I’m very bad at finishing romlangs (and very good at starting them), but hopefully this thread will incentivise me to do something. I’ve been working on two romlangs to experiment with different ways of evolving ergativity from a nominative-accusative ancestor; the first (Vissard) is a galloromlang which evolved a split-S system encouraged by an earlier marked nominative; the second (Illyrian) is an Eastern romlang in which the ergative structure derives from an old passive perfect construction, which was later generalised to all aspects. Interestingly, the Latin nominative evolved to the Vissard ergative but the Illyrian absolutive; while Vissard’s absolutive derives from the accusative and Illyrian’s ergative from the dative. This post, however, is only about Vissard, because I've been working on it for longer and the phonology is more interesting. Illyrian will be coming soon though...
What is Vissard?
Vissard is a galloromlang spoken in the althist island of Visserland, which is in the North Sea about 60 miles off the coast of England. It's equivalent to the OTL Dogger Bank. Dogger Bank is the remnant of the ancient landmass Doggerland which connected England and Europe. In the Vissard timeline, the Dogger Bank moraine was about 100 feet higher and remained above water, forming an island 160 by 60 miles in size. Some Vissard linguists have proposed that Vissard's ergativity was influenced by a pre-Roman Visserlandic language; they suggest that Visserlandic was an ergative language, potentially a Vasconic one. However, ergativity likely developed too late for this to be tenable.
I’ll just start by outlining a bit of Vissard phonology. After working through a number of inventories, I’ve settled on this one, inspired mostly by French/Franco-Provençal/Occitan.
/p b t d k ɡ/
/f β s z ʃ ʒ h ʁ/
/m n ɲ ŋ/
/i y u/
/ɛ œ ɔ/
Unusually for a romlang, /h/ and /ŋ/ are both phonemic segments. /h/ derives from loan-words (e.g. Old Norse hulfr "holly" > /hylˈβɑ/) and /ŋ/ from analogised coda nasals (e.g. /dɔ ˈdɔŋ/ "I give" analogises to /duˈŋɛ/ "to give").
Major historical developments
Vissard of course underwent a number of sound changes, but the two most distinctive are the evolution of vowels and the palatalisation of velars.
Western Vulgar Latin (WVL) had seven vowels */i u e o ɛ ɔ a/ with long and short variants; originally allophonic but later made phonemic. Old Vissard (OV) merged short and long /i u a/, while long /eː oː ɛː ɔː/ were diphthongised to /ei̯ ou̯ i̯ɛ u̯ɔ/. In unstressed syllables, /ɛ ɔ/ merged in favour of /e o/. Word-final vowels (or unstressed vowels before a word-final /s/) other than /a/ were usually dropped except following a consonant cluster that did not follow the sonority hierarchy (e.g. PRESBYTERUM > /prɛstro/, PATREM > /patre/). /a/ was also sometimes dropped (AQUA > /a(ɣ)w/, CASA > /c͡çai̯s/). Some consonants precipitated a yod onto the preceding vowel (word-final stressed /l n s/ and all palatals).
Old Vissard vowels
/i u e o ɛ ɔ a/
/ei̯ ou̯ oi̯ ai̯ au̯ i̯ɛ u̯ɔ i̯ɛi̯ u̯ɔi̯/
atonic /i u e o a/
Middle Vissard (MV) then underwent a major monophthong shift, probably started by a raising of /e o/ to distinguish them from /ɛ ɔ/. This then impinged on /i u/’s vowel space; /u/ was fronted to /y/ (as is areally very common; it was probably /ʉ/ in Old Vissard) while /i/ was centralised to /œ/ (reminiscent of New Zealand English’s /æ ɛ ɪ/ → /ɛ ɪ ə/ shift). Then /e o/ were fully raised to /i u/. This also applied to unstressed /i u e o a/ which became /œ y i u a/. The diphthongs /ei̯ ou̯/ were spread to /ai̯ au̯/ and at least in one dialect area /i̯ɛ u̯ɔ/ likewise became /i̯a u̯a/.
Middle Vissard vowels
/i y u ɛ œ ɔ a/
/ai̯ au̯ i̯ɛ u̯ɔ i̯ai̯ u̯ai̯/
atonic /i y u œ a/
Modern Vissard then backed /a/ to /ɑ/ and smoothed the diphthongs /ai̯ au̯ i̯ɛ u̯ɔ/ to /ɛ ɔ i u/. The triphthongs /i̯ai̯ u̯ai̯/ became /i̯ɛ u̯ɛ/ and then /jɛ ɥɛ/. By reassignment of stress, atonic /ɛ ɔ/ were reindtroduced.
/i y u ɛ œ ɔ ɑ/
This means that the reflexes of WVL */e o/ and */ɛ ɔ/ are switched between open and closed syllables (i.e. long vs. short in WVL). Long /eː oː ɛː ɔː/ give /ɛ ɔ i u/, while short /e o ɛ ɔ/ give /i u ɛ ɔ/.
In Franco-Provençal, stress reassignment to the final syllable is common but irregular (e.g. FARĪNA > /farˈna/, CUBITUM > /kuˈdu/). In Vissard, this was completely regular and affected all words, including the swath of feminine nouns ending in /ɑ/. This shift occurred in the 18th century, perhaps due to French influence.
As in mainland French, Latin velars were palatalised in two positions. Firstly, before Latin Ī I Ē E Æ Œ, which happened in WVL. Secondly, before Ā A, which happened much later. The first palatalisation gave WVL */t͡ʃ d͡ʒ/; in French these became /t͡s d͡ʒ/ (later > /s ʒ/), while in Vissard they simply deaffricated to give /ʃ ʒ/.
CĒRAM > /ʃɛˈʁɑ/ “wax”
CĪVITĀTEM > /ʃœˈtɑ/ “city”
GINGĪVA > /ʒiˈɲœ/ “gum”
GENUCULUM > /ʒiˈnɛl/ “knee”
The second palatalisation gave some more interesting results. Pre-Old Vissard likely had simple velar + semivowel sequences */kj ɡj/; in OV they were most likely /c͡ç ɟ͡ʝ/. They were distinguished from /t͡ʃ d͡ʒ/ based on the category ±sibilant; this allowed them to front to MV /t͡θ d͡ð/. Finally, Vissard deaffricated them to /t d/, merging with existing /t d/. (Think that’s strange? Franco-provençal dialects have any of /θ ts s f st h ɬ fl/ as reflexes too).
CAPRAM > */kjapra/ > /tɑˈβʁɑ/ “goat”
CABALLUM > */kjaβalo/ > /c͡çaˈai̯l/ > /tɛl/ “horse”
GALBINUM > */ɡjalb(i)no/ > /dɛˈnu/ “yellow”
GAMBAM > */ɡjamba/ > /dɑŋˈbɑ/ “leg”
In stressed syllables /i y u ɛ œ ɔ ɑ/ are generally <i u o e ue ô a>; in unstressed syllables <e u o ê i ô a>. Semivocalic [j ɥ w] are <y (h)u w>. /p b t d f β ʃ ʒ h ʁ l m n ɲ ŋ/ are <p b t d f v ch j h r l m n gn n>, although <h> may also be silent. Where /t d/ derive from CA GA, they are irregularly <th dh>. /s z/ are generally written <s z> word-initially and <ss s> between vowels. “Soft” <c> (before <i e>) is /s/ while soft <g> is /ʒ/; this use of <c> is unetymological and derives from French. /k ɡ/ are <qu~k gu> before <i e> and <c g> otherwise, except the sequence /kɥ/ is <qu(h)>. <n> is always /ŋ/ word-finally or before a consonant and always /n/ word-initially or non-word-finally after a consonant; otherwise it may be ambiguously either of /n ŋ/ (e.g. dona /duˈŋɑ/ "he gives" vs. dôna /dɔˈnɑ/ "woman").
Well, I promised you two ergative romlangs and delivered one romlang without addressing its ergativity at all. We'll get there eventually!