1. Orthography
The Gaulish language has historically been recorded in a number of different scripts, Lepontic, Greek and Latin, with the Lepontic representing the earliest attested inscriptions, the Greek being used in the middle period of attestation, and the Latin being used for the latest records. These latest records are however also the most prolific ones, and the ones providing the most information about the language and the various changes at work in it. Furthermore, all of the important long running texts in the Gaulish language known today are written in the Latin script. Therefore, it seems reasonable to adopt the Latin script used in those texts, with a few adjustments to the orthography to reflect the demands of the modern language.
By way of graphic example, a line from the important Chateaubleau text (one of the longest, continuous running, well preserved inscriptions, discovered in 1997) looks like this in the original Gaulish:
neí anmanbe gniíou ape niteme ueííe
[“not by the names which are knowing about ownership may you desire it”]
The modern Gaulish language is written entirely phonetically, and to represent the sounds of the language it uses a version of the Latin alphabet which contains the following 20 basic characters:
a, b, c, d, e, f, g, h, i, l, m, n, o, p, r, s, t, u, v, w.
These are augmented by the following consonants:
ch, dh, gh, lh, nh, rh, th, gw, chw, sh
The vowel inventory is also extended to include the following long vowels:
á, é, í, ó, ú
and the following diphthongs:
au, ái, éi, ói, úi
The various sounds represented by the above are discussed below.
2. Phonology
The characters discussed above are used to represent the sounds of the language phonetically, in such a way that a given letter almost always represents a given sound. There are a few exceptions, which will be discussed further below.
Vowels
The vowels of the modern Gaulish language can be either long or short. These vowels are clear, as is the case in e.g. Welsh and Spanish.
The short vowels are: a e o u i
a = [a] as in at
e = [e] as in hell
o = [o] as in fog
u = [u] as in book
i = [i] as in it
The long vowels are: á é ó ú í.
á = [a:] as in father
é = [e:] as in lay, but without the yod sound
ó = [o:] as in tote
ú = [u:] as in shoe
í = [i:] as in see
The diphthongs are: au, ái, éi, ói, úi
áu = [au] as in how
ái = [a:j] as in goodbye
éi = [e:j] as in day
ói = [o:j] as in toy
úi = [u:j] as in French brouillard
All other vowels can occur adjacent to other vowels without combining their sounds into diphthongs, so that each vowel is pronounced independently with its own syllabic peak:
gwerúach = [gweru:ax] (feral)
bóé = [bo:e:] (cows)
Consonants
The basic consonants of the Gaulish language are:
p, t, c, b, d, g, f, v, n, r, l, m, s
Of these:
c is always [k]
g is always [g]
r is the rolling r of Scottish “borrow”.
s is [s] in word initial and word medial position, and [z] in word final position
As mentioned above, in addition to these regular consonants Gaulish also has:
ch, dh, gh, lh, nh, rh, sh, th, gw, chw, sh
ch = [x] as in Scottish loch
dh = [δ] as in English there
gh = [γ] as in Greek εγω
lh = [xl] an /l/ preceded by [x]
nh = [xn] as for l
rh = [xr] as for l
sh = [∫] as in English shine
th = [θ] as in English thin
gw = [gw]
chw= [xw]
Lastly, Gaulish has the semi-vowels w, i
w = [w] as in English will
i = [j] as in English you
Of these, w never occurs freestanding: it is always adjacent to other consonants, and never occurs intervocalically.
Conversely, the character i indicates the regular vowel [i] when directly following a consonant, or when preceding a consonant by itself:
mi = [mi], “me”
imi = [imi], “of-me”
However, i indicates the semi-vowel [j] when following another vowel, when preceding another vowel in word initial position, or when between two vowels:
ái = [a:j], “to go”
iár = [ja:r], “chicken”
áiedh = [a:jeδ], “face”
Many of the various sounds discussed above are the result of a process of sound changes, the beginnings of which can be discerned in the attested Gaulish material. These sound changes are discussed in detail below.
3. Sound Changes
Historical changes
The process of the changing of the sounds of the Gaulish languages is one of the two main processes which transform it from an ancient language to a modern language, the other one being apocope, or the loss of final syllables, of which more later. The changing of the sounds of a language is best thought of as a process of phonetic erosion: over a certain period of time, by virtue of living in the mouths of people, the sounds of languages change. In a Western Indo European context, sound changes have been historically observed, recorded and extrapolated in the Romance, Germanic and Celtic language families.
Very broadly speaking, in the Celtic language families there are two sets of sound changes that have affected the two separate surviving Celtic language families: the Brittonic (Welsh, Cornish, Breton) and the Goidelic (Irish, Scottish Gaelic, Manx). In the Brittonic language family, an important change that occurred was the voicing of intervocalic voiceless stops (* -ata- > * -ada-), and the spirantisation of voiced intervocalic stops (*-aba- > * -ava-). In a nutshell, pre-sound change intervocalic [p, t, c] became post-sound change intervocalic [b, d, g]; at the same time, pre-sound change intervocalic [b, d, g] became [v, δ, γ], the latter of which eventually disappeared; intervocalic [m] also became [v]. A number of changes also affected the Brittonic vowel system, but these are not relevant here.
In the Goidelic branch of the Celtic language family, intervocalic voiceless stops were spirantised: [t, c] became [θ, x] (Goidelic did not have the sound [p]). At the same time, intervocalic voiced stops became spirantised as well: [b, d, g] became [v, δ, γ], as in Brittonic. Goidelic also experienced a shift of emphasis towards the first syllable of a word, which, in words of more than two syllables, typically resulted in the loss of the second syllable, a process referred to as syncope.
The two sound changes outlined above are clearly illustrated in the respective treatment by the two language families of the word “Caratacos”, “beloved-like/ loveable”, which is attested in Classical Gaulish as well as in Welsh and Irish.
Classical “Caratacos” > Welsh “Caradog” (voicing of voiceless stops)
> Irish “Carthach” (spirantisation of voiceless stops)
For a detailed discussion of these sound changes, including this example, see Lambert 2003 (p. 28, 47).
Changes in Gaulish
The sound changes described above occurred in the respective language families around the 5th-6th centuries CE, again broadly speaking, in a period of considerable political and social upheaval in Western Europe, which appears to have triggered a period of linguistic innovation and change.
It can be reasonably expected that the Gaulish language, at that time, was subject to the same pressures that caused sound changes in its neighbouring and related languages, and therefore also experienced a process of sound change. In order to construct Gaulish as a modern language, it is important and essential to discern and to apply this sound change. Therefore, the available Gaulish material has been investigated for clues as to what this change might have been like. Below are presented epigraphic instances of genuine, attested Gaulish material that testify to changes at work in the phonology of the language. Most of these, if not all, can be found in Delamarre 2003, as per individual reference, unless otherwise specified; when available, their primary source is included and follows the page number in Delamarre 2003.
aballos > avallo; i.e. [b] > [v] (p. 29; Endlicher’s Glossary)
anman > anuan; i.e. [-nm-] > [-nw-] (p. 50; Larzac, Chateaubleau)
iouincos > ioinchus; i.e. [c] > [x] (p. 191, anthroponomy)
iaccos > iacchus; i.e. [c] > [x] (p. 185, anthroponomy)
riocatus > riochatus; i.e. [c] > [x] (p. 258, anthroponomy)
arganto- > arxanti; i.e. [g] > [γ] (Lambert 2003, p. 48; Suessons coin)
> arganthoneia; i.e. [t] > [θ] (p. 53; Galatian source)
(g)nata > gnatha, i.e. [t] > [θ] (p. 181; spindle whorl of Saint-Révérien)
litan- > lithan-, i.e. [t] > [θ] (p. 204; anthroponomy)
ate- > atha-, i.e. [t] > [θ] (p. 57, p. 214; anthroponomy)
couerto- > couerthi, i.e. [t] > [θ] (p. 317, anthroponomy)
rectu- > ogmireectherio, i.e. [t] > [θ] (p. 239, 7th century map)
*dacti- . uirodacthi, i.e. [t] > [θ] (p. 321)
uiriatus > uiriathos, i.e. [t] > [θ] (p. 321, Ellis 1998 p. 210)
catu- > cathubodua, i.e [t] > [θ] (p. 81, 111)
rectu- > rextu- > reitu, i.e. [c] > [x] > [i] before –t (p. 239)
*sisagsiou > siaxsiou, i.e. [g] > [γ] (p. 273; Chateaubleau)
*agat > axat, i.e. [g] > [γ] (p. 63; Marcellus of Bordeaux)
luge > luxe, i.e. [g] > [γ] (p. 210; Chamalieres)
ambio > ape, i.e. [amb] > [ãb] (Bernard Mees 2010, Bhrghros 2013, pers. com.;
Chateaubleau and Rom)
briuo > brio, i.e. [-w-] > [-ø-] (p. 89; Endlicher’s Glossary)
iouinc- > ioinc-, i.e. [-w-] > [-ø-] (p. 191, anthroponomy)
magiorix > maiorix, i.e. [-g-] > [-ø-] (Lambert 2003, p. 46; anthroponomy)
catugenus > catuenus, i.e. [-g-] > [-ø-] (Lambert 2003, p. 46; anthroponomy)
traget- > treide, i.e. [-g-] > [-i-] (p. 300; Endlicher’s Glossary)
[-t-] > [-d-], (p. 300; Endlicher’s Glossary)
*brogilos > breialo, i.e. [-g-] > [-i-] (p. 91; Endlicher’s Glossary)
badio > * baδio > baio, i.e. [-d-] > [δ] > [-ø-] (p. 63, anthroponomy)
sindi > sini, i.e. [nd] > [n] (Mees 2010, Chateaubleau)
Stirona > Ðirona > Sirona, i.e. [st-] > [ts-] > [s-] (p. 282, theonomy)
The above instances can be summarised as follows:
b > u (v)
c > ch (x)
d > δ
g > x (γ)
g > i
g > ø
m > u (w)
t > th (θ)
t > d
amb > ãb, assimilation of nasal to vowel
st > ts > s
In the above, the grapheme “x” is considered to represent the spirantisation of intervocalic “g” > [γ] (Delamarre 2003, p. 63, p. 210; Lambert 2003, p. 46, p. 48)
An analysis of the above data indicates that the intervocalic consonants of Gaulish were subject to a process of spirantisation or fricativisation, which can be represented as follows:
p > p
t > th [θ]
c > ch [x]
b > u [v]
* d > δ ?
g > x [γ], i [j] or ø
m > u [w]
đ > ss [s]
[mb > nasal-b]
The only exception to this appears to be the rendition of –t- as –d- in “treide”, which appears to indicate a voicing of an intervocalic consonant. Lambert (2003, p. 207) considers “treide” as having been treated as a word of Vulgar Latin and as such as having undergone Romance lenition, where intervocalic voiceless stops become voiced (Western Romance). In support, it is the opinion currently held by scholars in the field of Continental Celtic linguistics that the possibility that “treide” can be considered as indicative and representative of a genuine Gaulish sound change is negligeable (Bhrghros, Mees, Stifter, Gwinn pers. com. 2009).
The remainder of the data appears to uniformly point towards a process of spirantisation of intervocalic consonants. The existence of lenition, the process of weakening of pronunciation of consonants, in this case through spirantisation or fricativisation, in Gaulish has been proposed and defended by Gray (1944), accepted by Fleuriot (in Delamarre 2003, p. 63), rejected and objected against by Lambert (2003) among many others (e.g Watkins 1955), and cautiously and somewhat sceptically regarded as a possibility by Delamarre (2003). Nevertheless, as Eska (2008) emphatically suggests, proposes and defends, the data appear to point in this direction.
In addition to the above, there is widespread academic agreement on the disappearance of intervocalic –u- and –g- (Delamarre 2003; Lambert 2003). In regards to –g-, a position is here adopted where –g- becomes spirantised when occurring in consonant clusters, and becomes –i- in intervocalic position and word finally when preceded by a vowel:
-Cg- > -Cγ- (e.g. argant- > arxant-)
-gC- > -γC-
-VgV- > -ViV- (e.g. brogil- > breial-)
-Vg > -Vi
Diphthongs
The sound change discussed above gives rise to a number of modern diphthongs which are not derived from inherited Indo European or Proto Celtic diphthongs:
Classical Gaulish > Modern Gaulish
dago > dái (good)
aneg- > anéi (protect)
brogi > brói (country)
lugus > lúi (a god)
This therefore produces the diphthongs:
ái, éi, ói, úi
in addition to the vowel inventory discussed previously.
The Classical Gaulish diphthong –ou-, which itself was the result of an older –eu-, is also attested as undergoing an evolution:
touta > tota, i.e. [ou] > [o:] (people, tribe, nation)
crouco > crocu, i.e. [ou] > [o:] (hillock, hill, heap, pile)
This indicates that:
ou > o [o:]
Consonant clusters
Several examples of the data analysed above provide information on what happens to stops as part of consonant clusters:
argant- > arganthoneia: -nt- > -nth-
argant- > arxant-: -rg- > -rx-
iouinc- > ioinch-: -nc- > -nch-
This indicates clearly that the process of spirantisation was productive across consonant clusters.
Tau Gallicum
The Tau Gallicum in Gaulish, represented by a wide variety of graphemes including đ, đđ, θ, θθ, ds, dd, tth, th, ts, ss and ss with a horizontal bar through it, is a matter of some controversy: some disagreement exists about its actual phonetic value. However, most scholars in the field of Continental Celtic linguistics agree that the Tau Gallicum represented a dental affricate [ts] (Delamarre 2003; Lambert 2003; Mees 2008, 2010; Eska 2008). Regardless of its original phonetic value, there is universal agreement that over the course of its attestation the sound of the Tau Gallicum evolved to become [s], which may become apparent from comparison with modern related languages, as well as with late Gaulish material.
e.g.: neđđamon > nesaf (“next” - Welsh)
miđ > mis (“month” – Welsh)
Meθθilus > Medsilus > Messilus (Lambert 2003, p. 46)
Caraθθounios > Carassounios (Lambert 2003, p. 46)
adsedo- > ađđedo- > assedo- (“reside, establish”, Delamarre 2003, p. 33)
stir- > đir- > sir- (“star”, Delamarre 2003, p. 282)
Therefore, the change postulated for the Tau Galicum is:
st, ts, đ > s
The group –ct-
There is no direct evidence of the evolution of –ct- in Gaulish. However, the similarity of its evolution in French (due to the Gaulish substrate) and in Brittonic is considered strong enough by Lambert (2003, p. 49) to conclude that its treatment was:
-ct- > -xt- > -it-, with i = [j], second element of diphthong.
e.g.: lact- > lait (“milk”, French)
> llaeth (“milk”, Welsh)
where -act- > -ait- [a:jt]. This then undergoes spirantisation of the word final stop and becomes –aith [a:jθ].
The group -sc
There is in the attested material no indication of an evolution of the group –sc in Late Gaulish. Therefore, no soundchange is postulated for this group, which is maintained as it is.
e.g. mesco- > mesc (“drunk”)
Word initial u- before vowel
There is no direct indication of the evolution of word initial u- before vowel, thought to have been [w] in Classical Gaulish (Delamarre 2003, p. 309; Lambert 2003, p. 46; Mees, Stifter pers. comm. 2009). In the surviving related languages, u- before vowel becomes f- in Goidelic, and gw- in Brittonic.
A comparison with the treatment of word initial u- before vowel in French, however, gives us some indication of what may be acceptably proposed as a plausible evolution of this phoneme. The French language has known a change of word initial [w] to [gw], often in words of Germanic origin, but not restricted to them; to wit “guerre” (war), Guillaume (William), g(u)arde (ward), gant (want – i.e. mitten), and Gascogne (Wasconia). The French language came to replace the Gaulish language in much of its domain and in the process demonstrably absorbed a certain amount of Gaulish substrate influence (Lambert 2003). Since the evolution of Brythonic [w] to [gw] appears to be partially paralleled in French, it is deemed acceptable to postulate an evolution of word initial [w] to [gw] for Modern Gaulish.Support for this notion can be found in the placename Gap, attested in the toponymic record as derived from an older Uapincum (Delamarre 2003, p. 438)
While this may appear controversial because the above attestation is found in toponymy and not in a strictly Gaulish context, it is worth bearing in mind that the phenomenon of changing word initial [w] to [gw] through fortition or strengthening is “fairly common cross linguistically” (B. Mees pers. comm. 2009). While it happened in Brythonic across the three attested languages Welsh, Cornish and Breton, it didn’t occur in Breton until the eleventh century (UT n.d.), a date well past the linguistic and cultural unity of Breton with the other Brythonic languages. This appears to indicate that the fortition of word initial [w] to [gw] happened due to internal processes, as opposed to shared innovations across the Brythonic family that could possibly be postulated.
Furthermore, it is reasonable to posit that the loss of word final consonants (loss of –s, -n, -m attested) would have resulted in a situation where word initial u- ended up in an intervocalic environment, which would have led to its disappearance. It can therefore be postulated that a strengthened gw- developed to prevent the loss of word initial u-.
Therefore, in keeping with the arguments outlined above, a sound change of word initial u- before vowel to gw- is proposed for Modern Gaulish.
The treatment of “s”
Although there has been controversy around this, it is now widely agreed that “s” is maintained in Gaulish (D. Stifter, B. Mees, C. Gwinn, Bhrghros pers. com. 2009). There is no indication of it becoming [h] word initially, as in Brittonic. While there have been instances where “s” is thought to have disappeared intervocalically, it has been argued strongly and persuasively that these were exceptions (Stifter 2009). Therefore, the position is taken here that “s” is maintained unchanged.
However, there is evidence that adjacent to –l- “s” became absorbed into –ll- (coslo- > collo-, Delamarre 2003, p. 127). Therefore, in modern Gaulish, “s” does not occur in clusters with –l-. It is proposed here that in its word initial and word medial position “s” is [s], while word finally it is [z].
Finally, in a mutation context, when “s-“ is followed by a consonant, “s-“ is dropped, and the following consonant remains unmutated. This elision of “s-“ is marked with an apostrophe: e.g. scrívi > ‘crívi (to write).
The treatment of –nt-
In word medial position, it is clear that “–nt-“ becomes “–nth-“, as is indicated by arganto- > arganthoneia (Delamarre 2003, p. 53. For its word final treatment, it is postulated here that an intermediate “–nth” evolved to “-n” through phonetic erosion and for reasons of ease of pronunciation.
The treatment of –nd-
The treatment of –nd- is not clearly attested in the available material. However, three instances are known that appear to bear witness to some evolution of –nd-, from which tentative clues may be extrapolated.
1. the word “andecinga” (agricultural measurement, similar to an acre) was absorbed into medieval Latin and hence passed into Old French as “ansenge, ensenge”. As such, it appears to show an evolution of ande > an, that is an elision of –d-. However, this is a highly dubious example, as it testifies to an evolution in Vulgar Latin / Old French, and not in Gaulish.
2. the word “andounna” (“low water, spring, source”, Collias inscription, Delamarre 2003, p. 48) has been persuasively analysed by Lambert (1990, followed by Eska, 1992, in Delamarre 2003, p. 48) as “ande-udna > ande-unna > andeunna > andounna. In this proposed evolution, the word “udna” (“water”, “wave”, “tide”) has evolved to “unna” by assimilation of nasality, of [d] assimilating to [n].
3. the word “sindi” (this, that, demonstrative) attested at Larzac, Coligny and Chateaubleau also appears as “sini” at Chateaubleau. This appears to indicate an evolution of [nd] > [n]. While it is also attested as “sendi” at Chateaubleau, its occurrence fits in with the spelling variations internal to the Chateaubleau document, and indicates that, at the very least, there was a certain amount of confusion and/or variation in the way *sindi was pronounced. That “sini” is the same word as “sendi” is rendered probable by the fact that they both occur twice, both in a similar context, and both in conjunction with a form of the verb *ieg-, albeit two different forms, which may or may not have something to do with the structural difference between the two words (iegu-mi-sini <> iexstu-mi-sendi).
A position is therefore postulated by modern Gaulish that all instances of –nd-, word medial and word final, assimilate to –n(-).
.....The treatment of geminate consonants
No pertinent information is available about the evolution of geminate consonants. A position is adopted here that all geminate consonants are treated as single consonants, and as such are subject to spirantisation in word medial and word final position, if they are stops.
The treatment of semi-vowel -i-
In the attested material, a Latin “i longa” (“long i”) is often used to indicate a semi-vowel [j]. It is proposed here that the semi-vowel [j] is represented by the grapheme “i”, in keeping with its occurrence in e.g. “breialo” (Delamarre 2003, p. 91).
Intervocalic –g- will almost always become –i-, pronounced [j]
e.g.: trageto > tráieth [tra:jeθ] (foot)
Words starting with g- will, in combination with prefixes, weaken this –g- to a –i-:
e.g.: gar: to call
sú + gar: good + call = to laugh > *súgar > súiar [su:jar]
Word initial i- before a vowel will be pronounced as [j]:
e.g.: iár: chicken [ja:r]
iach: sane, healthy, well [jax]
ianthu: wish [janθu]
However, word medial –i- preceded by a consonant and followed by a vowel, is pronounced [i]:
e.g.: adhianthu: ambition [aδianθu]
The treatment of semi-vowel –w-
It is proposed here that in words where “u” is followed by another vowel and receives emphasis, this “u” is interpreted as the long vowel [u:]:
e.g. arúer: to please
However, in instances where the emphasis shifts away from “u”, it becomes the semi-vowel “w”:
e.g.: arwéra mi: I please
The treatment of word final open vowel -o
In the Late Gaulish material available, it is evident that word final open vowel –o [o:] evolves to –u [u:] (Delamarre 2003, p. 342; Lambert 2003, p. 62). Although apocope removes most Classical and Late word endings, there are instances where word final open vowel –o is retained, most usually for reasons of necessity of distinction between near homonyms. In these cases, this end open vowel changes to –u, and is usually short [u].
e.g.: panto > panthu (suffering)
The treatment of –p-
In the data at our disposal there is no indication that /p/ was subject to any erosive phonetic processes. Inherited Indo-European /p/ had been lost in Celtic at a very early stage, and is indeed diagnostic in the determination of the Celticity of languages in antiquity. At a later stage the labiovelar [kʷ] mutated permanently to /p/ in all positions, an innovation shared between Gaulish and Brittonic. The position is adopted here that this newly-acquired /p/ remained firm enough to resist spirantisation in word medial and word final environments. In word initial position /p/ will mutate regularly in analogy with other consonants.
The treatment of r-
The rendition of the sound /r/ shows signs of ambiguity in some contexts, specifically in word initial position, as evidenced in the testimony of some classical authors, though not, it needs to be noted, in Gaulish language texts. As such /r/ is found written /rh/ in a very few instances:
Rhenus, the name of the river Rhine, recorded as such by Caesar (Delamarre 2003, p. 257)
Rhenum, the name of a river in Gallia Cisalpina (Delamarre 2003, p. 257)
Rhodanus, the name of the river Rhône (Ellis 1998, p. 9)
rhaeda/rheda, the name for a Gaulish four-wheeled wagon, recorded by Marcus Terentius Varro
Rhetogenes, the name of a Celtiberian chieftain, recorded by Appian of Alexandria (Ellis 1998, p. 211)
It is thought that, while the grapheme /rh/ might represent a Roman hyper correction used in words of non-Latin origin, it could also represent a sound akin to a voiceless /r/, for which no equivalent in Latin existed. Modern Gaulish adopts the latter view, and incorporates this unusual phoneme in its system of consonant mutations, where it only ever occurs in word initial position, after mutation. It is interpreted as [xr] and, in keeping with its attestation in the classical data, it will be represented as /rh/.
Apocope
Apocope is the complete loss of word final syllables or word endings, including endings defining case. It is the second major change that has influenced the evolution of both Brittonic and Goidelic and contributed to their becoming uninflected modern languages. In the case of Goidelic, cases were redefined as revolving around the pronunciation of word-final consonants, and as such are retained to this day – nominative, genitive and dative, as well as a vocative which does not however involve the pronunciation of the final consonant of a word. However, Goidelic also did lose all case endings, as did Brittonic. In the case of Brittonic, the loss of case endings resulted in the complete loss of all cases, in a parallel with Romance and most of Germanic (excepting German, Icelandic and Faroese).
The available Late Gaulish material shows clearly that a process of erosion of word endings was underway by the 4th century CE (see e.g Stifter 2009, 2012). This is mentioned by Delamarre, who indicates the erosion of the case endings (p. 342-43):
nominative sing. –os > -o (loss of –s)
accusative sing. –on > -o (loss of –n)
-an > -in/ -im > -i (loss of –n)
Examples from the attested material include:
coro bouido < * coros bouidos (Mees 2010, Chateaubleau)
andoedo < * andoedon (Delamarre 2003, La Graufesenque)
beni < * benin/ *benim (Mees 2010, Delamarre 2003, Chateaubleau)
These are indications that the case system was being eroded. Inevitably, such erosion leads to loss of meaning and the collapse of the case system. It is reasonable to postulate that had Gaulish continued to evolve as a living language, the case system would have disappeared entirely through the phenomenon of apocope, the first stages of which can be observed in the examples cited above. As a result, the Gaulish language would have, quite likely, ceased to be an inflected language with cases, and would have, quite likely, evolved as an analytic language with prepositions. This is exactly what happened to the Brittonic languages.
Graphic representation of sound changes
The sound changes discussed above need to be able to be represented graphically in the written language. To recap, the graphemes used in the attested material, and their probable pronunciation values, are as follows:
p > p
t > th [θ]
c > ch [x]
b > u [v]
* d > δ ?
g > x [γ], i [j] or ø
m > u [w]
mb >p = nasal vowel +b
đ > ss [s]
In Latin, the orthographic conventions of which were applied to the Gaulish texts in which these graphemes are found, the grapheme “u” was used to indicate both [u], [v] and [w]. It is commonly held that “u” for lenited “b” indicates [v] (Delamarre 2003, p. 29), and that “u” for lenited “m” indicates [w] (Lambert 2003, p. 46). It is therefore possible to posit:
b > v [v]
m > w [w]
Similarly, the digraphs “th” and “ch” used in the Gaulish texts are drawn from the Latin orthographic inventory. They were incorporated into Latin originally to represent sounds of Greek words which were unknown to Latin; as such “th” and “ch” represented aspirated “t” (theta, θ) and “c” (chi, χ). While in early antiquity it is thought that these sounds represented aspirated stops, i.e. a [t] and [c] followed by a puff of air (“t-huh”, “c-huh”), by Late Antiquity the respective pronunciation of these digraphs had evolved to the fricatives [θ] and [x]. Therefore, it is reasonable to assume that the digraphs “th” and “ch” as used in the Gaulish texts are likely intended to represent a fricativised or spirantised sound. It is postulated here that the sounds represented by the digraphs were the fricatives mentioned above. This therefore gives:
c > ch [x]
t > th [θ]
The changes represented by p > p, and đ > ss do not represent a problem: they can be taken at face value and be represented by “p”, and “s” respectively:
p > p [p]
đ > s [s]
The change represented by mb > p is interpreted as mb > nas-b, with the grapheme /p/ being used to indicate [b] at a period where /b/ was used to indicate [v] (Bhrghros 2013). It is postulated here that this led to mb > nas-b > nas > /m/, i.e. mb > m, in analogy with the treatment of –nd-.
This leaves the challenge of how to represent the lenited sounds of “g” and “d”. In the case of “g”, it is agreed (Delamarre 2003, p. 63, p. 210; Lambert 2003, p. 46, p. 48) that the lenited form of “g”, written as “x”, is intended to represent the sound [γ]. In analogy with the lenited forms of “c” and “t”, i.e. “ch” and “th”, used in the attested Gaulish material, it is proposed here that the sound [γ] of lenited “g” be written as “gh”, indicating a spirantised pronunciation of “g”. This is further supported by the use of the digraph “gh” to represent an originally similar sound in Irish, Scottish Gaelic, Cornish and English. The representation of intervocalic “g” as “i”, as found in the original material, does not present a problem and can be maintained as such.
g > gh [γ]
g > i [j]
Finally, the spirantisation of “d” is not attested anywhere, and only hypothesised by Delamarre (2003, p. 63). For the sake of analogy and conformity, a position is here adopted that “d” underwent spirantisation to [δ] as well in intervocalic, word final and consonant cluster position, like all other stops, and that this will be represented by the digraph “dh”. This latter is used in Cornish to represent the same sound, and as such is deemed acceptable in the framework of Celtic phonological graphic representation. This therefore gives:
d > dh [δ]
The graphic representation of post-sound-change consonants can now be summarised as follows:
p > p [p]
t > th [θ]
c > ch [x]
b > v [v]
d > dh [δ]
g > gh [γ], i [j]
m > w [w]
mb > m [m]
Emphasis
The emphasis in Gaulish is thought to have been predominantly on the penultimate syllable (see Lambert 2003 for a discussion; also Eska 2008). This is therefore the default emphasis position adopted by modern Gaulish:
caran = friend
If words receive suffixes, the emphasis is shifted to accomodate for the extended word, so that the emphasis remains on the penultimate syllable:
caranach = friendly
Exceptions to the penultimate emphasis rule are fusions of prepositions with personal pronouns, where the emphasis falls on the last syllable in order to emphasise the person referred to:
canith = with you, emphasis on –ith
The fact of receiving emphasis can make a vowel change from short to long. This happens in two cases:
1. if the vowel in question is followed by a word final open vowel, regardless of syllable position:
e.g.: gwel: to want (verbal noun) > gwéla mi: I want
gar: to call (v.n.) > gára mi: I call
2. if the vowel in question is in the second or subsequent syllable, and is followed by either an open vowel or a spirantised stop followed by a vowel, opened or closed:
e.g.: lavar: to speak (v.n.)
> lavára mi: I speak – 2nd syllable, followed by open vowel
car: to love (v.n.)
> caráthu: loved – 2nd syllable, followed by spirantised stop and open vowel
> caráthach: loveable – 2nd syllable, followed by spirantised stop and closed vowel
> caran: friend – 1st syllable, not followed by spirantised stop or open vowel
> caranach: friendly – 2nd syllable, not followed by a spirantised stop or open vowel
However, there are words where vowel length may be etymologically defined. These will not be subject to these rules:
e.g.: gwir: man [gwir]
gwír: true [gwi:r]
In the above example meaning is solely conveyed by vowel length.
Applied example
The graphic and emphatic conventions outlined above can now be put into practice and be applied to the example used previously to illustrate the ways in which the two surviving branches of the Celtic languages were affected by their respective sound changes. Thus, Classical Caratacos can be put through the two changes outlined above as affecting Gaulish, i.e. first apocope, loss of the final syllable:
caratacos > caratac
followed by spirantisation of unvoiced intervocalic stops:
caratac > carathach
Lastly, the emphasis is placed on the penultimate syllable, which is followed by an aspirated stop and therefore becomes long, indicated by a diacritic:
carathach > caráthach
This can be compared to the permutations of the same word in Brittonic and Goidelic:
Gaulish: Caráthach (spirantisation of voiceless stops, apocope)
Welsh: “Caradog” (voicing of voiceless stops, apocope)
Irish: “Carthach” (spirantisation of voiceless stops, apocope and syncope)
Mutated consonants in word initial position
An important part of the way the modern Gaulish language works is the mutation of word initial consonants, a phenomenon shared among all the Celtic languages. While this will be discussed in great detail in a separate section, it is important, from a phonological point of view and for the sake of clarity, to list here the various sounds that various consonants mutate into, as some of these do not occur elsewhere.
Very briefly, the system of changes runs as follows:
p t c > b d g > v dh gh
m > w
f > fh [φ] (bilabial)
In addition, the following are included:
l > lh [xl]
n > nh [xn]
r > rh [xr]
s > sh [ʃ ]
gw > chw [xw]
These are essentially spirantised sounds which, with the exception of “chw”, only occur word initially, and only in mutation conditions.
Schwa
It is proposed here that the modern Gaulish language features the sound schwa in certain cases, where a word ends on a consonant followed by n, r or l. These are usually the result of apocope. In these cases, a schwa is pronounced between the two final consonants. However, as there is no graphic representation known for schwa in Gaulish, this schwa is never written, and is deduced from the word final consonant configuration:
e.g.: attested word “sedlo” (seat) > modern Gaulish “sedhl”
a schwa is pronounced between –dh- and –l: sedhl = [seδəl]
petru (four) > pethr = [peθər]
anectlo (protection) > anéithl = [ane:jθəl]
louernos (fox) > lóern = [lo:erən]
In conjunction with the pluralising suffix -é this schwa sound disappears:
anéithlé: protections [[ane:jθle:]
lóerné: foxes [lo:erne:]
As scha only ever occurs as the ultimate vowel, emphasis is always on the vowel before schwa.
Summary of sound changes
Having explored in detail the sound changes which can be discerned as having been at work in Late Gaulish, and those which may reasonably have been expected to occur had Gaulish continued to this day as a living language, it is now possible to summarise these in table form. This provides the necessary background information for the exposition of modern Gaulish grammar which makes up the remainder of this document.