Class 2: Contrast and neutralization (part 1) Adam Albright - - PowerPoint PPT Presentation

Class 2: Contrast and neutralization (part 1)

Adam Albright (albright@mit.edu)

LSA 2017 Phonology University of Kentucky

Goals for today

▶ Address residual questions from last time (Constraints, Ranking,

OT)

▶ Recap: the function of phonology, illustrated ▶ Arguing for rankings ▶ Filling in some pieces

▶ Phonological distributions, linking distributions to rankings

▶ Warm-up problem: Navajo

▶ Grounding markedness constraints in perceptual considerations

Warm-up References 1/27

Navajo

ʔazit ‘liver’ kʰélã́ːt ‘tips of toes’ páːx ‘bread’ péːʃ ‘flint, metal’ tʃátí ‘antelope’ ʔaʒiʔ ‘torso’ xato ‘heat’ ɣótax ‘up above’ tʃeːʃóː ‘turkey buzzard’ lók’aːʔ ‘reed’ ɬáːɬ ‘to raise hell’ t’ĩ́ːɬ ‘to act, do’ tɬ’iʃ ‘to soak, dampen’ ʃaːʒ ‘knot’ sik’az ‘cold (object)’ tápãːx ‘shore’ tʰáláɣoʃ ‘soap’ ʔiːkex ‘wedding’ níɣiz ‘round and slender’ kax ‘rabbit’ k’at ‘now’ tʃeːɬit ‘soot’ ʔaʒí ‘voice’ ʔakot ‘knee’ ʔats’oːs ‘blood vessel’ t’átɬ’it ‘water moss’

Before vowels

Unaspirated, aspirated, ejective, affricates Labial, coronal, dorsal

Not before vowels: t, ʔ

Warm-up References 2/27

Navajo

ʔazit ‘liver’ kʰélã́ːt ‘tips of toes’ páːx ‘bread’ péːʃ ‘flint, metal’ tʃátí ‘antelope’ ʔaʒiʔ ‘torso’ xato ‘heat’ ɣótax ‘up above’ tʃeːʃóː ‘turkey buzzard’ lók’aːʔ ‘reed’ ɬáːɬ ‘to raise hell’ t’ĩ́ːɬ ‘to act, do’ tɬ’iʃ ‘to soak, dampen’ ʃaːʒ ‘knot’ sik’az ‘cold (object)’ tápãːx ‘shore’ tʰáláɣoʃ ‘soap’ ʔiːkex ‘wedding’ níɣiz ‘round and slender’ kax ‘rabbit’ k’at ‘now’ tʃeːɬit ‘soot’ ʔaʒí ‘voice’ ʔakot ‘knee’ ʔats’oːs ‘blood vessel’ t’átɬ’it ‘water moss’

▶ Before vowels

▶ Unaspirated, aspirated, ejective, affricates ▶ Labial, coronal, dorsal

▶ Not before vowels: t, ʔ

Warm-up References 2/27

Contextual neutralization

▶ Before vowels, stops may be:

▶ Unaspirated, aspirated, ejective, affricates ▶ Labial, coronal, dorsal, glottal

Contrast

▶ Not before vowels (i.e., word-finally), oral

stops must be:

▶ Unaspirated, and coronal or glottal

Neutralization

Neutralization = predictability

Segments with a stop component are predictably

▶ Coronal ▶ Not aspirated, not ejective ▶ Not affricates

…when not before a vowel

Warm-up References 3/27

Contextual neutralization: aspiration

▶ Markedness constraints

▶ *[+spread glottis]/

#: no aspirated segments word-finally

▶ *[+spread glottis]

▶ Faithfulness constraint

▶ Ident([±spread glottis])

/kʰat/ *[+s.g.]/ # Ident([±s.g.]) *[+s.g.]

☞

kʰat * kat *! /katʰ/ *[+s.g.]/ # Ident([±s.g.]) *[+s.g.] katʰ *! *

☞

kat *

▶ This ranking allows aspiration in the output when before a vowel,

and changes it to unaspirated otherwise

▶ Crucial:

▶ Ident([±s.g.]) ≫ *[+s.g.] ▶ *[±s.g.]/

# ≫ Ident([±s.g.])

Warm-up References 4/27

Formulating markedness constraints

▶ Constraint *[+s.g.]/

# suffices to capture the Navajo distribution, but why this constraint, and not some other?

▶ Why *[+s.g.], and not *[−s.g.]? ▶ Why does neutralization target word-final position?

▶ We could just stipulate: aspiration ([+s.g.]) is ‘marked’

▶ Markedness constraints penalize [+s.g.] segments, either in

general, or in specific contexts ▶ However, is not only unsatisfying, but also inadequate

▶ Doesn’t explain why aspiration is especially marked in final

position

▶ We’ll see below: value that’s ‘marked’ in some contexts may be

‘unmarked’ in others

The next example illustrates these points, with a restriction on voicing

Warm-up References 5/27

Review/background: manner of articulation

In increasing amount of airflow obstruction: Stops Obstruents Affricates Fricatives Nasals Sonorants Liquids Glides

Warm-up References 6/27

Review/background: manner of articulation

In increasing amount of airflow obstruction: Stops Obstruents Affricates Fricatives Nasals Sonorants Liquids Glides

Warm-up References 6/27

Review/background: manner of articulation

Bilab Lab Inter- Alv Alv- Retr Pal Vel Uvu Phar Glot dent Pal Stop p b t d k ɡ ʔ Fric Non-lat f v θ ð s z ʃ ʒ h ɦ Lat Affric tʃ dʒ Nas m ɱ n ŋ T ap/flap ɾ Approx Non-lat (w) ɹ j (w) Lat l

Warm-up References 7/27

Lithuanian voicing (Steriade 1997)

(Warning: data combines examples from Steriade and Lingea online dictionary)

▶ Initial singletons (i.e., before vowels)

▶ pulti ‘attack’, busti ‘wake up’

▶ Initial clusters (i.e., before sonorants and v)

▶ praba ‘hallmark’, brandus ‘mature’ ▶ platus ‘wide’, blaivus ‘sober’

sl, ʃl, ʒl, pl, bl, kl, ɡl sr, pr, br, tr, dr, kr, ɡr sm, ʃm, ʒm, sn, ʃn, ʒn, kn, ɡn sv, ʃv, ʒv, kv, ɡv, tv, dv

▶ Final:

▶ dauk ‘many, much’, kat ‘that’ ▶ *b#, *d#, *g#, *z#, *ʒ#

▶ Alternations: dauk ‘much (adj.)’ ∼ daug-elis ‘much (n.)’ ▶ Similar to distribution of aspiration in Navajo: neutralized /

#

▶ Heightens the suspicion that we may not want to leave contexts of

neutralization to coincidence

Warm-up References 8/27

Lithuanian voicing

▶ Constraints

▶ *

[ −sonorant +voice ]

# (=*D#) (contextual), *

[ −sonorant +voice ]

(general)

▶ Ident([±voice])

▶ Ranking for initial contrast, final neutralization

▶ Ident([±voice]) ≫ *

[ −sonorant +voice ]

▶ *

[ −sonorant +voice ]

# ≫ Ident(±voice) ▶ Devoicing instead of epenthesis or deletion: additional

faithfulness constraints

▶ Dep: don’t epenthesize ▶ Max: don’t delete ▶ Ranking for Lithuanian: Dep, Max ≫ Ident([±voice])

Warm-up References 9/27

Lithuanian voicing

Medial clusters: before sonorants and v Voiceless Voiced vikrùs ‘agile’ vedrus ‘glutton’ áukleː ‘nanny’ auglingas ‘fruitful’ nuknisti ‘swipe, steal’ dreːɡnas ‘damp brukneː ‘cowberry’ buːɡnas ‘drum’ riːtmetĩːs ‘morning’ bãdmetys ‘year of famine’ silpnas ‘weak’ skobnis ‘table’ akmuõ ‘stone’ augmuõ ‘growth’ àtminti ‘to remember’ liũdnas ‘sad’ asmuõ ‘person’ baʒnitʃʲa ‘church’ teʃmuõ ‘udder’ ʒʲeʒmuo (place name)

Warm-up References 10/27

Lithuanian voicing

Medial clusters: before obstruents Voiceless Voiced slaptas ‘secret’ — (*bt) aktas ‘act, deed’ — (*gt) kurapka ‘partridge’ — (*bk) agrastas ‘gooseberry’ — (*zt) aʃtrus ‘sharp’ — (*ʃd) — (*pd) stabdiːs ‘brake’ — (*kd) smaragdas ‘emerald’ — (*sd) barzda ‘beard’ — (*ʃd) iʒdas ‘exchequer’

Warm-up References 11/27

Lithuanian voicing

Alternations lap-kotis ‘petiole’ lap-as ‘leaf’ dirp-ti ‘work’ dirb-antiːsis ‘worker’ lak-ta ‘perch’ lak-ioti ‘fly around’ dek-ti ‘burn’ deg-inti ‘burn (trans.)’ at-auga ‘outgrowth’ ad-bulas ‘reversed’

▶ Voicing contrast is tolerated before vowels and sonorants,

neutralized before obstruents

▶ Outcome of neutralization depends on context: sometimes

voiceless, sometimes voiced

▶ Voiceless value is unmarked in many contexts (finally, before

voiceless) but marked before voiced obstruents

▶ Value that’s ‘marked’ in some contexts may be ‘unmarked’ in

• thers

Warm-up References 12/27

The typology of laryneal neutralization

Steriade (1997, p. 9) (O = obstruent, R = sonorant, including V)

# O, O # R O R # R R R T

• tontepec Mixe

no contrast no contrast no contrast no contrast contrast Lithuanian no contrast no contrast no contrast contrast contrast French no contrast no contrast contrast contrast contrast Shilha (T ašlḥiyt) no contrast contrast contrast contrast contrast Khasi contrast — contrast contrast contrast

▶ Asymmetry: contrast before/after obstruents implies contrast

before/after sonorants

Warm-up References 13/27

Licensing by cue

First step: differentiating contexts

▶ Why are contrasts for voicing, aspiration (etc.) neutralized more

readily in non-pre-sonorant position?

▶ Hypothesis: sonorants provide an advantageous context for

voicing and aspiration contrasts because they provide more/clearer perceptual cues

▶ Voicing cues (from Wright 2004, p. 39)

Warm-up References 14/27

Licensing by cue

▶ Cues, by context

Context Constraint Cues # O, O # (O O) *voice/[−son] [−son], closure voicing, closure dura- tion *voice/# [−son], *voice/[−son] # V [−son] *voice/V [−son] closure voicing, closure dura- tion, preceding V duration, F0 and F1 in preceding V V # *voice/V # closure voicing, closure dura- tion, preceding V duration, F0 and F1 in preceding V, burst duration and amplitude (if re- leased) V [+son] *voice/V [+son] closure voicing, closure dura- tion, preceding V duration, F0 and F1 in preceding V, burst duration and amplitude, F0 and F1 in following vowel

▶ Fixed ranking: *voi/[−son]

[−son] ≫ *voi/V [−son] ≫ *voi/V # ≫ *voi/V [+son]

Warm-up References 15/27

Analysis of Lithuanian

*voi/[−son] [−son] ≫ *voi/V [−son] ≫ *voi/V #

|

Ident([±voi])

|

*voi/V [+son]

▶ T

• p constraints all involve voicing before non-sonorants

▶ Shorthand: *voi/

¬[+son]

▶ We compare here the distribution before obstruents vs. before

Sonorants

▶ *voi/

¬[+son] ≫ *voi/

[+son] ▶ Neutralized word-finally ▶ Neutralized before obstruents

Warm-up References 16/27

Lithuanian: voicing contrast before sonorants

Contrast before sonorants

▶ Before vowels

/ta/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. ta b. da *! W * W /da/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. ta *! W L

☞

b. da *

▶ Before sonorant consonants

/tra/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. tra b. dra *! W * W /dra/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. tra *! W L

☞

b. dra *

Warm-up References 17/27

Lithuanian: voicing neutralization elsewhere

▶ Word-finally

/rat/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. rat b. rad *! W * W /rad/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. rat * b. rad *! W L

▶ Before obstruents

/ratka/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. ratka b. radka *! W * W /radka/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son]

☞

a. ratka * b. radka *! W L

Warm-up References 18/27

Voicing neutralization: [−voice] vs. [+voice]

▶ Analysis so far correctly characterizes the contexts that favor

neutralization, but it predicts that neutralization should also be to the voiceless value

▶ *voice/V

[−son] ▶ Recall that output actually depends on the following context

▶ Voiceless word-finally: /daug/ → [dauk] ▶ Voiceless before voiceless obstruent: /dirb-ti/ → [dirpti] ▶ Voiced before voiced obstruent: /at-bulas/ → [adbulas]

▶ Perhaps a higher-ranked constraint

▶ Agree([±voice]): adjacent obstruents must agree in voicing

▶ Steriade (1997): *voice means have no voicing value at all

(underspecification)

Warm-up References 19/27

Underspecified outputs

▶ If outputs are sets of feature values, then there’s no logical

necessity that a representation contain values for all features for all segments

▶ In fact, subsets of feature values are often assumed ▶ E.g., vowel features vs. consonant features

▶ Underspecified representations

Features t

  −continuant −nasal −voice  

d

  −continuant −nasal +voice  

t∅ voice

[ −continuant −nasal ]

▶ Proposal: *[αvoice] penalizes both [+voice] and [−voice], favors

candidates with no voicing specification

Warm-up References 20/27

Candidates with underspecification

/rat/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. rat *! W L b. rad *! W *

☞

c. rat∅voi * /rad/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. rat *! W * b. rad *! W L

☞

c. rat∅voi *

Warm-up References 21/27

Candidates with underspecification

/ratka/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. ratka *! W L * b. radka *! W * *

☞

c. rat∅voika * * d. rat∅voik∅voia ** W L /radka/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. ratka *! W * * b. radka *! W L *

☞

c. rat∅voika * * d. rat∅voik∅voia ** W L

Warm-up References 22/27

Candidates with underspecification

/ratga/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. ratga *! W L * b. radga *! W * *

☞

c. rat∅voiga * * d. rat∅voik∅voia ** W L /radga/ *voi/

¬[+son]

Ident([±voi]) *voi/[+son] a. ratga *! W * * b. radga *! W L *

☞

c. rat∅voiga * * d. rat∅voik∅voia ** W L

Warm-up References 23/27

Pronouncing underspecified outputs

▶ This ranking favors outputs with underspecification

▶ [rat∅voi], [rat∅voika], [rat∅voiga]

▶ However, phonetically these outputs sound voiceless or voiced

▶ [rat], [ratka], [radga]

▶ Steriade’s proposal: phonetic interpolation

▶ Normally, voicing (glottal vibration) ceases during stop closure,

unless special measures are taken to keep it going

▶ However, before a voiced obstruent, it is articulatorily simpler to

continue voicing through the closure than to let it stop and then restart it (fewer laryngeal gestures)

▶ Result: underspecified [∅voice] normally has no glottal vibrations,

except before another voiced obstruent1

1This logic is clearest between a voiced segment like a vowel and a voiced

• bstruent. In initial position (#t∅voiga), it’s less obvious that interpolation would favor

voicing throughout the stop closure. There are not many languages that allow such clusters to test this, but in Yiddish, regressive voicing assimilation does not normally affect word-initial obstruents: [kduʃə]/*[ɡduʃə] ‘holiness’, [pgam]/*[bgam] ‘blemish’.

Warm-up References 24/27

Richness of the Base (ROTB)

▶ This ranking eliminates voicing specifications in neutralizing

contexts in the output of the grammar

▶ In theory, there could also be input representations with

underspecification

▶ However, before sonorants, obstruents are consistently voiced or

voiceless (contrast)

▶ Eliminating outputs like [t∅voicea] requires an additional

constraint, and an additional ranking (what are they?)

Warm-up References 25/27

T aking stock

▶ The ‘licensing by cue’ hypothesis grounds the set of markedness

constraints in considerations of perception

▶ Pay-off: derive asymmetries (contrast in a perceptually

disadvantageous context implies contrast in perceptually more advantageous context)

▶ Interpretation of markedness as penalizing any specification

directly captures insight that the markedness eliminates contrasts (not specific feature values)

▶ Caveat: both of these hypotheses are controversial ▶ Alternative: find other ways to limit set of markedness

constraints, or simply allow a wide range of markedness constraints and seek typological explanations elsewhere (class 8)

Warm-up References 26/27

References

Steriade, D. (1997). Phonetics in phonology: The case of laryngeal neutralization. UCLA ms.

Warm-up References 27/27