gemination as non-local lengthening anne pycha, uc berkeley
TRANSCRIPT
Gemination as non-local lengthening
Anne Pycha, UC Berkeley
Geminates
Phonology Phonetics C C C C
/t/ [t:] [rel = ]
• Complex segments with internal detail– Characterize what gemination “does”
Overview
• Problem: Release features…– Seem to play no role in length contrasts– Even though they “should”
• Phonetic study: Hungarian…– Source of lengthening comes from the right– Most likely to lengthen frication, but doesn’t
• Phonological problems, and possible solutions…– Affricate representations– Geminate representations
Overview
• Proposal: Gemination as morpheme strengthening– Degrees of fortification
– Degrees of lengthening
• Predictions– Cross-linguistic
– Hungarian-internal
Release features
• Problem (Part 1): Release features seem to play no role in length contrasts.
• Closure duration as primary perceptual cue to singleton-geminate contrast.– Lisker 1958: Swedish, Marathi, Telugu– Pickett & Decker 1968: English– Obrecht 1965: Arabic– Repp 1983: English
• Suggests diminished role for release.
Affricates
• Reasonable: Release features play bigger (or different) role when they are distinctive
• Example: Affricates, where release corresponds to frication
• Expectation: Frication portion of affricate might lengthen under gemination
Affricates
Shilluk (Eastern Sudanic, Sudan)“it should be clarified that the lengthening
[of t] is evidenced on the closure phase” (Gilley 1992: 27).
Anejom (Malayo-Polynesian, Vanuatu)“Geminate /t/ also occurs, with the stop
onset, but not the fricative release, being lengthened – thus [t:]” (Lynch 2000: 24)
Affricates
Attested: C C
T S
Unattested?: *C C
T S
Affricates
Problem (Part 2) Reasons to think frication should lengthen under gemination
1. Affricates can pattern like fricatives– Hungarian, Yucatec Maya
– Segmental status for frication (S)
– Fricative segments lengthen under gemination, so frication should too
Affricates
2. Perception of affricate does not require stop closure portion
– Fricatives: gradual rise in noise– Affricates: abrupt rise in noise– Noise alone suffices for affricate percept
• English listeners (Repp et al. 1978)• Hungarian listeners (Tarnóczy 1987)
– Suggests ‘independence’ of S
Affricates
3. Listeners appear to need it!– Pattani Malay: singleton/geminate contrast in
initial position– Abramson (1986 et seq.):
Listeners make length distinction in utterance-initial position
• True for all consonants, even voiceless stops where no apparent cues are present, as well as fricatives
• Exception: Affricates, at 50%• Why not lengthen the S?
Release features
Turkish (Lahiri & Hankamer 1988)
• Articulatory data: closure duration: significant VOT: significant
Phonetic study
• Goal: test reality of constraint on lengthened S within an affricate
• Context:– Affricates in geminate environment
– Source of gemination: rightmost (S) side
– Most likely to produce lengthened frication.
• Method: duration measurements.
Phonetic study
• Language: Hungarian
• Affricates: [ts, t, dz, d, ty, dy]
• Previous research– Magdics (1969)
– Szende (1974)
– Tarnóczy (1987)
Phonetic study
Affixal gemination in Hungarian
Root Instrumental
kert ‘hat’ kert-tel
piros ‘red’ piros-sal
baj ‘trouble’ baj-jal
ketrec ‘cage’ ketrec-cel
etc…
Phonetic study
Affixal singletons
Superessive case
[tat-on] ‘buckle-Sup’,
[va-on] ‘iron-Sup’
[kat-on] ‘fringe-Sup’
Affixal geminates
Instrumental case
[tat-tal] ‘buckle-Instr’,
[va-al] ‘iron-Instr’
[kat-tal] ‘fringe-Instr’
Stimuli (from Papp 1969)
• Noun roots ending in… – Affricates /t, ts/
– Corresponding obstruent /t/
– Corresponding sibilants /s, /– Stop-sibilant clusters /ps, p, ks, k/
• Monosyllabic roots: /kat/ ‘fringe’
• Disyllabic roots: /pamat/ ‘mop’
Stimuli
8 word shapes (CVC, CV:C, etc)
x 5 segment types /ts, t, t, s, /x 2 repetitions of each shape
x 3 speakers
= 240
189 noun roots
Stimuli: Clusters
All noun roots ending in clusters 11
Shapes:
CVCC /gips/
CVNCC /skunks/
C(C)VC(C)VCC /kyklops/
Data: Environments
Each noun root (n=200) in two different environments:
Intervocalic singleton: /kat-on/(Super)
Intervocalic geminate: /kat-Cal/ (Instr)
Results: Raw durations
0
50
100
150
200
250
T T: S S: TS T:S
All consonants
Results: Raw durations
0
20
40
60
80
100
120Within affricates
T T: S S:
Calculation: Ratio in disyllable
k a t o n
(Subject 4) t = 0.1 aton
= 0.1 aton
Results: Ratio in disyllable
0% 20% 40% 60% 80% 100%
Plain S
Plain T
Affricate S
Affricate T
Results: Ratio in disyllable
0% 20% 40% 60% 80% 100%
Plain S
Plain T
Affricate S
Affricate T
Discussion
Affricates under affixal gemination
• Duration of T changes
• Duration of S stays basically the same
…even in “rightmost” environment that (according to locality) should affect S
Constraint on lengthened S seems to be real
Discussion
Typical account:– Instrumental suffix has empty slot, /-Cal/
– Spreading fills C with features
* C V C - C V C
k a t a l– Locality problem
Rethinking affricates
• Re-think representation of affricates?
• Traditional representation is ordered:
C
T S
Rethinking affricates
Phonology: Unordered representation (Lombardi 1990) T
C
SPhonetics: Universal ordering = TS
Rethinking affricates
Evidence: “anti-edge effects” (Lombardi 1990)
• Sensitivity to T from right– Basque
– Turkish
• Sensitivity to S from the left– Yucatec Maya MSCs
– Hungarian
Rethinking affricates
• Gemination as an “anti-edge effect”?
• Source of lengthening: right (next to S)
• Target of lengthening: left (T)
….TS-al
• Problem: Gemination can target both T and S independently (not just T)
• Unordered representation doesn’t help
Rethinking affricates
C C
T
[rel = S]
Closure feature with dependent release
Rethinking affricates
Problem: we lose unity of behavior between affricates and fricatives
C versus C
T S
[rel = S]
Rethinking affricates
Root node spreadingX X
T SProblem: lost fact of lengthened T
Rethinking affricates
• No good solution for affricate representation
• Geminate representation: Is the C-slot the problem?
Rethinking geminates
• Alternative: suffix -al triggers strengthening in the root– Intuition: -al is “weak”
– Converse: Root is “strong”
– Suppose that: Strong-weak relationships are manifested during morpheme concatenation
– Manifestation is violable
Rethinking geminates
Strength relationship
Roots > Suffix -al
kat > al
Manifestation:Fortification, and/or
Lengthening
Rethinking geminates
• Multiple ways for roots to be fortified– Have stress (cf. Smith 2001)
– Segments have more stricture:• J• S• T
Rethinking geminates
• Multiple ways to for roots to be longer:– Have a mora (cf. Hayes 1992)
– Have a coda
– Have a longer segment
Analysis
Proposal for Hungarian• Length requirement for roots:
– Have a coda
• Strength requirement for roots:– Have coda = T (most stricture)
• Implemented as subcategorization frame[…VC]σ-al
T
Analysis
[…VC]σ-al
T
/lat-al/ lat.Cal lat.tal
Continuous syllabification to template (Itô 1986)
Analysis
Stricture requirement is violable:lat lat.talvas vas.sal *vat.salbaj baj.jal *bat.jal
Faith [stricture] >> T“Keep underlying stricture.”
Analysis
Stricture requirement is violable:
/n:-el/ *n:tel
Dep [stricture] >> T
“Do not insert stricture.”
Analysis
Stricture requirement becomes apparent…
S S Phonetics
/kaC-al/ kaC Cal kaC Cal
T T T S
Analysis
Clusters/gips-el/ gip.sel
Analysis
• Alignment: Requires morpheme and syllable edges to coincide– Simple segments (same)
– Affricates (unclear)[kat][al] kat.Xal *kat.al,
*kat.al
– Clusters (different) [gips][el] gips.Xel *gips.sel
Predictions
1. Morphology as determining factorRoots > Suffixes:Meithei (Tibeto-Burman, India)Acooli (Nilotic; Uganda)Ibibio (Eastern Sudanic; Nigeria)Hup and Yuhup (Maku; Brazil)Maithili (Indo-Iranian; India)Mokilese (Malayo-Polynesian, Micronesia)
Predictions
2. Preference for strong strictures“The presence of a geminate continuant
consonant in the segment inventory implies the presence of a corresponding non-continuant” (Kirchner 2001) Language 1: TT Language 2: TT, SS*Language 3: SS
Predictions
3. Gemination is one degree of lengthening
• Cross-linguistic evidenceThese (Nilotic, Sudan; Yip 2004)
à-kw ‘I plant’
-kw ‘you (sg) plant’
á-kw ‘I planted’
• Hungarian evidence
Predictions
/gips-el/ gip.sel• C-slot analysis:
– No gemination because *CCC– No root lengthening
• Lengthening analysis:– No gemination because σ templates satisfied– Degrees of lengthening could still occur– Target = [p]
Data: Clusters
Hungarian noun roots ending in clusters PS, KSPS KS/gips/ /skunks//tap/ /teks//naps/ /boks//mumps/ /vok//tritseps/ /uviks//kyklops/
Results: Ratio in disyllable
-20% 0% 20% 40% 60% 80% 100%
S
P
“Non-lengtheners”:2/3 of cluster tokens
Results: Ratio in disyllable
-20% 0% 20% 40% 60% 80% 100%
S
P
“Non-lengtheners”:2/3 of cluster tokens
Results: Ratio in disyllable
0% 20% 40% 60% 80% 100%
S
P
“Lengtheners”:1/3 of cluster tokens
Results: Ratio in disyllable
0% 20% 40% 60% 80% 100%
S
P
“Lengtheners”:1/3 of cluster tokens
Conclusions
• Problems for affricate representation remain (/t- t/ t:)
• Gemination as morpheme strengthening addresses locality problem in Hungarian
• Makes testable predictions – Cross-linguistic patterns of morpheme
combinations– Cross-linguistic patterns of preference for T over S– Gemination as a degree of lengthening