Maria Spychalska and Cara Oster
Affiliation: Max Planck Institute for Psycholinguistics, Ruhr University Bochum
Category: Linguistics
Keywords: script-knoweldge, temporal implicature, ERPs, EEG, conjuctions
Date: Thursday 4th of September
Time: 18:00
Location: Room 161 (161)
View the full session: Pragmatics
Our descriptions of the past often align with the chronological order in which the events occurred, even though language provides tools for non-chronological descriptions, such as the connectives before and after. For instance, truth-conditionally equivalent sentences “Julia finished her PhD and published her results” and “Julia published her results and finished her PhD” imply a different order of events, that is consistent with the order in which the events are mentioned. In Gricean pragmatics, this phenomenon is explained as a temporal implicature: the implied temporal sequence arises from the Maxim of Manner, which requires speakers to be 'orderly' in their communication (Grice, 1975). Although temporal implicatures arise for any conjoined past-tensed event reports, even if there is no direct link between the events (Levinson, 2000), sentences describing activities from everyday life often depict temporally related events, or context may imply a link between the events, prompting the listeners to construct a temporal frame. In real life, some events normally occur in a certain temporal order, e.g., “She washed her hair and dried it.” The relation between the events may be even causal, which makes the order conceptually irreversible, e.g., “He drank poison and died”. Alternative pragmatic accounts have explained temporal implicatures in terms of complex interactions between syntax, context, script or world knowledge, discourse or narration structure and the semantics of “and” (Kamp & Reyle, 1993; Lascarides & Asher, 1993; Lascarides & Oberlander, 1993; Oberlander & Lascarides, 1992). While evidence from behavioural and neurolinguistic studies shows that readers and listeners prefer sentences and discourses that report past-tense events in the same order in which they happened in real life (Baggio et al., 2008; Bott et al., 2006; Claus & Kelter, 2006; Mandler, 1986; Münte et al., 1998; Nieuwland, 2015; Noveck et al., 2009; Politzer-Ahles et al., 2017; Raisig et al., 2010), it has not yet been investigated how script knowledge interacts with discourse structure in conjunctions. Using EEG, we investigate how script knowledge modulates the processing of temporal order inferences associated with conjunctive past-tense sentences, and whether the scripts can be revised by ad-hoc scenarios. In Experiment 1, we test the hypothesis that the order of events consistent with the script knowledge should be more expected by the processor, resulting in a lower N400 ERPs. This aligns with Nieuwland (2015) who reported an increased N400 in sentences with “before/after” that violated the expected temporal order with respect to world knowledge. However, if temporal order is not integrated at the meaning level, reversing the order of events would lead to structural reprocessing of the script, resulting in an increased P600. We constructed three-event scenarios, e.g., washing, drying, braiding hair, that followed a routine known from everyday life. While it is possible to first dry hair and then wash it, it would be rather unusual. The participants read word-by-word conjunctive sentences reporting two events, either the initial or final script-fragment, in a correct or reversed order (Fig1). They were also asked to judge the naturalness of the sentence order. On the second verb, we observed a late P600-like positivity for the reversed order relative to the correct order sentences; however, only for the script-initial sentences (ba vs ab), supporting the hypothesis that temporal order in conjunctions is processed at the level of correct discourse structure. In Experiment 2, we investigate whether the encoded scripts can be revised based on visual scenarios. The three-event scenarios were presented in form of sequences of illustrations in either the usual (chronological scenario: abc) or reversed order (non-chronological scenario: cba, Fig2.). Afterwards, the participants read a sentence word-by-word, which mentioned either the first two or the last two activities, following the script order (correct) or not (reversed). There was no task related to sentence evaluation, but occasionally the participants were asked: 1) about the non-mentioned activity (to make it natural that only two out of three displayed activities were mentioned in a sentence) 2) whether the sentence mentioned the activity shown in the first/second/third picture. In the context of chronological scenario, on the first verb, we observe an increased N400 for sentences mentioning the last script event (cb) as first relative to all alternative forms, i.e., forms starting from the first event (ab) and those starting from the middle one: (bc) and (ba). On the second verb, we observe both the N400 and P600 for the reversed-order script-initial sentences (ba) relative to all other conditions. Thus, mentioning the first event as last (ba) triggered both the N400 and P600 relative to all other cases (Fig3). In sum, both mentioning the second event in the first-verb position as well as the first event in the second position created a violation of expectancy at the meaning level, reflected in the increased N400. As this effect was not observed in the design without visual scenarios, it appears to be driven by the visual encoding of the scenario’s order in short-term memory. In addition, similar as in Experiment 1, a late positivity is observed for first events reported in the second verb position (ba). By contrast, script-final reversed-order sentences (cb) did not trigger any late positivity, suggesting that the main violation occurs when the first verb refers to the final event (c). Since c is expected last, its early mention signals a disruption immediately, reducing the need for later reanalysis. Embedding the sentences in a non-chronological context (cba) led to attenuation of these effects, meaning that ad-hoc scenario enabled revising the encoded script, modulating the expectations regarding the temporal discourse organization. Only at the second verb, we observed an increased N400 for sentences reporting the first script event (a) as second (ba), even though in the scenario this event was shown as last. Overall, our results show that script knowledge constrains the events’ order in the linguistic description, leading to reanalyses of sentences violating the expected temporal structure. The order is further shown to modulate meaning-related predictive processes but only if it is supported by a visual context, while the script can be revised in a different order is visually encoded.