Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
The colon as a separate prosodic category:Tonal evidence from Paicî(Oceanic, New Caledonia)
Florian [email protected]
Princeton University
WCCFL 36, UCLA, 20-22 April 2018
1 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Introduction
Prosodic Hierarchy (PH, up to Prosodic Word):
Prosodic Word (ω) [{(σσ)Ft(σσ)Ft}κ]ω|
Foot (Ft) (σσ)Ft(σσ)Ft|
Syllable (σ) σσσσ|
Mora (µ)
2 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
PH + Colon = Constituent made of two feet: {Ft+ Ft}κ[Stowell 1979; Halle and Clements 1983: 18-19;Hammond 1987; Hayes 1995: 119; Green 1997; a.o.]
Prosodic Word (ω) [{(σσ)Ft(σσ)Ft}κ]ω|
Colon (κ) {(σσ)Ft(σσ)Ft}κ|
Foot (Ft) (σσ)Ft(σσ)Ft|
Syllable (σ) σσσσ|
Mora (µ)
3 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Justification for the colon (κ): mostly tertiary stress.“It is clear that in order to derive four levels of stress(primary, secondary, tertiary, unstressed), four levels ofstructure (prosodic word, colon, foot, syllable) are calledfor.” (Green 1997: 102)Example: Hungarian [Hammond 1987]
Pr.Word: [x ]Colon: {x }{x }Foot: (x .)(x .)(x .)(x )
kíʃ.kuːn.fèː.lɛɟ.hâː.zaː.bɔǹ‘in Kiskúnfélegyháza’
4 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Colon-based analysis proposed for a dozen languages sofar:
Passamaquoddy [Stowell 1979; Hayes 1995: 215-216;Green 1997: 104-109]Tiberian Hebrew [Dresher 1981]Garawa [Halle and Clements 1983: 20-21; Halle andVergnaud 1987: 43; Hayes 1995: 202]Hungarian [Hammond 1987; Hayes 1995: 330; Green1997: 102-104]Maithili [Hayes 1995: 149-162]Eastern Ojibwa [Hayes 1995: 216-218; Green1997: 109-112]Asheninca [Hayes 1995: 288-296; Green 1997: 112-114]Neo-Štokavian (Serbo-Croatian) [Green 1997: 115, 116]3 Goidelic varieties [Green 1997: 120-133]
5 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
The colon is controversial:Unnecessary in Garawa and Maithili [Green1997: 116-120]Empirical evidence contested in at least two cases:
Passamaquoddy [LeSourd 1993]Hungarian [Siptár and Törkenczy 2000: 21-22]
In general tertiary stress is controversialThus rejected by most recent metrical theories:
either implicitly [Elenbaas and Kager 1999; Hyde 2002]or explicitly [Martínez-Paricio 2012, 2013;Martínez-Paricio and Kager 2015, 2016].
6 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
In this paper:
Claim: we need to include the colon in the PH.
Evidence: tonal processes, including an unusual downsteppattern, attested in Paicî (Oceanic, New Caledonia).
This is the first piece of tonal evidence in favor of thecolon.
7 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Road map:
1 Downstep in Paicî
2 Colon-based Analysis
3 Evaluating the colon analysis and alternatives
4 Conclusion
8 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
1. Downstep in Paicî
9 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Paicî
Oceanic, New Caledonia, tonalAll data from Jean-Claude Rivierre’s research [Rivierre1974, 1983; Bensa and Rivierre 1994]Data confirmed in Dec. 2017 in Tchamba (NC) with nativespeaker Hélène Nimbaye
10 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Paicî: tonal inventory
Two contrastive tones: H vs. LRivierre Reanalysis
H HM LL ꜜL
TBU: moraMost lexical items are monotonal, i.e. either all H or all L.Monomorphemic lexical items are rarely longer than 5µ.
11 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Tonal processes
H tones are stable, not affected by any tonal processesL tones are targeted by two tonal processes:
Downstep(H-spread, as we will see later)
12 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Paicî: Downstep
1 ∼ 3µ words: no downstepµ ̀: ù ‘breath’ pwʌ̰̀ ‘turtle’µ ̀µ̀: nèè ‘name’ cʌ̰m̀ì ‘to plant’µ ̀µ̀µ̀: pwààì ‘tree sp.’ ùdʌr̀ɨ ̀ ‘to disjoin’
4µ+ words: downstep after 2nd mora:µ ̀µ̀ꜜµ̀µ̀: àùꜜkɔɔ̀̀ ▲ ))) ‘cagou’ pʌɟ̀àꜜɟìì ‘molar tooth’µ ̀µ̀ꜜµ̀µ̀µ ̀: ɛà̀ꜜàràbwà ‘crab sp.’ pwèrèꜜtɔɔ̀t̀ɨ ̀ ▲ ))) ‘wind’
13 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Paicî: Downstep
Lower register is then maintained throughout theutterance:
/è tèàpàà nàà̰̰ cáb̰à/ [e³ te³a³pa¹a¹ na¹̰a¹̰ ca³̰ba¹] ▲ )))(s)he arrive at Tchamba ⇓
[è tèàꜜpàà nàà̰̰ cáb̰à]
14 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Paicî: Downstep
Questions to solve:Why downstep after 2nd mora?
= initial bimoraic foot (µ ̀µ̀)Why no downstep after 2nd mora in 2 ∼ 3µ words, i.e. whydoes there need to be at least 4µ?
15 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
2. Analysis
16 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Hypothesis 1: foot-based
Hypothesis 1a: Downstep after initial L-toned bimoraicfootDoes not work:
Prediction Facts1µ: µ ̀ µ̀ pwʌ̰̀ !2µ: (µ̀µ̀)ꜜ µ̀µ̀ cʌ̰m̀ì *3µ: (µ̀µ̀)ꜜµ ̀ µ̀µ̀µ ̀ ùdʌr̀ɨ ̀ *4µ: (µ̀µ̀)ꜜ(µ ̀µ̀) µ̀µ̀ꜜµ ̀µ̀ pʌɟ̀àꜜɟìì !
17 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Hypothesis 1: foot-based
Hypothesis 1b: Downstep between first two L-tonedbimoraic feet (OCP)Works with data presented so far: Prediction Facts1µ: µ ̀ µ̀ pwʌ̰̀ !2µ: (µ̀µ̀) µ̀µ̀ cʌ̰m̀ì !3µ: (µ̀µ̀)µ ̀ µ̀µ̀µ ̀ ùdʌr̀ɨ ̀ !4µ: (µ̀µ̀)ꜜ(µ ̀µ̀) µ̀µ̀ꜜµ ̀µ̀ pʌɟ̀àꜜɟìì !
So far: no need for colon
18 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Problem: Juncture H-spread
Head+Complement phrasesGenitiveAspect + Verb (a.o.)Derivational pfx + Verb/Noun
Cv̀ head + L-toned complement: Cv+̀LL... → Cv+̀HL...H-tone morpheme between Hd and Cp, realized on Cp./rʌ̀ wàd̰ò/ → [rʌ̀ wàd̰ò] ▲ )))they drink ‘They drink.”/rʌ̀ pì- +H+ wàd̰ò/ → [rʌ̀ pì-wád̰ò] ▲ )))they mid- drink ‘They are getting drunk.”
19 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Problem: Juncture H-spread
1 ∼ 3µ: L→H only affects initial µ
1µ /pì-+H+cɔ/̀ [pì-cɔ]́ ‘move forward’2µ /pì-+H+wàd̰ò/ [pì-wád̰ò] ▲ ))) ‘get drunk’3µ /pì-+H+tʌm̀ʌ̰r̀i ̰/̀ [pì-tʌḿʌ̰r̀i ̰]̀ ‘give birth’
4µ+: L→H only affects first two µ
4µ /pì-+H+nʌ̰ɟ̀àìrì/ [pì-(nʌ̰ɟ́á)ìrì] ‘to curse’/pì-+H+tɔ̰ɔ̀ ̰ẁàr̰i ̰/̀ [pì-(tɔ̰ɔ́ ̰)́wàr̰i ̰]̀ ▲ ))) ‘reimburse’
Conclusion: Juncture-H spreads within foot: *(HL) →(HH)
20 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Problem: Juncture H-spread
Hypothesis 1 (foot-based): wrong predictionsPrediction Facts
H+1µ̀: µ́ µ ́ pì-cɔ́ !H+2µ̀: (µ ́µ́) µµ̀́ pì-wád̰ò *H+3µ̀: (µ ́µ́)µ̀ µµ̀́µ̀ pì-tʌḿʌ̰r̀i ̰̀ *H+ 4µ̀: (µ ́µ́)(µ ̀µ̀) µ ́µ́µ̀µ ̀ pì-tɔ̰ɔ́ ̰ẃàr̰i ̰̀ !
Conclusion:There is no foot parsing in 1 ∼ 3µ words:Foot parsing requires at least 4µ, i.e. at least two feet.
21 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Analysis (2): colon-basedSolution: Foot parsed only if dominated by a colon:
∗κ
∗Ft Ft
2 ∼ 3µ̀: *( µ̀ µ̀ ) µ̀
κ
Ft Ft
4µ ̀+: {( µ̀ µ̀ ) ( µ̀ µ̀ )}22 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Analysis (2): colon-based
Fully accounts for downstep AND Juncture H-spread:Prediction Facts
H+1µ̀: µ́ µ́ pì-cɔ́ !H+2µ̀: µ́µ ́ µ́µ̀ pì-wád̰ò !H+3µ̀: µ́µ ́µ̀ µ́µ̀µ ̀ pì-tʌḿʌ̰r̀i ̰̀ !H+ 4µ̀: {(µ ́µ́)(µ ̀µ̀)} µ́µ́µ ̀µ̀ pì-tɔ̰ɔ́ ̰ẃàr̰i ̰̀ !
23 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
OT account
*Ft ̸⊂ κ: penalizes any foot that is not licensed by a colon
*LFtLFt: two adjacent feet cannot both be L (= OCP)
24 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
OT account
25 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
OT account
26 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
3. Evaluation and alternatives
27 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Potential weaknesses
Weakness 1: Why make use of the colon if words are atmost 5 ∼ 6µ long?
Actually: PrWd = lexical item + all following tonalenclitics (functional words)
28 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
/tɔṕwɔ́ =boo =naa̰̰ =wee/→ [tɔṕwɔ́ =bóó =náá̰̰ =wéé]ω
put down at there
/gèè =mḛ =naa̰̰ =nḭ =boo/→ [{(gèè) =ꜜ(mḛ̀ =nà)̰}à̰ =ni ̰̀ =bòò]ω
cross to.here at here down
However: downstep limited to one per PrWd:
→ *[{(gèè) =ꜜ(mḛ̀ =nà)̰} {(à̰ =ni ̰)̀ =ꜜ(bòò)}]ω
29 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Potential weaknesses
Weakness 2: The colon is used only as a licensorIt is not targeted by any processIt does not itself directly trigger any process
Addition of a new prosodic category for very little work...
30 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 1: Binary Foot + extrametricality
Alternative 1a:Downstep after initial L-toned bimoraic foot (µ ̀µ̀)ꜜLast two µ are extrametrical
Prediction Facts1µ ̀: ⟨µ̀⟩ µ ̀ pwʌ̰̀ !2µ ̀: ⟨µ̀⟩⟨µ̀⟩ µ ̀µ̀ cʌ̰m̀ì !3µ ̀: µ ̀⟨µ̀⟩⟨µ̀⟩ µ ̀µ̀µ̀ ùdʌr̀ɨ ̀ !4µ ̀: (µ̀µ̀)ꜜ⟨µ̀⟩⟨µ ̀⟩ µ ̀µ̀ꜜµ̀µ ̀ pʌɟ̀àꜜɟìì !
Problem: only ONE extrametrical prosodic categoryallowed at the edge of a posodic domain
31 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 1: Binary Foot + extrametricality
Alternative 1b:Downstep after initial L-toned bimoraic foot (µ ̀µ̀)ꜜLast Foot is extrametrical
Prediction Facts1µ ̀: ⟨µ̀⟩ µ̀ pwʌ̰̀ !2µ ̀: ⟨µ̀µ̀⟩ µ̀µ̀ cʌ̰m̀ì !3µ ̀: µ ̀⟨µ̀µ ̀⟩ µ̀µ̀µ ̀ ùdʌr̀ɨ ̀ !4µ ̀: (µ̀µ̀)ꜜ⟨µ̀µ̀⟩ µ̀µ̀ꜜµ ̀µ̀ pʌɟ̀àꜜɟìì !
32 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 1: Binary Foot + extrametricality
Problem: contradiction!Downstep is triggered by a bimoraic footWhat we need extrametricality for: prevent the last twomorae from being parsed into a foot (which would feeddownstep)This is not extrametricality!
33 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 1: Binary Foot + extrametricality
Rescue attempt:final foot is not extrametricalbut simply stipulated not to count for the application ofdownstep
Still problematic: Implies existence of two different kindsof feet, parsed at two different edges of the PrWd:
word-final foot (inert)word-initial foot (causes downstep)
34 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 1: Binary Foot + extrametricality
Conclusion: Foot+extrametricality analysis ismore stipulativeunnecessarily complicatedexplanatorily weak
35 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 2: Recursive Foot parsing
The relevant consitutent is not a separate category, but atype of foot = superfoot (aka “layered foot”)Recursive foot-parsing: µµµ → (µµ)Ftµ → ((µµ)Ftµ)Ft[Selkirk 1980; Prince 1980; McCarthy 1982; Shih 1986;Martínez-Paricio 2012, 2013; Martínez-Paricio and Kager2015, 2016, a.o.]Slight disadvantage: violates the Strict Layer HypothesisAdvantage: does not add a new category.
36 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 2: Recursive Foot parsing
Prosodic Word (ω) [((σσ)Ft σ)Ft]ω| Superfoot ((σσ)Ft σ)Ft
Foot (Ft)| Foot (σσ)Ft σ
Syllable (σ) σσσ|
Mora (µ)
37 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Alternative 2: Recursive Foot parsing
Paicî downstep: double recursionµ ̀µ̀µ̀µ ̀ → (µ̀µ̀)µ̀µ̀ → ((µ̀µ̀)µ̀)µ̀ → (((µ ̀µ̀)µ̀)µ̀)
⇓(((µ̀µ̀)ꜜµ ̀)µ̀)
What is the generalization? Downstep occurs between a footand the next mora within the same superfoot, only if it isfollowed by another mora parsed within that same superfoot?
UnrestrictedNo evidence for the intermediate iteration (first recursion)StipulativeUnexplanatory
38 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Conclusion
39 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
Conclusion
Disadvantages of the colon analysis:Adds a new category in the Prosodic HierarchyThe colon does very little work in Paicî
Advantages:Both descriptively and explanatorily adequateRestrictive: parsing can only be conditional on thepresence of at most one adjacent foot.Beats all the alternatives (proposed so far)
Paicî might be the most solid argument to date in favor ofthe colon
40 / 41
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
...
.
Introduction Paicî Colon-based Analysis Alternatives Conclusion
THANK YOU!Special thanks to Hélène Nimbaye: olé!
(References upon request)
41 / 41
Downstep in PaicîColon-based AnalysisEvaluating the colon analysis and alternativesConclusion