The brain response to inferential meanings: a detour through different figures of speech
In this talk I will try to provide a coherent picture of the main processing differences observed for non-literal language through ERP studies. Following the idea that N400 effects are linked to lexical-semantic processing while P600 effects are mainly linked to inference and interpretation, we can account for some of the differences observed in a series of studies on figurative language. Lexical-semantic mechanisms are switched off to some extent (reflected in N400 reduction) for idiomatic expression. Such mechanisms are instead strongly involved during metaphor processing, supporting hypotheses that assume a fundamental role of the literal meaning of the expression as an anchor for the ad hoc conceptual representation of the metaphor. This seems to be especially the case when metaphors occur in minimal sentence context or when they are particularly difficult, as in the case of literary metaphor. By contrast, when a sufficiently supportive context is provided, lexical-semantic effort is reduced and only inferential processes may be involved reflected by the P600 component. Interestingly, irony is associated with effects on the LPC component rather than the N400, suggesting that irony comprehension heavily hinges on inferential processes. Also humor strongly relies on late positivities, crucial for the re-interpretation and resolution of the joke incongruity.
In sum, different modulations of the LPC/P600 component are ubiquitous across all kinds of figurative language and strongly point to the common role of later inferential processes, which are needed to adjust or revise the mental representation of the discourse context and grasp the intended meaning.
Neurocomputational Semantics and Pragmatics
In this talk, I will outline how a neurocomputational model of the electrophysiology of language comprehension can inform formal semantic and pragmatic theories, and vice versa. We will start from an explicit neurocomputational model of the N400 and the P600 in language processing, and show that it quantitatively accounts for a wide range of signature processing phenomena (i.e., semantic anomaly and expectancy, syntactic violations, garden-paths, and reversal anomalies). On an incremental, word-by-word basis, this model constructs rich utterance meaning representations in terms of situation models. We will decompose these representations and show that they are neurally plausible (can be implemented in neural hardware), expressive (capture negation, quantification, and modality), compositional (capture complex propositional meaning as the sum of its parts), graded (are probabilistic in nature), and inferential (allow for inferences beyond literal propositional content). Crucially, differences in semantic encoding within these representations directly lead to differential effects on the N400 and P600 estimates of the model. As such, the neurocomputational model offers a quantitative means for the evaluation of formal semantic and pragmatic theories (e.g., regarding implicature, quantification, inference, reference, pragmatic enrichment, etc.) in terms of the two most salient electrophysiological correlates of language processing.
What neuroimaging can add to experimental pragmatics?
Experimental Pragmatics emerged as a means to test theories through empirical means. That semantic accounts are among those tested in Experimental Pragmatics is, of course, a positive development. That said, semantic theories are subtle and arguably too subtle for the kinds of psychological effects that experimental pragmatics seeks to expose. In this talk, I cover the way Gricean approaches have led to experimental pragmatic findings that are edifying to pragmatic accounts and that bring together insights from allied fields. Following up on my last visit to Bochum, I will cover how EEG research allows one to distinguish between two types of pragmatic inference, one that I call voluntary and the other imposed. I will also discuss how Grice’s original insights, with respect to intention reading, can be foundational to neuroimaging research.
Challenges in linking electrophysiological data to the theory of scalar implicatures
The past decade has seen an explosion of research using neuroscientific methods to probe the cognitive processing and representation of many aspects of language, including pragmatic aspects. For the most part, however, the influence of neurophysiological data on theoretical accounts of pragmatics has been negligible. Rather, much neuro-pragmatics research attempts to identify cortical regions or brainwave components that are implicated in some specific type of pragmatics, without having clear implications for theories or models of pragmatics or pragmatic processing. In this talk I will discuss this issue, focusing specifically on neurolinguistics research into scalar implicatures as an example. In this research area, as in many others, there is often difficulty in determining whether brain responses observed in experiments directly reflect something about pragmatic processing, or whether they represent downstream, domain-general processes that may be triggered by many things other than pragmatics. I will argue that a more productive strategy, rather than effect-unspecific studies probing for "neural correlates of X", is to adopt an effect-specific instrumental approach which adopts brain measures with well-known domain-general properties to test specific predictions of cognitive models of pragmatic processing, when such predictions are available.
A neurocognitive perspective on compositionality
Evidence from language comprehension asks for a reassessment of models of meaning adjustment. Online studies reveal discrete processing patterns for different types of metonymy calling i) for a new classification of metonymic expressions as well as ii) a generalization across phenomena at the semantics-pragmatics interface. Electrophysiological data may further help us to disentangle the processes underlying compositional operations. In this talk, I will present a series of ERP studies on meaning adjustment and discuss their implications for a model of compositionality.
Markus Werning
The Interaction of Bayesian Pragmatics and Lexical Semantics in Linguistic Interpretation: Hearers’ Probabilistic Predictions in Discourses
We contrast two views of how contextual influence on sentence comprehension can be explained. The Semantic Similarity View maintains that discourse context affects the comprehension of a sentence mainly because of the semantic similarity between the words in the discourse context and in the sentence. According to this view, discourse mainly facilitates (or impedes) the retrieval of lexical information, but does not affect semantic composition. It is thus consistent with a rigorous interpretation of the principle of compositionality as favored by semantic minimalists and according to which the truth-evaluable semantic content of a sentence is fully determined by its syntactic structure and lexical content where only a small number of lexical items (e.g., indexicals and anaphors) allow for a context-sensitive meaning contribution. The Free Pragmatic View, in contrast, challenges a rigorous notion of compositionality and maintains that pragmatic aspects of the discourse context can affect sentence meaning composition directly: Discourse can modulate sentence meaning at every node of the semantic composition tree. This phenomenon can be quantitatively modelled by Bayesian Pragmatics. We introduce a Predictive Completion Task in which the hearer at every moment in a communicative situation has to generate a probabilistic prediction about how a discourse being uttered by the speaker is continued. We test the predictions of the two views in EEG using the well-established observation that the conditional probability of a word given a context is negatively correlated with the amplitude of its N400 component.