Sub- & Cross-Phonemic Priming in Vowel Shadowing 1. Memory - - PowerPoint PPT Presentation

• 1. Memory Types and Respresentation of Language
• 2. Experiment-Setup
• 3. Results

1. Sub-Phonemic 2. Cross-Phonemic

• 4. Explanation

1. Exemplar Model 2. Sub-Phonemic 1. Early-Articulation 2. Planning-Interaction 3. Cross-Phonemic

Sub- & Cross-Phonemic Priming in Vowel Shadowing

• 1. Memory Types and Respresentation of Language
• 2. Experiment-Setup
• 3. Results

1. Sub-Phonemic 2. Cross-Phonemic

• 4. Explanation

1. Exemplar Model 2. Sub-Phonemic 1. Early-Articulation 2. Planning-Interaction 3. Cross-Phonemic

Sub- & Cross-Phonemic Priming in Vowel Shadowing

• brain remembers phonetic detail of utterances
• store seperate speech events

earlier:

• brain abstracts away from speaker/utterance-specific details and
• store abstract lexical items

“tree” [t̪ɹiː] “leave” [l ̴iːv]

today:

Memory

[tʰʀi] [t̪ɹiː] [tʀi] [tʰri] [t̪riː] [tɹiː] “tree” [l ̴iːv] [liːv] [liːf] [l ̴iv] “leave”

Perceptual Memory:

• long-term memory for visual, auditory and other perceptual information

Motor Memory:

• overt conscious appreciation of memory not necessary
• tasks done automatically without paying attention

Memory

Question: when and how are perceptual representations mapped to motor representations?

[tʰʀi] [t̪ɹiː] [tʀi] [tʰri] [t̪riː] [tɹiː] “tree” [tʀi] [t̪ɹiː] vocal tract shaping vocal tract shaping “tree” [t̪ɹiː] recent perceptions: not influence production recent perceptions: may influence production

⇒ ⇒

repr repr

• 1. Memory Types and Respresentation of Language
• 2. Experiment-Setup
• 3. Results

1. Sub-Phonemic 2. Cross-Phonemic

• 4. Explanation

1. Exemplar Model 2. Sub-Phonemic 1. Early-Articulation 2. Planning-Interaction 3. Cross-Phonemic

Sub- & Cross-Phonemic Priming in Vowel Shadowing

Basic Idea

to be examined:

• how do episodic memories (exemplars) influence speech production

Question:

• influence of recent perception of subphonemic details on subsequent production

[a]* [a]*

produce [a] Do people produce more centralized /a/ when hearing centralized [a]* prime than when hearing a normal [a] prime?

?

/ do they play a r

Experiment Set-Up

General Setup participants: American task: repeat last vowel as quick as possible cue: target: stimuli: beep, [a], [i], [a]*, [i]* beep, [a], [i] (12 participants: 6m, 6f) trials: 25-40 blocks, each 32 trials concordant: discordant: control: cue = target cue ≠ target no cue subphonemic differences between [a]/[i] and [a]*/[i]* condition: within-subject

white noise cue target reponse

• 1. Memory Types and Respresentation of Language
• 2. Experiment-Setup
• 3. Results

1. Sub-Phonemic 2. Cross-Phonemic

• 4. Explanation

1. Exemplar Model 2. Sub-Phonemic 1. Early-Articulation 2. Planning-Interaction 3. Cross-Phonemic

Sub- & Cross-Phonemic Priming in Vowel Shadowing

Centralization for [a]: F2 higher + F1 lower Centralization for [i]: F2 lower + F1 higher

Results - Subphonemic Effects

• significant mean F1, F2 differences between normal and centralized

concordant trials:

• most: trend for centralization
• some: trend against centralization

Results - Subphonemic Effects

Centralization for [a]: F2 higher + F1 lower Centralization for [i]: F2 lower + F1 higher

• significant mean F1, F2 differences between normal and centralized

concordant trials:

• most: trend for centralization
• some: trend against centralization

Results - Subphonemic Effects

Centralization for [a]: F2 higher + F1 lower Centralization for [i]: F2 lower + F1 higher

• most: trend for centralization
• some: trend against centralization
• significant mean F1, F2 differences between normal and centralized

different patterns: control similar to normal control notacibly distinct from normal control between normal and centralized

concordant trials:

[a]

[a]-F1

• most: negligable differences in F1
• some: rel. lower F1 in first half (→ centralization)
• some: rel. higher F1 in some parts

[a]-F2

• most: higher contours (→ centralization)

Centralization for [a]: F2 higher + F1 lower — normal

central

—

different patterns: begin and remain central begin central, eventually converge begin near, then diverge (and reconverge)

Results - Subphonemic Effects

[i]

[i]-F1

• most: clear centralization, esp. first half

[i]-F2

• most: centralization
• some: more peripheral

Centralization for [i]: F2 lower + F1 higher — normal

central

—

different patterns: begin and remain central begin central, eventually converge begin near, then diverge (and reconverge)

Results - Subphonemic Effects

Results - Crossphonemic Effects

Dissimilation for [a]: F2 lower + F1 higher Dissimilation for [i]: F2 higher + F1 lower

• dissimilation

[a] cue + [a] → [a] (nearer to i) [i] cue + [a] → [a] (further from i)

• significant mean F1, F2 differences between concordant and discordant

concordant vs. discordant:

Results - Crossphonemic Effects

Dissimilation for [a]: F2 lower + F1 higher Dissimilation for [i]: F2 higher + F1 lower different patterns: control similar to concordant control similar to discordant control between concordant and discordant

• dissimilation

[a] cue + [a] → [a] (nearer to i) [i] cue + [a] → [a] (further from i)

• significant mean F1, F2 differences between concordant and discordant

• separation-conver
• pr

initially similar

[a]-F1

• most: negligable differences in F1
• some: rel. higher F1 (→ dissimilation)
• some: rel. lower F1 in first half (→ assimilation)

[a]-F2

• most: lower contours (→ dissimilation)

Dissimilation for [a]: F2 lower + F1 higher

different patterns: begin and remain dissimilar begin dissimilar, eventually converge begin near, then diverge (and reconverge)

[a] — concordant

discordant

—

Results - Crossphonemic Effects

Dissimilation for [i]: F2 higher + F1 lower

different patterns: begin and remain dissimilar begin dissimilar, eventually converge begin near, then diverge (and reconverge)

[i] — concordant

discordant

—

[i]-F1

• most: clear dissimilation, esp. first half

[i]-F2

• most: dissimilation
• some: assimilation

Results - Crossphonemic Effects

• 1. Memory Types and Respresentation of Language
• 2. Experiment-Setup
• 3. Results

1. Sub-Phonemic 2. Cross-Phonemic

• 4. Explanation

1. Exemplar Model 2. Sub-Phonemic 1. Early-Articulation 2. Planning-Interaction 3. Cross-Phonemic

Sub- & Cross-Phonemic Priming in Vowel Shadowing

Explanation - Exemplar Theory

exemplar theory

• individual speech utterances are stored in the mind as seperate examples
• exemplars are activated during production and perception

→ lexical representation: during perception: identification of words: relative activation level in candidate distribution

• distributions of remembered exemplars
• updated by experiences of individual percepts

exemplar:

• set of associations between auditory properties and category labels
• linguistic information
• non-linguistic contextual information

[ˈdønɐ] [dœˈnʌː]

• word = “döner"
• speaker:
• young
• male
• context: school yard
• word = “döner"
• speaker:
• female
• journalist
• context: Tagesschau

[ˈdoʊnər]

• word = “döner"
• speaker:
• female
• American
• episodic memory of a speech event

activation based on similarity during production: activation of category

Explanation - Exemplar Theory - Perception

• different base level of activation
• recency of use
• frequency of use

→ decay over time classification of new item:

• activation according to similarity to new item
• activation of category = total activation of exemplars in the category

↳ determines whether item belongs to category →

acustic signal

[i] token [a] token [ə] token

[ə] [i] [a] [ə] [i] [a] classification: [ə]

→ →

acustic signal [ə] [i] [a] [ə] [i] [a] classification: [a]

→

[ə] [i] [a]

→

recent input

effect of recent inputs:

determination of production target:

• choose phonological category to produce
• randomly select exemplar from the category
• recency of use
• activation-weighted averageing:

averaging of phonetic characteristics of nearby group of exemplars

→

→

producation target: [a]

[i] token [a] token [ə] token

[ə] [i] [a] [ə] [i] [a] [a] [ə] [i] production:

→

activation-weighted averageing

→

Explanation - Exemplar Theory - Production

↳ more recent: more influence

prior to cue:

• both categories activated to equal extend
• both targets are planned

after cue perceived and stored: after target stimulus perceived: articulation and target planning potentially in parallel → effect in mid production

[i] [a]

→

cue signal [i] [a] [i] [a]

planned targets response

→

target signal

1 2 1 2 2 3 1 3

Explanation - Exemplar Theory in Prime Shadowing

[a]* [a]

• activations of cue due to likelyhood
• target replanned
• corresponding vowel: fully activated
• new production target planned and executed

2 accounts planning interaction account early articulation account difference: when and where shifts happen cause of differences: planning of speech targets cause of differences: pre-configuration of vocal tract

[a] [i]

2 3 1 3

Explanation - Subphonemic Effects

can happen in parallel

• priming effects due to changes in the vocal tract configuration, after cue stimulus
• pre-shapeing of vocal tract
• can reflect subphonemic details
• then influence sub-sequent production

Explanation - Early Articulation Account

• differences between normal and central cue due to:
• pre-configuration of vocal tract
• incorporation of subphonemic details of cue

→

cue signal

→

target signal [i] [a]

1 2 1 2

[a] [i]

2 3 1 3

[i] [a]

response

[a]*

→ →

pediction:

• differences in vocal tract geometry are maintained through (part of) response

decrease stay constant → differs most at beginning of response pattern

Explanation - Early Articulation Account

⇒ cannot be only explanation ↳ F-differences may: increase more complex patterns ↳ F-differences may not:

diverge later on

[i] [a]

1 2 1 2

[a] [i]

2 3 1 3

[i] [a]

• priming effects due to changes in the planning of speech targets

Explanation - Planning Interaction Account

• differences between normal and central cue due to:
• relatively high activation of cue (during planning)
• cue contributes disproportionally to target planning bc of recency
• vocal tract does not vary before perception of target stimulus
• only speech targets themself have been altered

→

cue signal

→

target signal

response

→ →

planned targets

pediction:

• phonetic values conceptualized as vectors

specify values in an accoustic dimension over time

• exemplar targets are actually trajectories

Explanation - Planning Interaction Account

t

• acc. dim.

[a] [a] [i] [a]

1 2 1 2

[i]

1 3 2 3

[i]

Explanation - CrossPhonemic

→

cue signal

→

target signal

response deactivation

cause 1: target planning cause 2: articulation → →

• diminished on concordant trials
• increased on discordant trials

articulator movement amplitude and velocity:

• diminished for exemplars similar to the cue (on dicordant trials)

activation:

→ subphonemic differences in the cue: → - perceived → - influence planning of targets → - influence subphonemic details of articulation ⇒ episodic memory does play significant role in speech planning and production → cross-phonemic priming: → - dissimilatory effects

Conclusion

• how does sub-phonemic memory interact with cross-phonemic priming?
• collect articulatory data to find out whether preshaping happens

Future

→ cause: →- early articulation →- planning of targets → can happen in parallel

• why are the internal dynamics of the productions so complex and speaker dependent?