How does GPT-4.1 comprehend conversational implicatures? Reasoning with contextual alternatives in discourse frames

This study examines the pragmatic competence ofGPT-4.1, specifically its ability to recover meaning conveyed by Levinsonian generalisedconversational implicatures (GCIs) in the context of interaction with general users. Through nine open-ended queries, we evaluated whether GPT-4.1 can infer implicatedmeaning as predicted by Levinson’s Q-, M-, and I-principles, and investigated themechanisms underlying its performance. Our findings indicate that GPT-4.1 effectivelyrecovers implicated meanings consistent with theoretical predictions. Analysis of themodel’s responses reveals a reliance on evidence-based, context-sensitive interpretationof lexical items and their alternatives, rather than on the direct application of abstractpragmatic principles. While the chatbot demonstrates notable competence in processingimplicated meaning, it primarily operates through contrastive meaning analysis of lexicalalternatives in context, backed by large-scale statistical associations, rather than explicitrule-based reasoning. These results lend support to a growing body of evidencesuggesting that LLMs approach human-like performance in pragmatic inference, whilealso highlighting areas for further research into their ability to use abstract pragmaticrules.