1. No category

IASCL 2005
Prosodic and frequency effects
on the development of
syllable structure
in European Portuguese
S. Frota*, M. J. Freitas*, M. Vigário+ & F. Martins^
*Universidade de Lisboa – FL and ONSET-CEL,
+
Universidade do Minho – ILCH,
^ Universidade de Lisboa – FL and ILTEC
1. Introduction
• Background
– Acquisition of syllable structure in EP (overview)
– The impact of frequency on language development
– Frequency studies for EP
• Goals: establish the effects of syllable type frequency and the
role of prosodic prominence in syllable development in EP
•
•
•
•
Data: input (ADS & CDS), and child data
Input frequencies and syllable types in child speech
Prosodic prominence effects: word edges, stress
Summary of main findings and discussion
2. Background: Acquisition of syllable structure
• Onsets (Freitas 1997):
C and Ø > CC
-
non-branching onsets (C or Ø)
are represented in the system
from the beginning of production
- empty onsets are used as a
default structure to deal with
problematic target onsets
- G may replace a segmentally
problematic C target
- branching onsets are the last
structure to be acquired
CV, V, ØV, GV > CC
• Rhymes V > (surface VG>)
-
VCfric > VCliq / VG
non-branching nuclei are, as
expected, represented in the system
from the beginning of production
surface VG structures are present
since early stages of production, but
the mastery of branching nuclei
occurs in late stages of development,
by the time syllable-final liquids are
acquired; V and VG seem to be
interpreted similarly by children
(Correia, 2004)
V, VG > VC
2. Background: Acquisition of syllable structure
•
Children early productions (examples)
a.
pato
/'pa.tu/
dá
/'da/
quer
/'kER/
b.
água
/'a.gwå/
é
/'E/
água
/'a.gwå/
c.
vês
/'veS/
flor
/'flor/
d.
quer
/'kER/
papel
/på.'pE¬/
e.
pato
/'pa.tu/
chupeta /Su.'pe.tå/→
dá
/'da/
cão
/'kå)w)/
→
['tå]
(João: 0;11.6)
‘duck’
→
['da]
(Inês: 0;11.14)
‘give’
→
['ke]
(Marta: 1;2.0)
‘wants’
→
['a.Bå]
(João: 0;11.6)
‘water’
→
['E]
(Inês: 1;0.25)
‘is’
→
['a.wå]
(Marta: 1;2.0)
‘water’
→
['eS]
(Marta: 1;3.8)
‘see’
→
['olˆ]
(Inês: 1;9.19)
‘flower’
→
['kE.Rˆ]/['kE.Ri]
(Inês: 1;10.29)
‘wants’
→
[på.'pE.lˆ]
(Marta: 2;2.17)
‘paper’
→
[å.'tå]/['tå]
(João: 0;11.06)
‘duck’
[å.'pi]/['pi]
(Inês: 1;01.30)
‘pacifier’
→
[å.'da]/['da]
(Inês: 1;0.25)
‘give (me)’
→
[å.'kå)w)]/['kå)u]
(Marta: 1;02.0)
‘dog’
2. Background: the impact of frequency
•
There is a growing interest in
determining the importance of
frequency information in linguistic
behaviour
•
Vigário & Falé, 1994; Viana et al.,
1996); phonetic segments and stress
patterns (Viana et al., 1996); words
(Bybee, 2000 & 2001; Bybee & Hooper,
2001; Jurafsky, Bell & Girand, 2002;
Moates, Bond & Stockmal, 2002;
Pierrehumbert, 2002)
•
The same is true for the acquisition
and development of linguistic
systems
(Fikkert & Freitas, 1998; Lleó &
Demuth, 1999; Beckman e Edwards,
2000; Roark & Demuth, 2000; Demuth
& Johnson, 2003; Prieto, 2004; Vigário,
Freitas & Frota 2005)
Frequency in EP (adult speech):
syllable types (Andrade & Viana, 1994;
undergoing phonetic reduction
(Vigário, 2003); minimal words and
patterns of cliticization (Vigário,
Martins & Frota, 2005)
•
Frequency in EP (child speech):
development of syllable structure
(Fikkert & Freitas; Vigário, Freitas &
Frota, 2003); development of word
shapes (Vigário, Freitas & Frota, 2005)
2. Background: the frequency tool FreP
• FreP (Vigário, Martins & Frota, 2005)
-
-
Automatically extracts syllable types
from written texts;
Semi-phonological: includes lexical
phonological processes (e.g. glide
insertion to break a hiatus); ignores
all optional processes (deletion of
unstressed vowels, optional gliding);
Includes glides in rising diphthongs
that are obligatory (post-tonic in
proparoxitons) and V positions
between consonants that violate
syllable construction principles
(Mateus & Andrade 2000);
-
Glides between vowels are treated as
ambisyllabic
• Goals
-
Establish the EP-specific frequency
distributions of syllable types in the
input
- Measure the effect of input
frequencies on syllable type
development
- Assess the role of prosodic
prominence:
. position in the word:
word-edges/monosylw > word internal
. stressed > unstressed
- Discuss the impact of frequency on
development
3. Data (number of syllables by syllable types)
¾
Spontaneous data from 3 monolingual Portuguese
children (CS): João aged 0;10.2 – 1;8.13
(n=1003)
(Freitas 1997)
Inês
Marta
aged 0;11.14 - 1;10.29
aged 1;2.0 - 2;2.17
¾
Child-directed speech data (CDS):
¾
Spontaneous Adult data (ADS):
Inês
Marta
(n=3619)
(n=6090)
(n=24867)
(n= 10985)
the Português Falado corpus (CLUL – CDRom) from 90s (n=41826)
(Vigário, Martins & Frota 2005)
Data obtained with FreP (Reliability was above 99.5%)
V
G
V
30
VG
N
G
VC
C
VN
C
VG
C
C
VC
C
C
C
V
C
C
VN
C
C
VG
N
C
VG
N
C
VN
C
C
C
VG
C
C
VC
C
G
VC
C
C
G
V
G
C
V
C C
C
VN
C
40
VC
45
G
V
G
VG
N
C
VG
C
C
VN
VG
C
VC
V
C
VN
C
VG
C
VG
N
C
frequency (%)
4. Input frequencies
ADS
50
CV, V, CVC = 73%
35
CV>V> (C)VC>(C)VN>(C)VG(N)>CCV
25
All
20
15
10
5
0
Syllable types
Prediction:
Prediction:order
orderof
ofemergence
emergenceof
of
syllable
types
in
CS
syllable types in CS
4. Input frequencies
CV>V>CVC>CVG/CVN>CCV
CV>V>CVC>CVG/CVN>CCV
CDS
50
45
CV, V, CVC = 78%
40
CV>V>(C)VC>(C)VN/(C)VG(N)>CCV
30
AllMarta
AllInês
25
20
CDS more V (+5%)
15
less types (25 vs. 29)
10
less freq. lower %
5
VG
G
VG
N
G
VC
C
VN
C
VG
C
C
C
V
C
C
VN
C
C
VG
N
C
VG
N
C
C
C
VG
C
C
VC
C
G
VC
VC
C
G
V
G
VG
N
C
VG
C
G
V
C
VN
VG
C
VC
V
C
VN
C
VG
C
VG
N
V
0
C
frequency (%)
35
Syllable types
4. Development of syllable types
CV>V>CVC>CVG/CVN>CCV
CV>V>CVC>CVG/CVN>CCV
CV
CV/ /VV>>(C)VN
(C)VN>>(C)VG,
(C)VG,GV
GV
João - Syllable types
70
60
CV
V
CVN
CVG
CVGN
VG
VN
CGV
GV
GVG
VGN
CGVG
frequency (%)
50
40
30
20
10
0
0_10_2
0_11_6
1_0_12
1_0_27
1_2_1
1_3_4
1_3_11
age
1_4_5
1_5_13
1_6_18
1_7_24
1_8_13
CV>V>CVC>CVG/CVN>CCV
CV>V>CVC>CVG/CVN>CCV
CV
CV/ /VV>>(C)VN
(C)VN>>(C)VG
(C)VG>>(C)VC
(C)VC
4. Development of syllable types
Inês - Syllable types
100
90
80
frequency /%)
70
60
50
40
30
20
10
0
0_11_14
1_0_25
1_1_30
1_3_6
1_4_9
1_5_11
age
1_6_6
1_7_2
1_8_2
1_9_19
1_10_29
CV
V
CVN
CVG
CVGN
VG
CVC
VN
CGV
GV
GVG
GVGN
CVGC
VC
GVN
CVGG
VGN
CGVGN
GVC
CVNC
CV>V>CVC>CVG/CVN>CCV
CV>V>CVC>CVG/CVN>CCV
CV
CV/ /VV, ,(C)VN,
(C)VN,(C)VG
(C)VG>>(C)VC
(C)VC
4. Development of syllable types
Marta - Syllable Types
70
60
frequency (%)
50
40
30
20
10
0
1_2
1_3_8
1_4_8
1_5_17
1_6_23
1_7_18
1_8_18
age
1_10_4
1_11_10
2_0_26
2_1_19
2_2_17
CV
V
CVN
CVG
CVGN
VG
CVC
VN
CGV
GV
GVG
GVGN
CVGC
VC
GVN
CVGG
VGN
CGVGN
GVC
CVNC
VGC
CVCC
CCV
CGVGNC
CCVN
5. Prosodic Prominence and frequency
• Two mismatches
• Prosodic Prominence (EP)
- Development of syllable
- Word-edges (Vigário 2003):
types and input frequencies: Prosodic word initial position
CV and V
often gets emphatic stress
(C)VG/(C)VN before (C)VC
vowels are exempt from regular
- Hypothesis: prosodic
reduction in stressless positions
e.g. [e / E]rguEr vs. ro[ˆ]dOr
prominence (stress and/or
[o / ç]piniÃo vs. mi[u]lInho
word-edges) may have a role
to enhance V, and CVG/N
Prosodic word final position
relative to CVC in the input
tends to show more variable/complex
e.g. Echols 1987; Peters 1977,1983
structures (more V reduction/deletion)
- Stress: stressed syllables are acoustically
longer; no vowel reduction
Syllable
Syllablediversity
diversity&&complexity:
complexity:
Final>Monow>Initial>Internal
Final>Monow>Initial>Internal
5. Input: word-edges vs. internal position
ADS - Syllables by position in w
18
16
14
frequency (%)
12
10
8
V
V
CVC
6
4
CVG/N
CVG/N
CVC
2
0
Initial
Final
Internal
position in w
Monow
CV
V
CVN
CVG
CVGN
VG
CVC
VN
CGV
GV
GVGN
CVGC
VC
VGN
GVC
CVNC
VGC
CVCC
CCV
CCVN
CCVGN
CVGNC
VNC
CCVG
CCVC
CGVC
CCGV
GCVC
CCVNC
Syllable
Syllablediversity
diversity&&complexity:
complexity:
Final>Monow>Initial>Internal
Final>Monow>Initial>Internal
5. Input: word-edges vs. internal position
CDS - Syllables by position in w
20
2 speakers = patterns
CDS more V
V
18
16
frequency (%)
14
12
10
8
V
CVC
6
CVG/N
CVC
CVG/N
4
2
0
InitialM
InitialI
FinalM
FinalI
InternalM
Position in w by speaker
InternalI
MonowM
MonowI
CV
V
CVN
CVG
CVGN
VG
CVC
VN
CGV
GV
GVGN
CVGC
VC
CVGG
VGN
GVC
CVNC
VGC
CCV
CCVN
CCVGN
CVGNC
CCVG
CCVC
CGVC
5. Word-edges vs. internal position
• Input
• Input
- V syllables in ADS, CDS
- CVG/N and CVC in ADS, CDS
Initial+monopw
54%, 56%
Initial+monow
90%, 95%
- V clitics are proclitics and occupy
pw initial position (don’t reduce if
followed by C and never delete –
[u] carro; [u / w] aluno])
- Initial V syllables are not reduced
even if stressless
- V syllables in monopw are
obviously stressed
Final+monopw
CVGN
88%, 100%
CVC
62%, 74%
- Final syllables and monosyllabic
words display more diversity and
complexity of types
- CVGN and CVC in monopw are
obviously stressed
Most V syllables occur in a
prominent word position
Most CVGN and CVC syllables
occur in a prominent word
position, but CVGN does so
more often (mainly in CDS)
5. Word-edges vs. internal position
• CS: CV / V > (C)VN > (C)VG >
-
-
-
(C)VC
So, prosodic prominence may have a
role to enhance V, and CVG/N
relative to CVC in the development
of syllables types
Initial w-position displays more V
syllables than all other positions (as
in the input but with higher
frequency)
Syllable diversity and complexity in
CS closely mirrors the input:
Final, Monow>Initial>Internal
• CS
- More complex types appear first
in final w-position and
monosyllabic words and are
more frequent in these positions
Types
CV
V
(C)VG
(C)VN
(C)VGN
(C)VC
(C)VGC
CC...
INITIAL
BEGIN
BEGIN
J, I, M
M
---------M
FINAL
BEGIN
BEGIN
M
J, I
I, M
M
I
MONOW
BEGIN
BEGIN
M
M
J, I, M
I, M
I, M
INTER
BEGIN
BEGIN
M
-------------------
5. Word-edges – initial position
Inês - Initial Position in w
30
25
CV
V
CVN
CVG
VG
CVC
VN
CGV
GV
GVG
VC
GVN
frequency (%)
20
15
10
5
0
0_11_14
1_0_25
1_1_30
1_3_6
1_4_9
1_5_11
age
1_6_6
1_7_2
1_8_2
1_9_19
1_10_29
5. Word-edges – final position
Inês - Final Position in w
35
30
CV
V
CVN
CVG
CVGN
VG
CVC
CGV
GV
GVG
GVGN
CVGC
VC
frequency (%)
25
20
15
10
5
0
0_11_14
1_0_25
1_1_30
1_3_6
1_4_9
1_5_11
age
1_6_6
1_7_2
1_8_2
1_9_19
1_10_29
5. Word stress
• ADS
• CDS
-
-
-
We saw that word position may
have a role to enhance V, and
CVG/N relative to CVC in the
development of syllables types.
What about stress?
CV, V, CVC more frequent in
unstressed position
CVG, CVN more frequent in
stressed position
- CVN 82%; CVG(N) 66%
Stress strengthens the word
position effect
-
CV, V more frequent in unstressed
position
CVG, CVN and CVC more frequent
in stressed position
- CVG(N)
- CVN
- CVC
-
98%
78%
60%
Still CVG and CVN occur more
frequently in stressed position !
Overall, prosodic prominence may play
a role to enhance CVN/G relative to
CVC
5. Word stress: CV, V
ADS - The effect of stress
40
35
30
frequency (%)
25
CV
V
20
15
10
5
0
stressed
unstressed
position
5. Word stress: CVN, CVG
ADS - The effect of stress
7
CVC
6
frequency (%)
5
CVGN
4
3
CVN
VC
CVG
2
VN
1
0
stressed
unstressed
position
CVN
CVG
CVGN
VG
CVC
VN
CGV
GV
GVGN
CVGC
VC
VGN
GVC
CVNC
VGC
CVCC
CCV
CCVN
CCVGN
CVGNC
VNC
CCVG
CCVC
CGVC
CCGV
GCVC
CCVNC
6. Summary
•
EP frequency distributions of
syllable types in the input (ADS
and CDS) predict the following
order of emergence of syllable
types: CV>V>CVC>CVG/CVN>CCV
• CS: CV / V > (C)VN, (C)VG > (C)VC
• Two mismatches:
CV and V
(C)VG/(C)VN before CVC
- Role of prosodic prominence offers
an explanation to the mismatches:
word-edges vs. internal position
stressed vs. unstresssed position
•
ADS & CDS: Most V syllables
occur in word-initial position
• ADS & CDS: Most CVGN syllables
occur in final position and CVGN
outranks CVC in this position
• ADS & CDS: CVG(N), CVN more
frequent in stressed position
9 Structural information (prosodic
prominence) and frequency together
predict the correct order of
emergence of syllable types
CV > V > CVC > CVG/CVN >CCV
CV / V > CVG/CVN > CVC >CCV
IASCL 2005
Prosodic and frequency effects
on the development of
syllable structure
in European Portuguese
S. Frota*, M. J. Freitas*, M. Vigário+ & F. Martins^
*Universidade de Lisboa – FL and ONSET-CEL,
+
Universidade do Minho – ILCH,
^ Universidade de Lisboa – FL and ILTEC
[email protected]; [email protected];
[email protected]; [email protected]
• Our data: ‘semi-phonological’ • And:
syllables
- Hypothesis: properties in the
signal that cue the type of
• Vowel deletion [ ˆ , u ]
-
Promotes surface C clusters
Demotes CV
• Against the syllable type
patterns shown in CS
• But:
-
V deletion only in internal and final
position (not in initial position)
V deletion only in unstressed
position
syllabic grammar, namely
rhythmic properties that place
EP in the group of syllable
timed languages
Vigário,Frota & Freitas 2003