Transcript Syllabification of American English

Syllabification of American
English
David Eddington (BYU)
Dirk Elzinga (BYU)
Rebecca Treiman (WU)
Mark Davies (BYU)
Brigham Young University and Washington University
We are NOT interested in:
• Devising prescriptive rules of
syllabification
• How clever linguists syllabify words
• Syllabification of the written word
We ARE interested in:
• Describing how actual speakers syllabify
• Syllabification based on sounds not
spelling
Don’t we know how English is
syllabified already?
• Dictionaries often disagree in their phonetic
transcription (e.g. silly, hollow, balance)
– Merriam Webster
– Cambridge
si . lly
ho . llow
ba . lance
sill . y
ho . llow or holl . ow
bal . ance
Previous empirical studies
• (Derwing 1992; Derwing and Neary 1991; Fallows 1981; Treiman,
Bowey, and Bourassa 2002; Treiman and Danis 1988; Treiman,
Gross, and Cwikiel-Glavin 1992; Treiman, Staub, and Lavery 1994;
Treiman and Zukowski 1990; Zamuner and Ohala 1999)
• Focused principally on bisyllabic words with one medial
consonant (e.g. lemon)
• Considered mainly monomorphemic words such as
about not polymorphemic words such as lightly (light +
ly)
• Used only 20-40 test words
Findings of previous studies
• Stress attracts the medial consonant
– lémon > lém . on
– attáck > a . ttáck
Findings of previous studies
• Initial lax vowels attract the consonant to
the coda of the first syllable
– letter > lett . er
• Initial tense vowels repel the consonant
into the onset of the second syllable
– laser > la . ser
Findings of previous studies
Sonorants are attracted to the coda of the
first syllable
– silly > sill . y
Findings of previous studies
Sonorants are attracted to the coda of the
first syllable
– silly > sill . y
Obstruents are pushed into the onset of the
second syllable
– aghast > a . ghast
Problem
• Vowel quality, stress, and medial
consonants interact with each other and
must be considered together
• Other factors may influence as well
Our innovations
• We considered:
– Words with 1-4 medial consonants
– Both monomorphemic and polymorphemic
words
– The syllabification of many words
– Legality of consonants word initially and finally
– Types of morpheme boundaries
BYU Syllabification Survey
•
•
•
•
•
5,000 bisyllabic test items
841 participants
average of 22 responses per test word
survey completed online
choices based on sounds not spelling
BYU Syllabification Survey
• Choose the syllabification of ‘victim’ that
sounds best to you
– VI / KTUHM
– VIK / TUHM
– VIKT / UHM
– I’m not sure
BYU syllabification survey
Factors we considered:
– Vowel quality (tense or lax)
– Medial consonant quality (nasal, liquid, obstruent, etc.)
– Morphological boundaries
• Compound word (sheepdog)
• Transparent (actor)
• Opaque (fifteen)
– Stress placement
– Legality of consonants word initially and finally
BYU syllabification survey
• This is a work in progress
– We may consider other factors
– Statistical analysis still underway
(interactions)
Interpreting Results
C
O
N
S
T
A
V1 C1 C2 C3 V2
N
T
Interpreting Results
• Multiple regression performed
• Only p < .05 results discussed
• I’ll avoid numbers and focus on
interpretation of them instead
Results for VCV Words
• 60% of the variance is accounted for
• Factors that favor V . CV
–
–
–
–
C is licit word initially (/b/ versus /Ŋ/
V1 is tense (able)
C is an obstruent (placard)
A transparent or compound morpheme boundary
precedes the consonant (tC, cC; rebirth, lawsuit)
• Transparent >> compound
Results for VCV Words
• Factors that disfavor V . CV
– The word is written with a geminate consonant
– The consonant is /r/, /l/, or a nasal
– A transparent, compound, or opaque morpheme
boundary follows the consonant (Ct, Cc, Co)
• Compound >> transparent >> opaque
• (getup, noisy, shadow)
Results for VCV Words
• 60% of the variance is accounted for
• Factors that favor VC . V
– The word contains a geminate consonant
– C1 is /l/, /r/, or a nasal
– A compound, transparent, or opaque morpheme
boundary follows the consonant (Cc, Ct, Co)
• Compound >> transparent >> opaque
• (getup, noisy, shadow)
Results for VCV Words
• Factors that disfavor VC . V
– V1 is tense
– A transparent or compound morpheme boundary
precedes the consonant (tC, cC)
• Transparent >> compound
• (prewar, seashore)
Interactions
(monomorphemic)
Final Stress
V.CV VC.V
Lax
86
13
Tense 90
8
?
1
2
Example
buffoon
erect
Initial Stress
V.CV VC.V
62
37
86
13
?
1
1
Example
govern
private
Results for VCCV Words
• 75% of variance accounted for
• Factors that favor V . CCV syllabification
–
–
–
–
CC is licit word initially (br- versus *dl-)
C2 is /r/, /l/ or a glide
V1 is tense (council)
Any kind of morpheme boundary precedes CC (+CC)
• Compound >> transparent >> opaque
• (seaplane, biplane, supplant)
Results for VCCV Words
• Factors that disfavor V . CCV syllabification
– CC is licit word finally
– C1 is /r/, /l/, or nasal
– Any kind of morpheme boundary appears between
CC (C+C)
• Transparent >> compound >> opaque
• (subjoin, bloodshed, fifteen)
– A compound boundary appears after CC (CCc)
Results for VCCV Words
• 69% of variance accounted for
• Factors that favor VC . CV syllabification
– C1 is /r/, /l/, or nasal
– Any kind of morpheme boundary separates the CC
(C+C)
• Compound >> transparent >> opaque
• (subjoin, bloodshed, fifteen)
Results for VCCV Words
• Factors that disfavor VC . CV syllabification
– CC is licit word initially
– C2 is /r/, /l/, or a glide
– Any kind of morpheme boundary falls before CC
(+CC)
• Transparent >> compound >> opaque
– A compound or transparent morpheme boundary falls
after CC (CCc, CCo)
• Compound >> transparent
Results for VCCV Words
• 34% of variance accounted for
• Factors that favor VCC . V syllabification
– CC is licit word finally
– A compound or transparent morpheme boundary
follows CC (CCc, CCt)
• Compound >> transparent
• (handout, actor)
Results for VCCV Words
• Factors that disfavor VCC . V syllabification
– CC is licit word initially
– C1 is /r/
– A transparent or compound boundary separates CC
(CtC, CcC)
• Compound >> transparent
• (prewar, myself)
Interactions
(monomorphemic)
Final Stress
V.CCV VC.CV V.CCV ?
Example
Lax
Licit Onset Clusters
Illicit Onset Clusters
76
3
22
94
2
3
0
0
afflict
vaccine
Tense
Licit Onset Clusters
Illicit Onset Clusters
76
8
17
86
3
5
5
2
digress
maintain
Initial Stress
V.CCV VC.CV V.CCV ?
Example
Lax
Licit Onset Clusters
Illicit Onset Clusters
34
4
60
89
5
7
1
1
saffron
after
Tense
Licit Onset Clusters
Illicit Onset Clusters
74
10
23
84
1
5
1
1
program
flounder
Largest percentage in each category appears in boldface.
Syllabification by cluster
• Several ways to determine permissibility of CCC
– Are all three consonants licit word initially?
• str- is licit given strange
• rtr- is illicit *rtrange
– Are all three consonants licit word finally?
• -nst is licit given rinsed
• -lnm is illicit *talnm
– Are the final two consonants licit word initially?
• nfl- is licit given fling
• rtl- is illicit *tling
– Are the first two consonants licit word finally?
• -rkr is licit given work
• -bst is illicit *rabs [bs]
Results for VCCCV Words
• 70% of variance accounted for
• Factors that favor V . CCCV
– CCC is licit word initially (astray)
– V1 is tense (cambric)
– A compound or transparent boundary precedes CCC
(cCCC, tCCC)
• Compound >> transparent
• (drawstring, rescript)
Results for VCCCV Words
• Factors that disfavor V . CCCV
– The first two consonants of CCC are licit word finally
• (margrave, masthead)
– There is a transparent boundary (CtCC; nonstop)
Results of VCCCV Words
• 77% of the variance accounted for
• Factors that favor VC . CCV
– The last two consonants of CCC are licit word initially
• (coxcomb, contract)
– A compound or transparent boundary occurs CcCC or
CtCC
• Compound >> transparent
• (outgrow, nonstop)
Results of VCCCV Words
• Factors that disfavor VC . CCV
– CCC is licit word initially or finally
• (bowstring, curtsey)
– The first two consonants are licit word finally
• (talesman)
– The following morpheme boundaries
• cCCC, tCCC, CCcC, CCtC, CCCc
Results of VCCCV Words
• 81% of variance accounted for
• Factors that favor VCC . CV
– The first two consonants are licit word finally
• (darksome, sixty, empty, sanctum)
– The following morpheme boundaries exist:
• CCcC, CCtC, CCoC, and CCCc (??)
• (porthole, endless, children, huntsman
[hʌnsmʌn])
Results of VCCCV Words
• Factors that disfavor VCC . CV
– CCC is licit word initially
– Final two consonants are licit word initially
– The following morpheme boundaries:
• CcCC, CtCC, CCCo
• (outgrow, nonstop, centrist)
Results of VCCCV Words
• 17% of variance accounted for
• Factors that favor VCCC . V
– CCC is licit word finally (monster)
– CCCo morpheme boundary (centrist)
Results of VCCCV Words
• Factors that disfavor VCCC . V
– tCCC, CcCC, CCcC morpheme boundaries
• (rescript, outgrow, porthole)
Syllabification by cluster
• Responses tend to break up illicit clusters
– por . trait, chand . ler
• Licit CCC, CC C, and C CC clusters following a
tense vowel are kept intact
– pro . scribe, pro . scribe, poul . try
• Licit CCC, CC C, and C CC clusters following a
lax vowel favor C.CC even if it breaks a licit
cluster
– as . tray, res . t(au)rant, pil . grim
Results of VCCCCV Words
• 36% of the variance accounted for
• Factors that favor V . CCCCV
– NONE
Results of VCCCCV Words
• Factors that disfavor V . CCCCV
– The first three consonants are licit word finally
Results of VCCCCV Words
• 79% of the variance accounted for
• Factors that favor VC . CCCV
– The final three consonants are licit word initially
• (abstract)
Results of VCCCCV Words
• Factors that disfavor V . CCCCV
– A CCtCC morpheme boundary
• (transplant)
Results of VCCCCV Words
• 68% of the variance accounted for
• Factors that favor VCC . CCV
– NONE
Results of VCCCCV Words
• Factors that disfavor VCC . CCV
– The final three consonants are licit word initially
• (abstract)
– A CCCcC morpheme boundary
• (marksman, textbook)
Results of VCCCCV Words
• 83% of the variance accounted for
• Factors that favor VCCC . CV
– A CCCcC morpheme boundary
• (worldwide, swordsman)
Results of VCCCCV Words
• Factors that disfavor VCCC . CV
– The final two consonants are licit word initially
Result of VCCCCV Words
• No one syllabified any word VCCCC . V
Summary of Results
• Syllabification was highly influenced by:
–
–
–
–
–
–
Morpheme boundaries
Legality of consonant clusters word initially and finally
V1 quality
Medial consonant quality
Stress
Interactions among these factors