Irony comprehension in developmental perspective

As children move from early to middle childhood and develop more complex social motivations (Carver et al., Reference Carver, Yunger and Perry2003; Del Giudice, Reference Del Giudice2014), they learn to use language in more sophisticated ways for various social pragmatic goals. Irony comprehension is thought to be a major milestone in the development of children’s social cognition and social-communicative competence (Aguert et al., Reference Aguert, Le Vallois, Martel and Laval2018; Peterson et al., Reference Peterson, Wellman and Slaughter2012). It is a skill that develops quite late in a child’s development, as compared to other aspects of language such as semantics or syntax (e.g., Ackerman, Reference Ackerman1983; Demorest et al., Reference Demorest, Silberstein, Gardner and Winner1983, Reference Demorest, Meyer, Phelps, Gardner and Winner1984). Although much has already been discovered about the developmental aspects of irony comprehension, not all aspects have been fully mapped. Many studies indicate that irony comprehension begins before the age of six, whereas some researchers claim that it emerges around the age of ten (Ackerman, Reference Ackerman1983; Dews et al., Reference Dews, Winner, Kaplan, Rosenblatt, Hunt, Lim, McGovern, Qualter and Smarsh1996; Filippova & Astington, Reference Filippova and Astington2010; Harris & Pexman, Reference Harris and Pexman2003; Pexman & Glenwright, Reference Pexman and Glenwright2007; Sullivan et al., Reference Sullivan, Winner and Hopfield1995). Earlier research suggested that even 13-year-old children were not always able to understand irony, and they were unable to distinguish it from cheating (Demorest et al., Reference Demorest, Silberstein, Gardner and Winner1983, Reference Demorest, Meyer, Phelps, Gardner and Winner1984). However, more recent studies show that this emerging skill can be observed much earlier: in 3- and 4-year-olds (Banasik-Jemielniak & Bokus, Reference Banasik-Jemielniak and Bokus2019; Loukusa & Leinonen, Reference Loukusa and Leinonen2008; Loukusa et al., Reference Loukusa, Leinonen and Ryder2007; Recchia et al., Reference Recchia, Howe, Ross and Alexander2010). Pexman et al. indicated that basic comprehension is established by five to six, but a more demanding skill–appreciation of its pragmatic functions such as humor–continues to unfold across childhood (e.g., Pexman & Glenwright, Reference Pexman and Glenwright2007).

The ability to understand irony develops with the child’s age (Bosco & Bucciarelli, Reference Bosco and Bucciarelli2008; Demorest et al., Reference Demorest, Meyer, Phelps, Gardner and Winner1984; Dews et al., Reference Dews, Winner, Kaplan, Rosenblatt, Hunt, Lim, McGovern, Qualter and Smarsh1996; Dews & Winner, Reference Dews and Winner1997; Filippova & Astington, Reference Filippova and Astington2008). Comprehension that a speaker is being ironic (or that the speaker means something other than what they are literally saying) precedes the comprehension of what the speaker intends by being ironic (sometimes referred to as irony appreciation).

For example, studies of Finnish children aged three to seven years found significant differences between older and younger children’s explanations in response to a simple ironic statement (Loukusa et al., Reference Loukusa, Ryder and Leinonen2008) – older children showed higher levels of comprehension than younger children. Other studies show that children aged six, eight, and nine are better at dealing with a simple form of irony, i.e., statements that are the opposite of contextual information, than with complex ironic statements (Bosco & Bucciarelli, Reference Bosco and Bucciarelli2008). Irony comprehension could be divided into two main categories. The first is basic comprehension, when a child understands that a speaker is being nonliteral. The second refers to a more advanced appreciation, when a child understands the goal of speaking that way or is able to appreciate the function of this form. The level of irony comprehension in children is related not just to age, but to various mental competencies that tend to develop with age, such as cognitive flexibility and theory of mind. For instance, Zajączkowska and Abbot-Smith (Reference Zajączkowska and Abbot-Smith2020) examined the children’s understanding of both simple and complex irony and found that in younger children, aged six to eight, cognitive flexibility and theory of mind accounted for unique variance in their interpretation of simple irony. In contrast, for older children, aged ten to twelve, cognitive flexibility predicted irony comprehension, but theory of mind did not. This also shows the complexity of the phenomenon of irony comprehension and of the trajectory: with age, one predictor can become insignificant, whereas another one only gains significance at a certain age.

Metapragmatic development

Irony comprehension is linked to metapragmatic abilities (Szücs & Babarczy, Reference Szücs and Babarczy2017), also known as the ability to explicitly reflect on pragmatic rules (Collins et al., Reference Collins, Lockton and Adams2014). According to Gombert (Reference Gombert1992: 186), there are four phases of metalinguistic development. In the first phase, there is no reflection on the language itself. In the second stage, which takes place during the preschool years, epilingual (or epilinguistic) control appears. The term epilingual refers to a behavior which is comparable to metalinguistic, but unlike metalanguage, which might be realized with a decontextualized task, is usually situated in a rich communicative context. Epilinguistic abilities are described as more subconscious than metalinguistic abilities.

The development of the third stage begins when the child is around five to six years old. Reflection on linguistic processes may occur in several ways, including plays on words (rhymes) and interest in knowing how words are written and why things are described one way and not another. The fourth stage, called process automation, allows the individual to face, reflect on, and monitor the actions that must be carried out to solve some linguistic problems, such as separating words from the communicative context to analyze and reflect on their meaning. The latter two stages are relevant to irony comprehension.

Metalinguistic ability, therefore, is directly related to the ability to analyze the words themselves, making it essential for the correct interpretation of non-literal language, and specifically, verbal irony, where the literal meaning of the words emitted does not coincide with that of the communicative intention of the speaker. To distinguish between epilinguistic and metalinguistic abilities, the following example could be used: during a shared picture book reading, a caregiver says, “Look, this shelf is so high.” The child, noticing that “high” sounds like “hi,” begins playing with the two words. “Hi, shelf!” This epilinguistic ability occurs within the context, in the here and now. The child, prompted by the caregiver pointing out the high shelf, connects “hi” (as in “hello”) with “high” (as in “tall”). A separate example illustrates metalinguistic ability: unprompted, a child says, in conversation, “There are some words that sound the same, but they mean different things, like “hi” and “high”!

Laval (Reference Laval2003) found that some metapragmatic knowledge is apparent as early as the age of six years and is related first to the contextual features of the communicative situation (such as an illustration accompanying a sentence that may be interpreted either literally or non-literally). Metapragmatic knowledge of linguistic conventions (i.e., the link between the literal meaning and the idiomatic meaning) usually emerges at the age of nine. In our study, we looked at children at the age of eight. According to Laval’s findings, these children do not yet exhibit metapragmatic knowledge of linguistic conventions. Our goal was to examine whether children produce metapragmatic references when presented with an ironic utterance followed by an open-ended question.

Additionally, we explored if girls and boys differed from each other in this aspect. Our motivation was twofold. Firstly, past research shows parents use more emotion and social, moral words with girls compared to boys (Bosacki et al., Reference Bosacki, Harwood and Sumaway2012; Maccoby, Reference Maccoby1998). Secondly, there are only a few studies on gender differences in middle childhood in metapragmatics and irony comprehension (Berglund et al., Reference Berglund, Eriksson and Westerlund2005; Huttenlocher et al., Reference Huttenlocher, Haight, Bryk, Seltzer and Lyons1991; Pınar et al., Reference Pınar, Ozturk, Ketrez and Özçalışkan2020; Recchia et al., Reference Recchia, Howe, Ross and Alexander2010; Rothermich et al., Reference Rothermich, Caivano, Knoll and Talwarin press; Tomasi & Volkow, Reference Tomasi and Volkow2012) and while some of them indicate the differences in fact are noted, others show similar performance by girls and boys (Gabbatore et al., Reference Gabbatore, Bosco, Mäkinen, Leinonen and Loukusa2021). Research shows that adult men use irony more frequently than women (e.g., Milanowicz & Bokus, Reference Milanowicz and Bokus2020), but to our best knowledge, no research has found differences in irony comprehension between girls and boys in middle childhood.

Method

The study was part of a larger project (see Banasik-Jemielniak et al., Reference Banasik-Jemielniak, Bosacki, Mitrowska, Wyrębek Walters, Wisiecka, Copeland, Wieland, Popovic, Piper and Siemieniuk2020), where children and their parents, respectively, were tested on irony comprehension (children) and use (parents), and humor styles. The sample included 46 families from a large city in Poland, where there was at least one eight-year-old child. Inclusion criteria for families were the presence of one eight-year-old child. The sample was homogenous in terms of SES, race and ethnicity: it included middle-class, Caucasian parents of Polish background. The majority of the parents had completed higher education (university degree). Before starting the research, ethical clearance was received from the participating university (Faculty of Psychology, University of Warsaw, 2017). Written informed consent by parents of all participating children was obtained. The final sample included 24 girls, 22 boys, 46 mothers, and 41 fathers. Children’s mean age was eight years; five months (SD = 3.9). All children attended state schools and were in second grade at the time of testing. All children were monolingual speakers of Polish who had not been diagnosed with any atypicalities in language or cognitive development. The families were recruited through social media and snowballing. We also randomly chose responses to matching questions of 45 five-year-old children, whose data was gleaned from a larger project (Banasik-Jemielniak et al., Reference Banasik-Jemielniak, Bosacki, Mitrowska, Wyrębek Walters, Wisiecka, Copeland, Wieland, Popovic, Piper and Siemieniuk2020). Older children were tested in their homes, and younger children were tested on the preschool premises after having been previously introduced to the researcher.

Materials

The ICT-sv was created by shortening the Irony Comprehension Task (ICT) (Banasik, Reference Banasik2013). The short version showed a satisfactory external validity (Banasik-Jemielniak & Bokus, Reference Banasik-Jemielniak, Bokus and Kyuchukov2022). The shortened task is composed of three stories describing situations in which one of the characters makes an ironic comment to a child. The verbal irony in the characters’ comments is based on the inversion of meaning, i.e., they say: “We are so lucky today” when everything goes wrong or “My favorite food”, when the most hated food is served. Such use of verbal irony is often referred to as conventional (Alba-Juez, Reference Alba-Juez2014). The stories are read to the child by the researcher while showing pictures on a computer screen. After each story, the child is asked questions about the ironic comment: an open question “Why do you think X said Y?”, and a comprehension (closed-ended) question which was counted as a score of irony comprehension: “When X said Y, did they mean Y or Z?” of every character who is a speaker. An example of a closed-ended question is after the character of the story said: ‘We are so lucky today!’ (please see appendix 1, story 2) the question directed to the child was: “When mommy said ‘We’re so lucky today!’, did she mean that they were lucky and everything went well or did she mean that they were not lucky and nothing went well?”.

The participants were then asked two questions to assess how funny the ironic comment was (1) and how nice/mean the character uttering it was (2). For the quantitative score of simple comprehension, the first question (closed question) was scored for each of the presented stories, meaning that a total score of 3 could be obtained. The stories used can be found in appendix 1. Data from five-year-old children were gleaned from a larger study, in which they responded to 12 questions of the unshortened task. For the purpose of the analysis, we only included responses to the questions matching those used in the short task.

Coding scheme

The responses to the open-ended questions were analyzed within the content analysis framework, which involves classifying the responses into categories distinguished on the basis of patterns of meaning (Hsieh & Shannon, Reference Hsieh and Shannon2005). The categories were created following Pike’s emic approach to language analysis (Reference Pike1967), according to which criteria are selected from an emergent perspective. Our findings emerged from the data and were not made a priori or before the study. Based on the data gathered, seven categories of children’s responses were distinguished, which are presented in Table 1.

Table 1.

Coding scheme for children’s responses to open-ended question

Table 2.

Categories of responses given by eight-year-olds after the exclusion of failed accuracy questions.

If the numbers differ in comparison to the data prior to the exclusion, we added the previous result in brackets.

A comment was categorized into a rephrasing statement (1) when the child repeated part of the story. This response does not allow us to infer either comprehension of the intended meaning of the speaker by the child or the lack of the comprehension. The literal statement label (2) was given to the responses in which the child referred to the character’s utterance as if it was intended literally, in a non-figurative way. Giving a response that falls into this category may indicate a lack of comprehension of the intended meaning of the speaker.

When children referred to the inversion of the literal meaning, which in the given task is equated with referring to the intended meaning, it was categorized as a reference to the intended meaning (3), and when they mentioned emotions, such as “She is really angry the bus went away,” it was categorized as a reference to emotions (4). Both (3) and (4) may be indicators of emerging irony comprehension since they manifest the recognition of an additional meaning behind the literal one. When the child talked about language conventions, but without the use of the term “irony” or “sarcasm”, we categorized the response as a metalinguistic reference without label (5). When they used the label “irony” or “sarcasm”, we classified the reply as a metalinguistic reference with label (6). If the child did not respond, said “I don’t know’’ or gave another answer that could not be classified within the 6 above, we marked it as “no response/other” (7).

While categories 3-6 can be treated as a scale where a higher number refers to a higher emergent level of irony comprehension, category 2 indicates no irony comprehension and categories 1 and 7 entail no obvious interpretation.

Once the categories were created, two authors independently coded 20% of the children’s responses. Next, Cohen’s kappa was run to determine if there was agreement between the coders. There was a strong agreement k = .89, p<.005. Any disagreements were resolved by discussion. The remaining data were coded by the first author.

Power analysis

Using G*Power (Faul et al., Reference Faul, Erdfelder, Buchner and Lang2009) a sensitivity analysis was conducted to determine that our sample of 46 with an alpha level of 0.05 and a medium effect size of 0.3 would achieve power for the study 0.67, which is not satisfactory. This is one of the shortcomings of the quantitative part of our study.