LabPhon15 Program


Download Conference Program booklet as a PDF

Oral Sessions
Production dynamics (Thursday morning)
How local computation leads to global structure: the dynamics of gesture and word
Invited speaker: Khalil Iskarous
[short abstract]
Production Effects in Light of Perceptual Evaluation: Tempo Effects for Phonologization
Kenneth de Jong and Kyoko Nagao
[short abstract] [full abstract]
Probing the interaction of dynamic stability with grammar: Evidence from Russian
Marianne Pouplier, Stefania Marin, Conceição Cunha and Alexei Kochetov
[short abstract] [full abstract]
Phonetic reduction, perceptual illusions, and phonotactic legality
Alex McAllister and Matthew T. Carlson
[short abstract] [full abstract]
Invited discussant: Elizabeth Zsiga

Perception dynamics (Thursday afternoon)
How are forms we rarely hear, understood so easily?
Invited speaker: Meghan Sumner
[short abstract]
Speech Prediction from Subphonemic Production
Donald Derrick and Daniel Bürkle
[short abstract] [full abstract]
Timing lag matters in the perception of Georgian stop sequences by native speakers
Ioana Chitoran and Harim Kwon
[short abstract] [full abstract]
Relative cue weighting in perception and production of a sound change in progress
Jianjing Kuang and Aletheia Cui
[short abstract] [full abstract]
Invited discussant: Jennifer Hay

Prosodic organization (Friday morning)
What does prosodic variation tell us about prosodic organization?
Invited speaker: Yiya Chen
[short abstract]
The Nature of Variation in the Tone Sandhi Patterns of Shanghai Wu
Hanbo Yan and Jie Zhang
[short abstract] [full abstract]
Non-canonical Word Order in Russian: Processing and Acoustic Parameterization
Tatiana Luchkina
[short abstract] [full abstract]
Individual specificity, redundancy and the evolution of phonological systems: Intonation in a tone language
Francesco Cangemi, Christian Weitz, Kieu-Phuong Ha, Marc Brunelle and Martine Grice
[short abstract] [full abstract]
Invited discussant: Martine Grice

Lexical dynamics and memory (Friday afternoon)
Representational dynamics in sound structure planning
Invited speaker: Matthew Goldrick
[short abstract] [full abstract]
Coalescing sources of bias in perception: Lexical and prelexical influences on the processing of phonological features
Alexander Martin and Sharon Peperkamp
[short abstract] [full abstract]
Modelling Phonetic and Phonological Variation with 'Small' Data: Evidence from Kaqchikel Mayan
Ryan Bennett and Kevin Tang
[short abstract] [full abstract]
Epenthesis into nonnative consonant clusters: phonetic factors eclipse gradient phonotactics
Colin Wilson and Lisa Davidson
[short abstract] [full abstract]
Invited discussant: Keith Johnson

Acquisition and life-span (Saturday morning)
More is more: how trying to learn multiple aspects of language at once can actually help
Invited speaker: Sharon Goldwater
[short abstract]
The effect of indexical cues on the distributional learning of sound categories
Masaki Noguchi and Carla Hudson Kam
[short abstract] [full abstract]
L1 influence on L2 assimilation: An EPG study of English /n/+stop sequences
Laura Colantoni, Alexei Kochetov and Jeffrey Steele
[short abstract] [full abstract]
/l/ in clusters: an articulatory-acoustic study of children’s productions
Susan Lin, Sharon Inkelas, Lara Mcconnaughey and Michael Dohn
[short abstract] [full abstract]
Invited discussant: Stefan Frisch

Social network dynamics (Saturday afternoon)
Social dynamics and phonological representations: Observations from speech and society in Scotland
Invited speaker: Jane Stuart-Smith
[short abstract]
What do you expect from an unfamiliar talker?
Dave F Kleinschmidt and T. Florian Jaeger
[short abstract] [full abstract]
Effects of phonetic reduction and social factors on cross-modal lexical priming
Zack Jones and Cynthia G. Clopper
[short abstract] [full abstract]
Network structural equivalence and the reversal of the Southern Vowel Shift
Robin Dodsworth and Richard Benton
[short abstract] [full abstract]
Invited discussant: Erik Thomas

POSTERS (alphabetical by first author):

Coarticulation magnitude in German children
Dzhuma Abakarova, Elina Rubertus, Khalil Iskarous, Mark Tiede, Jan Ries and Aude Noiray
[short abstract] [full abstract]
Speaking rate effect on consonant-vowel coarticulation and on stop consonant classification
Mohammad Abuoudeh and Olivier Crouzet
[short abstract] [full abstract]
The Role of Tongue Root in Laryngeal Contrasts: An Ultrasound Study of English, Spanish, Hindi, and Korean
Suzy Ahn
[short abstract] [full abstract]
An acoustic investigation of the categorical and gradient spread of pharyngealization in Urban Najdi Arabic
Abdulaziz Alarifi and Benjamin V. Tucker
[short abstract] [full abstract]
Effects of Position, Stress and Manner of Articulation on Consonant-Vowel Co-occurrence in Three Languages
Eleonora Albano
[short abstract] [full abstract]
Regressive Coarticulation of Pharyngealization in Arabic Spoken Word Recognition
Sawsan Alwabari and Tania Zamuner
[short abstract] [full abstract]
Perception, mental representations and production of non-native prosodic contrasts
Yuki Asano
[short abstract] [full abstract]
Is there cross-linguistic evidence for prosodic bootstrapping of word order?
Angeliki Athanasopoulou and Irene Vogel
[short abstract] [full abstract]
Category typicality in perceptual learning
Molly Babel, Michael McAuliffe, Zoe Lawler and Carolyn Norton
[short abstract] [full abstract]
Using Developmental Data to Explore Frequency and Neighborhood Density Effects in Production
Melissa Baese-Berk and Katherine White
[short abstract] [full abstract]
Rapid adaptation to foreign-accented speech: A web-based replication of Clarke and Garrett (2004)
Larisa Bainton, Emily Rowe, Zach Burchill, Linda Liu, Kodi Weatherholtz and T. Florian Jaeger
[short abstract] [full abstract]
The role of palate shape in individual articulatory and acoustic variability
Sarah Bakst
[short abstract] [full abstract]
Structured Variation across Sound Contrasts, Talkers, and Speech Styles
Hye-Young Bang and Meghan Clayards
[short abstract] [full abstract]
Deformation-based articulatory representations of speech sounds
Marissa Barlaz, Ryan Shosted, Christopher Carignan, Maojing Fu, Zhi-Pei Liang and Brad Sutton
[short abstract] [full abstract]
Final lengthening in German
Malte Belz, Oxana Rasskazova, Anja Riemenschneider, Jelena Krivokapić, Melanie Weirich and Christine Mooshammer
[short abstract] [full abstract]
Word-final geminates in a Modern South Arabian language: Phonetics and Phonology
Sabrina Bendjaballah and Rachid Ridouane
[short abstract] [full abstract]
Language-internal behavior of typologically rare sounds: Production, perception, and distribution of breathy sonorants in Marathi
Kelly Berkson
[short abstract] [full abstract]
Style-shifting and phonetic alignment in non-native discourse
Grant M. Berry and Mirjam Ernestus
[short abstract] [full abstract]
Morpho-phonology without semantics? The roles of lexical memory and experience in influencing the nature of lexical representations in a rote learning context
Siti Syuhada Binte Faizal and Ghada Khattab
[short abstract] [full abstract]
Cider apples and jeder Abend: the interplay of /r/-sandhi and word-initial glottalisation in English-accented German
Maria Paola Bissiri and James M. Scobbie
[short abstract] [full abstract]
Categorical perception of lexical stress: A cross-linguistic study
Natalie Boll-Avetisyan, Saioa Larraza, Aislyn Rose, Sylvie Margules, Ranka Bijeljac-Babic, Thierry Nazzi and Barbara Höhle
[short abstract] [full abstract]
When intonation fails to phonologize: the case of Southern Vietnamese
Marc Brunelle
[short abstract] [full abstract]
Effect of phrasal accent on vocalic and consonantal nuclei
Lia Saki Bucar Shigemori, Marianne Pouplier and Štefan Beňuš
[short abstract] [full abstract]
Levels of processing and their interaction in speech production
Adam Buchwald and Michele Miozzo
[short abstract] [full abstract]
Building a proto-lexicon: Does input variability matter?
Helen Buckler and Elizabeth Johnson
[short abstract] [full abstract]
Modeling new conceptions of functional load with perceptual confusability
Zachary Burchill
[short abstract] [full abstract]
Dynamic hyper-articulation: Effects of context and lexical representations
Esteban Buz, Scott Seyfarth and T Florian Jaeger
[short abstract] [full abstract]
Uncovering the Origins of Nucleus Raising in Liverpool English: Dynamic Analysis of Diphthongs
Amanda Cardoso
[short abstract] [full abstract]
L1 USE PREDICTS IMITATION OF METRICAL FEATURES IN A TYPOLOGICALLY DIFFERENT L2
Rossana Cavone and Mariapaola D'Imperio
[short abstract] [full abstract]
Accent-meter/tune alignment in Japanese vocal music
Sunghye Cho
[short abstract] [full abstract]
An empirical and computational study of generalized adaptation to natural talker-specific VOT
Eleanor Chodroff, Alessandra Golden and Colin Wilson
[short abstract] [full abstract]
Estimating segments' cost using cross-linguistic information
Uriel Cohen Priva and Emily Gleason
[short abstract] [full abstract]
Perception of acoustic, informational and structural prominence in English, French, and Spanish
Jennifer Cole, Jose Ignacio Hualde, Caroline Smith, Christopher Eager, Tim Mahrt and Ricardo Napoleão de Souza
[short abstract] [full abstract]
Acoustic-phonetic modelling of historical and prehistoric sound change
John Coleman
[short abstract] [full abstract]
Vowel lengthening in syllables without vowels
Jamison Cooper-Leavitt and Rachid Ridouane
[short abstract] [full abstract]
Like a square peg in a round hole: Why contour shape matters for learning new intonation patterns
Mariapaola D'Imperio and James Sneed German
[short abstract] [full abstract]
When oui becomes ou[i̥]: The role of vowel type, preceding consonant and lexical frequency on total final vowel devoicing in Continental French
Amanda Dalola
[short abstract] [full abstract]
Perception of anticipatory labial coarticulation by Belgian French blind listeners: A comparison with sighted listeners in audio-only, visual-only and audiovisual conditions
Véronique Delvaux, Kathy Huet, Myriam Piccaluga and Bernard Harmegnies
[short abstract] [full abstract]
Dynamics and articulatory control in Amharic ejectives
Didier Demolin
[short abstract] [full abstract]
Processing of coarticulatory nasalized vowels and phonological nasal vowels in Canadian French
Félix Desmeules-Trudel and Tania Zamuner
[short abstract] [full abstract]
Prosodic phonological characteristics of speech directed to adults and to infants with and without hearing impairment
Laura Dilley, Evamarie Burnham, Elizabeth Wieland, Derek Houston, Maria Kondaurova and Tonya Bergeson
[short abstract] [full abstract]
Social dynamics and phonological strength: Post-nasal devoicing in Tswana
Grzegorz Dogil, Jagoda Bruni, Daniel Duran, Justus Roux and Andries Coetzee
[short abstract] [full abstract]
Acquisition of word-level prominence in L2 English by Canadian French speakers
Guilherme Duarte Garcia and Natalia Brambatti Guzzo
[short abstract] [full abstract]
Phonology Modulates the Illusory Vowels in Perceptual Illusions
Karthik Durvasula, Ho-Hsin Huang, Sayako Uehara, Qian Luo and Yen-Hwei Lin
[short abstract] [full abstract]
Using ANNs for vowel identification from V-to-V coarticulation in non-harmonic VCV sequences
Indranil Dutta, Irfan S. and Harsha K.R.S.
[short abstract] [full abstract]
Contrast preservation at the level of the individual: Evidence from Spanish plosive lenition
Christopher D. Eager
[short abstract] [full abstract]
Rethinking reduction on the basis of phonetic variation in a discourse marker
Mirjam Ernestus and Rachel Smith
[short abstract] [full abstract]
Articulatory uniformity in Suzhou fricative vowels
Matthew Faytak
[short abstract] [full abstract]
The effects of listener age and language experience on talker identification
Natalie Fecher, Katrina Aranas and Elizabeth K. Johnson
[short abstract] [full abstract]
Variable aspiration of Spanish coda /s/: Laboratory evidence and Stochastic OT modeling
Valentyna Filimonova and Kelly Berkson
[short abstract] [full abstract]
The phonological representation of Japanese vowel devoicing
Marco Fonseca, Maria Cantoni and Thaïs Cristófaro Silva
[short abstract] [full abstract]
Regional variation in formant dynamics and the phonologization of pre-velar raising in American English
Michael J. Fox and Jeff Mielke
[short abstract] [full abstract]
Prosodic Organization of Spontaneous Spanish-English Bilingual Speech
Melinda Fricke, Marianna Nadeu and Michael Maslowski
[short abstract] [full abstract]
Prosodic organization and microprosodic effects in Shanghai Chinese
Jiayin Gao and Pierre André Hallé
[short abstract] [full abstract]
Embodied dynamics: A unified approach to local, non-local and global coarticulation
Bryan Gick, Chenhao Chiu and Ian Stavness
[short abstract] [full abstract]
Shanghai Chinese obstruent durations vary with voicing: A phonological or phonetic effect?
Pierre Hallé and Jiayin Gao
[short abstract] [full abstract]
Phonological Influence in Third Language Acquisition: L2 Spanish Effects on the Production of L3 Portuguese Voiced Stops
Sarah Harper
[short abstract] [full abstract]
Patterns of vowel laxing and vowel harmony in Peninsular Spanish
Nicholas Henriksen
[short abstract] [full abstract]
Representations of Place and Airstream Mechanism: A real-time MRI study of Tigrinya ejectives
Zainab Hermes, Mao-Jing Fu, Sharon Rose, Ryan Shosted and Brad Sutton
[short abstract] [full abstract]
The Phonetics and Phonology of Fataluku Intonational Downstep
Tyler Heston
[short abstract] [full abstract]
Post-pubescent long-term exposure to non-rhoticity causes qualitative and quantitative changes in the realization of postvocalic /r/
Marie-Christin Himmel and Baris Kabak
[short abstract] [full abstract]
Downstep in Japanese revisited: Lexical category matters
Manami Hirayama and Hyun Kyung Hwang
[short abstract] [full abstract]
Adaptive dispersion: a perceptual motivation for sound change
Phil Howson and Philip Monahan
[short abstract] [full abstract]
English coda [m] adaptations in Standard Mandarin loanwords: Corpora data vs. bilingual and monolingual experimental results
Ho-Hsin Huang and Yen-Hwei Lin
[short abstract] [full abstract]
A crosslinguistic Study of Vowel Categorization: Data from Canadian English, Korean and Japanese
Hyun Kyung Hwang
[short abstract] [full abstract]
Stress clash avoidance by 6- to 7-month-olds
Barbara Höhle, Natalie Boll-Avetisyan and Jürgen Weissenborn
[short abstract] [full abstract]
Language-specificity in Speakers’ Strategies of Focus Expression
Martin Ho Kwan Ip and Anne Cutler
[short abstract] [full abstract]
Intonation and Sentence Type: The Emergence of Conventions for Attitudinal Meanings
Sunwoo Jeong
[short abstract] [full abstract]
The perception of stop/sibilant clusters in Modern Hebrew
Kyle Jones
[short abstract] [full abstract]
Incomplete neutralization and the (a)symmetry of paradigm uniformity
Abby Kaplan
[short abstract] [full abstract]
Focus, accentuation and phonetic variability in Greek
Argyro Katsika and Amalia Arvaniti
[short abstract] [full abstract]
Lenition and segmentation
Jonah Katz and Melinda Fricke
[short abstract] [full abstract]
Targetless /u/ in Tokyo Japanese
Shigeto Kawahara, Jason Shaw and James Whang
[short abstract] [full abstract]
Intonational qualities of strong and weak imperatives
Megan Keough, Elise Kedersha McClay, Molly Babel and Lisa Matthewson
[short abstract] [full abstract]
Locality and variability in cross-word alternations: a production planning account
Oriana Kilbourn-Ceron, Michael Wagner and Meghan Clayards
[short abstract] [full abstract]
ePGG, Pio, airflow and acoustic data on the phasing of glottal opening and three-way phonation contrast: implications for laryngeal features
Hyunsoon Kim, Shinji Maeda, Kiyoshi Honda and Lise Crevier-Buchman
[short abstract] [full abstract]
Lexical access and stereotypical 'word age' in Korean
Jonny Kim
[short abstract] [full abstract]
Linguistic contrast enhancement under prosodic strengthening in L1 and L2 speech
Sahyang Kim, Jiyoun Choi and Taehong Cho
[short abstract] [full abstract]
Individual differences in second language speech perception across tasks and contrasts
Donghyun Kim, Meghan Clayards and Heather Goad
[short abstract] [full abstract]
Prosodic Accommodation in Seoul Korean Accentual Phrases
Jiseung Kim
[short abstract] [full abstract]
Perception of syllable stress varies by listener
Amelia Kimball and Jennifer Cole
[short abstract] [full abstract]
Obstruent voicing, aspiration, and tone: implications for laryngeal phonology
James Kirby
[short abstract] [full abstract]
Attentional modulation and individual differences in explaining the changing role of f0 in the Korean laryngeal stop perception
Eun Jong Kong and Hyunjung Lee
[short abstract] [full abstract]
Vocal tract and manual gesture coordination in prosodic structure
Jelena Krivokapic, Mark Tiede and Martha Tyrone
[short abstract] [full abstract]
The perceptual representation of place and voice in Russian. Evidence from eye-tracking
Martin Krämer and Natalia Mitrofanova
[short abstract] [full abstract]
Spontaneous imitation in a second language is different from native language imitation
Harim Kwon
[short abstract] [full abstract]
Variability in the French Mid Vowels: Vowel Harmony, Syllable Structure, and the Creation and Effects of Phonological Representations
Jeffrey Lamontagne
[short abstract] [full abstract]
Vowel dynamics and social meaning in York, Northern England
Daniel Lawrence
[short abstract] [full abstract]
Dealing with 'inperfection': Affixes, allomorphy, and dual-route parsing
Laurel Lawyer and David Corina
[short abstract] [full abstract]
Subcategorical contrasts in Korean affricates: Implications for English loanword adaptation
Yongeun Lee and Matthew Goldrick
[short abstract] [full abstract]
Prosodic Convergence During and After a Cooperative Maze Task
Yoonjeong Lee, Samantha Gordon Danner, Benjamin Parrell, Sungbok Lee, Louis Goldstein and Dani Byrd
[short abstract] [full abstract]
Ambiguous rhoticity in Glasgow: Short term exposure promotes perceptual adaptation for experienced and inexperienced listeners
Robert Lennon, Rachel Smith and Jane Stuart-Smith
[short abstract] [full abstract]
Imitation of non-native clusters: the role of transitional schwa
Tomas Lentz, Marianne Pouplier, Ioana Chitoran and Phil Hoole
[short abstract] [full abstract]
A bad feeling or a bad filling? The influence of social network size on speech perception
Shiri Lev-Ari
[short abstract] [full abstract]
What reaction times reveal about listener groups: L1 Aboriginal English and Standard Australian English responses to a prelateral merger-in-progress
Deborah Loakes, Janet Fletcher, John Hajek and Joshua Clothier
[short abstract] [full abstract]
Individual Differences in Perceptual Compensation and Lexical Effects and Implications for Sound Change
Yanyu Long
[short abstract] [full abstract]
Ambivalent Consonantal Effects on F0
Qian Luo, Karthik Durvasula and Yen-Hwei Lin
[short abstract] [full abstract]
The limits of inductive learning: The case of Modern Irish mutation
Ruth Maddeaux and Yoonjung Kang
[short abstract] [full abstract]
Sonority profile and temporal organization of clusters: evidence from Russian
Stefania Marin, Marianne Pouplier and Alexei Kochetov
[short abstract] [full abstract]
Seeing focalization: The role of visual information from lip movements in the natural referent vowel bias
Matthew Masapollo, Linda Polka and Lucie Ménard
[short abstract] [full abstract]
Production Influences on Phonological Representation in an Emergentist Grammar
Tara Mcallister Byun and Anne-Michelle Tessier
[short abstract] [full abstract]
A system for unified corpus analysis, applied to polysyllabic shortening across 12 languages
Michael Mcauliffe, Morgan Sonderegger and Michael Wagner
[short abstract] [full abstract]
Cross-linguistic gender priming in speech processing
Grant Mcguire, Molly Babel and Alexandra Bosurgi
[short abstract] [full abstract]
Gesture and velocity in Brazilian Portuguese devoiced vowels: a preliminary EMA study
Francisco Meneses, Denise Pozzani, Nicole Wong, Zainab Hermes, Torrey Loucks and Ryan Shosted
[short abstract] [full abstract]
Production dynamics and phonetic motivations for English raised /æ/ and intrusive [l]
Jeff Mielke, Christopher Carignan and Erik Thomas
[short abstract] [full abstract]
The Effect of Phonological Context on the Perception of Strong Place Assimilation in Nasal and Stop Consonants
Mercedeh Mohaghegh and Craig Chambers
[short abstract] [full abstract]
Vowel Lengthening Effects in Natural Speech: Learning under sparse data and high variance
Becca Morley
[short abstract] [full abstract]
Interactions of speaking rate and prosodic organization in non-native speech production
Tuuli Morrill and Melissa Baese-Berk
[short abstract] [full abstract]
How does deep brain stimulation affect regulation in speech motor control?
Doris Muecke, Anne Hermes, Timo B. Roettger, Johannes Becker and Michael Barbe
[short abstract] [full abstract]
How native and non-native listeners process schwa reduction in French: A combined eye-tracking and ERP study
Kimberley Mulder, Sophie Brand and Mirjam Ernestus
[short abstract] [full abstract]
Phonotactically-mediated Compensation for Coarticulation
Kevin Mullin
[short abstract] [full abstract]
The Perceptual Effects of Phonotactic Rareness and Partial Allophony in Canadian French
Patrick Murphy, Philip Monahan and Margaret Grant
[short abstract] [full abstract]
Perception, Production, and the Implementation of Phonological Opacity in the Bengali Vowel Chain Shift
Traci Nagle
[short abstract] [full abstract]
Corrective focus in conversational French
Ricardo Napoleão de Souza and Caroline Smith
[short abstract] [full abstract]
An acoustic study of Punjabi tone and stress (Doabi dialect)
Kiranpreet Nara
[short abstract] [full abstract]
How vowel variability relates to vowel perception
Nhung Nguyen, Jason A. Shaw, Catherine T. Best and Michael D. Tyler
[short abstract] [full abstract]
ERPs reveal that exemplar effects are driven by episodic memory instead of the mental lexicon
Annika Nijveld, Kimberley Mulder, Louis ten Bosch and Mirjam Ernestus
[short abstract] [full abstract]
On the relation between speech perception and loanword adaptation: new evidence from Korean
Mira Oh, Robert Daland and Lisa Davidson
[short abstract] [full abstract]
Investigating the perceptual hypocorrection hypothesis with sibilant harmony
Avery Ozburn
[short abstract] [full abstract]
Investigating Conflicting Aerodynamic Requirements in CC Clusters
Manfred Pastätter and Marianne Pouplier
[short abstract] [full abstract]
Phonetic devices and the construction of the phonological space
Elinor Payne, Brechtje Post, Nina Gram Garmann and Hanne Gram Simonsen
[short abstract] [full abstract]
Sleep-dependent consolidation in the learning of natural vs. unnatural phonological rules
Sharon Peperkamp and Alexander Martin
[short abstract] [full abstract]
Sonorancy of the rhotic /ɣ/ in Sgaw Karen
Pittayawat Pittayaporn and Sujinat Jitwiriyanont
[short abstract] [full abstract]
On the role of manner and place in Kurtöp tonogenesis
Sarah Plane and Gwendolyn Hyslop
[short abstract] [full abstract]
Car-talk: How physical environment influences speech production and perception
Ryan Podlubny, Jen Hay and Megan McAuliffe
[short abstract] [full abstract]
Infants prefer vowels with infant vocal resonances: Evidence for an “articulatory filter” bias
Linda Polka, Matthew Masapollo and Lucie Menard
[short abstract] [full abstract]
‘/t,d/ Deletion’: Articulatory Gradience in Variable Phonology
Ruaridh Purse and Alice Turk
[short abstract] [full abstract]
A phonologically weak contrast can induce phonetic overlap
Margaret Renwick, Ioana Vasilescu, Camille Dutrey, Lori Lamel and Bianca Vieru
[short abstract] [full abstract]
How is lexical gemination transposed in Tashlhiyt whistled speech?
Rachid Ridouane and Julien Meyer
[short abstract] [full abstract]
Acoustic salience and input frequency in L2 lexical tone learning
Katherine Riestenberg
[short abstract] [full abstract]
On the link between glottal vibration and sonority
Megan Risdal, Ann Aly, Adam Chong, Patricia Keating and Jesse Zymet
[short abstract] [full abstract]
Representation of intonation categories: evidence from speech perception
Joe Rodd and Aoju Chen
[short abstract] [full abstract]
High frequency prototypes do not facilitate phonotactic generalizations
Timo Roettger and Dinah Baer-Henney
[short abstract] [full abstract]
Preserving speech dynamics in Parkinson's disease: an acoustic study of the production of glides by Belgian French patients
Virginie Roland, Véronique Delvaux, Kathy Huet, Myriam Piccaluga, Marie-Claire Haelewyck and Bernard Harmegnies
[short abstract] [full abstract]
Blending of articulator activation in a dynamical model of phonological planning
Kevin Roon and Adamantios Gafos
[short abstract] [full abstract]
Plural predictability and OCP influence plural morpheme duration in English
Darcy E. Rose
[short abstract] [full abstract]
An experimental approach to perceptual salience
Hanna Ruch
[short abstract] [full abstract]
Lexical knowledge is available, but not always used, very early
Amanda Rysling, John Kingston, Adrian Staub, Andrew Cohen, Jeffrey Starns and Anthony Yacovone
[short abstract] [full abstract]
The adaptation of Mandarin falling diphthongs in Heritage Korean in China: The interaction of linguistic and sociolinguistic factors
Na-Young Ryu, Yoonjung Kang and Sung-Woo Han
[short abstract] [full abstract]
ERP evidence for the ecological validity of artificial language learning
Lisa Sanders, Claire Moore-Cantwell, Joe Pater, Robert Staubs and Benjamin Zobel
[short abstract] [full abstract]
Comparing neighborhood density and clear speech effects in the French vowel system
Rebecca Scarborough and Cecile Fougeron
[short abstract] [full abstract]
Influence of Syllable Structure on Musical Text Setting
Murray Schellenberg
[short abstract] [full abstract]
How much does the talker matter? Depends who's listening: Age-related variability in the use of social information in speech perception
Jessamyn Schertz, Yoonjung Kang and Sungwoo Han
[short abstract] [full abstract]
L1 influence on the identification of intonational contours
Elaine Schmidt, Carmen Kung, Brechtje Post, Ivan Yuen and Katherine Demuth
[short abstract] [full abstract]
Cross-language differences in vowel inherent spectral change - evidence from Polish learners of English
Geoff Schwartz, Jarosław Weckwerth, Kamil Kazmierski, Aperlinski Grzegorz, Malarski Kamil and Mateusz Jekiel
[short abstract] [full abstract]
Dynamic listening: temporal expectations guide perception of phonetic detail
Jason Shaw
[short abstract] [full abstract]
The articulatory space of oral and nasal vowels in Brazilian Portuguese
Ryan Shosted, Denise Pozzani, Francisco Meneses, Nicole Wong, Zainab Hermes and Torrey Loucks
[short abstract] [full abstract]
Intrinsic pitch of diphthongs in lexical tone perception
Jessica Siddins and Eva Reinisch
[short abstract] [full abstract]
Representation of Acoustic Detail
Michelle Sims and Benjamin V. Tucker
[short abstract] [full abstract]
A production-internal learning bias against large changes to the base
Amy Smolek and Vsevolod Kapatsinski
[short abstract] [full abstract]
Characterizing vocal tract dynamics with real-time MRI
Tanner Sorensen, Asterios Toutios, Louis Goldstein and Shrikanth Narayanan
[short abstract] [full abstract]
Vowel movement as a function of voicing in simple CV sequences
Stavroula Sotiropoulou, Tanner Sorensen, Stephen Tobin and Adamantios Gafos
[short abstract] [full abstract]
The role of echoic memory in the initial learning of a second dialect: the case of bilinguals
Laura Spinu, Jiwon Hwang and Renata Lohmann
[short abstract] [full abstract]
Phonetic Shift /ɔr/ Phonemic Change? American English mergers over 40 years
Joseph A. Stanley and Margaret E. L. Renwick
[short abstract] [full abstract]
The effect of realtime visual feedback on vocalic targets
Elizabeth Stelle, Eric Vatikiotis-Bateson and Caroline L. Smith
[short abstract] [full abstract]
Investigating the origins of pre-consonantal /s/-retraction: acoustic, perceptual and articulatory evidence from English
Mary Stevens and Jonathan Harrington
[short abstract] [full abstract]
Gradient phonological relationships: Evidence from vowels in French
Sophia Stevenson and Tania Zamuner
[short abstract] [full abstract]
Influence of palatalization on tongue-tip velocity in trills
Taja Stoll, Philip Hoole and Jonathan Harrington
[short abstract] [full abstract]
Articulatory dynamics of degemination in Dutch
Patrycja Strycharczuk and Koen Sebregts
[short abstract] [full abstract]
The link between anterior lingual gesture delay and loss of coda /r/: an ultrasound study
Jane Stuart-Smith, Eleanor Lawson and James Scobbie
[short abstract] [full abstract]
Lack of evidence for subphonemic contrasts motivating exceptional behavior in vowel harmony
Daniel Szeredi
[short abstract] [full abstract]
Children's sensitivity to degrees of mispronunciation: Enriching the preferential looking paradigm with pupillometry.
Katalin Tamasi, Cristina McKean, Adamantios Gafos and Barbara Hoehle
[short abstract] [full abstract]
Lexical specificity and temporal decay in intraspeaker priming of sociolinguistic variables
Meredith Tamminga
[short abstract] [full abstract]
Kinematic aspects of L2 production in an imitation task
Mark Tiede, Christine Mooshammer, Dolly Goldenberg and Douglas Honorof
[short abstract] [full abstract]
MALD: Massive Auditory Lexical Decision
Benjamin Tucker and Daniel Brenner
[short abstract] [full abstract]
The Function of Duration and Stress- vs. Syllable-Timing
Irene Vogel and Angeliki Athanasopoulou
[short abstract] [full abstract]
Asymmetries in English Liquid Production and Vowel Interactions
Rachel Walker, Michael Proctor, Caitlin Smith and Ewald Enzinger
[short abstract] [full abstract]
Investigating the interaction between speaker dialect and listener differences across two tasks.
Abby Walker, Andrew Burlile and Katherine Askew
[short abstract] [full abstract]
Constraints on cross-talker generalization of foreign-accent adaptation
Kodi Weatherholtz, Linda Liu and T. Florian Jaeger
[short abstract] [full abstract]
Evidence for vowel targets in formant distributions and within-syllable adjustments
D. H. Whalen
[short abstract] [full abstract]
Illusory epenthesis and recoverability-conditioned sensitivity to phonetic detail
James Whang
[short abstract] [full abstract]
Spectral Trajectories of Spanish /s/: Temporal Variability, Vowel Context, and Duration
Eric Wilbanks
[short abstract] [full abstract]
The Generation of Prosodic Frames in Speech Production: An Experimental Approach
Hilary Wynne, Linda Wheeldon and Aditi Lahiri
[short abstract] [full abstract]
When dynamics conflict: Flap dynamics and palatalization in Japanese
Noriko Yamane, Phil Howson, Masaki Noguchi and Bryan Gick
[short abstract] [full abstract]
Effects of boundary tones on the coordination of lexical tones
Hao Yi and Sam Tilsen
[short abstract] [full abstract]
Articulation and Representation of Laterals in Australian-accented English
Jia Ying, Jason Shaw, Catherine Best, Michael Proctor, Donald Derrick and Christopher Carignan
[short abstract] [full abstract]
The emergence of an inflectional edge tone morpheme in Samoan
Kristine Yu
[short abstract] [full abstract]
Planning of inserted /ɹ/ in the speech of Australian English-speaking children
Ivan Yuen, Felicity Cox and Katherine Demuth
[short abstract] [full abstract]
Different effects of production on spoken-word recognition for adults versus children
Tania Zamuner, Stephanie Strahm, Elizabeth Morin-Lessard and Mike Page
[short abstract] [full abstract]
The role of fundamental and formant frequency information on voice and speaker perception in children with Autism and Attention Deficit Hyperactivity
Georgia Zellou, Santiago Barreda, Nancy McIntyre, Lindsay Swain-Lerro, Matthew Zajic and Peter Mundy
[short abstract] [full abstract]
An Experimental Investigation of Positionally Conditioned Tone Sandhi in Hailu Hakka
Jie Zhang, Hanbo Yan, Yuwen Lai and Shao-Ren Lyu
[short abstract] [full abstract]
Infants’ use of phonological detail during foreign-accented word recognition
Marieke van Heugten, Dena Krieger, Melissa Paquette-Smith and Elizabeth Johnson
[short abstract] [full abstract]
I can't understand - The perception of native and non-native can and can’t by native and non-native listeners of English
Margot van Mulken, Huib Kouwenhoven and Mirjam Ernestus
[short abstract] [full abstract]
An acoustic analysis of laryngeal contrasts in Korean stops across three groups of speakers
Ruben van de Vijver and Hae-Eun Cho
[short abstract] [full abstract]

SHORT ABSTRACTS (alphabetical by first author):

Coarticulation magnitude in German children
Dzhuma Abakarova, Elina Rubertus, Khalil Iskarous, Mark Tiede, Jan Ries and Aude Noiray

The present study explored developmental differences in coarticulation magnitude in three cohorts of monolingual German preschoolers age 3, 4 and 5 and a group of adults taken as reference. CVCV nonwords with C=/b/, /d/, /g/, or /z/ and V=/i/, /y/, /u/, /a/, /e/, and /o/ were elicited in a repetition task. Movement from the tongue was recorded via ultrasound imaging to track the spatial and temporal organization of lingual gestures within syllables. To assess developmental differences in CV coarticulation, we employed 1) standard Locus Equation measures (LE) which test differences of coarticulation magnitude and resistance as a function of consonant place of articulation and 2) Mutual Information (MI), a method used to quantify the articulatory dependence versus independence between phonemes

Results highlighted similar order of coarticulation magnitude as a function of consonant place of articulation (labial >velar>alveolar) but differences in degree of coarticulatory magnitude and resistance between groups. Overall, children’s patterns were more variable than adults suggesting that at 4 and 5 years of age intra-syllabic patterns of lingual coarticulation are not yet stable. [full abstract]

Speaking rate effect on consonant-vowel coarticulation and on stop consonant classification
Mohammad Abuoudeh and Olivier Crouzet

The aim of this research is to investigate the impact of vowel duration variation due to different speaking rates on :1) consonant and vowel quality as well as CV coarticulation and 2) the extent of this influence on stop consonant classification in voiced and voiceless environments. Many studies expose that the CV dynamic portion (formant transitions) is varying in function of speaking rate. As for the static portions (consonant or vowel targets), the impact of temporal variation seems to be more ambiguous. As speaking rate increases, vowels and consonants tend to lose quality [Lindblöm, 1963], we hypothesize that this modification of information may be compensated by the temporal information. In other words, it is predicted that temporal dynamic information may neutralize the variation of spectral properties caused by speaking rate. Five male native speakers of French produced C1VC real words embedded in a carrier sentence in slow, normal and fast speaking rates.

According to our results, speaking rate influences vowel target whereas consonant target remains globally stable. The CV relation is modified by speaking rate, with similar descriptive tendencies for voiced and voiceless consonants. Moreover, in LDA tests, adding duration as a third predictor increases classification accuracy in fast and normal rates. [full abstract]

The Role of Tongue Root in Laryngeal Contrasts: An Ultrasound Study of English, Spanish, Hindi, and Korean
Suzy Ahn

This research investigates the articulatory expression of laryngeal contrasts in English, Spanish, Hindi, and Korean using ultrasound imaging. Many languages have a voicing contrast in stops, but phonetic implementation of voicing varies cross-linguistically (e.g. phonation during closure for Spanish and short-lag VOT for English). This study focuses on tongue root advancement since it could potentially serve two different purposes: promoting phonation during closure and enhancing short-lag VOT.

This study aims to answer the question of (1) whether laryngeal contrasts are reflected in tongue position in each language by comparing stops in utterance-initial position and (2) if the same acoustic realization can have different tongue positions depending on whether a language shows phonological contrasts or not by comparing languages with different laryngeal contrasts (e.g. Hindi vs. English) and different phonetic realizations of the same phonological contrasts (e.g. English vs. Spanish). Results show that tongue root advancement can facilitate phonation as well as short-lag VOT, the latter of which has not received much attention in previous literature. Articulatory differences align with a more abstract two-way laryngeal distinction that is the same for Spanish and English even though acoustics are different. However, in Hindi, tongue root advancement can be used only for phonation. [full abstract]

An acoustic investigation of the categorical and gradient spread of pharyngealization in Urban Najdi Arabic
Abdulaziz Alarifi and Benjamin V. Tucker

Arabic has a set of phonemes that are produced with a secondary constriction located in the pharynx in addition to a primary one in the dental/alveolar areas. It is well established that this secondary constriction, referred to as pharyngealization, affects adjacent vowels by mainly lowering their F2 values (e.g. Al-Ani, 1970; Ghazali, 1977; Yeou, 2001). Interestingly, the pattern of pharyngealization spread varies greatly across Arabic dialects. This study provides an acoustic investigation of the spread of pharyngealization in Urban Najdi Arabic (UNA), a dialect spoken in the central region of Saudi Arabia. Data from five male and five female speakers reveals that pharyngealized consonants affect, with varying degrees, all vowels in bisyllabic and trisyllabic words both regressively and progressively. However, progressive spread is more sensitive to distance, whereas pharyngealization seems to be less susceptible to distance effects in regressive spread, especially for female speakers. Overall, these findings demonstrate that pharyngealization in UNA displays both gradient and categorical effects, which is in line with other phonological patterns such as regressive voice assimilation in Dutch (Jensen, 2007) and Romanian nasal devoicing (Tucker and Warner, 2010). [full abstract]

Effects of Position, Stress and Manner of Articulation on Consonant-Vowel Co-occurrence in Three Languages
Eleonora Albano

This study follows-up on Albano (2016), comparing type and token frequencies of consonant vowel pairs in three large databases of British English, Latin American Spanish, and Brazilian Portuguese. The aim was to investigate whether token frequencies reproduce the CV co-occurrence biases found in type frequencies. Indeed, CV pair type and token frequencies are highly correlated in all three languages. Moreover, the biases as well as the conditions that enhance them are comparable across counts and languages. The same attraction and rejection patterns appear everywhere, namely: coronal Cs attract front Vs and dorsal Cs attract central and back Vs. These patterns are likely to originate in biomechanics, and can be explained through the DAC model of coarticulation (Recasens, Pallarès, & Fontdevila, 1997). However, their greater strength in certain positions and prosodic environments suggests that they interact with speech planning constraints. Interpreting this fact requires recourse to additional sources from the Laboratory Phonology literature.

References:

Albano, E. C. (2016). Conditions favoring biomechanically driven CV co-occurrence biases in lexicons. Journal of Phonetics, 55, 78–95.

Recasens, D., Pallarès, M.D., & Fontdevila, J. (1997). A model of lingual coarticulation based on articulatory constraints. Journal of the Acoustical Society of America, 102, 544–561. [full abstract]

Regressive Coarticulation of Pharyngealization in Arabic Spoken Word Recognition
Sawsan Alwabari and Tania Zamuner

The study investigates listeners sensitivity to phonetic-level variation of pharyngealization during online lexical processing by monitoring listeners’ eye movements. Segments are pharyngealized phonemically (e.g. /tˤ dˤ sˤ/) or phonetically as a consequence of coarticulation. There is a lack of evidence of whether spoken-word recognition (SWR) is influenced by the vocalic cues independently from the consonantal ones. Arabic minimal pairs contrastive in pharyngealization (e.g. [nasˤer] ‘victory’ and [naser] ‘eagle’) were cross-spliced so that regressive coarticulation was either congruent or incongruent with the phonemically pharyngealized consonant. Consistent with typical pharyngealization characteristics, vowels preceding pharyngealized consonants had higher F1 and F3 and lower F2 than those in a non-pharyngealized environment. Twenty Arabic speakers participated in visual-world paradigm experiment where the Target image matched the phonemic cues of pharyngealization and Competitor matched the phonetic cues in addition to two unrelated images. Word recognition is significantly faster when phonetic and phonemic information are congruent. Target fixations are phonetically driven, with congruent coarticulation incurring significantly higher Target fixation than incongruent coarticulation throughout the entire trial. Using Growth Curve Analysis, the study models the timely modulation of lexical processing as acoustic cues of pharyngealization change. These results are discussed in relation to existing SWR models. [full abstract]

Perception, mental representations and production of non-native prosodic contrasts
Yuki Asano

The study investigates which stages in L2 perception, storage and production contribute to foreign accents in L2 prosody. A speeded AX perception experiment with short and long inter-stimulus intervals (300 ms. vs. 2500 ms) was used to discriminate consonant and vowel length contrasts. Additionally, an immediate and delayed imitation experiment to produce those contrasts was conducted, testing the same participants (48 German learners of Japanese, 24 German non-learners and 24 Japanese L1 speakers). Only consonant length contrasts were non-native for German participants. The results showed that even learners and non-learners were able to perceive the non-native segmental length contrasts when task demands were lowest, simply through reliance on the acoustic correlates of stimuli. Once phonological representations were required, the sensitivity of non-learners. Even the learners, who were establishing L2 phonological representations, showed difficulties in automatizing their L2 perception processing and were easily affected by distracting factors. Finally, both learners and non-learners showed deviant L2 productions already in the imitation condition with lowest task demands, suggesting that speech production inevitably requires access to a speaker’s phonological representations. The study showed that mental representations cause non-native speakers’ difficulties when processing L2 prosody. [full abstract]

Is there cross-linguistic evidence for prosodic bootstrapping of word order?
Angeliki Athanasopoulou and Irene Vogel

According to the Prosodic Bootstrapping Hypothesis, babies use prosodic cues to assist in identifying their language’s syntactic structures. The Phrasal Prominence Hypothesis, incorporating the Iambic-Trochaic and Complement Laws, predicts specifically that the manifestation of phrasal prominence provides cues regarding the basic word order of a language. Thus, according to Nespor and others, French (VO) and Turkish (OV) will primarily signal Strong elements with greater vowel duration or pitch and intensity, respectively. Since the PBH crucially depends on the reliability of acoustic cues to identify, not simply correlate with, particular syntactic properties, we test the robustness of this claim with additional data from French and Turkish, as well as data from three other VO languages (Finnish, Greek, Portuguese, Spanish) and one with relatively free word order, Hungarian. Specifically, using Binary Logistic Regression Analyses we confirmed that the strongest prominence cue in Turkish is F0, followed by Intensity. This was true in Hungarian, too, but Duration was also a significant cue. We found that Duration was the main prominence cue as expected in the VO languages, however, with the exception of French. Thus, our findings suggest caution in using prominence manifestation as a prosodic cue for basic word order. [full abstract]

Category typicality in perceptual learning
Molly Babel, Michael McAuliffe, Zoe Lawler and Carolyn Norton

Perceptual learning in speech hinges on at least two steps: (i) a listener must be sensitive to the phonetic detail of the novel pronunciation and (ii) the novel pronunciation has to be apprehended as a recognizable lexical item. Of interest here is the range of variation that promotes perceptual learning. For example, when the category is not perfectly ambiguous, and instead biased towards one phonetic category over the other, do we still see perceptual learning? We address this question with a lexically-guided perceptual learning experiment that uses exposure stimuli that vary in terms of category typicality. Twenty multisyllabic words with /s/ in the onset of a non-initial syllable (e.g., castle) were morphed with /ʃ/ versions of the words (e.g., /kæʃl/) and pre-tested. Steps which corresponded to 70% (Typical), 50% (Ambiguous), and 30% (Atypical) word endorsement rates were used as exposure stimuli in a standard perceptual learning paradigm. Atypical pronunciations show attenuated perceptual learning compared to ambiguous stimuli, while more typical (yet still not canonical) pronunciations show no evidence for perceptual learning. Together, these results suggest that perceptual learning relies on an interplay between confidence in phonetic and lexical predictions and category typicality of the acoustic signal. [full abstract]

Using Developmental Data to Explore Frequency and Neighborhood Density Effects in Production
Melissa Baese-Berk and Katherine White

Previous research has suggested that lexical properties of words, including frequency and neighborhood density, can influence their production. Various hypotheses have been put forth for these lexical effects. We focus on two accounts: exemplar and production-internal accounts. Under exemplar accounts, speakers store episodic memories from perception and use these episodes (exemplars) to update their production targets. Speakers hyper- or hypo-articulate words that they have heard pronounced this way. However, under production-internal accounts, lexical effects are driven by the dynamics of the production system and competition induced by the activation of related forms.

In the present study, we contrast these two accounts using developmental data. In Experiment 1, we compared children’s productions of words that were high and low frequency for both children and adults to words that were high frequency for one group but low frequency for the other. In Experiment 2, we compared words with high or low neighborhood density for both groups with words that were high density for one group and low density for the other. Preliminary results suggest that children’s productions reflect the density and frequency patterns of their own lexicon rather than the adult lexicon, supporting production-internal accounts of lexically conditioned phonetic variation. [full abstract]

Rapid adaptation to foreign-accented speech: A web-based replication of Clarke and Garrett (2004)
Larisa Bainton, Emily Rowe, Zach Burchill, Linda Liu, Kodi Weatherholtz and T. Florian Jaeger

Understanding speech can be difficult, especially accented speech. To examine the effect of exposure to Mandarin-accented speech, we implemented a novel web-based paradigm, replicating and extending on an existing study’s claim that one minute of exposure significantly reduced accent-related difficulty (Clarke and Garrett 2004). We used reaction times as a measurement of participants’ processing speed in a sentence-listening task. Subjects who were exposed to an accented speaker were initially slower, but this disadvantage quickly decreased, and they improved significantly compared to subjects exposed to a native speaker. After exposure, both groups were tested on accented speech, and those who had not yet heard the Mandarin accent fared much worse (192 ms slower). All groups showed clear signs of task adaptation during exposure, but in the test phase there was an advantage seen in groups that had listened to an accented speaker. Higher error rates in the accented conditions suggest that not all of the difficulty introduced with an accented speaker went away. This is contrary to Clarke and Garrett’s conclusion and gives us a more complex understanding of early adaptation to accented speech. [full abstract]

The role of palate shape in individual articulatory and acoustic variability
Sarah Bakst

This ultrasound and acoustics study considers the role of individual anatomy in articulatory and acoustic variation and variability in American English /r/ and /s/. Individual variation in the production of these consonants is a well-documented phenomenon (Mielke et al., 2010; Lawson et al., 2011; Bladon and Nolan 1977). In their study of front vowels, Brunner et al. (2009) found that flatter palates require greater articulatory precision than domed palates to achieve acoustic consistency because, all else being equal, smaller changes in articulation result in greater changes in acoustics for flatter palates than for domed. The study here found that the segment determines the type of variability: palate shape was correlated with articulatory variability for /r/ but acoustic variability for /s/; speakers were similar to each other in their /r/ acoustics and /s/ articulation. Further, speakers fell into two main groups: one group had flatter palates and tended to have less articulatory variability in their /r/ production, but more acoustic variability in their /s/ production. The other group tended to have more domed palates, and while they were acoustically consistent in their /s/, they were slightly more articulatorily variable in their /r/. [full abstract]

Structured Variation across Sound Contrasts, Talkers, and Speech Styles
Hye-Young Bang and Meghan Clayards

Individual differences in talkers have been observed for vowel formant frequencies, VOT of stops, and the location of center of gravity (CoG) for sibilant fricatives. One question is whether the observed differences across talkers is constrained or structured in a way that facilitates listeners to quickly build expectations about how a talker sounds. Based on a scripted dialogue task, we examined individual talker and token variability in the production of several acoustic cues across different speech sounds (/p t s ʃ) and prominence conditions. We extracted VOT of /p/ & /t/, and CoG of /s/ and /ʃ/, and duration for word initial segments. We further computed Locus Equations (LE) slopes and speaking rate for each talker and each consonant. Contrastive stress (prominence) was elicited by “mishearing” some portion of the sentence. We found correlations in talker variability across cues and sound contrasts, and that the differences between talkers are in the same direction as the effect of prominence. These findings shed light on the issue of talker and speaking style related variation by documenting that some of the inter-speaker variability is structured systematically and may be due to a single parameter: how much talkers hyper-articulate. [full abstract]

Deformation-based articulatory representations of speech sounds
Marissa Barlaz, Ryan Shosted, Christopher Carignan, Maojing Fu, Zhi-Pei Liang and Brad Sutton

The use of advanced imaging techniques allows understanding of the dynamic organization of what are traditionally considered to have static production. Our study aims to provide a means of exploring and quantifying image-based articulatory information as a function of time, while also preserving spatial resolution. Pursuant to this goal, we analyze vocal tract deformations in a corpus of real-time magnetic resonance (rt-MR) images. We regard a deformation as the horizontal and vertical, pixel-by-pixel transformations of one image into another. We study nasal–oral vowel pairs of Northern Metropolitan French (NMF). We calculate the non-linear deformations between MR images at the beginning and end of each vowel. This results in a horizontal and vertical deformation vector for each pixel in the image. PCA was performed on these vectors in a region of interest, in order to understand the magnitude of movement from the beginning to the end of the vowel. The results of our analysis show that there are systematic time-varying differences in the articulatory configuration of both NMF oral and nasal vowels. We observe differences in the position of the tongue blade and tip between the beginning and end of these vowels, suggesting that movement has occurred within each time-course. [full abstract]

Final lengthening in German
Malte Belz, Oxana Rasskazova, Anja Riemenschneider, Jelena Krivokapić, Melanie Weirich and Christine Mooshammer

Prosodic variation affects tonal aspects as well as temporal properties of individual segments. Previous studies found that vowel gestures tend to be longer at strong prosodic boundaries (e.g. Byrd 2000). It also has been shown that tense vowels in German stretch in stressed syllables and compress for fast speech rate. However, up to now the interaction of tenseness and final lengthening has not been studied. In the present study we investigate the effects of phrasal boundaries on the temporal characteristics of the preceding syllable. According to the π-gesture model of Byrd and Saltzman (2003), the lengthening effect should gradually increase towards the boundary independent of the segmental content. Therefore, we expect an effect of lengthening for both tense and lax vowels. Acoustic and tongue movement data of six German subjects have been analysed for the correlation between tenseness, boundary position and speech rate. The results show that lax vowels are lengthened in phrase-final positions, although less than tense vowels. For faster speech rates, however, lax vowels tend to be incompressible, confirming previous findings. [full abstract]

Word-final geminates in a Modern South Arabian language: Phonetics and Phonology
Sabrina Bendjaballah and Rachid Ridouane

While word-final geminates are common across Arabic dialects, they are claimed to be rather unusual in other Semitic languages. In this study we examine the status of final geminates in Mehri, an endangered Modern South Arabian language. We show, contra to previous accounts, that these segments exist both at the phonological and phonetic levels.

Evidence for phonological word-final geminates is drawn from their behavior relative to stress assignment: we examine the class of biliteral verbs and show that the final consonant of these verbs patterns with a sequence of two consonants. We conclude it is an underlying geminate.

The distinction between a single linked segment and its double linked counterpart is generally understood as predicting a distinction of consonant length. We show that final consonants in Mehri biliteral verbs are acoustically longer than their singleton counterparts. This difference is statistically significant at p<.001 for both stops and fricatives. Importantly, stops are systematically released so that the durational cue of a geminate stop is preserved in this position. Finally, we observe that the singleton/geminate contrast is shaped by word position, with final segments, be they singletons or geminates, systematically longer than their medial counterparts. [full abstract]

Modelling Phonetic and Phonological Variation with 'Small' Data: Evidence from Kaqchikel Mayan
Ryan Bennett and Kevin Tang

This work confronts the challenge of 'big data' research in under-resourced languages by demonstrating that small, well-annotated spoken and written corpora can reveal systematic patterns of phonetic and phonological variation. The empirical focus is Kaqchikel, a Guatemalan Mayan language. We provide a proof-of-concept that psycholinguistic and phonetic norms computed from less-than-ideal corpora can be used to study fine-grained phonetic and phonological phenomena, including the robustness of phonemic contrasts, allophonic variability, and word-frequency effects on sub-phonemic detail.

As there were no previous structured corpora for Kaqchikel (apart from dictionaries) we compiled two new corpora, one written and one spoken. The spoken corpus contains just ~40,000 word tokens (~4 hours of spontaneous spoken Kaqchikel). The written corpus contains ~1 million word tokens, constructed from existing religious texts, spoken transcripts, government documents and educational books.

We show that even these small corpora are sufficient to replicate phonetic reduction effects related to word frequency, and to investigate a more cutting-edge topic, the effect of functional load on vowel acoustics and vowel mergers. The practical conclusion of this work is that interpretable, reliable, and theoretically meaningful corpus research can be carried out minority and under-resourced languages for which only sparse, noisy corpora may be available. [full abstract]

Language-internal behavior of typologically rare sounds: Production, perception, and distribution of breathy sonorants in Marathi
Kelly Berkson

Breathy sonorants are cross-linguistically rare, occurring in just 1% of the languages indexed in the UCLA Phonological Segment Inventory Database (UPSID). While prior work has shed some light on their acoustic properties, much remains to be learned about the perception and language-internal distribution of these sounds. The current work focuses on breathy sonorants in Marathi, an Indic language which contains breathy nasals, approximants, laterals and rhotics (Dhongde & Wali 2009). The results of three studies—an instrumental acoustic analysis, a perception experiment, and investigation of phonotactic frequencies in the 2.2 million word Marathi portion of the EMILLE-CIIL corpus—reveal that: (1) the acoustic differences associated with phonation type contrasts in sonorants are sensitive to contextual factors such as word position and vowel quality in a way that those associated with obstruents are not; (2) phonation type contrasts in obstruents are perceived more accurately than those in sonorants; and (3), breathy sonorants are severely under-represented language internally, representing just 0.7% of the consonants in the corpus. [full abstract]

Style-shifting and phonetic alignment in non-native discourse
Grant M. Berry and Mirjam Ernestus

We investigate the plasticity of phonological boundaries in discourse in a lingua franca by tracking the production of 34 Spaniards conversing with two Dutch confederates in English across two speech styles, and we focus on incremental changes in two key English vowel contrasts. The first, /i/-/ɪ/, is not made by Spanish speakers in aggregate but reliably produced by the Dutch confederates. The second, /ɛ/-/æ/, is produced by Spanairds, but not by Dutch confederates. We assessed degree of merger in each of ten normalized time bins with the Pillai score. Results indicate that Spaniards align with the Dutch confederates in their production of these contrasts, merging /ɛ/ and /æ/ and gradually separating /i/ and /ɪ/, rather than adopting standard English production. We found greater merger overall in informal speech, but an interaction with time for the /i/-/ɪ/ contrast indicates that /i/ and /ɪ/ gradually separate in informal speech. Conversely, there is no effect of time for the /ɛ/-/æ/ contrast: Spaniards merge /ɛ/ and /æ/ when conversing with the Dutch confederates significantly more in informal speech, but the magnitude is static throughout each interview. Finally, proficiency modulates alignment: the most proficient speakers separate /i/-/ɪ/ and merge /ɛ/-/æ/ more than less proficient speakers. [full abstract]

Morpho-phonology without semantics? The roles of lexical memory and experience in influencing the nature of lexical representations in a rote learning context
Siti Syuhada Binte Faizal and Ghada Khattab

Word learning is strengthened through building semantic representations alongside morpho-phonological structures (e.g. McGregor et al., 2002). Rote Qur’anic memorisation by non-Arabic-speaking pupils puts this notion to the test: word meanings are not provided for memorisers, and the non-concatenative nature of Arabic morphology (McCarthy, 1981) presents a stark contrast with these learners’ L1. The goal of this study was to investigate whether non-Arabic-speaking Qur’anic memorisers implicitly gain non-linear morpho-phonological representations while reading and memorising the Qur’an, and thus, are primed by Arabic roots and word patterns during lexical processing. We also investigated whether root priming interacts with Qur’an vocabulary knowledge and amount of memorisation. 153 participants were sampled from religious schools and the general public in Singapore.

Using a visual unmasked priming paradigm in a lexical decision task, we varied the prime-target relationship across conditions to test for priming of root, phonological-relatedness, or both. Preliminary results showed no statistical significance across the experimental conditions, but interesting trends showed that the root priming effect is the largest in participants who have more vocabulary knowledge and greater amount of memorisation. This suggests that root morphemes convey semantic information, and that semantic knowledge needs to be supported with statistical exposure to strengthen morpho-phonological representations. [full abstract]

Cider apples and jeder Abend: the interplay of /r/-sandhi and word-initial glottalisation in English-accented German
Maria Paola Bissiri and James M. Scobbie

The aim of the present study is to investigate word-final /r/ and word-initial glottalisation in English-accented German and in English in relationship to phrasing and focus. This is the first study on the interplay between external sandhi and glottalisation in relationship to prosodic structure, comparing native and non-native language, carried out by means of ultrasound tongue-imaging.

Some scholars argue that glottalisation can block external sandhi, however, a previous study of English /l/ found that glottalisation of vowel-initial words could co-occur with different amounts of sandhi (Scobbie and Pouplier 2010). The transfer of external sandhi in the interlanguage has been seldom investigated and with conflicting results (Zsiga 2011).

We recorded native English speakers with German as L2 and German native speakers with simultaneous tongue-ultrasound and audio recordings. The following hypotheses were formulated: a) glottalisations can overlap with /r/-sandhi, and phrase boundaries, not glottalisations, block sandhi, b) /r/-sandhi is most frequent in phrase-medial position and before deaccented words, while glottalisations are most frequent at phrase boundaries and before accented words, and c) transfer of glottalisation and sandhi occurs from the native to the non-native language.

Results will be reported and discussed in light of the models presented above. [full abstract]

Categorical perception of lexical stress: A cross-linguistic study
Natalie Boll-Avetisyan, Saioa Larraza, Aislyn Rose, Sylvie Margules, Ranka Bijeljac-Babic, Thierry Nazzi and Barbara Höhle

This study explored the categorical perception of lexical stress by German (a language with contrastive lexical stress) versus French (no lexical stress) listeners. We created an 8-step lexical stress continuum of the nonword /gaba/ from trochaic ['gaːbaː] to iambic [gaː'baː]. In a discrimination experiment, participants heard AXB triplets in which X was identical with either A or B. As expected, the Germans where better at perceiving whether X was equal to A or B than the French, but only, if the triplets were from the middle “between-categories” area of the continuum and not when they were from a “within-category” edge. Next, in an identification experiment, participants heard AXB triplets in which A and B were always either step1 or step8, and X could be any token on the continuum. Both groups had difficulties indicating whether X sounded more like A or like B when X was from the middle (i.e., at the boundary) of the continuum, while they readily associated all other steps with the more similar stress pattern, an ability that was more enhanced in the German than the French listeners. Together, these results suggest a categorical perception of lexical stress if the native language uses it contrastively. [full abstract]

When intonation fails to phonologize: the case of Southern Vietnamese
Marc Brunelle

This poster addresses the realization of intonation in Southern Vietnamese, a language in which each syllable bears one of five complex tones and in which communicative functions and sentence modality are largely encoded with final particles. An eight-hour corpus of spontaneous and semi-spontaneous speech was collected from 19 speakers. The corpus was transcribed and annotated for syntactic clauses and disfluencies, and for 13 types of sentence modalities/communicative functions (74,090 syllables grouped into 20,783 intonational phrases). F0, intensity and duration were measured and used to locate the boundaries of prosodic domains, to try to uncover intonational targets, and to determine if phrase-final particles affect the realization of intonation.

Syllable lengthening at the end of syntactic clauses suggests the existence of intonational phrases, but the lack of phonetic variability inside these phrases argues against the existence of other accentual or intonational domains. No significant phonetic difference between communicative functions or sentence modalities was uncovered, even after taking speaker-specific effects into account. Finally, the absence of phrase-final particles does not lead to greater intonational contrasts. Put together, these results suggest that the intonation of Vietnamese, though clearly present, is not conventionalized into discrete categories (i.e. not phonologized). [full abstract]

Effect of phrasal accent on vocalic and consonantal nuclei
Lia Saki Bucar Shigemori, Marianne Pouplier and Štefan Beňuš

This study examined the effect of phrasal accent on vowels and consonants occupying the syllable nucleus. Previous studies have revealed that prosodic emphasis on vowels is achieved by sonority expansion and hyperarticulation. In Slovak, the consonants /l, r/ can occupy the nucleus of the stressed syllable allowing us to investigate whether prosodic emphasis has the effect of enhancing the contrast between nucleus and syllable edge positions in terms of sonority rather than just inducing hyperarticulation. The contrast enhancement hypothesis predicts a weakening of consonantal constrictions in the nucleus in order to increase sonority, whereas hyperarticulation predicts articulatory strengthening of the nucleus, regardless of its vocalic or consonantal identity. We hypothesized that especially for /l/ the contrast enhancement hypothesis would come into effect under accentuation to facilitate sonority expansion. Unexpectedly, acoustic results revealed that neither the narrowing for the vocalic nucleus /e/ nor the constriction at the tongue tip for the consonantal nucleus /l/ are weakened in the accented condition, confirming the hyperarticulation account. Formant measures suggest that the posterior part of the tongue likewise shows hyperarticulation during nucleus /l/. The results therefore support a hyperarticulation rather than contrast enhancement account of prosodic emphasis. Preliminary articulatory results also confirmed these results. [full abstract]

Levels of processing and their interaction in speech production
Adam Buchwald and Michele Miozzo

Accounts of spoken production in psycholinguistics typically include a more abstract level that encodes context-independent representations (such as phonemes/gestures) and a more detailed level that encodes context-dependent representations that are used to generate motor programs for the articulators (indicating syllabification/gestural coordination). In this paper, we demonstrate the importance of understanding the spoken production system as containing levels that interact with one another, but remain distinct and identifiable through errors. We present data from four speakers with acquired deficits subsequent to stroke that affect speech sound production who make deletion errors in consonant clusters. On the basis of error acoustics, we identify individuals as having errors that arise at the context-independent level or at a level where context-dependent representations have been specified. While the presence of deficits affecting distinct levels (tautologically) confirms that these are indeed distinct levels, we also show that the level that the errors arise predicts whether individuals respond to basic speech motor learning treatment; individuals with errors arising at the more abstract context-independent level do not respond to speech motor treatment whereas individuals with errors at the context-dependent level that encodes articulatory detail benefit from these treatments, further indicating the importance and “neurocognitive reality” of this distinction. [full abstract]

Building a proto-lexicon: Does input variability matter?
Helen Buckler and Elizabeth Johnson

Word form representations in the infants’ proto-lexicon are crucially dependent on the distribution of sounds the infant is exposed to in the language environment. Variability in the language environment may have implications for the structure of lexical representations in the proto-lexicon. This study investigates how exposure to accent variability may affect the specificity of early lexical representations.

Using the Headturn Preference Procedure we tested 6-month-olds’ sensitivity to mispronunciations of their own name, one of the earliest word forms acquired. Mono-accented infants’ parents spoke Canadian English, and multi-accented infants had at least one parent who spoke a different variety of English. Listening time to correct pronunciations of the child’s name (e.g. Sam) were compared to listening times to vowel or consonant mispronunciations (e.g. *Sim or *Tam). Mispronunciations were distinct from the accented caregivers’ pronunciation of the name. Mono-accented infants were sensitive to vowel, but not consonant, mispronunciations. Multi-accented infants were not sensitive to vowel mispronunciations. This data indicates that the exposure to greater linguistic variability affects the specificity of early lexical representations, and multi-accented infants’ vowel representations are less well specified than those of the mono-accented infant. [full abstract]

Modeling new conceptions of functional load with perceptual confusability
Zachary Burchill

In linguistics, functional load was first proposed almost a century ago as a way of predicting historical sound changes, namely phoneme loss. Traditionally, accounts of functional load have eschewed behavioral/perceptual data, but more recently, theories have begun to embrace the stochastic nature of phone recognition. Current work suggests that language-wide phonological structure and their evolution in language change can affect perceptual similarity/confusability of phones (e.g. Hall, 2009), but other studies suggest the converse is also true: that perceptual similarity, through phonotactics, can influence language change (Tsui, 2012). In this paper I outline a simple exploratory model capable of simulating the complex interactions of perceptual confusability, the lexicon, and language change, ultimately using this model to produce a measure similar to traditional conceptions functional load, but with three key differences. I outline the theoretical advantages this model offers, illustrating these advantages with short demonstrations of the model’s capabilities. These theoretical advantages primarily stem from the model’s inclusion of perceptual data and its gradient, probabilistic conception of phone recognition. [full abstract]

Dynamic hyper-articulation: Effects of context and lexical representations
Esteban Buz, Scott Seyfarth and T Florian Jaeger

Speakers hyper-articulate words when phonetically-similar alternatives are salient in the speech context. For example, they elongate the aspiration in the target word 'pill' when 'bill' is available as an alternative word. This effect may derive either from competition between words during phonological encoding, or from audience-oriented enhancement of the target word. The first account predicts durational effects based on the position of the contrast in the word, while the second predicts effects that enhance the perceptual distinction between the two words. In two studies, we investigate (1) whether this hyper-articulation is conditioned by the position of the contrast, or by the phonetic cues that are relevant to the contrast; and (2) whether this hyper-articulation is lexically-mediated, or whether it is dependent on the communicative context. We find that hyper-articulation does specifically enhance the phonetic cues to relevant contrasts, but find no evidence that this hyper-articulation is lexically-mediated. The contrast-specific and context-dependent nature of these changes suggest that hyper-articulation may be communicative rather than derived from speaker-internal lexical competition. [full abstract]

Individual specificity, redundancy and the evolution of phonological systems: Intonation in a tone language
Francesco Cangemi, Christian Weitz, Kieu-Phuong Ha, Marc Brunelle and Martine Grice

Speakers encode phonological contrasts redundantly, i.e. using multiple cues which are distributed across different time domains. When multiple cues are available, different speakers are known to capitalize on different cues. This suggests that individual-specific strategies in the encoding of contrasts might be more frequent when contrasts are particularly redundant. Individual specificity in phonetic encoding might thus be expected to be particularly pervasive for contrasts which are usually encoded through non-phonetic devices, but which also have incipient, fading or non-grammaticalised phonetic encoding. We thus explore phonetic encoding of non-lexical or sentence-level contrasts in Northern Vietnamese, a tone language, where they are usually encoded through sentence final particles.

We develop an infrastructure for the quantification of speaker-specific variability, and use it to provide an in-depth analysis of a multi-speaker corpus. By performing both a top-down analysis (hypothesis-testing) and a bottom-up analysis (unsupervised clustering), we document unusual patterns in the phonetic encoding of sentence modality and affect. These patterns document the presence of intonation in a tone language, and suggest that speaker-specific effects might indeed be pervasive in the phonetic encoding of contrasts usually expressed through other devices. [full abstract]

Uncovering the Origins of Nucleus Raising in Liverpool English: Dynamic Analysis of Diphthongs
Amanda Cardoso

English varieties have developed PRICE nucleus raising before voiceless consonants, so that the vowels in the words 'tight' and 'tide' are produced differently but are perceived as the same. Different proposals to account for the origins of these processes have been suggested, including new-dialect formation (Trudgill 1986) and asymmetric assimilation (Moreton & Thomas 2007). The present paper examines PRICE nucleus raising in Liverpool English (Knowles 1973, Cardoso 2015) using a dynamic acoustic analysis of the PRICE vowel to better understand the origins of these processes.

Trudgill (1986) suggests that PRICE nucleus raising results from new-dialect formation. Moreton & Thomas (2007) propose that PRICE nucleus raising is the result of different co-articulatory pressures from the following consonant, which are reanalysed by future generations.

My findings suggest that PRICE nucleus raising is not present in the oldest speakers of the current sample, but that PRICE nucleus raising before voiceless consonants and nucleus lowering before voiced consonants emerge gradually in apparent time, which results in a clear difference in the height of the nucleus before voiceless consonants compared to voiced consonants for young speakers. Therefore, PRICE nucleus raising in Liverpool English is likely not due to new-dialect formation, but may be due to asymmetric assimilation. [full abstract]

L1 USE PREDICTS IMITATION OF METRICAL FEATURES IN A TYPOLOGICALLY DIFFERENT L2
Rossana Cavone and Mariapaola D'Imperio

During the process of second language acquisition, both L1 phonemic and intonational categories might be subject to transfer from the L1 to the L2. Among prosodic features, the early-acquired metrical properties of the L1 might shape L2 prosody and be partly the basis of the perception of Foreign Accent (FA). Moreover, we know that the amount of L1 use can directly impact FA degree in perception for late and advanced L2 learners. This hypothesis predicts that a higher degree of L1 use would result in an increase in prosodically induced FA. Here we investigate the impact of Italian L1 metrical features on the direct imitation of French L2 primary stress and final lengthening, in both a production (Direct Imitation) and a perception (XAB test) study, and whether L1 use is a better predictor of FA than traditional language proficiency scores. Our results show that thee degree of metrical FA can be reduced through limited exposure to L2 input (Baseline vs Imitation). Besides, the newly learnt patterns were retained in memory, during the Generalization Task. Finally, these results give support to the hypothesis that L2 phonetic learning is dependent of the amount of L1 use, and not to standard L2 proficiency. [full abstract]

What does prosodic variation tell us about prosodic organization?
Yiya Chen

In this talk, I will use Chinese lexical tones as a lens to examine the prosodic representation and processing mechanisms required to transform incoming auditory stimuli so as to access the lexicon and comprehend the speech signal. Variation abounds in the multi-level realization of lexical tones in Chinese dialects, which signals indexical and lexical information and is further complicated by various types of tonal sandhi changes in connected speech. How does the Chinese brain deal with the complex patterns of tonal variation, and what can knowledge as such tell us about the prosodic organization of the mental lexicon? To address these questions, I will discuss tonal acoustic patterns, behavioral reaction time data, perceptual judgements, and neural responses from three sets of experiments, each concerning a different type of tonal variation: within-speaker idiolectal pitch variation, within-dialect context-specific tone sandhi variation, and pitch variation due to cross-dialect lexical tonal mapping. With evidence of both acoustic details and abstract categories in tonal processing, I will argue that lexical representation must contain abstract tonal categories stored along with their allophonic variants, while at the same time maintaining relatively weak episodic effects in lexical processing. Furthermore, I will suggest that the findings are compatible with a hierarchical account of how acoustic details are accessed during lexical activation but transformed into abstract representations for the later stage of lexical meaning processing.

Timing lag matters in the perception of Georgian stop sequences by native speakers
Ioana Chitoran and Harim Kwon

This study tests the relevance of perceptual recoverability to phonological grammar, with respect to the relative timing of adjacent consonantal gestures. We conducted two perception experiments in which native Georgian listeners were tested with Georgian C1C2 sequences, along with acoustic/articulatory (EMA) analyses of the stimuli. Production of Georgian consonant sequences was characterized as front-to-back sequences (bg) having a high degree of overlap, and back-to-front sequences (gd) having significantly less overlap (Chitoran, Goldstein, Byrd 2002). This variation was interpreted as speaker-controlled strategy for increasing C1 perceptibility in C1C2 contexts because longer lag favors an audible C1 release, and the presence of a C1 vocalic release favors clearer C1 formant transitions. Based on these findings, we test two hypotheses: (H1) Longer timing lag between C1 and C2 facilitates the recovery of C1 gestures; (H2) C1 vocalic releases facilitate C1 recovery. The results support H1, but not H2. Native Georgian listeners are sensitive to differences in timing lag: longer onset lag facilitates C1 recovery in front-to-back sequences and longer release lag does so in back-to-front sequences; the presence of a vocalic release does not facilitate C1 recovery. The results support the inclusion of timing lag rather than recoverability constraints in the grammar. [full abstract]

Accent-meter/tune alignment in Japanese vocal music
Sunghye Cho

There is an ongoing debate about the nature of pitch accents. Even in Tokyo Japanese (TJ), which is the most widely cited pitch-accent language, the pitch accent is analyzed in two distinct ways: an accentual approach and a tonal one. In the accentual approach, an accent marks the position of a pitch drop from High to Low, whereas in the tonal approach, tones are directly aligned with morae without accent. The accentual approach predicts that accented syllables have a phonological prominence similar to stress in English, whereas the tonal approach expects that the position claimed to be accented is not prominent. My research question is whether a pitch accent in TJ marks a prominence. If it does, this would suggest that the accentual analysis is on the right track. One way to answer this question is to examine how Japanese lyrics align with musical tunes. It is well known that stressed syllables occur at strong beats of the meter in vocal music of stress languages, such as English. If an accent in TJ is a phonological prominence similar to stress, accented morae occur at strong beats. In this study, I examine this question with traditional Japanese songs. [full abstract]

An empirical and computational study of generalized adaptation to natural talker-specific VOT
Eleanor Chodroff, Alessandra Golden and Colin Wilson

Adaptation to the speech of novel talkers could be facilitated by knowledge of acoustic covariation across phonetic categories. Strong correlations have been documented for voice onset time (VOT) of word-initial aspirated stop consonants: e.g., talkers with longer mean VOT values for [pʰ tʰ] tend to have longer values for [kʰ] (Theodore et al., 2009; Chodroff et al., 2015). Previous studies of perceptual adaptation and phonetic imitation have provided evidence that listeners employ these correlations to generalize talker learning across categories (e.g., Theodore & Miller, 2010; Nielsen, 2011), but their implications are limited by the use of exaggerated stimulus values, extensive exposure to novel talkers, or limited stimulus variability. The present study investigated rapid perceptual adaptation and generalization with more natural and variable stimuli. In each condition, listeners were exposed to a relatively longer or shorter VOT distribution for two aspirated stops (e.g., long [pʰ tʰ]) and performed a two-alternative forced choice task on the held-out stop (e.g., long vs. short [kʰ]). Listeners rapidly generalized talker VOT to the held-out stop in all but one condition. A Bayesian model of the experimental results was developed with a prior distribution on talker-specific phonetic realization that explicitly encodes VOT covariation across aspirated stops. [full abstract]

Estimating segments' cost using cross-linguistic information
Uriel Cohen Priva and Emily Gleason

Phonology and phonetics often appeal to binary hierarchies between phonological and phonetic contrasts, e.g. voiced obstruents > voiceless obstruents, which stem from observed linguistic patterns. We propose that such phonological and phonetic hierarchies can be deduced from aggregating the distributions of sounds within multiple languages. In each language we calculate the information content of individual sounds, and average the resulting estimates across multiple languages to smooth out language-specific noise. The resulting cost estimate system has several desirable properties: (a) It predicts multiple well-known phonological and phonetic contrasts (e.g. in place, manner, voicing). (b) It is real-valued rather than binary, therefore making more concrete testable predictions and allowing better integration with Harmonic Grammars. (c) It does not take the phonological patterns it aims to predict as input, e.g. the input does not "know" that final-devoicing exists, but predicts that voicing is more costly. Finally, (d) it correlates better with within-language predictability for held-out data than typological cross-linguistic frequency with within-language frequency. The results signify the importance of information theoretic properties of segments in phonology, and provide a way of making more concrete predictions for which phonological processes are likely affect which language, providing a partial solution to the actuation problem. [full abstract]

L1 influence on L2 assimilation: An EPG study of English /n/+stop sequences
Laura Colantoni, Alexei Kochetov and Jeffrey Steele

This paper examines how gradient/variable patterns of nasal place assimilation in English (Ellis & Hardcastle, 2002) are acquired by advanced L2 English learners, whose native languages display different canonical realizations of nasal + stop sequences. Linguopalatal contact data using electropalatography (EPG) were collected from 3 speakers each of French, Japanese, and Spanish as well as from 2 native English controls. Sentences included words with a final /n/ followed by a word-initial /k/ (target items) or /t/ and /h/ (control items). Measurements of tongue front/back position and degree of palatal contact were extracted from the acoustically-defined nasal interval. The results showed that the L2 production of /n/+/k/ sequences was relatively uniform within each language group, and to a large extent, resembled the corresponding L1 patterns observed in data collected previously from the same speakers. These findings are consistent with previous studies in revealing that even advanced L2 learners continue to use L1 coarticulatory patterns, particularly across words (Zsiga, 2003). This, in turn, provides support for positionally-based acquisition models (Flege, 1995). [full abstract]

Perception of acoustic, informational and structural prominence in English, French, and Spanish
Jennifer Cole, Jose Ignacio Hualde, Caroline Smith, Christopher Eager, Tim Mahrt and Ricardo Napoleão de Souza

Nuclear prominence is assigned to a word based on information status in some languages, while its location is fixed at the end of a phrase in others. We test how this difference affects prominence perception, comparing English, Spanish and French, languages that differ in the strength of the link between informational, positional and acoustic prominence. Using the method of Rapid Prosody Transcription, we compare prominence perception in English, Spanish and French in relation to phrasal position and word frequency (a correlate of information status), and by directing listeners’ attention to acoustic criteria or to informational (“meaning-based” criteria). Prominence annotations were collected for spontaneous speech excerpts from 30 listeners of each language. Statistical results of mixed-effect regression show that word frequency as an informational factor most strongly influences prominence ratings for English, where prominence is the primary expression of information status. But despite differences in the phrasal location of nuclear prominence among these languages, the structural factor of adjacency to a prosodic boundary uniformly influences prominence perception based on acoustic criteria in all languages. Listeners in all three languages tend to perceive an acoustically-cued structural prominence on the phrase-final word, suggesting the primacy of a structural nuclear prominence in prosodic theory. [full abstract]

Acoustic-phonetic modelling of historical and prehistoric sound change
John Coleman

The comparative method reconstructs hypothesized ancestral forms of words based upon their modern or historical written forms. Computational modeling of language phylogenies enables ancestral forms, sound change rules, phylogenetic trees and chronologies to be tested, but the data used is text: alphabetic transcriptions or features representing traits. Recently, progress has been made in phylogenetic modeling of continuous functions: curves and surfaces. As speech parameters such as formant frequencies, amplitude contours, and spectrograms can be represented using continuous functions, it is becoming possible to model linguistic history and prehistory using these methods and to reconstruct audible sound files instantiating hypothesized spoken forms from the past, including distant ancestral pronunciations and intermediate forms at each generation. These acoustic simulations of sound change are testable against observations or other sources of data. Gradual, incremental, acoustic modeling of sound changes enables us to estimate the varying rates at which words have changed over millennia. We present and discuss a selection of temporal maps of changing pronunciations in the long history of English over timescales of hundreds and thousands of years, with audio simulations of sounds from the distant past. [full abstract]

Vowel lengthening in syllables without vowels
Jamison Cooper-Leavitt and Rachid Ridouane

There is a pragmatic context in Tashlhiyt, ‘of-course’ response, which provides a context where vowel lengthening plays an important prosodic role. It is used when the responder believes the answer should already be known. Speakers show this by an exaggerated lengthening of the vowel of the final syllable. What is lengthened in items where final syllables are composed of consonants only?

Results show that the location of lengthening is determined by several factors, leading to a probabilistic distribution. In addition to the systematic lengthening of the final syllable when its nucleus is a vowel, lengthening occurs in 100% of utterances on the final heavy syllable when its nucleus is a liquid or nasal, in 48% of utterances on the final light syllable when its nucleus is a liquid or nasal, and in 20% of utterances on the final syllable when its nucleus is an obstruent. Our findings also show that in the absence of a vowel in the final syllable, lengthening is realized on the initial vowel, on a schwa-like element within the final syllable, or on a nucleus consonant of the final syllable.

These findings bear on Tashlhiyt syllable structure and its relation to sonority, weight and schwa vowel. [full abstract]

Like a square peg in a round hole: Why contour shape matters for learning new intonation patterns
Mariapaola D'Imperio and James Sneed German

This study uses dialect imitation across typologically different prosodic systems to explore the type of phonetic detail that is accessible from recently experienced intonation patterns. Specifically, Singapore English (SgE) speakers imitated Yes/No questions produced by an American English (AmE) speaker. Although both varieties use a final rising contour in this context, the AmE contour involves a rightward inflection (S-like curvature), while the SgE contour is concave. We therefore used a 3rd-order polynomial fitting analysis to compare the degree of inflection in the SgE baseline tokens, the AmE contour, and the imitation tokens. The SgE speakers were clearly able to approximate the overall f0 scaling of the AmE pattern across different regions of the contour, however, they showed no robust tendency to approximate the higher order shape of the contour. This suggests that speakers could not hear the shape differences due to perceptual assimilation effects, in which case the imitated tokens are variants of native categories with adjustments to scaling parameters that better match the targets. Alternatively, SgE speakers may lack the articulatory practice needed to produce the inflection points. In either case, these findings reveal that new contour shapes are not immediately accessible to the perception and/or the production system. [full abstract]

When oui becomes ou[i̥]: The role of vowel type, preceding consonant and lexical frequency on total final vowel devoicing in Continental French
Amanda Dalola

Phrase-final vowel devoicing (PFVD), e.g. mais oui_hhh, is a phenomenon in Continental French (CF) in which utterance-final vowels devoice partially or completely to yield fricative-like whistles. Studies have examined the phonological conditioning of partially devoiced vowels, but none has examined total vowel devoicing (TVD). This is problematic because TVD is attested outside of phrase-final position in varieties other than CF with different segmental conditioning than combined PFVD. This study seeks to isolate occurrences of TVD and study their segmental and structural preferences, namely vowel type, voice and manner of the preceding consonant and lexical frequency.

Devoiced vowels were elicited from native CF speakers in a two-part reading and roleplaying task that targeted phrase-final tokens of /i/, /y/ and /u/ with preceding consonants that were voiced and voiceless and of different manners of articulation. Target words were of varying frequencies.

Results revealed that high vowels do not undergo TVD at similar rates. There was no main effect for lexical frequency or voicing of the preceding segment, but there was a significant interaction between /u/ and voicing. These findings reveal that TVD in CF behaves neither entirely like the larger combined phenomenon of PFVD, nor like the crosslinguistic phenomenon of TVD. [full abstract]

Perception of anticipatory labial coarticulation by Belgian French blind listeners: A comparison with sighted listeners in audio-only, visual-only and audiovisual conditions
Véronique Delvaux, Kathy Huet, Myriam Piccaluga and Bernard Harmegnies

Our aims were to investigate the dynamics of the perception of the /i-y/ contrast by adult Belgian French blind listeners, and to position the performances of the blind listeners with respect to the range of performances demonstrated by sighted controls in audiovisual speech perception. 8 blind and 8 sighted listeners completed two tasks on pairs of stimuli gated from original [agi] and [agy], a two-alternative forced choice identification task and an AX discrimination task. Both tasks were performed on stimuli which were presented in quiet vs. (acoustically-) noisy conditions, combined with three sensory modalities: audio-only (for blind and sighted listeners), audio-visual and visual-only (for sighted listeners only). Results: (i) in the audio-only condition, blind listeners overall outperformed sighted listeners, even more so for a lower signal-to-noise ratio; (ii) overall, sighted listeners exhibited strong visual enhancement, i.e. better performances in the audio-visual modality than in the visual-only and audio-only, and stronger visual enhancement when acoustical noise was added; (iii) to some extent, the enhanced performances of blind listeners were mediated by the perceptual task to be performed, i.e. discrimination vs. identification; (iv) complex interactions were observed between groups of listeners, conditions and modalities, in terms of relative timing between perceptual scores and the date of the earliest available information in the audio and visual streams [full abstract]

Dynamics and articulatory control in Amharic ejectives
Didier Demolin

Ejectives are produced with a glottal closure at the larynx and another at various places of the vocal tract. The temporal timing and coordination of the gestures made at the glottis and in the vocal are crucial to understand how these sounds are produced and how they differ from pulmonic stops. Ejectives are produced by an elevation of the larynx that makes the space between the glottal and oral closures smaller. This volume reduction triggers and elevation of the pressure in the cavity sealed between the glottis and the oral closure. When the oral closure is released a sharp and intense burst is produced.

Amharic has a set of ejective consonants [pʼ, tʼ, kʼ, tʃʼ, sʼ] that can be short and long. There are several interesting questions related to their production mechanism and to their difference with pulmonic stops: (1) is the glottal closure released separately or simultaneously with the oral closure? (2) How are ejectives fricatives and affricates produced? (3) Is the production of geminated alveolar ejectives fricatives really possible? If so what makes the difference with singleton ejectives fricatives? This study presents results of experiments made by using aerodynamic and acoustic parameters on Amharic ejectives. [full abstract]

Speech Prediction from Subphonemic Production
Donald Derrick and Daniel Bürkle

Some North American English speakers demonstrate anticipatory coarticulation in VɾVɾV sequences such that during the first vowel, these speakers move and hold the tongue tip higher if the last vowel is rhotic (as in "editor") than they do if the last vowel is non-rhotic (as in "edit a"). Such anticipation of the tongue position of the final vowel shows articulatory planning at a level lower than the syllable, phoneme, or feature, analogous to Rosenbaum’s end-state comfort effect in hand grasps.

This paper presents results from an eye-tracking study that suggest this anticipatory coarticulation is used in speech perception. We asked North American perceivers to identify the spoken word or phrase in VɾVɾV sequences produced by both speakers who produce this coarticulation and those who do not. We find that the tongue-tip differences are perceptible - anticipatory coarticulation allows perceivers to identify the word or phrase more quickly. This result provides evidence for increased speed of listener’s speech perception as another benefit of anticipatory sub-phonemic coarticulation. [full abstract]

Processing of coarticulatory nasalized vowels and phonological nasal vowels in Canadian French
Félix Desmeules-Trudel and Tania Zamuner

Research has shown that speakers use coarticulation to anticipate segments in the speech stream. Some models of spoken word recognition postulate that phonetic integration is continuous, while others assume a buffer in which phonetic information is gathered and organized before identification. In Canadian French (CF), vowel nasalization can be coarticulatory or phonologically specified. We investigate how the phonetic information of both vowel types is integrated. CF listeners were tested using an eyetracking visual world paradigm with unambiguously and ambiguously nasalized stimuli. Results revealed that participants appropriately categorized the unambiguous stimuli and that longer vowel nasalization on ambiguous stimuli led to higher proportions of identification and fixation to words containing phonologically nasal vowels. However, participants were generally biased towards the word containing an oral vowel (in a VN sequence) in ambiguous conditions. The results provide support for both models of phonetic integration, i.e. coarticulation is continuously integrated and phonologically specified information that potentially originates from multiple sources is buffered. Consequently, we suggest that vowel orality at vowel onset is a strong cue that biases the participants towards the identification of a vowel as phonologically oral when nasalization timing is ambiguous. [full abstract]

Prosodic phonological characteristics of speech directed to adults and to infants with and without hearing impairment
Laura Dilley, Evamarie Burnham, Elizabeth Wieland, Derek Houston, Maria Kondaurova and Tonya Bergeson

Infant-direct (ID) and adult-directed (AD) speech are distinguished via a variety of acoustic-prosodic characteristics, but little is known about how these differences map onto phonologically relevant constructs, such as pitch accents and phrasal boundaries, or how a child’s hearing impairment affects prosody. In two studies, trained analysts coded prosodic characteristics in corpora of mothers reading to their children (ID condition) or another adult (AD condition); acoustic characteristics (e.g., fundamental frequency, speech rate) were also measured. In Study 1, 48 mothers read a storybook to their typically-developing infants and an experimenter when the infants were approximately 3, 9, 13, or 20 months old. In Study 2, 11 mothers read a storybook to their hearing-impaired child or an experimenter. Study 2 also included mothers of typically-developing children; each normal-hearing control child was matched to a hearing-impaired child. ID speech was associated with a greater density of pitch accents and prominences than the AD style. There was no difference in the distribution of phrase boundaries across speech styles. Hearing status did not mediate the effect of speech style on prosody. The results suggest that acoustic differences distinguishing ID speech and AD speech map onto a combination of phonological structural and gradient paralinguistic characteristics. [full abstract]

Network structural equivalence and the reversal of the Southern Vowel Shift
Robin Dodsworth and Richard Benton

Network structural equivalence is explored as a factor in the reversal of the Southern Vowel Shift (SVS) in Raleigh. Structural equivalence refers to the extent to which nodes inhabit similar positions within a social network, as determined by the ties they have in common. The network data consist of a bipartite network of schools and speakers, thus approximating childhood and adolescent network. In Raleigh, the 5 front vowels all began retreating from their SVS positions around 1950, concurrent with white collar in-migration. F1 and F2 were measured at 25% of vowel duration in force-aligned conversational interviews with 155 native Raleigh speakers. A matrix of structural equivalence distances between pairs of speakers was calculated on the bipartite network. This matrix is a fixed effect in each of a set of quadratic assignment procedure (QAP) regressions; the dependent variable in each model is a matrix of differences in mean Z2-Z1 for a given vowel. For three vowels, the interaction among age differences, occupation, and structural equivalence is significant: pairs of speakers far apart in age are more different linguistically than speakers close in age, and greater network distance between 2 speakers corresponds to greater linguistic difference, especially among white collar speakers. [full abstract]

Social dynamics and phonological strength: Post-nasal devoicing in Tswana
Grzegorz Dogil, Jagoda Bruni, Daniel Duran, Justus Roux and Andries Coetzee

This study describes the influence of social and political changes in the South African phylum. We focus on Tswana, which has a phonetically marked post-nasal strengthening process. Post-nasal devoicing is often considered a phonetically unintuitive phenomenon. In analyzed corpus data, eighty percent of post-nasal stops are devoiced. Moreover, twenty-five percent of those voiceless plosives are produced as ejectives. In the present study we use computational simulations in order to test predictions about the direction of a sound change apparently happening in Tswana. We use corpus data as a society sample and put the two observed speech varieties (voiced stop vs. strong post-nasal devoicing) through several multi-agent simulation settings modeling social dynamics as well as speech production-perception and memory organization at the individual level. In order to reflect the parochial society structure corresponding to largely closed language communities, we use a social network with a number of distinct parishes (clusters) with relatively small social distances within each parish but relatively large distances across parish boundaries. The linguistic situation of Tswana after 1994 (opening to influence of other languages) is modeled by a network which has small distances over the entire population. [full abstract]

Acquisition of word-level prominence in L2 English by Canadian French speakers
Guilherme Duarte Garcia and Natalia Brambatti Guzzo

Canadian French (CF), unlike European French, is argued to have word-level prominence (Walker, 1984; Paradis & Deshaies, 1990). However, word-internal constituency in CF is disputed: whereas some authors argue that French has foot structure (e.g., Goad and Buckley, 2006), others argue that it has no such a constituent (e.g., Jun & Fougeron, 2000; Özçelik, 2011). English, on the other hand, has word-level stress, which is phonetically correlated with pitch and duration (Fry, 1955; Lehiste, 1976), and influenced by syllable weight. With regard to word-level constituency, English is a trochaic language, and feet are built iteratively (Hayes, 1982). Given the differences between CF and English, we investigate how CF speakers adapt to the stress and rhythmic patterns in English as a second language. Specifically, we report on a production experiment that examines (a) whether CF has properties related to prominence and word-internal constituency and (b) how CF speakers whose L2 is English (advanced proficiency) realize English stress and rhythmic patterns. Our data show no consistent phonetic evidence for word-level prominence nor word-internal constituency in CF. However, in L2 English, CF speakers clearly make use of duration to signal word-level prominence/constituency, which mirrors the patterns found in native controls. [full abstract]

Phonology Modulates the Illusory Vowels in Perceptual Illusions
Karthik Durvasula, Ho-Hsin Huang, Sayako Uehara, Qian Luo and Yen-Hwei Lin

Native listeners perceive illusory vowels when presented with sound sequences that do not respect the phonotactic constraints of their language (Dehaene-Lambertz, Dupoux, & Gout, 2000; Dupoux, Kakehi, Hirose, Pallier, & Mehler, 1999). For example, when a Japanese listener is presented with [ebzo], they may actually perceive [ebuzo] given that [bz] is an illicit consonant sequence in Japanese. It has been argued that the illusory vowel can only be the phonetically minimal vowel in the language (Dupoux, Parlato, Frota, Hirose, & Peperkamp, 2011), and that multiple illusory vowels cannot be perceived during speech perception (Uffman, 2006). Here, we show that there can indeed be different illusory vowels in different contexts, and that the quality of the illusory vowels is itself modulated by the phonological patterns in the language. Specifically, when presented with phonotactically illegal alveopalatal coda consonants, Mandarin listeners (but not English listeners) perceived an illusory [i]; however, in illegal alveolar stop coda contexts, they perceived a [ə]. This is the pattern of illusory vowels expected based on the phonological patterns on the language. [full abstract]

Using ANNs for vowel identification from V-to-V coarticulation in non-harmonic VCV sequences
Indranil Dutta, Irfan S. and Harsha K.R.S.

Acoustic variation due to vowel-to-vowel (henceforth, V-to-V) coarticulation, once phonologized, has been shown to be a conditioning factor for development of vowel harmony patterns (Przezdziecki, 2000; Ohala, 1994). The resultant reduction in phonetic distinctiveness between vowels, however, is known to be compensated perceptually (Beddor et al., 2002). In this study, we show that while Telugu vowel harmony patterns advanced from the anticipatory direction leading to neutralization of vowel contrasts between /iCu/->[uCu] sequences, the extent of carryover coarticulation is still greater in non-harmonic contexts. We report on results from a single-layered Artificial Neural Network (ANN) model that pre- dicts the identity of V1 from V2 (carryover) consistently better than the V2 from V1 (anticipatory), in Telugu non-harmonic #. . . V1CV2. . . # sequences. The greater extent of carryover coarticulation exerts an assimilatory effect that may aid perception, and maintenance of contrast. Our findings suggest that while anticipatory coarticulation, and failure of perceptual compensation may have lead to the development of right-to-left harmony/assimilation in Telugu, it is in the carryover direction that an assimilatory process may indeed serve to maintain contrast. [full abstract]

Contrast preservation at the level of the individual: Evidence from Spanish plosive lenition
Christopher D. Eager

Contrast preservation has been proposed as a constraint on the phonetic realization of phonemes, and as a motivating factor in sound change. Recent quantitative diachronic analyses have shown that lenition rarely results in neutralization and that contrasts with higher functional loads are significantly less likely to be lost. The results of these studies imply a synchronic role for contrast preservation at the level of the individual speaker for which little previous research exists. In this study, synchronic individual variation in the ongoing simultaneous lenition of intervocalic /ptk/ and /bdg/ in 56 speakers from 3 dialects of Spanish is examined. Mixed effects regressions are used to obtain point estimates of the voiceless period duration and intensity difference of /p/, /t/, and /k/, and the intensity difference of /b/, /d/, and /g/, and a generalized additive model is used to show that, in a gradient manner, the more an individual voices /p/, /t/, and /k/, the more the same individual weakens the constriction of the corresponding voiced consonant, and vice versa. The results support the hypothesis that contrast preservation plays a role in individual speakers’ phonologies. [full abstract]

Rethinking reduction on the basis of phonetic variation in a discourse marker
Mirjam Ernestus and Rachel Smith

Many word tokens are produced with fewer or weaker segments in spontaneous conversations than in formal speech. This study reports on detailed research on the phonetic variation of the Dutch word eigenlijk /'ɛιxәlәk/ 'actually', based on 159 tokens. The hand crafted transcriptions show that the variation in the pronunciation of eigenlijk is substantial, both at the segmental level (pronunciations range from /'ɛιxәlәk/ to /'ɛιxk/ to /'ɛιk/) and at the subsegmental level (e.g. /l/ is realized in full, devoiced, as lateral frication or as a simple F2 dip). We also observed two prosodic effects so far undocumented for reduction. First, reduction is also extensive for tokens carrying primary or secondary accent. Second, the monosyllabic tokens of eigenlijk are more often followed by an unstressed syllable than polysyllabic tokens, suggesting an effect of phrasal rhythm on reduction. These results raise questions among other topics about 1) Should the full form or the most frequently occurring one be considered as canonical? 2) which forms are stored in the mental lexicon and with which detail? 3) how do listeners process this variation? Further research has to reveal to what extent our findings are specific to (Dutch) discourse markers. [full abstract]

Articulatory uniformity in Suzhou fricative vowels
Matthew Faytak

Using uniform articulatory strategies for producing similar acoustic contrasts is a factor known to shape phonological inventories to a limited extent. This study examines fricative noise production strategies in the fricatives /s ɕ/ and "fricative vowel" of Suzhou Chinese; the latter is a phonemic vocoid produced with fricative noise often approaching, but not identical to, a syllabic [ʑ]. A preliminary ultrasound study of eight speakers (4F, 4M, median age 20.5, max. age 51) was carried out to assess tongue shape in the production of the aforementioned sounds. Results suggest a substantial minimization of distinct strategies, with the fricative vowel nearly identical to the articulation of one of the fricative consonants within each speaker. More tentatively, data from the oldest speaker also suggest that the fricative vowel's articulatory strategy has changed in apparent time, "skipping" from an [i]-like tongue posture in his productions directly to a [ɕ]-like tongue posture in younger speakers. I discuss this apparent "re-use" of articulatory strategies (first from /i/, then from /ɕ/) in light of existing theories of parsimony and economy in phonological inventory structures. [full abstract]

The effects of listener age and language experience on talker identification
Natalie Fecher, Katrina Aranas and Elizabeth K. Johnson

Previous research has shown that adults identify talkers better in a familiar versus an unfamiliar language (the language familiarity effect, LFE), and that adult-like talker identification capabilities do not emerge in children until about age 10. Here, we use two different tasks to investigate whether increasing phonological competence can account for both of these observations. We tested native English-speaking 6-year-olds and adults on an AX talker discrimination task (Experiments 1 and 3), as well as on a voice line-up talker recognition task (Experiment 2). Recordings of four English-Polish bilingual talkers were used in all experiments. We predicted that the strength of the LFE would increase with age, but this is not what we found. In the voice line-up task, we observed a strong LFE in both adults and children. In the same-different task, however, we observed no LFE for either age group. In all three experiments, we replicate past studies showing that adults identify talkers more accurately than do children. We speculate that no LFE was found in the discrimination experiments because this sort of same-different discrimination task taps into a lower level of processing than voice line-up tasks. Our findings offer new insights into the source of the LFE. [full abstract]

Variable aspiration of Spanish coda /s/: Laboratory evidence and Stochastic OT modeling
Valentyna Filimonova and Kelly Berkson

The focus of this study is Spanish coda /s/ aspiration, a complex phenomenon known for its high levels of inter- and intra-dialectal variation. As the long tradition of dialectology and sociolinguistic accounts keeps identifying multiple social and linguistic factors that favor certain realizations, a growing number of acoustic analyses are revealing additional variants previously overlooked by theorists and applied linguists. One such variant is a voiced glottal fricative [ɦ]. In this study, acoustic analysis of data elicited in a controlled experimental design reveals that this variant is the preferred realization of coda /s/ in both Argentinean and Puerto Rican Spanish. Four other realizations were also found. Two stochastic Optimality Theoretic models were generated, and accurately predicted the real-life probabilities of each variant in each dialect. In the results of this modeling, a clear phonological hierarchy of manner over place over voicing constraints emerged, with each constraint tier characterized by dominance of markedness over faithfulness constraints. Taken together, the fine-grained acoustic analysis and the stochastic modeling contribute to a more nuanced description of this complex variable phenomenon and provide insight into the factors that govern Spanish coda /s/ aspiration. [full abstract]

The phonological representation of Japanese vowel devoicing
Marco Fonseca, Maria Cantoni and Thaïs Cristófaro Silva

Japanese vowel devoicing offers an important contribution to the understanding of sound change and the nature of phonological representations. Some studies report that Japanese vowel devoicing results in a deleted vowel, suggesting that it is an abrupt phenomenon. Other studies suggest that Japanese vowel devoicing is a gradual phenomenon, in the sense that temporally reduced vowels are observed before deletion takes place. Whether Japanese vowel devoicing is abrupt or gradual impacts the nature of phonological representations: is the devoiced vowel deleted or does it leave a trace in the phonological representations? The goal of this study is to provide a better understanding of Japanese vowel devoicing as a sound change phenomenon and contribute towards the debate on the dynamic nature of phonological representations. It will be argued that lab speech is an important tool to understand how phonological representations are built and evolve in time. We suggest that the results presented in this paper provide evidence for the dynamic nature of phonological representations. [full abstract]

Regional variation in formant dynamics and the phonologization of pre-velar raising in American English
Michael J. Fox and Jeff Mielke

The phonologization of phonetic effects is central to our understanding of phonological phenomena. To understand how a mechanical phonetic effect is exaggerated and incorporated into the phonological system, it is instructive to compare phonetic and phonological versions of the same pattern when they can be observed coexisting in related language varieties. Herein we examine the formant dynamics of /æ/ in two American English varieties (Wisconsin and North Carolina) that differ in the raising effect of a following /ɡ/, to access whether raising is a consequence of coarticulation or its phonologization. Read speech data come from 84 speakers from Northwest Wisconsin and 40 speakers from Western and Central North Carolina. GAMMs were fit to measurements of F1 and F2 at 50-equidistant time-points for all /æ/ vowels (n = 16,588). The movement of the transition into the vowel for bag found in Wisconsin is an indication that the transition is no longer a coarticulatory effect from the following /ɡ/, but rather is a phonologized property of the class of words containing /æɡ/. This claim is strengthened by the fact that the time-point of the transition varies by duration in Wisconsin, but is stable in North Carolina, indicating a dialect-specific pattern of phonologization. [full abstract]

Prosodic Organization of Spontaneous Spanish-English Bilingual Speech
Melinda Fricke, Marianna Nadeu and Michael Maslowski

This study uses Rapid Prosodic Transcription (Cole et al., 2010) to derive probabilistic measures of perceived prosodic prominence and boundary strength for stretches of English, Spanish, and codeswitched spontaneous speech. We selected 127 utterances from the Bangor Miami Corpus (Deuchar et al., 2014) and presented them to 20 highly proficient English-Spanish bilinguals who reported codeswitching on a daily basis, and who had no prior training in prosodic annotation. Listeners performed two labeling tasks (with task order counterbalanced), marking either prosodic prominences or boundaries by clicking on words on a computer screen. Mixed effects logistic regression was used to predict the probability that any given word would be perceived as prominent, or as the end of a prosodic constituent. Preliminary analyses returned main effects of syntactic category and codeswitch status, along with an interaction between these factors. Codeswitched words were more likely to be considered the beginning of a new prosodic chunk (though only for certain syntactic categories). In addition, listeners tended to perceive codeswitched nouns and determiners as prosodically prominent, consistent with the idea that codeswitching can be used to signal narrow focus (Olson, 2012). Additional analyses are underway to determine the acoustic correlates of perceived prosodic structure. [full abstract]

Prosodic organization and microprosodic effects in Shanghai Chinese
Jiayin Gao and Pierre André Hallé

This study bears on the F0 perturbation effect caused by intervocalic obstruents in Shanghai Chinese (SHC). Previous studies on SHC report a systematic F0 lowering effect with voiced stops, as is found in many other languages. Our study focuses on the impact of prosodic context on F0 perturbation. Prosodic word-tone in SHC is determined by the tone of the first syllable (S1): it is falling vs. slightly rising for S1 in tone T1 vs. T2, respectively. We compared the F0 contour in the second syllable (S2) of disyllabic words in these two word-tone contexts: S1 in tone T1 vs. T2 (post-T1 vs. post-T2). We found an F0 lowering effect with voiced obstruents in the post-T2 but not in the post-T1 context. We propose that the timing of the F0 contour of a prosodic word is planned independently from the segmental timing. On this account, F0 height at S2’s rime onset may be influenced, independently, by both its time location relative to the word-level F0 contour (a “ballistic” effect) and by the voicing of S2’s onset (a “depressor” effect). Ballistic and depressor effects run in opposite vs. same directions for falling vs. rising word-tones, respectively, explaining our data. [full abstract]

Embodied dynamics: A unified approach to local, non-local and global coarticulation
Bryan Gick, Chenhao Chiu and Ian Stavness

Joos (1948, Language Monogr. 23) pioneered the modern study of coarticulation with his theory of Overlapping Innervation Waves, postulating that the complex dynamics of interactions between nearby speech sounds could be understood as simple overlap, or superposition, of muscle activation patterns. The present paper describes an implementation of Joos’ proposals within a modular neuromuscular framework (see Gick & Stavness 2013, Front. Psych. 4, 977) using the ArtiSynth (www.artisynth.org) biomechanical modeling toolset. This implementation, using additive muscle activations built on top of embodied dynamics, is applied as a solution to three superficially distinct instances of coarticulation, over several timescales. Coarticulation is construed here as occurring locally (as with immediately adjacent speech sounds), non-locally (as with the long-distance interactions observed in harmony systems), or globally (as with articulatory settings, where a setting may affect every sound in a language [see Gick et al. 2004, Phonetica 61, 220-233]). Simulations show how superposed activations of neuromuscular modules can produce complex coarticulatory outcomes with no extrinsic model of coarticulation (i.e., with only embodied dynamics). Embodied dynamics can be thus shown to provide a unified, context-independent model for local CV coarticulation, long-distance harmony and articulatory settings, as well as other speech and non-speech interactions. [full abstract]

Representational dynamics in sound structure planning
Matthew Goldrick

Lexical neighbors (non-target words overlapping in form with the target) induce a range of effects on phonetic realization. When response preparation is difficult or disrupted, non-target representations serve to attract target articulations, resulting in productions. that reduce the contrast between the target and co-activated neighbors. When response preparation is facilitated, neighbors serve to repel the target, resulting in enhancement of contrasts between targets and neighbors. Parallel effects have been observed in bilingual speech, as well as in non-communicative motor movements (saccadic eye movements and reaching). This suggest that enhancement and reduction reflect general principles of motor planning. A dynamical framework for lexical access can account for this diverse array of effects. Reduction is the consequence of coactivation. Spreading activation enhances the activation of neighbors, causing the output of lexical access to be a blend of target and non-target properties. The resulting presence of neighbor articulatory properties during target production produces reduction. Building on exemplar accounts  and adaptive speaker models, enhancement reflects learned inhibition of competing neighbor forms. This interacts with general selection processes to produce strong suppression of non-target forms, greatly reducing lingering effects of coactivation. This produces enhancement relative to cases that lack learned inhibition [full abstract]

More is more: how trying to learn multiple aspects of language at once can actually help
Sharon Goldwater

The term "bootstrapping" appears frequently in the literature on child language acquisition (especially with respect to syntax and semantics), but is often defined vaguely (if at all) and may mean different things to different people. In this talk, I define bootstrapping as the use of structured correspondences between different aspects of linguistic structure as a way to aid learning, and discuss how probabilistic models can be used to investigate the nature of these correspondences and how they might help the child learner. I will present example models of early language acquisition tasks involving word segmentation, phonetic learning, and word meaning. My work with these models on naturalistic corpora illustrates how jointly learning multiple aspects of language at once can actually make the learning problem easier, rather than more difficult, and suggests the need for more research integrating multiple levels of linguistic structure.

Shanghai Chinese obstruent durations vary with voicing: A phonological or phonetic effect?
Pierre Hallé and Jiayin Gao

Voiced obstruents are shorter than voiceless obstruents in many languages. This cross-linguistically robust finding has usually received phonetic interpretations in terms of aerodynamic constraints in the interaction between the laryngeal and supralaryngeal articulations, which indeed make sense when [+voice] is realized as glottal pulsing, that is, phonetic voicing. We examine here whether this effect is also found when obstruent phonological voicing is defined at an abstract level and may not surface as phonetic voicing. Shanghai Chinese (SHC) offers such a situation, whereby the phonological voicing of obstruents is indexed by tone height and not phonetic voicing in word-initial position and vice-versa in word-medial position. We find that the effect of voicing on obstruent duration holds for SHC obtruents regardless of within-word position. This suggests that the duration effect cannot be solely explained by phonetic constraints and therefore must reflect phonological representations. A perceptual experiment suggests that the pattern of durations in obstruent onset plus rime syllables affects their categorization in terms of onset phonological voicing. This further support our proposition that onset-rime duration pattern is part of the phonological representation of onset voicing. [full abstract]

Phonological Influence in Third Language Acquisition: L2 Spanish Effects on the Production of L3 Portuguese Voiced Stops
Sarah Harper

This study examines the influence of a learner’s L1 and L2 on the acquisition of sounds in a third language (L3), expanding upon previous research to investigate whether the tendency for greater L2 influence on the L3 holds for allophonic contrasts acquired relatively late by adult L2 learners. Specifically, we examine how native English speakers’ knowledge of voiced stop lenition patterns in their L2, Spanish, influences their production of voiced stops in an L3, Brazilian Portuguese, that lacks this lenition process. An acoustic analysis incorporating relative intensity measurements commonly used in voiced stop lenition studies compared the production of voiced stops by L2 and L3 Portuguese speakers in each of their languages, with the results providing strong evidence for L2 Spanish influence on the production of L3 Portuguese voiced stops. Unlike the L2 Portuguese speakers, who produced stops of relatively similar intensity in both English and Portuguese, L3 Portuguese speakers produced voiced stops with higher relative intensity measurements in Portuguese than in English, and with values extremely similar to those observed in Spanish. This is taken as strong support for an L2 status effect in which L2 occupies a privileged position as a source of influence in L3 acquisition. [full abstract]

Patterns of vowel laxing and vowel harmony in Peninsular Spanish
Nicholas Henriksen

This study examines the factors that contribute to vowel laxing and vowel harmony in relation to the presence or absence of word-final /s/ in Spanish. Our data come from two varieties of Iberian Spanish: Eastern Andalusian Spanish (EAS), an /s/-deletion variety, and North-Central Peninsular Spanish (NCPS), an /s/-retention variety. Twenty-four speakers participated in two carrier phrase reading task experiments. In Experiment 1 we show that EAS (but not NCPS) speakers lax the low and mid vowels /a e o/ (i.e., approximating [æ ɛ ɔ], respectively); our data suggest further that the laxing process does not happen in order to compensate for the loss of a morphological /s/. In Experiment 2 we show that EAS (but not NCPS) speakers harmonize the mid vowels /e o/ when followed by laxed /a e o/, suggesting that this right-to-left harmony constitutes a phonological process of vowel laxing rather than a phonetic coarticulation process with word-final lax vowels. Furthermore, we do not find evidence for small V2-to-V1 coarticulation effects in NCPS, contrary to the hypothesis that phonological harmony derives from V-to-V coarticulation and that such effects are manifested as small V2-to-V1 coarticulation effects in non-harmonzing dialects. [full abstract]

Representations of Place and Airstream Mechanism: A real-time MRI study of Tigrinya ejectives
Zainab Hermes, Mao-Jing Fu, Sharon Rose, Ryan Shosted and Brad Sutton

This study examines how aerodynamics is integrated into phonetic and phonological representations of speech sounds, as suggested in the supralaryngeal configuration of the vocal tract to achieve a glottalic egressive airstream during Tigrinya ejectives. Real-time magnetic resonance imaging is used to describe the articulatory configurations associated with the geminate, ejective and pulmonic, velar stops /kk’/–/kk/, and to describe the articulatory correlates of ejectivity above the larynx. In ejective /kk’/, MR images show retraction of the tongue root along with forward expansion of the upper posterior pharyngeal wall. Neither gesture is observed during pulmonic /kk/. Forward expansion (dilation) of the upper posterior pharyngeal wall has been suggested to result from the raising of the larynx and may also be achieved independently from the larynx. Tongue root retraction and upper pharyngeal constriction clearly decrease the volume of the supralaryngeal cavity behind the oral constriction. We argue that these gestures are implicated in achieving the high pressure associated with glottalic egressives and that this results in clear place differences between pulmonic and ejective stops. Results from this study describe a synchronic phonological process, and informs the discussion around the diachronic evolution of original Proto-Semitic emphatics into modern-day Semitic ejectives and pharyngealized consonants. [full abstract]

The Phonetics and Phonology of Fataluku Intonational Downstep
Tyler Heston

Intonational analysis of any language must deal with the crucial question of how many discrete tonal levels are necessary to account for the observed prosodic behavior. Though the autosegmental-metrical theory of intonational phonology would ideally address all intonational systems with a binary distinction between high (H) and low (L) tonal targets, more recent research in the AM framework has suggested that levels such as mid (M), downstepped high (!H) or superhigh (^H) may be required.

In this talk, I present a phonetic description of the intonational high peaks occurring in the underdocumented Papuan language Fataluku (ISO 639-3 ddg, IPA [fatáluku]). Previous work has shown that Fataluku utterances are characterized by an intonational peak on the second mora of each accentual phrase (AP). I here focus on the variability of the absolute f0 of intonational peaks in broad focus utterances, using linear mixed effects models to determine the contribution of continuous (e.g., time) and categorical (e.g., position) factors in the realization of intonational tones. I conclude that only one level of high tone needs to be posited, and differences in absolute f0 values can be explained by a phonological process of downstep. [full abstract]

Post-pubescent long-term exposure to non-rhoticity causes qualitative and quantitative changes in the realization of postvocalic /r/
Marie-Christin Himmel and Baris Kabak

English postvocalic /r/ is subject to variation not only diachronically and synchronically, but also within bilingual grammars. L2 rhoticity is known to influence L1 non-rhoticity, the reverse contact scenario (non-rhotic -> rhotic) remaining unexplored. Furthermore, whether subsequent changes to postvocalic /r/ within an individual’s life span follow universal principles as well as developmental and diachronic paths is largely unknown. In our experiments, we explore developmental trajectories in adult bilinguals from a rhotic L1 upon long-term post-pubescent exposure to a non-rhotic L2. Twelve American-English/German bilinguals and a monolingual control group performed various speech elicitation tasks. Using a phonetic and binary analysis of the presence and absence of rhoticity in [Vr] sequences, we show that, as compared to monolinguals, bilinguals (i) show greater variability in production and (ii) produce rhotic [Vr] sequences with a higher maximum F3.

We take our results to show that non-rhoticity in the L2 influences rhoticity in the L1 in terms of (i) the gradual loss of post-vocalic /r/, as well as (ii) the acoustic quality of [Vr] sequences. We suggest that the adaptation of rhoticity within an individual’s lifespan mirrors diachronic and synchronic patterns commonly observed in English varieties, as well as the general laws of phonetics. [full abstract]

Downstep in Japanese revisited: Lexical category matters
Manami Hirayama and Hyun Kyung Hwang

It has been widely acknowledged that an F0 after an accented word is noticeably lower than after an unaccented word in Japanese (i.e., downstep). In pursuing research concerning the syntax-phonology interface, Selkirk and Tateishi (1991) call attention to the lack of downstep at the left edges of maximal projections of syntactic categories (XPs), and propose that the left edges of XPs are mapped onto the left edges of the major phrase boundaries that block downstep. This study aims to test the effect of different lexical categories on downstep, as the distinction between different categories is frequently ignored in literature regarding the Japanese downstep.

The results from a production test reveal that distinct lexical category does in fact influence the presence or absence of downstep in Japanese. Furthermore, the occurrence of downstep in the NP condition casts doubt on the Selkirk and Tateishi’s generalization. Moreover, this research has an important implication regarding the methodology: lexical category should be carefully controlled for in studying Japanese downstep. [full abstract]

Adaptive dispersion: a perceptual motivation for sound change
Phil Howson and Philip Monahan

Previous research has indicated that speakers prefer contrasts which maximize the perceptual difference between the segments (Adaptive Dispersion, AD). This effect has been mostly observed in vowel contrasts; however, this study examines this effect through native Czech speakers' perception of Czech and Polish sibilant contrasts. Twenty-two participants heard CV sequences of sibilants followed by a low back vowel, /a/, in three blocks: Czech only; Polish only; Czech and Polish sibilants. Statistical analysis was done using a repeated measures ANOVA on the d-prime data. Multidimensional scaling was also performed on the reaction times. Overall, the results indicate that Czech speakers are much better at perceiving the Polish sibilant contrast than the Czech one. The results do support an AD approach to sound change. The old Polish consonant was much more crowded and had a contrast between alveolars, palateoalveolars, alveolopalatals and a post-alveolar trill-fricative. The crowded space caused a great deal of perceptual confusion, resulting in the coalescence of the trill-fricative and the eventual change into modern Polish. The perceptual distance between the three segments has been maximized, increasing perceptual distinctness. However, the Czech system is much more crowded, causing a much poorer overall perceptibility. [full abstract]

English coda [m] adaptations in Standard Mandarin loanwords: Corpora data vs. bilingual and monolingual experimental results
Ho-Hsin Huang and Yen-Hwei Lin

This study investigates how English coda [m] is adapted into Standard Mandarin (SM) loanwords both in the existing corpora and in perceptual similarity adaptation data from monolingual SM and bilingual SM-English speakers. The nasal [m] in coda position is prohibited in SM. Deletion, nasal place change ([m]->[n]/[ŋ]) and vowel epenthesis are the possible repair strategies. The generalizations identified in the corpora indicate that deletion never occurs (c.f. Preservation Principle from Paradis 1996, Paradis & Lacharité 1997). Vowel epenthesis appears in SM when English coda [m] is in word-medial and word final positions. Nasal place change appears when English coda [m] is followed by a labial obstruent. The experimental results show that (i) the bilingual experimental strategies for nonce word adaptations are similar to the patterns observed in the SM loanwords corpora and (ii) monolinguals’ adaptation patterns are more variable due to greater dependence on perceptual cues. The fact that monolinguals and bilinguals differ in the extent to which they employ perceptual cues and phonological features/constraints for loanword adaptations challenges a pure perception-based account of loanword adaptation. [full abstract]

A crosslinguistic Study of Vowel Categorization: Data from Canadian English, Korean and Japanese
Hyun Kyung Hwang

The perception pattern of non-native sounds provides much insight into language-specific aspects of the native phonology of a language. In exploring the influence of language-specificity on perception, it is important to directly compare languages having distinct sound systems. This study is concerned with the English [æ] -[ɛ] and [æ]-[a] distinctions for speakers of Canadian English, Korean and Japanese. By utilizing vowel stimuli with varying F1 and F2, I explore how language-universal perceptual sensitivities are modified, and which acoustic cue the speakers were more attentive to for vowel categorization.

The results from vowel identification tests reveal that both the F1 and F2 cues of the ranges tested are perceptually salient to the speakers of Canadian English. Interestingly, Japanese speakers are only sensitive to vowel height (signaled by F1), whereas Korean speakers show sensitivity only to the backness of a vowel (signaled by F2). These results suggest that language-specificity can modify the sensitivity to different acoustic cues in perceptual processing. The current study considers two possible language-specific factors—the L1 phonology and phonetic variations—in directing native speakers’ attention to different acoustic cues. [full abstract]

Stress clash avoidance by 6- to 7-month-olds
Barbara Höhle, Natalie Boll-Avetisyan and Jürgen Weissenborn

This study investigates whether infants use stress clash avoidance as a cue for speech segmentation such that they separate co-occurring syllables if they are both stressed. Moreover, we explore the link between speech segmentation performance in infancy and later language abilities and specific language impairment (SLI). We tested 119 German-learning 6-7 month old infants, 33 of which had a family risk for SLI. Using the head-turn preference paradigm, infants were familiarized with trisyllables (BATEko, PEGAdi) stressed on the first two syllables and unstressed on the last (strong-strong-weak, henceforth SSw). In a consecutive test phase, infants either heard S#Sw items (BA_TEko or PE_GAdi) with a pause between the two strong syllables or SS#w items (BATE_ko or PEGA_di) with a pause before the weak syllable. Results indicated a novelty preference for the unsegmented SS#w over S#Sw items. This suggests that infants use stress clash avoidance as a speech segmentation cue. A family risk for SLI, however, had no effect. The lack of a link between a risk for SLI and segmentation preferences might suggest that infants with later language difficulties are still influenced by perceptual biases on prosodic processing. An ongoing analysis of data from later language measures will explore this further. [full abstract]

Language-specificity in Speakers’ Strategies of Focus Expression
Martin Ho Kwan Ip and Anne Cutler

The experiment we report here forms part of a larger cross-language project examining processing of prosodic focus in English and Mandarin. In the present production component, we investigated individual variation in speakers’ expressions of focus by drawing on an extensive collection of focus production data from 48 speakers (24 native speakers of Mandarin Chinese; 24 native speakers of Australian English) engaging in two structured dialogues that differed in discourse contexts. The first dialogue involved a conversation between a buyer (participant) and a street vendor (experimenter), while the second dialogue involved an interrogation by a police inspector (experimenter) to his witness (participant). All participants engaged in the two dialogues in their native language, and each dialogue contained pairs involving the same words in a focused vs. an unfocused realization (19 out of the 21 word pairs in each language were either monosyllabic or disyllabic). Analyses revealed that the speakers’ intra- and inter-speaker variability patterned different between the two languages. At the same time, how language-specific factors affects speaker variability also depends on the discourse contexts of the dialogues. Our results provide evidence that speakers’ production of prosodic focus may be more variable, language-specific, and context-dependent than previously assumed. [full abstract]

How local computation leads to global structure: the dynamics of gesture and word
Khalil Iskarous

One of the great steps in science was the development of the idea of a dynamical system by Isaac Newton in the late 17th century. He realized that one can describe highly complex global trajectories of particles through a completely local constraint on how the position of a particle at each instant in time could relate to the position at neighboring instants. This deepest computational insight of dynamical systems analysis has been applied to all kinds of dependent variables other than particle position, and many independent variables other than time, in the physical, biological, and social sciences. This work is on the deep relation between two dynamical models of sound structure: the articulatory phonology approach to gestural dynamics (Browman and Goldstein) and the harmonic phonology dynamical approach to the metrical structure of words (Goldsmith, Larson, Prince). Both models, which treat different levels of sound structure, will be shown to be minimally different instances of one local computational principle leading to global gestural and metrical patterns at different time scales. First principles dynamic arguments will be presented on how this local principle predicts many well-attested phenomena relating to gestures (structural vs. non-strictural contrasts) and metrical structure (demarcativity, culminativity, rhythm, and extrametricality).

Intonation and Sentence Type: The Emergence of Conventions for Attitudinal Meanings
Sunwoo Jeong

This paper argues for the existence of intonational conventions that convey attitudinal meanings such as speaker authority, politeness, etc., across many different contents and contexts for a given sentence type. The evidence for this argument comes from a perception experiment that used a series of spoken English stimuli, chosen to represent a diverse range of interpretational biases (information seeking, request, invitation, etc.) and sentence types (polar interrogative, wh-interrogative, declarative, imperative), as well as acoustically manipulated to represent three terminal contours (rising, falling, level). In the experiment, subjects answered a series of questions pertaining to the likely attitudinal meanings conveyed by the utterances, as well as the most likely speech acts associated with them. The results demonstrate the consistent effects of intonation on subjects’ interpretations of attitudinal meanings, and show that such effects carry across sentences with different interpretational biases conveying different speech acts. Crucially, such consistent effect of intonation on attitudinal meanings seems to hold only within a given sentence type, but differ depending on the sentence type, suggesting that the conventions on attitudinal meanings operate via linguistic representations that tap into both the intonational features of a given utterance and the sentence-type related morphosyntanctic features of the utterance. [full abstract]

Effects of phonetic reduction and social factors on cross-modal lexical priming
Zack Jones and Cynthia G. Clopper

Different phonological variants of the same lexical item prime matching targets in lexical decision tasks, but canonical and familiar forms are more robust primes than less canonical and less familiar forms, suggesting that multiple phonological variants may be encoded in listeners’ lexical representations. The goal of the current study was to explore the role of subphonemic vowel variation due to phonetic reduction and social factors in lexical representation. In a cross-modal lexical decision task, single words extracted from naturally-produced read passages by talkers from the Midland and Northern American English dialect regions served as primes. Matching and unrelated primes were balanced for talker dialect, talker gender, speaking style, lexical frequency, and phonological neighborhood density. Responses were slower overall to Northern primes and the largest priming effects were observed for high-frequency and high-density primes produced by male talkers in plain speech. Northern and male primes were less intelligible and high-frequency and plain speech primes were more reduced, suggesting that processing more difficult auditory primes (i.e., less intelligible, more reduced, and/or those with more lexical competitors) interferes with immediate, subsequent processing of an unrelated visual target. This processing cost reflects a mismatch between a difficult prime and the relevant lexical representation. [full abstract]

The perception of stop/sibilant clusters in Modern Hebrew
Kyle Jones

In binyan hitpa’el, the reflexive and reciprocal verbal conjugation in Modern Hebrew, the /t/ of the /hit-/ prefix categorically metathesizes with a following sibilant (/s/, /z/, /ʃ/, or /t͡s/, giving forms like [histakel] instead of expected forms like *[hitsakel]. It has been theorized that this metathesis may be perceptual, serving to place the /-t-/ in prevocalic position where it can be better perceived by listeners, the direction of metathesis being the more common sibilant + stop sequence in Modern Hebrew (Hume 2004), or that it may be auditory, based on a tendency for the sibilant noise to decouple from the rest of the speech stream, resulting in listener confusion about the place of the sibilant within the word (Blevins & Garrett 2004). Based on data from a speech perception experiment using English speakers, who listened to masked stimuli similar to hitpa’el verbs, I argue that Blevins & Garrett (2004)’s account is correct, with English speaking listeners evincing a tendency to metathesize stop + sibilant sequences into sibilant + stop sequences, despite the higher phonotactic probability of stop + sibilant sequences in English. [full abstract]

Incomplete neutralization and the (a)symmetry of paradigm uniformity
Abby Kaplan

It is well known that neutralization can be incomplete, such that speakers produce a small difference between apparently neutralized segments for at least some cues. Most famously, this has been shown for final devoicing in many languages. This phenomenon has been described as a kind of paradigm uniformity, such that devoiced obstruents are under pressure to resemble voiced obstruents elsewhere in the paradigm. If true, this account raises the possibility that those non-neutralized obstruents might themselves devoice slightly due to their devoiced counterparts elsewhere. I test this hypothesis in two languages with neutralizing patterns: Afrikaans (final devoicing) and Russian (vowel reduction). In both cases, I find that neutralization is indeed incomplete for some cues, but that non-neutralized segments that alternate with neutralized segments are no different from non-neutralized segments that do not alternate. These results suggest that sub-phonemic paradigm uniformity effects are asymmetrical: a non-neutralized form can influence a neutralized form, yielding ordinary incomplete neutralization; but effects in the opposite direction are not attested. [full abstract]

Focus, accentuation and phonetic variability in Greek
Argyro Katsika and Amalia Arvaniti

Three accents used in focal position in Greek declaratives – H*, L+H*, and H*+L – were studied to determine their pragmatics and phonetic realization. These accents were examined in polysyllabic words as a function of phrase length, stress location and tonal crowding. The Lucero et al. (1997) nonlinear time warping technique was used to compute the normalized alignment of the F0 signals per condition, and the resulting averaged signals were compared across conditions. Greek speakers used systematically distinct F0 shapes in response to prompts differing in information structure, confirming that H* is used for new information, L+H* for contrastive focus and H*+L for information that the speaker believes should have been in the common ground. The data show systematic variation that is not limited to tonal target scaling and alignment, but evidence non-localized effects on F0 as well as effects on other phonetic categories (e.g. F0 slope, duration). Nonlinear time-warping also revealed both variability across contexts and non-localized effects in the realization of the accents, with such effects extending to preceding and following syllables. The results support a view of accents as distributions of values (in line with other phonetic categories) rather than as invariable prototypes or sets of discrete “allotones”. [full abstract]

Lenition and segmentation
Jonah Katz and Melinda Fricke

This paper presents experimental evidence bearing on the functional roots of lenition/fortition patterns. We focus on the characteristically-intervocalic lenition patterns of spirantization and voicing, which are attested in dozens of unrelated languages. One proposed explanation for such patterns is that they help listeners detect prosodic boundaries by minimizing auditory disruption internal to prosodic domains and maximizing it at prosodic boundaries. A basic premise of such accounts is that listeners are biased to posit boundaries in the vicinity of more fortis (e.g. unvoiced, plosive) consonants relative to their lenis (voiced, continuant) counterparts. We tested the prediction that lenition/fortition patterns reinforce segmentation by using an artificial-language word segmentation task. We hypothesized that subjects would more quickly and/or robustly learn repeating units (‘words’) in artificial languages with lenition-like phonetic patterns than in languages with non-lenition-like patterns. The first two experiments confirmed this prediction for spirantization but not for voicing. We proposed post-hoc that this may be because consonant duration is a more important factor in 'voicing' lenition than voicing per se. A third experiment, currently being analyzed, attempts to test this hypothesis by simultaneously manipulating voicing and duration. [full abstract]

Targetless /u/ in Tokyo Japanese
Shigeto Kawahara, Jason Shaw and James Whang

Japanese [u] shares some properties with English schwa. It is the shortest and most central vowel in the language, it functions as an epenthetic vowel in some loanword contexts, and it can be devoiced between voiceless consonants. Although there is a large body of literature on the acoustics and perception of (devoiced) [u], few studies have investigated the lingual articulation of this vowel. To fill this gap, this study uses electromagnetic articulography (EMA) to explore the lingual gesture of [u] by native speakers of Tokyo Japanese in devoicing and non-devoicing contexts. We found that the tongue dorsum trajectory for [u] can take the form of a linear interpolation between surrounding vowels. For example, in an [e-u-o] sequence, the tongue dorsum does not rise from [e] to [u], but steadily declines from [e] to [o] instead. On the basis of this data, we argue that /u/ lacks a lingual height target, at least in some environments and for some speakers. We discuss statistical methods for analyzing targetlessness and the implications of the result for the interpretation of “perceptual illusions”, i.e., Japanese listeners’ propensity to perceive /u/ between non-native consonant clusters. [full abstract]

Intonational qualities of strong and weak imperatives
Megan Keough, Elise Kedersha McClay, Molly Babel and Lisa Matthewson

Work in theoretical semantics has suggested that English imperatives be classified as “weak” or “strong”, where weak imperatives correspond roughly to may statements and strong imperatives to should statements. It has been claimed that rising intonation on the imperative cues the interlocutor to a weak reading, and falling intonation cues a strong reading (Portner, to appear). The present work offers a clarification: specifically, the H*L-L% contour signals a weak imperative and L*L-L% a strong imperative. In a set of experiments we test these predictions, seeking to enrich our empirical understanding of how speaker-listeners of English perceive and produce imperatives. We report on a perception experiment where listeners were auditorily presented with a context and asked to pick which auditorily presented imperative (strong or weak) is a better fit for the presented context. Listeners (n=26) associated the predicted intonational contours with the strong imperatives 81% of the time and weak imperatives 67% of the time. Production data naturally produced in response to the auditorily presented contexts from a different set of participants (n=20) is currently being analyzed. [full abstract]

Locality and variability in cross-word alternations: a production planning account
Oriana Kilbourn-Ceron, Michael Wagner and Meghan Clayards

The alternation in English between coronal stops /t,d/ and alveolar flap when intervocalic is nearly categorical when the VTV sequence is within a word, but is variable when a word boundary intervenes and occurs only rarely across a large boundary such as a clause edge. This pattern is common across many processes cross-linguistically – but why are segmental processes at word edges often more variable, and what influences the rate of variability? Previous literature that addressed phonological variability has proposed that phonological rules have to make reference to syntactic structure or that phonological process are tied to certain prosodic domains. In contrast, we propose that phonological variability is only indirectly influenced by syntax and prosody through the locality of production planning (LPP). This hypothesis is motivated based on psycholinguistic models of speech production, and we test its predictions for English flapping in a corpus study and a production experiment. Results show that syntax may have an effect above and beyond prosodic boundary strength, and that the lexical frequency of the following word has a significant influence on rate of flapping, both effects being consistent with and explained by the LPP hypothesis. [full abstract]

ePGG, Pio, airflow and acoustic data on the phasing of glottal opening and three-way phonation contrast: implications for laryngeal features
Hyunsoon Kim, Shinji Maeda, Kiyoshi Honda and Lise Crevier-Buchman

This paper is concerned with the phasing of glottal opening and three-way phonation contrast in the Korean lenis (/p t k/), aspirated (/ph th kh/) and fortis (/p’ t’ k’/) plosives based on simultaneous recordings of articulatory, aerodynamic and acoustic data, in an attempt to refine our understanding of laryngeal features involved in the phonation contrast. For this purpose we have obtained a new non-invasive technique called external lighting and sensing photoglottograph (ePGG) as well as Pio (intra-oral air pressure), airflow and acoustic data, and the following investigations were made: (a) the timing relations among glottal opening onset and peak, airflow onset and peak, and aspiration onset in relation to acoustic events such as a consonant release onset and a vowel onset; (b) how much the peak of glottal opening area and airflow peak height occur; and (c) what acoustic and Pio conditions arise in accordance with the three-way phonation contrast. [full abstract]

Lexical access and stereotypical 'word age' in Korean
Jonny Kim

Recent research demonstrates that listeners' perceptual categorization of speech segments is affected by social information attributed to the talker. The current study examines its effect on lexical-level processing, focusing on Korean words stereotypically associated with either younger or older people. Recognition response times were measured using a lexical decision paradigm. Based on preliminary-survey respondents' evaluations on stereotypical associations and self-reported usage frequencies across age groups, lexical items in the experiment were assigned with two sets of continuous variables as indices for "word age": stereotype scores and usage-age scores. Response times were significantly decreased when a word was spoken in a voice whose age matched the word's stereotype score. Lexical access was also facilitated when speaker age matched the usage-age score, but the effect was marginal. The results provide evidence that lexical representations are not only indexed to the distributional properties of lexical use in episodic memories, but also closely related to explicitly idealized stereotypes for words and social groups. It is argued that linguistic stereotypes enhance the activation of social information, which spreads to the lexical representations via social indices, and thus lexical access is facilitated when the words are heard in a voice of the stereotypically associated age group. [full abstract]

Linguistic contrast enhancement under prosodic strengthening in L1 and L2 speech
Sahyang Kim, Jiyoun Choi and Taehong Cho

This study explores how vowel contrast is phonetically enhanced under prominence- and boundary-related strengthening in both L1 (English) and L2 (by Koreans), by investigating phonetic encoding patterns of vowel contrasts (/i/-/ɪ/, /ԑ/-/æ/) along the temporal and the spectral (F1/F2) dimensions. Prominence conditions were obtained with phonologically-focused (bed-bad), lexically-focused (bed-chair), and unfocused; and boundary conditions were phrase-initial vs. phrase-medial. Prominence-induced strengthening was found to make reference to the phonological system, maximizing phonological contrast by virtue of enhancing phonetic features such as F1/F2 and V-duration in both L1 and L2, though with no clear difference between phonological focus and lexical focus. Boundary-induced strengthening showed less robust enhancement only along the spectral dimension, suggesting that both L1 and L2 speakers encode different aspects of prosodic structure in phonetically different ways. Furthermore, L2 (Korean) speakers’ strengthening patterns indicate that prosodic enhancement of phonological contrast is further modulated by the speakers’ L1 experience: i.e., Koreans exploit the temporal dimension more efficiently in L2 production as they are less sensitive to the spectral differences of English vowels due to their L1 experience (relatively sparse vowel space). The results thus illuminate the nature of the phonetics-prosody interface in L1 and L2 in connection with linguistic contrast and L1 experience. [full abstract]

Individual differences in second language speech perception across tasks and contrasts
Donghyun Kim, Meghan Clayards and Heather Goad

The present study examines whether individual differences in second language (L2) learners’ perceptual cue weighting strategies are related to their discrimination abilities and how cue weights are related across contrasts for individual learners. Twenty-four native Korean learners of English completed a two alternative forced-choice identification task on /ɪ/-/i/ and /ɛ/-/æ/ contrasts varying orthogonally in formant frequency and duration to determine their perceptual cue weights. They also completed a two-talker AX discrimination task on natural productions of the same vowels. In the cue weighting task, we found that individual L2 learners vary greatly in the extent to which they rely on particular phonetic cues. However, individual learners’ perceptual weighting strategies are stable across contrasts. We also found that more native-like performance on this task – reliance on spectral differences over duration – is related to better recognition of naturally produced vowels in the discrimination task. Therefore, although the present study confirms earlier reports that learners vary in the extent to which they rely on particular phonetic cues, our results additionally demonstrate that these individual differences are not random. Instead, they reflect how well individual L2 learners discriminate contrasts and indicate that learners use a stable cue weighting strategy across contrasts. [full abstract]

Prosodic Accommodation in Seoul Korean Accentual Phrases
Jiseung Kim

The goal of this study is to examine prosodic accommodation, specifically testing accommodation at prosodic boundaries for speakers of Seoul Korean. Sixteen native speakers of Seoul Korean participated in a sentence completion task where they were asked to complete a target sentence after reading (in the baseline condition) or listening to (in the test condition) a context sentence. In both cases, participants completed the sentence by speaking. The auditory context sentences had artificially manipulated prosody. The manipulation lowered the f0 of the phrase-final syllables that were associated with the Accentual Phrase-final rise, which is a characteristic intonational property of Seoul Korean. The f0 values – f0 maximum, minimum, mean, and range – were extracted from the AP-final syllables of the participants’ responses, and were compared between the baseline and test conditions. The results of the linear mixed-effects model analysis on 14 speakers showed evidence of convergence. The results suggest that effects of accommodation may be manifested in the pitch levels associated with a prosodic boundary. Further analysis is needed to separate the accommodation effects from the effect of pitch declination over the course of experiment, which was also found to be significant. [full abstract]

Perception of syllable stress varies by listener
Amelia Kimball and Jennifer Cole

The realization of stress in English utterances is theorized to derive from a prosodic hierarchy or metrical grid. If stress patterns are regular and straightforwardly assigned from the metrical structure of an utterance, listeners who hear the same utterance should perceive stress in the same location. However, experiments involving metalinguistic stress marking tasks indicate that listeners disagree on the perceived location of stress. In light of these results, we ask: Is there variability in the perception of stress location, above and beyond variability due to task effects?

We use a metalinguistic stress reporting task and focus our analysis on participants who consistently mark stress correctly in individual words or patterned speech such as poems. Our results indicate a lack of agreement on the location of stressed syllables, even among those participants who have proven accuracy in this metalinguistic task. Based on these results, we argue that the realization of phrasal stress is complex not only in the varying constraints that underlie the acoustic realization, but also in listener-to-listener variation in perception. Overall, we suggest that in order to understand phrasal stress, listener perception must be modeled in addition to modeling metrical structure and the acoustic spell-out of stress. [full abstract]

Obstruent voicing, aspiration, and tone: implications for laryngeal phonology
James Kirby

We study the onset voicing effect (OVE) on vowel F0 in Khmer, Central Thai and Northern Vietnamese, three languages with a three-way laryngeal contrast in initial obstruents. In contrast to previous studies, we find that onset F0 is higher following voiceless aspirated stops in all three languages, although the effect is attenuated in the two tone languages compared to non-tonal Khmer. Moreover, in Khmer, an OVE is observed for both aspirated and unaspirated voiceless stops, suggesting that both types of stop may involve laryngeal adjustments to inhibit voicing in this language. Thus, while gestures that suppress phonation are likely to underlie the OVE in general, the magnitude and temporal extent of the F0 perturbations are a matter of language-specific phonetic implementation, involving differences in the laryngeal gestures themselves rather than simply the timing of glottal and supraglottal articulations. The fact that the degree of F0 raising is most pronounced following aspirated stops in all three languages suggests that a reappraisal of the role played by aerodynamic constraints in conditioning the OVE may also be appropriate. [full abstract]

What do you expect from an unfamiliar talker?
Dave F Kleinschmidt and T. Florian Jaeger

Speech perception is made much harder by variability between talkers. As a result, listeners need to adapt to each different talker's particular acoustic cue distributions. Thinking of this adaptation as a form of statistical inference, we explore the role that listeners' prior expectations play in adapting to an unfamiliar talker. Specifically, we test the hypothesis that listeners will have a harder time adapting to talkers whose cue distributions fall outside the range of normal variation across talkers. We also show that it is possible to infer listeners' shared prior expectations based on patterns of adaptation to different cue distributions. This provides a potentially powerful tool for directly probing listeners' prior expectations about talkers that does not rely on speech produced by many different talkers, which is costly to collect and annotate, and only indirectly related to listeners' subjective expectations. [full abstract]

Attentional modulation and individual differences in explaining the changing role of f0 in the Korean laryngeal stop perception
Eun Jong Kong and Hyunjung Lee

While the enhanced role of the f0 cue in distinguishing a three-way laryngeal contrast in the Korean stops has been examined in many production studies in the context of the sound change, there has not been strong consensus on whether f0 is perceptually primary over VOT in the category distinction. We aimed to define the primacy between f0 and VOT by examining attention demand in utilizing the cues in processing the stops, known to indicate a cue primacy in the speech perception. The 3-alternative forced-choice identification task (/t’ t th/) was conducted over 28 adults speaking Seoul Korean in their twenties. Audio stimuli were created by combining 7-step VOTs and 5-step f0s. The task was blocked by the attentional condition: Distractor block (perception task with arithmetic calculation) vs. No-distractor (perception task only). Findings showed that the presence of the distractor did not affect the Korean listeners’ f0 use in identifying the stops, indicating that f0 did not operate with high level of attention. Rather, it was VOT that was consistently modulated with attentional resource. Unlike the acoustic findings, the current results do not encourage us to argue that f0 is a solely primary cue in processing the Korean stops. [full abstract]

Vocal tract and manual gesture coordination in prosodic structure
Jelena Krivokapic, Mark Tiede and Martha Tyrone

The coordination of speech gestures and manual pointing gestures in prosodic prominence is examined using a novel approach that supports the simultaneous recording of audio, speech articulator and body movement data.

We manipulate stress (placed on the first or on the second syllable) and boundary (word, intermediate phrase, Intonation Phrase). The boundary adjacent target words are the names MIma or miMA (capitalization indicates stress). To elicit gestures, participants are asked to point to a picture of a doll (named either miMA or MIma) while reading the target word. Vocal tract gestures, the audio signal, and manual gestures are recorded concurrently using electromagnetic articulometry (EMA) and a motion capture system (Vicon).

The gestures of interest are the lip aperture consonant and tongue dorsum vowel gestures for the target word, the finger pointing gesture, and f0. The temporal lags between landmarks of the vocal tract gestures and landmarks of the manual gesture are calculated. We examine the duration of these lags (to test for the closeness of the speech gestures and manual gestures) and the standard deviations of the lags (to indicate stability of coordination). [Supported by NIH DC002717 to Douglas Whalen and NIH DC-012350 to Mark Tiede.] [full abstract]

The perceptual representation of place and voice in Russian. Evidence from eye-tracking
Martin Krämer and Natalia Mitrofanova

On this poster we present the results of an eye-tracking experiment that tests how Russian children identify place of articulation (PoA) and voicing in stops in online perception. We compare three hypotheses about lexical representations: Exemplar Theory predicts that listeners identify voiced stops faster than voiceless stops, because the negative VOT of voiced stops is available early on in the signal whereas different PoA (labial vs. coronal vs. dorsal) should be identified with the same speed since the acoustic cue for each PoA becomes available at the same time, i.e., in the onset of the formants of a following vowel. A more abstract phonological full specification approach predicts that all stops are identified with the same speed. An underspecification approach (e.g., FUL – Featurally Underspecified Lexicon theory) predicts that the segments with the unmarked value for each feature are identified faster, i.e., voiceless stops and coronal stops. Our results support the FUL theory of underlying phonological representations regarding the laryngeal contrast, because in our 2AFC picture identification task our subjects identified words starting in voiceless stops faster and more robustly than those starting in voiced stops. The results on PoA are less clear-cut. [full abstract]

Relative cue weighting in perception and production of a sound change in progress
Jianjing Kuang and Aletheia Cui

This study investigates the mapping between production and perception during an ongoing sound change. The tense vs. lax register contrast in Southern Yi vowels is typically distinguished by phonation, and additional cues (F1, F2, and F0) have been introduced by coarticulation between glottal and supraglottal settings. Two perception experiments and one production experiment were conducted to explore the relative importance of each cue in producing and perceiving the tense vs. lax contrast. The same 41 speakers participated in all three experiments. The results show that Southern Yi is undergoing sound change, and the register contrast is shifting from using phonation as the primary cue to using formants. The change is more advanced in perception than production. These results support Ohala’s (1993) hypothesis that change is initiated when listeners reinterpret coarticulated cues as inherent in perception. Similar to Harrington et al. (2012), production and perception become misaligned during the sound change. Nevertheless, the perceptibility of the phonological contrast is maintained because listeners establish perceptual equivalence between coarticulated cues (Beddor 2009). This study presents compelling evidence for the importance of perceptual bias at the initial stage of sound change and illustrates how sound change is perpetuated from perception to production. [full abstract]

Spontaneous imitation in a second language is different from native language imitation
Harim Kwon

This study examines how two acoustic cues for Seoul Korean aspirated stops – high post-stop F0 and long VOT – operate when proficient bilinguals of Seoul Korean and English spontaneously imitate English voiceless stops. Previous studies have shown that English speakers imitate extended VOTs of voiceless stops (e.g., Nielsen, 2011). Furthermore, Seoul Korean speakers imitate aspirated stops with an enhanced non-primary cue (longer VOT) by exaggerating the primary cue for phonological aspiration in the language (post-stop F0), suggesting that the imitative pattern can be language-specific (Kwon, 2015). In this study, Seoul Korean-English bilinguals heard English /t/-initial words that were produced by a native speaker of American English and manipulated to have either extended VOT (Experiment 1) or raised post-/t/ F0 (Experiment 2). Participants’ own English /t/ productions before, during, and after exposure were compared. After hearing /t/ with extended VOT, participants produced /t/ with longer VOT, but not with higher post-stop F0. F0-raised English /t/, on the other hand, did not induce imitative changes in the two acoustic properties measured. These findings suggest that proficient bilingual speakers do not draw on their L1 cue primacy in performing imitation tasks in L2, but adjust the phonetic properties relevant to the L2. [full abstract]

Variability in the French Mid Vowels: Vowel Harmony, Syllable Structure, and the Creation and Effects of Phonological Representations
Jeffrey Lamontagne

Canadian French (CF) has the potential to generate a laxness contrast due to laxing alternations triggered by syllable structure and due to optional high vowel laxing harmony. These phenomena closely resemble two tendencies undergone by the mid vowels: the loi de position (LDP), a trend for mid-low vowels to surface in closed syllables and for mid-high vowels to surface in open syllables, and vowel harmony (VH), a tendency for mid vowels’ heights to be influenced by the phonological identity of the following syllable’s vowel (eg. Nguyen et al. 2004). Despite the surface similarity, the mid vowels are described as being distinguished by height and not by laxing (Nguyen et al. 2004). This corpus study analyses over 25 000 non-final-syllable mid vowels from spontaneous speech using mixed-effects linear regression and finds that, while syllable structure does not significant predictor mid vowel height (p=0.1877), coarticulatory tendencies are stronger with increasing phonological similarity. In young adults, this included laxing (p=0.0012), suggesting laxness may be emerging as a featural specification for high vowels in CF. This work illustrates that phonetic processes could provide suggestive evidence of featural specifications or phonemic status where few or no clear minimal pairs are present. [full abstract]

Vowel dynamics and social meaning in York, Northern England
Daniel Lawrence

This paper explores how dynamic properties of vowel variation may be available to listeners as a social-indexical cue. It presents data from a perceptual experiment testing listeners' sensitivity to the fronting and diphthongization of /o/ in York, Northern England. Listeners heard resynthesized speech tokens representing different degrees of /o/ fronting and diphthongization, as well as different temporal implementations of fronting (targeting the vowel onset or off-glide). Participants matched these samples to a set of characters which varied systematically in terms of their age, social class, and urban/rural identity. Multilevel logistic regression modeling allows a comparison of the relative impact of each acoustic dimension on selections for each social dimension, as well as a comparison of the relative sensitivity of individual listeners to each acoustic dimension. The results suggest that diphthongization is consistently recognized as a cue to socioeconomic status across age groups, while the social interpretation of fronting varies -- younger speakers hear back diphthongs as working-class and rural, while older speakers tend to interpret them as socially unmarked. These findings demonstrate how social meanings may attach to multiple dimensions of phonetic variation, and provide evidence of the role of social indexicality in ongoing processes of change in this community. [full abstract]

Dealing with 'inperfection': Affixes, allomorphy, and dual-route parsing
Laurel Lawyer and David Corina

This study explores prefix allomorphy and morphological parsing using behavioral and electrophysiological measures of lexical access. Subjects indentified words containing either 'in-' or 'un-' prefixes attached to labial or coronal stems. In half of these items, the nasal of the prefix was replaced to form mispronounced forms (eg. UMP: 'umproblematic', UMT: 'umtenable', INP: 'inprecise', IMT: 'imtangible'). Subject responses to UMT forms were more accurate and faster than to other stimuli, and triggered a robust LAN (left anterior negativity) component in the ERP data. In contrast, responses to UMP, INP, and IMT forms were less accurate, slower, and did not trigger an ERP LAN response. These results are explored using a dual-route model of morphological processing whereby some forms achieve successful lexical access via a decompositional route (as in the INP/IMT cases, where 'in-' and 'im-' prefixes provide access to existing allomorphs of the relevant prefix). Others may have full access forms due to previous exposure and are therefore processed continuously (as in the UMP cases). On the other hand, responses to UMT items suggest these items are less likely to achieve lexical access, thus in these cases, demonstrating a failure of parsing using either decompositional or continuous routes. [full abstract]

Subcategorical contrasts in Korean affricates: Implications for English loanword adaptation
Yongeun Lee and Matthew Goldrick

Loanwords containing English affricates are often orthographically adapted in Korean as a sequence of an alveolar affricate plus a palatal glide (e.g, <ʦjʌsɨtin> “Justin”, Kang, 2013). This is an unexpected adaptation since post-affricate /j/ is systematically not pronounced on the surface in Korean (Kang, 2013). We hypothesized that the use of an affricate-glide sequence in spelling English loanwords reflects a subcategorical phonetic distinction in Korean affricates. To test this, the relative constriction locations of the affricates before <j> + /ʌ/ vs. those before /ʌ/ were estimated by two spectral moments of the frication noises. The fricative portion of affricates preceding <j> + /ʌ/ had lower centers of gravity and higher skewness values than their counterparts preceding /ʌ/, suggesting that Korean affricates followed by <j> are produced in a small but statistically more posterior position than other Korean affricates, analogous to the English affricates. This suggests that the adaptation of English affricates into Korean could be modulated by a subtle phonetic contrast present in the native language. Such effects are consistent with recent empirical results and theories of loanword adaptation emphasizing the role of the perception of fine-grained phonetic details in adapting nonnative sounds (Wilson, Davidson, & Martin, 2014). [full abstract]

Prosodic Convergence During and After a Cooperative Maze Task
Yoonjeong Lee, Samantha Gordon Danner, Benjamin Parrell, Sungbok Lee, Louis Goldstein and Dani Byrd

Convergence effects have been found in many aspects of communication between speakers in conversation. In this study, we are interested in convergence in both qualitative and quantitative aspects of prosodic structure, and we measure whether these effects persist in individual speech tasks after conversation concluded. We examined: 1) sentence durations, 2) sentence-final word durations, 3) f0 extrema in target words, and 4) sentence-final boundary tone quality (H% or L%) for two speakers, before, during and after participating in a cooperative speaking task. Speakers (F1, F2) exhibited distinct temporal and intonational patterns before the cooperative maze task. Our results show that many phonetic properties of F2’s speech became more similar to F1’s speech properties over the course of the experiment; in contrast, F1 did not show any significant prosodic modification during or after the cooperative task. We found that convergence effects can persist even after a conversational speech interaction between speakers has ended. Moreover, we show that there were convergence effects in prosodic boundary strength, use of boundary tones, and quantitative properties of those boundary tones. Our study suggests that conversing speakers converge not only in broad measures like speech rate, but also in qualitative and quantitative aspects of prosodic structure. [full abstract]

Ambiguous rhoticity in Glasgow: Short term exposure promotes perceptual adaptation for experienced and inexperienced listeners
Robert Lennon, Rachel Smith and Jane Stuart-Smith

Sociophonetic research indicates a trend towards the loss of postvocalic /r/ in working class Glaswegian speech, leading to ‘derhoticisation’ (Stuart-Smith 2007). Misperception can occur when listeners hear minimal pairs such as e.g. cut/curt, as working class speakers realise postvocalic /r/ as a pharyngealized variant, produced with a similar place of articulation to the preceding vowel. This research tested groups of listeners who had varying familiarity with Glaswegian on their ability to distinguish between these pairs in a two-alternative-forced-choice (2AFC) task. Listeners then heard a short story spoken by a native Glaswegian, before redoing the 2AFC task.

Signal detection analysis indicated that long-term experience of Glaswegian improves listeners' sensitivity to differences between minimal pairs such as cut/curt. Log-transformed reaction times were best fit by a mixed-effects model which included the 3-way interaction of ListenerGroup (i.e. level of long-term experience), Test (Pretest-Posttest performance), and Coda (whether the stimulus word canonically had an /r/: i.e. 'cut' or 'curt'). Overall, the least familiar listeners got faster at identifying words with /r/, following exposure to the story. These results appear to support the hypothesis that short-term learning can promote perceptual changes, possibly indicating the early stages of ongoing learning. [full abstract]

Imitation of non-native clusters: the role of transitional schwa
Tomas Lentz, Marianne Pouplier, Ioana Chitoran and Phil Hoole

Using EMA, we compared consonant cluster overlap for German and Georgian native speakers imitating Georgian and German clusters. A native speaker of German and one of Georgian produced the model productions that were to be imitated. The Georgian model features less overlap than the German model. In clusters only occurring in Georgian (bg, dg, gb and gd), Germans imitate the Georgian model's low overlap, but not the fine-grained overlap differences between the clusters. Georgian speakers do imitate these fine grained relative differences, but overall they produced the clusters with far more overlap than the model. The cluster /bl/ is familiar to both Germans and Georgians. However, the German model produced it with more overlap than the Georgian one. The participants differed in their imitations: Germans imitated the each model's respective timing pattern, while Georgian speakers imitated both models with roughly the same amount of overlap. The results suggest that German speakers are sensitive to the Georgian model's low overlap values (possibly because they perceive a transitional schwa) and focus on reproducing this feature of the model, while Georgian speakers ignore the actual overlap value, but seem to apply Georgian-specific relative timing patterns. [full abstract]

A bad feeling or a bad filling? The influence of social network size on speech perception
Shiri Lev-Ari

Infants and adults learn new phonological varieties better when exposed to multiple rather than a single speaker. Does having a larger social network similarly facilitate phonological performance? Study 1 shows that people with larger social networks are indeed better at speech perception in noise, indicating that the benefit of exposure to multiple speakers extends to real life experience and to adult native speakers. Furthermore, the study shows that this association is not due to differences in amount of input or to cognitive differences between people with different social network sizes. Using computational simulations, Study 2 reveals that the effect of social network size on speech perception is fully mediated by the fact that having a larger social network leads to smoother sampling of the central areas of the phonemes. Furthermore, the simulations reveal that in contrast to previous assumptions, variability itself does not boost performance. The simulations also show that the effect of social network size is independent of amount of input but is modulated by the ratio of intra- to inter-individual variability. Together, these studies show how properties of our social network influence our speech perception. They thus show how aspects of our life-style can influence our linguistic performance. [full abstract]

/l/ in clusters: an articulatory-acoustic study of children’s productions
Susan Lin, Sharon Inkelas, Lara Mcconnaughey and Michael Dohn

This study uses audio and ultrasound data to illuminate young children’s gestural coordination in consonant+/l/ clusters. The hypothesis is that children who are not yet capable of producing a fully adult-like /l/ will require more time (duration) to produce consonant clusters in which the other consonant also involves a lingual articulation. Analysis of audio recordings and associated lingual ultrasound video of five English-learning children producing onset singleton laterals (e.g., lip) and laterals in /kl-/ and /sl-/ clusters (e.g., clip, slap) in an elicited imitation task showed that /l/ duration was significantly longer in /sl-/ and /kl-/ clusters overall, but covaried with articulation. Subjects with adult-like /l/ productions (posterior and anterior constriction) exhibited shorter /l/ duration in clusters than subjects with less adult-like, more advanced tongue position for /l/, a difference we attribute to the time needed for the tongue to move from a posterior position for /k/ and /s/ to the anterior position of the less adult-like /l/ productions. This result suggests that children’s productions of clusters requiring gestural coordination depend on covert articulatory details of the consonants involved; these details, exposed by ultrasound imaging, vary across individuals, contributing to our understanding of why children’s phonological patterns are so variable. [full abstract]

What reaction times reveal about listener groups: L1 Aboriginal English and Standard Australian English responses to a prelateral merger-in-progress
Deborah Loakes, Janet Fletcher, John Hajek and Joshua Clothier

This study investigates perception of an Australian English vowel merger, which occurs for some speaker-listeners. In Australia, /el/->[æl] in Victoria (a state in the south-east), where an isogloss is located. Aboriginal English and Standard Australian English listeners from Warrnambool (a sound change community) took part in a two-alternative forced-choice identification task to determine how well they identified Australian English vowels in varying conditions, and also how long they took to process their responses. Standard Australian English listeners from Albury-Wodonga (no sound change) also took part. We report on the accuracy and timing of responses for "control condition" het-hat stimuli, and "merger condition" hell-Hal stimuli. We show that all listener groups accurately respond to control condition stimuli, but the listeners from Warrnambool, where sound change occurs, find /æl/ particularly troublesome. Aboriginal English listeners have a preference for /el/ at both ends of the continuum, while the Standard Australian English listeners answer at random for /æl/. Where reaction times are concerned, Albury-Wodonga listeners are fastest across the board, and all groups have comparatively faster reaction times to /æt/ stimuli. On average, the listener groups from Warrnambool have around double the processing time for /æl/ stimuli compared to /æt/, and they very often responded wrongly – highlighting the confusion in this prelateral context. Aboriginal English listeners have longer reaction times than the Standard Australian English listeners for all stimuli. [full abstract]

Individual Differences in Perceptual Compensation and Lexical Effects and Implications for Sound Change
Yanyu Long

Following Yu’s pioneering research that related Perceptual Compensation (PC), Autism Spectrum Quotient (AQ) and social profiles of innovators in sound change, this study investigated how individuals with different AQ differ in their PC and lexical effects (LE). The motivation is that, unlike what assumed by Yu, minimal compensation alone does not necessarily result in misperception because lexical knowledge can help to restore the intended sound. Thus, a potential innovator should be poor at making use of both phonetic and lexical contexts to normalize speech variation.

30 participants with low/high AQ were selected to form a low AQ group and a high AQ group, and their PC and LE were tested. The major results are that: 1) the low AQ group shows significantly lower PC and non-significantly lower LE; 2) PC and LE positively correlate with each other, suggesting people who compensate less are less influenced by lexical knowledge. These people also tend to have low AQ. Their perceptual behaviors suggest that they may introduce new variants through constant misperception, and their low AQs imply social profiles that may facilitate propagation of new variants. This supports the view that low AQ individuals are the best candidates as innovators of sound change. [full abstract]

Non-canonical Word Order in Russian: Processing and Acoustic Parameterization
Tatiana Luchkina

In this study, we test if prosodic augmentation of ex-situ constituents in Russian, a free word order language, is conducive to sentence processing and leads to faster recall and recognition of the ex-situ word. Results of production and probe recognition tasks reveal that augmented prosodic expression and a change in word order may co-occur in spoken language use. Cross-application of these cues facilitates subsequent recognition of the ex-situ word only when it is aligned with a natural prominence landing site, such as the utterance-final position in Russian. [full abstract]

Ambivalent Consonantal Effects on F0
Qian Luo, Karthik Durvasula and Yen-Hwei Lin

Previous research shows that aspirated consonants and sonorants can have inconsistent consonantal effects on F0 across languages (Xu and Xu 2003, Zee 1980, Maddieson 1984). This study is interested in such inconsistent consonantal influences in tonal languages and investigates the effects of aspiration and sonorancy on F0 in tonal languages: (1) what factors condition these consonantal effects on F0? (2) what insights can the results provide for phonological feature theories and historical phonology?

The target languages, namely Mandarin and Cantonese, are selected because there are reports of inconsistent consonantal effects on F0, and their different historical devoicing and tone merger processes. A series of production experiments were conducted: Standard Mandarin speakers and Standard Cantonese speakers were asked to read CV stimuli. The stimuli have onsets grouped by different consonant types (i.e. aspirated/unaspirated obstruents and sononants), and also by different Cantonese or Mandarin lexical tone types. Our findings suggest that (a) consonantal effects on F0 are language-specific and conditioned by lexical tones; (b) consonantal effects do not seem to be a good argument for feature specification; (c) the historical tone merger may have resulted from the interaction between consonant type and tone type. [full abstract]

The limits of inductive learning: The case of Modern Irish mutation
Ruth Maddeaux and Yoonjung Kang

We report the findings from an experiment that examines the roles of 1) inductive learning from statistical patterns (Ernestus & Baayen 2003, Zuraw 2010, Daland, et al. 2010, Hayes et al. 2011), and 2) speakers’ awareness of universal preferences in phonological patterns (Wilson 2006, Moreton 2008, Berent, et al. 2008, Becker, et al. 2011, 2012). A nonce word experiment examines Irish consonant mutation as a test case. In Irish, word-initial [d] and [ɡ] in non-lenition contexts alternate with [ɣ] in lenition contexts, while [dj] and [ɡj] alternate with [j]. The experiment probes Irish speakers’ preference between the two alternation options (coronal stop ~ [ɣ]/[j] vs. dorsal stop ~ [ɣ]/[j]). We examine to what extent the preferences reflect the statistical distribution of existing alternations and what role, if any, phonological universals play in shaping speaker’s choices. The results confirm the tenet of inductive learning that speakers possess implicit knowledge of the statistical patterns of their language, but we also find evidence for a substantial role of universal phonological preference. We conclude that Irish speakers have implicit knowledge of the statistical patterns in the distribution of these sounds but the generalizations they draw are filtered through the sieve of universal phonological preferences. [full abstract]

Sonority profile and temporal organization of clusters: evidence from Russian
Stefania Marin, Marianne Pouplier and Alexei Kochetov

It is well-known that consonant clusters obeying the sonority sequencing principle are universally preferred over clusters violating it; what is less clear is the status of sonority violating clusters in languages that have them. These clusters could betray their typological “markedness” by differing from sonority-obeying clusters in the same language. However, it may actually be the case that once they are part of a language’s grammar, sonority-violating clusters are indistinguishable from sonority-obeying ones. In the current research, we address this issue by comparing the temporal organization of onset clusters with different sonority profiles in Russian – sonority-raising /bl-/, /gl-/, sonority falling /lb-/, /lg-/, and sonority-flat /kt-/, /tk-/. Our results show that, at least on the measures used, the sonority profile of a cluster does not affect its temporal organization. Russian sonority-violating clusters exhibit the same timing patterns as sonority-obeying clusters and they all exhibit the timing predicted for complex onsets, and generally observed for onsets in other languages (a c-center organization). This indicates that grammar has a stabilizing function: once sonority-violating onset clusters are part of a language’s grammar they behave indistinguishable from sonority-obeying ones. If anything, language internal cluster frequency may be a factor influencing cluster timing. [full abstract]

Coalescing sources of bias in perception: Lexical and prelexical influences on the processing of phonological features
Alexander Martin and Sharon Peperkamp

Phonological features are known to be processed differentially during word recognition. Focusing on French listeners, who pay less attention to voicing than to place and manner, we explore two possible sources of these differences. (1) We consider universal bottom-up acoustic-perceptual influences, using an ABX discrimination task. We find that differences in manner are more reliably discriminated than differences in place or voicing. We attribute this to the stark acoustic difference between stops and fricatives. (2) We consider the influence of language-specific lexical knowledge by proposing a novel method for measuring functional load (FL), the amount of work a contrast does in a phonological system. We find that in French, place has a higher FL than manner or voicing. Taken together, these two results can explain why voicing is less important than the other two features for word recognition in French. We propose that listeners are biased not only by bottom-up acoustics (therefore lending attention to manner differences), but also by their knowledge of their native lexicon (for French, lending attention to place differences), making word recognition more efficient. [full abstract]

Seeing focalization: The role of visual information from lip movements in the natural referent vowel bias
Matthew Masapollo, Linda Polka and Lucie Ménard

Considerable research in cross-language speech perception indicates that perceivers (both adult and infant) are universally biased to attend to vowels with extreme articulatory/acoustic properties (peripheral in F1/F2 vowel space). This bias is evident in phonetic discrimination tasks which show robust directional asymmetries, such that perceivers consistently perform better at detecting a change from a relatively less to a relatively more peripheral vowel, compared to the same change presented in the reverse direction. Yet, the nature of this perceptual phenomenon (i.e., the natural referent vowel [NRV] bias) is not fully understood. The present research investigates whether this bias is attributable to general auditory processes or to phonetic processes that track articulatory information available across sensory modalities. Using a same-different (AX) task, we examined whether there are directional asymmetries in English-speaking adults’ discrimination of less versus more extreme (i.e., hyper-articulated) /u/ vocalic articulations when only acoustic- or only visual-phonetic information is available. The results revealed asymmetric perceptual responses favoring the more peripheral /u/ tokens, regardless of the modality of stimulus presentation. These findings suggest that the NRV bias derives from an ingrained sensitivity for detecting extreme vocalic gestures, which may be specified in the optic, as well as in the acoustic, signal. [full abstract]

Phonetic reduction, perceptual illusions, and phonotactic legality
Alex McAllister and Matthew T. Carlson

This study probed the relationship between automatic phonotactic repair and speech production, by asking whether the repair structure (a prothetic vowel) may be susceptible to reduction in speech. Spanish productively repairs word-initial /s/-consonant clusters (henceforth #sC) with a prothetic /e/ in both production and perception. We asked whether the initial vowel in Spanish #esC words like espalda ‘back’, which matches the default repair vowel, is more prone to reduction than other initial vowels, such as in aspirina ‘aspirin’. We explore this question in the speech production of 15 speakers of Andalusian Spanish who produced half #esC and half #asC words in isolation (578 tokens). Outright vowel deletion was uncommon, but was more likely with initial /e/ (5%) than initial /a/ (0.3%, one token). Moreover, when the /s/ was realized with greater duration (cf. the common tendency to lenite syllable-final /s/ in Andalusian), shortening of /e/, but not /a/, was observed. These findings provide evidence that reduction may be enabled when the reduced material can be perceptually repaired, leading to the occurrence of apparently illicit sequences in actual speech, e.g. espalda produced as [spalda]. The influence of articulatory, frequency, and other factors on reduction is also evaluated. [full abstract]

Production Influences on Phonological Representation in an Emergentist Grammar
Tara Mcallister Byun and Anne-Michelle Tessier

Historically, there has been tension between performance-based accounts that attribute child speech patterns to physical constraints on production and competence-based accounts that invoke rule or constraint formalism. Research in the laboratory phonology tradition suggests that it is neither necessary nor desirable to segregate speech-motor development from grammatical development when modeling speech acquisition. This paper focuses on new insights that can be made available by bringing together two lines of inquiry that have evolved largely independently. The first is the literature assembling behavioral and neural evidence that speech strings that have been articulated are represented more robustly in memory than strings that have been heard but not produced (e.g., Keren-Portnoy et al., 2010). The second is the theoretical and experimental literature positing that abstract elements of phonology—segments, features, and constraints—can be understood to emerge from generalizations over stored memory traces at a more holistic level former (e.g., Pierrehumbert, 2003; Werker & Curtin, 2005). We argue that when the production advantage for learning is modeled within an emergentist model of phonology, the production biases that are known to shape the earliest stages of lexical development are expected to propagate up through higher levels of abstraction, explaining their fundamentally intertwined nature. [full abstract]

A system for unified corpus analysis, applied to polysyllabic shortening across 12 languages
Michael Mcauliffe, Morgan Sonderegger and Michael Wagner

A huge and growing pool of speech data annotated at least with an orthographic transcription exists in the world (e.g. speech corpora, lab speech). This pool has significant scientific potential, by “scaling up” phonetic and phonological investigations to use many datasets. However, doing so requires software for unified corpus analysis: integrating speech datasets and querying across them. We present an open-source software package for unified corpus analysis, Montreal Corpus Tools (MCT): speech datasets are imported and stored in a common database format (implemented in Neo4J/SQL) based on the annotation graph formalism (Bird & Liberman, 2001); the database is queried to find linguistic objects (words, phones) meeting some criteria (e.g. utterance-final stops); information about the objects (positional, durational, and acoustic) is then exported to CSV. We exemplify MCT in a case study examining polysyllabic shortening (and the closely related Menzerath’s Law) in 12 typologically diverse languages, using the GlobalPhone and TIMIT corpora (~1,830 speakers, 170 hours). We find that in utterance-final position, mean syllable duration and first vowel duration decrease in a similar way as word length increases across all languages, suggesting that the pattern is part of a more general effect, as predicted by Menzerath’s Law. [full abstract]

Cross-linguistic gender priming in speech processing
Grant Mcguire, Molly Babel and Alexandra Bosurgi

Socio-indexical properties affect linguistic processing as learned associations between speech patterns and social meaning. Strand (2000) showed that stereotypical female faces facilitated the processing of female voices, but stereotypical male voices did not show the equivalent effect in English. This indicates that learned social stereotypes are more active when processing women’s voices. It’s possible that these stereotypes are not a social bias but more of a necessary strategy for processing the less dense harmonic structure of female voices. Listeners may rely on female stereotypes to manage this increased challenge. As an initial inquiry, we tested gender identification in Mandarin with Mandarin and English listeners. On each trial listeners were presented with a face followed by a voice and identified the gender of the voice. In a subliminal prime condition attractive female faces facilitated processing of female voices more than the unattractive female faces for both Mandarin and English listeners. In an overt priming condition, attractive faces for both genders facilitated processing of same-gender voices. This suggests that female voice stereotypes are used early in voice identification for Mandarin and English listeners. With more overt primes, both attractive male and female faces show evidence of face-voice associations. [full abstract]

Gesture and velocity in Brazilian Portuguese devoiced vowels: a preliminary EMA study
Francisco Meneses, Denise Pozzani, Nicole Wong, Zainab Hermes, Torrey Loucks and Ryan Shosted

Vowel devoicing in certain phonological contexts is a synchronic phenomenon in Brazilian Portuguese. Meneses and Albano (2015) argue that the vocal tract configuration during devoiced vowels is narrower than it is during voiced vowels, as there is no time to reach the opening necessary for a significant drop in supraglottal pressure. EMA data clarify the articulation of such devoiced vowels. Lingual position data were collected using the NDI Wave system. Three sensors were glued on the anterior portion of the tongue. Seven speakers produced utterances in a carrier phrase including an initial voiced /i/ and a utterance-final /i/, which was expected to devoice. One speaker (S9) consistently devoiced the final vowel of this phrase. The average position and velocity of the sensors was measured using synchronized audio signals as a guide. For S9, the position of the two posterior tongue body sensors was higher than it was for other speakers, who devoiced less frequently. The velocity of these sensors was greater during the devoiced vowel. Increased velocity and tongue body elevation are in agreement with data analyzed by Meneses and Albano (2015). Further investigation will provide more controlled results on the coordination between the preceding consonant and the devoiced vowel. [full abstract]

Production dynamics and phonetic motivations for English raised /æ/ and intrusive [l]
Jeff Mielke, Christopher Carignan and Erik Thomas

We describe a method for examining the temporal dynamics of speech production directly from ultrasound video without tracing. We employ this technique to investigate the phonetic motivations of /æ/ raising and intrusive [l] in North American English. Principal component analysis is applied to filtered ultrasound images, as described by Hueber et al. (2007), to identify independent axes of variation within a set of speech data. We use formant measurements and linear regression to transform the PC score vectors into meaningful articulatory parameters. To generate an articulatory signal representing the lingual contribution to /æ/ raising, the PC score vectors are transformed to correlate with the front diagonal of the acoustic vowel space (Z2-Z1) at the same time points. We are able to see clearly that /æ/ raising before /m n/ involves tongue body raising that is timed to the vowel nucleus. We find that F1 lowering in pre-nasal /æ/ is accounted for by tongue raising in all of our speakers who raise /æ/ pre-nasally. Raising before /ɡ ŋ/ involves anticipating the velar closure to different degrees, and there is incomplete overlap between the speakers exhibiting this phonological pattern and the speakers who appear to exhibit the strongest phonetic motivation for it. [full abstract]

The Effect of Phonological Context on the Perception of Strong Place Assimilation in Nasal and Stop Consonants
Mercedeh Mohaghegh and Craig Chambers

Two forced-choice identification experiments and two eyetracking experiments examined the recognition of words whose final consonants have undergone place assimilation (‘phone button: /n/→[m],‘cat button’: /t/→[p]). Identification scores and fixation patterns were measured for assimilated or unassimilated words that were presented either in isolation or in sentences containing the triggering phonological context for assimilation (i.e., next word began with a labial consonant). With isolated words, listeners tended to (mis)identify assimilated coronals as labials. However, when the context was present, perception of assimilated consonants shifted towards a coronal interpretation, even though complete compensation was not achieved, especially for words ending in nasal consonants. Interestingly, measures of online comprehension (eye gaze) showed a priming effect for assimilated nasals (but not stops) whereby a recent act of compensating for assimilation facilitated on-line perceptual processing of a subsequent word-final assimilated nasal. These findings reflect the idea that phonological context plays an important, if sometimes weak, role in perception of assimilated sounds during spoken word recognition, even for lexically ambiguous forms in the absence of a disambiguating semantic/syntactic context. Further, we argue that the precise nature of compensation may vary according to the characteristics of the particular sound class that is affected by the assimilation process. [full abstract]

Vowel Lengthening Effects in Natural Speech: Learning under sparse data and high variance
Becca Morley

Vowel lengthening before voiced obstruents has been a much studied phenomenon for several decades. However, there remains no general consensus on the phonetic basis of the effect, and considerable variability in reported results. Perceptual experiments suggest that vowel length differences alone are enough to cue the voicing contrast in word-final stops in American English. Production studies, however, raise the question of how such a robust relationship can be induced from the learner's widely varying input. This paper uses the Buckeye Corpus of Conversational Speech to explore the distribution of this phonetic cue in detail. Despite the large number of tokens available, from a statistical perspective the corpus exhibits a pervasive data sparsity problem: virtually no items were balanced for either speaker, vowel, place of articulation, or word frequency across the voiced/voiceless contrast. Furthermore, the length difference between pre-voiced versus pre-voiceless vowels was smaller than the durational differences due to speaking rate and word frequency; and was completely masked by a floor effect in intrinsically shorter words. These results bear on questions of whether, or how much, listeners must normalize their input, and how synchronic (phonetic) variation might lead to (phonological) sound change. [full abstract]

Interactions of speaking rate and prosodic organization in non-native speech production
Tuuli Morrill and Melissa Baese-Berk

This study investigates both speaking rate and intonation in non-native speech. In addition to speaking more slowly than native speakers, non-native speakers of English exhibit greater variability in rate across utterances when reading (Baese-Berk & Morrill, 2015). We suggest that an interaction between rate and prosodic organization is a potential source for the increased variability. The current study examines read speech from native Korean (n = 9), Mandarin (n = 8), and English speakers (n = 10); all were reading The Little Prince in English (data from the ALLSSTAR corpus). We used Smoothing Spline ANOVA to model pitch contours and examine differences between language groups. Findings revealed that: (1) In certain utterances, non-native speakers followed native intonation contours closely, including pitch accents, while in other utterances, non-native speakers diverged from native intonation contours, only exhibiting overlap at utterance boundaries. (2) As expected, non-native speakers were slower in rate than native speakers. (3) The utterances where non-native speakers were slower than native speakers were those in which they most closely matched the native intonation contours; when non-native speakers were faster, they failed to match native intonation contours. These findings suggest a relationship between speaking rate and prosodic organization in non-native speech. [full abstract]

How does deep brain stimulation affect regulation in speech motor control?
Doris Muecke, Anne Hermes, Timo B. Roettger, Johannes Becker and Michael Barbe

Chronic deep brain stimulation of the nucleus ventralis intermedius is an effective treatment for patients with medication resistant Essential Tremor. However, these patients report that stimulation has a deleterious effect on their speech. The present study investigates gestural coordination patterns in the speech motor system by using Electromagnetic Articulography. We analyse onset coordination patterns in syllables with low and high complexity, such as /li/ and /kli/, within the coupling hypothesis of syllable structure (Broman & Goldstein 2000). The results reveal stimulation-induced effects on the regulation in speech motor control in ET patients. Under stimulation, we found a detrimental decrease in peak velocities, stiffness and acceleration phase of oral constriction gestures leading to imprecise articulation in the production of stop consonants: movements are slower, longer and less stiff. Furthermore, we found timing deficits in the phonetic realization of competing coupling relations for complex onsets in the patients’ data. For syllables with high complexity, /kli/, a delay would have been expected between the activation of the initial consonants. However, in Essential Tremor patients both consonants are activated at the same time. The competitive coupling coordination, a phonological pattern that has to be learnt, is reduced to simple in-phase relations and breaks down. [full abstract]

How native and non-native listeners process schwa reduction in French: A combined eye-tracking and ERP study
Kimberley Mulder, Sophie Brand and Mirjam Ernestus

In contrast to native listeners, non-native listeners of a language often have problems understanding reduced forms (e.g., English /jɛʃeɪ/ for /jɛstədeɪ/ yesterday). We investigated, in a combined EEG and eye-tracking experiment with native and Dutch non-native listeners of French, to which extent and why non-natives suffer more from reduction than natives do. We focused on schwa reduction in the first syllable of French nouns (e.g., /rkɛ̃/ for /rəkɛ̃/ requin ‘shark’) that were presented in the middle of sentences.

Participants listened carefully to the spoken sentences and saw a display of four line drawings. Each display consisted of the target word (e.g., requin), a phonological competitor (e.g., rideau /rido/ 'curtain'), and two neutral distractors.

The EEG data show a more negative N400 for reduced than for full forms in the non-natives only. The eye tracking data reveal that the non-natives seriously considered other lexical candidates (mainly phonological competitors) during the lexical search. In addition, they considered other lexical candidates for a longer stretch of time than the natives did. Taken together, the combined eye-tracking and EEG data suggest that non-natives suffer more from reduction than natives do because they have more problems in accessing the meaning representations of reduced words. [full abstract]

Phonotactically-mediated Compensation for Coarticulation
Kevin Mullin

This experiment finds positive evidence for a speech perception model that is parallel-interactive with top-down linguistic knowledge influencing lower level phonetic representations. Thus, a serial-autonomous model of perception is disconfirmed. A categorization experiment finds a first-order effect of the *spup constraint in English influencing a [p-t] continuum (‘t’ bias in a [spu_] context, no bias in a [pu_] context), a first-order effect of compensation for coarticulation on a [i-u] continuum (‘i' bias in the [p] context, ‘u’ bias in [t] context), and a second-order effect of phonotactics on the [i-u] continnum (‘u’ bias in the [spu_] context with an ambiguous stop). The second-order phonotactically-mediated compensation is only predicted by the parallel-interactive model. The experiment was presented online auditorily with subjects making categorization responses for both [p-t] and [i-u] in four possible choices. The two speech perception models are formalized as parallel and stratal multi-level representational Maximum Entropy perceptual grammars. [full abstract]

The Perceptual Effects of Phonotactic Rareness and Partial Allophony in Canadian French
Patrick Murphy, Philip Monahan and Margaret Grant

In Canadian French, coronal stops turn to affricates (/t, d/ to [ts, dz]) before high front vowels and glides (/i, y/ and /j, ɥ/), e.g. petit "small" [pətsi]. Non-allophonic affricates also exist and they lack the same environmental restrictions (e.g. tsé "y'know", mouche tsé-tsé "tse-tse fly", tsar [dzaʁ] "tsar"), but they are rare. Our paper is an experimental investigation of how these affrication patterns affect French speakers' perceptual tendencies. Results from Canadian French speakers (n=13) in Experiment 1 suggest that the rareness of non-high-front vowels after affricates biases them towards hearing high front vowels after affricates. In Experiment 2, their perception of an affricate-stop continuum ([ts] ... [t] and [dz] ... [d]) was more categorical before the non-high-front /e/ (where stops and affricates are contrastive) than before the high front /i/. Preliminary results on an exploratory sample of European French speakers living in Quebec (n=6) suggest a similar effect in Experiment 1 but not Experiment 2. The European French speakers lack affrication in their own dialect but they encounter it in the speech of others from living in Quebec, and so these findings could indicate an effect of dialect exposure. [full abstract]

Perception, Production, and the Implementation of Phonological Opacity in the Bengali Vowel Chain Shift
Traci Nagle

Opaque phonological patterns, such as the synchronic vowel-height chain shift attested in the Bengali verb system, pose difficulties for surface-based theoretical analyses, and their very existence has been challenged by proponents of natural generative phonology. The research reported here experimentally investigates the phonetic and phonological status of the chain shift pattern in Bengali verbs, in which the height of the vowel in a verb stem raises under the influence of a subsequent high affix vowel. The experimental results, in which Bengali speakers seem ambivalent between generalizing the chain shift pattern to nonce verbs and leaving the verbs’ stem vowel unchanged, suggest that this pattern is evidence of neither a fully productive phonological process nor a purely phonetic phenomenon of vowel-to-vowel coarticulation. These results echo those of experimental tests in other languages in which chain shifts resist generalization to nonce words. The present results also offer support for some scholars’ skepticism about the nature and learnability of opaque interactions and are consistent with propositions that supposedly opaque phonological patterns may instead be evidence of allomorphy. [full abstract]

Corrective focus in conversational French
Ricardo Napoleão de Souza and Caroline Smith

This study investigates the prosodic marking of focus in a large corpus of spoken Parisian French, the Nijmegen Corpus of Casual French (Torreira, Adda-Decker, and Ernestus 2010). A previous study of left dislocations in the same corpus (Smith and Napoleão de Souza 2015) had shown minimal prosodic marking, compared to dislocations elicited in a laboratory setting. For the present investigation we searched the corpus to identify utterances that are examples of corrective focus, in order to examine whether they are clearly marked prosodically, or whether the lack of marking observed in the dislocations may be more generally true of spontaneous speech. Measurements were made of f0 and durations for the entire utterance, the focus phrase, and where applicable, post-focus material. Our results show that there is an almost complete absence of the cues to focus that have been proposed in the literature, with the possible exception of initial f0 rises in the subset of corrections that contain negation of a previous statement. These results suggest that prosodic marking may be attenuated in conversational French, and that speakers and listeners may rely on non-prosodic strategies to establish what information is being focused in spontaneous speech. [full abstract]

An acoustic study of Punjabi tone and stress (Doabi dialect)
Kiranpreet Nara

An acoustic analysis of Punjabi stress and tone was conducted to determine the acoustic cues of stress and the pitch contours of the three tones (rising, falling, and default). Duration was determined to be the acoustic cue for stress, which is similar to Hindi (Nair et al., 2001). In words with rising or falling tone, f0 is also a cue for stress. There is only primary stress in Punjabi and it is attracted to long vowels followed by syllables with geminates or nasal stop codas. In the absence of long vowels, geminate codas, or nasal stop codas, stress appears on the penultimate syllable. As for tone, falling tone is realized entirely on the stressed syllable while peak delay is observed for rising tone except when the rising tone appears on the word final syllable. This suggests that peak delay is constrained by word boundary. Peak delay has also been observed for Mandarin (Xu, 2001). Tone association of rising or falling tone happens when there is a LH or HL tonal contour which associates with the bimoraic stressed syllable. In the case of no underlying tonal contour, a default low associates with the word and results in a default tonal contour. [full abstract]

How vowel variability relates to vowel perception
Nhung Nguyen, Jason A. Shaw, Catherine T. Best and Michael D. Tyler

Studies on the relationship between vowel variability and vowel perception usually estimate vowel variability based on how spread-out the tokens of a particular vowel category are in F1-F2 vowel space. This method fails to account for the fact that, all else being equal, the degree of variability of a vowel formant is systematically related to the magnitude of the mean formant value. In our study, we investigated how vowel variability in a large corpus of Australian English relates to vowel perception in a vowel categorization task by Australian listeners, and in particular which index of variability better predicts accuracy. We compared the standard index of variability, standard deviations of formants, with a new alternative based on the residuals of regression lines fit to the means and standard deviations of F1 and F2 values. Results show that (1) vowel variability correlates with vowel categorization accuracy; (2) the nature of the relationship between vowel variability and categorization accuracy depends on the location of the vowel, with non-low non-back vowels behaving differently than others; and, (3) that the variability index based on residuals is a better predictor of accuracy than the standard variability index. [full abstract]

ERPs reveal that exemplar effects are driven by episodic memory instead of the mental lexicon
Annika Nijveld, Kimberley Mulder, Louis ten Bosch and Mirjam Ernestus

This study investigates the conditions under which exemplar effects arise. We conducted two experiments (with 33 participants in each), in which a prime and target represented the same word type and were spoken by the same or a different speaker. In Experiment 1, participants performed an old/new judgment task, which requires their use of episodic memory. In Experiment 2, participants performed a semantic classification task (animacy judgment), for which they can rely on their lexical memories rather than their episodic memories. We recorded participants' response times, response accuracy and electro-encephalography (EEG). We observed a difference between words repeated by the same and a different speaker (exemplar effects) in Experiment 1, but not Experiment 2. Importantly, this shows that participants' use of episodic memory enhances exemplar effects, and suggests therefore that clouds of exemplars are represented in episodic memory rather than in the mental lexicon. Interestingly, these exemplar effects only clearly surfaced in the EEG data. This shows that event-related potentials may capture exemplar effects which remain invisible in behavioral measures. This indicates that we should be careful in disregarding (previous and future) null findings in exemplar experiments that were obtained with behavioral measures. [full abstract]

The effect of indexical cues on the distributional learning of sound categories
Masaki Noguchi and Carla Hudson Kam

The distributional learning theory of phonetic category acquisition assumes that reliable distributional cues for the categorization of speech sounds exist in the input. In the real world, however, not all talkers produce exactly the same distribution of speech sounds, and this variability may undermine the reliability of distributional cues. Studies have demonstrated that listeners are sensitive to talker-specific information in the acoustic signal and that indexical information (i.e., the identity of talkers) is an integral part of linguistic knowledge, suggesting a potential solution to the reliability problem. In this study, we tested whether adults can learn two novel phonetic categories from input in which talker-dependent variation introduced potential ambiguities into the categorization. Specifically, the input was produced by multiple talkers. Each talker provided reliable distributional cues for the learning of two categories, but the locations of the distributional peaks varied from talker to talker such that the overall aggregate distribution of speech sounds did not have a bimodal shape. The results showed that adults could learn two novel phonetic categories from this input, suggesting that they could overcome the variation in the input by using indexical information to sort out statistical information. [full abstract]

On the relation between speech perception and loanword adaptation: new evidence from Korean
Mira Oh, Robert Daland and Lisa Davidson

The illusory vowel effect refers to the failure to perceive the presence/absence of a vowel in a context where it is phonotactically required (e.g. Japanese ebuzo/*ebzo; Dupoux et al. 1999). In loanword adaptation, vowel epenthesis is the preferred repair for consonant clusters that are phonotactically illicit; it is therefore tempting to posit that vowel epenthesis in loanword adaptation arises from the illusory vowel effect. We address this hypothesis with a cross-linguistic speech perception experiment with Korean and American listeners. It is shown that the illusory vowel effect is sufficient to account for some but not all instances of vowel adaptation in loanword adaptation.

The results of both a discrimination and an identification experiment show that Koreans exhibit the illusory vowel effect when the initial consonant of the cluster contains an audible burst release or other frication noise. However, in the absence of such a burst release, Koreans discriminate (illegal) stop-nasal clusters from an epenthetic repair, and judge the illicit cluster as more perceptually similar to the native/assimilatory repair rather than the epenthetic/loanword (e.g. [pʰak̚na] ~> [pʰaŋna], *[pʰakʰɨna]). Loanword adaptation must be both perceptual and phonological. [full abstract]

Investigating the perceptual hypocorrection hypothesis with sibilant harmony
Avery Ozburn

While Ohala (1993) suggests that assimilation and dissimilation result from perceptual hypocorrection and hypercorrection respectively, evidence suggests that listeners hypocorrect in liquid-vowel-liquid sequences, which typologically tend to dissimilate (Abrego-Collier 2013). This study examines such perceptual effects in sibilant harmony. A [s]-[ʃ] categorization task was conducted using CVCV continua, where one consonant belonged to a [s]-[ʃ] continuum and the other was one of [s,ʃ,z,tʃ,n,m]. Spliced (identical [sa]-[ʃa] continuum in each context) and non-spliced (natural coarticulation) stimuli were used. Categorization results were analyzed to determine whether sibilant harmony properties (similarity-sensitivity, trigger asymmetry, and regressive directionality) are mirrored by asymmetries in categorization of perceptually ambiguous stimuli, as we expect if sibilant harmony originates from hypocorrection. With [__aʃa] and [__atʃa] compared to neutral contexts, participants heard more of the continuum as [s]; they ‘hypercorrected’ real or perceived coarticulatory influences of the following sibilant. In contrast, participants heard more of the continuum as [ʃ] in the [ʃa__a] context compared to neutral contexts; this directionality shows ‘hypocorrection’. Spliced and non-spliced stimuli produced nearly identical results. Consequently, phonologized misperception of coarticulation alone is unlikely to explain the development of sibilant harmony, and the relationship between perception and phonological patterns is more complicated than previously believed. [full abstract]

Investigating Conflicting Aerodynamic Requirements in CC Clusters
Manfred Pastätter and Marianne Pouplier

We investigated articulography data of six native speakers of Polish in terms of the temporal coordination of stop and nasal onset clusters with varying C2. Previous research reported that C1C2 timing depends on prosodic alternation (i.e. more overlap of cluster consonants when they occurred in unaccented than in accented position) as well as on cluster composition. For instance, less articulatory overlap has been shown for stop+/n/ than for stop+/l, s/ clusters. This has been hypothesized to be conditioned by the aerodynamic incompatibility of stops and nasals, since an early lowering of the velum would interfere with the acoustic stop burst. Based on these findings, we expected for clusters with conflicting aerodynamic requirements (/pn/ and /mʃ/) comparably less C1C2 overlap and a lesser flexibility under prosodic variation compared to clusters without pronounced aerodynamic conflict (/pl, pʃ/ and /mn, ml/). For /pn/ we observed qualitatively less variation as a function of prosody than for /pl, pʃ/. However, against our expectations we found for /mʃ/ a clearly increased C1C2 overlap in unaccented position in spite of the potential aerodynamic conflict. In sum, the current results suggest that a complex interaction of language and cluster-specific effects determine C1C2 timing, rather than general aerodynamic requirements. [full abstract]

Phonetic devices and the construction of the phonological space
Elinor Payne, Brechtje Post, Nina Gram Garmann and Hanne Gram Simonsen

Languages differ in the way structure is phonetically implemented, making different use of a range of ‘phonetic devices’, which may each signal multiple structures. For example, the relative timing of VC sequences is an important cue to coda obstruent voicing in English (e.g. beat/bead), but the duration of the vowel in itself may also be an associative cue to vowel quality (e.g. bit/beat). Similarly, in Norwegian, VC timing also plays role in cueing [voice] in coda obstruents, but is additionally a key marker of syllable weight (e.g. hat/hatt). In learning to speak a language, children thus have to learn both the appropriate phonetic devices and how these are correctly mapped. This study examines how this cross-linguistic micro-variation in the ecology of phonetic mappings influences the acquisition pathway. We discuss the implications of our findings for the acquisition and dynamic nature of phonological representations, and argue that these are not acquired in isolation but emerge as part of a complete, dynamic system. During acquisition, by exploring the possible mappings and how they may be in competition, the child both seeks out the phonetic shape as projected by the phonological space of her ambient language, and constructs that phonological space for herself. [full abstract]

Sleep-dependent consolidation in the learning of natural vs. unnatural phonological rules
Sharon Peperkamp and Alexander Martin

Phonological rules tend to be phonetically ‘natural’: they reflect constraints on speech production and perception. Substance-based phonological theories predict that a preference for phonetically natural rules is encoded in synchronic grammars and translates into learning biases. This study explores this by focusing on the learning of palatal vowel harmony (phonetically natural) compared to vowel disharmony (phonetically unnatural). In addition, we investigate the role of memory consolidation after sleep on rule learning. Using an artificial language learning paradigm with test and retest after twelve hours, we exposed participants to singular/plural pairs where the plural suffix was either harmonic or disharmonic with the stem vowels. Participants not only learned the natural rule better than the unnatural one (contrary to previous findings), but also showed evidence for consolidation after sleep of the natural, but not of the unnatural rule. That is, during retest, learners who had first been tested in the evening and hence had slept, showed increased performance if they had been exposed to the natural rule but decreased performance if they had been exposed to the unnatural rule, while learners who had first been tested in the morning and hence had not slept showed decreased performance regardless of the rule. [full abstract]

Sonorancy of the rhotic /ɣ/ in Sgaw Karen
Pittayawat Pittayaporn and Sujinat Jitwiriyanont

With its two contrastive rhotic sounds /ɾ/ and /ɣ/, Sgaw Karen is a good case study for testing the phonetic basis of treating /ɣ/ as a sonorant. Three acoustic properties related to sonorancy were measured: harmonics-to-noise ratio, intensity, and the difference in band limited energy at lower frequencies (100-400 Hz) between the consonant and the maximum energy within the word. The results show that /ɣ/ is characterized by having the lowest HNR, lowest intensity, and the greatest change in low-frequency amplitude among the sonorants. Furthermore, two mixed effects logistic regressions were performed to determine whether any of the acoustic properties are predictors of sonorancy. The results reveal that HNR, intensity, and lower-frequency energy are all statistically significant. However, a model that treats /ɣ/ as an obstruent fits the data much better than one that treats it as a sonorant. While it is a rhotic sonorant in phonology, phonetically it is ambiguous but better classified as obstruents. This discrepancy suggests that most “phonetically sonorous” fricatives may fall on either side of the “sonorancy threshold” imposed by each individual language. The rhotic segments thus can be characterized as central and oral segments that are phonologized as sonorants in the language. [full abstract]

On the role of manner and place in Kurtöp tonogenesis
Sarah Plane and Gwendolyn Hyslop

According to the established model of tonogenesis, a contrast in consonant type conditions pitch on the following vowel. Voiceless onsets condition high pitch and voiced onsets condition low pitch. However, the manner in which this change spreads through the language is unknown. This study builds on Hyslop (2009), showing that in the Tibeto-Burman language Kurtöp, place and manner of consonant trigger play a role in the development of tone. Results from a production study with five speakers show that Kurtöp voiceless stops condition a fundamental frequency approximately 30 hertz higher than voiced stops, and that this difference is maintained over the duration of the vowel. With regard to voicing in the obstruents, we now see that the dental fricative is more often devoiced than any of the stops, suggesting that tonogenesis is further along in that environment. In regards to VOT of stops, phonologically “voiced” stops are indeed often realized as voiceless, however not consistently. Retroflex stops are most likely to be realized as voiceless, followed by the palatals and velars, and finally by the labials and dentals, which are most likely to still be phonetically voiced. [full abstract]

Car-talk: How physical environment influences speech production and perception
Ryan Podlubny, Jen Hay and Megan McAuliffe

Speech produced under sufficiently noisy conditions shows shifts in pitch, intensity, duration and formant values (known as the Lombard effect). Through a pair of speech-in-noise experiments, we explore two ways in which past experience with the Lombard effect might influence subsequent production and perception behaviours.

Experiment 1 investigates the possibility of noisy conditions also influencing speech perception: We ask if listening in noise leads listeners to adjust perceptual boundaries for vowels in ways that match previous experience with Lombard speech (i.e., predictably shifted formants). Experiment 2 explores a potential influence of physical environment (i.e., a laboratory vs. a parked car). Time spent in vehicles typically involves engine-sounds and road-noise; as a result, one’s past experience of speaking and listening in a car should be dominated by Lombard-type speech. We hypothesize that being physically positioned within a car may trigger situation-specific speaking and listening modes.

Results indicate perceptual boundaries for vowels are shifted in ways that resemble typical Lombard speech. We also find that Lombard-like changes to production and perception can be driven by experienced-based associations between noise and environment, and such changes can thus be triggered by a ‘typically noisy’ physical environment, even in the absence of background noise. [full abstract]

Infants prefer vowels with infant vocal resonances: Evidence for an “articulatory filter” bias
Linda Polka, Matthew Masapollo and Lucie Menard

Recent findings show that pre-babbling infants (at 4-6 months) preferentially attend to infant vowel sounds over adult vowel sounds (Masapollo, Polka, & Ménard, 2015). Here, we examined whether infant vocal resonance properties alone are sufficient to induce this preference. In three experiments, 5- to 7-month-olds were tested in a sequential preferential listening procedure using synthetic /i/ vowels. Experiment 1 revealed a robust preference for vowels with infant over adult resonance when f0 values of both vowel types were matched (315 or 360 Hz). Importantly, this preference was maintained even when f0 values were modulated, confirming that this bias is not attributable solely to higher f0. In Experiment 2, infants preferred vowels with infant formants and a relatively high f0 (400 or 450 Hz) over vowels with female formants and a relatively low f0 (315 or 360 Hz). In Experiment 3, infants preferred vowels with infant formants and a relatively low f0 (315 or 360 Hz) over vowels witih female formants and a relatively high f0 (400 or 450 Hz). Collectively, these results are compatible with the view that the infant vowel preference derives from an “articulatory filter” that biases infants’ attention toward vocalic signals that resemble their own productions. [full abstract]

Probing the interaction of dynamic stability with grammar: Evidence from Russian
Marianne Pouplier, Stefania Marin, Conceição Cunha and Alexei Kochetov

We investigate whether there is an interaction between speech motor stability induced by grammar and physiological stability. We ask whether in situations in which universal production constraints have been argued to emerge (articulatory reorganizations in forced rate-scaling tasks), reorganization patterns may vary in a language-specific fashion, depending on a language's phonotactics. Specifically, cluster-like structures may emerge under speech rate pressure. Using Russian, we study whether the likelihood of cluster emergence is influenced by the existence of a lexical correspondence cluster and whether the overlap pattern of emergent clusters mirrors those of lexical clusters. Lexicality had neither a significant influence on the frequency of cluster emergence nor on the emergent overlap patterns. This supports the view these types of tasks trigger extra-linguistic, physiologically universal stability patterns. A more fine-grained analysis revealed a preference for obstruent-sonorant clusters to emerge over sonorant-obstruent clusters, in agreement with their lexical frequency in Russian. Moreover, overlap differences between cluster types seem to be enhanced by existence of a lexical correspondence cluster. We interpret our results as showing how grammars may capitalize on physiologically given patterns, yet at the same time the powerful mechanism of learning allows languages to overcome these patterns in a non-deterministic fashion. [full abstract]

‘/t,d/ Deletion’: Articulatory Gradience in Variable Phonology
Ruaridh Purse and Alice Turk

t,d/ Deletion – the deletion of post-consonantal word-final coronal stops (CCCOR#) – is a paradigm variable in sociolinguistics and variable phonology. Until now, despite much debate in the literature, no real evidence has been uncovered concerning the articulatory nature of the phenomenon. In this study, we use Electromagnetic Articulography (EMA) data from the ESPF DoubleTalk Corpus of Scottish and Southern British English to look at the articulatory realizations of these stops. We present evidence that categorical /t,d/ Deletion (with no evidence of tongue tip raising) does exist but it is extremely rare, occurring in only 9 of the 289 tokens included in the analysis. Two of these tokens had no tongue tip tangential velocity target corresponding to the coronal stop, and 7 exhibited tongue tip lowering. For 25 other tokens, there was no audible reflex of a coronal stop, but demonstrable tongue tip raising from a neutral position that either undershot in its trajectory towards closure or was masked by surrounding segments. Further, analysis of tongue tip height corresponding to all /t,d/ tokens showed that Southern British English speakers, but not Scottish English speakers, produced stops with a lower tongue tip in monomorphemes than in simple past forms (2-4mm differences). [full abstract]

A phonologically weak contrast can induce phonetic overlap
Margaret Renwick, Ioana Vasilescu, Camille Dutrey, Lori Lamel and Bianca Vieru

The Romanian central vowels /ɨ, ʌ/ are marginally contrastive, with few minimal pairs and low functional load. We show that in continuous speech, unlike in laboratory speech, their contrast is phonetically weak, with strong overlap between /ɨ, ʌ/. A laboratory speech corpus (5,261 vowel tokens) is compared to broadcast speech data (104,456 tokens). Despite occurring in many frequent function words, the central vowels are rare; [ɨ, ʌ] are only 10.5% of vowels in the broadcast dataset. Their distribution is highly complementary: [ɨ] is common word-initially and pre-nasally, while [ʌ] is often word-final; morphology and word stress also condition the vowels. In lab speech, the central vowels are distinct; but in broadcast data [ɨ] lowers, and the vowels’ distributions are highly overlapping in the F1 dimension. The datasets show that although the central vowels have different phonological representations, their marginal contrast is severely diminished in continuous speech, leading to a phonetic near-merger. The lack of a strong distinction in production suggests that speakers’ cognitive representations of /ɨ, ʌ/ are not separable from the contexts in which each appears. We suspect that many languages maintain contrasts via speakers’ knowledge of context rather than acoustic distinction, a phenomenon little acknowledged by phonological theory. [full abstract]

How is lexical gemination transposed in Tashlhiyt whistled speech?
Rachid Ridouane and Julien Meyer

This paper addresses a traditional language practice based on whistling – called whistled

speech – in Tashlhiyt, a variety of Moroccan Amazigh. As part of a larger database on various phonological aspects of the language, we examine how the key acoustic properties of gemination contrast for voiced and voiceless stops is rendered in whistled speech and how position in the word shapes variability in the way this contrast is transposed. Compared to spoken forms, whistling, while adapting to the specific constraints imposed by the medium, seems to transpose the basic strategies used in normal speech to convey lexical gemination contrast. As results from three whistlers show, the clearest cue to gemination in word-intervocalic and word-final positions was closure duration. Supplementary cues are conveyed which may serve to enhance the primary correlate by contributing additional acoustic properties increasing the perceptual distance between the two lexical categories. These enhancing cues may take on distinctive function in cases where the primary correlate – duration – is not implemented. This is, for instance, the case of higher frequency values in word-initial position where duration differences cannot be acoustically implemented using whistled speech. [full abstract]

Acoustic salience and input frequency in L2 lexical tone learning
Katherine Riestenberg

Studies of second language (L2) lexical tone learning have found that learners are better at perceiving and producing tones that are acoustically salient, highly frequent in the input, or both. In many tone languages, however, tones with high acoustic salience are the least frequent across the lexicon while tones with low acoustic salience are the most frequent. This begs the question of how learners make use of these competing factors during tone learning. This poster presents the initial results of a longitudinal study investigating L2 lexical tone development using a classroom corpus and periodic tests of lexical tone production. The learners in the study are Spanish-speaking children learning Macuiltianguis Zapotec as a L2 through an after-school language revitalization program in Oaxaca, Mexico. The results show that the learners employed a default tone strategy using mid tones, the tone type that was the most frequent in the input. At the same time, some learners showed early accuracy producing dipping tones, which have high acoustic salience, on frequent lexical items. Overall the results suggest that both acoustic salience and input frequency are relevant for L2 lexical tone learning, but that they drive different aspects of development. [full abstract]

On the link between glottal vibration and sonority
Megan Risdal, Ann Aly, Adam Chong, Patricia Keating and Jesse Zymet

Adequate characterization of the physical manifestation of sonority is currently debated and so far does not make clear connections to inherent source-filter dependencies (Parker 2002). We replicated and extended Mittal et al. (2014), a prior investigation into the effects of vocal tract constriction on glottal vibration. We rely on the Contact Quotient (the proportion of a complete vibratory cycle for which vocal fold contact area is greater than a specified threshold) from EGG signals and a Strength of glottal Excitation measure (the relative amplitude of impulse-like excitation derived from the instant of significant excitation of the vocal-tract system during production of speech; Murty and Yegnanarayana (2008)). A lower Contact Quotient (CQ) indicates a more spread glottis (breathier voicing), and a lower Strength of Excitation (SoE) indicates less acoustic energy from voicing. The results of our study indicate that CQ and SoE, contrary to our null hypothesis, do vary depending on the degree of oral constriction, exemplifying effects of the filter on the source. From the perspective of sonority, while CQ and SoE together accurately capture the ends of the sonority scale, they make independent distinctions within the major segmental categories. [full abstract]

Representation of intonation categories: evidence from speech perception
Joe Rodd and Aoju Chen

The question of whether intonation events have a categorical mental representation has long been a puzzle in prosodic research, and one that experiments testing production and perception across category boundaries have failed to definitively resolve. This paper takes the alternative approach of looking for evidence of structure within a postulated category by testing for a Perceptual Magnet Effect (PME). PME has been found in boundary tones but has not previously been conclusively found in pitch accents. In this investigation, perceived goodness and discriminability of re-synthesised Dutch nuclear rise contours (L*H H%) were evaluated by naive native speakers of Dutch. The variation between these stimuli was quantified using a polynomial-parametric modelling approach (i.e. the SOCoPaSul model) in place of the traditional approach whereby excursion size, peak alignment and pitch register are used independently of each other to quantify variation between pitch accents. Using this approach to calculate the acoustic-perceptual distance between different stimuli, PME was detected: (1) rated “goodness” decreased as acoustic-perceptual distance relative to the prototype increased, and (2) equally spaced items far from the prototype were less frequently generalised than equally spaced items in the neighbourhood of the prototype. These results support the concept of categorically distinct intonation events. [full abstract]

High frequency prototypes do not facilitate phonotactic generalizations
Timo Roettger and Dinah Baer-Henney

Phonotactic regularities are learned rapidly. Even adult learners encode phonotactic information and generalize these to new instances in short periods of time. In most experimental investigations on this topic, learners are exposed to a balanced number of items. However, evidence from both non-linguistic and linguistic categorization tasks suggests that the structure of exposure has an impact on the learned generalization. It has been found that learning a category is facilitated if learners are initially trained on items that are presented more frequently. The present study investigates this aspect of categorization with respect to phonotactic learning to shed light on the underlying categorization mechanisms involved. While learners in previous studies were exposed to balanced distributions of tokens per category, the present study exposes learners to skewed distributions of tokens per category. Results indicate that skewed distributions do not facilitate phonotactic generalizations. This suggests that learners do not simply assemble concrete exemplars that they base their generalization on, but rather form hypotheses about possible generalizations right away. Alternatively, the null result may be due to a ceiling effect. Learning rates for the balanced condition were already very high leaving not enough room for significant facilitation in the skewed condition. [full abstract]

Preserving speech dynamics in Parkinson's disease: an acoustic study of the production of glides by Belgian French patients
Virginie Roland, Véronique Delvaux, Kathy Huet, Myriam Piccaluga, Marie-Claire Haelewyck and Bernard Harmegnies

In this study, we investigate the dynamics of supra-laryngeal articulators in Parkinson's Disease (PD). We report on an acoustic study of the production of glides and steady vowels by 9 Belgian French non dysarthric PD speakers (6 male, 3 female, aged 52-77) and 10 control speakers. They were administered 3 self-assessment questionnaires (VHI; PDQ-39; MHAVIE 4.0) as well as speech tasks including the production of sustained oral vowels [a,i,u] and V1[a,u]C[glide]V2[a,i,u] pseudo-words, the repetition of CV(C)CV pseudo-words and the reading of a short text. Acoustic measurements include duration and formant frequencies (all vocoid segments), slopes of F2 trajectories (glides), and time from V1 onset to the point where F2 reached its extreme within the glide (VCV pseudo-words).

Results suggest that non dysarthric Parkinsonian speakers maintain an accurate production of glides in VCV pseudo-words at the expense of articulatory undershoot in the surrounding vowels, and some assymetry between the V1-to-glide and glide-to-V2 articulatory movements. We discuss how these results both support and challenge the accuracy-timing trade-off hypothesis (Ackermann & Ziegler, 1991), and more generally how laboratory studies of the dynamics of supra-laryngeal articulators in vocoids may shed light on speech motor control in PD, especially when dysarthria is infraclinic. [full abstract]

Blending of articulator activation in a dynamical model of phonological planning
Kevin Roon and Adamantios Gafos

In response-distractor tasks, speakers hear distractor syllables while preparing to produce responses. The concurrent use of perception and production makes these tasks especially apt for addressing the assembly of representations at the microchronic timescale. A model of phonological planning has been developed on the basis of such tasks (Roon & Gafos 2016). Its key unit is that of the dynamic field, a distribution of activation over the range of values associated with each phonological parameter, which evolves in time via a dynamical system. We show that the same model can account for phonetic modulations (not response times) in different tasks. Yuen et al. (2010) reported increased alveolar contact for /k/-initial responses preceded by /t/-initial distractors, compared to /k/-initial responses preceded by /k/-initial distractors. Crucially, responses were time-locked to a cue. In our model, parameters such as (a distractor's) constriction location contribute activation peaks at their intended values which compete with those of the response via lateral inhibition. Because responses were externally elicited via the cue, inhibition between response and distractor does not completely suppress the distractor activation (as in the case when readout of the evolving activations is not externally enforced), thus leading to the observed effects of multiple articulation. [full abstract]

Plural predictability and OCP influence plural morpheme duration in English
Darcy E. Rose

This corpus study uses the phonetics of English plural /s/ to explore the effects of two novel influences on the phonetic realization of single-segment morphemes. Holding previously reported effects constant, we demonstrate a significant long distance gradient OCP effect, and a significant effect of contextual morphological predictability.

While long-distance OCP effects have been found in English, previous studies have focused on categorical phenomena such as the presence or absence of segments which violate the OCP. While the conditioning factors in some of these studies have been gradient, with violations occurring more or less often depending on the specific violation, the present study demonstrates that OCP effects can be manifested through phonetically gradient realizations of segments which violate the OCP.

The effect of contextual morphological predictability provides evidence in support of the idea that morphological structure is relevant to the phonetic realization of segments, contributing to the debate around whether morphemes are psychologically real entities independent of the words in which they occur. [full abstract]

An experimental approach to perceptual salience
Hanna Ruch

Perceptual salience is considered to play a role in phonetic accommodation and sound change in the sense that more salient features are more likely to be adopted (or avoided) in linguistic accommodation. This study presents a method how perceptual salience of dialect-specific segments can be quantified. Perceptual salience was operationalized following MacLeod (2015) by measuring their contribution to dialect recognition. In a forced-choice perception experiment, Zurich German listeners were asked to categorise real words spoken by Grison or Zurich German speakers while accuracy and reaction time (RT) were measured. The stimuli contained either none, one or two segmental dialect differences. Independent of word duration, certain dialect-specific segments and a higher number of dialect-specific segments led to a higher accuracy and a shorter RT. According to the results, /k(h)-(k)x/ and unstressed word-final /ɐ-ə/ appeared to be the most salient dialect differences. This finding was in line with Zurich German speakers' explicit knowledge about Grison German pronunciation, but only partly with what would be predicted by auditory principles. The results support the view by Auer (2014) that acoustic (e.g. intensity, transition), cognitive (e.g. the listener's experience with other dialects) and sociolinguistic factors (e.g. attitudes towards a dialect) jointly contribute to perceptual salience. [full abstract]

Lexical knowledge is available, but not always used, very early
Amanda Rysling, John Kingston, Adrian Staub, Andrew Cohen, Jeffrey Starns and Anthony Yacovone

Lexical knowledge informs categorization as early as possible, but is not recruited unless listeners develop an appropriate response strategy. Participants categorized a /s-f/ continuum in lexical /s/-biasing /aɪd/ (side, *fide), /f/-biasing /aɪl/ (file, *sile), and no biasing /aɪm/ (*sime, *fime) contexts.

In Exp. 1 (free-response), participants responded "f" and "s" more in the appropriate biasing contexts. In Exp. 2 (response-signal), participants responded within a 300ms window when prompted at 375, 675, 975, and 1350ms following stimulus onset (375ms = biasing consonant first audible). The proportion of "f" responses was greater in the /f/-biasing context, this lexical bias was present at the earliest prompt, and increased at later prompts. Exp. 3 used 175 and 375ms prompts. The distribution of responses to the 375 ms prompt differed between Exps. 2 and 3, being confined to the second half of the interval in Exp. 2 but spread across the entire interval in Exp. 3. However, lexical knowledge only biased responses during the second half of the response interval in Exp. 3, i.e. on trials when they waited as long to respond as participants in Exp. 2. When participants must respond earlier than comfortable, but can wait for lexical information, whether they wait depends on task demands. [full abstract]

The adaptation of Mandarin falling diphthongs in Heritage Korean in China: The interaction of linguistic and sociolinguistic factors
Na-Young Ryu, Yoonjung Kang and Sung-Woo Han

The paper investigates the adaptation of Mandarin falling diphthongs into a heritage Korean dialect from linguistic and sociolinguistic perspectives. This study is based on loanword data, established loans and on-line adaptations, collected from seven ethnic Koreans in Dandong, China, who are bilingual speakers of Korean and Mandarin. Two major adaptation patterns emerge in the data: coalescence into a monophthong (“substitution”) and retention of the diphthong (“importation”). The monophthongal adaptation obeys the constraint against falling diphthongs in native phonology and is the dominant pattern, while the diphthongal adaptation retains the foreign structure and preserves the syllable count at the expense of violating a native constraint. We found an effect of duration difference of tones on adaptation, with a longer tone (T3) inducing more diphthongal adaptations than shorter tones. In addition, word-final position, where vowels are longer, is more likely to elicit a diphthongal adaptation than non-final position. These findings suggest that phonetic factors in a source language play a considerable role in adapting loanwords. Also, the younger speakers produced more diphthongal realization than middle-age speakers in on-line adaptations, indicating that speakers may internalize different adaptation patterns depending on their level of bilingualism. [full abstract]

ERP evidence for the ecological validity of artificial language learning
Lisa Sanders, Claire Moore-Cantwell, Joe Pater, Robert Staubs and Benjamin Zobel

The experimental study of artificial language learning has become a widely used means of investigating the predictions of theories of phonology and of learning (see Moreton and Pater (2012a; 2012b) for a review). Because of this, it is becoming increasingly important to understand how lab-learned phonological patterns are cognitively represented, and in particular if that representation is similar to naturalistically learned phonological generalizations. We use neurological data to argue that lab-learned phonotactic patterns are encoded abstractly, similarly to native-language grammatical generalizations both at the phonological and at the syntactic level. Brain responses to violations of lab-learned phonotactics were examined. Novel words that violated a learned phonotactic constraint elicited a larger Late Positive Component (LPC; also known as P600) than novel words that satisfied it. The LPC is also associated with syntactic violations and with violations of musical expectations - knowledge which is considered to be abstract in nature. We argue that the presence of an LPC in response to a lab-learned phonotactic pattern constitutes evidence that that pattern (a) is abstractly represented, similarly to a syntactic generalization, and (b) is represented in a cognitively similar way to native language phonotactics. [full abstract]

Comparing neighborhood density and clear speech effects in the French vowel system
Rebecca Scarborough and Cecile Fougeron

We investigate phonetic variation in French vowels as conditioned by two factors associated with hyperarticulation and/or reduction: communicative context (demand for clarity) and phonological neighborhood density (ND). Native French speakers dictated 144 phonetically controlled high-ND and low-ND disyllabic French words to a confederate listener in a worksheet fill-in task (producing prompts of the type “It’s the word XXX on the yellow star.”). A second version of the same task was completed during exposure to conversational noise, increasing demands of clarity. Both in high ND words and in the clear task, results showed larger F1/F2 vowel space and greater dispersion, less variability within vowel categories, and lower vowel misclassification rates in linear discriminant analysis based on F1 and F2, as well as based on F2 and F3 for front vowels (where rounding is contrastive). Thus, both communicative contexts requiring clarity and potential lexical competition yielded clearer productions. However, this clarity was not identical in the two cases. High ND words appear to be produced such that vowel-specific dimensions are specifically enhanced, while in the clear task, adjustments are less sensitive to the specific system of contrasts. These multi-dimensional results contribute to a more comprehensive understanding of the notion of hyperarticulation. [full abstract]

Influence of Syllable Structure on Musical Text Setting
Murray Schellenberg

Following claims that the syllable structure of a language may influence how a composer sets that language to music, this paper looks at whether there is a correlation between syllable structure and the use of multiple notes for a single syllable (melismata) in 28 art songs in 4 different Indo-European languages: German, Italian, English and French (7 songs in each language). The number of notes per syllable and the syllable structure (open or closed) were recorded for the first and last 25 syllables of each song, excluding repetitions of lyrics.

There is no significant difference between Italian and English but both French and German are significantly different from English. Syllable type does not make a significant difference. The results support claims that Italian and German are different. However, English, which patterns with German in terms of allowable phonotactic complexity, appears to pattern with Italian having an almost equal number of melismatic syllables despite permitting much more complex syllable structure. French, which might be expected to pattern with Italian, trails behind with almost a quarter of the number of melismatic syllables of either English or Italian. These results call into question previous claims that syllable type influences meslismatic text setting. [full abstract]

How much does the talker matter? Depends who's listening: Age-related variability in the use of social information in speech perception
Jessamyn Schertz, Yoonjung Kang and Sungwoo Han

Listeners' knowledge of sociophonetic variation has been shown to affect speech perception, but the extent and nature of this influence is not well understood. The current work examines how listeners adjust their phonetic categorization when listening to talkers from different dialects. We present data from 124 native Korean listeners from Hunchun and Dandong, China, who categorized stop-initial syllables varying in VOT and f0. Although stimuli were acoustically identical, listeners were told that the talker was either from their local community (in one block) or from Seoul (in another block). Listeners showed different cue-weighting strategies when listening to the different "dialects," with increased reliance on f0 and decreased reliance on VOT when they believed the talker to be from Seoul, and the age and/or gender of the talker also influenced categorization. However, the influence of these "talker" effects was not equal for all listeners, with younger listeners carrying the effects. We hypothesize that younger members of these communities have a heightened awareness of how subtle differences in pronunciation are linked to social variables, and in turn make greater use of this information in speech perception. Implications for the effects of dialectal variation on perceptual representations are also discussed. [full abstract]

L1 influence on the identification of intonational contours
Elaine Schmidt, Carmen Kung, Brechtje Post, Ivan Yuen and Katherine Demuth

In this study we investigated the influence of L1 (here Mandarin Chinese) on the identification of intonational contours in an L2 (here Australian English). We hypothesised that Mandarin learners would be sensitive to subtle intonational differences as a function of their tonal language background. However, we predicted that this sensitivity would be modulated by the position in which pitch changes occur, thus reflecting the influence of L1 on L2 perception. Our results demonstrate that Mandarin learners of English are sensitive to pitch height of the boundary tone as a factor in determining the discourse function of an utterance. Specifically, higher pitch in the boundary tone gradiently shifts listeners’ interpretation toward an interrogative response. In contrast, lower pitch height at the boundary of a rise results in more declarative responses. As such, the identification pattern found here is comparable to that of Australian English listeners. However, Mandarin listeners were more likely to interpret rises that differed only in pitch accent as interrogatives. This suggests that for Mandarin learners the perceptual space of rises is more likely to be reserved for the linguistic category of interrogatives. The results demonstrate the influence of L1 on the identification of intonational contours in an L2. [full abstract]

Cross-language differences in vowel inherent spectral change - evidence from Polish learners of English
Geoff Schwartz, Jarosław Weckwerth, Kamil Kazmierski, Aperlinski Grzegorz, Malarski Kamil and Mateusz Jekiel

As yet, research into vowel inherent spectral change (VISC) has primarily concentrated on English as an L1; there have not been many cross-language studies, and VISC has remained outside of the focus of L2 speech research. Since languages show systematic differences in the relative degree of spectral dynamics, VISC may qualify as an additional dimension in the definition of cross-language phonetic similarity, a crucial concept for models of L2 speech This paper will present acoustic data documenting VISC in the speech of Polish users of English. The Polish vowel system is characterized by relatively stable formant patterns, leading to a hypothesis that acquisition for Polish learners must entail mastery of native-like patterns of VISC in English. We compare students in their first year of university English-language studies with highly proficient Polish users of English, lecturers and professors employed in a university English department. An analysis of the FLEECE and TRAP vowels revealed significantly greater formant excursion and formant slope measures in the speech of the more proficient group, while the less proficient group shows higher measures of formant stability. This latter finding is compatible with patterns observed for L1 vowels that appear as substitutes in Polish-accented speech. [full abstract]

Dynamic listening: temporal expectations guide perception of phonetic detail
Jason Shaw

Processing advantages found for familiar talkers have been sometimes interpreted as support for lexical representations incorporating both phonological and indexical information, although the specific phonetic parameters that contribute to familiarity advantages are poorly understood. In this study, we investigated how phonetic parameters relating to time, i.e., gesture duration and gesture coordination, influence talker discrimination, a task requiring perception of talker-specific phonetic detail. Australian English listeners heard a sequence of one to six words, and were asked to detect a change in talker. Exposure and test talkers differed in temporal typicality, defined as the degree to which that talker’s timing patterns can be described by population-level distributions (as opposed to talker-specific distributions). Results indicate, firstly, that listeners rapidly adapt to the exposure talker regardless of temporally typicality, and, secondly, that temporal differences do not aid in talker detection. On the contrary, it is harder to recognize a change in talker when the test talker is temporally distinct. We interpret this result as evidence that processing of talker-specific phonetic detail is guided by temporal expectations. Listeners appear to adjust expectations on a talker-specific basis for when, in time, task-relevant information will become available. [full abstract]

The articulatory space of oral and nasal vowels in Brazilian Portuguese
Ryan Shosted, Denise Pozzani, Francisco Meneses, Nicole Wong, Zainab Hermes and Torrey Loucks

This study uses electromagnetic articulography (EMA) data from seven speakers of Brazilian Portuguese (BP) to highlight articulatory differences between nasal vowels and their oral congeners. Recent studies have demonstrated that oral–nasal vowel pairs differ with respect to their oro-pharyngeal configuration (beyond velopharyngeal opening) and that these changes may serve to enhance acoustic features of nasality. Comparisons of the oral–nasal low vowels in BP show that differences are due primarily to jaw position, not tongue position, as previously supposed. Additional investigations using time-dynamic SSANOVA demonstrated diphthongization of the nasal vowels /ĩ, ẽ/. A principal component analysis (PCA) captured the raising and backing of the nasal low vowel, associating nasality with decreased frontness and increased height. Articulatory-based PCs were used in a linear discriminant analysis to differentiate vowel categories in oral and nasal spaces. Oral vowels were classified more precisely along these articulatory parameters than nasal vowels, recalling centralization, one of the acoustic consequences of velopharyngeal opening. For BP, reduction of acoustic difference in the nasal vowel space is mirrored in articulation and can be considered articulatory enhancement of the oral–nasal vowel distinction, though it increases the potential for collapse of vowel distinctions internal to the nasal vowel space. [full abstract]

Intrinsic pitch of diphthongs in lexical tone perception
Jessica Siddins and Eva Reinisch

Listeners can distinguish between two determinants of fundamental frequency: lexical tone (phonological) and vowel height with associated intrinsic f0 (phonetic). The interplay between these factors is especially relevant for users of complex tone languages such as Hong Kong Cantonese but has rarely been studied. Intrinsic pitch is a well-known perceptual normalisation process whereby listeners factor out the (phonetic) effects of intrinsic f0. We tested Cantonese listeners’ perception of a tone continuum from low-falling via low-level to low-rising on closing and opening diphthongs. We predicted that if listeners normalise for intrinsic f0, there should be more low-falling responses for the closing diphthong (i.e. falling intrinsic pitch) and more low-rising responses for the opening diphthong (i.e. rising intrinsic pitch). However, results showed no effect of diphthong (i.e., intrinsic pitch) on the perception of the low-rising tone and an effect in the opposite direction for the low-falling tone. Thus, it is likely that any intrinsic f0 in the signal is (mistakenly) parsed as part of the lexical tone. This finding is important for our understanding of tone processing, as it indicates that phonetic variation can be misparsed as being phonological not only in the segmental but also the suprasegmental domain. [full abstract]

Representation of Acoustic Detail
Michelle Sims and Benjamin V. Tucker

This study investigates whether acoustic detail provides a link between production and processing. Our previous research has found that morphological information correlates with the production of vowel duration in irregular English verbs. In the present study, we test whether this production-based correlation is of consequence to processing (i.e. whether production and processing are linked). We test for the existence of a production-processing link in two auditory speech recognition paradigms: lexical decision and morphological decision. We find that production and processing are linked, though the link is weak and incomplete. Our results indicate that the production-based correlation between morphology and vowel duration affects processing only in certain tasks and within particular morphological conditions. We interpret this weak production-processing link in terms of the mental representation of acoustic detail. We conclude that acoustic detail is captured in a learned network of statistical associations between a word’s stored meaning and its produced form. Thus, acoustic detail acts as a link between form and meaning, though the strength of that link is dependent on the strength of its statistical association in the learning network. [full abstract]

A production-internal learning bias against large changes to the base
Amy Smolek and Vsevolod Kapatsinski

Paradigm uniformity refers to the force militating against stem changes, or, in other words, against having multiple forms of the same stem. Especially abhorrent to paradigm uniformity are cases in which the allomorphs of the stem are phonologically dissimilar. Previous explanations for paradigm uniformity have invoked perceptual similarity. We propose a novel explanation that instead relies on articulatory similarity. Associations between dissimilar stimuli or representations have been shown to be harder to learn than associations between similar stimuli or representations. We argue that learning to produce an alternation between X and Y involves learning associations between production representations of X and Y: XY and YX. Once the associations are built up, a wug test participant can be given a form containing X and generate from it a form containing Y, and vice versa. Without the associations, the presented form is simply copied into the speaker’s output. The proposed theory predicts that the primary difficulty in acquiring an alternation is learning to change the segments that should be changed in production, and that a hard-to-produce stem change may nonetheless be perceived as acceptable. We present evidence for this prediction from miniature artificial language learning of palatalization. [full abstract]

Characterizing vocal tract dynamics with real-time MRI
Tanner Sorensen, Asterios Toutios, Louis Goldstein and Shrikanth Narayanan

Real-time magnetic resonance imaging (rtMRI) provides information about the dynamic shaping of the vocal tract during speech production and valuable data for creating and testing models of speech production. We report on the development of a dynamical system in the framework of Task Dynamics which controls vocal tract constrictions and induces deformation of the air-tissue boundary. The system parameterizes the midsagittal air-tissue boundary of the vocal tract as a combination of independent tongue, lip, and jaw factors extracted from rtMRI videos using an anatomically guided factor analysis. Weights on these factors vary in time to change vocal tract shape. We introduce a locally linear forward map, based on a hierarchical clustering algorithm, from these factors of vocal tract shape to constriction degrees along the vocal tract. This forward map quantitatively characterizes the relation between vocal tract shape and vocal tract constrictions. We use the forward map to build a speaker-specific task dynamical system. This is the first task dynamical system explicitly derived from speech kinematic data which can be used to quantitatively validate aspects of the dynamical system through error analysis. [full abstract]

Vowel movement as a function of voicing in simple CV sequences
Stavroula Sotiropoulou, Tanner Sorensen, Stephen Tobin and Adamantios Gafos

One of the articulatory and acoustic correlates of the stop voicing contrast is to be found in the velocity of the vocalic gesture of the tongue. Earlier ultrasound studies based on limited stimuli and subjects suggest that these vocalic gestures are faster and achieve their targets earlier after voiced than voiceless stops. However, it is not yet known whether the initiation of the vocalic gesture is sensitive to differences in voicing and in particular to variation in voice-onset-time (VOT). Using acoustic and articulatory data from four native speakers of American English, we examined vowel initiation in CV sequences as a function of consonant voicing at different places of articulation in low and high vowel contexts (e.g. beach-peach, dock-tock). Our results show that the vocalic gesture starts earlier and unfolds more quickly after voiced than voiceless stops. Additionally, this binary contrast extends to within-category VOT variation among voiceless aspirated stops. Longer VOTs were associated with later initiation of the following vowel, and slower achievement of the vocalic target. Thus, initiation time and velocity of the vocalic gesture are systematically conditioned by the presence of the laryngeal gesture and its timing with the oral constriction. [full abstract]

The role of echoic memory in the initial learning of a second dialect: the case of bilinguals
Laura Spinu, Jiwon Hwang and Renata Lohmann

Research suggests that bilinguals are generally better at learning foreign languages as adults, e.g. in terms of auditory processing or novel word acquisition. One of the most difficult tasks for the adult second language learner has to do with the phonetic and phonological details of the target language, for which native-like proficiency is typically not attained. The question we address in this study is whether the ”bilingual advantage” also applies to phonetic and phonological learning. We compare the performance of 17 monolinguals and 12 bilinguals in a production experiment with two tasks: imitation and spontaneous reproduction of novel foreign accents. We analyzed acoustically aspects of their production related to intonation, vocalic space, and consonant realization. Our results reveal different behaviors for the two groups and support the idea that there is a bilingual advantage in phonetic and phonological learning. We interpret these findings in light of recent psycholinguistic work and conclude that echoic memory strategies, underlain by stronger subcortical encoding of sound in bilinguals, may account for our results by facilitating the re-mapping between existing mental representations of sounds and existing articulatory command configurations. [full abstract]

Phonetic Shift /ɔr/ Phonemic Change? American English mergers over 40 years
Joseph A. Stanley and Margaret E. L. Renwick

We present a real-time study of a speaker of Utah English, to show phonemic change through time. Born in a time and place where the cord-card merger (/or, ɔr/ ≠ /ar/) was widespread, our speaker is expected to initially have this merger, though we hypothesize that over time he developed the hoarse-horse merger (/or, ɔr/ ≠ /ar/) to match changes in the surrounding community. From 36 hours of his religious sermons given over 42 years (1973–2015) we analyze formant measurements extracted from 7,678 potential cord-card words, defined as words with pre-rhotic [o] or [a] in stressed position. Using vowel plots and overlap measures, we show that our speaker surprisingly had the unusual three-way split (/or/ ≠ /ɔr/ ≠ /ar/) in his 50s. However, between 1973 and 1990, his vowel system moves closer to a hoarse-horse merger, as expected, before a reversal occurs and it reverts back towards the three-way split between 1995 and 2015. We argue that this speaker's shift toward the more mainstream hoarse-horse merger represents not just phonetic change, but also reshaping of the phonological inventory over time.  [full abstract]

The effect of realtime visual feedback on vocalic targets
Elizabeth Stelle, Eric Vatikiotis-Bateson and Caroline L. Smith

During speech production, realtime perturbations to a speaker's auditory feedback cause compensatory shifts in the production of the perturbed phonological category (e.g. Houde & Jordan, 1998), demonstrating the dynamic nature of phonological contrasts. Visual speech information plays an important role in speech perception, but there is growing evidence that it is also relevant to speech production (e.g. Chesters et al., 2015). The present research extends this work by testing the hypothesis that realtime visual feedback can be used during speech production to maintain accurate speech targets. We compared the effect of visual feedback on speech produced with and without a bite block, measuring acoustic and articulatory differences. Model comparisons of mixed effects models revealed that visual feedback significantly increased vowel contrast and significantly decreased vowel dispersion. Optical flow analysis was used to infer magnitudes of motion of the lower face. For a subset of participants, this motion increased with visual feedback for non-high vowels, and overall there was a significant positive correlation between motion and acoustic vowel contrast for non-high vowels. Our results support dynamic accounts of phonological targets; targets can be refined by novel visual feedback, with productions changing to enhance visually salient aspects of articulation. [full abstract]

Investigating the origins of pre-consonantal /s/-retraction: acoustic, perceptual and articulatory evidence from English
Mary Stevens and Jonathan Harrington

The sound change involving /s/-retraction in pre-consonantal position has taken place in several languages but has been the subject of little experimental phonetic study. We take English /st/ as a test case and consider acoustic, perceptual and physiological data in an effort to shed light on whether /s/-retraction in /sC/ can be considered a phonetically motivated sound change and, if so, to locate its origins in speech production or perception. The acoustic data involved native Australian English speakers’ (n =20) productions of seam and steam; results showed that the spectral centre of gravity was lower over the entire temporal middle half of the /s/ in steam than seam. This acoustic difference was audible to listeners (n=22): spliced into a pre-vocalic context, sibilants spliced from steam elicited more /ʃ/ responses than those spliced from seam. The remaining part of the study investigates the possible sources of retraction in English /st/, considering (1) physiological (coarticulatory) conditioning and (2) extension of /str/-retraction. Physiological (EMA) data showed that acoustic retraction in /st/ is speaker-specific and can be attributed to articulatory tongue tip retraction. That is, the acoustic retraction to which listeners were shown to be sensitive may have an articulatory basis. [full abstract]

Gradient phonological relationships: Evidence from vowels in French
Sophia Stevenson and Tania Zamuner

The dichotomy of contrastive and allophonic phonological relationships has a long-standing tradition in phonology, but research points to phonological relationships that fall between contrastive and allophonic (Hall, 2013). Measures of lexical distinction and predictability of distributions were applied to Laurentian French vowels to quantify three degrees of contrast between pairs: high, mid, and low contrast. According to traditional definitions, both the high and mid contrast pairs are classified as phonologically contrastive, and low contrast pairs as allophonic. As such, a binary view of contrast predicted that high and mid contrast pairs would pattern together on tasks of speech perception, and low contrast pairs would show a different pattern. The gradient view predicted all vowel pairs would fall along a continuum. Thirty-two speakers of French participated in an AX task and a similarity-rating task. The high, mid, and low contrast vowel pairs patterned differently across the experiments, providing evidence for what is being more frequently acknowledged in the theoretical literature, namely that there are phonological relationships that fall between purely allophonic or purely contrastive. This nuanced view can better represent the range of relationships between categories of speech sounds and further our understanding of sound patterns in human language. [full abstract]

Influence of palatalization on tongue-tip velocity in trills
Taja Stoll, Philip Hoole and Jonathan Harrington

Palatalized rhotics, unlike palatalized laterals, are unstable sounds from the diachronic point of view. This is probably due to their complex articulation: secondary palatalization likely disrupts conditions necessary for trilling (Ladefoged and Maddieson 1996, Kavitskaya 1997). High tongue-tip velocity might be one of such necessary conditions.

First, we compare the velocity profiles of the tongue tip in trills and laterals. Second, we analyze whether the tongue-tip peak velocity in trills is influenced negatively by palatalization, unlike laterals.

Five native Russian speakers were recorded by means of EMA. We measured tongue-tip peak velocity, maximum displacement, and stiffness in palatalized and plain trills and laterals. Peak velocity and stiffness are greater in rhotics, as compared to laterals. While showing similar maximum displacement, the peak velocity and stiffness are less in the palatalized rhotic, as compared to /r/. In contrast, the peak velocity and maximum displacement are greater in the palatalized lateral, as compared to its plain (velarized) counterpart.

Provided that the high tongue-tip velocity is important for trilling (the tongue tip might have to block the mouth cavity quickly and thus enable the rapid intraoral pressure accumulation), palatalization seems to impair the tongue tip in trills. [full abstract]

Articulatory dynamics of degemination in Dutch
Patrycja Strycharczuk and Koen Sebregts

In Dutch, derived clusters of identical consonants (“fake geminates”) are traditionally thought to undergo categorical degemination, though for fricative clusters it has been shown to be gradient durationally. Since Standard Dutch displays a pattern of strongly categorical /r/ allophony with coda-r a post-alveolar approximant and onset-r a uvular fricative/trill, /r#r/ clusters in a fake geminate context should show phonetically strongly distinct allophones (e.g. [ɻ#ʀ]). We investigate the spatial and temporal characteristics of coarticulation and possible degemination in these phonemically identical but phonetically disparate sequences.

Articulatory (high-speed ultrasound) and acoustic data from 8 speakers of Standard Dutch were analysed using dynamic (principal components analysis of pixel intensity data) and static (SS-ANOVA comparisons of tongue contours) measures, plus linear mixed-effects regression modelling of vowel+rhotic duration.

The results show that fake geminate-r entails an intermediate articulation between coda-r and onset-r, combining coda-like bunching with onset-like dorsal raising (gestural blending), with spatial reduction of both gestures vis-à-vis non-geminate onsets and codas. Durationally, however, degemination appears categorical, as the fake geminate context and singleton onset /r/ are non-distinct.

We discuss the problems these facts raise for both an Articulatory Phonology framework without a “phonemic” level, and more abstractionist models with discrete phonological and phonetic levels. [full abstract]

Social dynamics and phonological representations: Observations from speech and society in Scotland
Jane Stuart-Smith

An important bundle of factors in constraining and shaping phonological representations seems to be the social dynamics which speakers and listeners are constantly engaged in. In this talk I will consider some different representations of social dynamics, and how these might relate to speech, from macro-social categories, to more intimate links formed through social networks, and to micro-social processes as speakers construct shared social and linguistic practices (Labov 2001; Eckert 2012). Also relevant for investigating possible links between social dynamics and phonological representation, is how social dynamics may be captured and/or inferred through different kinds of sampling and speech elicitation. Recent sociophonetic work on Scottish English, which includes a further dynamic of social and phonological change over time, provides illustrations through three complementary examples:

-- articulatory investigation of gestural and timing differences in coda /r/ in socially-stratified laboratory recordings of Scottish English, alongside dynamic acoustic diachronic consideration of coda /r/ in spontaneous Scottish English speech;

-- dynamic acoustic representations of /s/ in spontaneous Glaswegian vernacular speech with respect to the social dynamics of gender in a changing city over time;

-- consonantal and vocalic variation in spontaneous Glasgow Asian speech, with rich representations of social dynamics relating to ethnic identities.

The link between anterior lingual gesture delay and loss of coda /r/: an ultrasound study
Jane Stuart-Smith, Eleanor Lawson and James Scobbie

We investigated the contribution that anterior lingual gesture delay makes to the lenition of postvocalic /r/, using a socially-stratified, audio-ultrasound corpus of adolescent Glaswegian English (16 speakers), containing word-list recordings from two sociolects; one with /r/ weakening (working-class speech) and the other with /r/ strengthening (middle-class speech). We quantified the temporal difference between the maximum of the anterior lingual /r/ gesture and either the offset of voicing in CVr word, or the onset of a following labial consonant in CVrC words, events that could mask the /r/ gesture. This articulatory measure we refer to as ‘lag’. F1-F5 were measured at the same location, or as close as possible, as the anterior lingual maximum. Rhotic strength of /r/ tokens was rated by the three authors.

Results showed middle-class and working-class speakers produced negative and positive lags respectively. Linear Mixed Effects analysis showed sociogender group significantly affected lag in our corpus, as did the presence of a pre-rhotic checked vowel. Correlation tests showed strong correlations between a long, positive articulatory lag, low /r/ index score (i.e. weak audible rhoticity) and a high F3. These results suggest that anterior lingual gesture lag is a key factor in the lenition of postvocalic /r/. [full abstract]

How are forms we rarely hear, understood so easily?
Meghan Sumner

Episodic theories of representation and lexical access are now strongly supported with a wide range of data. With this in mind, we might wonder how listeners seemingly navigate a variable signal with ease. For example, while there are clear benefits for frequently experienced variants and forms during speech perception, rarely experienced variants and word forms (oftentimes idealized) seem to have a benefit both in terms of spoken word recognition, and detailed memory retention. This pattern runs counter to our hypotheses and theories about the atypical, infrequent linguistic unit, from an episodic perspective (and, the specific memories associated with these infrequent forms run counter to notions of variation-independent abstract representations). In this talk, I provide an overview of this bias, illuminate instances in which our assumptions about the phonetic composition of a word have contributed to the bias, and show that even after such considerations, these less common pronunciation types do indeed have a processing benefit that does not match with our theoretical expectations. To address this discrepancy, I show that the effects we see in both word recognition and memory can be attributed to differences in speech perception. Specifically, all linguistic events, or instances of spoken words, are not treated equally by the perceptual system, resulting in a domino effect, weighting form-based representations, available for subsequent processing, differentially depending on both linguistic and social factors. This approach explains the idealized-form benefit, while shedding some insight on the representations and processes that underlie speech perception more generally.

Lack of evidence for subphonemic contrasts motivating exceptional behavior in vowel harmony
Daniel Szeredi

This paper discusses the plausibility of small subphonemic contrasts motivating exceptional allomorphy in vowel harmony. Three experiments were devised to understand whether antiharmonicity in Hungarian vowel harmony can be explained by segmental level phonetic differences. Hungarian antiharmonic stems contain non-low unrounded front vowels, but consistently take back suffixes (eg. híd `bridge', dative híd-nak and not *híd-nek).

An acoustic study showed that the difference between harmonic (regular) and antiharmonic (irregular) stems is at most minimal (contrary to what Benus and Gafos 2007 found). A perception experiment showed that these small differences are not distinguishable by Hungarian speakers. Finally, a wug study showed that perceivable differences do not motivate Hungarian speakers in categorizing stems as harmonic and antiharmonic. Based on these experimental results the paper rejects the phonetic level explanation for exceptionality and proposes a sublexical analysis instead. [full abstract]

Children's sensitivity to degrees of mispronunciation: Enriching the preferential looking paradigm with pupillometry.
Katalin Tamasi, Cristina McKean, Adamantios Gafos and Barbara Hoehle

Recent findings indicate that infants can detect sub-segmental changes in words, suggesting the encoding of sub-segmental detail. However, the degree of granularity of this sensitivity remains unclear: Evidence is inconclusive on whether infants give differential responses to small versus large degrees of mismatch. The current study extends previous work by combining the preferential looking paradigm with a measure automatically collected in such studies: pupil dilation. Our data found degree of mispronunciation (0-3 feature changes and novel label) to be a predictor of both looking preference and pupil dilation. Further, time-course analyses explored the evolution of the two measures in response to mispronunciation. Steady target preference was observed for correct and less robust but still significant target preference for Δ1F labels, whereas preferences flipped from distractor to target for Δ2F and Δ3F labels, and distractor preference was observed for novel labels. In the pupil measure, late differentiation of the Δ3F labels and early differentiation of the novel labels was observed. Our results indicate sensitivity to degrees of mispronunciation and, as such, corroborate previous work that found word recognition modulated by degree of mismatch. We thus establish that pupillometry can be used in combination with preferential looking paradigms to study lexical processing. [full abstract]

Lexical specificity and temporal decay in intraspeaker priming of sociolinguistic variables
Meredith Tamminga

Self-priming effects in morphophonological variation are taken as a window on the psycholinguistic processes involved in sociolinguistic production. Data from sociolinguistic interviews in Philadelphia show that within both ING (workin’ ~ working) and TD (ol’ ~ old), words where the variable is coterminous with a verbal suffix (work-ing, kick-ed) behave differently from words where the word containing the variable is monomorphemic (ceiling, old). Variant choice in polymorphemic words does not prime variant choice in subsequent monomorphemic words, and vice versa; morphologically-matched prime-target pairs, however, do show priming. Further, in the polymorphemic conditions the priming effect generalizes across different lexical items and decays significantly over about a minute, while monomorphemic TD shows long-lasting priming but only when the prime and target are the same word. I argue that variation in the polymorphemic cases involves morphological alternations, while the variation in the monomorphemic cases arises from phonological variation. When the variable is a suffix, the allomorphs are stored abstractly in the lexicon and are subject to repetition priming in lexical access, just like non-variable lexical items. Phonological variation, however, is retained only as part of episodic memories of the details of specific instances of whole words. [full abstract]

Kinematic aspects of L2 production in an imitation task
Mark Tiede, Christine Mooshammer, Dolly Goldenberg and Douglas Honorof

In this work we examine the L2 production of American English (AE) by native speakers of Mexican Spanish (MS) observed using electromagnetic articulometry (EMA) during an imitation task. Six MS speakers were age and gender matched with native AE speakers. The two speakers of a given pairing were seated face-to-face two meters apart, each with a computer screen for stimulus presentation. EMA sensors were glued to each speaker's midsagittal tongue dorsum and tip, upper and lower lips, and lower incisors, together with reference sensors to correct for head movement. Using a randomized alternation, one of the two speakers was presented with a target sentence for production (the initiator), and the other (the imitator) was instructed to repeat this sentence as carefully as possible. Sentence material was selected to include clusters, to observe patterns of overlap, and coda velars, whose trajectories are known to be sensitive to coproduced vowel quality. Timing landmarks were identified using velocity minima associated with points of maximum constriction. In imitations, the relative phasing of velars within clusters was observed to shift systematically in the direction of the initiator. Perceptual evaluation by native listeners of initiated vs. imitated L2 productions is used to assess whether this phasing change results in improved apparent naturalness. [full abstract]

MALD: Massive Auditory Lexical Decision
Benjamin Tucker and Daniel Brenner

The current presentation introduces and illustrates some of the uses of the MALD database (Massive Auditory Lexical Decision). The first phase of MALD, described in this presentation, contains lexical decision responses to 26,800 words and 9,600 pseudo-words produced by a single male speaker of Western Canadian English. Both response time and accuracy are available as dependent variables with common predictors such as frequency, neighborhood density, number of syllables, stress location, and uniqueness point available. We provide a description of the dataset and include several example analyses using the data. Different frequency measures have been be compared using this dataset. When we compare frequency counts from the Google Unigram corpus to the Corpus of Contemporary American English (COCA, counts across all genres compared to the spoken only counts) we find an effect of frequency for all the frequency measures (response latency decreases as frequency increases). Using model comparison techniques, we find that COCA frequencies (both spoken and all) better predict response times than those derived from the Google corpus. [full abstract]

The Function of Duration and Stress- vs. Syllable-Timing
Irene Vogel and Angeliki Athanasopoulou

While the literal interpretation of syllable- vs. stress-timed isochrony has been amply disconfirmed, the impression of this distinction persists, and alternative means of capturing it phonologically and phonetically have been proposed. Instead of measurements of durations of specific segments and intervals within the speech stream, we propose that the intuitive rhythmic distinction is a function of the extent to which vowel durations resist enhancement under conditions of prosodic prominence, specifically stress and focus. This is, the Stress Ratio (SR=duration stressed V/unstressed V) and the Focus Ratio (FR=focused V/unfocused V) will be close to 1 in syllable-timed languages, but not in stress-timed languages. We test this with Arabic, French, Greek, Finnish, Hungarian, Spanish, Portuguese and Turkish. While we find that Spanish, unsurprisingly, has a SR close to 1, its FR (1.3) shows some modification of duration. French, Greek and Portuguese have the highest SR, showing the most lengthening under stress, but only Portuguese also has a high FR; French and Greek are close to Spanish. The other languages show smaller differences in SR and FR. We additionally demonstrate the use of FR to express the rhythmic properties of non-stress languages, Indonesian, Korean, Vietnamese, all of which are under 1.15. [full abstract]

Asymmetries in English Liquid Production and Vowel Interactions
Rachel Walker, Michael Proctor, Caitlin Smith and Ewald Enzinger

We present new real-time MRI data and new methods of analysis to examine production of liquid consonants in General American English. We investigate two hypotheses: (i) /ɹ/ is defined by a tongue body gesture that exerts more global influence over tongue shaping and constriction than /l/; and (ii) /l/ shows greater variance in timing and place of tongue body articulation than /ɹ/. Under these hypotheses, [ɹ] is predicted to show less variance than [l] in lingual center of gravity (CoG) across vocalic contexts and syllable positions, and to show less variance in mean midsagittal lingual deformation in vocalic transitions. Four participants produced liquids in contrastive vocalic contexts. Lingual outlines were identified, and changes in articulation into and out of neighboring vowels were tracked using CoG and lingual displacement. Preliminary results indicate that rhotics constrain global tongue shaping more consistently across syllable positions than laterals. Achievement of tongue-tip and tongue body constriction targets is less synchronous in laterals, where tongue body constriction location differs more between onsets and codas, compared to rhotics. These stricter articulatory dictates could make finer vocalic distinctions more difficult to produce before /ɹ/, shedding light on maintenance of tense/lax vowel contrasts before coda /l/ but not /ɹ/. [full abstract]

Investigating the interaction between speaker dialect and listener differences across two tasks.
Abby Walker, Andrew Burlile and Katherine Askew

In this paper we test whether the same factors are behind poor performance with non-standard dialects in two different listening tasks - a false memory task, and a listening in noise task - by comparing the way in which individual differences across listeners do (not) interact with non-standard dialect performance in each task. We find that while non-native speakers elicited the worst performances in both tasks, the response to Southern American English relative to Standard American English was equal for the listening in noise task, but worse overall for the false memory task. Similarly, individual factors interacted with the size of dialect differences in both experiments, but not the same ones: most interestingly, participants who self reported being accented did relatively better with both non-standard dialects when listening to a noisy signal, and listeners who had an anti-Asian bias did worse with both non-standard dialects in terms of less accurate phonetic encoding of clear speech. While the results of the two studies are not directly comparable because they involved different population samples, these preliminary results suggest that different factors may be responsible for poor performance with the same dialects across different tasks, and highlight the value of investigating individual differences. [full abstract]

Constraints on cross-talker generalization of foreign-accent adaptation
Kodi Weatherholtz, Linda Liu and T. Florian Jaeger

This study investigated adaptation to foreign-accented speech and the effect of exposure conditions on cross-talker generalization. Using a novel web-based crowd-sourcing paradigm, participants listened to noise-masked sentences produced by Mandarin-accented English speakers (or by native American English speakers in the control condition) and transcribed what they heard. Consistent with previous studies (Bradlow & Bent, 2008), results showed that listeners rapidly adapted to talker-specific accent variability: recognition accuracy for foreign-accented sentences at test was higher following talker-specific training than following training on native-accented speech. This result validates our paradigm for studying accent adaptation. We additionally found reliable talker-specific differences in how a talker’s accent affected participants: some talkers were hard to understand but easy to adapt to, others were easier to understand but provided less benefit from additional exposure. In contrast to previous results, we found robust cross-talker generalization following both single-talker and multi-talker exposure. Thus, exposure to multiple talkers with the same accent is not a necessary condition for listeners to adapt to cross-talker accent variation. In follow-up experiments, we are investigating whether cross-talker generalization following single-talker exposure is based on pair-wise talker similarity. [full abstract]

Evidence for vowel targets in formant distributions and within-syllable adjustments
D. H. Whalen

Variability is an inescapable aspect of language yet languages continue to convey content and indexical information. While some variability is conditioned (e.g. coarticulation), some is presumed to be random. How variability aligns with targets is undetermined. If normally distributed, it indicates a central target. If the distribution becomes flat between two values, the target should be, instead, a region. Recent evidence suggests that the target exists even within a (variable) production, with extreme values tending toward the mean after 50 ms. The present experiment explores vowel targets and variability in a large sample. 500 repetitions of four target CVCs (5 words) (“heed”, “owed/ode”, “geek”, “dote”) are collected across five days. Every subset of 8 words includes one exemplar of the target CVC, along with a random selection of 4 filler items. Thus while the words are spoken repeatedly, they are never spoken in succession. Preliminary analysis of the first session indicates that the distribution is likely to be normal. Changes within a syllable frequently moved toward the mean, but 58% of F1 and 35% of F2 trajectories overshot the mean. The results shed light on the target for vowels and how often online adjustments are made toward that target. [full abstract]

Illusory epenthesis and recoverability-conditioned sensitivity to phonetic detail
James Whang

Japanese speakers systematically devoice or delete high vowels [i, u] between two voiceless consonants. Japanese listeners also report perceiving the same high vowels between consonant clusters even in the absence of a vocalic segment. Although perceptual vowel epenthesis has been described primarily as a phonotactic repair strategy where a phonetically minimal vowel is epenthesized by default, few studies have investigated how high vowel reduction affects the identity of the epenthesized vowel. The present study uses a forced-choice labeling task to test how sensitive Japanese listeners are to coarticulatory cues of high vowels [i, u] and non-high vowel [a] in reducing and non-reducing contexts. To test the effects of recoverability, reducing contexts were further divided into high-predictability contexts where only one of the high vowels is phonotactically legal and low-predictability contexts where both high vowels are allowed. Results reveal a strong tendency towards [u] perception as previous studies have found, but also sensitivity to coarticulatory cues that override the default [u] epenthesis, particularly in low-predictability reducing contexts. Recoverability-conditioned gestural coordination has been proposed for production, and this study provides evidence that perception is conditioned by recoverability as well. [full abstract]

Spectral Trajectories of Spanish /s/: Temporal Variability, Vowel Context, and Duration
Eric Wilbanks

Recent work on the time-varying nature of sibilants in English and Japanese (Iskarous et al., 2011; Reidy, 2015) have expanded our knowledge of the articulatory and acoustic characteristics of these dynamic, turbulent segments. Expanding this analysis to a corpus of naturalistic Spanish data, I examine the temporal properties of Spanish /s/ in inter-vocalic positions, focusing on the role of duration, speaker sex, and vocalic context on the realization of this segment. For all 2,797 tokens in the current data set, nine 30-ms Hamming windows were calculated following Iskarous et al. (2011) and Center of Gravity (COG) measurements were taken at each window. A series of Generalized Additive Mixed Models (Wood, 2011) were fit to the data to model the spectral trajectories. The model results indicate a significant effect of vowel context, with adjacent /i/ conditioning a higher COG. Interestingly, model predictions indicate that, aside from producing /s/ with higher COG than men, women tend to produce /s/ with a greater duration and also produce their peak COG later within the segment. It is unclear whether these differences in peak COG realization arise from vocal tract differences due to sexual dimorphism, sociophonetic pressures, or some combination of factors. [full abstract]

Epenthesis into nonnative consonant clusters: phonetic factors eclipse gradient phonotactics
Colin Wilson and Lisa Davidson

Errors in cross-language speech processing could potentially have several sources, including phonetic decoding, phonological repair, and production planning and articulation. The contribution of these factors to epenthesis into nonnative consonant clusters was investigated with production and transcription experiments. English listeners heard novel words beginning with a range of nonnative clusters, and matched fillers containing schwas, recorded by a native Russian speaker. When participants attempted to produce the nonwords, an intrusive schwa often appeared between the members of the cluster (e.g., bdafa ➝ [bədafa]). The rate of epenthesis was modulated by the voicing of the initial consonant, with more epenthesis after voiced stops, as well as by a number of other phonological and phonetic properties of the clusters. A separate group of participants performed forced-choice transcription of the same stimuli. While the production and transcription results were parallel in many respects, the strong effect of voicing observed in production was essentially nullified in transcription. These results suggest that the voicing asymmetry in production does not reflect greater perceptual similarity of voiced-stop releases to English schwa, or greater phonotactic markedness of nonnative clusters beginning with voiced obstruents, but instead arises from spontaneous voicing when nonnative clusters are produced with insufficient gestural overlap. [full abstract]

The Generation of Prosodic Frames in Speech Production: An Experimental Approach
Hilary Wynne, Linda Wheeldon and Aditi Lahiri

However, theories are not transparent about the prosodic status of compounds: although a noun-noun compound in English consists of two lexical words (and therefore two prosodic words), it can also act as a single prosodic item by exhibiting main stress on the first unit and carrying inflection. Thus the question remains controversial - should these items be treated as a single prosodic unit, similar to a monomorphemic word, or as two distinct units for the purpose of post-lexical representation? Recursive word formation may suggest that compounds are a single unit. Psycholinguistic evidence measuring speech onset latency in native speakers of Dutch and Portuguese also shows compounds being treated as single prosodic units (Wheeldon & Lahiri 1997, 2002; Vigario, 2010). This poster reports the results of two psycholinguistic experiments designed to test the planning of compounds in native British English speakers. [full abstract]

When dynamics conflict: Flap dynamics and palatalization in Japanese
Noriko Yamane, Phil Howson, Masaki Noguchi and Bryan Gick

Previous studies have indicated that there is a cross-linguistic tendency to avoid secondary palatalization and rhotics. It has been suggested that there is an articulatory constraint on the tongue dorsum, which is generally required to be retracted during rhotic production. However, the tap in Japanese does not have any clear dorsum target and has thus been argued to have no dorsum component. Even still, palatalization of taps is avoided in Japanese, despite the fact it has a phonemic tap. This study examines this issue in Japanese through the use of ultrasound technologies. The tap and palatalized tap were compared in the environments i_i, o_o, e_e for 6 native speakers of Japanese. Geminates of the palatalized alveolar nasal, tap and alveolopalatal affricate were also taken to compare when the palatalization gesture was achieved for each class of sounds. The results indicate that the palatalization gesture for the tap is achieved late, compared to the nasal and alveolopalatal affricates. However, when compared to the plain tap, there is a clear tongue body raising which occurs prior to the tap contact. This suggests a dynamic difficulty in the timing of both the tap contact and the palatalization gesture. [full abstract]

The Nature of Variation in the Tone Sandhi Patterns of Shanghai Wu
Hanbo Yan and Jie Zhang

Shanghai Wu has two different tone sandhi patterns, tonal extension and tonal reduction, that can apply variably in a disyllabic sequence. Syntactic structure, semantic transparency, and lexical frequency have all been documented to affect the sandhi application. However, the exact influence of each factor on the variation pattern or how these factors interact was never made explicit. We report a goodness rating experiment for the variant forms with native Shanghai speakers in tandem with semantic transparency and subjective frequency ratings from the same speakers to shed light on the nature of Shanghai tone sandhi variation. The results show that disyllabic tone sandhi application in Shanghai is primarily determined by syntactic structure, with a strong tonal extension preference for modifier-noun compounds and a tonal reduction preference for verb-noun phrases. The effect of semantic transparency is only observed for verb-noun phrases, with semantically less transparent phrases preferring tonal extension more. The lexical frequency effect is only operative insofar as it enhances the prediction of compoundhood: a more frequent modifier-noun compound is more compound-like and hence has a greater tendency to undergo tonal extension, while a more frequent verb-noun item is more phrase-like and hence has a stronger tendency for tonal reduction. [full abstract]

Effects of boundary tones on the coordination of lexical tones
Hao Yi and Sam Tilsen

Recent studies have shown that lexical tone gestures are tightly coordinated with consonantal and vocalic articulatory gestures in Mandarin Chinese, but it is unclear whether prosodic tones influence this coordinative pattern. Conceptualizing both lexical tones and prosodic tones as articulatory gestures, we conducted an Electromagnetic Articulography (EMA) study to assess the influence of intonation on the intra-syllabic coordination between oral articulatory gestures and lexical tone gestures in Mandarin. Specifically, we compared the consonant-vowel-F0 coordinative patterns at phrase-final position and at phrase-medial position. We found that boundary tone gestures were less tightly integrated into the intra-syllabic coordinative network, which seems to either favor the notion that boundary tones, consistent with their post-lexical status, are implemented at a different phonological level than lexical tones, or deny the existence of boundary tones altogether. However, the picture is complicated by the evidence that the intra-syllabic coordinative patterns were altered by the presence of pitch-accent gestures. We argue that pitch accents, while being post-lexical like boundary tones, interact with lexical tones locally. However, the nature of boundary tone coordination is a question left unanswered. [full abstract]

Articulation and Representation of Laterals in Australian-accented English
Jia Ying, Jason Shaw, Catherine Best, Michael Proctor, Donald Derrick and Christopher Carignan

This study investigated the dynamics of /l/ lateralization in Australian-accented English (AusE) using 3D electromagnetic articulography (EMA). Coils were placed both mid-sagitally, on the tongue tip, tongue mid and tongue back and para-sagitally on tongue blade left and right. We varied the vowel preceding /l/ between /ɪ/ and /æ/ and between syllable-initial and -final position of /l/. By comparing /l/ production in the context of these two vowels, we investigate how local variation in tongue shape impacts the time course of lateralization. By comparing /l/ across syllable positions, we investigated how known variation in the timing and magnitude of mid-sagittal movements are related to para-sagittal dynamics. We conducted three analyses of the data: a mid-sagittal analysis designed to replicate past work, a para-sagittal articulatory analysis designed to explore the consequences of mid-sagittal movements for lateral side-branch formation, and a tongue angular analysis designed to provide converging evidence about tongue shape. Our mid-sagittal analysis showed synchronous timing between tongue tip and tongue dorsum gestures in onset position and a timing lag between these gestures in coda position, regardless of preceding vowel. Unlike the mid-sagittal analysis, para-sagittal dynamics were influenced by the preceding vowel in a way that interacted with syllable position. [full abstract]

The emergence of an inflectional edge tone morpheme in Samoan
Kristine Yu

Although intonational tones that occur at phrase edges are typically assumed to mark prosodic constituency, the position of a tone is not diagnostic of its source. Samoan is a non-tonal VSO language where absolutive case on the direct object of a transitive sentence and the subject of an intransitive sentence has been said to be unmarked. However, it was previously observed in intonational fieldwork that a high edge tone (H-) occurs on the final mora of the phonological material preceding absolutive (ABS) arguments. Samoan also has other high edge tones that reliably appear: at the right edge of a fronted argument and the right edge of a coordinated argument preceding a conjunction or disjunction. Our study provides evidence from intonational fieldwork suggesting that the ABS H- is distinct from the other H-’s in how it comes into prosodic organization because it is a tonal morpheme, and that it diachronically emerged from tonal reassociation from stress on the preposition ia. [full abstract]

Planning of inserted /ɹ/ in the speech of Australian English-speaking children
Ivan Yuen, Felicity Cox and Katherine Demuth

In non-rhotic Australian English (AusE), adults commonly separate certain heterosyllabic V1.V2 hiatus contexts with /ɹ/-insertion, but it is not known how children resolve this context. This study therefore investigated whether, like adults, children use /ɹ/-insertion, and if so, whether this hiatus breaking segment is planned, as evidenced by F3 lowering on V1. Thirteen AusE-speaking children (7F, 6M) with a mean age of 6;1 participated in an elicited production task. The stimuli included four test words (linking –r context: door, floor; intrusive –r context: paw, claw) followed by of in a carrier sentence (e.g., ‘This is the paw of a cat.’). After familiarization containing auditory and visual (picture) prompts, the children produced test sentences upon presentation of visual prompts alone. Eight children produced /ɹ/-insertion, whereas the remainder used glottalization. The incidence of /ɹ/ did not vary across linking or intrusive contexts and inserted /ɹ/ exhibited a low F3. Compared to control items (without /ɹ/) , F3 was also lowered at the onset of V1 when /ɹ/ was inserted. To sum, some 6-year-old AusE-speaking children use /ɹ/ insertion and those who do show evidence of planning ahead. Implications for phonological and lexical representations, and the development of speech planning processes are discussed. [full abstract]

Different effects of production on spoken-word recognition for adults versus children
Tania Zamuner, Stephanie Strahm, Elizabeth Morin-Lessard and Mike Page

Speech production may serve a variety of functions, such as allowing for the creation of articulatory representations or the generation of auditory and sensori-motor feedback (Hickok, 2014). In previous work with adults, we explored the impact of production on the recognition of newly learned non-words. Adults’ had faster recognition of non-words that they produced compared with heard-only during training (Zamuner et al., in press). In the current research, we tested whether similar effects are found with children. Previous research has indicated links between production and learning (Schwartz & Leonard, 1982; Vihman et al., 2014); thus, it was predicted that children would show similar effects as adults. However, the opposite pattern was found. Children showed better recognition of non-words that they heard-only rather than produced during training. Although unexpected, the results with children are in-line with previous work with adults that has found that production does not always show beneficial learning effects, and this depends on the linguistic characteristics of the stimuli (Kaushanskaya & Yoo, 2011). The effect of production may also depend on the developmental stage of the learner and the difficulty of the task, as seen in other domains of language development (Curtin et al., 2011). [full abstract]

The role of fundamental and formant frequency information on voice and speaker perception in children with Autism and Attention Deficit Hyperactivity
Georgia Zellou, Santiago Barreda, Nancy McIntyre, Lindsay Swain-Lerro, Matthew Zajic and Peter Mundy

For typical listeners, pitch (f0) and formant frequencies (FFs) of voices are used to infer indexical speaker characteristics (i.e., size and gender). Prior studies report that children with Autism (ASD) and/or Attention deficit hyperactivity disorder (ADHD), highly comorbid with ASD, exhibit linguistic processing abnormalities. However, little is known about how children with ASD/ADHD use acoustic information in determining apparent speaker characteristics. We examine the effect of systematic differences in f0 and FF on voice-differentness and speaker perception in children with ASD and/or ADHD. Children heard word pairs across four voice conditions: same f0/FF levels between voices, f0 difference, FF difference, and both f0/FF difference. Children rated voice similarity and whether the same or different speakers produced the pair. With respect to voice-similarity perception, ASD children display diminished perceptual sensitivity for f0; yet, they show enhanced use of formant frequency (FFs). Meanwhile, ASD and typical children, displayed a tendency to disassociate their perception of voice “differentness” with their assessments of speaker identity. ADHD children did not display this disassociation to such an extent. This study contributes to our understanding of speech perception in children with different cognitive profiles, as well as the developmental trajectory of perceptual representations for words and voices. [full abstract]

An Experimental Investigation of Positionally Conditioned Tone Sandhi in Hailu Hakka
Jie Zhang, Hanbo Yan, Yuwen Lai and Shao-Ren Lyu

Hailu Hakka, a Chinese dialect spoken in Taiwan, has a neutralizing tone sandhi pattern that turns a low-rising 13 into a mid-level 33 in non-phrase-final positions. This paper presents the results of a nonce-probe experiment and an auditory lexical decision experiment that investigated the productivity of this sandhi pattern and the lexical representations of the tone-sandhi words used in spoken word recognition.

Nineteen Hailu speakers participated the nonce-probe experiment, in which they produced disyllabic real words and novel words with 13 as an underlying tone for the first syllable. Growth curve analysis showed that there was no significant f0 difference in the sandhi realization between real and novel words, suggesting that the sandhi applied productively in the novel words. Thirty-two Hailu speakers participated in the lexical decision experiment, in which disyllabic tone-sandhi words were preceded by a monosyllabic prime that matched the first syllable of the disyllable in either the underlying or the surface tone. Results showed that only the underlying-tone primes elicited significant priming compared to control primes, suggesting that the underlying tone was accessed in the spoken word recognition of tone-sandhi words. Collectively, these results indicate that this sandhi pattern involves a computational mechanism in the speakers’ phonology. [full abstract]

Production Effects in Light of Perceptual Evaluation: Tempo Effects for Phonologization
Kenneth de Jong and Kyoko Nagao

The current paper presents a set of interlocking production and perception experiments, 1) a production study of how tempo modulation affects the articulation of the American English plosive laryngeal contrasts in prevocalic and postvocalic contexts, and 2) a perceptual study of how such productions are identified. Productions of syllables were elicited through a motor control task entraining speakers to metronomic pacers inducing extreme tempo variation. Acoustic analyses revealed extreme and pervasively non-proportional variation of VOT and other properties as a function of tempo, and different variation patterns for pre-vocalic and post-vocalic contexts. To determine the effect of these tempo modulations, three-syllable productions were subjected to identification experiments. While listeners were generally readily able to account for the tempo variation, the extremely fast tempi induced many errors. The paper investigates the roots of these errors, and notes that they are in opposite directions for pre-vocalic and post-vocalic stops. Such opposite directional biases mirror the typical effects of syllabic position in phonological systems; onset segments tending toward the voiceless and aspirated variants, and coda segments tending toward voiced and unaspirated variants. Thus, production and perception experiments in an interlocking design seem useful for uncovering what production effects might contribute to phonological systems. [full abstract]

Infants’ use of phonological detail during foreign-accented word recognition
Marieke van Heugten, Dena Krieger, Melissa Paquette-Smith and Elizabeth Johnson

The pronunciation of words differs tremendously across accents. Nonetheless, infants learn to cope with unfamiliar regional accents in the months preceding their second birthday. Less is known about the effects of foreign accents on early speech perception. Using the Headturn Preference Procedure, we here examined when and how infants start recognizing foreign-accented words. English-learning 15-, 18-, and 22-month-olds were presented with lists containing known (bottle, kiss) and nonsense words (bocky, lath) in a French accent. Infants’ preference for known over nonsense words emerged by 18 months, corroborating early reports that the ability to contend with accents becomes robust in the months prior to children’s second birthday. Moreover, the preference for known words held even when the nonsense items were replaced by mispronounced versions of the known words (bittle, koss) in the same French accent. This suggests that children have not simply become tolerant of all phonemic substitutions when listening to accented speech. Infants’ word recognition is thus simultaneously flexible and specific. [full abstract]

I can't understand - The perception of native and non-native can and can’t by native and non-native listeners of English
Margot van Mulken, Huib Kouwenhoven and Mirjam Ernestus

This study investigates how native and non-native listeners with different language backgrounds cope with lexical ambiguity that results from the reduction of a single segment. Four groups of participants differing in language background listened to short phrases produced in American English or Spanish English and indicated whether these included can or can’t. Spanish English reduced can't tokens typically contain less duration information than the American English tokens. American English and Dutch listeners only had problems correctly identifying these Spanish English reduced can’t tokens. This strongly suggests that these listeners base their identification on subsegmental information. The Spanish and the Mandarin listeners, who are not familiar with /t/-reduction in their native languages, often misidentified both the American English and Spanish reduced can’t tokens, which suggests that they did not rely on the American English subsegmental cues distinguishing can from can't. These results also show that listeners who in a non-native language often delete segments to make words comply with the phonotactic constraints of their own native languages do not easily reconstruct these segments while listening. We conclude that it is difficult for non-native listeners to learn to interpret subsegmental information, which is necessary to efficiently understand casual speech. [full abstract]

An acoustic analysis of laryngeal contrasts in Korean stops across three groups of speakers
Ruben van de Vijver and Hae-Eun Cho

We examined the phonetic realization of the three-way laryngeal contrast of phrase-initial stops in the speech of three groups of Korean women. The three groups are (1) older speakers living abroad (2) younger speakers living in Seoul and (3) younger heritage speakers. We found that all speakers realize a low F0 on vowels after plain stops and a high F0 on vowels after aspirated and tense stops. There were differences in the realization of stops across speaker groups. Older native speakers living abroad produced intermediate VOTs for plain stops, long VOTs for aspirated stops and short VOTs for tense stops. Younger native speakers living in Seoul and younger heritage speakers produced long VOTs for plain and aspirated stops and short VOTs for tense stops. Our findings suggest that older native speakers living abroad produce distinct VOTs in order to realize the three-way laryngeal contrast. In contrast, native speakers living in Korea and heritage speakers use VOT and F0: plain stops have a long VOT and a low F0, aspirated stops have a long VOT and a high F0, and tense stops have a short VOT and a high F0; a three-way contrast is being reanalyzed as a double two-way contrast. [full abstract]