Models and Theories of Speech Production

310 Pages·2020·53.851 MB·English

by  Adamantios Gafos, Pascal van Lieshout

MODELS AND THEORIES OF SPEECH
PRODUCTION
EDITED BY : Adamantios Gafos and Pascal van Lieshout
PUBLISHED IN : Frontiers in Psychology and Frontiers in Communication
Frontiers eBook Copyright Statement About Frontiers
The copyright in the text of
individual articles in this eBook is the Frontiers is more than just an open-access publisher of scholarly articles: it is a
property of their respective authors
pioneering approach to the world of academia, radically improving the way scholarly
or their respective institutions or
funders. The copyright in graphics research is managed. The grand vision of Frontiers is a world where all people have
and images within each article may
an equal opportunity to seek, share and generate knowledge. Frontiers provides
be subject to copyright of other
parties. In both cases this is subject immediate and permanent online open access to all its publications, but this alone
to a license granted to Frontiers.
is not enough to realize our grand goals.
The compilation of articles
constituting this eBook is the
property of Frontiers. Frontiers Journal Series
Each article within this eBook, and
the eBook itself, are published under The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access,
the most recent version of the online journals, promising a paradigm shift from the current review, selection and
Creative Commons CC-BY licence.
The version current at the date of dissemination processes in academic publishing. All Frontiers journals are driven
publication of this eBook is
by researchers for researchers; therefore, they constitute a service to the scholarly
CC-BY 4.0. If the CC-BY licence is
updated, the licence granted by community. At the same time, the Frontiers Journal Series operates on a revolutionary
Frontiers is automatically updated to
invention, the tiered publishing system, initially addressing specific communities of
the new version.
scholars, and gradually climbing up to broader public understanding, thus serving
When exercising any right under the
CC-BY licence, Frontiers must be the interests of the lay society, too.
attributed as the original publisher
of the article or eBook, as
applicable. Dedication to Quality
Authors have the responsibility of
ensuring that any graphics or other Each Frontiers article is a landmark of the highest quality, thanks to genuinely
materials which are the property of collaborative interactions between authors and review editors, who include some
others may be included in the
CC-BY licence, but this should be of the world’s best academicians. Research must be certified by peers before entering
checked before relying on the a stream of knowledge that may eventually reach the public - and shape society;
CC-BY licence to reproduce those
materials. Any copyright notices therefore, Frontiers only applies the most rigorous and unbiased reviews.
relating to those materials must be
Frontiers revolutionizes research publishing by freely delivering the most outstanding
complied with.
research, evaluated with no bias from both the academic and social point of view.
Copyright and source
acknowledgement notices may not By applying the most advanced information technologies, Frontiers is catapulting
be removed and must be displayed
scholarly publishing into a new generation.
in any copy, derivative work or
partial copy which includes the
elements in question.
What are Frontiers Research Topics?
All copyright, and all rights therein,
are protected by national and Frontiers Research Topics are very popular trademarks of the Frontiers Journals
international copyright laws. The
above represents a summary only. Series: they are collections of at least ten articles, all centered on a particular subject.
For further information please read With their unique mix of varied contributions from Original Research to Review
Frontiers’ Conditions for Website
Use and Copyright Statement, and Articles, Frontiers Research Topics unify the most influential researchers, the latest
the applicable CC-BY licence. key findings and historical advances in a hot research area! Find out more on how
ISSN 1664-8714 to host your own Frontiers Research Topic or contribute to one as an author by
ISBN 978-2-88963-928-1
DOI 10.3389/978-2-88963-928-1 contacting the Frontiers Editorial Office: [email protected]
Frontiers in Psychology 1 August 2020 | Models and Theories of Speech Production
MODELS AND THEORIES OF SPEECH
PRODUCTION
Topic Editors:
Adamantios Gafos, University of Potsdam, Germany
Pascal van Lieshout, University of Toronto, Canada
Citation: Gafos, A., van Lieshout, P., eds. (2020). Models and Theories of Speech
Production. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-928-1
Frontiers in Psychology 2 August 2020 | Models and Theories of Speech Production
Table of Contents
05 Editorial: Models and Theories of Speech Production
Adamantios Gafos and Pascal van Lieshout
09 Emergence of an Action Repository as Part of a Biologically Inspired
Model of Speech Processing: The Role of Somatosensory Information
in Learning Phonetic-Phonological Sound Features
Bernd J. Kröger, Tanya Bafna and Mengxue Cao
26 Variability and Central Tendencies in Speech Production
D. H. Whalen and Wei-Rong Chen
35 Modeling Dimensions of Prosodic Prominence
Simon Roessig and Doris Mücke
54 The Emergence of Discrete Perceptual-Motor Units in a Production
Model That Assumes Holistic Phonological Representations
Maya Davis and Melissa A. Redford
73 Motoric Mechanisms for the Emergence of Non-local Phonological Patterns
Sam Tilsen
98 Bridging Dynamical Systems and Optimal Trajectory Approaches
to Speech Motor Control With Dynamic Movement Primitives
Benjamin Parrell and Adam C. Lammert
115 Modeling Sensory Preference in Speech Motor Planning: A Bayesian
Modeling Framework
Jean-François Patri, Julien Diard and Pascal Perrier
129 The Morphogenesis of Speech Gestures: From Local Computations
to Global Patterns
Khalil Iskarous
140 Economy of Effort or Maximum Rate of Information? Exploring
Basic Principles of Articulatory Dynamics
Yi Xu and Santitham Prom-on
162 Native Language Influence on Brass Instrument Performance: An
Application of Generalized Additive Mixed Models (GAMMs) to Midsagittal
Ultrasound Images of the Tongue
Matthias Heyne, Donald Derrick and Jalal Al-Tamimi
188 Noggin Nodding: Head Movement Correlates With Increased
Effort in Accelerating Speech Production Tasks
Mark Tiede, Christine Mooshammer and Louis Goldstein
205 Spatially Conditioned Speech Timing: Evidence and Implications
Jason A. Shaw and Wei-rong Chen
222 The Role of Temporal Modulation in Sensorimotor Interaction
Louis Goldstein
234 Spoken Language Development and the Challenge of Skill Integration
Aude Noiray, Anisia Popescu, Helene Killmer, Elina Rubertus, Stella Krüger
and Lisa Hintermeier
Frontiers in Psychology 3 August 2020 | Models and Theories of Speech Production
251 A Simple 3-Parameter Model for Examining Adaptation in Speech
and Voice Production
Elaine Kearney, Alfonso Nieto-Castañón, Hasini R. Weerathunge,
Riccardo Falsini, Ayoub Daliri, Defne Abur, Kirrie J. Ballard,
Soo-Eun Chang, Sara-Ching Chao, Elizabeth S. Heller Murray,
Terri L. Scott and Frank H. Guenther
267 Timing Evidence for Symbolic Phonological Representations
and Phonology-Extrinsic Timing in Speech Production
Alice Turk and Stefanie Shattuck-Hufnagel
287 Speech Sound Disorders in Children: An Articulatory Phonology Perspective
Aravind Kumar Namasivayam, Deirdre Coleman, Aisling O’Dwyer and
Pascal van Lieshout
Frontiers in Psychology 4 August 2020 | Models and Theories of Speech Production
EDITORIAL
published:19June2020
doi:10.3389/fpsyg.2020.01238
Editorial: Models and Theories of
Speech Production
AdamantiosGafos1* andPascalvanLieshout2*
1DepartmentofLinguisticsandExcellenceAreaofCognitiveSciences,UniversityofPotsdam,Potsdam,Germany,
2DepartmentofSpeech-LanguagePathology,OralDynamicsLaboratory,UniversityofToronto,Toronto,ON,Canada
Keywords:speechproduction,motorcontrol,dynamicalmodels,phonology,speechdisorders,timing
EditorialontheResearchTopic
ModelsandTheoriesofSpeechProduction
Spokenlanguageisconveyedviawell-coordinatedspeechmovements,whichactascoherentunits
of control referred to as gestures. These gestures and their underlying movements show several
distinctivepropertiesintermsoflawfulrelationsamongtheparametersofduration,relativetiming,
range of motion, target accuracy, and speed. However, currently, no existing theory successfully
accountsforallpropertiesofthesemovements.Eventhoughmodelsinspeechmotorcontrolin
thelast40yearshaveconsistentlytakeninspirationfromgeneralmovementscience,someofthe
comparisonsremainill-informed.Forexample,ourpresentknowledgeonwhetherwidelyknown
principlesthatapplytolimbmovements(e.g.,thespeed-accuracytradeoffknownasFitts’law)also
holdtrueforspeechmovementsisstillverylimited.Anunderstandingoftheprinciplesthatapply
tospeechmovementsiskeytodefiningthesomewhatelusiveconceptofspeechmotorskillandto
assessingandinterpretingdifferentlevelsofthatskillinpopulationswithandwithoutdiagnosed
speechdisorders.Thelatterissuetapsintofundamentaldebatesaboutwhetherspeechpathology
assessmentparadigmsneedtoberestrictedtocontrolregimesthatarespecifictothoseunderlying
Editedandreviewedby: typicalspeechproductions.Resolutionofsuchdebatescruciallyreliesonourunderstandingofthe
NielsO.Schiller, natureofspeechprocessesandtheunderlyingcontrolunits.
LeidenUniversity,Netherlands Unlikemovementsinlocomotionoroculomotorfunction,speechmovementswhencombined
*Correspondence: into gestures are not mere physical instantiations of organs moving in space and time but, also,
AdamantiosGafos have intrinsic symbolic function. Language-particular systems, or phonological grammars, are
[email protected] involvedinthepatterningofthesegestures.Grammarconstraintsregulatethepermissiblesymbolic
PascalvanLieshout
combinationsasevidencedviaelicitingjudgmentsonwhetheranygivensequenceiswell-formedin
[email protected]
anyparticularlanguage(thesamesequencecanbeacceptableinone,butnottheotherlanguage).In
whatwaystheseconstraintsshapespeechgesturesandhowthesefitwithexistinggeneralprinciples
Specialtysection:
ofmotorcontrolis,also,notclearlyunderstood.
Thisarticlewassubmittedto
LanguageSciences, Furthermore,speechgesturesarepartsofwordsandthusonewindowintounderstandingthe
asectionofthejournal natureofthespeechproduction1systemistoobservespeechmovementsaspartsofwordsorlarger
FrontiersinPsychology chunks of speech such as phrases or sentences. The intention to produce a lexical item involves
Received:14April2020 activatingsequencesofgesturesthatarepartofthelexicalitem.Theregulationintimeoftheunits
Accepted:12May2020 insuchsequencesraisesmajorquestionsforspeechmotorcontroltheories(butalsofortheories
Published:19June2020
1Oneofourreviewersnotesthatinthefieldofpsycholinguisticsthetermspeechproductionisusedmorebroadly(thanin
Citation:
theuseofthetermimpliedbythecontributionstothisResearchTopic)and,pointsouttheneed,aptlystated,“tobridge
GafosAandvanLieshoutP(2020)
thegapbetweenpsycholinguisticallyinformedphoneticsandphoneticallyinformedpsycholinguistics.”Wefullyconcurand
Editorial:ModelsandTheoriesof lookforwardtofutureresearcheffortsandperhapsResearchTopicsdevotedtosuchbridging.Forarecentspecialissue
SpeechProduction. onpsycholinguisticapproachestospeechproduction,seeMeyeretal.(2019)andforamorefocusedreviewoftheissues
Front.Psychol.11:1238. pertinentto“phoneticencoding”(aterminpsycholinguisticsroughlyequivalenttoouruseofthetermspeechproductionin
doi:10.3389/fpsyg.2020.01238 thepresentResearchTopic)seeLaganaro(2019).
FrontiersinPsychology|www.frontiersin.org 15 June2020|Volume11|Article1238
GafosandvanLieshout Editorial:ModelsandTheoriesofSpeechProduction
ofcognitionandsequentialactioningeneral).Majorchallenges componentofabstract,symbolicphonologicalrepresentations
are met in the inter-dependence among different time is kept apart from the way(s) in which these representations
scales related to gestural planning, movement execution areimplementedinquantitativetermswhichincludesurface
andcoordinationwithinandacrossdomainsofindividuallexical durationspecificationsandattendanttimingmechanismsfor
items. How these different time scales interact and how their achievingthese.
interaction affects the observed movement properties is for the ShawandCheninvestigatedtowhatdegreetimingbetween
mostpartstillunknown. gestures is stable across variations in the spatial positions of
In this special issue, we present a variety of theoretical and individualarticulators,aspredictedinArticulatoryPhonology.
empiricalcontributionswhichexplorethenatureofthedynamics Using Electromagnetic Articulography with a group of
ofspeechmotorcontrol.Forpracticalpurposes,weseparatethese MandarinspeakersproducingCVmonosyllables,theyfound
contributionsintwomajorthemes: acorrelationbetweentheinitialpositionofthetonguegesture
for the vowel and C-V timing. In contrast to the original
1) Modelsandtheoriesofspeechproduction.
hypothesis,thisindicatesthatinter-gesturaltimingissensitive
2) Applications.
tothepositionofthearticulators,suggestingacriticalrolefor
Following is a short description of each paper as listed under somatosensoryfeedback.
thesethemes. Roessig and Mücke study tonal and kinematic profiles
of different degrees of prominence (unaccented, broad,
1) Modelsandtheoriesofspeechproduction
narrow and contrastive focus) from 27 speakers of German.
The speech signal is simultaneously expressed in two
Parameters in both the tonal and kinematic dimensions are
information-encoding systems: articulation and acoustics.
shown to vary systematically across degrees of prominence.
Goldstein’s contribution addresses the relation between
A dynamical approach is put forward in modeling these
representationsinthesetwoparallelmanifestationsofspeech
findings.Thisapproachembracesthemultidimensionalityof
while focusing not on static properties but on patterns of
prosodywhileatthesametimeshowinghowbothdiscreteand
change over time (temporal co-modulation) in these two
continuousmodificationsinfocusmarkingcanbeexpressed
channels.Todoso,Goldsteinquantifiestherelationbetween
withinoneformallanguage.Themodelcapturesqualitatively
rates of change in the parallel acoustic and articulatory
the observed patterns in the data by tuning of an abstract
representations of the same utterance, produced by various
control variable which shapes the attractor landscape over
speakers, based on x-ray microbeam data. Analysis of this
the parameter space of kinematic and tonal dimensions
relation indicates that the two representations are correlated
consideredinthiswork.
via a pulse-like modulation structure, with local correlations
Iskarous provides a computational approach to explain
beingstrongerthanglobalones.Thismodulationseemslinked
the nature of spatiotemporal particulation of the vocal tract,
tothefundamentalunitofthesyllable.
as evidenced in the production of speech gestures. Based
It is widely assumed that acoustic parameters for vowels
on a set of reaction-diffusion equations with simultaneous
are normally distributed, but it is rarely demonstrated that
TuringandHopfpatternsthecriticalcharacteristicsofspeech
this might be the case. Whalen and Chen quantified the
gesturesrelatedtovocaltractconstrictionscanbereplicatedin
distributionsofF1andF2valuesof/i/and/o/intheEnglish
supportofthenotionthatmotorprocessescanbeseenasthe
words“heed,”“geek,”“ode”/“owed,”and“dote”producedbya
emergence of low degree of freedom descriptions from high
singlespeakeronthreedifferentdays.Analysisbasedonahigh
degreeoffreedomsystems.
numberofrepetitionsofthesevowelsindifferentconsonantal
Patri et al. address individual differences in responses to
contexts indicates that distributions are generally normal,
auditoryorsomatosensoryperturbationinspeechproduction.
whichinturnsuggestsconsistentvowel-specifictargetsacross
Two accounts are entertained. The first reduces individual
different contextual environments. The results add weight
differencestodifferencesinacuityofthesensoryspecifications
to the widely-held assumption that speech targets follow a
while the second leaves sensory specifications intact and,
normaldistributionandtheauthorsdiscusstheimplications
instead, modulates the sensitivity of match between motor
fortheoriesofspeechtargets.
commandsandtheirauditoryconsequences.Whilesimulation
TurkandShattuck-Hufnageladdressthenatureoftiming
results show that both accounts lead to similar results, it is
inspeech,withspecialattentiongiventomovementendpoints,
arguedthatmaintainingintactsensoryspecificationsismore
which as they argue relate to the goals of these movements.
flexible, enabling a more encompassing approach to speech
Theargumentispresentedthatthesepointsrequirededicated
variability where cognitive, attentional and other factors can
control regimes. Evidence for this argument is derived
modulateresponsestoperturbations.
from work in both speech and non-speech motor control. Oneofthefoundationalideasofphonologyandphonetics
It is also argued that in contrast to the Articulatory is that produced and perceived utterances are decomposed
Phonology/TaskDynamicsview,wheregesturaldurationsare into sequences of discrete units. However, evidence from
determined by an intrinsic dynamics, duration must be an development indicates that in child speech utterances are
independently controlled variable in speech. A phonology- holistic rather than segmented. The contribution by Davis
extrinsiccomponentisthusproposedtobenecessaryandacall and Redford offers a theoretical demonstration along with
ismadefordevelopingandtestingmodelsofspeechwherea attendantmodelingthatthepositedunitscanemergefroma
FrontiersinPsychology|www.frontiersin.org 26 June2020|Volume11|Article1238
GafosandvanLieshout Editorial:ModelsandTheoriesofSpeechProduction
stageofspeechwherewordsorphrasesstartoffastime-aligned Parrell and Lammert develop a synthesis of the dynamic
motoric and perceptual trajectories. As words are added and movement primitives model of motor control (Schaal et al.,
repeatedlyrehearsedbythelearner,motorictrajectoriesbegin 2007; Ijspeert et al., 2013) with the task dynamics model
todeveloprecurrentarticulatoryconfigurationswhich,when of speech production (Saltzman and Munhall, 1989). A key
coupledwiththeircorrespondingperceptualrepresentations, element in achieving this synthesis is the incorporation of
give rise to perceptual-motor units claimed to characterize a learnable forcing term into the task dynamics’ point-
maturespeechproduction. attractorsystem.Thepresenceofsuchatunabletermendows
In their contribution, Kearney et al. present a task dynamics with flexibility in movement trajectories. The
simplified version of the DIVA model, focusing on three proposed synthesis also establishes a link to optimization
fitting parameters related to auditory feedback control, approaches to motor control where the forcing term can be
somatosensory feedback control, and feedforward control. seen to minimize a cost function over the timespan of the
The model is tested through computer simulations that movementunderconsideration(e.g.,minimizingtotalenergy
identify optimal model fits to six existing sensorimotor expendedduringareachingmovement).Thedynamicsofthe
adaptation datasets, showing excellent fits to real data across proposed synthesis model are explicitly described and their
differenttypesofperturbationsandexperimentalparadigms. effects are demonstrated in the form of proof of concept
An active area in phonological theory is the investigation simulationsshowingtheconsequencesofperturbationsonjaw
of long-distance assimilation where features of a phoneme movementtrajectories.
assimilatetofeaturesofanothernon-adjacentphoneme.Tilsen 2) Applications
seeks to identify mechanisms for the emergence of such Noiray et al. present a study in which they examined
non-local assimilations in speech planning and production whether phonemic awareness correlates with coarticulation
models. Two mechanisms are proposed. The first is one degree, commonly used as a metric for estimating the size
where a gesture is either anticipatorily selected in an earlier of children’s production units. A speech production task
epoch or is not suppressed (after being selected) so that its was designed to test for developmental differences in intra-
influence extends to later epochs. The second is one where syllabic coarticulation degree in 41 German children from
gestures which may be active in one epoch of a planning- 4 to 7 years of age, using ultrasound imaging. The results
level dynamics, even though not selected during execution, suggestthattheprocessofdevelopingspokenlanguagefluency
maystillinfluenceproduction ina differentepoch.Evidence involvesdynamicalinteractionsbetweencognitiveandspeech
forthesemechanismsisfoundinbothspeechandnon-speech motordomains.
movementpreparationparadigms.Theexistenceofthesetwo Tiede et al. describe a study in which they tracked
mechanisms is argued to account for the major dichotomy movements of the head and speech articulators during
betweenassimilationphenomenathathavebeendescribedas an alternating word pair production task driven by an
involving the extension of an assimilating property vs. those accelerating rate metronome. The results show that as
thatcannotbesodescribed. production effort increased, so did speaker head nodding,
Xu and Prom-on contrast two principles assumed to and that nodding increased abruptly following errors. The
underlie the dynamics of movement control: economy of strongest entrainment between head and articulators was
effort and maximum rate of information. They present observedatthefastestrateundercodaalternationconditions.
data from speakers of American English on repetitive Namasivayam et al. present an Articulatory Phonology
syllable sequences who were asked to imitate recordings of approach for understanding the nature of Speech Sound
the same sequences that had been artificially accelerated Disorders (SSDs) in children, aiming to reconcile the
and to produce meaningful sentences containing the same traditional phonetic-phonology dichotomy with the concept
syllables at normal and fast speaking rates. The results of interconnectedness between these levels. They present
show that the characteristics of the formant trajectories evidence supporting the notion of articulatory gestures at
they analyzed fit best the notion of the maximum rate of the level of speech production and how this is reflected in
informationprinciple. control processes in the brain. They add an overview of
Kröger et al.’s contribution offers a demonstration that a how an articulatory “gesture”-based approach can account
learning model based on self-organizing maps can serve as for articulatory behaviors in typical and disordered speech
bridge between models of the mental lexicon and models production, concluding that the Articulatory Phonology
of sensorimotor control and that such a model can learn approach offers a productive strategy for further research in
(from semantic, auditory and somatosensory information) thisarea.
representationalunitsakintophonetic-phonologicalfeatures. Heyne et al. address the relation between speech
At a broad level, few efforts have been made to bridge and another oral motor skill, trombone playing. Using
theory and modeling of the lexicon and motor control. ultrasound, they recorded midsagittal tongue shapes from
The proposed model aims at addressing that gap and New Zealand English and Tongan-speaking trombone
makes predictions about the specificity and rate of growth players. Tongue shapes from the two language groups were
of such representational features under different training estimated via fits with generalized additive mixed models,
conditions (auditory only vs. auditory and somatosensory while these speakers/players produced vowels (in their
trainingmodes). native languages) and sustained notes at different pitches
FrontiersinPsychology|www.frontiersin.org 37 June2020|Volume11|Article1238
GafosandvanLieshout Editorial:ModelsandTheoriesofSpeechProduction
and intensities. The results indicate that, while airflow models with other aspects of cognition; and finally, to the
production and requisite acoustics largely constrain vocal potential of theoretical models in informing applications of
tract configuration during trombone playing, evidence for a speechproductionindisorderedspeechandmotorskillsinother
secondaryinfluencefromspeechmotorconfigurationscanbe oralactivitiessuchasplayingmusicalinstruments.
discernedinthatthetwogroupstendedtousedifferenttongue
AUTHOR CONTRIBUTIONS
configurations resembling distinct vocalic monopthongs in
theirrespectivelanguages.
Allauthorslistedhavemadeequalcontributionstotheworkand
The papers assembled for this Research Topic attest to the approveditforpublication.
advantages of combining theoretical and empirical approaches
to the study of speech production. They also attest to the ACKNOWLEDGMENTS
value of formal modeling in addressing long-standing issues in
speechdevelopmentandtherelationshipbetweenmotorcontrol AG’sworkhasbeensupportedbytheEuropeanResearchCouncil
andphonologicalpatterns;totheimportanceofsomatosensory (AdG249440)andtheDeutscheForschungsgemeinschaft(DFG,
and auditory feedback in planning and monitoring speech German Research Foundation) - Project ID 317633480 - SFB
productionandtheimportanceofintegratingspeechproduction 1287,ProjectC04.
REFERENCES
Schaal,S.,Mohajerian,P.,Ijspeert,A.J.,Cisek,P.,Drew,T.,andKalaska,J.F.
(2007).Dynamicssystemsvs.Optimalcontrolaunifyingview.InProgressin
Ijspeert, A. J., Nakanishi, J., Hoffmann, H., Pastor, P., and Schaal, S. (2013). BrainResearch165,425–45.doi:10.1016/S0079-6123(06)65027-9
Dynamical movement primitives: learning attractor models for motor
behaviors.NeuralComputation,25,328–73.doi:10.1162/NECO_a_00393 ConflictofInterest:Theauthorsdeclarethattheresearchwasconductedinthe
Laganaro, M. (2019). Phonetic encoding in utterance production: a review of absenceofanycommercialorfinancialrelationshipsthatcouldbeconstruedasa
open issues from 1989 to 2018. Language Cognit. Neurosci. 34, 1193–1201. potentialconflictofinterest.
doi:10.1080/23273798.2019.1599128
Meyer, A. S., Ardi, R., and Laurel, B. (2019). Thirty years of speaking: an Copyright©2020GafosandvanLieshout.Thisisanopen-accessarticledistributed
introductiontotheSpecialIssue.LanguageCognit.Neurosci.34,1073–1084. underthetermsoftheCreativeCommonsAttributionLicense(CCBY).Theuse,
doi:10.1080/23273798.2019.1652763 distribution or reproduction in other forums is permitted, provided the original
Saltzman, E. L., and Munhall, K. G. (1989). A dynamical approach to author(s)andthecopyrightowner(s)arecreditedandthattheoriginalpublication
gestural patterning in speech production. Ecological Psychology, 1, 333–82. in this journal is cited, in accordance with accepted academic practice. No use,
doi:10.1207/s15326969eco0104_2 distributionorreproductionispermittedwhichdoesnotcomplywiththeseterms.
FrontiersinPsychology|www.frontiersin.org 48 June2020|Volume11|Article1238
fpsyg-10-01462 July9,2019 Time:17:38 #1
ORIGINALRESEARCH
published:10July2019
doi:10.3389/fpsyg.2019.01462
Emergence of an Action Repository
as Part of a Biologically Inspired
Model of Speech Processing: The
Role of Somatosensory Information
in Learning Phonetic-Phonological
Sound Features
BerndJ.Kröger1*,TanyaBafna2andMengxueCao3
1NeurophoneticsGroup,DepartmentofPhoniatrics,Pedaudiology,andCommunicationDisorders,MedicalSchool,RWTH
AachenUniversity,Aachen,Germany,2MedicalSchool,RWTHAachenUniversity,Aachen,Germany,3SchoolofChinese
LanguageandLiterature,BeijingNormalUniversity,Beijing,China
A comprehensive model of speech processing and speech learning has been
established. The model comprises a mental lexicon, an action repository and an
articulatory-acoustic module for executing motor plans and generating auditory and
somatosensory feedback information (Kröger and Cao, 2015). In this study a “model
Editedby: language” based on three auditory and motor realizations of 70 monosyllabic words
AdamantiosGafos,
has been trained in order to simulate early phases of speech acquisition (babbling
UniversitätPotsdam,Germany
and imitation). We were able to show that (i) the emergence of phonetic-phonological
Reviewedby:
JoanaCholin, features results from an increasing degree of ordering of syllable representations within
BielefeldUniversity,Germany
the action repository and that (ii) this ordering or arrangement of syllables is mainly
JasonW.Bohland,
BostonUniversity,UnitedStates shapedbyauditoryinformation.Somatosensoryinformationhelpstoincreasethespeed
*Correspondence: oflearning.Especiallyconsonantalfeatureslikeplaceofarticulationarelearnedearlierif
BerndJ.Kröger auditoryinformationisaccompaniedbysomatosensoryinformation.Itcanbeconcluded
[email protected];
that somatosensory information as it is generated already during the babbling and the
[email protected]
imitation phase of speech acquisition is very helpful especially for learning features
Specialtysection:
like place of articulation. After learning is completed acoustic information together with
Thisarticlewassubmittedto
LanguageSciences, semanticinformationissufficientfordeterminingthephonetic-phonologicalinformation
asectionofthejournal from the speech signal. Moreover it is possible to learn phonetic-phonological features
FrontiersinPsychology
like place of articulation from auditory and semantic information only but not as fast as
Received:09January2019
whensomatosensoryinformationisalsoavailableduringtheearlystagesoflearning.
Accepted:07June2019
Published:10July2019
Keywords: neural model simulation, speech production and acquisition, speech perception, neural self-
Citation: organization,connectionismandneuralnets
KrögerBJ,BafnaTandCaoM
(2019)EmergenceofanAction
RepositoryasPartofaBiologically INTRODUCTION
InspiredModelofSpeechProcessing:
TheRoleofSomatosensory
Speakingstartswithamessagewhichthespeakerwantstocommunicate,followedbyanactivation
InformationinLearning
ofconcepts.Thisprocessiscalledinitiation.Subsequentlyconceptsactivatewordswhichmaybe
Phonetic-PhonologicalSound
Features.Front.Psychol.10:1462. inflected and ordered within a sentence with respect to their grammatical and functional role.
doi:10.3389/fpsyg.2019.01462 Thisprocessiscalledformulationandstartswiththeactivationoflemmasinthementallexicon
FrontiersinPsychology|www.frontiersin.org 19 July2019|Volume10|Article1462