In order to understand most, if not any communicative act, the listener needs to make inferences about what the speaker intends to convey. This perspective-taking process is especially challenging in the case of nonliteral uses of language such as verbal Irony (e.g., "Thanks for your help!" uttered to someone who has not provided the expected support). Children have been shown to have difficulties with the comprehension of Irony well into the school years, but the factors that hamper or facilitate children's perspective-taking in Irony comprehension are not well understood. This study takes as its starting point the relevance-theoretic analysis of verbal Irony, and focuses on two of Irony's distinctive features as defined by this theory: (i) the normative bias and (ii) the characteristic tone of voice. In this study, we investigated the comprehension of Irony in children aged 3-8. We manipulated these two factors, namely, the violation of different types of norms and the use of different tones of voice - to see how they affected children's processing and interpretation of Irony. Using an Irony comprehension task that combined picture selection and eye-tracking, we found that the type of norm violation affected 4-to 5-year-olds' offline understanding of Irony, with a better performance on moral compared with social norm violations. Tone of voice had an effect on gaze behavior in adults, but not children, although a parodic, pretense-oriented tone of voice tended to lead to more looks to the angry compared with the happy emoticon at the offset of the ironical utterance, potentially facilitating children's Irony understanding. Our results show that the understanding of Irony can be detected on explicit measures around age 6 - with the emergence of second-order perspective-taking abilities - but that a sensitivity to some of Irony's features can be detected several years earlier. Finally, our study provides a novel input to the debate on the existence of a so-called literal stage in pragmatic development, in particular regarding 3-year-olds' differential performance on the offline and online measures of Irony understanding, suggesting that they are not naively mistaking ironical utterances for "ordinary" literal ones.
