Upload
others
View
4
Download
0
Embed Size (px)
Citation preview
Syllabification in English:
Contextualizing LOI and LOF
San Duanmu (端木三)
University of Michigan
March 3, 2010
City University of Hong Kong
1
Abstract
• A common assumption for locating syllable boundaries is the Onset requirement
• Onset in turn is based on the Law of Initials (LOI) and the Law of Finals (LOF).
• However, ambiguities remains under current formulations of LOI and LOF.
• I propose that LOI and LOF should be applied in context, i.e. in phrase-medial positions rather than in isolated words.
• The CLOI and CLOF minimize ambiguities in syllabification and offers better results than Onset.
2
Outline
1. Syllable boundaries and Onset
2. LOI and LOF
3. Problems with LOI and LOF
4. CLOI and CLOF
5. Additional ambiguities
6. CLOI and CLOF with Max Onset
7. CLOI and CLOF with VX and WSP
8. Evidence for WSP in CELEX
9. Conclusions
10. Further issues
3
1. Syllable boundaries and Onset
• Given a polysyllabic word, how do we know where
the medial syllable boundary is?
• Examples
city ci.ty, ci.ty
whisky whi,sky, whis.ky, whisk.y
aspect a.spect, as.pect, asp.ect
4
Onset (Maximal Onset)
• Onset (Pulgram 1970, Kahn 1976):
Put as many consonants in the onset as is allowed by the given language.
Word Analysis Reasonextra [ɛk][strə]
atlas [æt][ləs], not [æ][tləs] No word-initial [tl]
after [æf][tɚ], not [æ][ftɚ] No word-initial [ft]
singing [sɪŋ][ɪŋ], not [sɪ][ŋɪŋ] No word-initial [ŋ]
aspect [æs][pɛct], not [æ][spɛct] No word-final [æ]
5
Max Onset
• So Max Onset must consider whether the resulting
rime and onset are both good for the language
• Rime:
aspect [æs][pɛct], not [æ][spɛct] [æ] is not a rime
• Onset:
singing [sɪŋ][ɪŋ], not [sɪ][ŋɪŋ] [ŋ] is not an onset
• But how do we know whether a rime or onset is
good or not?
6
2. LOI and LOF (Vennemann 1988)
• The Law of Initials (LOI):
Word-medial onsets should resemble word-initial onsets.
• The Law of Finals (LOF):
Word-medial rimes should resemble word-final rimes.
7
Examples of LOI and LOF
Word Analysis Reason
atlas [æt][ləs]
*[æ][tləs] *LOI: [tl]; *LOF: [æ]
after [æf][tɚ]
*[æ][ftɚ] *LOI: [ft]; *LOF: [æ]
singing [sɪŋ][ɪŋ]
*[sɪ][ŋɪŋ] *LOI: [ŋ]
aspect [æs][pɛct]
*[æ][spɛct] *LOF: [æ]
8
3. Problems with LOI and LOF
• Some times LOI and LOF cannot both be satisfied,
or neither is satisfied (Blevins 2004).
• Analysis of lemon and city
[lɛ][mən] *LOF: [lɛ] (no word-final [ɛ])
[lɛm][ən] *LOI: [ən] (should be [ʔən])
[sɪɾ][i] *LOF: [sɪɾ]; *LOI: [i] (should be [ʔi])
[sɪ][ɾi] *LOF: [sɪ]; *LOI [ɾi]
9
Consequences
• Lack of clear speaker judgment on syllable boundaries in such words
– E.g.: lemon, city, happy, etc.
• Lack of agreement among linguists
– E.g. analyses of city
[sɪ[t]i] Kahn (1976); ambisyllabic
[sɪt][i] Selkirk (1982)
[sɪt][ti] Burzio (1994); geminate [tt]
[sɪ][ti] Jensen (2000)
10
4. CLOI and CLOF
• Consider Blevin’s analysis of lemon and city again [lɛ][mən] *LOF: [lɛ] (no word-final [ɛ])
[lɛm][ən] *LOI: [ən] (should be [ʔən])
[sɪɾ][i] *LOF: [sɪɾ]; *LOI: [i] (should be [ʔi])
[sɪ][ɾi] *LOF: [sɪ]; *LOI [ɾi]• Questions
– Is initial V always realized as [ʔV]?
– Can word-final [ɾ] ever occur?
11
Reconsidering the problems
• Initial V not as [ʔV]
get Ann [gɛɾ][æn]
*[gɛt][ʔæn] unless exaggerated
• Word-final [ɾ] get Ann [gɛɾ][æn]
it is [ɪɾ][ɪz]
• So, if we consider words in context, we need to reinterpret LOI and LOF, and reevaluate syllabification as a result.
12
13
Defining CLOI and CLOF
• Contextualized Law of Initials (CLOI):
Word-medial onsets should resemble phrase-
medial word-initial onsets.
• Contextualized Law of Finals (CLOF):
Word-medial codas should resemble phrase-
medial word-final codas.
14
Syllabification with CLOI and CLOF
whisky [wɪs][ki] miss, key
*[wɪ][ski] *CLOF: [ɪ]
atlas [æt][ləs] at, Las (Vegas)
*[æ][tləs] *CLOF: [æ]; LOI: [tl]
*[ætl][əs] *CLOF: [ætl]
lemon [lɛm][ən] gem, an
*[lɛ][mən] *CLOF: [ɛ]
15
Syllabification with CLOI and CLOF
city [sɪɾ][i] sit, east
*[sɪ][ɾi] *CLOF: [ɪ]; ??CLOI: [ɾ]
Can [ɾ] word initially?
go tomorrow ??[go][ɾəmɒro]
Summary
• Since medial syllables are adjacent, LOI and LOF
should not be applied to isolated words, but to
words in sequence.
• LOI and LOF should be contextualized.
• CLOI and CLOF can always be satisfied at the
same time.
• Well-formed syllabification is always possible.
16
5. Additional ambiguities
whisky [wɪs][ki] miss, key
[wɪsk][i] risk, even
tiny [tai][ni] tie, negotiate
[tain][i] fine, even
Canada [kæn][ə][də] can, a, Dakoda
[kæn][əd][ə] can, admit, ballad, a
Extra [ɛkst][rə] text, remote
[ɛks][trə] decks, tradition
[ɛk][strə] deck, strategic
17
6. CLOI and CLOF with Max Onset
whisky [wɪs][ki] CLOI, CLOF, Onset
*[wɪsk][i] CLOI, CLOF, *Onset
tiny [tai][ni] CLOI, CLOF, Onset
*[tain][i] CLOI, CLOF, *Onset
Canada [kæn][ə][də] CLOI, CLOF, Onset
*[kæn][əd][ə] CLOI, CLOF, *Onset
Extra [ɛk][strə] CLOI, CLOF, Onset
*[ɛkst][rə] CLOI, CLOF, *Onset
*[ɛks][trə] CLOI, CLOF, *Onset
18
Summary
• CLOI and CLOF are not enough to determine
syllabification
• Adding Max Onset can resolve the ambiguities
• However, there is certain overlap, because Max
Onset already assumes LOI and LOF.
19
7. CLOI and CLOF with VX and WSP
• What is VX?
• What is the WSP?
20
21
What is VX?
• VX is a limit on rime length.
• It claims that the maximal rime length is VX in
English , i.e. either VV or VC.
• It is based on the fact that most nonfinal rimes in
English are VX (Borowsky 1986, Duanmu 2009)
• VX is an independent fact that must be recognized
• VX cannot be derived from Onset, CLOI, or CLOF
22
Syllabification with CLOI, CLOF, and VX
whisky [wɪs][ki] CLOI, CLOF, VX
*[wɪsk][i] *VX: [ɪsk]
tiny [tai][ni] CLOI, CLOF, VX
*[tain][i] *VX: [ain]
23
Ambiguity?
• The word extra seems to render ambiguity under CLOI, CLOF, and WSP:
extra [ɛk][strə] CLOI, CLOF, VX
[ɛks][trə] [ɛks] is VX if [ks] = affricate
*[ɛkst][rə] *VX: [ɛkst]
• Wiese (1996): affricates do not have to have the same place of articulation
• However, additional evidence can help us choose between [ɛk][strə] and [ɛks][trə].
• For example, corpus data show that medial [sCC] onsets are rare, which support [ɛks][trə].
What is the WSP?
• The Weight-Stress Principle (WSP)
WPS-(a): Stressed syllables are heavy.
WSP-(b): Unstressed syllables are light.
• The WSP is an independent fact that must be
recognized
• The WSP cannot be derived from Onset, CLOI, or
CLOF
24
Syllable weight
• Rime and syllable weight in English
Rime Weight Example
VV heavy [bai] buy, [bi:] bee
VC heavy [bæn] ban, [bɛt] bet
V light [ðə] the, [ə] a
C light [n̩] ’n (reduced and)
25
Syllabification with CLOI, CLOF, and WSP
Canada [kæn][ə][də] CLOI, CLOF, WSP
*[kæn][əd][ə] *WSP: [əd]
26
Summary
• VX and WSP are independent principles that must
be recognized.
• CLOI, CLOF, VX, and WSP can all be satisfied at
the same time.
• When CLOI and CLOF are used with VX and WSP,
syllabification is unambiguous.
• There is no need for Onset.
27
8. Evidence for WSP in CELEX
• Evidence for VX have been discussed in Borowsky
(1989) and Duanmu (2009)
• Quantitative evidence for WSP in the English
lexicon remains to be shown.
28
CELEX
• Three lexicons
Lexicon Entries
Words 160,595
Lemmas 52,447
Simplex 7,401
• Syllabification: Max Onset without LOF
29
CELEX
30
20 abattoir 2 P '{-b@-tw#R [V][CV][CCVVC] [&][b@][twA:r*] U
21 abbe 1 P '{-b1 [V][CVV] [&][beI] M
22 abbess 2 P '{-bEs [V][CVC] [&][bEs] O
23 abbey 1 P '{-bI [V][CV] [&][bI] M
24 abbot 1 P '{-b@t [V][CVC] [&][b@t] M
25 abbreviate 2 P @-'bri-vI-1t [V][CCVV][CV][VVC] [@][bri:][vI][eIt] O
• Sample entries
Exceptions for WSP in CELEX
• Coding
S stressed
U unstressed
H heavy
L light
SL exception for WSP
SH predicted by WSP
UL predicted by WSP
UH exception for WSP
31
Extrametricality and final syllable
Stress Weight Example
S H must [mʌs]<t>
U L must [mə]<st>
S H some [sʌm]
U L some [sə]<m> or [sm ̩]S H a [ei]
U L a [ə]
• Therefore, final syllables show no violation of WSP
32
Stress-weight types
• SL (nonfinal stressed and light) lemon
• SH (nonfinal stressed and heavy) culprit
• WL (nonfinal unstressed and light) today
• UH (nonfinal unstressed and heavy) compete, ignite
• F (word final) that, lemon, emeritus
33
Exceptions for WSP in CELEX
• Data in polysyllabic simplex words (3,567 in all)Type Count All Nonfinal Same-stress
SL 1,402 17.3% 30.8% 43.2%
SH 1,846 22.8% 40.6% 56.8%
UL 1,010 12.4% 22.2% 77.8%
UH 288 3.5% 6.3% 22.2%
F 3,567 44.0%
All 8,113 100.0% 100.0%
Nonfinal SL and UH syllables 1,690
Among all syllables 21%
Among nonfinal syllables 37%
34
Re-examine exceptions to WSP
• Context of SL syllables (S is syllabic C)
Context Count Example (British English)
V][CCCV 10 mistress, ostrich
V][CCS 4 crystal
V][CCV 146 vista,
V][CS 245 heaven, apple
V][CV 989 any, happy
V][V 8
All 1,402
• General pattern
V][CX
35
Re-examine SL syllables
• The exception V][CX can be reanalyzed as VC][X
Context Count Example (British English)
VC][CCV 10 mistress, ostrich
VC][CS 4 crystal
VC][CV 146 vista,
VC][S 245 heaven, apple
VC][V 989 any, happy
• V][CX violates LOF (CLOF)
• VC][X observes CLOF and CLOI and has been proposed by many people, including those who assume Onset (e.g. Kahn 1976, Selkirk 1982, Hammond 1999).
36
Re-examine UH syllables
• Four cases of UH syllables
Case Count
Rime has a full V 156
Rime has [V:] 29
Rime has [əN] or [ɪN] 79 (N is a nasal)
Other 24
All 288
37
Re-examine UH syllables
• Case Analysis Example
Rime has a full V 2nd stress sardine, bamboo
Rime has [V:] 2nd stress duet
short V create, verbena
Rime has [əN] syllabic N combine
Rime has [ɪN] 2nd stress employ, invite
Other 24
All 288
38
Summary on WSP
• Most SL syllables are found in the context of V][CV,
where V is a stressed lax V
• According to CLOI and CLOF, V][CV should be
analyze as VC][X
• Most UH syllables can be accounted for; they either
have secondary stress or a syllabic C.
• Most apparent exceptions to WSP are not real.
39
9. Conclusions
• Current formulation of LOI and LOF often cannot be both satisfied
• Current use of Onset sacrifices LOF
• LOI and LOF should be contextualized
• Both CLOI and CLOF are satisfiable in every word
• VX and WSP are independently motivated and must be recognized
• CLOI, CLOF, VX, and WSP offer unambiguous syllabification
• Onset is entirely superfluous
40
10. Further issues
• Are there languages where Onset is not superfluous?
• Can VX be derived from WSP?
• Can CLOI and CLOF be derived from WSP?
• Can the proposed analysis account for allophonic variation, such as aspiration and flapping in American English?
• Can the proposed analysis account for the CV effect?
41
ReferencesBorowsky, Toni. 1986. Topics in the lexical phonology of English. Doctoral dissertation, University of Massachusetts,
Amherst.
Davis, Stuart. 2009. On the foot-based analysis of aspiration in American English. Paper presented at the CUNY Conference
on the Foot, City University of New York Graduate Center, New York, January 15-17.
Duanmu, San. 2009. Syllable structure: the limits of variation. Oxford: Oxford University Press.
Jensen, John. 2000. Against ambisyllabicity. Phonology 17: 187-235.
Kahn, Daniel. 1976. Syllable-based generalizations in English phonology. Doctoral dissertation, MIT, Cambridge, Mass.
Prince, Alan. 1990. Quantitative consequences of rhythmic organization. CLS 26, Papers from the 26th Regional Meeting of
the Chicago Linguistic Society Volume 2: The Parasession on the Syllable in Phonetics and Phonology, Chicago
Linguistic Society, 1992, 355-398.
Prokosch, Eduard. 1939. A comparative Germanic grammar. Philadelphia, PA: Linguistic Society of America.
Vennemann, Theo. 1988. Preference laws for syllable structure and the explanation of sound change. Berlin and New York:
Mouton de Gruyter.
Wiese, Richard. 1996. The phonology of German. Oxford and New York: Oxford University Press.
42
Syllabification and allophonic variation
at ease city potato
[æɾ.íz] [sɪɾ́i] [phә.théiɾo] Kahn (1976)
[æɾ.íz] [sɪɾ́.i] [phә.théiɾ.o] Borowsky (1986)
(æɾ).(íz) (sɪ.́ɾi) (phә.(théi.ɾo)) Jense, Davis
(æɾ).(íz) (sɪɾ́.i) [phә.(théi.ɾo)] WSP
• Summary: Syllabification under WSP can fully
account for allophonic rules.
43
44
The CV effect
• Jakobson (1958: 21): ‘There are languages lacking
syllables with initial vowels and/or syllables with
final consonants, but there are no languages devoid
of syllables with initial consonants or of syllables
with final vowels.’
• Hooper (1976: 199): CV syllables are ‘optimal’
universally
• Steriade (1982: 78): CV is ‘maximally unmarked’.
45
Explaining the CV effect
• Max Onset: VCV V.CV always, which explains
the CV effect
• CVX and WSP: There is no requirement for the
onset. How can we explain the CV effect?
46
Deriving CV effect from CVX and WSP
Two sources of CV syllables:
1. Unstressed rhyme is V, leaving following C as an onset
2. Stressed V can become VV, leaving following C as an onset
String Initial stress Final stress
CVCV [CVC][v], [CV:][cv] [cv][CV:]
CVCVC [CVC][v](c), [CV:][cv](c) [cv][CVC], [cv][CV:](c)
VCVC [VC][v](c), [V:][cv](c) [v][CVC], [v][CV:](c)
VCV [VC][v], [V:][cv] [v][CV:]