Class 5: Faithfulness and alternations (part 2) Adam Albright - - PowerPoint PPT Presentation

Class 5: Faithfulness and alternations (part 2)

Adam Albright (albright@mit.edu)

LSA 2017 Phonology University of Kentucky

Announcements

▶ T

wo resources posted in the ‘background’

▶ If you are taking this class for credit…

▶ Option 1: assigment 1 feedback posted on Canvas, assignment 2

due next Monday by PDF on Canvas

▶ Option 2: finding a topic for the short squib

▶ T

• day: continue discussion of faithfulness constraints and

alternations

Perceptual asymmetries: The P-Map hypothesis References 1/40

Picking up from last time

Which consonant deletes in CC cluster simplification?

▶ Structural privilege

▶ Onset C’s, word-initial C’s ▶ Positional faithfulness = less likely to delete

▶ Perceptual privilege

▶ C/

V > C/ # > C/ C

▶ Further to the left = better cued, less likely to delete

▶ Continuing the argument for perceptual privilege

▶ Jóola: C1 deletion even when both consonants are codas ▶ Jóola: Deletion when both consonants are onsets ▶ More generally: explaining divergent final CC# outcomes across

languages

Perceptual asymmetries: The P-Map hypothesis References 2/40

Shizuoka Japanese emphatic adjectives

Adjective Emphatic Gloss a. hade hande ‘showy’ b.

• zoi
• nzoi

‘terrible’ c. jowai joɴwai ‘weak’ d. hajai haɴjai ‘fast’ e. kaɾai kaɴɾai ‘spicy’ f. nagai naŋgai ‘long’ g. kanaʃiː kanːaʃiː ‘sad’ h. amai amːai ‘sweet’ i. katai katːai ‘hard’ j.

• soi
• sːoi

‘slow’ k. takai takːai ‘high’ l. atsui atːsui ‘hot’ Adjective Emphatic Gloss m. kitanai kitːanai ‘dirty’ n. kusai kusːai ‘stinky’

• .

ikai ikːai ‘big’ p. zonzai zoːnzai ‘impolite’ q. kandaɾui kaːndaɾui ‘languid’ r.

• nzokutai
• ːnzokutai

‘ugly’ s. supːai suːpːai ‘sour’ t.

• kːanai
• ːkːanai

‘scary’ u.

• ͡ɪʃiː
• ͡ːɪʃiː

‘delicious’ v. kiːɾoi kiːɴɾoi ‘yellow’ w. toːtoi toːtːoi respectable

Perceptual asymmetries: The P-Map hypothesis References 3/40

Shizuoka Japanese emphatic adjectives

▶ Three strategies for emphatic adjective formation

▶ Lengthen consonant: kitanai → kitːanai ▶ Lengthen vowel: supːai → suːpːai ▶ Insert nasal C: hade → hande

▶ Choice of strategy is partially determined by general phonological

restrictions

▶ No voiced geminates: hade ̸→ hadːe ▶ No ‘overlong’ consonants: supːai ̸→ supːːai

▶ However, phonological restrictions underdetermine the choice

▶ Why kitanai → kitːanai, not *kiːtanai? ▶ Why hade → hande, not *haːde?

▶ Priority: certain changes favored over others

Perceptual asymmetries: The P-Map hypothesis References 4/40

OT is good at this!

• 1. Lengthen the consonant

▶ Unless it’s already a geminate or it’s not a possible geminate, in

which case…

• 2. Insert a nasal

▶ Unless there’s already NC or Cː, in which case…

• 3. Lengthen the vowel

These preferences follow from the ranking of the relevant markedness and faithfulness constraints. (Working this out is the purpose of assignment 2.)

Perceptual asymmetries: The P-Map hypothesis References 5/40

Maybe too good…

Another process we could easily have described

• 1. Lengthen the vowel

▶ Unless it’s already long, in which case…

• 2. Insert a nasal

▶ Unless there’s already NC or Cː, in which case…

• 3. Lengthen the consonant

(In actuality this never proves necessary in native words, because there are no VːN syllables)

Perceptual asymmetries: The P-Map hypothesis References 6/40

Shizuoka emphatic adjectives

▶ Generalization: lowest sonority change that’s consistent with

phonological requirements

▶ obstruent lengthening > nasal > vowel lengthening ▶ ‘Least conspicuous change’

▶ Is there anything significant about this?

Perceptual asymmetries: The P-Map hypothesis References 7/40

The general question

▶ Why do languages prefer one repair over another? ▶ A prior question: how do we know that they do?

▶ Alternations: compare morphologically related outputs

(allomorphs)

▶ German voicing alternations

Sg. Pl. Gloss diːp diːbə ‘thief’ liːt liːdɐ ‘song’ kriːk kriːɡə ‘war’ braːf braːvə ‘good’ (adj.) kraɪs kraɪzə ‘circle’ Sg. Pl. Gloss pʁɪntsiːp pʁɪntsiːpjən ‘principle’ gəbiːt ɡəbiːtə ‘area’ blɪk blɪkə ‘glance’ ʃaːf ʃaːfə ‘sheep’ ɡlaɪs ɡlaɪsə ‘track’

▶ Contrast: stem-final obstruent voicing is unpredictable before

vowels (contrastive)

▶ Contextual neutralization: only voiceless word-finally ▶ Loanword adaptation: compare source and loan form ▶ Cantonese (Silverman, 1992; Yip, 1993)

English Cantonese adaptation Gloss sæləd saː

55løt 35

salad kʰaːd kʰaːt

55

card

Perceptual asymmetries: The P-Map hypothesis References 8/40

The P-Map hypothesis (Steriade, 2001)

Conjecture: faithfulness favors perceptually indistinct changes

▶ Ident(big change) ≫ Ident(smaller change) ▶ Max(well cued consonant) ≫ Max(poorly cued consonant) ▶ Dep(well cued consonant) ≫ Dep(poorly cued consonant)

NB: it’s not actually clear whether we’d expect a change with morphological function, such as emphatic adjective formation, to favor less perceptually salient or more perceptually salient change (Löfstedt, 2010). We’ll side-step this question in the discussion that follows by focusing on phonologically motivated changes.

Perceptual asymmetries: The P-Map hypothesis References 9/40

Epenthesis

Recall the problem posed by epenthesis:

▶ We were interested in epenthetic consonants under the idea that

they might tell us something about the relative markedness of different place features, since Ident doesn’t differentiate /aɪ/ Ident(place) *#V *Dors *Cor *Glottal a. aɪ *! W L b. kaɪ *! W L c. taɪ *! W L

☞

d. ʔaɪ *

▶ Our hopes for a cross-linguistically valid ranking of general place

markedness constraints were dashed by discrepancies in inventories and contextual neutralizations

▶ However, epenthetic segments do show a greater degree of

consistency

Perceptual asymmetries: The P-Map hypothesis References 10/40

Initial epenthesis

German (Indo-European, Germany) /aʊf/ [ʔaʊf] ‘on’ /aɪn/ [ʔaɪn] ‘one’ /ɛkə/ [ˈʔɛkə] ‘corner’ /oft/ [ʔoft] ‘often’ /ideː/ [ʔiˈdeː] ‘idea’

▶ Epenthesis applies to all V-initial roots, even in the middle of a

word, which can make it hard to determine whether it’s underlying or epenthetic ([ˈʃnaps.ʔiˌdeː] ‘stupid idea that you have when you’ve drunk too much schnapps’)

▶ Epenthesis can depend on speech rate, position in phrase, and

vowel quality (see Pompino-Marschall and Żygis 2010)

▶ German also has [ʔ] epenthesis in VV sequences when V2 is

stressed ([ka.ˈʔo.tiʃ] ‘chaotic’), but not stressless ([ˈka.os] ‘chaos’)

▶ Cross-linguistically Blevins (2008): C epenthesis at phrase edges

is usually [ʔ], occasionally [h] (more often finally than initially)

Perceptual asymmetries: The P-Map hypothesis References 11/40

Favoring [ʔ]

▶ P-Map approach: ∆(V∼ʔV) < ∆(V∼pV, ∆(V∼tV)), ∆(V∼kV),

∆(V∼hV), ∆(V∼jV), etc.

▶ Stops at oral places introduce CV formant transitions, ʔ does not ▶ Glide introduces distinct formant target, large CV transitions to

vowels other than [i]

▶ Among glottals, ʔ has fewer internal cues than h (i.e., none), so

most like silence of word edge

▶ Faithfulness hierarchy

▶ Dep(oral stop)/#

V ≫ Dep(h)/# V ≫ Dep(ʔ)/# V

▶ Faithfulness constraints must care about quality of the segment

and context, since both determine perceptual distance to ∅

Perceptual asymmetries: The P-Map hypothesis References 12/40

Favoring [ʔ]

[pa] [ha] [ta] [ja] [ka] [ʔa]

Perceptual asymmetries: The P-Map hypothesis References 13/40

Intervocalic epenthesis

Lou (Eastern Admiralty Islands branch of Malayo-Polynesian, Papua New Guinea; data from Blust 1998, cited in Blevins, 2008)

▶ “Rising sonority” VV sequences repaired with epenthetic glides

▶ Rising sonority = higher V + lower V

/tia-n/ [tijan] ‘his/her abdomen’ /kea/ [keja] ‘swim’ /moloa-n/ [molowan] ‘his/her shadow/spirit’ /suep/ [suwep] ‘digging stick’

▶ “Falling sonority” VV sequences do not

▶ Falling sonority = lower V + higher V

/wei-n golom/ [weiŋgolom] ‘your saliva’ /mween/ [mwεεn] ‘man, male’ /kapeun/ [kaβeun] ‘bitter’

Perceptual asymmetries: The P-Map hypothesis References 14/40

Dep/V V

▶ Between identical V’s: no change in formants, [ʔ] is predicted to

be favored

▶ Higher V + lower V

▶ Intuitively: iV, uV transitions very similar to jV, wV ▶ All have rising F1 ▶ iV, jV: falling F2 (to varying degrees) ▶ uV, wV: rising F2, F3

▶ Rankings reflect relative perceptual similarity

▶ Dep(j)/Vi

Vi ≫ Dep(ʔ)/Vi Vi (except maybe i i, though *ji may block)

▶ Dep(ʔ)/i

V ≫ Dep(j)/i V ▶ Height of V1

▶ ∆(aV∼ajV) > ∆(eV∼ejV) > ∆(iV∼ijV) ▶ Large excursion > small excursion ▶ Asymmetrical implication: glide epenthesis in /ea/ implies

epenthesis in /ia/

Perceptual asymmetries: The P-Map hypothesis References 15/40

Epenthesis in a few other contexts

▶ NC, e.g., Latin

▶ /suːm-s-iː/ → [suːmpsiː] ‘take-prt-1sg’ ▶ /suːm-t-us/ → [suːmptus] ‘take-ppl’

▶ NL, e.g., in English

▶ Historically: thimble ▶ Synchronic variation: family [fæmli] ∼ [fæmbli]

▶ ‘Shielding’: /VNV/ → VndV, VdnV, VdndV ▶ ‘Excrescent stops’: epenthesis favors a stop at same place of

articulation as preceding nasal, and same voicing as following segment

▶ Voiced stops are not the least marked segments in any global

sense (i.e., can’t posit *ʔ ≫ *voiced stop)

▶ But they are the most similar to nasals

Perceptual asymmetries: The P-Map hypothesis References 16/40

The upshot

▶ Epenthesis in different contexts motivated by different

markedness constraints

▶ Epenthetic C depends on the contexts ▶ In all(?) cases, the C that is favored is the one that is most similar

to the surrounding segments

▶ Consistent with the P-Map hypothesis: faithfulness constraints

penalizing perceptually more salient changes ≫ those penalizing less salient changes

Perceptual asymmetries: The P-Map hypothesis References 17/40

Inverted deletions

‘Intrusive consonants’

▶ [r]: some dialects of British English, New England American

English, etc.

▶ He saw[r] it.

(Seen in various other languages, too)

▶ [l]: Midland American English

▶ ▷

He saw[l] it. ▶ [n]: Dutch dialects, Swiss German dialects (Ortmann, 1998)

These cases are different, and share a specific property: the ‘intrusive’ segment is illegal and deleted before consonants and pauses.

Perceptual asymmetries: The P-Map hypothesis References 18/40

Inverted deletions

▶ Consonants weaken (less/shorter constriction) when not before

vowels ⇒ “vocalization”

▶ Essentially, diphthongs: Vɹ̞, Vl̞ (lowered/more open)

▶ Coalescence with certain nuclei: əɹ̞ → ə, ɔl̞ → ɔː

▶ Consequence: /Vl/, /V/ surface identically unless prevocalic

▶ Reanalysis: insertion rather than deletion

Perceptual asymmetries: The P-Map hypothesis References 19/40

Inverted deletions

Where do ‘intrusive’ consonants come from, after reanalysis?

▶ Perhaps a reanalysis of underlying forms?

▶ /sɔ/ ⇒ /sɔɹ/ or /sɔl/ ▶ Need some way of ensuring that all relevant morphemes have

final ɹ/l (UR’s next time) ▶ Perhaps sets up an “unnatural” ranking Dep(glide)/V

V ≫ Dep(ɹ)/V V

Perceptual asymmetries: The P-Map hypothesis References 20/40

Tiene (Ellington, 1971; Hyman and Inkelas, 1997)

/dim-s-/ [diseb-], *[dinem-] ‘extinguish-caus’ /suɔmɔ-s-/ [sɔsɔb-], *[sɔnɔm-] ‘lend-caus’ /kam-l-/ [kanam-], *[kalab-] ‘be turned over-stat’

▶ Suffix is realized in the middle of the verb root (infixed) to satisfy

a constraint requiring CVC[cor]VC[non-cor]

▶ Nasal harmony requires that C2 and C3 agree in [±nasal] (nasal

harmony)…but which one changes?

▶ Stridents don’t nasalize ▶ Laterals readily nasalize

▶ Conjecture: follows from perceptual distance

▶ ∆(s∼n) > ∆(b∼m) > ∆(l∼n) ▶ Ident([±strident]) ≫ Ident([±son])/[−contin] ≫

Ident([±nas])/[+lateral] ▶ Alternative: Ident([±strident]) ≫ Ident([±nas])/Root ≫

Ident([±nas])

▶ This is more in line with Inkelas and Hyman’s analysis

Perceptual asymmetries: The P-Map hypothesis References 21/40

The ‘too many solutions’ problem (Steriade, 2001, 2008)

Reminder:

▶ Many languages lack laryngeal contrasts in final/non-presonorant

position

▶ German, Cantonese: aspiration contrast prevocalically, no

contrast finally

▶ Thai: voiced ∼ aspirated ∼ plain contrast prevocalically, only

aspirated ∼ plain before sonorant C, only plain finally

▶ Korean: aspiration and ‘tense’ contrast #

V and V V, no contrast finally ▶ Ranking: *

[ −son +voi ]

/̸ V ≫ at least one Faithfulness constraint

Perceptual asymmetries: The P-Map hypothesis References 22/40

The ‘too many solutions’ problem

▶ In many cases, this leads to alternations ▶ Recall German final devoicing Sg. Pl. Gloss diːp diːbə ‘thief’ liːt liːdɐ ‘song’ kriːk kriːɡə ‘war’ braːf braːvə ‘good’ (adj.) kraɪs kraɪzə ‘circle’ Sg. Pl. Gloss pʁɪntsiːp pʁɪntsiːpjən ‘principle’ gəbiːt ɡəbiːtə ‘area’ blɪk blɪkə ‘glance’ ʃaːf ʃaːfə ‘sheep’ ɡlaɪs ɡlaɪsə ‘track’ ▶ Ranking: Max(C), Dep(V), Ident([±F|F̸=voi) ≫ Ident([±voi])

Perceptual asymmetries: The P-Map hypothesis References 23/40

The ‘too many solutions’ problem

The observation

▶ Many languages have laryngeal alternations in response to

contextual laryngeal neutralization

▶ /tæb/ → [tæp]

German, Russian, Korean, Limbu

▶ Few languages (if any) eliminate final voiced obstruents through

nasalization, deletion, vowel epenthesis, etc.

▶ /tæb/ ̸→ [tæm], [tæw], [tæ], [tæbə], [dæp], etc.

▶ Surprising, because it would be easy to construct rankings that

would derive these patterns

▶ Ident([±voice]), Max(C), Dep(V) ≫ Ident([±nasal]) ▶ Ident([±voice]), Ident([±nasal]), Dep(V) ≫ Max(C) ▶ Ident([±voice]), Ident([±nasal]), Max(C) ≫ Dep(V)

Perceptual asymmetries: The P-Map hypothesis References 24/40

The P-Map solution

▶ Hypothesized asymmetry: ∆(Vb# ∼ Vm#) > ∆(Vb# ∼ Vp#)

▶ Nasality within C, but also coarticulatory nasalization of preceding

C ▶ Projects a partial ranking among Faithfulness constraints:

Ident([±nas]/V #) ≫ Ident([±voi]/V #)

Perceptual asymmetries: The P-Map hypothesis References 25/40

Evidence for similarity asymmetries

Devoicing vs. nasalization

▶ Imperfect rhymes

▶ Frequent: voicing mismatches (died ∼ light) ▶ Infrequent: nasality mismatches (mid ∼ sin)

▶ Judged similarity (Kawahara and Garvey, 2010)

▶ bV∼pV > bV∼mV ▶ b∼p even less distinct (→more similar?) /

#

▶ Extrapolation: expect same asymmetry word-finally

▶ Also potentially relevant: confusabilty

Perceptual asymmetries: The P-Map hypothesis References 26/40

Similarity is contextual

▶ Final devoicing vs. regressive nasalization

▶ ∆(Vb#∼Vm#) > ∆(Vb#∼Vp#) ▶ ∆(VbNV∼VpNV) > ∆(VbNV∼VmNV)

▶ Though a puzzle: colloquiual German (Duden; Steriade 1997)

▶ /sigmatiʃ/ → optional [sɪk.ma.tɪʃ], never *[sɪŋ.ma.tɪʃ]

▶ Steriade (2008) leaves open the possibility that some similarity

relations are not universally fixed

Perceptual asymmetries: The P-Map hypothesis References 27/40

Evidence from loanword adaptation

Broselow (2004)

▶ Mandarin lacks final obstruents altogether (regardless of voicing) ▶ Mandarin speakers learning English as a second language must

learn to tolerate final obstruents

▶ Intermediate stage: final voiceless stops, but not voiced ones

▶ Final voiced stops realized as voiceless

▶ T

wo pieces of ‘unmotivated’ knowledge

▶ *

[ +voice −sonorant ]

/ #

▶ Ident([±nasal]) ≫ Ident([±voice]) ▶ Neither(?) is necessary to capture any aspects of native Mandarin

phonology

Perceptual asymmetries: The P-Map hypothesis References 28/40

An Artificial Grammar experiment (Albright and Do, in prep)

▶ T

each participants a language in which stem-final stops alternate, sometimes by devoicing and sometimes by nasalization

▶ Devoicing: goʊp ∼ ɡoʊbi, pɹiːk ∼ pɹiːɡi ▶ Nasalization: kɹum ∼ kɹubi, tuŋ ∼ tuɡi

▶ Present equal numbers of devoicing and nasalization items ▶ T

est: novel stems ending in voiced stops

▶ Prediction: generalize devoicing more often

Perceptual asymmetries: The P-Map hypothesis References 29/40

Martian plurals

▶ Training: 54 pairs presented auditorily, and in ‘Martian

• rthography’

▶ Periodically orthography was not given, and participants were

required to enter what they had heard

▶ Items + fillers presented once per participant, in random order

▶ T

est phase: untrained words [ɡoʊp] [ɡoʊbi]

Perceptual asymmetries: The P-Map hypothesis References 30/40

Implicit vs. explicit tasks

▶ Goal: discourage task strategies that are unlike natural language

learning

▶ A common approach: explicit task to distract from pattern of

interest

▶ Explicit task

▶ Vowel harmony: -iː after front vowel, -uː after back vowels

▶ Implicit task

▶ Alternations involving word-final voiced stops (devoicing,

nasalization)

Perceptual asymmetries: The P-Map hypothesis References 31/40

Experimental design

Stem-final C Number Voiced stops devoice 6 nasalize 6 Voiceless stops 18 Nasals 18 Liquids 9

▶ Voiced stop items show alternations

▶ Trained on two places of articulation, tested on all three

(counterbalanced) ▶ Non-alternating voiceless stops and nasals show that there is no

phonotactic preference for devoicing vs. nasalization

▶ Harmony: equal numbers of front and back vowels in final syllable ▶ T

est items

▶ Harmony: 8 singulars (4 front V, 4 back V), forced choice for plural ▶ Stem alternations: 6 plurals with voiced stops, forced choice for

singular

Perceptual asymmetries: The P-Map hypothesis References 32/40

Participants

▶ 90 self-reported native speakers of English recruited from U.S. IP

addresses via Amazon Mechanical T urk, paid $1 for participation (median time: 22 mins)

Perceptual asymmetries: The P-Map hypothesis References 33/40

Results: Explicit task

▶ Participants strongly prefer correct harmony (72%) ▶ No significant differences depending on which places

trained/withheld

Perceptual asymmetries: The P-Map hypothesis References 34/40

Results: Implicit task

▶ T

ested voiced stops in word-final position These are nahzooboo. This is a .

▶ nahzoob (non-alternating) ▶ nahzoop (devoicing) ▶ nahzoom (nasalization)

▶ General preference for non-alternation

No Labial No Coronal No Dorsal Prob(alternation) 0.0 0.2 0.4 0.6 0.8 1.0 nasality voicing non-alternating

▶ In spite of the fact that all voiced stops alternated in training ▶ More on this later…

Perceptual asymmetries: The P-Map hypothesis References 35/40

Results: Implicit task

▶ T

ested voiced stops in word-final position These are nahzooboo. This is a .

▶ nahzoob (non-alternating) ▶ nahzoop (devoicing) ▶ nahzoom (nasalization)

▶ Among alternating responses: preference for devoicing

No Labial No Coronal No Dorsal Prob(alternation) 0.0 0.2 0.4 0.6 0.8 1.0 nasality voicing

▶ In spite of the fact that both processes were seen equally often in

training

Perceptual asymmetries: The P-Map hypothesis References 35/40

Results: Implicit task

▶ Mixed effects poisson regression, modeling count of response

types

▶ Non-alternating vs. alternating ▶ Nasality vs. voicing

▶ Significant preference for nonalternation (p(|z|) < .0001) ) ▶ Significant preference for devoicing over nasalization (p(|z|) <

.05)

▶ No significant differences depending on which place was withheld

in training

Perceptual asymmetries: The P-Map hypothesis References 36/40

An untrained preference

▶ Participants preferred final devoicing over final nasalization

▶ Both occurred equally often in the training data ▶ Neither occurs in English

▶ Evidence for a faithfulness ranking?

▶ Greater probability of devoicing could reflect higher ranking of

Ident([±voice])

▶ Preference is weaker than typological asymmetry ▶ However, participants received strong evidence for final

nasalization (eventually, evidence overcomes bias?)

▶ Conjecture: even small learning biases lead to large typological

asymmetries over time

Perceptual asymmetries: The P-Map hypothesis References 37/40

Summing up

▶ Various sources of evidence that speakers have systematic

preferences regarding how to repair ungrammatical structures

▶ Alternations, loanword adaptation, asymmetries in learning

▶ In OT, these preferences are encoded with rankings among

faithfulness constraints

▶ The P-Map hypothesis: these rankings respect or are guided by

considerations of perceptability

▶ Faith(more salient change) ≫ Faith(less salient change)

▶ We now know how to capture alternations

▶ One morpheme → different outputs, depending on context

But how do we decide on the representations of actual morphemes? Next time: underlying forms

Perceptual asymmetries: The P-Map hypothesis References 38/40

References

Blevins, J. (2008). Consonant epenthesis: natural and unnatural histories. In J. Good (Ed.), Language universals and language change, pp. 79–107. Oxford: Oxford University Press. Broselow, E. (2004). Unmarked structures and emergent rankings in second language

• phonology. International Journal of Bilingualism 8, 51–65.

Ellington, J. (1971). Aspects of the Tiene language. Ph. D. thesis, University of Wisconsin, Madison. Hyman, L. and S. Inkelas (1997). Emergent templates: The unusual case of Tiene. In

• V. Miglio and B. Morén (Eds.), Selected phonology papers from H-OT-97, Number 5

in University of Maryland Working Papers in Linguistics, pp. 92–116. Kawahara, S. and K. Garvey (2010). T esting the P-map hypothesis: Coda devoicing. Rutgers University ms. Löfstedt, I. (2010). Phonetic Effects in Swedish Phonology: Allomorphy and Paradigms.

• Ph. D. thesis, University of California, Los Angeles.

Ortmann, A. (1998). Consonant epenthesis: Its distribution and phonological

• specification. In W. Kehrein and R. Wiese (Eds.), Phonology and Morphology of the

Germanic Languages, Number 386 in Linguistische Arbeiten, pp. 51–76. Tübingen: Max Niemeyer.

Perceptual asymmetries: The P-Map hypothesis References 39/40

References

Silverman, D. (1992). Multiple scansions in loanword phonology: evidence from

• Cantonese. Phonology 9, 289–328.

Steriade, D. (2001). Directional asymmetries in place assimilation: a perceptual

• account. In E. Hume and K. Johnson (Eds.), Perception in Phonology, pp. 219–250.

Academic Press. Steriade, D. (2008). The phonology of perceptibility effects: the P-map and its consequences for constraint organization. In K. Hanson and S. Inkelas (Eds.), The Nature of the Word: Essays in Honor of Paul Kiparsky. MIT Press. Yip, M. (1993). Cantonese loanword phonology and optimality theory. Journal of East Asian Linguistics 2, 261–291.

Perceptual asymmetries: The P-Map hypothesis References 40/40