BBN ANG 243 Phonological analysis Laryngeal …seas3.elte.hu/analysis/analysis03-slides-h.pdflaryngeal contrast voicing types of voicing & devoicing based on aerodynamics 1. spontaneous(“modal”)

BBN–ANG–243 Phonological analysis

Laryngeal contrast in English consonants

Zoltán G. Kiss, Attila Starcevic, Péter Szigetvári, Miklós Törkenczy

Dept. of English Linguistics, Eötvös Loránd University

z. g. kiss (elte|delg) analysis | laryngeal contrast 1 / 97

aims

lecture outline

◮ contrast among English consonants (obstruents)

◮ laryngeal (“voicing”) contrast

◮ phonological modelling and phonetic implementation of the contrast

◮ when contrast disappears: neutralization

◮ voicing assimilation


revision

revision of some basic concepts

◮ phonology = mapping between sounds and contrastive words

◮ the generative model of phonology

◮ phonemes, allophones, features


revision

when you know a language, you have this ability:


revision

phonology = mapping between sounds & contrastive units

◮ when you know a language, you can isolate di=erent meaningful unitsin the speech signal

◮ = you can recognise and produce contrastive words, morphemes

◮ = you can map between chunks of sound sequences and meaningfulunits (in both directions)

◮ this is your phonological competence, or simply your phonology

◮ phonological competence is closely linked to the notion of contrast


revision

how can we model this competence?


revision

the generative model of language

phonology is one of the modules of linguistic competence (“grammar”)


revision

the typical generative model of language


revision

the lexicon

◮ stores a list of contrastive words/morphemes

◮ they are made up of abstract contrasting segments which we callphonemes

◮ plus any information about them which needs to be memorized(e.g., meaning)


revision

the phoneme principle

Every language has a limited set of phonemes;and every word in the language is built up from those phonemes.


revision

an extract from an English speaker’s lexicon

◮ /kat/ – ‘a small domesticated carnivorous mammal’, . . .

◮ /dIp/ – ‘put something into liquid’, . . .

◮ /sal@d/ – ‘a cold dish of raw or cooked vegetables’, . . .

◮ . . .


revision

an example for phonological mapping

/ s a l @ d /

[ s a l @ d ]

◮ the top level is called the underlying/phonological/phonemicrepresentation (UR), it contains the contrastive phonemes that makeup the di=erent words in the lexicon

◮ the bottom level is called the surface/phonetic representation (SR), itcontains the phoneme-realizations: the allophones


revision

an example for phonological mapping

/ s a l @ d /| | | | |

[ s a l @ d ]

◮ what phonology does is map each phoneme to a speech sound, andeach speech sound to a phoneme

◮ it defines how a phoneme has to be articulated (“derivation”), and ittells which meaningful unit a sound belongs to


revision

positional variation

◮ sometimes 2 or more allophones belong to the same phoneme

◮ in this case mapping to the appropriate allophone is decided by aphonological rule

◮ e.g.: “if /l/ is followed by a V, map it to clear L [l]”

◮ “elsewhere, map it to dark L [ë]”


revision

mapping of the Ls in eleven films

UR: /IlEv@n fIlmz/

L-Darkening Rule: l ë

SR: [IlEv@n fIëmz]


features

what is it exactly that makes the di=erence between /d/ and /s/, etc,possible?


features

features

◮ /d/ and /s/ are phonemes because e.g., dine ⇐⇒ sign

◮ we can explicitly express why they contrast with the distinctive features

they contain

/d/ ⇐⇒ /s/voicing: [+voice] [–voice]

nasality: [–nasal] [–nasal]place: [+alveolar] [+alveolar]manner: [+stop] [–stop]


features

features

◮ a phoneme = a sum of these features

◮ e.g., /d/ = voiced + oral + alveolar + stop

◮ features are usually defined based on articulation (phonetics)

◮ they can be thought of as the instructions the brain sends to thespeech organs to implement sounds


features

phonetic implementation of distinctive features

you want to say the word sign:

Phonological representation =⇒ Phonetic implementation

[−voice] ‘do not vibrate vocal folds’ +[−nasal] ‘do not lower the velum’ +

/s/ =[+alveolar] ‘raise tongue to alveolar ridge’ +[−stop] ‘do not create complete closure’


consonant contrasts

consonant contrasts in English


consonant contrasts

the consonant inventory (contrastive consonants)

Bil. Lab-den. Dental Alveolar Palatal Velar Glottal

P p | b t | d k | gO F f | v T | D s | z S | Z

A Ù Ã

N | m | n | NS L | l | r

G | w | j h |

Bil. = bilabial, Lab-den. = labio-dentalP = stop/plosive, F = fricative, A = a=ricate, N = nasal, L = liquid, G = glideO = obstruent, S = sonorant| = left of line: “voiceless”, right of line: “voiced”


consonant contrasts

obstruent contrasts

◮ place contrast: e.g., /t/ vs. /p/: tin–pin; /t/ vs. /k/: tan–can

◮ manner contrast: e.g., /t/ vs. /s/: tin–sin

◮ “voicing” contrast: /t/ vs. /d/: time–dime


laryngeal contrast

“voicing” contrast in obstruents

OBSTRUENTS

STOPS FRICATIVES

/t/ – /d/ /s/ – /z/

tie – die sip – zipwriter – rider missle – mizzlebeat – bead bus – buzz

‘voiceless’ ‘voiced’ ‘voiceless’ ‘voiced’


laryngeal contrast

laryngeal contrast in obstruents

◮ the phonological contrast of “voicing” is signalled (= cued)by a complex of features

◮ there are several correlates of this contrast = there are many“concomitant” features for the contrast

◮ vocal fold vibration is only one of them


laryngeal contrast

laryngeal contrast in obstruents

◮ let’s call the phonological contrast between tie–die, writer–rider,bus–buzz, etc. laryngeal contrast

◮ voicing is a narrowly used phonetic term: vocal fold vibration(also called: phonation)


laryngeal contrast

some phonetic correlates of laryngeal contrast

in obstruents

◮ voicing/phonation: vocal fold vibration

◮ Voice Onset Time (VOT)

◮ relative length of preceding vowel

◮ glottalization

◮ release noise/burst: intensity & length


laryngeal contrast voicing

larynx: vocal folds + glottis



vocal folds: periodic vibration



states of the vocal folds



cross-section of the larynx



vocal fold vibration: the Bernoulli e=ect



the aerodynamics of voicing: steps of vocal fold vibrationvocal fold vibration happens because of air pressure changes

(aerodynamic reasons):

1. vocal folds loosely close

2. air pressure increases below vocal folds

3. air pressure blows vocal folds apart (glottis opens)

4. speed of air particles increases through narrow glottis

5. air pressure decreases below/within vocal folds ⇒ vocal folds suckedtogether (Bernoulli e=ect)

6. vocal folds are closed again, a cycle like this repeats itself approx.100–300 times/second

7. the cycles last until the state of glottis changes (e.g., opens to producea voiceless sound)

– phonation can start and can continue when air pressure is higher below

the vocal folds than above it



types of voicing & devoicing based on aerodynamics

1. spontaneous (“modal”) voicing: open oral cavity – this helps to startand maintain voicing because air pressure will be low in the mouth buthigh below vocal folds ⇒ vowels, sonorants

2. passive devoicing: closure/constriction in mouth – this creates highair pressure above vocal folds, which inhibits vocal fold vibration⇒ obstruents



voicing in obstruents

◮ obstruents easily get devoiced

◮ but there exist voiced obstruents, how?

◮ passive voicing

◮ active voicing



passive voicing

◮ when obstruents are between two vowels or sonorants, voicing fromthe preceding vowel/sonorant continues throughout the obstruent

◮ this is what we call passive (“lenis”) voicing







passive voicing

◮ English “voiced” obstruents are passively voiced: “lenis”

◮ examples: rider,bandit,rabid,gamble,begin,English,gadget,nostalgia,fuzzy,palsy,Magda,exam /Igzam/. . .



active voicing

◮ additional, “extra” voicing articulation-strategies are used to delaydevoicing in obstruents

◮ e.g., lower the larynx, enlarge the oral cavity

◮ Hungarian, French, Russian, etc. obstruents are like this: activelyvoiced



active devoicing

◮ between vowels/sonorants, obstruents are passively voiced

◮ to produce voiceless obstruents in such positions, extra articulatorye=ort is needed

◮ this is called active devoicing

◮ examples from English: city, lucky, tempo, etc.



types of voicing

summary

Type A=ected sounds

spontaneous voicing vowels, sonorant consonants

passive devoicing obstruents

passive voicing obstruents between Vs/son. (English, etc.)

active voicing obstruents (Hungarian, etc.)

active devoicing voiceless obstruents between Vs/son.



some important consequences of voicing types

◮ vowels and sonorants are usually only voiced (no voiceless pair)

◮ if an obstruent is passively voiced (as in English), its voicing isdependent on its environment

◮ passively voiced obstruents are only fully voiced betweenvowels/sonorants or another passively voiced obstruent

◮ elsewhere they are usually devoiced, e.g.:

◮ word-initial position: back, demon, game, juice. . .◮ word-final position: rob, lead, vague, bridge. . .◮ next to a non-lenis obstruent:

anecdote, Afghan, Agfa, Aztec. . .

◮ in Hungarian: “voiced” obstruents are typically voiced in all positions(initially and word-finally, too): bab, babos, méz, rúzs. . .(except in voicing assimilation position: méztol [st] – see later)



some important consequences of voicing types

◮ Hungarian obstruents contain the feature [±voice] in their underlyingrepresentation

◮ English obstruent do not contain the feature [±voice] in theirunderlying representation

◮ so what feature makes the contrast possible between Englishobstruents? for example: tip – dip?

◮ VOT to the rescue!


laryngeal contrast VOT

articulation phases of stops

◮ for example: repel /rIpEl/

1. vowel/sonorant2. closure and hold3. release4. transition into the vowel5. vowel



Voice Onset Time, VOT

◮ voicing of stops can be characterized by the timing between the release

and the beginning (“onset”) of voicing of the next vowel/approximant

◮ we call this timing relationship Voice Onset Time (VOT)



Voice Onset Time, VOT

◮ three major VOT possibilities:

1. voicing begins right after or only shortly after the release:zero VOT/short lag VOT

2. voicing begins later than the release, there is a relatively long lagof voicing after the release: positive/long lag VOT

3. voicing is already underway during the closure and release:negative VOT/VOT lead

◮ these three VOT options give three phonetic laryngeal categoriesof stops



three Voice Onset Time options



three Voice Onset Time options



voicing and aspirating languages

“voicing” “aspirating”

voiced voiceless voiceless

unaspirated aspirated

[d] [t] [th]

Hawaiian [t]Hungarian ⇐⇒[d] [t]English ⇐⇒[t] [th]Thai ⇐⇒ ⇐⇒[d] [t] [th]Cl. Greek: βας /bas/ – πας /pas/ – ϕας /phas/




voicing (zero VOT ⇔−VOT) languages

most Romance and Slavic languages (Spanish, Italian, French, Russian,Polish, Slovak, etc.) but also Dutch, Yiddish, Scottish English, andHungarian

aspirating (+VOT ⇔ zero VOT) languages

most Germanic (English, German, Danish, Icelandic, Norwegian, Swedish,etc.) but also some Turkic languages




voicing (zero VOT ⇔−VOT) languages: [±voice]

– di=erence between obstruents is due to voicing

– e.g., /p/ = voiceless, [−voice]; /b/ = voiced, [+voice]

aspirating (+VOT ⇔ zero VOT) languages: [±fortis]

– di=erence between obstruents is due to fortisness

– e.g., /p/ = fortis, [+fortis]; /b/ = lenis, [−fortis]



fortis vs. lenis obstruents

Fortis [+fortis] Lenis [−fortis]

never voiced have passive voicing, only voicedbetween Vs/sonorants/lenis obst.’s

can be aspirated never aspirated

can shorten the preceding vowel(“Pre-Fortis Clipping”)

never shorten preceding vowel

can be glottalized can never be glottalized

“stronger” articulation “weaker” articulation



fortis vs. lenis obstruent contrast in English

English obstruent phonemes

Stops A=r. FricativesFortis [+fortis] /p t k/ /Ù/ /f T s S/Lenis [−fortis] /b d g/ /Ã/ /v D z Z/

◮ lenis phonemes have two allophones: (partially) voiceless and voiced,the voiced allophone is the most limited

◮ for example: /b/

[b]between Vs/son.

[p] or [b˚

]elsewhere

◮ bin /bIn/ phonetically: [pIn] or [b˚

In]Robin /rObIn/ phonetically: [rObIn]

◮ fortis phonemes have various allophones (aspirated, unaspirated,glottalized), as we will see (also see your previous studies!)



phonetic implementation of [±fortis] in Englishexample: tip – dip:


[+fortis] ‘aspirate/+VOT’ +[−nasal] ‘don’t lower the velum’ +

/t/ =[+alveolar] ‘raise tongue to alveolar ridge’ +[+stop] ‘create complete closure’

= [th]

[−fortis] ‘don’t do anything/∅VOT’ +[−nasal] ‘don’t lower the velum’ +

/d/ =[+alveolar] ‘raise tongue to alveolar ridge’ +[+stop] ‘create complete closure’

= [d˚

] = [t]



phonetic implementation of [±voice] in Hungarianexample: tél ‘winter’ – dél ‘noon’:


[−voice] ‘don’t vibrate vocal folds/∅VOT’ +[−nasal] ‘do’nt lower the velum’ +

/t/ =[+alveolar] ‘raise tongue to alveolar ridge’ +[+stop] ‘create complete closure’

= [t]

[+voice] ‘vibrate vocal folds/−VOT +[−nasal] ‘don’t lower the velum’ +

/d/ =[+alveolar] ‘raise tongue to alveolar ridge’ +[+stop] ‘create complete closure’

= [d]


laryngeal contrast positions

laryngeal contrast of English stops in various positions

◮ as we saw, there can be various phonetic implementations of [+fortis]and [−fortis]

◮ they depend on the position of the obstruent

1. between sonorants, before a stressed vowel: repél – rebél2. word-initial, before a stressed or unstressed vowel: tíe – díe,

políte – Bolívia3. between sonorants, before an unstressed vowel: wríter – ríder



1. between Vs/sonorants, before a stressed vowel

Word in spelling: repél rebélUnderlying repr.: /p/ /b/

[+fortis] [−fortis]

Phonetic implem.:w

w

�

w

w

�

aspirated/+VOT unaspirated/∅VOTplus: passive voicing

= [ph] = [b]

– on the surface, both aspiration and phonetic voicing make the contrastpossible, we can call it a strong position for the obstruent contrast



2. word-initial, before vowel (stressed or unstressed)

Word in spelling: tíe díeUnderlying repr.: /t/ /d/


Phonetic implem.:w

w

�

w

w

�

aspirated/+VOT unaspirated/∅VOT= [th] = [d

˚] or [t]

– only aspiration is active in this position for the contrast on the surface



3. between sonorants, before an unstressed vowel

Word in spelling: wríter ríderUnderlying repr.: /t/ /d/


Phonetic implem.:w

w

�

w

w

�

unaspirated/∅VOT unaspirated/∅VOTplus: passive voicing

= [t] = [d]

◮ only voicing is active in this position for the contrast on the surface

◮ note: 1. fortis stops may be weakly aspirated in this position, too2. length of stops is relatively short here, and voicing may continuethroughout the stop: /t/ and /d/ may become a flap [R] in AmericanEnglish, but not /p/–/b/ or /k/–/g/: rápid – rábid still contrast


laryngeal contrast word-final

absolute word final position: beat – bead

◮ in this position, voicing is di;cult to maintain

◮ since nothing follows the stop, aspiration is also impossible


laryngeal contrast word-final

no contrast in beat – bead?

Word in spelling: beat beadUnderlying repr.: /t/ /d/


Phonetic implem.:w

w

�

w

w

�

unaspirated/∅VOT unaspirated/∅VOTNO passive voicing here!

= [t] = [d˚

] or [t]

◮ has English given up contrast in word-final position? = neutralization

◮ or maybe there are features other than aspiration or voicing that getactivated here to maintain the contrast. . .


laryngeal contrast neutralization

neutralization: the beer goggle e=ect



neutralization

The disappearance of contrast under a given condition.

= The local suspension of a phonological oppositionbetween two or more contrastive sound segments; only one segmentcan appear in that position (but not its contrastive counterpart(s)).



neutralization: the beer goggle e=ect

◮ opposition: the attractiveness of people is perceived di=erently

◮ condition: being drunk

◮ output: the di=erence in attractiveness disappears (all people areperceived as attractive)



neutralization example: vowel reduction

◮ a wide range of vowels can appear in a stressed syllable but inunstressed syllables, vowel contrast is reduced to a handful of vowels(primarily the schwa)

◮ senténtial ∼ séntence E ∼ @

systémic ∼ sýstem E ∼ @

morálity ∼móral a ∼ @

symbólic ∼ sýmbol O ∼ @

atómic ∼ átom O ∼ @

harmónious∼ hármony @w∼ @

mystérious ∼mýstery I: ∼ @

dráma ∼ dramátic A: ∼ @

sulphúrious ∼ súlphur j0:∼ @

◮ opposition: full vowels, condition: unstressed syllable, output: /@/



laryngeal neutralization for word-final obstruents?

◮ beat – bead, back – bag, loose – lose, leaf – leave, etc.

◮ opposition: laryngeal contrast of obstruents,condition: word-final position,output: only voiceless-unaspirated obstruents

◮ based on this, beat and bead are supposed to be pronounced the sameway:

◮ BUT this does not seem to be the case!



“redundant” features to the help

◮ correlates of laryngeal contrast so far: voicing and aspiration

– but they are not active in word-final position

◮ however, there are other correlates of the laryngeal contrast

◮ they do not seem to play a role in other positions (e.g., word-medially),they are “redundant”

◮ but they seem to emerge more saliently when contrast is in danger (asin word-final position):

◮ relative length of preceding vowel◮ glottalization◮ other features: release noise, articulatory strength/e=ort/force


laryngeal contrast vowel length

relative length of preceding vowel

◮ experimental evidence: relative length of vowel to the length of theobstruent is an important factor for categorizing the obstruent as‘fortis/voiceless’ or ‘lenis/voiced’, especially in word-final position

◮ generally: short vowel + longer obstr. ⇒ ‘fortis/voiceless’long vowel + shorter obstr. ⇒ ‘lenis/voiced’



experiments: manipulating voicing & length



relation of voicing categorization and voicing

amount+vowel length in word-final position



conclusions

◮ if there is little or no voicing in the obstruent, the preceding vowel hasto be around twice as long as the obstruent for it to be categorized as‘lenis/voiced’

◮ if this vowel is not this long, the obstruent is categorized as‘fortis/voiceless’ = Pre-Fortis Clipping



Pre-Fortis Clipping in English in word-final position

◮ in English vowels (+ sonorants) are significantly shorter (clipped)before fortis obstruents than before lenis obstruents, where they aresignificantly longer

◮ Vowel (+ son.) is shorter Vowel (+ son.) is longerbefore fortis obstr. before lenis obstr.

mate maderope robewrite rideroot rudecap cab

speak speedloose loseleaf leavefont fond

dense dens



contrast in beat – bead



Phonetic implem.:w

w

�

w

w

�

unaspirated/∅VOT unaspirated/∅VOT+ shorter/clipped vowel + longer vowel

= [Vclippedt] = [Vlongerd˚

]or [Vlongert]


laryngeal contrast glottalization

pre-glottalization/glottal reinforcement

◮ glottal closure quickly closes down the voicing of the vowel, followedby the oral closure of the fortis stops & a=ricate

◮ happens word-finally or when they are followed by another consonant

◮ it only happens for the fortis consonants:mate [mEjPt] – made [mEjt],seat [sIjPt] – seed [sIjt]

◮ it is another indicator of the fortis – lenis contrast!

◮ it happens where the contrast between fortis & lenis stops couldpotentially disappear (word-finally)

◮ note: glottalization may well be just a more salient/forceful versionof pre-fortis clipping: the vowel is cut by glottal closure



contrast in beat – bead



Phonetic implem.:w

w

�

w

w

�

unaspirated/∅VOT unaspirated/∅VOT+ shorter/clipped vowel + longer vowel

+ glottalization + no glottalizaton= [VclippedPt] = [Vlongerd

˚]

or [Vlongert]



summary

phonetic implementation of [+fortis] in stops

Phonetic implementation In which position?

[+fortis] =⇒ (strong) aspiration word-initiallyor before a stressed vowel

=⇒ vowel clipping word-finally=⇒ glottalization word-finally=⇒ weak/no aspiration word-medially

before an unstressed vowel

phonetic implementation of [−fortis] in stops

Phonetic implementation In which position?

[−fortis] =⇒ voicing between vowels/son.=⇒ no aspiration, no voicing elsewhere


laryngeal contrast fricatives

correlates of laryngeal contrast for English fricatives

◮ examples for the laryngeal contrast of fricatives: thigh – thy, feel – veal,file – vile, sip – zip, leaf – leave, bus – buzz, etc.

◮ /T/ – /D/, /f/ – /v/, /s/ – /z/, /S/ – /Z/

◮ voicing/phonation, length di=erences (preceding vowel length, lengthof the fricative), and intensity signal the contrast

◮ aspiration or glottalizatione do not play a role



1. medial, between sonorants, before a stressed V

◮ conféction – convéction, defíed – divíde

◮ absence/presence of voicing/phonation is the primary cue, no dangerfor contrast

◮ /T f s S/: voiceless

◮ /D v z Z/: voiced/phonated



2. word-initial, before a V

◮ sip – zip, cellar – Zellar, fain – vein, fault – vault, feel – veal,sheet /SIjt/ – gite /ZIjt/, thigh /TAj/ – thy /DAj/

◮ the lenis fricatives in this position have lower intensity

◮ some research suggests that lenis fricatives /D v z Z/ in initialposition are relatively voiced, unlike lenis stops, the contrasts aboveare due to voicing/phonation (e.g., sip is voiceless, zip is voiced)

◮ /T f s S/: voiceless, higher intensity

◮ /D v z Z/: partially or fully voiceless, lower intensity



3. medial, between sonorants, after a stressed V

◮ míssle – mízzle, grístle – grízzle, rífle – ríval, Óphir /@wf@/ – óver /@wv@/,Áisha – Ásia, Ásher – ázure, tréssure – tréasure, Confúcian – confúsion

◮ absence/presence of voicing/phonation is the primary cue, no dangerfor contrast

◮ /T f s S/: voiceless

◮ /D v z Z/: voiced/phonated



4. absolute word-final position

◮ leaf – leave, brief – breve, calf – calve, safe – save, bus – buzz, race – raise,hiss – his, ruche /r0wS/ – rouge /r0wZ/, teeth /tIjT/ – teethe /tIjD/,loath – loathe

◮ for similar reasons as for stops, vocal fold vibration in this position isdi;cult to maintain

◮ relative vowel and consonant length emerge to maintain the contrast

◮ /T f s S/: have a shorter vowel before them and they are articulatedlonger with more intensity than

◮ /D v z Z/: preceding vowel is relatively longer and they are articulatedrelatively shorter and with less intensity


laryngeal contrast fric + stop

fortis fricative + stop clusters

◮ so far we have not seen neutralization of the laryngeal contrast foreither stops or fricatives

◮ fortis fricative + stop clusters:

◮ /s/ + stop: speak, sport, spring, stéreo, stúpid, string, school,scheme, sketch, discóver, displáy, expláin. . .

◮ /f/ + stop: caftán, fiftéen◮ /S/ + stop: gestált

◮ the laryngeal contrast is completely neutralized in this position:only an unvoiced-unaspirated stop may occur here



the traditional analysis: exception!

◮ spelling suggests that the stop in speak, discover, etc. is fortis, just likein peak, recover, etc.

◮ but then all stops after fortis fricatives and before a (stressed) vowelare exceptionally unaspirated

◮ Aspirated Not aspirated

péak spéakrecóver discóverattáin sustáinmátter áster

◮ why?



analysis 1: syllable structure

◮ ‘fortis stops are only aspirated at the beginning of a syllable’

◮ so: /p/ in peak is aspirated but not in speak, and /k/ in re.co.ver isaspirated but not in di.sco.ver (. signals syllable boundary)

◮ problem: putting the fricative and the stop in the same syllable isquestionable: a.fter, ca.ftan, fi.fteen, ge.stalt. . .



analysis 2: articulation of vocelessness

◮ ‘the fortisness/voicelessness of the fricative (open vocal folds)“expires” by the time we reach the end of the stop (= it cannot involvemore than two sounds)’

◮ so peak /phIjk/ (one sound) but not in speak /spIjk/

◮ problem: why does this not happen when there is only one sound?

◮ this idea implicitly suggests that the stop after the fricative is actuallynot fortis but lenis. . .



perception of stops after fortis fricatives

– what do native speakers hear when the /s/ of school is deleted?



stops after /s/ are perceived as lenis

– what do native speakers hear when the /s/ of school is deleted?



analysis 3: the stop is lenis, not fortis

◮ stops after fortis fricatives seem to be lenis and NOT fortis

◮ they are not aspirated because they are never aspirated

◮ they are not voiced either because they are only (passively) voicedbetween vowels/sonorants/other lenis obstruents but not next to fortisobstruents

◮ if transcription was following this phonological fact, then:stop = /sdOp/, discover = /dIsg@v@/, etc.

◮ why don’t we use this transcription then? – probably because it wouldbe very misleading for language learners coming from voicinglanguages. . ./sbo:t/, /s@sdEjn/, /asb@:g@z/. . .

◮ = sport, sustain, Asperger’s. . .

◮ note Welsh spelling: sbecto ‘spectacles’, sgyrt ‘skirt’, Sbaen ‘Spain’,sblasio ‘splash’. . .


RVA

regressive voicing assimilation (RVA) in Hungarian

◮ a phonotactic and morpho-phonological pattern: two adjacentobstruents must have the same voicing (compulsory rule)

◮ the voicelessness or voicing of the second obstruent governs thevoicelessness or voicing of the preceding obstruent (= “voicingspreads backward”)


RVA

regressive voicing assimilation (RVA) in Hungarian

◮ méz [z] ‘honey’ méz-ig [z] méz-tol [st] méz-bol [zb]‘to h.’ ‘from h.’ ‘of h.’

mész [s] ‘lime’ mész-ig [s] mész-tol [st] mész-bol [zb]‘to l.’ ‘from l.’ ‘of l.’

dob [b] ‘throw’ dob-ás [b] dob-tam [pt] dob-d [bd]‘throwing’ ‘I threw’ ‘you throw (imp.)’

kap [p] ‘get’ kap-ás [p] kap-tam [pt] kap-d [bd]‘getting’ ‘I got’ ‘you got (imp.)’

◮ notice that Hungarian is a voicing language, voicing spreads, makingvoiceless obstruents voiced; and voicelessness spreads the same way

◮ RVA is neutralizing: the voicing contrast between obstruentsdisappears


RVA

voicing assimilation can be modelled as feature-spreading

– mész+bol ‘of lime’ /s/+/b/ → /zb/:


RVA

spontaneously voiced sounds do not cause RVA

◮ mésznek ‘for the lime’ /s/ + /n/ → *[zn]

◮ képnek ‘for picture’ [pn] *[bn], töröknél [kn] *[gn] ‘at Turk(ish)’,zokni ‘socks’ [kn] *[gn]

◮ reason: sonorants are spontaneously voiced, not actively, this kind ofvoicing cannot spread to other sounds

◮ these sounds do NOT contain a [+voice] feature in their underlyingrepresentation that could spread

◮ in English too: batman [tm] *[dm], putney [tn] *[dn], replay [pl] *[bl]


RVA

passively voiced sounds do not cause RVA either

◮ since English is an aspirating language, neither fortis nor lenisobstruents have a phonologically active voiceless or voiced feature thatcould spread

◮ therefore we predict that English does not have RVA

◮ the pronunciation of words before obstruents does not change


RVA

no RVA from fortis /p t k T f s S Ù/ obstruents to lenis

obstruents in English

◮ is /z/ [z˚

] is Tom going? /zt/ [z˚

th]live /v/ [v

˚] live show /vS/ [v

˚S]

grade /d/ [d˚

] grade four /df/ [d˚

f]bead /d/ [d

˚] bead pack /dp/ [d

˚ph]

◮ the first words do not change at all, no spreading of voicelessness,hence no voicing neutralization either


RVA

no RVA from lenis /b d g D v z Z Ã/ to fortis obstruents

in English

◮ up /p/ [p] update /pd/ [pd˚

] (not *[bd]!)beat /t/ [t] beat band /tb/ [tb

˚] (not *[db]!)

beat /t/ [t] beat Zoë /tz/ [tz˚

] (not *[dz]!)birth /T/ [T] birthday /Td/ [Td

˚] (not *[Dd] or *[zd]!)

base /s/ [s] baseball /sb/ [sb˚

] (not *[zb]!)match /Ù/ [Ù] matchbox /Ùb/ [Ùb

˚] (not *[Ãb]!)

anec /k/ [k] anecdote /kd/ [kd˚

] (not *[gd]!)

◮ again, the first words do not change at all, no spreading of voicing,hence no voicing neutralization either


PVA?

progressive voicing assimilation?!

◮ traditional analysis of the regular past tense: 3 allomorphs:/t/, /d/, /Id/

◮ /Id/ after /t/ or /d/: vomited

◮ “voiceless” /t/ after “voiceless” sound: locked /kt/, laughed /ft/

◮ “voiced” /d/ after “voiced” sound: logged /gd/, carved /vd/

◮ “progressive voicing assimilation”

◮ but there is NO voicing assimilation in English!


PVA?

analysis of the regular past tensethere is just one past tense morpheme: /d/ (look at the spelling!)and 2 (not 3) allomorphs: /Id/ (after /t d/, and /d/ elsewhere:

Word in spelling: locked loggedUnderlying repr.: /k/+/d/ /g/+/d/

[+fortis]+[−fortis] [−fortis]+[−fortis]

Phonetic implem.:w

w

�

w

w

�

both: unasp.& voiceless both: unasp.& voiceless+ shorter/clipped vowel + longer vowel

= [Vclippedkd˚

] = [Vlongerg˚

d˚

]

= [Vclippedkt] = [Vlongerkt]

– same analysis for the regular plural/genitive/3Sg; do the analysis at homefor: locks vs. logs!


Documents

BBN ANG 243 Phonological analysis Laryngeal …seas3.elte.hu/analysis/analysis03-slides-h.pdflaryngeal contrast voicing types of voicing & devoicing based on aerodynamics 1. spontaneous(“modal”)