Eye-tracking reveals the effects of perceptual learning on neighboring phonemes Bridget Smith The Ohio State University
Transcript
Slide 1
Slide 2
Eye-tracking reveals the effects of perceptual learning on
neighboring phonemes Bridget Smith The Ohio State University
Slide 3
Using time-course data to view phonological processes Bridget
Smith The Ohio State University
Slide 4
Background The interaction of speech perception and production
in sound change Recreating sound change in a laboratory Measures
changes in perception via lexical decision, identification tasks
(RT and Accuracy), and eye-tracking paradigm 3
Slide 5
Theory of sound change Phonetically-conditioned sound change
begins with phonetic variation and ends with systematic change
e.g., O.E. kirke, kiken -> M.E. church, chicken What happens in
between? How gradual is it? Do lexical differences exist? Can it be
conditioned by other non-phonetic factors? Can we reproduce it in a
laboratory? 4
Slide 6
Research Question Can we use perceptual learning and shadowing
to reproduce sound change in a laboratory? After participants are
exposed to a pronunciation variant, do they exhibit a change in
perception and production consistent with a sound shift? (If yes,
then there are many interesting questions about sound change to
look at) 5
Slide 7
Perceptual Learning When exposed to a pronunciation variant in
familiar words, listeners incorporate the variant into their mental
representation, temporarily (or sometimes long-term) thus changing
the representation of that sound e.g., Norris, McQueen, &
Cutler 2003 Ambiguous /s-f/ sound replaced segment in words with
/s/ and words with /f/, boundary shift depended on which words.
6
Slide 8
Shadowing/convergence When saying a word after hearing it
pronounced, talkers change their productions to be more similar to
those of the model talker. e.g., Goldinger et al 1998 Measured
similarity using AXB task Later studies measured variables include
VOT, F0, amplitude envelope, mean spectral frequency (center of
gravity) 7
Slide 9
Sub-questions Do participants undergo perceptual learning? Does
this extend to new talkers? new words? Do participants undergo
convergence with the trainers voices while shadowing? Is the change
in pronunciation generalized to familiar words that they did not
hear during training? Does the change in one sound affect other
neighboring sounds? 8
Slide 10
Experiment design Needed a source of variation that was not
known to have any indexical value or be a sound change in progress
Affrication of /tw/ Phonetically natural: stops before approximants
frequently become affricated Historical precedent in English: /tr/
and /tj/ 9
Slide 11
Experiment design Two likely trajectories of change: front or
retracted frication: tsw- or tchw- tchw- common in observed
variation parallels with of /tr/ and /tj/ physiological basis -
rounding gesture for /w/ tsw- also possible, especially with dental
/t/ c.f. OHG /t/ -> /ts/, even in front of /w/, e.g. zwei, or
Japanese /t/ -> /ts/ before // 10
Slide 12
tchw 11
Slide 13
tsw 12
Slide 14
tw 13
Slide 15
tw (not as affricated) 14
Slide 16
Ambient variation or sound change-in- progress? Known TV
personalities who now say things like chwenty and chwitter and
betchween: Rachel Maddow Michael Savage Michael Ian Black 15
Slide 17
Links to videos http://www.youtube.com/watch?v=XS3rR5N5
vAA&t=49s http://www.youtube.com/watch?v=XS3rR5N5 vAA&t=49s
http://youtu.be/D4rq53Ztvbg?t=4m18s 16
Slide 18
Experiment 1 Design 45 /tw-/ initial English words 30
familiar-somewhat familiar 15 highly unfamiliar/archaic
Necessitates different paradigm than traditional perceptual
learning methods (e.g., Norris, McQueen & Cutler 2003). 17
Slide 19
Experiment 1 Design Task 1: pre-training production and
familiarity rating Participants read familiar and unfamiliar words
at self controlled pace subset of training words, plus others: tw-,
en-, vi-, t-, tr-, and str- words Rate familiarity of each 18
Slide 20
Experiment 1 Design Familiarity Ratings: very familiar I know
this word and use it somewhat familiar I know this word, but I may
or may not use it myself neither familiar nor unfamiliar I may know
this word, but do not use it somewhat unfamiliar I may have heard
this word before, but have never used it very unfamiliar I have
never heard or used this word before 19
Slide 21
Experiment 1 Design Task 2: training/shadowing (en- vi- tw-
words): Participants see the word on the screen Hear the word
pronounced by trainers over headphones Say the word out loud after
hearing it Hear the word again Silently read a definition See,
hear, and say the word again Repeat in blocks of definitions and
then sentences with the word in context 20
Slide 22
Experiment 1 Design 21
Slide 23
Experiment Design Hear each word 6 times, shadow 4 times 2
trainers for each word 1 male, 1 female Total 8 trainers 4 male, 4
female 3 conditions: Front tsw- Retracted tchw- Control tw- 22
Slide 24
Experiment Design Task 3 - Lexical decision task: Using button
box, choose whether stimulus is word or non-word 4 new talkers (2
male, 2 female) /tw/ target words: 28 trained, 15 untrained Half
front tsw- variant, half retracted tchw- Non-words with variant,
also vi- and en- words and non-words 23
Slide 25
Experiment Design Task 4 Post-training production Participants
read words off the screen for comparison to before pronunciations
Task 5 Identification task Participants hear a stimulus and select
whether they heard two chew or tsu Tests whether adaptation is
extended to related environment /tu/ 24
Slide 26
Overview Results Experiment 1 Lexical decision showed greater
acceptance of variant that participants were trained on RT varies
greatly by subject and other unknown factors, and cannot be
directly compared Identification task showed generalization of
training variant to /tu/ by boundary shifts Production results show
convergence 25