Upload
morris-robertson
View
215
Download
0
Embed Size (px)
Citation preview
Background: Speakers use prosody to distinguish between the meanings of ambiguous syntactic structures (Snedeker & Trueswell, 2004). Discourse also has a syntax-like structure (Asher & Lascarides 2003, Kehler 2002), which can be ambiguous. We sought to establish that prosodic variation corresponds systematically to manipulations in discourse structure (also see den Ouden et al., 2009, for correlational data). We then examined whether listeners can identify which interpretation the speaker originally intended.
Research Questions: Can speakers use prosody to communicate the intended interpretation of this ambiguous discourse? Can listeners tell which meaning the speakers intended when uttering the lexically ambiguous discourse?
References:• Asher, N., & Lascarides, A. (2003). Logics of Conversation. xxii+526pp, Cambridge, UK: Cambridge U Press.• Den Ouden, H., Noordman, L., & Terken, J. (2009). Prosodic realizations of global and local structure and rhetorical relations in read aloud news reports. Speech Communication, 51(2), 116-129.• Kehler, A. (2002). Coherence, Reference, and the Theory of Grammar. Stanford, California: Center for the Study of Language and Information.• Snedeker, J., & Trueswell, J. (2003). Using Prosody to Avoid Ambiguity: Effects of Speaker Awareness and Referential Context. Journal of Memory and Language, 48(1), 103-130.
Conclusions:•Speakers can, in these circumstances, produce the intended meaning of an ambiguous discourse with distinctive prosody.•Listeners, upon hearing the speakers’ productions, are unable to systematically retrieve the speaker’s intended meaning, though they are more confident in their answer when the prosody reflects it.
Prod. Procedure: Presented with a three-sentence discourse with two interpretations, alongside a picture representing each interpretation, participants explained how the discourse and pictures were related. Then, one or the other picture flashed and they read the three sentences aloud, verbatim, trying to convey the highlighted interpretation.
Speakers Use Prosody to Communicate the Intended Interpretation of an Ambiguous Discourse
Jason Kahn (UNC Chapel Hill), Jason Kahn, [email protected] Tyler (University of Michigan), [email protected]
Jennifer E. Arnold (UNC Chapel Hill), Jennifer, [email protected]
Questions for Future Research: Would similar production patterns show up in a more natural context? If only the productions that showed the overall production patterns were presented to listeners, would they be able to identify the intended meaning?
I ate some breakfast.I had a bagel.
And I indulged in an ice cream cone.
Discourse Ambiguity 1(2nd sentence subordinated)
2x2 Design (12 participants): Main manipulation (condition): coord vs. subord interpretation; Two types of structure: sentences 2 and 3 subord, or 2 subord and 3 coord
I drove to Maryland.I went to the beach.
And I saw an old friend from college.
Discourse Ambiguity 2(2nd & 3rd sentences
subordinated)
Sentence 1
Pause 1
Sentence 3
Sentence 2
Pause 2
Pause 1
Methodological Contributions:1. Most discourse prosody research has explored large-scale correlations. This study
manipulated discourse structural variation while keeping lexical and sentential content constant.
2. Bringing psychological methods to theoretical discourse research: a) Show whether speakers (and listeners?) actually use the relations specified by theory b) Allow for adjudication among theoretical claims in discourse semantics, e.g. claims about the inventory of discourse relations, the hierarchical organization of discourse, and discourse segmentation.
Pause 2
Durational Measures (Production Data)
Discourse ambiguity type 2
Discourse ambiguity type 1
Theoretical Contributions:1. Extends our understanding of the interface of prosody and
linguistic structure to the level of discourse by demonstrating that prosody can be used to disambiguate discourse structures, not just syntactic structures.
2. Shows that listeners are not able in these circumstances to recover a speaker’s intended meaning of an ambiguous discourse, though their confidence is affected by their accuracy.
Longer pause durations here indicate speakers
put space between discourse segments at
the same level
Shorter pause durations, by contrast, indicate that the segments belong in a hierarchical relationship
Speakers continue to coordinate
segments with short pauses
When they want to move from one subordinated segment to
another, speakers do not pause. When they move from a
subordinated to a coordinated segment (a discourse “pop”),
they pause significantly.
Sentence 1 was significantly shorter in the
subordinated condition
Sentence 2 was also significantly shorter in
the subordinated condition
For the subordination interpretations, Sentence 3 was shorter when also
subordinated (Discourse Ambiguity 2) and longer when coordinated to
Sentence 1 following the subordination of Sentence 2
(Discourse Ambiguity 1, a discourse "pop"). For the coordination
interpretations of all discourses, Sentence 3 was medium length.
(significant interaction)
Comp. Procedure: Presented with a three-sentence discourse with two potential interpretations, alongside a picture representing each interpretation, participants listeners to a production and chose which interpretation they though the speaker intended and how confident they were in their choice.Design (21 participants): Listeners judged a subset of the recordings from the production experiment, two judgments per item (coord + subord)
Comprehension DataListener’s Confidence in their Judgment
Confidence ~ DiscourseType(Centered) + Logduration.Silence1 + Logduration.Silence2(Centered)
+ Logduration.Silence2*DiscourseType + (1|Subject)+(1|Item)
Fixed EffectsEstimate SE t-value pMCMC
Intercept 73.81 3.24 22.8 0.0001DiscourseType(C) 0.73 1.34 0.54 0.59
Logduration.Silence1 -1.21 0.58 -2.09 0.035Logduration.Silence2(C) 2.66 0.59 4.48 0.0001Logduration.Silence2(C)
*DiscourseType(C) -2.86 1.21 -2.36 0.02
Random EffectsVariance SD
Subject 171.06 13.08Item 0.97 0.99
Residual 188 13.71
Statistical Model of Listener Confidence
*
*
*