Abstract: Successful communication relies on perspective-taking, which is the ability to infer a speaker’s intentions by representing their beliefs and attitudes. A central theoretical issue concerns whether this process is automatic or effortful, reflecting the broader debate between serial and constraint-based models of language processing. While serial models posit an initial egocentric bias followed by later adjustment, constraint-based models assume that perspective is integrated in parallel with other linguistic cues from the outset.
To examine how social context and task demands shape perspective-taking during irony comprehension, this study employed a 3 (speaker’s belief: classically gender-stereotypical, counterstereotypical, neutral) × 2 (utterance type: ironic, literal) within-participants design. Participants read scenarios composed of six segments (Background, Belief, Event, Target, Spillover, and Conclusion). The experimental materials were identical across conditions except for the Belief and Event segments, in which critical variations were introduced. Two self-paced reading experiments were conducted. In Experiment 1, participants provided explicit irony judgements on a 5-point scale and then answered yes/no comprehension questions. Experiment 2 used only the comprehension questions and omitted the rating task. Individual differences were also assessed across multiple domains, including implicit and explicit gender attitudes (via the IAT and CMSS, respectively), social cognitive abilities (via the Interpersonal Reactivity Index Scale, IRI), and executive functions (via the Stroop task and a working memory capacity task), to examine their respective contributions to irony comprehension.
In Experiment 1, an irony effect (longer reading times for ironic utterances than for literal utterances) was observed in the target region when the speaker's beliefs were counterstereotypical or neutral, but no such effect emerged when beliefs were classically stereotypical. The absence of an irony effect in the classically stereotypical condition reflects the fact that a highly predictable belief constitutes a strong contextual constraint, which facilitates the integration of ironic meanings without incurring additional processing costs. Individual difference analyses further revealed that only the subscale of the IRI, Empathic Concern (EC), modulated irony processing. Specifically, in the spillover region, higher IRI-EC was associated with facilitated processing of ironic utterances under both the classically stereotypical and neutral conditions. In Experiment 2, processing costs were confined to the neutral condition and appeared only in the spillover region and not in the target region. Individual difference analyses further revealed that only implicit gender attitudes moderate reading times in the neutral condition. Specifically, in neutral contexts, readers with stronger implicit gender stereotypes exhibited more cautious processing of literal utterances, as reflected in prolonged reading times in the spillover region.
Three main conclusions emerge: (1) Speakers’ classical social beliefs generate strong expectations that facilitate irony processing without costs, whereas unconventional or neutral beliefs generate weak expectations and thus incur processing costs, which is consistent with a constraint-based parallel processing account. (2) Task demands modulate the timing and sensitivity to social conflict; explicit tasks amplify and accelerate the cost of counterstereotypical beliefs, whereas the condition that omits the explicit irony-rating task (while retaining comprehension questions) delays that cost. (3) Individual differences in implicit attitudes and social-cognitive abilities further moderate perspective-taking during irony comprehension.

Key words: perspective-taking, irony, gender stereotypes, pragmatics, individual differences