Poster D44, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB

Negation, prediction and truth-value judgments: evidence from ERPs

Maria Spychalska1, Viviana Haase2, Jarmo Kontinen2, Markus Werning2;1University of Cologne, 2Ruhr University Bochum

Prediction has been shown beneficial in various cognitive domains and argued to play a key role in linguistic comprehension [1,2]. Predicting sentence continuation as it unfolds seems straightforward in the case of affirmative sentences, but it is less clear how negation modulates prediction. First, it is generally more difficult to predict something that is absent. Second, the time-course of integrating negation into the semantic representation of a sentence is still not well-understood. In our experiments we investigated how negation affects predictive processing. To this aim we compared the modulation of the N400 component in true and false, affirmative and negative sentences, such as “Julia has chosen/not chosen the plum”. The N400 was time-locked to the presentation of content nouns (e.g. “plum”), whose cloze probability in a sentence was a function of the presented context scenario, i.e. of whether the scenario provided alternative referents that could be mentioned in the sentence. The scenarios contained three cards depicting three different objects, one or two of which were then selected (framed green) or rejected (framed red) by a virtual player. After the cards were marked, an affirmative or negative sentence (in German) was presented phrase-by-phrase: “X hat (nicht) den/die/das Y ausgewählt” (“X has (not) chosen Y”). In the first round of the experiments, all stimuli sentences provided a true description of the scenario (fillers of the opposite truth-value were added), while in the second round, all stimuli sentences gave a false description of the scenario. In the unique conditions, there was only one object that could complete a true (respectively false) sentence, i.e. one selected object in the affirmative condition or one rejected in the negative condition. In the multiple conditions, there were two alternative objects that could complete a true (respectively false) sentence. It is well-established that the N400 is inversely correlated with the cloze probability of the triggering word. Our model assumes that the processor has to split the probability weights between the number of alternative objects available in the scenarios, leading to lower probability weights for objects that are presented with alternatives, and furthermore that the expected probability is higher for those objects that make the sentence true (truth-makers) rather than false (false-makers). In line with these assumptions, we observed larger N400 ERPs for words of lower probability weights (multiple conditions) than for words of higher probability weights (unique conditions), both for affirmative and negative sentences. This effect indicates that the processing of the unique referent is facilitated relative to the case where multiple referents are available in the scenario. Additionally, we observed a sustained positivity effect for the negative compared to affirmative conditions, which indicates possible reanalysis mechanisms related to the processing of negation. The effects were similar for true and false sentences: for true sentence the presence of multiple truth-makers increases the N400 amplitude, for false sentences it is the presence of multiple false-makers that increases the N400 amplitude. [1] Kuperberg and Jaeger. Language, Cognition, and Neuroscience, 2015. [2] Van Petten and Luka. International Journal of Psychophysiology, 2012.

Topic Area: Meaning: Discourse and Pragmatics