You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster C46, Thursday, November 9, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

On-line expectation management during discourse comprehension

Geertje van Bergen1, Marlou Rasenberg1,2, Joost Rommers3;1Max Planck Institute for Psycholinguistics, 2Radboud University, 3Donders Institute for Brain, Cognition and Behaviour

INTRODUCTION. Discourse markers are linguistic expressions that relate the sentence in which they occur to the surrounding discourse situation (e.g., Fischer, 2000; Levinson, 1983; Schiffrin, 1987). Discourse markers typically occur in spontaneous conversational interaction; they are overt linguistic indices of the cognitive, expressive and/or social organization of the discourse (e.g., Maschler & Schiffrin, 2015). While DMs have been analyzed theoretically and in terms of their licensing conditions in production, little is known about how they are processed in comprehension. The current study focuses on the Dutch discourse marker ‘eigenlijk’ (≈‘actually’), which expresses an interpersonal relation between conversational partners: it encodes the non-alignment between what the speaker says and what the hearer is assumed to expect on the basis of the (extra-)linguistic context (van Bergen et al., 2011). We investigate to what extent the presence of this high-level, interactional common ground-managing device modulates predictions about incoming lexical input during incremental language comprehension. We concentrate on two event-related potentials (ERPs): (a) the N400, which is sensitive to the semantic predictability of a presented word, irrespective of the constraint of the preceding context (Federmeier et al., 2007; Kutas & Hillyard, 1984); (b) the Anterior Late Positivity, which is enhanced by words that disconfirm likely predictions in constraining contexts (Federmeier et al., 2007; Van Petten & Luka, 2012). METHOD. We performed an EEG experiment in which native Dutch participants read 216 short conversational interactions, consisting of an introductory context sentence (e.g., On a hot Summer day, Mark runs into Hanna on campus.), a question (e.g., Mark asks: “Are you joining us to the beach?”) and an answer (e.g., Hanna says: “I would very much/[eigenlijk] like to take a swim/nap this afternoon.”). We manipulated (a) the contextual constraint of the conversations (as measured by cloze probabilities), (b) whether the critical word (CW) in the answer confirmed or disconfirmed a likely prediction, and (c) whether answers contained ‘eigenlijk’ or a control adverb. RESULTS. Preliminary analyses of the results (N=16) suggest that predictable CWs elicited the smallest N400 amplitude, whereas CWs in weakly constraining conversations yielded the largest N400 amplitude. For predictable CWs, N400 amplitude increased if the answer contained ‘eigenlijk’; for unexpected CWs, we found no difference in N400 amplitude between answers with and without ‘eigenlijk’. Initial analyses furthermore suggest that CWs disconfirming a likely prediction elicited a more positive Anterior Anterior Late Positivity in answers without ‘eigenlijk’ than in answers with ‘eigenlijk’. At the same time, predictable CWs seem to yield a more positive Anterior Late Positivity if the answers contained ‘eigenlijk’ compared to answers without ‘eigenlijk’. CONCLUSION. Our preliminary results suggest that, upon encountering ‘eigenlijk’, comprehenders rapidly alter their predictions about incoming words: they start anticipating unpredictable input, which reduces semantic facilitation of incoming predictable words, and possibly reduces processing costs incurred by disconfirmed predictions. These findings show that high-level, interactional pragmatic cues have immediate effects on lexical predictions, and that expectation-managing particles provide a useful means to investigate the nature of discourse-based expectations in language processing.

Topic Area: Meaning: Discourse and Pragmatics

Back to Poster Schedule