Catalan phonology

Voiced obstruents undergo final-obstruent devoicing so that fred ('cold', m. s.) is pronounced with [t], while fredes ('cold', f. pl.) is pronounced with [ð].[18]

Voiced stops become lenited to approximants in syllable onsets, after continuants:[10] /b/ → [β], /d/ → [ð], /ɡ/ → [ɣ]. Exceptions include /d/ after lateral consonants and /b/ after /f/, e.g. ull de bou [ˈuʎ də ˈβɔw] (EC) [ˈuʎ de ˈβɔw] (WC) ('oeil-de-boeuf'), bolígraf boníssim [buˈɫiɣɾəv‿buˈnisim] (EC) [boˈliɣɾav‿boˈnisim] (WC) ('excellent ballpoint'). Additionally, /b/ remains unlenited in non-betacist dialects. In the coda position, these sounds are always realized as stops[19] except in many Valencian dialects, where they are lenited.[20]

In some Valencian dialects final /p, t, k/ can be lenited before a vowel: tot açò [ˈt̪oð‿aˈsɔ] ('all this').[21]

In some dialects (e.g. many Valencian accents) initial /ɡ/ can be lenited: gata [ˈɣat̪ə] (EC) [ˈɣat̪a] (WC).[22]

In many Catalan dialects (except Valencian), /b/ and /ɡ/ may be geminated in certain environments (e.g. poble [ˈpɔbːɫə] 'village', regla [ˈreɡːɫə] 'rule').[23][24]

In Majorcan varieties, /k/ and /ɡ/ become [c] and [ɟ] word-finally and before front vowels,[20] in some of these dialects, this has extended to all environments except before liquids and back vowels; e.g. sang [ˈsaɲc] ('blood').[10]

The phonemic status of affricates is dubious; after other consonants, affricates are in free variation with fricatives, e.g. clenxa [ˈkɫɛnʃə ~ ˈkɫɛɲt͡ʃə] (EC) [ˈklɛɲt͡ʃa] (WC) ('hair parting')[25] and may be analyzed as either single phonemes or clusters of a stop and a fricative.

• Alveolar affricates, [t͡s] and [d͡z], occur the least of all affricates.[26]
  • [d͡z] only occurs intervocalically: metzines [məˈd͡zinəs] (EC) [meˈd͡zines] (WC) ('toxic substances').[27]
  • Instances of [t͡s] arise mostly from compounding; the few lexical instances arise from historical compounding.[25] For instance, potser [puˈtse] (EC) [poˈtseɾ] (WC) ('maybe') comes from pot ('may') + ser ('be' inf). As such, [t͡s] does not occur word-initially; other than some rare words of foreign origin (e.g. tsar 'tsar',[28] tsuga 'tsuga'[29]), but it may occur word-finally and quite often in cases of heteromorphemic (i.e. across a morpheme boundary) plural endings: tots [ˈtots] ('everybody').[26]
• The distribution of alveolo-palatal affricates, [t͡ʃ] and [d͡ʒ], depends on dialect:
  • In Standard Eastern Catalan, word-initial [t͡ʃ] is found only in a few words of foreign origin (e.g. txec 'Czech',[30] Txaikovski 'Tchaikovsky') while being found freely intervocalically (e.g. fletxa 'arrow') and word-finally: despatx [dəsˈpat͡ʃ] (EC) [desˈpat͡ʃ] (WC) ('office').
  • Standard Eastern Catalan also only allows [d͡ʒ] in intervocalic position (e.g. metge 'medic', adjunt 'enclosed'). Phonemic analyses show word-final occurrences of /d͡ʒ/ (e.g. raig esbiaixat [ˈrad͡ʒ‿əzbiəˈʃat] (EC) [ˈrad͡ʒ‿ezbiajˈʃat] (WC) 'skew ray'), but final devoicing eliminates this from the surface: raig [ˈrat͡ʃ] ('ray').
  • In various other dialects (as well as in emphatic speech),[31] [tʃ] occurs word-initially and after another consonant to the exclusion of [ʃ]. These instances of word-initial [t͡ʃ] seem to correspond to [ʃ] in other dialects, including the standard (on which the orthography is based): xinxa ('bedbug'), pronounced [ˈʃiɲʃə] in the standard, is [ˈt͡ʃiɲt͡ʃa] in these varieties.[27]
  • Similarly, in most of Valencian and southern Catalonia,[26][32] most occurrences of [d͡ʒ] correspond to the voiced fricative [ʒ] in Standard Eastern Catalan: gel [ˈd͡ʒɛl] ('ice').

There is dialectal variation in regards to affricate length, with long affricates occurring in both Eastern and Western dialects such as in Majorca and few areas in Southern Valencia.[33] Also, intervocalic affricates are predominantly long, especially those that are voiced or occurring immediately after a stressed syllable (e.g. metge [ˈmed͡ʒːə] (EC) [ˈmed͡ʒːe] (WC) 'medic').[34] In modern Valencian [d͡ʒ] and [d͡ʒː] have merged into /d͡ʒ/.

/v/ occurs in Balearic,[31] as well as in Alguerese, standard Valencian and some areas in southern Catalonia.[35] Everywhere else, it has merged with historic /β/ so that [b] and [β] occur in complementary distribution.[36] In Majorcan, [v] and [w] are in complementary distribution, with [v] occurring before vowels (e.g. blava [ˈbɫavə] 'blue' f. vs. blau [ˈbɫaw] 'blue' m.). In other varieties that have both sounds, they are in contrast before vowels, with neutralization in favor of [w] before consonants.[37]

In some Valencian dialects, /s/ and /ʃ/ are auditorily similar such that neutralization may occur in the future.[38] That is the case of Northern Valencian where /ʃ/ is depalatalized to [js̠ ~ jsʲ] as in caixa ('box'). Central Valencian words like mig ('half') and lleig ('ugly') have been transcribed with [ts] rather than the expected [t͡ʃ], and Southern Valencian /t͡ʃ/ "has been reported to undergo depalatalization without merging with [t͡s]".[39] as in passets ('small steps') versus passeig ('promenade')

In Aragon and Central Valencian (the so called apitxat), voiced fricatives and affricates are missing (i.e. /z/ has merged with /s/, /d͡ʒ/ has merged with /t͡ʃ/, with only voiceless realizations occurring) and /v/ has merged with the [b ~ β] set.[40]

While "dark (velarized) l", [ɫ], may be a positional allophone of /l/ in most dialects (such as in the syllable coda; e.g. sòl [ˈsɔɫ] 'ground'),[41] /l/ is dark irrespective of position in Eastern dialects like Majorcan[42] and standard Eastern Catalan (e.g. tela [ˈtɛɫə]).

The distribution of the two rhotics /r/ and /ɾ/ closely parallels that of Spanish. Between vowels, the two contrast (e.g. mirra [ˈmirə] (EC) [ˈmira] (WC) 'myrrh' vs. mira [ˈmiɾə] (EC) [ˈmiɾa] (WC) 'look'), but they are otherwise in complementary distribution. [ɾ] appears in the onset, except in word-initial position (ruc), after /l/, /n/, and /s/ (honra, Israel), and in compounds, where [r] is used. Different dialects vary in regards to rhotics in the coda, with Western Catalan generally featuring [ɾ] and Central Catalan dialects like those of Barcelona or Girona featuring a weakly trilled [r] unless it precedes a vowel-initial word in the same prosodic unit, in which case [ɾ] appears (per [peɾ] in Western Catalan, [pər] in Central Catalan).[43] There is free variation in "r" word-initially, after /l/, /n/, and /s/, and in compounds (if /r/ is preceded by consonant), wherein /r/ is pronounced [r] or [ɹ], the latter being similar to English red: ruc [ˈruk ~ ˈɹuk], honra [ˈonrə ~ ˈonɹə] (EC) [ˈonra ~ ˈonɹa] (WC), Israel [isrəˈɛl ~ isɹəˈɛl] (EC) [israˈɛl ~ isɹaˈɛl] (WC).

In careful speech, /n/, /m/, and /l/ may be geminated (e.g. innecessari [inːəsəˈsaɾi] (EC) [inːeseˈsaɾi] (WC) 'unnecessary'; emmagatzemar [əmːəɣəd͡zəˈma] (EC) [emːaɣad͡zeˈmaɾ] (WC) 'to store'; il·lusió [ilːuziˈo] 'illusion'). A geminated /ʎː/ may also occur (e.g. ratlla [ˈraʎːə] (EC) [ˈraʎːa] (WC) 'line').[31] Wheeler (1979) analyzes intervocalic [r] as the result of gemination of a single rhotic phoneme: serra /ˈsɛɾɾə/ → [ˈsɛrə] (EC) /ˈsɛɾɾa/ → [ˈsɛra] (WC) 'saw, mountains' (this is similar to the common analysis of Spanish and Portuguese rhotics).[44]

Vowels of Standard Eastern Catalan, from Carbonell & Llisterri (1999:62)

Vowels of Valencian, from Saborit Vilar (2009:23)

Most varieties of Catalan contrast seven stressed vowel phonemes.[53] However, some Balearic dialects have an additional stressed vowel phoneme (/ə/); e.g. sec /ˈsək/ ('dry').[20] The stressed schwa of these dialects corresponds to /ɛ/ in Central Catalan and /e/ in Western Catalan varieties (that is, Central and Western Catalan dialects differ in their incidence of /e/ and /ɛ/, with /e/ appearing more frequently in Western Catalan; e.g. Central Catalan sec /ˈsɛk/ vs. Western Catalan sec /ˈsek/ 'dry, I sit').[53]

Contrasting series of the main Catalan dialects:

In Eastern Catalan, vowels in unstressed position reduce to three : /a/, /e/, /ɛ/ → [ə]; /o/, /ɔ/, /u/ → [u]; /i/ remains unchanged. However there are some dialectal differences: Alguerese merges /a/, /e/, and /ɛ/ with [a]; and in most areas of Majorca, [o] can appear in unstressed position (that is, /o/ and /ɔ/ are usually reduced to [o]).[54]

In Western Catalan, vowels in unstressed position reduce to five: /e/, /ɛ/ → [e]; /o/, /ɔ/ → [o]; /a/, /u/, /i/ remain unchanged.[55] However, in some Western dialects reduced vowels tend to merge into different realizations in some cases:

• Unstressed /e/ may merge with [a] before a nasal or sibilant consonant (e.g. enclusa [aŋˈkluza] 'anvil', eixam [ajˈʃam] 'swarm'), in some environments before any consonant (e.g. terròs [taˈrɔs] 'earthy'), and in monosyllabic clitics. This sounds almost the same as the Barcelonian open schwa [ɐ].[56] Likewise, unstressed /e/ may merge into [i] when in contact with palatal consonants (e.g. senyor [siˈɲoɾ ~ siˈɲo] 'lord').[57]
• Unstressed /o/ may merge with [u] before a bilabial consonant (e.g. cobert [kuˈβɛɾt] 'covered'), before a stressed syllable with a high vowel (e.g. conill [kuˈniʎ] 'rabbit'), in contact with palatal consonants (e.g. Josep [d͡ʒuˈzɛp] 'Joseph'), and in monosyllabic clitics.[58]

There are also a number of phonetic diphthongs and triphthongs, all of which begin and/or end in [j] or [w].[59]

In standard Eastern Catalan, rising diphthongs (that is, those starting with [j] or [w]) are only possible in the following contexts:[60]

• [j] in word-initial position, e.g. iogurt ([juˈɣur]) ('yoghurt').
• The semivowel ([j] or [w]) occurs between vowels as in feia ([ˈfɛjə] 'he/she was doing') or diuen ([ˈdiwən] 'they say').
• In the sequences [ɡw] or [kw] plus vowel, e.g. guant ('glove'), quota ('quota'), qüestió ('question'), pingüí ('penguin'); these exceptional cases even lead some scholars[61] to hypothesize the existence of rare labiovelar phonemes /ɡʷ/ and /kʷ/.[62]

Catalan denti-alveolar stops can fully assimilate to the following consonant, producing gemination; this is particularly evident before nasal and lateral consonants: e.g. cotna ('rind'), motlle/motle ('spring'), and setmana ('week'). Learned words can alternate between featuring and not featuring such assimilation (e.g. atles [ˈadɫəs ~ ˈaɫːəs] 'atlas', administrar [ədminisˈtɾa ~ əmːinisˈtɾa] 'to administer').[65][66]

Central Valencian features simple elision in many of these cases (e.g cotna [ˈkona], setmana [seˈmana]) though learned words don't exhibit either assimilation or elision: atles [ˈadles] and administrar [adminisˈtɾaɾ].[67]

Stress most often occurs on any of the last three syllables of a word (e.g. brúixola [ˈbɾuʃuɫə] (EC) [ˈbɾujʃola] (WC) 'compass', càstig [ˈkastik] 'punishment', pallús [pəˈʎus] (EC) [paˈʎus] (WC) 'fool').

Compound words and adverbs formed with /ˈment/ may have a syllable with secondary stress (e.g. bonament [ˌbɔnəˈmen] (EC) [ˌbɔnaˈment] (WC) 'willingly'; parallamps [ˌpaɾəˈʎams] (EC) [ˌpaɾaˈʎamps] (WC) 'lightning conductor') but every lexical word has just one syllable with main stress.[68]

Any consonant, as well as [j] and [w] may be an onset. Clusters may consist of a consonant plus a semivowel (C[j], C[w]) or an obstruent plus a liquid. Some speakers may have one of these obstruent-plus-liquid clusters preceding a semivowel, e.g. síndria [ˈsin.dɾjə] ('watermelon'); for other speakers, this is pronounced [ˈsin.dɾi.ə] (i.e. the semivowel must be syllabic in this context).[69]

Word-medial codas are restricted to one consonant + [s] (extra [ˈɛks.tɾə] (EC) [ˈɛks.tɾa] (WC)).[70] In the coda position, voice contrasts among obstruents are neutralized.[71] Although there are exceptions (such as futur [fuˈtuɾ] 'future'), syllable-final rhotics are often lost before a word boundary or before the plural morpheme of most words: color [kuˈɫo] (EC) [koˈloɾ] (WC) ('color') vs. coloraina [kuɫuˈɾajnə] (EC) [koloˈɾajna] (WC) ('bright color').[31]

In Central Eastern Catalan, obstruents fail to surface word-finally when preceded by a homorganic consonant (e.g. /nt/ → [n]). Complex codas simplify only if the loss of the segment doesn't result in the loss of place specification.[72]

When the suffix -erol [əˈɾɔɫ] is added to camp [ˈkam] it makes [kəmpəˈɾɔɫ], indicating that the underlying representation is /ˈkamp/ (with subsequent cluster simplification), however when the copula [ˈes] is added it makes [ˈkam ˈes]. The resulting generalization is that this underlying /p/ will only surface in a morphologically complex word.[73] Despite this, word-final codas are not usually simplified in most of Balearic and Valencian (e.g. camp [ˈkamp]).[74]

Word-initial clusters from Graeco-Latin learned words tend to drop the first phoneme: pneumàtic [nəwˈmatik] (EC) [newˈmatik] (WC) ('pneumatic'), pseudònim [səwˈðɔnim] (EC) [sewˈðɔnim] (WC) ('pseudonym'), pterodàctil [təɾuˈðaktiɫ] (EC) [teɾoˈðaktil] (WC) ('pterodactylus'), gnom [ˈnom] ('gnome').[75]

Word-final obstruents are devoiced; however, they assimilate voicing of the following consonant, e.g. cuc de seda [ˈkuɡ‿də ˈsɛðə] (EC) [ˈkuɡ‿de ˈsɛða] (WC) ('silkworm'). In regular and fast speech, stops often assimilate the place of articulation of the following consonant producing phonetic gemination: tot bé [ˈtod‿ˈbe] → [ˈtob‿ˈbe] ('all good').[76]

Word-final fricatives (except /f/) are voiced before a following vowel; e.g. bus enorme [ˈbuz‿əˈnormə] (EC) [ˈbuz‿eˈnorme] (WC) ('huge bus').[77]

In Majorcan and Minorcan Catalan, /f/ undergoes total assimilation to a following consonant (just as stops do): buf gros [ˈbuɡ‿ˈɡɾɔs] ('large puff').[78]