In this paper, we propose a fine-grained, empirically-driven formalisation of framing in discourse, with a particular focus on dialogical interaction. This formalisation provides an operationalization of framing that lays the foundation for subsequently (i) designing guidelines for the annotation of framing moves in large-scale dialogue datasets and (ii) implementing computational models that can be informed with the necessary cornerstones of linguistically-motivated models of framing. The core assumptions of our model are based on Entman (1993) who defines framing as the way “to select some aspects of a perceived reality and to make them more salient”. But we go beyond the work of Entman to show, in abstract, generalised terms, what it means to use framing to make things more salient and, furthermore, how such salience can be changed through framing dynamics. The paper describes the formal underpinnings of the ways that frames evolve and are reacted to in dialogue by way of seven framing moves. These moves characterise how interlocutors modify salience from one frame to another. Through our model, it becomes possible to capture both local and long-distance structures of framing in natural communication. We illustrate the workings of the model based on two real-life examples of framing in dialogue and highlight how this model can connect to contemporary pragmatic models of communication.
This special issue adopts a multifaceted approach to analyzing compliments from various perspectives. The introductory paper by Maíz-Arévalo (2026) organizes previous studies on compliments into three waves, providing a significant overview of the literature on compliments. The papers by Meluzzi and Castagneto (2026), Xie (2026) and Jia (2026) are focused on the under-researched area of the perception of compliments. From a cross-cultural perspective, Ravetto (2026) examines the lexical and syntactic patterns of explicit and implicit German and Italian compliments, identifying six different patterns of implicit compliment, two of which are culturally specific to Italian. Two papers address the sociolinguistic factor of gender in compliments: Alfonzetti and Scivoletto (2025) approach this topic from both production and perception perspectives, while Sorianello and Gurrado (2025) demonstrate that women’s compliments are emphatic at the lexical level, whereas men’s compliments are characterized by marked prosody. As ever, the pragmatics of this complex speech act is grounded at the interfaces of different forms, functions and linguistic levels.
Metaphors are central to health communication, especially for sensitive topics like cancer and depression. As AI tools like ChatGPT become increasingly prevalent in health conversations, their capacity to process and respond to metaphors remains underexplored. This study investigates whether ChatGPT processes metaphors pragmatically and what this reveals about the pragmatics of AI-generated language. Utilizing metaphors commonly found in cancer and depression discourse, we interacted with ChatGPT-4o mini using a 2 × 2 design, comparing responses across illness type (cancer versus depression) and input metaphoricity (metaphorical versus literal). We evaluated how ChatGPT organizes its responses though pragmatic strategies, such as situation re-construction, alternative reasoning, and agreement with the user, as well as its handling of metaphors via reuse, replacement, or creation. Poisson and Negative Binomial regression analyses were conducted to examine how illness type and input metaphoricity influenced ChatGPT’s responses. Results show that responses to metaphorical prompts contained more pragmatic elements and metaphors than those to literal prompts. Situation re-construction and alternative reasoning were the most frequent pragmatic strategies, regardless of illness type or input metaphoricity. Depression-related literal prompts elicited longer responses and more metaphors than cancer-related literal prompts. A chi-square test further revealed a significant association between metaphor use and pragmatic elements, indicating that metaphor handling is systematically linked to the pragmatic organization of ChatGPT’s responses. Overall, both response structure and metaphor use are largely templated, lacking the flexibility and nuance characteristic of human interaction. These findings highlight both the potential and limits of ChatGPT’s pragmatic competence in health communication.
