Class 4: Faithfulness and alternations (part 1) Adam Albright - - PowerPoint PPT Presentation

Class 4: Faithfulness and alternations (part 1)

Adam Albright (albright@mit.edu)

LSA 2017 Phonology University of Kentucky

Announcements

▶ Assignment 1 due today for Option 1 (by Canvas, if you haven’t

already turned it in)

▶ Assignment 2 posted, with some preliminary discussion to set it

up at the end of class today

▶ Last week: markedness asymmetries

▶ Why are contrasts disfavored for some segment types, or in some

contexts, more than others?

▶ Licensing by cue hypothesis: contrasts disfavored when lacking

cues ▶ Goal today: move on to discussion of faithfulness asymmetries

Morphological privilege Positional privilege Perceptual privilege References 1/34

Where we are

▶ The function of phonology: concentrate probability on particular

phonological outputs

▶ Unconditioned (marginalized over inputs): what strings are

grammatical in the language

▶ Conditioned on inputs: how should specific morphemes be

produced? ▶ Concentrating probability = eliminating outputs

▶ Markedness: eliminate outputs containing particular substrings ▶ Faithfulness: eliminate outputs deviating the input in particular

ways ▶ Markedness

▶ Usually: penalize feature values ▶ Alternative: penalize perceptually difficult contrasts

Morphological privilege Positional privilege Perceptual privilege References 2/34

T wo things left over from last time

▶ Further evidence for diverse “small place inventories” ▶ Small inventories in epenthetic consonants: why consistent

places? (ʔ, h, j, w) These illustrate, in different ways, motivations for a more articulated theory of Faithfulness constraints

Morphological privilege Positional privilege Perceptual privilege References 3/34

Sources of evidence for markedness

Reminder: last time we asked: where should we look to discover what markedness constraints on place contrasts penalize?

▶ Languages with limited sets of place contrasts

▶ English contrasts p,t,k (*ʔ) ▶ Seneca contrasts t,k (*p, *ʔ) ▶ Hawaiian has only k,ʔ (*p, *t)

▶ Within a language, place contrasts may be restricted in specific

contexts

▶ Phonological: Navajo lacks *p, *k in final position ▶ Phonological: epenthesis ▶ Morphological: restrictions in affixes, etc.

We already saw by comparing Seneca and Hawaiian that there is no universal fixed hierarchy of preferred places. The next example underscores that conclusion, using evidence from morphologically restricted place contrasts.

Morphological privilege Positional privilege Perceptual privilege References 4/34

Morphologically restricted contrast: Lakhota consonants

Three-way laryngeal contrast, four-way place contrast

▶ Stops and affricates: three-way contrast

Voiceless unasp. p, t, t͡ʃ, k ka ‘there’ ʃʼaka ‘strong’ Aspirated pʰ, tʰ, t͡ʃʰ, kʰ kʰa ‘to mean’ mãkʰa ‘earth’ Ejective pʼ, tʼ, t͡ʃʼ, kʼ k’a ‘to dig’ t͡ʃikʼala ‘small’

▶ Fricatives: similar three-way contrast

Voiceless s, ʃ, x, h xã ‘scab’ ixa ‘to laugh’ Voiced z, ʒ, ɣ ɣã ‘messy hair’ hoɣã ‘fish’ Ejective/glottalized sʼ, ʃʼ, xʼ xʼã ‘to do’ ptuxʼa ‘to crumble’

▶ Sonorants: m, n, l, w, j

Morphological privilege Positional privilege Perceptual privilege References 5/34

Morphologically restricted contrast: Lakhota consonants

Prefixes

▶ Subject, object marking: wa-, ja-, ũ(k)-, mã-, nĩ-, tʃi- ▶ Argument structure: wa-, ki-, ki-, kʰi- ▶ Locatives: a-, o-, i- ▶ Instrumental: ja-, wa-, wo-, ju-, pa-, ka-, na-, na-, (pu-)

Systematically missing:

▶ Aspirated and ejective stops ▶ Coronal stops

Morphological privilege Positional privilege Perceptual privilege References 6/34

Positional faithfulness

▶ ‘Privileged positions’ associated with special faithfulness

constraints

▶ Structural privilege: strong positions

▶ Morphological category ▶ Phonological position

▶ Or perceptual privilege: better acoustic cues

▶ Adjacent to V (transitions) > not adjacent to V ▶ Before V (C release) > after V

Morphological privilege Positional privilege Perceptual privilege References 7/34

Background fact: *VV

Languages frequently ban VV (vowel hiatus) (Casali, 1997)

▶ Sometimes satisfied by epenthesizing a consonant: VʔV, VjV, etc. ▶ Sometimes satisfied by combining features of the two vowels

(coalescence)

▶ Attic Greek

/zdɛː -omen/ → [zdɔːmen] live 1pl

→ ‘live-1pl.pres.subj’

/tiːma-omen/ → [tiːmɔːmen] honor1pl

→ ‘honor-1pl.pres.ind’

▶ More often: keep V2

▶ Lakhota

/wa- ijukpã/ → [wijukpã] indef.obj grinder → ‘coffee mill’ /tʰa + isto/

→ [tʰisto]

ruminant arm

→ ‘foreleg of ruminant’

Morphological privilege Positional privilege Perceptual privilege References 8/34

Elision of V1

Diola Fogny (Sapir, 1965, p. 13) (a.k.a. Jóola-Fóoñi, Kujamaat Jóola; Northern Atlantic, Senegal) /si+əw/ [səw] ‘house flies’ /si+uːk/ [suːk] ‘knuckles’ /bu+it/ [bit] ‘rice field’ /fu+ɛ/ [fɛ] ‘egg’ /mu+ɔf/ [mɔf] ‘earth’ /ka+et/ [ket] ‘palm leaf’ /ka+unɡund/ [kunɡund] ‘yam’

Morphological privilege Positional privilege Perceptual privilege References 9/34

Affix/root asymmetries

T urkish possessives (data from Hankamer 2010) at atɨm ‘horse’ karpuz karpuzum ‘watermelon’ kemik kemi(ɣ)im ‘bone’ gøl gølym ‘lake’ baba babam ‘father’ tʃene tʃenem ‘chin’ ketʃi ketʃim ‘goat’ byro byrom ‘office’

▶ 1sg possessive suffix is -im/-ɨm/-ym/-um

▶ Vowel agrees in backness and rounding with preceding (vowel

harmony) ▶ V1 retained, contrary to general cross-linguistic tendency ▶ Morphological privilege: V1 is a root vowel

Morphological privilege Positional privilege Perceptual privilege References 10/34

Morphological or phonological position?

▶ Examples of V1 deletion in Diola above are ambiguous!

▶ E.g., /ka+et/ → [ket] ‘palm leaf’ ▶ V2 is in second position, but also a root vowel

▶ At root+suffix boundaries, coalescence preserves some features

• f the root (Sapir, 1965, p. 15)

/sibe+as/ [sibəs] ‘the cows’ /eturu+ɛj/ [eturəj] ‘the grass’

▶ e,u are ‘tense’; a,ɛ are ‘lax’ ▶ Coalescence preserves tenseness of root vowel

▶ However, the fact that the root vowel isn’t simply preserved

faithfully (all features of root) shows the importance of V2

▶ Or, a prefix/suffix asymmetry, as in Lakhota?

Morphological privilege Positional privilege Perceptual privilege References 11/34

Affix/root asymmetries

Morphological privilege: roots vs. affixes

▶ Ident(place)/root ▶ Hypothesis: no corresponding Ident(place)/affix ▶ Predicts asymmetry: contrast in affixes implies contrast in roots

▶ Mostly correct, though there are interesting exceptions

▶ Lakhota ranking

▶ Ident(place)/root ≫ *[+coronal] ≫ Ident(place) ≫ *other ▶ Coronal more marked than other places? (Featural markedness) ▶ Or: coronal/X contrast more marked (Dispersion)

▶ Consequence: place contrasts are restricted outside the root ▶ A wrinkle not covered by this: suffixes show more contrasts than

prefixes in Lakhota (though still restricted)

▶ Though some items called ‘suffixes’ may be independent roots…

Morphological privilege Positional privilege Perceptual privilege References 12/34

Affix/root asymmetries

Other examples of roots vs. affix asymmetries

▶ Jakobson (1949): Czech inflectional suffixes are restricted to a set

• f 8 (out of about two dozen) consonants

▶ Arabic: affixes do not contain pharyngeal consonants (McCarthy

and Prince, 1995, 365)

▶ Yiddish: inflectional suffixes do not contain non-coronal

consonants

▶ See Beckman (1998, chap. 4) for many more examples

Morphological privilege Positional privilege Perceptual privilege References 13/34

Part of speech

Spanish

▶ Position of stress is contrastive in nouns, within certain limits

(three-syllable window at right edge of the word, weight effects) sáβana ‘sheet’ saβána ‘savannah’ káskaɾa ‘husk’ kaskáða ‘waterfall’ tóɾtola ‘dove’ toɾtúɣa ‘turtle’ bíspeɾa ‘eve’ espéɾa ‘wait’

▶ Position of stress is predictable in verbs

▶ Penultimate or final, depending on tense

Morphological privilege Positional privilege Perceptual privilege References 14/34

Part of speech

T

• kyo Japanese

▶ Nouns and verbs/adjectives may be accented or unaccented ▶ Location of accent in accented nouns can vary, within limits

(much like Spanish stress)

▶ Location of accent in accented verbs is predictable

Morphological privilege Positional privilege Perceptual privilege References 15/34

Morphologically privileged contexts

Shona, Kinande, and many other Bantu languages

▶ Nouns and verbs may both be toned or toneless ▶ T

• ned nouns bear wide range of tonal melodies

▶ T

• ned verbs have predictable melody, in a fixed position

See Smith (2011) for more examples, and discussion.

▶ Smith’s suggestion: Faith/Noun ≫ Faith

Morphological privilege Positional privilege Perceptual privilege References 16/34

Positional faithfulness Beckman (1998)

▶ Observation: some parts of the word tend to show a greater

range of contrasts than others

▶ Stressed syllables, onsets, word-initial position

▶ Implementation: contextual faithfulness constraints

▶ Specific and general variants

Positional: Max(C)/σ́ Input consonants in the stressed syllable should have correspondents in the out- put General: Max(C) Input consonants should have corre- spondents in the output

▶ Clusters in stressed syllables only: Max(C)/σ́ ≫ *[CC ≫ Max(C)

/ˈpraklo/ Max(C)/σ́ *Coda *[CC Max(C) a. ˈpra.klo **! b. ˈprak.lo *! *

☞

c. ˈpra.ko * * d. ˈpa.klo *! * * e. ˈpa.ko *! **

Morphological privilege Positional privilege Perceptual privilege References 17/34

Positional privilege: stress

▶ English, Russian, Catalan, Palauan, etc.

▶ Full set of vowel contrasts in stressed syllables, fewer in stressless

syllables (vowel reduction)

▶ Caveat: faithfulness in stressed syllables, or a markedness

constraint against vowel contrasts in stressless syllables? ▶ Guaraní

▶ Vowel nasality is contrastive in stressed syllables (and spreads) ▶ Vowel nasality is predictable in stressless syllables (spreading

from adjacent nasalized vowels, or a nasal consonant)

Morphological privilege Positional privilege Perceptual privilege References 18/34

Positional privilege: initiality

▶ !Xóõ

▶ Clicks are contrastive only in word-initial position

▶ Yiddish

▶ Stressless vowels are generally reduced to [ə], except in absolute

word-initial position, where reduction is blocked ▶ Ancient Greek

▶ “Smooth” vs. “rough” breathing (essentially, V vs. hV) contrastive

• nly in absolute initial position

Morphological privilege Positional privilege Perceptual privilege References 19/34

Positional privilege: initial syllables

Shona, as reported in (Beckman, 1998)

▶ Initial syllable may contain any height, rounding ▶ Non-initial syllables lack mid vowels, unless previous syllable also

has a mid vowel (harmony)

charuk- ‘jump over’ katuk- ‘flicker (flame)’ tandanis- ‘chase’ kwazis- ‘greet’ ganhur- ‘limit’ fungat- ‘embrace’ pfugam- ‘kneel’ ruram- ‘be straight,’ buruk- ‘dismount’ dukup- ‘to be small’ kumbir- ‘ask for’ turikir- ‘translate’ bvinar- ‘fade’ findam- ‘tangle (intr.)’ minaik- ‘wriggle’ simuk- ‘stand up’ simudz- ‘lift’ kwipur- ‘uproot’ tonhor- ‘be cold’ nonok- ‘dally, delay’ nonot- ‘scold, abuse’ korokod- ‘itch (nostril)’ gobor- ‘uproot’ bover- ‘collapse inwards’ kobodek- ‘become empty’ pofomadz- ‘blind (trans.)’ pofomar- ‘be blind’ chonjomar- ‘hunker’ chenjer- ‘be wise’ chember- ‘grow old’ verer- ‘move stealthily’ vereng- ‘read; count’ pember- ‘dance for joy’ nyemwerer- ‘smile’ zendam- ‘lean on’ chenam- ‘snarl’ svetuk- ‘jump’ serenuk- ‘water (gums)’

Morphological privilege Positional privilege Perceptual privilege References 20/34

Positional privilege: initial syllables

T urkish

▶ Vowels in initial syllables contrast for height, backness, and

rounding

▶ Non-initial high vowels harmonize in rounding with preceding

vowel; otherwise, no round vowels in non-initial syllables Note that examples here all involve root-initial syllables, not necessarily word-initial syllable

Morphological privilege Positional privilege Perceptual privilege References 21/34

Perceptually privileged positions

Diola Fogny (Sapir, 1965): the C2 advantage

▶ Obstruents, nasals, and liquids may occur before a vowel, or at

the end of a word

tɔ ‘there (precise)’ utɛk ‘you hit’ basɪːt ‘sorghum’ kəsiːt ‘feather’ idʒaut ‘I did not come’ nɔ ‘at that time’ adʒaŋa ‘girl’ bərun ‘type of antelope’ bʊrʊŋ ‘road’ bəsikən ‘mortar’ lɔb ‘talk!’ jɛlɪnd ‘shadow’ fal ‘river’ ɛɡaːl ‘to suffer’ ɛfɔl ‘frog’

▶ Before another consonant, only nasals may occur

ndaw (man’s name) mba ‘or’ ekuːmpə ‘type of dance’ dʒɛnsʊ ‘undershirt’ fanfaŋ ‘lots’ ekɔndɔr ‘neck’ kaband ‘shoulder’ kəɡuːmp ‘ashes’

Morphological privilege Positional privilege Perceptual privilege References 22/34

Perceptually privileged positions

Diola Fogny (Sapir, 1965): the C2 advantage

▶ CC clusters through morpheme concatenation

▶ V Epenthesis: /CC/ → CVC

/ʊmaŋʊt+dʒa/ ʊmaŋʊtʊdʒa ‘if you don’t want’ /fʊlɛŋ+fʊlɛŋ/ fʊlɛŋʊfʊlɛŋ ‘each month’

▶ C deletion: /CC/ → C

/lɛt+kʊ+dʒaw/ lɛkʊdʒaw ‘they won’t go’ /ɛkɛt bɔ/ ɛkɛbɔ ‘death there’ ▶ Choice of epenthesis vs. deletion determined partly by speech

rate (epenthesis across words in hyper-careful speech) and partly by morphological context (as above)

▶ Deletion favored for prefixal C’s and word-final C’s?

▶ Fact of interest here: when C deletion applies, it is always C1 that

deletes

Morphological privilege Positional privilege Perceptual privilege References 23/34

Perceptually privileged positions

Diola Fogny (Sapir, 1965): the C2 advantage /lɛt+kʊ+dʒaw/ lɛkʊdʒaw ‘they won’t go’ /ɛ+rɛnt+rent/ ɛrɛrɛnt ‘it is light’ /na+mandʒ+mandʒ/ namamandʒ ‘he knows’ /ɛkɛt bɔ/ ɛkɛbɔ ‘death there’ /ban ɲa/ baɲa ‘finish now’ /ʊdʒʊk+dʒa/ ʊdʒʊdʒa ‘if you see’ (T

• p group are reduced even in careful speech, as far as I can tell from

Sapir’s description; bottom example is a context for deletion only in casual/‘rapid’ speech)

▶ *CC would be satisfied by C1 or C2 deletion ▶ What favors C1 deletion?

Morphological privilege Positional privilege Perceptual privilege References 24/34

C1 deletion

/lɛt+kʊ+dʒaw/ *CC Max a. lɛtkʊdʒaw *! W

☞

b. lɛkʊdʒaw *

☞

c. lɛtʊdʒaw *

▶ Need a constraint that favors keeping C2 (ranking unimportant)

What constraint has the effect of Max(C1)?

Structural privilege: Max/Onset Perceptual privilege: Max/ V

Either would be sufficient for this tableau, but Max for better-cued segments captures a wider range of the Jóola facts

Morphological privilege Positional privilege Perceptual privilege References 25/34

C1 deletion

/lɛt+kʊ+dʒaw/ *CC Max(C1) Max a. lɛtkʊdʒaw *! W

☞

b. lɛkʊdʒaw * c. lɛtʊdʒaw *! W *

▶ Need a constraint that favors keeping C2 (ranking unimportant) ▶ What constraint has the effect of Max(C1)?

▶ Structural privilege: Max/Onset ▶ Perceptual privilege: Max/

V ▶ Either would be sufficient for this tableau, but Max for better-cued

segments captures a wider range of the Jóola facts

Morphological privilege Positional privilege Perceptual privilege References 25/34

C1 deletion in codas

Some more casual/‘rapid’ speech deletions

• laɲ+m
• laɲʊm ∼ -lam

‘return’

• badʒ+ul
• bədʒul ∼ -bal

‘have from’

▶ In this context, neither C1 nor C2 is in onset position

▶ Max/Ons doesn’t adequately distinguish

▶ A familiar perceptual asymmetry

▶ C’s may be released word-finally, but not in CC clusters ▶ Sapir (1965, p. 5): “In final position and before a pause it is

• ptionally unreleased” (i.e., p, t, k, m, n; infer that other stops are
• bligatorily released in final position?)

▶ Max/

V ≫ Max/ # ≫ Max/ C

Morphological privilege Positional privilege Perceptual privilege References 26/34

C1 deletion in onsets

A more subtle argument

▶ Jóola has morphemes/words that begin or end with NC

▶ mba ‘or’, ndaw (man’s name) ▶ famb ‘annoy’, kaband ‘shoulder’, -mandʒ ‘know’, aŋk ‘be hard’

▶ No morphemes begin or end with an obstruent-initial Clusters

▶ *kta, *kna, etc.

▶ Different rankings of *CC and Max/Ons predict different outcomes

▶ Max/Ons ≫ *CC incorrectly allows initial clusters ▶ *CC ≫ Max/Ons bans clusters, but predicts tie between outputs

with C1 and C2

Morphological privilege Positional privilege Perceptual privilege References 27/34

C1 deletion in onsets

Is indecision harmful? /ktɔ/ *CC Max/Ons Max a. ktɔ *! W

☞

b. tɔ * *

☞

c. kɔ * *

▶ This grammar allows for the interesting possibility of morphemes

that surface sometimes with one consonant, and someone with another consonant

▶ Underlying representation: CC cluster with both consonants

▶ Jóola has no such morphemes ▶ A Richness of the Base (ROTB) issue

Morphological privilege Positional privilege Perceptual privilege References 28/34

Faithfulness in perceptually strong positions

/ktɔ/ *CC Max/ V Max a. ktɔ *! W

☞

b. tɔ * c. kɔ *! W *

▶ Max/

V consistently favors preserving C2 in C1C2V

▶ Note that context here must be determined w.r.t. the input

▶ If there was a morpheme /ktɔ/, it would always yield the same

• utput as /tɔ/

▶ More restrictive grammar: concentrates probability on the types

• f morphemes that we actually observe

▶ We’ll come back to this in class 6

Morphological privilege Positional privilege Perceptual privilege References 29/34

C2 deletion in final CC clusters

▶ Jóola shows C1 deletion in final clusters

▶ /-laɲ+m/ → [-lam] ‘return’, /-badʒ+(u)l/ → [-bal] ‘have from’

▶ Conjecture: this is related to the fact that final C’s are released

(at least in careful speech)

▶ Sapir (1965) expressly mentions the subset of consonants that are

• ptionally unreleased in phrase-final position

▶ However, in many languages, final consonants are obligatorily

unreleased

▶ Lakhota, Thai, T

agalog, Korean, Caribbean English

▶ Without release, these final consonants rely on a preceding vowel

for cues to their presence and features

▶ Final VC1C2: 1 better cued than C2

Morphological privilege Positional privilege Perceptual privilege References 30/34

C1 deletion in languages lacking final release

▶ Jamaican English (Patrick, 1992)

rest res act ak kept kep front frʌn

▶ Gyeongsang Korean (Kim, 2002; complications suppressed)

/kaps/ kap ‘price’ /nəks/ nək ‘spirit’ /hɨlk/ hɨl ‘soil’ /talk/ tal ‘chicken’

Morphological privilege Positional privilege Perceptual privilege References 31/34

T aking stock

▶ So far, we have focused on faithfulness asymmetrics between

different positions in the string

▶ Strategy: identify positions that favor greater faithfulness

(morphological, phonological)

▶ We turn now to differences between segments: why are some

alternations preferred over others?

Morphological privilege Positional privilege Perceptual privilege References 32/34

References

Beckman, J. (1998). Positional Faithfulness. Ph. D. thesis, UMass. Casali, R. F . (1997). Vowel elision in hiatus contexts: Which vowel goes? Language 73, 493–533. Jakobson, R. (1949). L’aspect phonologique et l’aspect grammatical du langage dans leurs interrelations. Reprinted 1963 in Roman Jakobson. Essais de linguistique gé né rale. Kim, Y . (2002). Coda cluster simplification and its interactions with other coda processes in Korean. Working Papers of the Cornell Phonetics Laboratory 14, 82–112. McCarthy, J. and A. Prince (1995). Faithfulness and reduplicative identity. In

• J. Beckman, S. Urbanczyk, and L. W. Dickey (Eds.), University of Massachusetts

Occasional Papers in Linguistics [UMOP] 18: Papers in Optimality Theory, pp. 249–384. Amherst, MA: GLSA. ROA 60. Patrick, P . L. (1992). Linguistic variation in urban Jamaican creole: A sociolinguistic study of Kingston, Jamaica. Ph. D. thesis, University of Pennsylvania. Sapir, J. D. (1965). A grammar of Diola-Fogny: a language spoken in the Basse-Casamance region of Senegal. Cambridge University Press.

Morphological privilege Positional privilege Perceptual privilege References 33/34

References

Smith, J. (2011). Category-specific effects. In M. van Oostendorp, C. J. Ewen, E. Hume, and K. Rice (Eds.), The Blackwell Companion to Phonology, Volume 4. Malden, MA: Wiley-Blackwell.

Morphological privilege Positional privilege Perceptual privilege References 34/34