Methods for defining categories in intonational phonology: A check on Italian data Barbara Gili...
-
Upload
lewis-hicks -
Category
Documents
-
view
213 -
download
0
Transcript of Methods for defining categories in intonational phonology: A check on Italian data Barbara Gili...
Methods for defining categories in intonational phonology:A check on Italian data
Barbara Gili FivelaBarbara Gili Fivela
Università del Salento – Lecce, ItalyUniversità del Salento – Lecce, Italy
CRIL – Centro di Ricerche Interdisciplinare sul LinguaggioCRIL – Centro di Ricerche Interdisciplinare sul Linguaggio
[email protected]@ateneo.unile.it
Universitat Autònoma de Barcelona - May the 14th 2007
Overview
Introduction Linguistics and paralinguistics
Intonation, meaning and categories Methods for defining categories in intonation
Italian data: production and perception Production, perception, and perception-production:
Different constraints in production and perception? Categorical perception in intonation
Prosody and intonation
Prosody is due to variation in: fundamental frequency (F0)-pitch duration-length intensity-loudness
speech tempo / speech rate rhythm
Intonation [Ladd, 1996] suprasegmental: F0, intensity, duration conveying meaning to phrases/utterances organized in terms of categorically distinct entities and relations
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Linguistics and paralinguistics
Linguistics: the scientific study of language [Crystal, ’91]
Paralinguistics: independent from the linguistic message…although it is ‘coordinated in time with the linguistic channel’ and influences the interpretation of the utterance [Ladd, 1996: 34] Interaction (solidarity, aggression)
speaker’s attitude Emotions (fear, surprise)
“intonation clearly ‘feels’ paralinguistic” [Ladd, 1996: 38] same features used for paralinguistic change
(e.g.,voice quality) over long stretches of speech (e.g. loudness) affective and interpersonal meaning (e.g. doubt, irony)
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Linguistics and paralinguistics in phonology/phonetics
Linguistics: categorical distinction Paralinguistics: gradual changes
Segmental level: /i/ vs. /u/ ex.it. mito vs. muto; sp. si vs. su /i/ vs. /i/ produced while smiling
Suprasegmental level Truth value of the utterance ex. In Saint Petersburg, OFFICERS always escort ballerinas
[Rooth, 1985] Sentence modality
ex. it. vai vs. vai?ex. cat. volen una nena vs. ¿volen una nena?
…..produced for conveying surprise
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Intonational meaning British tradition
functional units, such as head, nucleus, tail [Palmer, 1922] intonation “involves the occurrence of pitch patterns, each of
which is used with a set of relatively constant meaning, either on single words or on groups of words” [Cruttenden, 1986:9]
IPO approach pitch movements, defined through perceptual equivalence,
combined according to a grammar of intonation in configurations and contours [‘t Hart and Collier, 1990]
intonation features have no intrinsic meaning, its semantics may be related to syntax, in cases of ambiguity resolution [id.]
Autosegmental theories sequences of L and H tone targets, belonging to pitch accents
and edge tones [Bruce, 1977; Pierrehumbert, 1980] pitch accents and edge tones convey both linguistic and
paralinguistic meaning [Pierrehumbert and Hirschberg ,’90; Kohler,’91; Ladd,’96; Gussenhoven,’04]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Intonation conveys linguistic and paralinguistic meaning
The meaning of intonation is considered as a way to shed light on its form [Ladd,1996: 98]
linguistic entities paralinguistic cues imply modification of the
way phonological categories are realized
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Intonation conveys linguistic and paralinguistic meaning - II
Tune-based analysis and tone-based analysis Meaning conveyed by the whole contour
[Liberman and Sag, 1974]
Meaning derived from contour’s components [Pierrehumbert and Hirschberg, 1990]
Changes in the form of intonation Implying a change in category Gradual, for signalling paralinguistic changes Due to phonetic implementation
Problems with paralinguistic variation High-fall vs. low-fall treated as contrasting in some
analysis and as paralinguistic variants in others [O’Connor and Arnold, 1973 vs. Crystal, 1969]
Gradient form-meaning relations may be grammaticalized as discrete [Gussenhoven, 2002]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Biological codes
Frequency code [Ohala, 1983; Gussenhoven, 2002] Differences due to phonatory system
low = dominant-self confidence-assertive mode Grammaticalization: statement vs. questions
Effort code [Gussenhoven, 2002] Differences due to effort in production
High =important-surprise-emphasis-focus Grammaticalization: focus
Production code [Gussenhoven, 2002] Differences due to energy dissipation
Lowering = end of constituent – finality
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Which information are expressed by categorical elements?
Accentuation, focus, phrase signals for indicating sentence modality, function and meaning statement, yn-question [D’Imperio and House, 1997] yn- and wh-question, check, focus, continuation
[Cruttenden, 1986, Pierrehumbert and Beckman, 1986, Casper, 1998] check, query and accessibility [Grice and Savino, 2003]
introducing, committing to presence, and selecting from background [Gussenhoven, 1984]
new, salient, linked to mutual believes, to be interpreted with the following [Pierrehumbert, Hirschberg,’90]
finality-knowing, openess-realizing [Kohler, 1987; 1991]
direct, indirect speech acts [Liberman and Sag, 1974; 1975] topic and comment [Cresti, 2000]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
…and by gradient variations?
Emotions, attitude… Perception of paralinguistic form-function relation is
influenced by subject’s background [Chen, 2005] Degrees of meaning related to the linguistic unit
Openess-realizing and unexpectedness-opposing [Kohler, 1991; 2006]
Neutral statement and contraddiction [Ladd and Morton, 1997] two different meanings, categorically perceived
Variations in: pitch range, i.e. scaling of targets on the frequency scale alignment, i.e. synchronization with segmental chain
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Where do categories usually come from? Production
categories defined on the basis of speech recordings: read and (semi)spontaneous speech
Perception categories defined on the basis of perception
experiments: IPO approach
Production and perception Patterns observed in production Perception of different categories checked by means of
perception experiments
Semantic contrast is no longer sufficient to ‘proof’ a structural difference [Kohler, 1991; Ladd and Morton, 1997; Gussenhoven, 2006]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
In intonation contrast: both discrete and gradient
categorical = discrete linguistic = categorical
Production and perception mismatch Production data suggest syllable onset as reference
point for alignment [Caspers and van Heuven, 1992; van Santen and Möbius, 2000]
Ladd [1999] ‘segmental anchoring’ hypothesis, but see data discussed in the literature [Prieto and Torreira, in print; Loevenbruck and Welby, in print; Gili Fivela, 2004]
Perception data point to vowel onset as crucial for tone comparison [House, 1990:113]
Perceptual and acoustic tonal targets [D’Imperio, 2000]
‘Intonational’ and ‘categorical’
linguistic = discrete
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Problems with methods for testing (categorical) perception? ‘Even in situations where subjects can make sharp
distinctions between classes, they are still able to discriminate within a class’ [Ladd and Morton, 1997]
“The ideal experiment […] observes the subject’s behaviour in a situation as close to natural conversation as possible” [Kochanski, 2006]
‘Perception’ and ‘categorical’
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
‘Linguistic’ and ‘categorical’ Kohler [2006] observes that according to some theories
of intonational meaning: Linguistically relevant elements are discrete and
categorical Categories of intonation have to be distinguished from
paralinguistic modifications Only categories are linguist’s concern
Analyses of communicative functions and meanings few meanings in the linguistic domain (accentuation,
focus, phrasing, sentence modality) intonation is mainly concerned with paralanguage
(expressive and attitudinal;interactive; speaker evaluation of events - finality, openess; style)
-> “Categorical perception in the classical sense is therefore a special case and not essential for pitch categorization” (see also Newport [1982])
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Methods for defining categoriesProduction
Speech recordings, both monologues and dialogues read speech, proposing specific functions (semi)spontaneous speech obtained by means of
various tasks
• e.g., map-task, card games Analysis of F0 tracks, in relation to segmental events Invariant tonal events, independent of phonetic
modifications
“The ideal general methodology would then be some kind of cyclicity between test material and spontaneous speech using feedback from preceding studies” [Bruce and Touati, 1990].
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Methods for defining categoriesPerception - I
Perceptual equivalence [‘t Hart and Collier, 1990]Structural discreteness tested by means of speakerintuition of perceptual equality, i.e. ‘passable imitations’ of each other ‘passable imitation’ [Odé, 2005; Gussenhoven, 2006]
Categorical perception [Repp, 1984; Gussenhoven, 1999]
Identification task continuum between two phonological categories stimuli are assigned to either category abrupt shift
Discrimination task pairs of stimuli to be judged ‘same’ or ‘different’ expected grater distinction across perceptual boundary
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Categorical perception
CAT I
%
CAT IIContinuum of stimuli
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Methods for defining categoriesPerception - II
Perceptual magnet effect [Kuhl, 1991; Schneider et al., 2006] each category has a prototype lower discrimination sensitivity for its neighbours
Identification task Goodness rating
• rating as for very bad/very good exemplar • individuation of the prototype
Discrimination task • prototype and (not necessarily adjacent) neighbour
Semantic difference and scaling [Gussenhoven, 1999] Gradient and categorical judgments on the presence of a
meaning or its opposite (Grabe [1997] for discussion) Judgement on the extent to which a meaning is conveyed -
especially for paralinguistic On rating scales, see Chen [2005]; Rietveld and Chen [2006]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Methods for defining categoriesPerception - III
Imitation [Pierrehumbert and Steele, 1989; Gussenhoven, 1999]
continuum between two patterns subjects are asked to imitate each stimulus, paying
attention to the intonation pattern in case they produce the whole continuum, the
difference is gradient; in case of binomial distribution it is categorical
‘correcting’ not acceptable patterns [Gussenhoven, 2006] imitation of their own imitation [Brown et al., 2006;
Kochanski, 2006]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Check on Italian data Pisa Italian
Production Read speech
out of the blue within context utterances
(Semi)spontaneous Map-Task
Perception Perceptual equivalence
Passable imitation Categorical perception Perceptual magnet effect Semantic difference and scaling Imitation
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Production
Inventory of Pisa Italian tonal events [Gili Fivela, 2004]
Functions wh-question: query-wh yn-question: query-yn, checks and align statements: instruct focalization
Structurally distinct units three types of nuclear pitch accents: H*, H*+L, H+L* edge tones: L-L%, H-L%, L-H%, H-H%
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Examples
e dove dev(o) andare ?
H* H+L* L-L%
hai detto leggimelo ?
H+L* H-L%
and where should I go ?
did you say read it to me?
wh-question
yn-question / check
Examples
Allora ripartiamo riparti dalla partenza
H+L* L- H+L* L-L% H* H+L* L-L%
statement
then let’s start again you begin from the start
statement – narrow focus
[L+]H*+L L-L%
it will be a centimetersarà<aa> un centimetro
Production: categories and meaning Meaning of pitch accents difficult to delimit
pitch accents are be shared by different functions more than one pitch accent type may be exploited
for a specific function depending on pragmatic variation
general meaning, coherently with literature
Analysis of contrasting characteristics lead to the choice of transparent, either abstract or detailed, labels emphasizing the structurally distinctive
characteristics differentiating shared and dissimilar structural
properties low target point as starting point of a rise to H*
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
ExamplesMangia il melone (…)
statementbroad focusutterance final
s/he eats the mellon …Time (s)
0 1.22032-0.2469
0.1761
0
Time (s)0 1.22032
0
350
Time (s)0 1.52363
-0.1578
0.1326
0
Time (s)0 1.52363
0
350
Time (s)0 0.819728
-0.4992
0.3092
0
Time (s)0 0.819728
0
350
H+L* L-L%
[L+]H*+L L-L%
[L+]H* L-
statementcontrastive focusutterance final
statementbroad focus utterance initial /(narrow focus)
Pitch accents under investigation[Gili Fivela, 2002]
La pronuncia di lavaglielo non (la) ricordo mai The pronunciation of lavaglielo I never remember (it)
B
C
[L+]H*+L L-
[L+]H* L-
PerceptionPitch accents under investigation
Measurements of F0, latencies between targets and segmental points showed that the accents differ as for target alignment [L+] H* [L+] H*+L
ms
target scaling
Hz
syllable duration
ms
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Acoustic manipulation
Difference of mean values
Number of steps for gradually getting from one pattern to the other one: 8 alignment steps: 15 ms 2 scaling steps: 13Hz(L) - 17Hz(H) - 6 Hz(L) 5 repetitions
PRAAT – PSOLA resynthesis Perceval (Aix-en-Provence) for perception test
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Perception of peak accents
Identification test, manipulating the alignment and scaling characteristics of stimuli – 10 subjects
‘No. Ho detto velava velocemente’ I said velava quickly
Alignment
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
Med
ia p
unte
ggio
per
C
0 1 2 3 4 5 6 7 8
PR2
PR1
PR0
Align steps
• Is it a peremptory and
conclusive correction?
• Stimuli are ambiguous as
for pitch height
• Pitch height has a significant
influence on perception
Scaling
[Gili Fivela, 2005]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Summing up: peak accents
In identifying two peak accents, “S-shaped” plots in relation to alignment
but there was always an ambiguous pitch height value Extremes are categorically perceived
Pitch height has an influence on ‘when’ a different pattern is perceived
Discrimination task would be needed best with no ambiguous cues
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Peak accents: attention to cues
In the identification test subjects appear to rely on different cues
Align PitchRange
-,2
0
,2
,4
,6
,8
1
1,2
Mea
n of
ans
wer
s in
favo
ur o
f con
tras
tive
inte
rpre
tatio
n
0 1 2 3 4 5 6 7 8
PR2
PR1
PR0
Ar
-,2
0
,2
,4
,6
,8
1
1,2
Mea
n of
ans
wer
s in
favo
ur o
f con
tras
tive
inte
rpre
tatio
n
0 1 2 3 4 5 6 7 8
PR2
PR1
PR0
Fa
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Imitation task Imitation of same stimuli
(same steps of manipulation) contrastive base, 3 subj
Stimulus –
say number - beep –
target imitation Measurements of (L)HL target
height and latencies: vowel onset-to-H
-,01
,02
,04
,06
,09
,12
,14
,17
,19
Cel
l Mea
n fo
r 'd
v1H
'
al0 al1 al2 al3 al4 al5 al6 al7 al8
c-pr2
c-pr1
c-pr0
-,01
,02
,04
,06
,09
,12
,14
,17
,19
Cel
l Mea
n fo
r 'd
v1H
'
al0 al1 al2 al3 al4 al5 al6 al7 al8
c-pr2
c-pr1
c-pr0
ILL
IL
-,12
-,07
-,02
,03
,08
,13
,18
,23
,28
Cel
l Mea
n fo
r 'd
v1H
'
al0 al1 al2 al3 al4 al5 al6 al7 al8
c-pr2
c-pr1
c-pr0
NI
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
ExamplesMangia il melone (…)
statementbroad focusutterance final
s/he eats the mellon …Time (s)
0 1.22032-0.2469
0.1761
0
Time (s)0 1.22032
0
350
Time (s)0 1.52363
-0.1578
0.1326
0
Time (s)0 1.52363
0
350
Time (s)0 0.819728
-0.4992
0.3092
0
Time (s)0 0.819728
0
350
H+L* L-L%
[L+]H*+L L-L%
[L+]H* L-
statementcontrastive focusutterance final
statementbroad focus utterance initial /(narrow focus)
PerceptionPitch accents under investigation-II Absence/presence of a (close) low target preceding a
rise to peak Measurements of F0, latencies between targets and
segmental points showed that the accents differ as for target alignment H+L* [L+] H*+L
ms
target scaling
Hz
syllable duration
ms
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Acoustic manipulation
Difference of mean values
Number of steps for gradually getting from the broad focus pattern to the contrastive one: 5 alignment steps: 22 ms 4 scaling steps: 15 Hz (+ 1 step 7.5 Hz) 3 repetitions PRAAT (PSOLA) Perceval
Test on perceptual relevance of L+ target [Gili F., 2006]• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
-,1
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
Contr
astiv
e focus: m
ean o
f positi
ve a
nsw
ers
0 al1 al2 al3 al4 al5
H8
H6
H4
H2
0
Steps of align manipulation
Perception of falling accents
Identification test, manipulating the alignment and scaling characteristics, from a broad stimulus – 10 subj
‘Mangia il melone’ s/he eats the mellon
Alignment
• Does it correct a
preceeding utterance?
• Stimuli are ambiguous as
for pitch height
• Pitch height has a small
influence on perception
Scaling
[Gili Fivela, 2006]
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Falling accents: attention to cues
In the identification test subjects appear to rely on different cues
Align PitchRange
-,1
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
1,1
Con
tras
tive
focu
s: m
ean
of p
ositi
ve a
nsw
ers
0 al1 al2 al3 al4 al5
H8
H6
H4
H2
0
-,1
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
1,1
Con
tras
tive
focu
s: m
ean
of p
ositi
ve a
nsw
ers
0 al1 al2 al3 al4 al5
H8
H6
H4
H2
0
FV MA
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
-,1
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
Con
tras
tive
Foc
us: M
ean
of p
ositi
ve a
nsw
ers
0 al1 al2 al3 al4 al5
C-noL-H9
C-noL-H8
C-noL-H6
C-noL-H4
C-noL-H2
C-0
0
Is there a base effect ? Identification test, manipulating the alignment and
scaling characteristics, from a contrastive stimulus
12 subjects
Scaling Alignment
• No S-shaped plot• Stronger influence of pitch height • Extremes are categorically perceived
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Summary and comments In identifying two falling accents,
No S-shaped plots in relation to alignment but there was always an ambiguous pitch height value not even categorically perceived (broad base)
Pitch height alone has an influence, at least when a contrastive base is considered
other correlates? Syllable duration?
Possible reasons for these results: Ambiguity in function or meaning?
Same phonological categories not gradient variation in production coherent, at least partly, with Gussenhoven’s hypothesis
Not appropriated task more articulated context: question-answer sequence? sentence modality: question (check) vs statement?
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
0
,2
,4
,6
,8
1
1,2
Cel
l Mea
n fo
r an
swer
YE
S
0 al1 al2 al3 al4 al5
H9
H8
H6
H4
H2
C
0
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
Cel
l Mea
n fo
r M
edie
YE
S
0 al1 al2 al3 al4 al5
C-H9
C-H8
C-H6
C-H4
C-H2
C-0
0
What about a different task ?‘Question-answer’
Broad focus base Contrastive focus base
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Identification test, from both broad and contrastive base – 11/12 subjects
Sequence of question-answer Broad context: broad focus answer expected
Cosa succede what’s up? Is the answer adeguated to the question?
Same results !
What about a different task ?‘Is it a question?’
Identification test, from broad base - 13 subjects Utterance in isolation
Need to set up a quite complex context based on mutual believes
Could you interpret it as a check of information? Would you give a yes/no answer?
Subjects were actually giving answers that were exactly the opposit of expected ones !
Too difficult task
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
0
,1
,2
,3
,4
,5
,6
,7
,8
,9
1
1,1
Cel
l Mea
n fo
r A
nsw
ers
YE
S
0 al1 al2 al3 al4 al5
H9
H8
H6
H4
H2
C
0
-,1
-,08
-,05
-,03
0
,03
,05
,08
,1
,13
,15
Dis
tance o
f peak fro
m s
ylla
ble
onset (m
s)
0 al1 al2 al3 al4 al5
C-H9
C-H8
C-H6
C-H4
C-H2
C-0
0
-,04
-,02
0
,02
,04
,06
,08
,1
,12
,14
,16
Dis
tanc
e of
pea
k fr
om s
ylla
ble
onse
t (m
s)
0 al1 al2 al3 al4 al5
C-H9
C-H8
C-H6
C-H4
C-H2
C-0
0
-,08
-,05
-,03
0
,02
,05
,08
,1
,13
,15
,17
Dis
tanc
e of
pea
k fr
om s
ylla
ble
onse
t (m
s)
0 al1 al2 al3 al4 al5
C-H9
C-H8
C-H6
C-H4
C-H2
C-0
0
Imitation task Imitation of same stimuli
(same steps of manipulation) contrastive base – 3 subj
Stimulus - beep -
imitation – say a number -
target imitation Measurements of (L)HL target height
and latencies: syllable onset-to-H
ILNI
ILL• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Discrimination: falling accents Pairs: AB and AA
B with either higher or later peak [Ladd and Morton, 1997] From both broad and contrastive base
Same or different? 9 subjects
No discrimination ! Reaction times give no information
-,2
0
,2
,4
,6
,8
1
Cel
l Mea
n fo
r M
ean
NO
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
discrimination scores
1400
1450
1500
1550
1600
1650
1700
Cel
l Mea
n fo
r R
eact
Tim
e
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
reaction times
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
-,2
0
,2
,4
,6
,8
1
Cel
l Mea
n fo
r M
ean
NO
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
1400
1500
1600
1700
1800
1900
2000
2100
Cel
l Mea
n fo
r R
eact
Tim
e
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
IL
-,2
0
,2
,4
,6
,8
1C
ell M
ean
for
Mea
n N
O
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
discrimination scores
1000
1200
1400
1600
1800
2000
2200
2400
2600
2800
3000
3200
Cel
l Mea
n fo
r R
eact
Tim
e
C0-
C0
Cal
1-C
0
Cal
2-C
al1
Cal
3-C
al2
Cal
4-C
al3
Cal
5-C
al4
reaction times
ILNI
NI IL
Summary
Production data quite straightforwardly showed the existence of (patterns) pitch accents related to functions/meanings
Perception data seem to be far more problematic: Reason may be that they have being
investigated teasing alignment and scaling apart In any case, the H* vs H*+L contrast appears
to be perceptually more different that the H*+L H+L* contrast
Probably only one phonological distinction
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Summary
Two peak accents: [L+]H* vs [L+]H*+L “S-shaped” identification results
but not all speakers appear to pay attention to the same cues
Extremes are correctly identified 2 out of three speakers appears to perform the
imitation taskThey both imitate the continuum by creating
two different classes as for the peak distance to vowel onset
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Summary Two falling accents: [L+]H*+L vs H+L*
No S-shaped identification results and not all speakers appear to pay attention to the
same cues Extremes are correctly identified 2 out of three speakers appears to perform the task
One speaker creates two different classes as for the peak alignment
One speaker creates a continuum of alignment differences
Raction times, in general, give no information, but looking at the ‘discriminator’ speaker: Trend in the
direction of categorical peak and coherent information for reaction time
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
Improvements
Methods could be improved Gussenhoven proposal of ‘passable imitation’ Perceptual Magnet Effect rather than traditional
CP, in particular for the discrimination task
Experiments to be runned without teasing apart correlates
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
ExamplesMangia il melone (…)
statementbroad focusutterance final
s/he eats the mellon …Time (s)
0 1.22032-0.2469
0.1761
0
Time (s)0 1.22032
0
350
Time (s)0 1.52363
-0.1578
0.1326
0
Time (s)0 1.52363
0
350
Time (s)0 0.819728
-0.4992
0.3092
0
Time (s)0 0.819728
0
350
H+L* L-L%
[L+]H*+L L-L%
[L+]H* L-
statementcontrastive focusutterance final
statementbroad focus utterance initial /(narrow focus)
Discussion Problem due to the form of intonation
especially in the second sets of experiments, pitch accents were formally really different
Proportional change of all the features needed The resynthesis of continua is reasonable to apply for
similar patterns Problem due to meaning/function of intonation
general meanings functions expressed by the same pitch accent
Both production and perception should be taken into consideration in deciding whether there is a contrast Production could be more robust than perception
need of producing redundant features perception within the context of strictly linguistic
message
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
…and discussion Phonological perception should say something on
the properties of phonic chain However intonation has also being described in
terms of morphemes
Morphemes have meanings may select different meaning of a base
(message?) among their meanings, one may be selected
depending on the base (message?) Intonational units have a meaning/function even
though they are not categorically perceived ?
• Introduction
• Meaning and
Categories
• Methods
• Check on
Italian
• Production
• Perception
• Discussion
If a language does not show a contrast that is adapt for perceptual testing,
does it really mean that the language does not exploit a/that phonological contrast?
….in Pisa as well……
In Saint Petersburg, OFFICERS always escort ballerinas orIn Saint Petersburg, officers always escort BALLERINAS
THANK YOU!