French exhibits perhaps the most extensive phonetic changes (from Latin) of any of theRomance languages. Similar changes are seen in some of the northern Italian regional languages, such asLombard orLigurian. Most other Romance languages are significantly more conservative phonetically, withSpanish,Italian, and especiallySardinian showing the most conservatism, andPortuguese,Romanian,Catalan, andOccitan showing moderate conservatism.[1]
French also shows enormous phonetic changes between theOld French period and the modern language. Spelling, however, has barely changed, which accounts for the wide differences between current spelling and pronunciation. Some of the most profound changes have been:
The loss of almost all final consonants.
The occasional elision of final/ə/, which caused many newly-final consonants.
The loss of the formerly strong stress that had characterized the language throughout much of its history and triggered many of the phonetic changes.
Significant transformations in the pronunciation of vowels, especiallynasal vowels.
Only some of the changes are reflected in theorthography, which generally corresponds to the pronunciation of c. 1100–1200 CE (theOld French period) rather than modern pronunciation.
A profound change in very late spoken Latin (Vulgar Latin, the forerunner of all theRomance languages) was the restructuring of thevowel system ofClassical Latin. Latin had thirteen distinct vowels: ten pure vowels (long and short versions of⟨a, e, i, o, u⟩), and threediphthongs (⟨ae, oe, au⟩).[2] What happened to Vulgar Latin is set forth in the table.[3]
Essentially, the ten pure vowels were reduced at first to nine vowels/aɛeiɪɔouʊ/ inProto-Romance. Then Italo-Western Romance shifted/ɪʊ/ to/eo/, yielding seven vowels/aɛeiɔou/, andvowel length was no longer a distinguishing feature. The diphthongs⟨ae⟩ and⟨oe⟩ fell in with/ɛ/ and/e/, respectively.⟨au⟩ was retained, but various languages (including Old French) eventually turned it into/ɔ/ after the original/ɔ/ fell victim to further changes.[citation needed]
The Vulgar Latin pronunciation reconstructions reflect Italo-Western Romance verb forms.
French first-person plural forms (with the exception of past historic) are not necessarily inherited from Latin, instead they are probably borrowed fromFrankish suffix*-ōmês, instead of yielding*chantens.
The French second-person plural imperative is not inherited from the Latin form with the same function, instead it is supplied by either second-person plural indicative or subjunctive present; comparechante —chantez, butaie —ayez (subjunctive present ofavoir), notevouille –vouillez (alternative imperative forms ofvouloir); as they have⟨-ez⟩ instead of the expected*chanté (this form would be the past participle).
The complex but regular French sound changes have caused irregularities in the conjugation of Old French verbs, like stressed stems caused by historic diphthongization (amer,aim,aimes,aime,aiment, butamons,amez), or regular loss of certain phonemes (vivre,vif,vis,vit). Later in Modern French, these changes were limited to fewer irregular verbs. Modern French also had lost the class of rather unpredictable-ier verbs (resulting from ejection of/j/ into the infinitive suffix-āre, which still exists in somelangues d'oïl), having been replaced by simple-er verbs plus⟨-i⟩ instead, as inmanier, but Old Frenchlaissier →laisser.
Vowel length became automatically determined by syllable structure, with stressedopen syllables having long vowels and other syllables having short vowels. Furthermore, the stress on accented syllables became more pronounced in Vulgar Latin than in Classical Latin. That tended to cause unaccented syllables to become less distinct, while working further changes on the sounds of the accented syllables. That especially applied to the new long vowels, many of which broke into diphthongs but with different results in each daughter language.[citation needed]
Old French underwent more thorough alterations of its sound system than did the other Romance languages.Vowel breaking is observed to some extent in Spanish and Italian: Vulgar Latinfocu(s) "fire" (in Classical Latin, "hearth") becomes Italianfuoco and Spanishfuego. In Old French, it went even further than in any other Romance language; of the seven vowels inherited from Vulgar Latin, only/i/ remained unchanged in stressed open syllables:[citation needed]
The sound of Latin short e, turning to/ɛ/ in Proto-Romance, became⟨ie⟩ in Old French: Latinmel, "honey" > OFmiel
The sound of Latin short o > Proto-Romance/ɔ/ > OF⟨uo⟩, later⟨ue⟩: cor >cuor >cuer, "heart"
Latin long ē and short i > Proto-Romance/e/ > OFei:habēre >aveir, "to have"; this later becomes/oi/ in many words, as inavoir
Latin long ō and short u > Proto-Romance/o/ > OFou, latereu:flōrem >flour, "flower"
Latin a, ā > Proto-Romance/a/ > OF/e/, probably through an intervening stage of/æ/;mare >mer, "sea". That change also characterizes the Gallo-Italic languages of Northern Italy (cf. Bolognese[mɛːr]).
Furthermore, all instances of Latin long ū > Proto-Romance/u/ became/y/, the lip-rounded sound that is written⟨u⟩ in Modern French. That occurred in both stressed and unstressed syllables, regardless of whether open or closed.
Latin au did not share the fate of/ɔ/ or/o/; Latinaurum > OFor, "gold": not *œur nor *our. Latin au must have been retained at the time such changes were affecting Proto-Romance.
Changes affecting consonants were also quite pervasive in Old French. Old French shared with the rest of the Vulgar Latin world the loss of final⟨-M⟩. Old French also dropped many internal consonants when they followed the strongly stressed syllable; Latinpetram > Proto-Romance*/ˈpɛtra/ > OFpierre; cf. Spanishpiedra ("stone").
Table of Old French outcomes of Latin vowels
Letter
Classical Latin
Vulgar Latin
Proto- Western Romance
Early Old French (through early 12th c.)
Later Old French (from late 12th c.)
closed
open
closed
open
a
/a/
/a/
⟨a⟩/a/
⟨e, ie⟩/æ,ie/
⟨a⟩/a/
⟨e, ie⟩/ɛ,jɛ/
ā
/aː/
ae
/a͡e̞/
/ɛ/
⟨e⟩/ɛ/
⟨ie⟩/ie/
⟨e⟩/ɛ/
⟨ie⟩/jɛ/
e
/e̞/
oe
/o͡e̞/
/e/
/e/
⟨e⟩/e/
⟨ei⟩/ei/
⟨oi⟩/oi/>/wɛ/
ē
/e̞ː/
i
/i/
/ɪ/
y
/ʏ/
ī
/iː/
/i/
⟨i⟩/i/
ȳ
/ʏː/
au
/a͡u/
/aw/
⟨o⟩/ɔ/
o
/o̞/
/ɔ/
⟨o⟩/ɔ/
⟨uo⟩/uo/
⟨o⟩/ɔ/
⟨ue⟩/wɛ/>/ø/
ō
/o̞ː/
/o/
/o/
⟨o⟩/o/
⟨ou⟩/ou/
⟨o(u)⟩/u/
⟨eu⟩/eu/>/ø/
u
/u/
/ʊ/
ū
/uː/
/u/
⟨u⟩/y/
In some contexts,/oi/ became/e/, still written⟨oi⟩ in Modern French. During the early Old French period, it was pronounced as the writing suggests, as/oi/ as afalling diphthong:/oi̯/. It later shifted to becomerising,/o̯i/, before becoming/o̯e/. The sound developed variously in different varieties of Oïl: most of the surviving languages maintain a pronunciation as/we/, but Literary French adopted a dialectal pronunciation,/wa/. The doublet offrançais andFrançois in modern French orthography demonstrates the mix of dialectal features.[citation needed]
At some point during the Old French period, vowels with a following nasalconsonant began to be nasalized. While the process of losing the final nasal consonant took place after the Old French period, the nasal vowels that characterize Modern French appeared during the period in question.[citation needed]
The following table shows the most important modern outcomes of Vulgar Latin vowels, starting from the seven-vowel system ofProto-Western Romance stressed syllables:/a/,/ɛ/,/e/,/i/,/ɔ/,/o/,/u/. The vowels developed differently in different contexts, with the most important contexts being:
"Open" syllables (followed by at most one consonant), where most of the vowels were diphthongized or otherwise modified.
Syllables followed by apalatal consonant. An/i/ usually appeared before the palatal consonant, producing a diphthong, which subsequently evolved in complex ways. There were various palatal sources: Classical Latin/jj/ (e.g.peior[4] "worse"); any consonant followed by a/j/ coming from Latin short/e/ or/i/ inhiatus (e.g.balneum "bath",palātium "palace");/k/ or/ɡ/ followed by/e/ or/i/ (e.g.pācem "peace",cōgitō "I think");/k/ or/ɡ/ followed by/a/ and preceded by/a/,/e/ or/i/ (e.g.plāga "wound");/k/ or/ɡ/ after a vowel in various sequences, such as/kl/,/kr/,/ks/,/kt/,/ɡl/,/ɡn/,/ɡr/ (e.g.noctem "night",veclum <vetulum "old",nigrum "black").
Syllables preceded by a palatal consonant. An/i/ appeared after the palatal consonant, producing a rising diphthong. The palatal consonant could arise in any of the ways just described. In addition, it could stem from an earlier/j/ brought into contact with a following consonant by loss of the intervening vowel: e.g.medietātem > Proto-Romance/mejjeˈtate/ > Gallo-Romance/mejˈtat/ (loss of unstressed vowels) > Proto-French/meiˈtʲat/ (palatalization) > Old French/moiˈtjɛ/ >moitié/mwaˈtje/ "half".
Nasal syllables (followed by an/n/ or/m/), where nasal vowels arose. Nasal syllables inhibited many of the changes that otherwise happened in open syllables; instead, vowels tended to be raised. Subsequently, the following/n/ or/m/ was deleted unless a vowel followed, and the nasal vowels were lowered; but when the/n/ or/m/ remained, the nasal quality was lost, with no lowering of the vowel. This produced significant alternations, such as masculinefin/fɛ̃/ vs. femininefine/fin/.
Syllables closed by/s/ followed by another consonant. ByOld French times, this/s/ was "debuccalized" into/h/, which was subsequently lost, with a phonemic long vowel taking its place. These long vowels remained for centuries, and continued to be indicated by an⟨s⟩, and later acircumflex, with alternations such asbette/bɛt/ "chard" vs.bête (formerly/bɛːt/) "beast" (borrowed frombēstiam). Sometimes the length difference was accompanied by a difference in vowel quality, e.g.mal/mal/ "bad" vs.mâle/mɑːl/ "male" (Latinmāsculum <*/ˈmaslə/). Phonemic length disappeared from Parisian French by the 18th century, but survived regionally (now especially inBelgian French).
Syllables closed by/l/ followed by another consonant (although the sequence-lla- was not affected). The/l/ vocalized to/w/, producing a diphthong, which then developed in various ways.
Syllables where two or more of the above conditions occurred simultaneously, which generally evolved in complex ways. Common examples are syllables followed by both a nasal and a palatal element (e.g. from Latin -neu-, -nea-, -nct-); open syllables preceded by a palatal (e.g.cēram "wax"); syllables both preceded and followed by a palatal (e.g.iacet "it lies"); syllables preceded by a palatal and followed by a nasal (e.g.canem "dog").
The developments in unstressed syllables were both simpler and less predictable. In Proto-Western Romance, there were only five vowels in unstressed syllables:/a/,/e/,/i/,/o/,/u/, as low-mid vowels/ɛ/,/ɔ/ were raised to/e/,/o/. These syllables were not subject to diphthongization and many of the other complex changes that affected stressed syllables. This produced many lexical and grammatical alternations between stressed and unstressed syllables. However, there was a strong tendency (especially beginning in the Middle French period, when the formerly strong stress accent was drastically weakened) to even out these alternations. In certain cases in verbal paradigms an unstressed variant was imported into stressed syllables, but mostly it was the other way around, with the result that in Modern French all of the numerous vowels can appear in unstressed syllables.
Table of modern outcomes of Vulgar Latin vowel combinations
medietātem > Vulgar Latin/mejeˈtate/ >/mejˈtʲate/ > Early Old French/meiˈtiɛθ/3 > Late Old French/moiˈtjɛ/ >moitié/mwaˈtje/ "half";cārum > Old Frenchchier/tʃjɛr/ >cher/ʃɛʁ/ "dear"
^1 "Context" refers to the syllable context at the Vulgar Latin or Gallo-Romance stage. The contexts are as follows:
An "open" context is a stressed syllable followed by at most a single consonant at the Vulgar Latin stage.
A "closed" context is any other syllable type (unstressed, or followed by two or more consonants).
A "late closed" context is a context that is open at the Vulgar Latin (Proto-Romance) stage but becomes closed in the Gallo-Romance stage due to loss an unstressed vowel (usually/e/ or/o/ in a final syllable).
A "palatal" context is a stressed syllable where thepreceding consonant has a palatal quality, causing ayod/j/ to be generated after the preceding consonant, before the stressed vowel.
Changes that occurred due to contexts that developed during the Old French stage or later are indicated in the "Modern French" column. In particular, "+#" indicates a word-final context in modern French, which generally evolved due to loss of a final consonant in Old French or Middle French. For example, loss of/θ/ inaimé "loved" (originally/aiˈmɛθ/) occurred in Old French, while loss of/t/ insot "silly" occurred in Middle French (hence its continuing presence in spelling, which tends to reflect later Old French).
^2 Both/œ/ and/ø/ occur in modern French, and there are a small number ofminimal pairs, e.g.jeune/ʒœn(ə)/ "young" vs.jeûne/ʒøn(ə)/[ʒøːn(ə)] "fast (abstain from food)". In general, however,/ø/ only occurs word-finally, before/z/, and usually before/t/, while/œ/ occurs elsewhere.
^3 However, the sequences*/ueu/ from multiple origins regularly dissimilate to/jɛw/ (and later/jœ/,/jø/) except after labials and velars (Latinlocus →/lueu/ →lieu/ljø/, but*volet →[vuoɫt] →[vueɫt] →[vueut] →veut/vø/).[6]
^4 The changes producing Frenchmoitié/mwaˈtje/ were approximately as follows:[citation needed]
medietātem (Classical Latin form)
[medjeˈtaːtẽː] (pronunciation c. 1st century BC)
[mejjeˈtaːtẽː] (1st century AD:[dj] >[jj])
/mɛjjɛˈtaːteː] (2nd century AD, loss of nasalization)
[mejˈtaːde] (c. 3rd-6th century AD: loss of vowel quantity, new lengthening under stress, raising of[ɛ] to[e] in unstressed syllables, loss of intertonic[e]; firstlenition of intervocalic obstruents, causing voicing of the second[t]; the first[t] is preceded by the consonant[j] and so remains voiceless)
[mejˈtʲaːde] (c. 3rd-6th century AD: progressive palatalization in clusters of[j] + consonant)
[mei̯ˈtieðe] (5th-7th century AD:Bartsch's law, changing[aː] into[ie] after a palatal or palatalized consonant; second lenition of intervocalic stops, changing[d] into[ð] between vowels)
[mei̯ˈtieð] (7th century AD: loss of final unstressed[e])
[mei̯ˈtieθ] (7th century AD: final devoicing)
[mei̯ˈtieθ] (9th century AD, Early Old French)
[mei̯ˈtie] (12th century AD: loss of dental fricatives)
[moi̯ˈtie] (12th century AD, Later Old French: dissimilation of the diphthong[ei̯] to[oi̯])
[mweˈtje] (12th-13th century AD, Late Old French: replacement of fallingdiphthongs with rising diphthongs)
[mwɛˈtje] (13th century AD)
[mwaˈtje] (18th century AD, Classical French and Modern French)⟨moitié⟩
Loss of final/m/ (except in monosyllables: Modern Frenchrien < rem).[8]
/ns/ >/s/.[8] The preceding vowel was long as a result ofcompensatory lengthening (already in Classical Latin).
/rs/ >/ss/ in some words[8] (dorsum > Vulgar Latin*dossu > Modern Frenchdos) but not others (ursum > Modern Frenchours).
Fusion of the diphthongsae andoe to/ɛ(ː)/ and/e(ː)/ respectively (it is disputed whether the fusion ofae initially produced a short or long vowel, or postdated the neutralization of Latin vowel length). The diphthong/au/ was retained.
Development of Latin/w/ and intervocalic/b/ to a voiced labial fricative. The original phonetic realization of this sound was probably bilabial[β] (found today in Spanish as an allophone of/b/), but it eventually became labiodental[v] in French, and so will be transcribed/v/ below.
Introduction ofprosthetic short/i/ before words beginning with/s/ + consonant, which would later become closed/e/ with the Italo-Western vowel change (Spanishespina, Modern Frenchépine "thorn, spine" <espine <spīnam).
Vulgar Latin unstressed vowel loss: Loss ofintertonic (unstressed and in an interior syllable) vowels between/k/,/ɡ/ and/r/,/l/.
Final/-er/ >/-re/,/-or/ >/-ro/ (Spanishcuatro, sobre < quattuor, super).
Reduction of/e/ and/i/ in hiatus to/j/, which would eventually be followed bypalatalization of the resulting consonant +/j/ sequences.
Reduction of ten-vowel system to the seven vowels/aɛeiɔou/ (see table). In unstressed syllables,/ɛɔ/ become/eo/, resulting in only five distinct unstressed vowels.
Palatalization of/k/,/ɡ/ before the front vowels/ɛei/ (around the fifth century AD[10]). For simplicity, the outcomes can be transcribed as/kʲ/,/ɡʲ/; the steps involved in their subsequent phonetic development are debated.
/kʲ/ and/tj/ merge as an affricate/tsʲ/ (treated as a single sound). The double version of this affricate,/ttsʲ/, is the regular outcome of/kkj/, from earlier/kj/, from unstressed Latin/ki/ or/ke/ + vowel.
/j/,/dj/,/ɡj/,/ɡʲ/ have all merged as/j/ by this point. (A merge of some or all of these sounds is also widely seen in other Romance languages, but some languages show divergent developments in at least some words, particularly for/dj/.)
/ɡn/ and/nj/ become/ɲ/.
/ɡl/ and/kl/ become/ʎ/. The intermediate steps are disputed.
/kt/ >/jt/ and/ks/ >/js/; first going through/xt/ and/xs/, respectively.
First lenition (did not happen in a small area around the Pyrenees): chain shift involving intervocalic singleton consonants. Voiced stops and unvoiced fricatives become voiced fricatives, while unvoiced stops become voiced stops. Specifically, intervocalic/dɡ/ >[ðɣ] (Latin intervocalic/b/ had already become/v/); intervocalic/sf/ >[zv], and intervocalic/ptktsʲ/ >/bdɡdzʲ/. The dating is debated; it is sometimes placed as early as the 3rd century AD, but was probably not completed until later;[11] it seems to have been complete in Gaul by the end of the sixth century.[12] Consonants before/r/ are lenited, also, and/pl/ >/bl/. Final/t/ and/d/ are lenited when preceded by a vowel.
First unstressed vowel loss: Loss of intertonic (unstressed and in an interior syllable) vowels except/a/ when pretonic. That occurred at the same time as the first lenition, and individual words inconsistently show one change before the other. Hencemanica >manche butgrānica >grange.carricāre becomes eithercharchier orchargier in Old French. However, in some analyses, the standard for central French was initially for lenition to occur before the unstressed vowel loss, and patterns of the order being reversed, resulting in voiceless consonants, were loaned from the moreFrankish-influenced Northern dialects of Normandy, Champagne and Lorrain, eventually spreading to some other words byanalogy, leading to known cases of divergent development, such asgrange andgranche, andvenger and(re)vencher (the latter both from Latinvindicāre).[14]
Evidence of 9th century French phonology is relatively limited, being based largely on two short documents, theOaths of Strasbourg, written in 842 in what was likely a deliberately Latinized, archaic form of Romance, and theSequence of Saint Eulalia, written around 880 in some Romance vernacular of north central France,[15] not directly ancestral to modern French (the modern French formchose requires palatalization of/ka/ to have taken place before monophthongization of[au̯], whereas theSequence's "cose" shows only the latter of these two sound changes, as in modernPicard[16]). Nevertheless, the following sound changes may be identified as having taken place before or around this period:
Diphthongization of open-mid vowels/ɛ,ɔ/ in stressed open syllables (where the vowels had likely been allophonically lengthened to[ɛː,ɔː]), and also in stressed closed syllables if followed by a palatal sound (often later absorbed). They remain unchanged in other kinds of closed syllables, hencesemper >sempre[ˈsɛmprə] (Eulalia line 10[17]) andmort(u)a >morte[ˈmɔrtə] (Eulalia line 18[17]). Unstressed/ɛ,ɔ/ did not exist, having merged with/e,o/. The diphthongs resulting from this sound change are variously transcribed by modern linguists as[iɛ,uɔ] or[jɛ,wɔ].[18] Old French assonances and rhymes suggest that diphthongization initially producesfalling diphthongs such as[ie̯,uo̯] or[iɛ̯,uɔ̯], with[ie̯] later evolving into a rising diphthong ([jɛ,je]) and[uo̯] later evolving into a front rounded vowel[œ,ø] (possibly with[ue̯] or[wɛ,we] as intermediate steps). However, Porter 1960 reconstructs the rising diphthongs[jɛ,wɔ] as occurring already inEulalia.[19]
In stressed open syllables: Latinbona,caelum > Early Old Frenchbuona,ciel (Eulalia lines 1, 6)[20]
Followed by a palatal in stressed closed syllables:peior >>/ˈpejro/ >/ˈpiejro/ >>pire "worst";noctem >/ˈnojte/ >/ˈnuojte/ >>/nujt/nuit; buttertium >/ˈtertsʲo/ >>tierz.
Diphthongization in the contexts described above predates the earliest Old French texts, but the exact date of this sound change for northern Gallo-Romance languages is uncertain: some place it in the 6th or 7th century, others as early as the 3rd–4th centuries, although Loporcaro 2015 argues the early dating has not been established. Although diphthongization of/ɛ,ɔ/ is a widespread sound change in Romance languages (suggesting it arose relatively early, possibly within a shared community of Late Latin speakers) the conditions in which it occurs are not uniform between languages: for example,/ɛ,ɔ/ diphthongize in both closed and open stressed syllables in Romanian and Spanish,[18] and nowhere in Portuguese.
Second lenition of intervocalic voiced stops (not in all Gallo-Romance):
Between vowels, singleton[bdɡ] (from Latin[ptk]) become[vðɣ]. As before, intervocalic[brdrɡr] were also affected:patrem, capra,[21]sacrāmentum >[ˈpaːdre,ˈkaːbra,saɡraˈmento] > EOF[ˈpæðrə,ˈtʃie̯vrə,saɣraˈment] > Frenchpère, chèvre, serment. Cf.soure[sovrə] 'over' (Eulalia, line 12).[22] The labialized velar[ɡʷ] (from Latin[kʷ] between vowels) became[jw] after stressed front vowels, as insequit > *[ˈsiejwet] >siut, or with metathesis,suit. This could be strengthened to[i̯v] when followed by a vowel or by[r], as insequere > *[ˈsiejwrə] >sivre~siure[ˈsivrə]~[ˈsiwrə]. Northern and eastern dialects show[w] rather than[jw] in this context, hencesequere, sequit >seure~sieure, seut~sieut.[23]
This lenition does not affect[d] that had come into contact with a preceding consonant via intertonic vowel loss, even in cases where that preceding consonant eventually becomes lost or vocalized, as inadcubitāre > Frenchaccouder,*subitānum > Frenchsoudain,*placitāre > Frenchplaider,adjūtāre > Frenchaider.[24] (Pope 1952 interprets forms such as OFaidier,sodain,bondir as showing voicing of[t] to[d] by progressive assimilation after[j,β,b].[25])
In contrast, the glide that develops from diphthongization of[eː] (see below) does not protect a following consonant, as seen inmonēta >[moˈneːda] >[moˈnei̯ðə] > Old Frenchmoneie,monoie.
Palatalization of velars before/a/:
[k,ɡ] before/a/ become palatal affricates[tʃ,dʒ] (late fifth to early sixth century).[26] Very few words failed to palatalize:cavea >>cage, not**chage.
[ɣ] before/a/ becomes a palatal glide[i̯] when preceded by an unrounded vowel.[27]
[ɣ] also becomes a palatal glide[i̯] in the suffixes-īcum and-(i)ācum (amīcum >ami,Tornācum >Tournai, but not in the wordscaecum >cieu andgraecum >grieu).
[pʲ] and[fʲ] become[tʃ];[bʲ] and[vʲ] become[dʒ];[mʲ] becomes[ndʒ]. This development was also seen in Occitan and Ligurian.[28]
When not preceded by a vowel,/j/ becomes[dʒ]. The ultimate source can be Late Latin/dj/,/ɡj/,/ɡ(e,i)/, or word-initial/j/:
Where intertonic vowel loss had brought[j] into contact with following[dtnr], it palatalized them to[dʲtʲnʲrʲ] (as indicated by the development of a following/a/ in a stressed originally open syllable). The preceding vowel develops to a diphthong ending in the glide[i̯].[30] Examples:
double[ddz(ʲ)], which developed to Old French[dz] (as in OFdoze, treze, seze,[35] from[doddze],[36][treddze],[seddze])
[ʎ] (although in writing[ʎ] was represented by "il" or "ill"[37]).
Morphemic[-arʲ-] in inherited words becomes[-ie̯r-] instead of[-ai̯r-], henceoperārium >[obˈraːrʲo] >[obˈrie̯ro] (not*[obˈrai̯ro]) >>ouvrier "worker", butārea >>aire "area" did not participate.
Diphthongization of/e,o/ and fronting of/a/ in stressed, originally open syllables. In other words, these changes affect long[aː,eː,oː], which were either allophones of/a,e,o/ (if it is assumed that diphthongization preceded degemination and final vowel apocope) or distinct phonemes (if degemination and final vowel apocope preceded diphthongization). There is disagreement about the relative ordering of these sound changes.[38] Diphthongization did not affect vowels followed by a palatal glide or palatalized sound.[39] This diphthongization can be dated to the fourth century;[40] it did not occur in all Gallo-Romance.
[oː] becomes[ou̯].
[eː] becomes[ei̯] when not preceded by a palatal sound.
After a palatal or palatalized consonant,[eː] evolves instead to/i/[41] (likely via simplification of[ie̯i̯];[42] see below). Examples:cēra > OFcire,pagēnsem > OFpaïs,placēre > OFplaisir,[41]iacēre > OFgesir
[aː], when not followed by a nasal or preceded by a palatal sound, becomes a vowel that can be transcribed as/æ/.[43] Its actual phonetic quality is debated: in Early Old French, it is usually written⟨e⟩ but does not assonate with either/ɛ/ or/e/. It evolves later in French to[ɛ] in a closed syllable,[e] in an open one.[44] A diphthong such as[aɛ̯] may have been a stage in its development, but alternatively it may have simply developed by fronting of[aː] to[æː], resulting in a phonemic distinction between the four vowel qualities/a/,/æ/,/ɛ/ and/e/. Another common interpretation supposes that[aː] evolved to/eː/ or/ɛː/, a distinctively long vowel in contrast to short/ɛ/ and/e/, although this would be the only phonemic length contrast in the Early Old French vowel system.[45]
Before a nasal,[aː] evolves instead to/ai̯/ when not preceded by a palatalized consonant:manum, amat > OFmain, aime
After a palatalized consonant (including the affricates[tʃ,dʒ,tsʲ] as well as[tʲ,dʲ,nʲ,rʲ]),[aː] evolves instead to[ie̯]. This is known asBartsch's law, and can be dated to the sixth or seventh century.[46] Examples:*cugitāre >[kujeˈtaːre] >[kujeˈdaːre] >[kujˈdaːre] >[kui̯ˈdʲaːre] >>[kui̯ˈdie̯r] OFcuidier "to think",mansiōnātam >[mazʲoˈnaːda] >[mazʲˈnʲaːda] >[mai̯zˈnie̯ðə] > OFmaisniée "household".
Other vowel changes:
[au̯] >[ɔ]. This takes place after the palatalization of velars before/a/.
[ie̯i̯,uo̯i̯], from Proto-Italo-Western Romance *[ɛ,ɔ] before a palatal glide, are simplified to[i,ui̯]. (Alternatively, Pope 1952 explains the development of the second as[ue̯j] >[ye̯j] >[yi̯j] >[yi̯].[47])
Compare the development of[eː] to[i] when preceded by a palatal or palatalized consonant, described above.
Similarly,[ai̯] becomes[i] when preceded by a palatal consonant:iacet > OFgist[d͡ʒist],cacat > OFchie[ˈtʃiə].
Second unstressed vowel loss (complete by around 700 AD)[48]:
In unstressed final syllables, all vowels except/a/ are lost, unless this loss would result in an impermissible final cluster. In that case, the vowel is retained as[ə]:mortem >mort, butvincere >veintre (Eualia).[48]
Unstressed/a/ becomes[ə] in final syllables[49] and in open word-medial syllables.
Other consonant changes:
/h/ (one of the first consonants lost from Classical Latin) is reintroduced in borrowings from Germanic languages.[50]
Single intervocalic[dzʲ] is eventually deaffricated to[zʲ], upon which it merges with the outcome of/sj/. There is conflicting evidence of the date of this sound change. The consonant derived from Latin/k/ before a front vowel seems to have still been a palatalized affricate[dzʲ] or[i̯dz] when the following vowel was lost in a final syllable, resulting in word-final[i̯ts] in Early Old French (spelled "iz"), later simplified to[i̯s].[51][52] In contrast, the consonant derived from Latin/t/ + yod seems to have become a palatalized fricative by the time the following vowel was lost in a final syllable, resulting in word-final[i̯s] in Early Old French.[53] In the Sequence of Saint Eulalia, the letter⟨z⟩ may represent[dz] in the words "domnizelle" and "bellezour"[54] (from Latin *domnicella and *bellatiorem).
Degemination of obstruents: At some point after the lenition of single intervocalic[bdɡdzʲ] to[vðɣzʲ~i̯z], geminate obstruents are simplified to single consonants. This change is variously dated from the 7th-9th century.[56] Since diphthongization of/ɛɔ/, diphthongization of/e,o/ and fronting of/a/ (discussed above) occur only in originally open syllables, some analysts assume that degemination must postdate all of these sound changes. However, it is possible that the distinction at the time of these sound changes was not in the length of the consonant, but in the length of the vowel.[57]
Intervocalic/v/ (probably still pronounced as bilabial[β]) is lost when followed, or sometimes when preceded by a rounded vowel:
[ɫ] is simply deleted after/i/ (filz/ˈfilts/ >fiz[ˈfits]).
[ɫ] was also deleted after rounded vowels[ou̯],[u], and[y] (multum >[moɫt] >molt~mult[mut]).
[ɫ] became[u̯] after the remaining vowels (alterum >[aɫtrə] >autre[au̯trə]).
However,/æ/ (< open/a/) and/ɛ/ when affected by this process, became sequences[i̯eu̯] and[ɛ̯au̯], respectively (pālus >[pæɫs] >pieus[pi̯eu̯s], andbellus >[bɛɫs] >beaus[bɛ̯au̯s]).
[ou̯] >[eu̯], however this is blocked if a labial consonant follows, in which case the segment remains[ou̯], ultimately becoming[u] later.[60] (lupa > OFlouve.)
[uo̯] >[ue̯] (blocked by nasalization; see below).
[u̯ou̯] >[u̯eu̯] >[eu̯] after labials and velars, but[i̯eu̯] elsewhere (Latinlocus →/lueu/ →lieu/ljø/, but*volet →[vuoɫt] →[vueɫt] →[vueut] →veut/vø/).[61]
/a/ develops allophone[ɑ] before/s/, which later develops into a separate phoneme.
Loss of/θ/ and/ð/.[62] When it results in a hiatus of/a/ with a following vowel, the/a/ becomes a schwa/ə/.
Loss of/s/ before voiced consonant (passing first through/h/), with lengthening of preceding vowel. That produces a new set of long vowel phonemes, as is described more completely in the following section.
/u/ >/y/. (This shift, along with the later/o/ >/u/, is an areal feature common to mostGallo-Romance languages.)
Nasal/wɛ̃/ segments, for which there had dialectal variation with nasal/ũ/ previously, are all shifted (or returned) to/ũ/ (ultimately becoming/ɔ̃/) before this can occur.
Rising diphthongs develop when the first element of diphthong is/u/,/y/,/i/.
Stress shifts to second element.
everywhere
Hence[yi̯] >[ɥi]
[oi̯] >/we/
everywhere
Later,/we/ >/ɛ/ in some words likefrançais; note doubletFrançois.
[ai̯] >/ɛ/
everywhere
afterward,⟨ai⟩ is a common spelling of/ɛ/, regardless of origin.
Consonants in coda position word-internally underwent weakening and loss (Gess 1996). This affected/S/ ([z] before voiced consonants and[s] before voiceless ones),/N/ (=nasal consonants), and to some extent the most sonorous coda consonant,/r/. Syllable-final/s/ reduced to[h] before deleting. Borrowings into English suggest that the process occurred first when the following consonant was voiced but not when it was unvoiced (this explains the English pronunciationsisle vs.feast). This process was accompanied by compensatory lengthening of the preceding vowel. Preconsonantal⟨s⟩ was retained as a marker of vowel length (sometimes non-etymologically) until being substituted by⟨ˆ⟩. Syllable-final nasal consonants nasalized and then were absorbed into the preceding vowels, leading to phonemic nasal vowels. Where syllable-final/r/ was weakened and lost word-internally, it was mostly later restored with the notable exception of morphemic-er.
Changes here affect oral and nasal vowels alike, unless otherwise indicated.
⟨ei⟩/ei/ >/ɛ/ (the[ei̯] diphthong is maintained inQuebec French:neige "snow"[nei̯ʒ(ə)] or[naɪ̯ʒ(ə)]).
Loss of final consonants before a word beginning with a consonant. That produces a three-way pronunciation for many words (alone, followed by a vowel, followed by a consonant), which is still maintained in the wordssix "six" anddix "ten" (and until recentlyneuf "nine"), e.g.dix/dis/ "ten" butdix amis/dizaˈmi/ "ten friends" anddix femmes/diˈfam(ə)/ "ten women".
Subject pronouns start to become mandatory because of loss of phonetic differences between inflections.
Medievalapical s, as insaint, merges into deaffricated⟨c⟩ as inceint, thus merging soft⟨c⟩ and⟨s⟩.
Loss of final consonants in a word standing alone. That produces a two-way pronunciation for many words (in close connection with a following word that begins with a vowel), often still maintained:nous voyons/nuvwaˈjɔ̃/ "we see" vs.nous avons/nuzaˈvɔ̃/ "we have". That phenomenon is known asliaison.
⟨oi⟩/we/ >/wa/[68] (see above –To Late Old French) or/ɛ/ (étoit >était; the spelling was not changed until the 19th century). This also affects certain other instances of/we~o̯e/; e.g.moelle/mwal/,poêle/pwɑl/.
Change into/ɛ/ is relatively rare in standard French, it occurs notably in the imperfect tense suffixes (-ais,-ait, and-aient), the adjectival suffix-ois >-ais, as well as other few words, e.g.connoître >connaître.
The pronunciation/we/ is preserved in some forms of Quebec and Acadian French, especially by old speakers.
Instances of/h/ were again deleted in the late seventeenth century. The phoneme/h/ had been reintroduced to the language through the absorption of loanwords, primarily of Germanic origin, and these are the/h/ instances that were lost this time around.[69][50] However a Germanic⟨h⟩ usually disallows liaison:les halles/le.al(ə)/,les haies/le.ɛ/,les haltes/le.alt(ə)/, whereas a Latin⟨h⟩ allows liaison:les herbes/lezɛrb(ə)/,les hôtels/lezotɛl/.
/r/ becomes a uvular sound ("Guttural R"), realized as either a trill/ʀ/ or fricative/ʁ/, in most accents. The alveolar trill is maintained in Acadia, Louisiana, some parts of Québec and in Francophone Africa.
Merger of/ʎ/ (spelled⟨il⟩ inœil andtravail) into/j/, which had begun in the seventeenth and eighteenth centuries, reaches completion around the beginning of the nineteenth century[70] (seeMouillé)
Elision of final/ə/, and occasionally of/ə/ elsewhere, unless a sequence of three consonants would be produced (such constraints operate over multiword sequences of words that are syntactically connected).Occitan French tends to be more conservative, while the elision of final/ə/ does not occur in Francophone Africa.
Changing use ofliaison, which overall becomes rarer.
In Metropolitan French, gradual merging of/œ̃/ and/ɛ̃/, both are realized as[æ̃], but the distinction is maintained in Southern France, Switzerland, Belgium, Quebec and Francophone Africa.
In Metropolitan French, loss of the phoneme/ɑ/, merged with/a/, both are realized as[ä], but the distinction is maintained in French Switzerland, Belgium, Quebec and Francophone Africa.
In Metropolitan French, loss of the phoneme/ə/, merged with/ø/, both are realized as[ø], but the distinction is maintained in Quebec French.
In Metropolitan French, loss of the phoneme/ɛː/, merged with/ɛ/, both are realized as[ɛ], but the distinction is maintained in Northern French, Switzerland, Belgium, Quebec and Francophone Africa.
In Metropolitan French, merger of/ɔ/ into/o/ when word-final, but the distinction is maintained in Belgian French.
Nasalization of vowels before/n/ or/m/ occurred gradually over several hundred years, beginning with the low vowels, possibly as early as 900, and finishing with the high vowels, possibly as late as c. 1300. Numerous changes occurred afterwards that are still continuing.
The following steps occurred during the Old French period:
Nasalization of/a/,/e/,/ɔ/ before/n/ or/m/ (originally, in all circumstances, including when a vowel followed).
Nasalization occurs before and blocks the changes/ei/ >/oi/ and/ou/ >/eu/. However, the sequence/õĩ/ occurs because/oi/ has more than one origin:coin "corner" <cvnevm. The sequences/ĩẽn/ or/ĩẽm/, and/ũẽn/ or/ũẽm/, also occur, but the last two occur in only a few words, in each case alternating with a non-diphthongized variant:om oruem (ModFon), andbon orbuen (ModFbon). The version without the diphthong apparently arose in unstressed environments and is the only one that survived.
Lowering of/ẽ/ and/ɛ̃/ to[æ̃] but not in the sequences/jẽ/ and/ẽj/:bien,plein. The realization of/ẽ/ to[æ̃] probably occurred during the 11th or early 12th century and did not affect Old Norman or Anglo-Norman. Ultimately[æ̃] merged into/ã/.
Nasalization of/i/,/u/,/y/ before/n/ or/m/.
It is not clear if the third-person plural ending-ent contained a nasalized schwa/ə̃/; although then is consistently kept in writing, by Early Modern French at the latest it had become non-nasal/ə/.
The following steps occurred during the Middle French period:
Lowering of/ũ/ >/õ/ >/ɔ̃/. (/ũ/ usually comes from original/oN/, as original/u/ became/y/.)
Denasalization of vowels before/n/ or/m/ followed by a vowel or semi-vowel. (Examples likefemme/fam/ "woman" < OF/ˈfãmə/ <fēminam anddonne/dɔn/ "(he) gives" < OF/ˈdũnə/ <dōnat, with lowering and lack of diphthongization before a nasal even when a vowel followed, show that nasalization originally operated in all environments.)
Deletion of/n/ or/m/ after remaining nasal vowels (when preceding a consonant or word-final):dent/dɑ̃/ "tooth" <*/dãt/ < OFrdent/dãnt/ < EOFr*/dɛ̃nt/ <dentem.
The following steps occurred during the Modern French period:
/ĩ/ >/ẽ/ >/ɛ̃/ >[æ̃] ([ẽɪ̯̃] in Quebec French). That also affects diphthongs such as/ĩẽ/ >/jẽ/ >/jɛ̃/ (bien/bjɛ̃/ "well" <bene);/ỹĩ/ >/ɥĩ/ >/ɥɛ̃/, (juin/ʒɥɛ̃/ "June" <iūnium);/õĩ/ >/wẽ/ >/wɛ̃/, (coin/kwɛ̃/ "corner" <cuneum). Also,/ãĩ/ >/ɛ̃/, (pain/pɛ̃/ "bread" <pānem);/ẽĩ/ >/ɛ̃/, (plein/plɛ̃/ "full" <plēnum).
/ã/ >/ɑ̃/ >[ɒ̃], but the[ã] sound is maintained in Quebec French.
/ɔ̃/ >[õ] ([õʊ̯̃] in Quebec French)
/ỹ/ >/œ̃/ ([ɚ̃] in Quebec French). In the 20th century, this sound has low functional load and has tended to merge with/ɛ̃/.
That leaves only four nasal vowels:/ɛ̃/,/ɑ̃/,/ɔ̃/, and/œ̃/, the last often no longer being distinguished from the first.
^The changes occurred in the majority of Vulgar Latin, specifically theItalo-Western Romance area, which underlies the vast majority of Romance languages spoken in Italy, France, Belgium, Spain, Portugal, and Andorra. However, different vowel changes occurred elsewhere, in the Vulgar Latin underlying modernRomanian,Sardinian,Corsican, and a few modern southern Italian varieties.
^Found as⟨pēior⟩ "worse" in many 19th and 20th century editions, but was actually pronounced/ˈpej.jor/, with a short/e/ followed by a geminate/jj/; writing the macron is a convention to mark the resulting syllable weight.
^Morin, Yves (2008). "235. Histoire interne du français: Histoire des systèmes phonique et graphique du français". In Ernst, Gerhard; Gleßgen, Martin-Dietrich; Schmitt, Christian; Schweickard, Wolfgang (eds.).Romanische Sprachgeschichte. Ein internationales Handbuch zur Geschichte der romanichen Sprachen. De Gruyter. p. 2917.
^abRobert McColl Miller; Larry Trask (20 February 2015).Trask's Historical Linguistics. Routledge.ISBN9781317541769.Between the fifth and eighth centuries, French borrowed a number of Germanic words with [h]... and [h] thus rejoined the French phonological system... the [h]s had disappeared by the eighteenth century.
Gess, Randall (1996)Optimality Theory in the Historical Phonology of French. PhD dissertation, University of Washington
Buckley, Eugene (2009). "Phonetics and phonology in Gallo-Romance palatalisation".Transactions of the Philological Society.107:31–65.doi:10.1111/j.1467-968X.2008.00212.x.
Fought, John (1979). "The 'Medieval Sibilants' of the Eulalia-Ludwigslied Manuscript and Their Development in Early Old French".Language.55 (4):842–858.doi:10.2307/412747.JSTOR412747.
Harris, Martin (1988), "French", in Harris, Martin; Vincent, Nigel (eds.),The Romance Languages, Oxford University Press, pp. 209–245,ISBN978-0195208290
Kibler, William (1984),Introduction to Old French, Modern Language Association of America,ISBN978-0873522922
Loporcaro, Michele (2015).Vowel Length from Latin to Romance. Oxford University Press.
