a processing-based account of acoustic reduction (or: reduction comes from facilitation of levels of...
Post on 19-Dec-2015
215 views
TRANSCRIPT
A Processing-based Account of Acoustic Reduction
(or: Reduction Comes From Facilitation of Levels of Language Production)
Jason M. Kahn & Jennifer E. ArnoldUNC-CH
ETAP Montreal!
Thanks
• Molly Bergeson, Andrés Buxó, Kellen Carpenter, Sam Handel, Leighanne Mcgill, Kayla Finch, Alyssa Ventimiglia, Liz Wagner for help with experiments
• The PLUG group at UNC for valuable critical commentary, as well as Scott Fraundorf, Florian Jaeger, Tuan Lam, and Joseph Tyler
Introduction
Body
Conclusion
Teacher: “Today we’re learning about the structure of a paper. What do these elements do?”
Zzz…
Students (in unison!): “The introduction lays out the problem, the body presents evidence, and the conclusion gives the take home message.”
Scenario 1
Introduction
Body
Conclusion
Teacher: “Today we’re talking about the introduction, the body, and the conclusion – what do they do?”
Zzz…
Students (in unison!): “The introduction lays out the problem, the body presents evidence, and the conclusion gives the take home message.”
Scenario 2
Repeated Mentions Get Reduced (e.g. Bard et al., 2000; Fowler & Housum, 1987)
Teacher:“Structure… elements…”
Linguistically Given Discourse Status
+givenness +predictability
Teacher:“introduction … body …
conclusion …”
Students:“introduction … body …
conclusion…”
Linguistically New Discourse Status
-givenness -predictability
Students:“introduction … body …
conclusion…”
Arnold (1998)
General Questions
• What mechanism drives speakers to reduce words in certain contexts?– Facilitation of multiple levels of processing (either the
representations themselves or the algorithms that operate on them)
• Does reduction occur with the listener in mind, or with respect to only the speaker’s internal state?– Probably a mix of both, but I’ll provide evidence of the
latter
Two Classes of Explanation
Discourse status– Discourse Status – defined as
the relative accessibility or givenness of a referent (Ariel, 1990; Gundel et al., 1993)
– Typically conceived as shared information (Clark & Haviland, 1977)
Speakers reduce when they can rely on common discourse status
Facilitated Processing– Hearing or reading words
activates representations associated with language processing (e.g. lemmas, phonemes)
Speakers reduce for themselves
ORSpeakers reduce for their
listener
Joint Discourse Status
CONCEPTUALIZATION STAGE
DISCOURSE STATUS(given vs. new)
(what speaker and listener both know)
FORMULATION STAGE
ARTICULATION STAGE
Adapted from Levelt (1989), Schmitt, Meyer & Levelt (1999), and van der Meulen, Meyer, & Levelt (2001)
CONCEPTUALIZATION STAGE
DISCOURSE STATUS(given vs. new)
(what speaker and listener both know)
“introduction”
Fowler & Housum, 1987; Prince 1992
Joint Discourse Status
Introduction
Body
Conclusion
Reduction!
CONCEPTUALIZATION STAGE
FORMULATION STAGE
ARTICULATION STAGE
Facilitation-based
Adapted from Levelt, 1989; c.f. Balota, Boland & Shields, 1989; Bard et al., 2000; Bell et al., 2009
CONCEPTUALIZATION STAGE
FORMULATION STAGE
“introduction”
Introduction
Body
Conclusion
Reduction!More Reduction!
Facilitation-based
Linguistic vs. Non-linguistic Givenness
“The accordion…”
Bard & Anderson, 1990; Clark & Marshall, 1981; Prince, 1992
CONCEPTUALIZATION STAGE
FORMULATION STAGE
CONCEPTUALIZATION STAGE
DISCOURSE STATUS(given vs. new)
(what speaker and listener both know)
FORMULATION STAGE
Joint Discourse Facilitation-based
Instruction-giving Task
Speaker
Approximately 12 feet
Listener
Experimental Paradigm
Speaker: “The accordion rotates right”
Speaker: “The toothbrush shrinks”
Speaker: “The belt expands”
Experiment 1: Priming Information
“The toothbrush;The belt;
The accordion”
Control Non-linguistic Linguistic
Joint Discourse Status predicts….
Facilitation-based predicts…
Predictions
Reduced Duration of the Object Word
Linguistic
Non-linguistic
Control
360 370 380 390 400 410 420
Linguistic < Non-linguistic < Control
*
*
Facilitation-based Account Explains the Results Naturally
Non-linguistic information led to reduction
Linguistic information led to more reduction
This task gave the priming information to both the speaker and the listener simultaneously…
Will Speakers Reduce For Their Listener?
• Discourse status says yes – but only when they share information
• Facilitate-for-the-listener says yes – whenever the listener has relevant information
• Facilitate-for-the-speaker says no – speakers will reduce whenever they have information
Instruction-giving Task
Speaker
Listener
1) Blocked trials
2) Icon at the top of the screen
3) Headphones
v v
v v
Both
Speaker
Listener
None
405 410 415 420 425 430 435 440 445 450 455
Reduced Object Word Duration
(Both, Speaker) < (Listener, None)
*X
X
Facilitation Once Again Provides a Natural Explanation
• Speakers reduced words when, and only when, they had relevant information
• This is contrary to a strong audience design account
• For evidence of listener attention on speaker’s acoustic decisions, see Elise Rosa’s talk tomorrow
• For evidence of listener-driven speaker attention on speaker’s acoustic decisions, see Jennifer Arnold’s talk tomorrow
Facilitation at Multiple Levels
• Experiments 1 and 2 provide support for a facilitation-based account, where facilitated levels lead to reduction
• It could be that facilitation might matter only at early stages of production, or it could be that facilitation at any level creates reduction
• Experiment 3 will try to prime a different level of representation – the articulatory level
c
c
Name these objects aloud
c
Name these objects silently to yourself
Or…
c
Then…
c
c
Or…
Spoken Aloud Cond. Silent Naming Cond.
Congruent Cond. Incongruent Cond.
Spoken Aloud Spoken Internally
Congruent Prime Facilitated articulation Unfacilitated articulationIncongruent Prime Facilitated articulators? Unfacilitated representations
CONCEPTUALIZATION STAGE
FORMULATION STAGE
ARTICULATION STAGE
Spoken Aloud Spoken Internally
Congruent Prime Facilitated articulation Unfacilitated articulationIncongruent Prime Facilitated articulators? Unfacilitated representations
CONCEPTUALIZATION STAGE
FORMULATION STAGE
ARTICULATION STAGE
Congruent Prime/Target
Incongruent Prime/Target
410 420 430 440 450 460 470
Spoken Internally
Spoken Aloud
*
*
*Interaction
Reduced Object Word Duration
Congruent+Aloud < (Congruent, Aloud) < Incongruent + Internal
Facilitation of Articulation Creates Reduction
• Speakers reduced after speaking the target aloud, relative to saying it internally
• Speakers also reduced after simply speaking aloud (even to incongruent targets)
Speaker-internal Facilitation Explains It All
• Speakers reduce more for linguistic than non-linguistic givenness• Speakers reduce when, and only when, they have priming
information• Speakers reduce more after saying the word aloud than saying it
to themselves
• This implies that we don’t need a discourse representation to account for these results
• It also implies that at least some acoustic reduction is entirely speaker-driven
• This speaker-driven reduction is plausibly explained by a multiple-levels-of-facilitation account