Categorically perceiving vs. Categorizing while perceiving:
The role of segments' recognition and lexical access
while categorizing the pragmatic function of pitch movements in speech.
Alessandra Zappoli 1, Massimo Grassi 2, Cinzia Avesani 3, Francesco Vespignani 4
Center for Cognitive Science of Language, University of Nova Gorica, Slovenia
2Department of General Psychology, University of Padova, Italy
3Institute of Cognitive Science and Technologies of Padova (CNR), Italy
4Department of Developmental Psychology and Socialisation, University of Padova, Italy
1
RESULTS
INTRODUCTION
• The fundamental frequency (F0), and its perceptual correlate, Pitch, is one of the prosodic cues that convey linguistic
(semantic, syntactic, and pragmatic) and paralinguistic (irony, mood, etc.) information.
• It is organized in phonological categories (Pitch Accents, Boundary Tones (BTs)) aligned to the text (segmental
information).
• In Italian, the opposition alone between Low BT vs. High BT can convey two different syntactic-pragmatic
interpretations of the same linguistic utterance: Statement vs. Yes-No Question (Gili Fivela, 2008).
• Segmental Phonological Categories showed to be Categorically Perceived (CP) (Liberman et al., 1967).
• Perception of BTs has been found Quasi-Categorical (Schneider, 2012).
• CP in speech and categorization as a cognitive mechanism affecting several domains of cognition is debated. (Holt &
Lotto, 2010).
• Pitch perception, conveying linguistic and non-linguistic information, in the presence and absence of recognizable
linguistic segmental information has not been yet fully investigated.
• We investigate whether the Categorization of BT, shifting from Low to High, is affected by the presence and absence of
linguistic, semantically accessible segmental information (3 degrees manipulation); and whether the presence of
recognizable linguistic segmental information affects the perceptual mechanism if presented before or after the stimuli
lacking recognizable linguistic segmental information.
CATEGORIZATION TASK
Correlation of Slope values across conditions.
Seq 1
Seq 2
Correlation of Midpoint values across conditions.
Seq 1
Seq 2
Slope values in Seq 1
Slope values in Seq 2
Midpoint values in Seq 1
Midpoint values in Seq 2
Seq 2 - Fixed and correlated random effects
Seq 2 - Fixed effect and random intercepts
FL (x; α,β)= 1/(1+exp(β(α-x))
α = Midpoint – Category Boundary
β = Slope - Degree of
Identification/Categorization
STIMULI
• 4 Conditions of stimuli on which manipulated F0 is aligned and varied from Statement-Category to Question-Category
through 11 levels of Boundary Tones, for a total step of 140 Hz.
• 1) Holophrasis in Italian (e.g. [aˈnɛmoni]).
• 2) Pseudo Word in Italian(e.g. [e’nimena]).
• 3) Foreign Word- Russian segments Italian phonotactics ([ɨ’lɨnʲəlɨ]).
• 4) F0 movement on de-lexicalized speech (Humming).
• The first presented condition of each sequence is repeated at the end of each sequence as a retest condition.
Statement F0 Contour.
Semi-spontaneous recorded speech.
Question F0 Contour &
Synthetic generation of intermediate levels.
SEQUENCES
• Seq 1 – From Word to Humming – Repetition of Word at the end (Retest).
• Seq 2 – From Humming to Word – Repetition of Humming at the end (Retest).
COVARIATE MEASURES
• Autism Spectrum Quotient (AQ).
• Interpersonal Reactivity Index (IRI).
PARTICIPANTS
• 34 Women from Trentino, NW Veneto, NE Garda Lake area with no knowledge of Russian. Information on Musical
Training and degree of experience in Foreign Languages acquisition has been collected.
• 5 participants were excluded from group analysis but included as informative results. (Impossible fit with a logistic
function in the Humming condition. Humming condition non-categorized.)
• 15 perceived Seq 1 – 14 perceived Seq 2
PSYCHOPHYSICS CURVES IN SEQ 2
Seq 2 – Prototypical Expected Categorization curves in presence of Categorical Perception
Seq 2 – Categorization Curves – Categorization Process
RESULTS OF PCA OF COVARIATE MEASURES
Scales
Questionnaire
• Attention to Details (Att_Dett)
• Attention Switching (Sp_Att)
• Social Abilities (Ab_Soc)
• Communication (Comunic)
• Imagination (Imag)
AQ
• Fantasy Scale (FS)
• Perspective Taking (PT)
• Empathic Concern (EC)
• Personal Distress (PD)
IRI
PC1 (25% variance) & Midpoint values in Seq 1
PT
Immag_30
Att_Dett_30
Ab_Soc_30
EC
Comunic_30
Sp_Att_30
FS
PD
DISCUSSION AND CONCLUSIONS
Sequence 1:
• Slope values are set in the 1st condition and reliably kept averagely constant along the sequence, adopting a highly
correlated categorization strategy across conditions.
• Midpoint (category boundary) values are set in the 1st condition and interact with the modulation of linguistic segmental
information in the signal but do not show a linear effect of order and training.
• The categorization strategy in the Humming condition shows to be unrelated to the strategy adopted in the first linguistic
(word) condition.
• LMMs show that random intercepts do not interact with the sequence – Pitch is linguistically categorized.
• Individual variability in the decoding of the interlocutor’s mental states correlates with the modeling of midpoint values
(25% of explained variance), thus with the criteria to establish the category boundaries.
Sequence 2:
• The categorization process is observable in the linear trend of slope and midpoint values along the linguistic conditions.
• Slope values linearly increase; Midpoint values linearly decrease and correlate only with the adjacent condition ->
Categorization Process.
• When the Humming condition must be categorized based on its pragmatic – syntactic interpretation without segmental
information 5 participants fail the task, and the rest of the group adopts a strategy that doesn’t correlate with the
following linguistic ones. It correlates only in the Retest condition.
The two sequences show that the Humming condition is categorized with two different strategies depending on when the
linguistic segmental information is accessed and retrieved or not.
The pragmatic function of pitch is not directly accessed if segmental information is missing.
The two pragmatic categories must be created by linking the acoustic information about pitch height with linguistic
information. Categories are formed through a categorization process along sequence 2. The information is available and
retrieved in sequence 1.
References
Liberman, A. M., Cooper, F. S., Shankweiler, D. P., & Studdert-Kennedy, M. (1967). Perception of the speech code. Psychol. Rev., 74, 431–461.
Gili Fivela B. (2008) Intonation in Production and Perception: The Case of Pisa Italian. Edizioni dell'Orso, Alessandria, ISBN 978-88-6274-066-1.
Holt, L., & Lotto, A. J. (2010). Speech perception as categorization. Attention, Perception, & Psychophysics, 72(5), 1218–1227.
Schneider, K. (2012). The German boundary tones: categorical Perception, perceptual magnets, and the perceptual reference space.
Gili Fivela, B., Avesani, C., Barone, M., Bocci, G., Crocco, C., D'Imperio, M., ... & Sorianello, P. (2015). Intonational phonology of the regional varieties of Italian. Intonation in Romance.
RESEARCH POSTER PRESENTATION DESIGN © 2012
www.PosterPresentations.com