| Keywords [eng] |
Veiklioji atmintis, pragmatinis samprotavimas, Racionalaus Šnekos Akto modelis, Psicholingvistika, Working memory, pragmatic reasoning, Rational Speech Act model, Psycholinguistics |
| Abstract [eng] |
A prominent formal approach in pragmatics to model the capacity of language users going beyond the literal se- mantic content of utterances and inferring speakers’ intended meanings in context is provided by the Rational Speech Act (RSA) framework (Frank & Goodman, 2012). Standard implementations of RSA typically assume idealized agents with unlimited cognitive resources, thereby neglecting the potential impact of processing constraints on pragmatic reasoning. The present study addresses this limitation by examining how working memory constraints influence the derivation of scalar implicatures, pragmatic inferences whereby a weaker expression (e.g., some) is interpreted as excluding stronger al- ternatives (e.g., all). To this end, the study employs a dual-task experimental paradigm, adapted from Marty et al. (2013), where participants are required to perform a primary task involving the interpretation of pragmatically underinformative utterances, while simultaneously maintaining, recalling, and mentally reversing short sequences of letters. The experimen- tal design is grounded in predictions generated from two probabilistic models of pragmatic reasoning: the Vanilla RSA model (Frank & Goodman, 2012) and the Global Intentions model (Franke & Bergen, 2020). By integrating these models, it is hypothesized that working memory constraints may reduce the depth or precision of recursive reasoning, effectively lowering the degree of rationality attributed to agents. Additionally, cognitive load may restrict the set of alternatives that listeners are able to consider, thereby altering the likelihood of deriving scalar implicatures. Through this dual-task paradigm, the study aims to disentangle which components of pragmatic reasoning, such as available alternatives or rationality, are most sensitive to cognitive constraints. By mapping model-based predictions onto empirical performance under varying levels of cognitive load, the research seeks to contribute to a more psychologi- cally realistic account of pragmatic inference, bridging the gap between formal models and human cognitive limitations. However, the experiment ultimately did not yield theoretically informative results. Several factors may have contributed to this outcome, including insufficient task sensitivity, inadequate manipulation of cognitive load, or high variability in participant strategies. In light of these limitations, the study concludes with a discussion of potential avenues for future research. Future work may benefit from integrating more nuanced computational models that explicitly incorporate re- source constraints, such as bounded rationality frameworks. By addressing these challenges, subsequent research can more effectively investigate how real-world cognitive limitations shape the mechanisms of human pragmatic reasoning. |
