps the
alternation.

**Open.** Whether Latin x blocks is undecided. Spanish *seis* from *sex*
suggests it produced a yod and blocked diphthongization; *diestro* from
*dextrum* suggests it did not, though that form is semi-learned. It currently
does not block, so *sex* gives **šeč**.

### Consonant developments

Beyond the inherited list, the following departures:

**ř from cr-, kr-, and chr.** The existing list already derives ř from a
velar-plus-r cluster; Latin's cr- is fed into the same rule, initially and
medially alike (*sacrum*, *crēdere*). Greek ch before r is the same cluster
shape by a different route, and keeps it: *pulchrum* gives **polřom**. But bare
ch with no following r, and initial rh-, do not — ř should mark one thing, the
cr- history, not "any velar the Greek stratum happens to produce." *Monachos*
gives accusative **mönhom**, not *mönřom*; *stomachos* gives **stömhom**, not
*stömřom*. The Greek stratum still supplies **h** freely wherever the cluster
doesn't hold — it only supplies ř when it does.

**The learned stratum.** cr- is retained unchanged in reborrowings from
Latin and in high-register, scholarly, and onomastic vocabulary, producing
doublets on the Romance pattern. These are maintained as a separate input
list and merged afterwards.

**Latin y gives ü.** The letter transcribes Greek upsilon, and since it occurs
only in the Hellenic stratum it hands the language a vowel that has otherwise
almost no lexical source: **lüra**, **butürom**, **dažül**, **tümpanom**,
**amüdala**, **abüss**. Eighteen occurrences in the lexicon, against one from
the *iu*/*ui* coalescence.

**Three reflexes of intervocalic d.** Before a high vowel it becomes z, by an
existing rule. Elsewhere between vowels it lenites to **ð**, on the Spanish
model. A third, unlenited d arises later from cluster simplification
(-rd-, -ld-, -gd-, -dd-): *smaragdos* gives **marád**. The ordering is load-bearing: lenition must precede
cluster simplification, or the newly created d lenites too and the
three-way split collapses.

**Final e lost after r.** A narrow change with a wide reach: it turns every
infinitive consonant-final (*potár, vedér, dormír*), which is what Verdurian's
own infinitives in -r, -n and -c call for, and it produces the correct Romance
outcome for the few nouns it also touches — *mare* gives **mar**.

**Syllabic r vocalised to î.** Where the preceding change leaves r after a
consonant, the standard vocalises it, much as German final -er does. See
Dialects above for the variety that does not. This and the loss of final s are
the language's two sources of î; ui joins iu in giving ü, so nothing else feeds
it.

**Glide sequences resolve.** **yë** gives **ë**, so *agnellos* is **anël** and
*herba* **hëba**; this keeps *ergō* → **ëgo** apart from *egō* → **yego**, which
the other outcome would have merged. **yi** and **iy** both give **i**, so
*fīlios* is **fi** and *pariēs* **pári**.

**The palatals do not stand alone in a coda.** Syllable-final ny and ly revert
to plain n and l, since neither palatal exists independently of its ordinary
counterpart: *magnos* → **man**, *longos* → **lön**. They survive only before a
vowel (*ponyár*, *folyom*, *anyóst*).

**qu → k.** Latin qu fills the second velar slot, and feeds the existing rule
palatalizing k to c before i — so *quīnque* palatalizes while *aqua* does not.
**gn → ny and li → ny's counterpart ly.** Latin gn gives the palatal nasal,
as in Spanish *leña* and *seña*: *lignum* → **lenyom**, *pugnāre* →
**ponyár**, *tegula* → **telya**. Latin -li- before a vowel, together with the
-g'l- clusters syncope creates, gives the palatal lateral in the same places
Italian has *gli*: *folium* → **folyom**, *tegula* → **telya**, *mulier* →
**molyer**. Before i the contrast is neutralized and the sequence stays *li*.

**tc gives č.** The suffix *-āticum* syncopates to a tc cluster that neither
Caďinor nor Verdurian has, and it resolves as ts does: *viāticum* → **viačom**,
*formāticum* → **formačom**, *silvāticum* → **selvačom**. The suffix therefore
lands where French *-age*, Italian *-aggio* and Spanish *-azgo* land, by the
same route.

**Cluster resolution.** Latin x, fed in as *cs*, resolves two ways. Word-finally
it gives **č** (*nox* → **nöč**, *rādīx* → **razič**, *laxos* → **lač**), on the
analogy of the inherited ts → č and of Caďinor's initial cť, which resolves the
same way. Everywhere else the stop debuccalises and the cluster reduces to a
plain sibilant: *dexter* → **dëster**, *maximos* → **masem**, *uxor* →
**osor**, *exemplom* → **ešemplom**. Caďinor's initial ps gives Verdurian fs,
so a stop before a sibilant weakening rather than assimilating is the pattern
the language already has; and it is what Romance did, giving Italian *massimo*
and *esempio*. Neither language has medial cs, and Verduroman now has none
either.

Medial -čz- and -žz- reduce to č and ž. Because syncope can create fresh
clusters after the earlier passes have gone by, the dz and ts rules are run a
second time below it. That second pass is currently inert: the annihilation of
*-os* removes the environment before syncope reaches it, so *sordidos* gives
**södeð** with no cluster left to resolve. It is kept against the environment
recurring.

**The dental-loss placeholder resolves to zero.** Three rules in the list turn
a final short i, e or a into a placeholder before a dental, and a fourth
deletes the dental; nothing in the transcription removes the placeholder
itself. It is taken here to vanish, on the model of ny and ly reverting in
coda: *tenet* → **čen**, *bibit* → **beb**, *dormit* → **dörm**. *Amat* is
untouched at this stage, since the a-rule requires a preceding vowel.

**Nasals assimilate to a following stop**: n → m before a labial, m → n before
a velar (*numquam* → **nonkam**). The first is currently inert — it was written
for the cluster in *manibus*, which no longer forms now that the dative plural
is annihilated before syncope can reach it — but is kept against the
environment recurring.

The last two are late, and are what carry Verduroman past reading as Latin:

**Final s debuccalised and lost.** It weakens to h and disappears, centralising
the vowel it leaves behind. Only a long vowel ever kept an s this far, and the
outcome depends on where that vowel sat: **-ās** and **-ōs** both give **-î**,
while **-ēs** and **-īs** give **-i** and **-ūs** gives **-u** — the peripheral
vowels holding, the central ones falling together. This is what makes î a
plural ending, as it is in Verdurian.

**Final t lost**, through an intermediate glottal stop. It takes the third
person singular and plural of every verb, and a handful of nouns (*caput* →
**capo**).

Latin's own c before front vowels already becomes s by an existing rule, so
*centum* and *caelum* come out sibilant with no intervention.

### Output inventory

Derived, not designed: the following is what a 207-item core lexicon actually
produced.

**Consonants.** p b t d c g k · f v s z š ž · č · ð · m n ny · l ly r ř · y

**Vowels.** a e i o u · ë ö ä ü, and the diphthongs surviving from the
pre-processing stage.

Several observations follow from the counts rather than from any intention.

*h is written but never pronounced*, as in French *homme* and *hier*. Two
sources fall together in the spelling: Latin's own initial h, which no sound
change ever sees, and the Greek stratum's ch by way of x. Verdurian inherits
the same split, with Caďinor's h and ȟ both ending as a silent letter — except
that Latin's initial h is common enough to fill the role Caďinor's ȟ filled
alone.

*x and Þ do not occur.* Both are reachable only through the Greek stratum, and
a core lexicon contains no Hellenisms, so x belongs exclusively to learned
vocabulary. Þ is a special case: the inherited rules convert it to ð between
vowels and to t elsewhere, so it never surfaces at all and should be struck
from the inventory as a phoneme.

*ř is marginal*, at two occurrences in 206 — as against 90 for r, still the
most frequent consonant in the language by nearly two to one over the next
(n, at 49). That is the price of sourcing it
from cr- alone. It is a defensible frequency for a phoneme, but if ř is meant
to be characteristic of the language rather than incidental to it, restoring
the rr source is the lever.

*The palatalized allophones are scarce.* Only three core words keep one —
*ponyár*, *anyóst*, *folyom* — since coda position strips them and ny
outnumbers ly two to one in what remains,
because Latin supplies gn far more freely than -li-. Both then diverge at the
Modern stage, where ny reduces to y and ly does not — so Verduroman ends with
the palatal lateral and without the palatal nasal, which is an unusual place
to land and one it arrives at without being steered there.

*î is morphological, not lexical.* It has two sources, neither in a root: the
vocalised syllabic r of the syncopated third-conjugation infinitive, and the
accusative plural in -ās and -ōs once final s is lost. The second is the
important one, because it makes î a plural marker.
That is exactly the distribution Verdurian has,
where î marks the nominative plural and so appears constantly in running text
while occupying almost no dictionary entries — and Verduroman reaches the same
slot by its own route.

*e is the dominant vowel*, at 118 occurrences against 86 for a and 53 for o,
largely because the Latin infinitives and the third-declension endings all
converge on it. This is strongly Romance in character and needs no correction.

*Two words in three end in a consonant* — 119 of the 206 core forms. Most are
second-declension nominatives reduced to the bare stem (*lop, an, ocol, nöf,
bön, mal, sec, aub, man, lön, capél, nas, fum, vën, parf*), and the rest are
infinitives. This is the single largest determinant of how the language looks
on the page, and it places Verduroman well away from the vowel-final Italian
and Portuguese shape it had before the endings were corrected, closer to
Catalan or Occitan. It also determines how much of the lexicon the Modern stage will reshape:
67 of the 90 r's now stand in coda — roughly a
third of the core vocabulary — all of which lose the consonant outright two
centuries on.

### Orthography

Verdurian romanization conventions throughout. Latin's initial h is silent but
kept in the spelling, as in French: **hömo**, **hyeri**, **hi**. It is
invisible to every sound change and is restored afterwards, so it never affects
a derivation — only the look of the word. Stress is marked with an acute
where it is not predictable. No writing system is proposed.

Neither Verdurian nor Verduroman has ɲ or ʎ. What they have is nʲ and lʲ,
palatalized allophones of plain n and l, written here as the digraphs **ny**
and **ly**. Before i the palatalization is automatic and the contrast is
neutralized, so the sequence is written **li**: *foliīs* gives *foli*, not
*\*folyi*. It is also why neither survives in a coda, where there is nothing to
condition it.

### Diacritic disambiguation

Where two words fall together, the one carrying more weight in the sentence
takes an acute and the other goes bare — Spanish *tú* against *tu*, Italian
*dì* against *di*. The accent is not a stress mark: in most of these pairs
both members are monosyllables whose stress is entirely predictable, and the
acute is doing nothing but distinguishing a tonic word from an atonic one.

Two pairs exist in the current lexicon, and the first is the Spanish case
exactly:

| tonic | atonic |
|---|---|
| **čú** 'you' (*tū*) | **ču** 'your' (*tuos*) |
| **séne** 'ashes' (*cinis*) | **sene** 'without' (*sine*) |

A third comes from the clitics: **la** 'her' falls together with **lá** 'wide'
(*lātos*), and is resolved the same way.

---

## Morphology

### Nominal declension

Seven model nouns, one per Latin declension, run through the full paradigm.
Nothing here is designed; the tables are what the sound changes returned.

| | rosa 1f | mūros 2m | templum 2n | pater 3m | nōmen 3n | manus 4f | diēs 5m |
|---|---|---|---|---|---|---|---|
| nom sg | rösa | mur | čemplom | pater | nomen | mano | di |
| gen sg | röse | muri | čempli | pače | nomne | manu | dëi |
| dat sg | röse | muro | čemplo | pači | nomni | manü | dëi |
| acc sg | rösam | murom | čemplom | pačem | nomnem | manom | dëm |
| nom pl | röse | muri | čempla | pači | nomna | manu | di |
| gen pl | rösarom | murorom | čemplorom | patrom | nomnom | manvom | dërom |
| dat pl | röseb | mureb | čempleb | pačeb | nomeneb | maneb | dëb |
| acc pl | rösî | murî | čempla | pači | nomna | manu | di |

**The ablative is gone**, and is not tabulated. In every paradigm the ablative
singular falls together with the nominative, genitive or dative, and the
ablative plural is identical to the dative plural without exception, so
Verdurian's loss of the case does not have to be imposed on Verduroman — the
sound changes deliver it unprompted. Its functions pass to the dative.

Where the two senses would contradict each other the difference is carried by a
preposition, since a bare ablative means *from* while a bare dative means *to*.

**Eight cells reduce to six** in most declensions — seven in *nōmen*, five
in *diēs*. What survives everywhere is the
accusative singular in -m, the genitive plural in -rum, the plural in -î or -i, and a
consonant-final dative plural in **-eb**. Everything else merges. This is a working case system rather than a collapsing one, and it is
the -s — first surviving, then leaving its vowel behind — that makes it so.

**Two forms are rebuilt by analogy**, and both repairs are ones Romance made
too.

*Nōmen* had the only accusative singular in the system not ending in -m, since
Latin neuters have nominative and accusative alike. The oblique stem supplies a
replacement: *nōminem* gives **nomnem**, which slots into the paradigm exactly
and takes it to seven distinct cells, the most of any declension. This is not a
coincidence — *nōminem* is an accusative Latin never had and Vulgar Latin
invented, and it is the source of Spanish *nombre*, French *nom* and Italian
*nome*. The cost is the word's neuter identity: once the nominative and
accusative differ, it is a masculine, which is what happened to it in Romance.

*Templōrum* was the only cell in its paradigm with an unpalatalised stem, the
stress having moved off the root and left the diphthong unformed. It levels to
**čemplorom**. Nominative singulars resist levelling — *pater* keeps its t
against *pače-* — because they carry more weight; an oblique plural has no such
protection.

**The dative plural generalises.** Only the third, fourth and fifth
declensions inherit *-ibus*; the first and second had *-īs*, which gives **-i**
and collides badly. In the second declension *muri* would otherwise serve as
genitive singular, nominative plural and dative plural at once, and in the
neuter *čempli* as genitive singular and dative plural. The **-eb** ending
spreads to both — *röseb*, *mureb*, *čempleb* — because it is the one ending in
the system that resembles nothing else: consonant-final, in a paradigm
otherwise built on vowels. Every declension then marks the dative plural the
same way, and the second declension and the neuter each recover a cell.

**Levelling is blocked only by the nominative singular.** Where stress moved
off the root in the genitive plural, the root failed to diphthongise or
palatalise, leaving an odd cell: *rosarom* against *rös-*, *templorom* against
*čempl-*. Both level, to **rösarom** and **čemplorom**, because nothing else in
their paradigms preserves the unaffected stem and speakers have no reason to
remember an exception for one cell.

*Patrom* survives unlevelled because *pater* does. The nominative singular
carries enough weight to keep the plain stem in circulation, and the genitive
plural shelters behind it. That is the whole of the principle: an alternation
lasts exactly as long as the nominative singular shows it.

**Zero exponence marks the second-declension nominative singular.** *Mūros*
gives bare **mur**, *lupos* **lop**, *annos* **an**, *oculos* **ocol**. This is
the unmarked case taking the unmarked form, which is where zero belongs and
what Latin's own third declension already did with *rēx*, *pater* and *nōmen*.
Verdurian's masculine nouns end in consonants for the same reason.

It also survives the Modern stage: post-syllabic r is lost there, but
monosyllables are exempt, and Rosenfelder cites *mur* itself as an example of
the exemption.

**The fourth declension is saved by vowel length.** Latin distinguishes short
*manŭs* from long *manūs*, and only Rosenfelder's Caďinor-specific -US rule
collapses them; declining to apply it to the long vowel is declining to erase
a contrast Caďinor never had. The fourth declension then keeps a characteristic
**-o** against the second's **null**, and survives as a declension in its own
right instead of folding into the second as it did in Romance.

Its dative singular is also the one place ü appears outside a handful of
lexical items: *manuī* gives **manü** and *frūctuī* **fružü**. The genitive
plural is **manvom**, a short u in hiatus before another u becoming a glide,
and the dative plural **maneb**.

**The fifth declension survives after all.** It distinguishes five of its eight
cells, fewer than the others but enough to function. Romance moved it into the first
because it had nothing left to distinguish; Verduroman does not have to.

It also carries a stem alternation that is authentically Latin rather than
introduced: *diēs* has a long ē and *diem* a short one, so the nominative gives
**di** while the accusative diphthongises to **dëm**. The paradigm is *di*
against *dë-* throughout, on the same pattern as *nöč* against *nožem*.

### Pronouns and determiners

#### Personal pronouns

| | nom | acc | dat |
|---|---|---|---|
| 1sg | yego | me | me |
| 2sg | čú | te | tebe |
| 3sg m | ele | elom | elü |
| 3sg f | ela | elam | eli |
| refl | — | se | sebe |
| 1pl | nî | nobi | nobi |
| 2pl | vî | vobi | vobi |
| 3pl m | eli | elî | elorom |
| 3pl f | ele | elî | elorom |

*Egō* gives **yego**, which is where Spanish *yo* and Italian *io* come from as
well.

The third person carries more load than the inherited forms can bear, and it
relieves it the same two ways Romance did. Left alone, *illī* would serve as
dative singular, masculine nominative plural and dative plural at once, all as
**eli**. Instead:

* The dative plural is taken over by the **genitive** *illōrum* → **elorom**,
which is Italian *loro* and the Romanian suffixed article *-lor*. A genitive
doing dative work is irregular, but it is what the load required.
* The masculine dative singular is rebuilt as *illuī* → **elü**, on the model
of *cui*. This is Italian and French *lui*, and it looks very much as though
Proto-Romance speakers coined the form for exactly this reason — to keep the
dative singular apart from the nominative plural.

The feminine keeps the inherited **eli**, which it can afford to because the
clash was masculine. **Elî** still covers both accusative plurals, the same
merger the nouns show.

#### The article

The article does not come from *ille*. It comes from *ipse*, reduced as a
proclitic — which is what Sardinian did, against the *ille* of every other
Romance language, and which is why Verdurian's articles begin with s.

| | m | f | n |
|---|---|---|---|
| nom sg | se | sa | so |
| acc sg | som | sam | so |
| dat sg | sü | si | sü |
| nom pl | si | se | sa |
| acc pl | sî | sî | sa |
| dat pl | sorom | sorom | sorom |

Against Sardinian *su, sa, sos, sas*: the same source, the same reduction. The
unreduced forms survive alongside as **efse**, **efsa**, **efsom** 'self', so
*ipse* keeps the reduced/unreduced contrast that *ille* has between article and
pronoun elsewhere in Romance.

The reduction is apheresis of the unstressed first syllable. A proclitic is not
subject to the changes that shaped the stressed forms, so the article keeps
material the free pronoun has lost. The masculine dative **sü** is *ipsuī*,
rebuilt on the same analogy that gave *illuī* → *elü*.

The article has no genitive: possession is carried by the possessives, which
cover that ground entirely. It does inflect for the dative, as Romanian's does.

#### Object clitics

*Ille* reduced is not wasted. It supplies the object clitics, which is again
what Sardinian does — article *su*, clitic *lu* — and what Spanish *lo, la, los,
las* and French *le, la, les* are.

| | m | f | n |
|---|---|---|---|
| acc | lom | lam | lo |
| dat | lü | li | lü |
| acc pl | lî | lî | la |
| dat pl | lorom | lorom | lorom |

Each form now has exactly one job: **ele** is the free pronoun, **lom** the
clitic, **som** the article.

#### Possessives and demonstratives

Singular:

| | m | f | n |
|---|---|---|---|
| my | më | mëa | mëom |
| your | ču | ča | čom |
| his, her | su | sua | suom |
| their | lorom | lorom | lorom |
| our | nöster | nöstra | nöstrom |
| your (pl) | vëster | vëstra | vëstrom |
| this | hec | hyec | höc |
| that | ete | eta | etom |
| self | efse | efsa | efsom |

Plural, in the nominative; the other cases follow the adjective declensions:

| | m | f | n |
|---|---|---|---|
| my | mëi | mëe | mëa |
| your | čui | če | ča |
| his, her | sui | sue | sua |
| their | lorom | lorom | lorom |
| our | nöstri | nöstî | nöstra |
| your (pl) | vëstri | vëstî | vëstra |
| this | hi | hye | hyec |
| that | eči | ete | eta |
| self | efsi | efse | efsa |

**Lorom** is invariable for gender and number, as Italian *loro* is, and is the
same form as the article's dative plural — one word doing the work of 'them'
and 'their', which is what *loro* does too.

*Tuos* gives **ču**, falling together with the pronoun *tū*; the pair is
distinguished as **čú** against **ču**, which is the Spanish *tú*/*tu*
distinction reached independently.

*Nostrae* and *vestrae* give **nöstî** and **vëstî** rather than the expected
*nöstre*: the final e is lost after r and the syllabic r vocalises, exactly as
in the third-conjugation infinitives. *Istī* gives **eči**, palatalising before
the high vowel, so 'that' alternates *ete*/*eči* across number.

The *hic* series survives only because the spelling keeps its h. *Hī* is **i**
phonologically — a single segment, unusable as a demonstrative — but is written
**hi**, and the paradigm stays legible. Romance let the letter go and had to
rebuild its demonstratives on *iste* and *ipse* instead.

#### Interrogatives

**ke** (*quis*), **ked** (*quid*), **cëm** (*quem*), **ci** (*quī*), **cë**
(*quae*), **köd** (*quod*), **kam** (*quam*), **coyo** (*cuius*), **cue**
(*cui*); **obe** (*ubi*), **kando** (*quandō*).

The two velars separate here on vowel length. *Quis* has a short i, which
merges to e and leaves the k alone, giving **ke**; *quī* has a long ī, which
survives to trigger the palatalisation, giving **ci**. Two forms of the same
stem, split by a distinction Caďinor never made.

### Numerals

#### Cardinals

| | | | |
|---|---|---|---|
| 1 | un, una, unom | 11 | unsem |
| 2 | duo, due | 12 | döðsem |
| 3 | tri, triá | 13 | treðsem |
| 4 | kator | 14 | katödsem |
| 5 | cüke | 15 | cíndesem |
| 6 | šeč | 16 | seðsem |
| 7 | šeftem | 17 | seftënesem |
| 8 | ožo | 20 | vižeti |
| 9 | növem | 100 | šetom |
| 10 | dësem | 1000 | mile |

The tens are regular on *-gintā*, which gives **-žeta**: **trižeta**,
**kadražeta**, **kinkažeta**, **sesažeta**, **seftuažeta**, **ožožeta**,
**nonažeta**. Twenty keeps *-gintī* and so ends in **-žeti**. The hundreds
contract sharply: *ducentōs* and *trecentōs* give **došë** and **trešë**.

*Ūnos* and *ūna* double as the indefinite article, as everywhere in Romance.
Two keeps its gender — **duo** against **due** — which Spanish and Italian lost
but Portuguese and Romanian kept.

Eighteen and nineteen are analytic, as in Vulgar Latin: *decem et octō* rather
than *duodēvīgintī*.

#### Ordinals

There are three: **prim**, **secón** and **čëti**. Everything above uses the
cardinal.

The inherited series could not survive its own phonology. *Sextos* falls
together with *sex* as **šeč**, *septimos* with *septem* as **šeftem**,
*decimos* with *decem* as **dësem**, and *nōnos* with *nōn* 'not' as **non** —
four collisions in the first ten, three of them between an ordinal and the
cardinal it is built on, which is the least useful place for a language to lose
a distinction. Spanish abandoned its ordinals above tenth for much milder
reasons; Verduroman abandons them above third.

Ordinal sense is carried by the cardinal, following the noun as in Spanish
*el capítulo tres*.

Two of the discarded forms survive lexicalised rather than as ordinals.
**Kar** and **ožáf** — with the feminines **karta** and **ožava** — keep the
senses of a quarter and an eighth, in fractions and in the printing formats
that gave English *quarto* and *octavo*. They no longer mean 'fourth' and
'eighth'. Alongside them **mëzi** (*medios*) gives a half.

### Adjectives

Adjectives decline exactly as the nouns of the corresponding class, so the
paradigms are not repeated here. Two classes survive.

The first, from Latin's first and second declensions, keeps three genders:
**bön**, **böna**, **bönom** (*bonos, bona, bonom*), with accusative **bönom**,
**bönam**, plural **bönî** and dative **böni**.

The second, from the third declension, has lost its gender distinction
entirely. *Gravis* and *grave* both give **grave**, as *fortis* and *forte*
both give **förte** — the same collapse behind Spanish and Italian *grave*.
Only the neuter plural stays apart, as **grávia**.

**Ācer** is worth its own line. The masculine nominative *ācer* gives **aser**,
with c palatalising to s before a front vowel; but the oblique stem *ācr-* has
the cluster instead, and takes the cr → ř rule, giving **aře**. The paradigm
therefore alternates *aser* against *aře-*, on exactly the pattern of *pater*
against *pače-*.

#### Comparison

Comparison is analytic, with **maže** (*magis*) before the adjective — the
choice Spanish, Catalan, Occitan and Romanian made, against the *plūs* of
French and Italian. **Plu** exists but is not the productive form. 'Than' is
**kam** (*quam*).

Four suppletive comparatives survive inherited and irregular, as they do
throughout Romance: **melyor**, **pëyor**, **mayor**, **menor** — *mejor*,
*peor*, *mayor*, *menor* exactly.

The synthetic superlative does not survive. Latin's *-issimus* series erodes to
forms that carry no obvious relation to their positives — *optimos*,
*pessimos*, *maximos*, *minimos* give **öftem**, **pëssem**, **masem**,
**menem** — and the superlative is expressed with the article and the
comparative, as in *se maže bön*. Romance did the same, and its *óptimo* and
*máximo* are later borrowings rather than inherited forms.

#### Adverbs

Two inherited adverbs remain in ordinary use, **bëne** and **male** (*bene*,
*male*), as *bien* and *mal* do.

The productive formation is the feminine adjective plus **-mëte**, from the
ablative *mente*: *böna* + *mëte* gives *bönamëte* 'well'. This is the only
place the ablative survives at all. The phrase fossilised before the case was
swallowed by the dative, so the adverb preserves a form the grammar otherwise
no longer has — which is why it is **-mëte** and not the dative **-mëti**.

### Verbal conjugation

The infinitives sort into five classes, not Romance's three.

| Class | Count | From | Examples |
|---|---|---|---|
| -ár | 14 | Latin -āre | *potár, volár, legár, lavár* |
| -ér | 11 | Latin -ēre | *vedér, tenér, temér, olér* |
| -ír | 4 | Latin -īre | *dormír, venír, ferír, azír* |
| -î | 11 | Latin -ere, syncopated | *bebî, vivî, caðî, luðî* |
| -Vr | 14 | Latin -ere, unsyncopated | *fender, vërter, pëler, dir* |

The Latin third conjugation's unstressed *-ere* splits in two according to
whether syncope applies. Where it does, the r is left syllabic and vocalises,
giving the **-î** class; where it does not, a vowel survives before the r and
the outcome falls in with the others. Romance found the whole ending unstable
and absorbed the third conjugation into the second and fourth, which is why
Spanish has *beber* and *vivir*. Verduroman instead keeps it, twice over.

Five classes is nearer to Caďinor's five than to Romance's three, and the
resemblance is not accidental in kind even though it is in detail: both
languages arrive at their count by letting sound change split an inherited
class rather than by tidying it.

#### Present indicative

| | amár | tenér | dormír | bebî | vërter |
|---|---|---|---|---|---|
| 1sg | amo | čénio | dörmio | bebo | vërto |
| 2sg | amî | čeni | dörmi | bebe | vërte |
| 3sg | ama | čen | dörm | beb | vërt |
| 1pl | amám | teném | dormím | bebem | vërtem |
| 2pl | amate | tenete | dormite | bebte | vërtete |
| 3pl | aman | čenen | dörmiu | bebu | vërtu |

All six persons are distinct in every class. The first person plural is bare
**-m**: Latin *-mus* is another ending in short -us, from PIE *-mos, so it is
annihilated like the rest and leaves only the nasal.

**Verduroman has reinvented the Spanish radical-changing verb, in two
flavours.** *Dörmio* against *dormím* is *duermo* against *dormimos* exactly,
and by the same route — stressed ŏ diphthongises, unstressed ŏ does not. But
*čénio* against *teném* is a **consonantal** alternation: the t palatalises
only where the stressed ié appears, so the stem alternates in its first segment
rather than its vowel. Spanish has no counterpart. Same conditioning, different
exponent.

#### Other tenses of amár

| | 1sg | 2sg | 3sg | 1pl | 2pl | 3pl |
|---|---|---|---|---|---|---|
| imperfect | amabam | amabî | amaba | amabám | amabate | amaban |
| future | amabo | amabe | amáb | amabem | amabte | amabu |
| perfect | amavi | amavesti | amáv | amavem | amaveste | amaveru |

The imperfect is regular throughout once length shields the stem vowel. The
perfect is the one paradigm still taking damage: *amāvistī* gives **amavesti**
and *amāvērunt* **amaveru**, and syncope has begun to obscure the -v- that
marks the tense. Whether to let analogy rebuild it is open.

---

## Syntax

Everything above was derived. This section is not, and cannot be: sound changes
do not produce a word order. What follows is argued instead — from what
Verduroman inherited, from what survived the phonology, and from what Romance
did under the same pressures — and where the argument is weak it is said so.

### Constituent order

The pressure that turned Latin's free order into Romance's fixed SVO was the
loss of case. Verduroman did not lose case: eight cells still yield six
distinct forms, and the accusative in **-m** is intact throughout. So the
pressure is much weaker here, and Verduroman keeps a good deal of Latin's
pragmatic ordering, with SVO as the unmarked default rather than the only
option. Fronting for topic and focus remains available in a way it is not in
Spanish or French.

### The article

The article is preposed, and this is one syntactic fact the derivation does
settle. **Se** and **sa** arise by apheresis of *ipse* and *ipsa*, and apheresis
of an initial unstressed syllable is a proclitic change — it only happens to a
word leaning on what follows it. A postposed article, as in Romanian, would
have reduced at the other end. The article therefore precedes its noun and
agrees with it in gender, number and case.

### Use of the cases

* **Nominative** — subject, and predicate nominal.
* **Accusative** — direct object, and the object of a preposition of motion.
* **Dative** — indirect object, plus everything the ablative used to do:
source, instrument, means, accompaniment. Where the two senses would
contradict each other a preposition carries the difference, since a bare
dative reads as *to* and the old ablative sense was *from*.
* **Genitive** — possession and partitive. Unlike Romance, which rebuilt
possession on *dē*, Verduroman still has the inherited genitive and uses it.
**De** exists and competes with it, as it did in late Latin.

### Prepositions

**de**, **ad**, **en**, **com**, **për**, **pro**, **sob**, **soper**, **ate**,
**pös**, **sene**, **eter**. They govern the accusative where motion or
extension is involved and the dative otherwise, which is Latin's own split with
the ablative's share transferred.

### Negation

**Non** precedes the verb, as in every Romance language. Verduroman has not
undergone the Jespersen reinforcement that produced French *ne … pas*; there is
no second element. **Nemo**, **nel** and **nonkam** (*nēmō*, *nihil*,
*numquam*) are the negative pronouns and adverb.

### Questions

Content questions front the interrogative: **ke**, **ked**, **ci**, **obe**,
**kando**, **konðo** (*quōmodo*). Polar questions are marked by intonation
alone — Latin's *-ne* and *num* left no reflex worth using, and Romance did the
same.

### Clauses

The complementiser is **ke**, from *quid* rather than *quod*, on the Romance
pattern that gave *que* and *che*. Relatives use the inherited **ci**, **cë**,
**cëm**, which still distinguish case and so need no resumptive pronoun — a
point where the surviving morphology buys Verduroman something Romance had to
give up.

Subordinators: **kando** 'when', **atkam** 'before', **pöstkam** 'after',
**dom** 'while', **kamvi** 'although', **ciá** 'because'.

### Second person

**Čú** and **vî** are singular and plural. Whether the plural has taken on
polite singular use — the T/V distinction Romance developed from *vōs* — is
undecided and is a matter of pragmatics rather than of grammar; nothing in the
derivation forces it either way.

---

## Example sentences

Chosen to exhibit how the language works rather than to read naturally: each
shows something the grammar does that Romance does not, or does differently.

**Se hömo veð som canem.**
&nbsp;&nbsp;*The man sees the dog.*
Article and noun both inflect; the article is the *ipse* one, as in Sardinian.
The object is marked by the accusative in **-m**
as well as by position, so the order is a default rather than a requirement.

**Veðém tri yekî.**
&nbsp;&nbsp;*We see three horses.*
The first person plural is bare **-m**; the accusative plural is the **-î** that
final s left behind. No article — Verduroman uses it as Romance does, not as
Latin did without one.

**Non do som levom si femne.**
&nbsp;&nbsp;*I am not giving the book to the woman.*
Negation is a single preverbal **non**. The recipient is a bare dative with no
preposition, which Romance cannot do.

**Se nomen sua yes lön.**
&nbsp;&nbsp;*Her name is long.*
Possession by a pronoun uses the possessive; possession by a noun uses the
surviving genitive. *Femne* is genitive singular, dative singular and
nominative plural all at once, and the article resolves the last two: **si
femne** is 'to the woman', **se femne** 'the women'. Where the noun has merged,
the article has not — which is much of what the article is for.

**Hyec aka yes maže frižða kam etom vinom.**
&nbsp;&nbsp;*This water is colder than that wine.*
Comparison with **maže** and **kam**. The adjective agrees in gender with a
feminine subject and the demonstratives agree with their own nouns.

**Si púeri vëniu de agreb com pači.**
&nbsp;&nbsp;*The boys come from the fields with their father.*
Two prepositions on the dative, which now carries what the ablative used to:
**de agreb** is source, **com pači** accompaniment. **Agreb** is the generalised
dative plural.

**Ke šrifs sam epéstolam?**
&nbsp;&nbsp;*Who wrote the letter?*
Content question, interrogative fronted. *Scrīpsit* gives **šrifs** — the ps of
the perfect stem takes the same fs treatment Caďinor's initial ps does.

**Non hab som levom cëm ču me deðesti.**
&nbsp;&nbsp;*He does not have the book that you gave me.*
The relative is accusative **cëm**, not the nominative **ci**, because it is the
object of the subordinate clause. Verduroman marks that on the pronoun; Romance
lost the distinction and repairs it with a resumptive.

**Čú veði ču pačem.**
&nbsp;&nbsp;*You see your father.*
The two **ču** are the pronoun and the possessive, distinguished only by the
acute — *tū* and *tuus*, exactly as in *tú* and *tu*.

**Se hömo da som levom; si hömeni dan sî levî.**
&nbsp;&nbsp;*The man gives the book; the men give the books.*
Singular against plural throughout: **da** against **dan**, **se** against
**si**, and the accusative plural in **-î** against the singular in **-m**.

**Dörmio, šed non dormím.**
&nbsp;&nbsp;*I sleep, but we do not sleep.*
The radical-changing verb in its vocalic form. Stress falls on the root in the
singular and diphthongises it; it moves to the ending in the plural and does
not. This is *duermo*/*dormimos* arrived at independently.

**Čénio som levom; teném sî levî.**
&nbsp;&nbsp;*I hold the book; we hold the books.*
The same alternation in its consonantal form. Here the stressed diphthong
palatalises the initial t rather than surfacing as a vowel, so the stem changes
in its first segment. Spanish has no counterpart to this.

**Bebî yes bön.**
&nbsp;&nbsp;*Drinking is good.*
The third-conjugation infinitive, whose syllabic r vocalised to **-î**. It is
used as a verbal noun and takes a masculine predicate adjective.

**Se puer bebe sam akam lëtamëte.**
&nbsp;&nbsp;*The boy drinks the water slowly.*
The adverb is the feminine adjective plus **-mëte**, a fossilised ablative —
the only place that case survives at all.

**Se veð en si ake.**
&nbsp;&nbsp;*He sees himself in the water.*
The reflexive **se**, homophonous with the masculine article but never in its
position — an article precedes a noun, a reflexive a verb. The preposition of
location takes the dative, which now covers what the ablative used to.

**Dic ke se cane yes man.**
&nbsp;&nbsp;*He says that the dog is big.*
The complementiser **ke**, from *quid* rather than *quod*, on the pattern that
gave *que* and *che*. *Magnos* gives **man**: the coda strips the palatal, so
the masculine loses the ny that the feminine **manya** keeps.

**Se folë caðu de si árbori.**
&nbsp;&nbsp;*The leaves fall from the tree.*
**Folya** was Latin's neuter plural *folia*, reanalysed as a feminine singular —
so this plural is a plural of a plural, built afresh on the new stem. That is
the *foglia* and *hoja* development, replayed.

**Ven sü dëi kator.**
&nbsp;&nbsp;*He came on the fourth day.*
There is no fourth: above third the cardinal does the work, following its noun,
because the inherited ordinals collided with the numbers they were built on.
The bare dative carries 'on', one of the functions it took over from the
ablative, so no preposition is needed.

**Ča mano yes frižða.**
&nbsp;&nbsp;*Your hand is cold.*
The fourth declension, saved by vowel length and feminine like its Latin
source, with an adjective agreeing. **Ča** is the feminine of the possessive,
distinct from **ču** and from **čú**.

**Kando vëniu si púeri?**
&nbsp;&nbsp;*When are the boys coming?*
Interrogative fronted; no auxiliary and no inversion beyond the fronting
itself, since the verb already marks person and the article marks number.

---

## Lexicon

The lexicon proper is a separate document, as Rosenfelder's are. It runs to
1,729 entries, includes the Swadesh list, and gives the Latin source for every
headword.

Because the nominative singular is so often a bare stem, it reveals little of
the rest of the word, and the lexicon gives the accusative singular wherever it
is not simply the nominative plus an ending. A quarter of the entries need it:
**pi** beside **pëðem**, **hömo** beside **hömnem**, **reč** beside
**režem**.

### Erosion and replacement

Latin words that eroded to almost nothing were generally not kept: Romance
replaced or reinforced them. *Ōs* 'mouth' gave way to *bucca*, *crūs* to
*gamba*, *rēs* to *causa*, *ignis* to *focus*; *auris*, *apis*, *agnus* and
*avis* were rebuilt with diminutive suffixes; *diēs* survived in Spanish but
was displaced by *diurnum* in Italian and French.

Verduroman needs the same policy, but the words requiring it must be found
from its own output rather than borrowed from Romance's list. Running the core
lexicon and isolating everything that falls to two segments or fewer gives:

* **Function words and numerals** — *ču, nî, vî, ec, ic, ke, un, ad, en, ye,
si*. These need no repair. Function words are supposed to be short, and
numerals are the class most tolerant of irregular shortening.
* **Content words** — *os* 'bone' → **ué**, *ōs* 'mouth' → **î**, *pēs* 'foot'
→ **pi**, *crūs* 'leg' → **řu**, *diēs* 'day' → **di**, *annos* 'year' →
**an**, *auris* 'ear' → **är**, *lātos* 'wide' → **la**, *vetos* 'old' →
**vë**.

Eight of those nine are exactly the items Romance replaced — *ōs* by *bucca*,
*os* remade as *ossum*, *crūs* by *gamba*, *diēs* by *diurnum*, *auris* by
*auricula*, *lātus* by *amplus*, *vetus* by the diminutive *vetulus* that gave
*viejo* and *vieux*. That is a strong sign the erosion threshold is a real
property of the changes rather than an artefact of either language's history.
*Annos* is the exception and needs no help: Romance kept it (*año*, *anno*,
*an*), and **an** is a perfectly ordinary monosyllable.

Verduroman's threshold does sit differently from Romance's, and in a specific
direction. **Romance had one form per noun; Verduroman has six.** A thin
nominative is not the crisis it was for a caseless language: *pēs* gives **pi**,
which looks untenable until the accusative *pedem* is set beside it as
**pëðem**. Romance reinforced *pēs* into *pedem*-based forms because it had
nowhere else to put the information; Verduroman does not need to.

That defence has narrowed, though. When the paradigm was collapsing to four
distinct cells the case system could carry very little; now that it holds six,
it carries more — but the late loss of s and t has thinned the citation forms
at the same rate. The replacement set is no longer the proper subset of
Romance's it was two stages ago. *Avis* → **ave** remains a genuine divergence,
kept where Romance rebuilt it as *avicellus*.

### Homophony

The core lexicon currently has no collisions, but it has had two, and both were
resolved the way Romance resolved them. *Vir* 'man' fell together with *vēros*
'correct' as **ver**, and was replaced by the *homō* descendant **hömo** —
exactly the substitution behind *hombre*, *homme* and *uomo*. *Dēns* and *diēs*
briefly fell together as **dës**; the late loss of final s separated them again
on its own, giving **dës** and **di**, so no replacement was needed. Romance
had reached for *diurnum* in the same position.

[Raphael]

I don't really have the linguistic or conlanging knowledge or understanding to properly form and opinion about this, but I like how the resulting language still has a bit of a Verdurian feel to it. Interesting contrast to actual modern Romance languages.

[bradrn]

Some of the phrasing there feels decidedly odd to me. Were LLMs used to generate any of that text?

(I sincerely hope that I am wrong in suggesting this and that it’s merely my LLM-dislike misfiring…)

[zompist]

This is really impressive work... very thorough, with an evident understanding of both Romance and Verdurian philology.

I'm curious how ipse would develop. Probably just into ifse. But abbreviated it might produce a parallel to Sardinian su/sa and Verdurian so/sa.