Kelsey Kraus • Department of Linguistics English Mirative Strategies [email protected] How intonation and discourse particles combine . Research Questions: i. Can the contribution of discourse particles and intonation be unified by reference to speaker expectation? ii. How are the meanings of discourse particle and prosodic contours composed? 1. A starting point 1.2 Sentence-Level Prosodic Contours Discourse particles and intonation Claim: Contours contributes not-at-issue commentary on how the Discourse particles signal a change in a speaker’s knowledge state, and speaker feels the discourse should be navigated: ➢ help participants navigate a discourse (Schiffrin, 1987). They can, for Neutral, falling H*L%: high expectation that the contribution is true ➢ example, mark acknowledgment, confusion, or signal a change in topic. Excited, High-falling H*L%: high expectation that the contribution is true with added pragmatic effect of excitement or surprise The performance of a discourse particle bears heavily on its interpretation. ➢ Surprise-Redundancy (SRC), falling (H-)L* H*L% (Ladd, 1980): speaker is We ask: How does the pragmatic interpretation of discourse particles surprised/believes the listener “should have known” a salient interact with intonation? proposition 1.1 Restricting the Space 2. Mirative Strategies Three discourse particles: Miratives grammatically encode a participant’s epistemic state at the ➢ oh — signals acknowledgement, speaker undergoes a ‘’change-of- time of utterance (DeLancey 1997, Aikenvahld 2012). Their functions include: state’’ (Aijmer, 2002, Gunlogson 2008) ➢ commenting on the common ground status of a proposition ➢ huh — speaker experiences something unexpected ➢ making clear a speaker’s surprise or (violated) expectations ➢ indicating to a listener how the speaker has integrated (or plans to ➢ what — wh-word with dual use, registering expectation violation

integrate) a proposition into their belief set Overlapping distributions: (1) A: The server is down. Oh, huh and what can all convey shades of violated expectations: B: Oh. / #Huh. / #What. No it isn’t. (4) A: I saw your parents yesterday. B: Oh. / Huh. / #What. + Neutral contour = accept, didn’t expect it (2) Looks outside to see that it has started to rain. Says to herself: Oh! / Huh! / What! + Excited contour = accept, high surprise A: #Oh!? / Huh!? / What!? It’s raining!? Oh? / Huh? / What? + SRC = do not accept, needs more info (3) A: Julie just got a puppy. Claim: oh, huh, what and the SRC are all mirative strategies in English, B: Oh! / Huh! / #What! She must be so happy! and contribute not-at-issue commentary on a discourse

An Experimental Investigation:. Impressionistically, differences in discourse particle usage or prosodic tune can be the difference between a discourse-incoherent utterance and a felicitous sentence. The experiment presented below attempts to outline a methodology for testing prosodic judgements, as well as test the felicity of three discourse particles in English in varying prosodic environments.

3. Experiment I 3.1 Results, cont.

Stimuli and Design Emotions categorizations

Question: Can listeners reliably perceive distinct speaker

propositional attitudes from particle and contour combinations?

3 x 3 x 3 design, crossing factors: PARTICLE X PARTICLE CONTOUR X SENTENCE CONTOUR [OH, HUH, WHAT] [NEUTRAL, EXCITED, SRC] [NEUTRAL, EXCITED, SRC] A sample item set with oh, presented only auditorily*: a. {Oh./Oh!/Oh!?} Another pile of papers to grade. [NEUTRAL] b. {Oh./Oh!/Oh!?} A package arrived for you! [EXCITED] c. {Oh./Oh!/Oh!?} You don’t know how to tango? [SRC] ➢670 English native speakers, recruited via Mechanical Turk. Each heard 3.3 Conclusions 2 non-identical clips, and answered 4 questions per recording, ✦Participants identify the three prosodic contours as distinct. including how natural the clips sounded, and the perceived emotion of ✦Lack of an interaction between PARTICLE and PARTICLE CONTOUR the speaker. suggests that these effects may be additive. ➢ Stimuli were produced by a single speaker & controlled for similar ✦SRC and excited contours shift the likelihood for the emotion rating acoustic properties (pitch, height, contour) toward a Surprise categorization. *Punctuation here is shorthand for contours. ‘.’ = Neutral, ‘!’ = Excited, ‘!?’ = SRC 3.1 Results 4. Future Work Naturalness Ratings Experiments In Progress ✦Two-way ANOVA conducted on the influence of PARTICLE and Experiment II: Controlling for Context PARTICLE CONTOUR on the naturalness ratings showed: The pilot study looked at the effects these contours had in neutral a. Significance within PARTICLE; planned comparisons show all contexts; this study presents the same contours and particles, but controls for cases of direct and indirect contradiction. three particles have distinct usages when paired with different Hypothesis: huh will be rated lower than oh when paired with a contours (oh vs what:p=0.02*, oh vs huh:p<0.001***, huh vs what:p<0.001***) NEUTRAL or EXCITED contours. b. Significance within PARTICLE CONTOUR; comparisons show Experiment III: Pulling apart Surprise & Redundancy NEUTRAL and SRC are rated more natural than EXCITED (p<0.001***), This experiment tests the hypothesis that pitch distinguishes the no significance between NEUTRAL and SRC. pragmatic effects of the SRC: higher pitch (relative to a single ✦T-tests show the combinations oh+NEUTRAL and huh+NEUTRAL to speaker) yields an effect of surprise, lower pitch leads a listener to rate more natural than what+NEUTRAL (p=0.03*,p<0.001*** respectively) interpret the redundancy or ‘’should have known’’ reading. ✦No interaction effect: particle+contour meanings are additive. References: Aijmer, K. 2002. English discourse particles: evidence from a corpus. Aikenvahld, A. 2012. The essence of SUMMARY/UPSHOT: Variations in naturalness ratings confirm impression- Mirativity. Linguistic Typology. Delancey, S. 1997. Mirativity: the grammatical marking of unexpected information. istic judgements: particle+contour combinations are sensitive to their Linguistic Typology. Gunlogson, C. 2008. A question of commitment. Commitment: Belgian Journal of Linguistics. Ladd, R. 1980. The structure of intonational meaning: Evidence from English. Sag, I. & M. Lieberman. 1975. The intonational prosodic environment. disambiguation of indirect speech acts. CLS. Schiffrin, D. 1987. Discourse Markers.

Historical Linguistics Summer School • Georg-August-Universität Göttingen • 24 July—5 August, 2017 Thanks to Pranav Anand, Donka Farkas, Jean E. Fox Tree, Grant McGuire, Paul Willis, Steven Foley, Deniz Rudin This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Grant No. DGE 1339067. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the NSF.