I was reading up about Gaulish (and by extension pre-Roman Celtic) phonetics as hypothesised by analysing Gaulish writing, and one writer brought up this interesting observation: at least in the Narbonensis dialect area of Gaulish, there is some confusion between which graphemes should represent particular phonemes (in both Greco-Gallic and Romano-Gallic writing).
Edit: (Source can be found on
JSTOR.)
1. Intervocalic /b t k/ are interchangeably represented by <B T C/K> (primarily) and <PH TH/Θ CH/Χ> (secondarily)
eg. akito- > Gr. <ΑΧΙΤΟ-> and La. <ACITO->; mako- > La. <MACO> and La. <MACHO>; katu- > La. <CATHU-> and Gr. <ΚΑΤΟΥ->; dubios > La. <DUPHIUS> and La. <DUBIUS>
2. Intervocalic /b d g/ are interchangeably represented by <B D/Δ G/Γ/∅> (primarily) and <P/Π T C/Κ> (secondarily)
eg. magiorixs > La. <MAIORIX> and La. <MAGIORIX>; -brigi > La. <-BRIGI> and La. <-BRICI>; arisiton > La. <ARISITUM> and La. <ARISIDUM>; ebidios > La. <EPIDIUS> and La. <EBIDEOS>
The conclusion made by this writer is this interchange did not indicate lenition per se, but at least indicated a contrast between the Greco-Roman phonologies and Gaulish phonology that is to be met.
Now the writer suggested that this interchange was to meet a supposed
missing fortis-lenis/tense-lax distinction in Gaulish; while Latin had a primary voicing distinction and a secondary fortis-lenis distinction (voiced=lenis, unvoiced=fortis, geminate=fortis), Gaulish rather may have had only lenis singleton plosives (voiced=lenis, unvoiced=lenis, geminate=fortis).
I want to explore this for a bit because I'm struggling to see how this plays out with what we know of fortis-lenis and diachronic change. As a brief tangent, let's look at the phonological systems we do know about (considering only the plosives at this time).
First, we know what Greek phonology and Latin phonology looked like at the time:
Greek was in the process of (or had completed) its voiced consonant lenition as of Roman times, but the Greek alphabet was borrowed into Western Europe (and indeed, as the predecessor of the Etruscan and Roman alphabets) before this occurred, so the Greek alphabet was interpreted according to the ancient phonology. There were two common variants of the Greek alphabet: western (the ancestor of the Italic alphabets) and eastern (the ancestor of Modern Greek; also used by Greek colonists); for efficacy, the western letters are written below with the Latin letters closest in form. Greek also had geminates, although these weren't quite so common as in Italic and Celtic.
Unvoiced: /p t k/ <P T K/Q> <Π Τ Κ> (<Q> was used for /k/ before back vowels, and subsequently was adopted for /kʷ/ in Italic)
Voiced: /b d g/ <Β D C> <Β Δ Γ> (<G> is innovated in Latin)
Aspirate: /pʰ tʰ kʰ/ <Φ Θ Ψ> <Φ Θ Χ> (<X> represented /ks/ in Western Greek and retained this value into Italic writing)
Geminate: /pp (tt) kk/ <PP TT KK> <ΠΠ TT ΚΚ>
In later Greek, the phonology changes thus:
/p t k/ unchanged; but NT > (N)D
/b d g/ > /β ð ɣ/
/pʰ tʰ kʰ/ > /ɸ θ x/
/pp kk/ > /p k/ (other geminates likewise merged with their singleton counterparts)
----
Latin phonology possessed much a similar system minus the aspirates, except Latin also had allophonic aspiration (in contrast to Ancient Greek): stops were mildly aspirated initially and near liquids. So the primary distinction was certainly voicing, but Latin also had a secondary fortis-lenis distinction. Latin also had a sizeable geminate series.
Unvoiced/fortis: /p t k kʷ/ <p t c qu>
Voiced/lenis: /b d g gʷ/ <b d g gu>
Geminate/fortis: /pp tt kk kkʷ/ <pp tt cc cqu>
In Western Romance, the phonology changes thus:
/p t k kʷ/ intervocalically > /b d g g(w)/ (>> /β ð ɣ ɣ(w)/) (initial unvoiced stops lose aspiration)
/b d g gʷ/ intervocalically > /β ð ɣ ɣ(w)/ (> ∅)
/pp tt kk/ > /p t k/ (including originally aspirated allophones of unvoiced stops!)
(Italo Romance and Eastern Romance don't change markably as a whole in this fashion and are rather either more conservative or palatal-leniting.)
----
The specifics of ancient Celtic phonology we don't know much about, but we do know what consonants were distinguished (even if not how exactly) and what were the later phonological evolutions to Brythonic and Goidelic. Proto-Celtic on the surface has a very similar system to Latin phonology:
Unvoiced: /t k kʷ/
Voiced: /b d g gʷ/
Geminate: /tt kk kkʷ/
In Brythonic Celtic, the phonology changes thus: (including the assumed changes: kʷ > p and gʷ > w)
/p t k/ intervocalically > /b d g/
/b d g/ intervocalically > /β ð ɣ/ (~> ∅)
/pp tt kk/ > /p t k/ > /ɸ θ x/ (merging with singleton reflexes from LT/XT/sT>hT clusters)
In Goidelic Celtic on the other hand:
/t k kʷ/ intervocalically > /θ x xʷ/ (except in LT/XT clusters)
/b d g gʷ/ intervocalically > /β ð ɣ w/ (~> ∅)
/tt kk kkʷ/ > /t k kʷ/
/bb dd gg ggʷ/ > /b d g gʷ/ (new geminates from syncope)
----
The obvious commonality across these languages firstly is the
spirantisation of voiced stops.
Gemination also generally tends towards simplification here (although Brythonic undergoes a second spirantising lenition, masking this).
The inconsistency lies with singleton voiceless stops. If on the one hand we compare Greek and Romance, the Romance aspirate stop reflexes resemble the stability of the Greek voiceless stops as well as the geminates in general, which re-affirms the case for these particular stops being considered fortis (if not equivalent in how they are fortis). Brythonic and Goidelic voiceless stops also resist lenition in similar if not identical word positions to Romance, which implies these particular stops are also fortis (at least at first, in the case of Brythonic).
Where not aspirated, the Romance voiceless stops all undergo
voicing lenition. This is in common with intervocalic Brythonic voiceless stops. On the other hand, the intervocalic Goidelic voiceless stops (and remaining/new Brythonic voiceless stops) undergo
spirantisation, which implies a difference in phonation (while still being lenis, or prone to weakening).
Moreover, this is all in stark contrast to what we see in Gaulish transcription; voiced stops are more likely to be written with unvoiced graphemes than the reverse, which upon first glance seems to imply a reverse movement if anything. We haven't considered Proto-Germanic here, which famously exhibits Grimm's law and Verner's law. Could Germanic influence be at play here? But at this point, I'm speculating.
Is there any way we can determine the exact phonations here in comparison with how other languages undergo lenition?
How many different ways can fortis-lenis, before lenition, be expressed if voicing is still the primary phonemic distinction? Evidently in Brythonic, there were
two sequential lenis phonations different from each other.
