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

Poster D49, Thursday, November 9, 6:15 – 7:30 pm, Harborview and Loch Raven Ballrooms

The fate of the unexpected: Downstream repetition effects for prediction violations

Melinh K. Lai1, Kara D. Federmeier1;1University of Illinois, Champaign, United States

Amid increasing interest in the nature and role of prediction in language comprehension, there remains a gap in our understanding of what happens when predictions are disconfirmed. Recent findings have shown that, even when disconfirmed, predictions still linger in memory (Joost & Federmeier, in prep), raising questions about what the impact of these lingering predictions might be for the processing of the unexpected word that was actually encountered. One possibility is these lingering representations interfere with those of the unexpected words. Alternatively, it is also possible that violating predictions strengthens the representations of unexpected words – e.g., by drawing attention and/or making them more distinctive. Here, we explore the consequences of prediction violations using the ERP repetition effect. If violating a prediction results in impaired processing of the unexpected word, then this should decrease repetition effects on the N400 and subsequent late positive complex (LPC); in contrast, enhanced processing should increase repetition effects. Thirty participants read 282 sentences for comprehension. The experimental sentences were weakly constraining (“Yesterday evening I saw the SMOKE”, cloze probability 0.01). In the Seen in Strong Constraint condition, the critical word had previously been presented in a sentence that strongly constrained for a different word (“The ship disappeared into the thick SMOKE”, cloze 0.01; expected word FOG, cloze 0.85). In the Seen in Weak Constraint condition, the critical word had previously been presented in a weakly constraining sentence (“Rachel was uncomfortable with all the SMOKE”, cloze 0.01). In the Not Previously Seen condition the critical word had not previously been presented. The lag between initial and repeated presentation was three sentences. Fillers ensured that 70% of the final words did not constitute a repetition. Repeated words elicited a reduced (more positive) N400 and enhanced (more positive) LPC. N400 reductions for repeated words have been associated with a superficial sense of familiarity, while LPC enhancements have been related to episodic retrieval of previously seen words. Critically, there was no effect of initial sentence constraint on the size of the repetition effect in either time window. Thus, words that were initially encountered as prediction violations seemed to be processed similarly to words that were simply repeated. Our results suggest that prediction violations are not particularly disruptive to word encoding, even in the face of lingering activation for the predicted word. At the same time, contrary to accounts that posit a role for prediction violations in learning, we also did not observe a memory benefit for prediction violations, at least when those words were encountered again after a short delay. Thus, prediction violations may be neither as costly – nor as critical – for language comprehension as has sometimes been assumed.

Topic Area: Meaning: Lexical Semantics

Back to Poster Schedule