Are Children Sensitive to Ironic Prosody? A Novel Task to Settle the Issue

Abstract

Ironic remarks are often pronounced with a distinctive intonation. It is not clear whether children rely on acoustic cues to attribute an ironic intent. This question has been only indirectly tackled, with studies that manipulated the intonation with which the final remark is pronounced within an irony comprehension task. We propose a new task that is meant to assess whether children rely on prosody to infer speakers’ sincere or ironic communicative intentions, without requiring meta-linguistic judgments (since pragmatic awareness is challenging for young children). Children listen to evaluative remarks (e.g., “That house is really beautiful”), pronounced with sincere or ironic intonation, and they are asked to identify what the speaker is referring to by selecting one of two pictures depicting an image corresponding to a literal interpretation (a luxury house) and one to its reverse interpretation (a hovel). We tested eighty children aged 3 to 11 years and found a clear developmental trend, with children consistently responding above the chance level from age seven, and there was no correlation with the recognition of emotions transmitted through the vocal channel.

1. Introduction

If Ann asks Leo to hand her a mug of beer, but he trips and spills it on her, Ann might comment “Good job!”. In uttering such a remark, Ann is being ironic: she surely does not think that Leo did a good job, and she intends to blame him, not to praise him. How is this figurative meaning communicated, and how can the interlocutor recognize it? As summarized in Garmendia (2018), there are three main pragmatic approaches to verbal irony, making different predictions as to the cues that the audience might rely on to detect irony. Classical accounts, such as Grice (1975), focus on the opposition between what is said and what is meant. Having noticed that the speaker said something blatantly false, the interlocutor will infer that they meant to communicate the opposite. According to the Echoic theory (Wilson & Sperber, 2012), on the other hand, the ironic remark is not used, but simply mentioned, to express the speaker’s contemptuous attitude towards the thought that is being echoed (e.g., that the addressee did a good job). Similarly, in the Mention theory (Clark & Gerrig, 1984), the speaker is just pretending to be someone who would utter that remark (e.g., “Good job!”), thus ridiculing whoever would be so foolish to believe it. In both cases, when the addressees recognize the ironic speaker’s mocking and derogatory attitude, they will understand that the speaker is not endorsing what they said and typically believe the opposite.

To detect irony, interlocutors might rely on different cues. Since all ironic remarks are evaluative statements (Partington, 2007) that are incongruent with the context (Attardo, 2000), if such discrepancy is acknowledged, it constitutes the strongest cue for the attribution of an ironic intent (Rivière et al., 2018; Deliens et al., 2018a, a.o.). If, on the other hand, the incongruence of the comment relative to the situation is not evident, communication failures might take place, with the interlocutor interpreting literally the speaker’s meant-to-be ironic remark. To avoid misunderstandings, ironic speakers might accompany their remarks with irony markers, that is, metacommunicative cues that alert the audience that the comment should not be taken at face value. The most studied irony marker is the so-called ironic tone of voice. Interestingly, the aforementioned theories make different predictions as to the ironic intonation. As highlighted by Wilson and Sperber (2012), the classical Gricean model cannot explain why, among figurative expressions, only irony is typically pronounced with a distinct intonation that has the function of making manifest the speaker’s attitude. In Sperber and Wilson’s Echoic theory, on the other hand, ironic speakers are expressing their negative, derogatory attitude towards the thought that is echoed, so a tone “of contempt” is expected; Clark and Gerrig’s Pretense theory hypothesizes that the ironic speaker is pretending to be another person, and thus the remark should be pronounced with a parodic intonation, which has “an exaggerated tone of enthusiasm or even worship” (Sperber, 1984, p. 135). Even if it is not possible to single out specific acoustic parameters that uniquely characterize the ironic intonation, there is evidence that adults do rely on prosody to attribute an ironic intent since they can correctly discriminate ironic remarks relying on prosody only (Bryant & Fox-Tree, 2005; Giustolisi & Panzeri, 2021), at least in their native language (Cheang & Pell, 2011; Bettelli et al., 2024).

When we turn to the acquisition of irony, the situation is less clear cut. The recognition of the speaker’s ironic intent and attitude is a late-developing skill. Children start grasping ironic criticisms (i.e., literally positive remarks ironically used to blame, as in the example discussed above) around six years of age, while ironic compliments (i.e., literally negative comments used to praise, as saying “That went really bad” to someone who just got straight As in all exams) are understood later (Harris & Pexman, 2003; Pexman & Glenwright, 2007). To account for younger children’s difficulties in irony comprehension, it has been hypothesized that children must possess sophisticated mindreading abilities (Sullivan et al., 1995), and indeed, several studies found a correlation between Theory of Mind skills and irony comprehension (Filippova & Astington, 2008; Nilsen et al., 2011; Massaro et al., 2013), even if other scholars disputed this conclusion (Bosco & Gabbatore, 2017; Szücs & Babarczy, 2017), highlighted the prominent role of linguistic competence (Angeleri & Airenti, 2014), or obtained mixed findings (Zajączkowska & Abbot-Smith, 2020).

More importantly, as discussed in Fuchs (2023), it is not clear whether children rely on intonation to attribute an ironic intent. As for the relative weight of context and prosody to detect irony, Vesper (1997, cited in Fuchs, 2023) found that children (aged six to ten) behave as adults, relying more on the incongruity between context and utterance compared to intonation. Other studies, on the other hand, revealed the opposite pattern. Laval and Bert-Erboul (2005) found that at five years of age, children need to rely on intonation to appreciate the role of the context, whereas, for seven-year-olds, context and prosody play the same role. Capelli et al. (1990) tested older (eight to twelve years of age) children, who could recognize sarcasm if pronounced with ironic intonation but failed to do so without the prosodic cue, even if the context was strongly incongruent. Other studies focus specifically on the role of prosody, manipulating the intonation with which the final remark is pronounced within an irony comprehension task. Participants are presented with a context that contradicts, is neutral, or consistent with the final remark, pronounced with a sincere/neutral or ironic intonation. The hypothesis is that if the presence of ironic prosody increases participants’ attribution of ironic intent, then they are sensitive to this phonological cue. The results are contrasting. Many studies found that six-year-olds make little use of the ironic tone of voice compared to older children (from eight years of age) and adults (Winner & Leekam, 1991; Demorest et al., 1984; Sullivan et al., 1995). Ackerman’s (1983) results indicate that six-year-olds do not seem to rely on intonation, whereas the presence of the ironic tone of voice (“sarcastic disapproving intonation”) guides eight-year-olds towards the correct detection of the attitude, but not of the communicative intent, of the speaker. In Winner et al. (1987), the presence of ironic intonation did not facilitate irony comprehension of even older children (eight- and ten-year-olds). In contrast, de Groot et al. (1995) and Keenan and Quigley (1999) have found that intonation does facilitate six-year-olds’ irony comprehension; along the same lines, Glenwright et al. (2014) demonstrated that five- and six-year-old children (and even adults) are sensitive to the magnitude of the reduction in pitch (fundamental frequency, F0) that characterizes ironic prosody: the larger the reduction, the higher the probability that an ironic intent is attributed to the remark. Finally, Köder and Falkum (2021) investigated the comprehension of irony in children aged 3–8, manipulating also the type of ironic tone of voice. The final remarks were pronounced either with a deadpan (flat, slower tempo, lower pitch level, and greater intensity) intonation, which should represent the dismissive attitude of ironic speakers in the Echoic theory (Wilson & Sperber, 2012), or with a parodic (imitative and exaggerated) contour, with the speaker mimicking another person, always with the goal of ridiculing them, as predicted by the Pretense theory (Clark & Gerrig, 1984). The results are mixed. The type of prosodic contour did not have an effect on children’s behavioral responses (the choice of the emoticon depicting the speaker’s attitude) but influenced children’s eye gazes, which converged more rapidly to the correct target in the presence of a parodic intonation, differently from what was found with adults.

It is important to note that in this kind of task, children’s attunement to ironic prosody is only indirectly assessed. Remarks are presented after a context that weakly or strongly biases towards a sincere or ironic interpretation of the remark, and the presence (or absence) of ironic intonation is meant to facilitate (or hamper) the attribution of ironic communicative intent. Filippova and Astington (2008), on the other hand, addressed the issue of children’s sensitivity to ironic prosody by testing whether accuracy in an irony recognition task correlated with performance in an emotional prosody discrimination task. It is well known that infants are highly sensitive to prosodic variations in the voice of their caregivers. Infant-directed speech is typically characterized by changes in pitch levels that convey emotionality, and by twelve months of age, children adapt their behavior to adults’ paralinguistic cues (Mumme & Fernald, 2003). Children’s ability to correctly categorize and label vocally transmitted emotions, though, develops over time. Children from the age of five are able to detect happiness, fear, and sadness (Allgood & Heaton, 2015), as well as even more complex emotions (amusement, contentment, relief, achievement, anger, disgust, fear, sadness, surprise, and neutral; Sauter et al., 2013), and they become adult-like by mid-childhood (Doherty et al., 1999) to early adolescence (Aguert et al., 2013; Zupan, 2015). Filippova and Astington’s (2008) idea is to verify whether those children who are more accurate in detecting emotions (happiness, sadness, anger, fear) communicated through intonation are also good at recognizing irony. If this is the case, attunement to expressive prosody might facilitate irony understanding. Thus, Filippova and Astington (2008) tested 5-, 7-, and 9-year-old children with a classic irony comprehension task, where the preceding context biases toward a literal (sincere) or reverse (ironic) interpretation of the final remark, and with an emotion recognition task, in which children had to recognize the (happy, sad, angry, or afraid) emotion encoded in the speech signal. Since no correlation was found between accuracy in these two tasks, they concluded that sensitivity to intonational cues does not contribute to children’s overall ability to draw inferences about a speaker’s ironic intent.

To the best of our knowledge, no task has directly assessed children’s detection of the ironic tone of voice in the absence of preceding context. To verify whether adult participants detect the ironic intonation in isolation, discrimination tasks are typically used. Participants are presented with remarks pronounced with a sincere or an ironic prosody, and they are asked whether the speaker is ironic, directly (using a Likert scale—Bryant & Fox-Tree, 2005, or a dichotomous judgment—Giustolisi & Panzeri, 2021, exp. 2) or indirectly (saying whether the speaker meant what they say—Mauchand et al., 2020). Deliens et al. (2018a, 2018b), though, argued that these tasks do not tackle irony comprehension, which requires the identification of the speaker’s goals and not simply the detection of a distinctive prosody. Moreover, these discrimination tasks require meta-linguistic judgments, that is, they require participants to consciously reflect on language and to report their intuitions about language—and thus, they should not be used with young children (Gillioz & Zufferey, 2020) because pragmatic awareness is challenging for them (Collins et al., 2014; Bernicot et al., 2007). As highlighted by Falkum and Köder (2024), tasks that require metapragmatic reasoning impose linguistic or metacognitive demands that might obscure children’s actual understanding of the phenomenon under investigation, as demonstrated by the fact that children’s performance on figurative language tests is better when they are asked to select the image depicting the non-literal interpretation of a metaphor, for instance, rather than verbally explaining it.

To address these issues, we designed a new Ironic Prosody Recognition task that is meant to assess whether children rely on prosody to infer speakers’ sincere or ironic communicative intentions. Children listen to evaluative remarks (e.g., “That house is really beautiful”), pronounced with sincere or ironic intonation, and their task is to identify what the speaker is referring to by selecting one of two pictures depicting an image corresponding to a literal interpretation (a luxury house) and one to its reverse interpretation (a hovel). We reason that if children select the correct picture (i.e., the luxury house if the remark “That house is really beautiful” is uttered sincerely, and the hovel in the presence of ironic intonation), they demonstrate sensitivity to prosody and rely on acoustic features to derive the speaker’s communicative intention.

We aimed to assess at which age children exploit phonological cues to recognize irony. Since, to the best of our knowledge, this is the first study that evaluates the detection of ironic prosody in children directly, without a preceding context, we did not have specific experimental hypotheses regarding the age at which children are sensitive to ironic intonation, even if the majority of previous studies suggest that six-year-olds do not exploit prosodic cues for irony comprehension (Ackerman, 1983; Winner et al., 1987; Winner & Leekam, 1991; Demorest et al., 1984; Sullivan et al., 1995). A second goal of our study was to verify whether attunement to ironic intonation is related to the detection of emotions transmitted through the acoustic channel. No previous study specifically investigated this question. Filippova and Astington’s (2008) results (no relation between irony comprehension and emotional prosody recognition) might suggest that the development of these abilities proceeds along different routes. We decided to test children on two prosody recognition tasks, one involving irony and the other emotions, to see whether Filippova and Astington’s conclusions are supported. To achieve these goals, we tested children aged 3 to 11 years with an Ironic Prosody Recognition task and an Emotional Prosody Recognition task.

2. Methods

2.1. Participants

We tested a total of eighty children. They all had Italian as their native language, and none of them was diagnosed with hearing or language problems. Twenty-seven were enrolled in preschool: eight were three-year-olds (P3); seven were four-year-olds (P4); and twelve were five-year-olds (P5). Fifty-three children were enrolled in primary school: ten in the first year (S1), ten in the second year (S2), twelve in the third year (S3), ten in the fourth year (S4), and eleven in the fifth year (S5). Demographic characteristics of the participants are shown in Table 1.

Table 1. Characteristics of the participants.

School	Grade	Tag	N	Sex	Mean Age (Year; Months)	Age Range (Year; Months)
Preschool	3-year-olds	P3	8	3 F, 5 M	3; 9	3; 5–3; 11
	4-year-olds	P4	7	3 F, 4 M	4; 8	4; 4–4; 11
	5-year-olds	P5	12	5 F, 7 M	5; 7	5; 2–6; 3
Primary school	1st	S1	10	4 F, 6 M	6; 9	6; 1–7; 3
	2nd	S2	10	2 F, 8 M	7; 9	7; 2–8; 1
	3rd	S3	12	5 F, 7 M	8; 11	8; 5–9; 2
	4th	S4	10	7 F, 3 M	9; 10	9; 6–10; 3
	5th	S5	11	6 F, 5 M	10; 10	10; 5–11; 2

2.2. Materials and Procedure

We prepared two new tasks that aimed at evaluating the detection of the ironic intent (Ironic Prosody Recognition task) and of emotions (happiness, anger, sadness, and neutral, Emotional Prosody Recognition task) transmitted through modulations of the tone of voice. The Ironic Prosody Recognition task consisted of remarks, uttered in Italian, which contained a literally positive evaluation (e.g., “Veramente bella quella casa”, ‘That house is really beautiful’), paired with two pictures, which represented a literal and a reverse interpretation. Thus, for instance, for the comment “That house is really beautiful”, one picture depicted a luxury house and the other a hovel, as illustrated in Figure 1. The remarks were pronounced either with sincere or ironic intonation. The participants’ task was to indicate the picture that represented what the speaker intended to refer to, in order to avoid metapragmatic judgments that might obscure children’s actual understanding of the speaker’s communicative intent, as suggested by Falkum and Köder (2024).

Figure 1. Pictures accompanying the remark “That house is really beautiful”.

To prepare the material, we first selected twelve evaluative adjectives (such as “beautiful”) or predicates (such as “have fun”) that had a positive polarity and then created three different statements for each adjective/predicate, choosing for each sentence the two accompanying pictures—one corresponding to the literal and one to the reverse interpretation. We then asked a female Italian-speaking adult to pronounce each of the thirty-six remarks with a sincere and an ironic prosody. To verify that the intended communicative intent was recognizable, we ran a first pilot study with twenty-one adult participants, who were presented with the seventy-two remarks (in a randomized order), and we asked them to select the image the speaker intended to refer to. The overall accuracy was 91.52%. Among the seventy-two items, we then selected twelve remarks (six sincere and six ironic) with different adjectives/predicates, choosing items that were correctly recognized by at least 19 participants out of 21 (accuracy of at least 90%). We then ran a second pilot study, with twenty adult participants (different from those of the first pilot study), who were presented with these twelve remarks only, which were correctly recognized 97.08% of the time. We thus used these twelve remarks (reported in Appendix A) as material for the Ironic Prosody Recognition task for children.

The Ironic Prosody Recognition task thus consisted of a PowerPoint presentation, in which participants listened to the twelve remarks, and their task was to select the image that the speaker was referring to. To familiarize children with the task, they were first told that someone might say things they do not actually believe, such as saying “You did a great job!” to someone who just spilled a glass of milk on the floor, and that this would constitute a sort of joke because they actually think that the other person did not do a good job. Indeed, these instructions alerted children that they should expect non-literal uses of the language. This was decided because children exhibit a tendency for a literal interpretation of language (Winner, 1988/1997; Levorato & Cacciari, 2002; Köder & Falkum, 2020). More specifically, Falkum and Köder (2024, p. 158) underline how children’s expectations related to the experimental context might influence their responses to irony comprehension tasks. Since children are sensitive to linguistic and social conventions, they might expect to be tested on purely linguistic abilities, and this might lead them to assume that the literal interpretation—which is in line with the conventional use of words and sentences—is more “correct” than the ironic one. For these reasons, we deemed these instructions necessary for children to understand that they should pay attention to the way the remarks were pronounced, and not simply stick to their literal interpretation. Likewise, we presented two practice items, one remark pronounced with sincere prosody and one with an ironic tone of voice, and invited children to select the image the speaker was referring to, correcting them in case they selected the wrong picture. For the subsequent twelve target items, no feedback was given.

For the task assessing the recognition of emotional prosody, we first selected twelve statements with a semantically neutral content (such as “It is quarter past four”) and then asked a female Italian speaking adult to pronounce them with a neutral prosody and with happy, angry, and sad intonation. To verify whether the emotions were recognizable, we ran a first pilot study in which twenty adult participants had to associate each of the forty-eight sentences, presented in a randomized order, with an emoji depicting a neutral, happy, angry, and sad facial expression (see Figure 2). The overall accuracy was 94.6%. We then selected twelve items (that were correctly recognized by at least 18 participants out of 20), one for each statement and three for each emotion. We then ran a second pilot study, presenting only those twelve items, to another group of forty-one participants, who responded correctly 94.31% of the time. We thus used these twelve items as material for the Emotional Prosody Recognition task, which comprises twelve semantically neutral sentences pronounced with a sad, angry, happy, or neutral tone of voice., The task was implemented as a PowerPoint presentation, and the children’s task was to select the emoji that best matched the emotion conveyed by the speaker (see Figure 2).

Figure 2. Example of an item of the Emotional Prosody Recognition task.

The twelve target items were preceded by four practice trials, in which children would get corrected in case of wrong answers.

3. Data Analysis

All analyses were run using R version 4.3.1 (R Core Team, 2023). Responses to the Ironic Prosody Recognition task were analyzed considering accuracy coded binomially (1 correct answer, 0 wrong answer). Accuracy was analyzed using mixed effects logistic regression entering as fixed effects type of prosody (ironic vs. sincere), mean-centered emotion recognition scores, their interaction, and age (mean centered). Participants and items were entered into the model as random intercepts, and prosody over participants as random slopes.

Moreover, we calculated the percentage of participants by class performing above the chance level. Considering that for each trial participants had a 50% probability of responding correctly and that the number of trials was 12, a performance was considered above chance if the score was greater than nine responses correct based on a binomial test (cumulative binomial probability = 0.02). Considering that participants might be biased towards one of the two possible responses, we also calculated d’ scores for each participant. D’ scores were analyzed with a linear model with age (mean centered) as predictor.

4. Results

As displayed in Figure 3, participants’ mean accuracy showed a clear developmental trend. The logistic regression analysis revealed that the interaction between the type of prosody and emotion recognition score was not significant; therefore, it was removed from the model without decreasing the model’s goodness of fit (χ²(1) = 0.88, p = 0.35).

Figure 3. Box plots showing participants’ mean accuracy in the Ironic Prosody Recognition task across classes, from preschool (P3, P4, P5) to school (S1, S2, S3, S4, S5) children in the two prosody conditions.

The reduced model confirms that accuracy increased as age increased (β = 0.63, SE = 0.15, z = 4.15, p < 0.0001), whereas the effects of the type of prosody and emotion recognition score were not significant (p = 0.14 and p = 0.97, respectively).

Figure 4 further specifies the age trend. The majority of children aged 3 and 4 (groups P3 and P4) performed below the chance level, and about 50% of 5- and 6-year-old children (groups P5 and S1) performed above the chance level. The increase in participants above the chance level continued at ages 7 and 8, and almost all children between the ages of 9 and 10 performed above the chance level.

Figure 4. Percentage of children who scored above the chance level in the Ironic Prosody Recognition task by class, from preschool (P3, P4, P5) to school (S1, S2, S3, S4, S5) children.

Confirmation of what has just been described also comes from applying signal detection theory, thus using d’ scores, a measure of sensitivity (Figure 5). The d’ score offers a robust index of discriminative ability, allowing in this case for a more precise evaluation of children’s sensitivity to prosodic cues. In fact, d’ is unaffected by response biases (APA, 2018), in our case, the presence of a possible bias towards a literal interpretation of the remarks (that is, the tendency to consider ironic remarks as sincere, and not the other way around). The correct detection of irony was considered a hit, and the correct detection of literal prosody was considered a correct rejection. Preschool children do not show high d’ scores (all are below 2, and when d’ = 0, the performance is at chance, whereas d’ = 3 is almost a perfect performance). It is from age 7 that some children display a high d’ score, and the developmental trend is clear and confirmed by the analysis (effect of age: β = 0.56, SE = 0.09, t = 6.18, p < 0.0001).

Figure 5. Box plots showing participants’ mean d’ score in the Ironic Prosody Recognition task across classes, from preschool (P3, P4, P5) to school (S1, S2, S3, S4, S5) children.

5. Discussion

We designed a new task to verify whether children relied on prosodic cues to recognize the ironic communicative intent. Evaluative remarks pronounced with sincere or ironic intonation were paired with two images depicting a literal and a reverse interpretation of the comment; the participants’ task was to select what the speaker was referring to with their remark. In the case of ironically pronounced remarks, the selection of the image corresponding to their reverse interpretation (i.e., the image of a hovel for the remark “That house is really beautiful”) was interpreted as the correct identification of the speaker’s ironic intent. As already mentioned, to perform this task, participants have to rely on prosodic cues alone: Remarks are presented in isolation, that is, without a preceding context that could bias a sincere or ironic reading. Moreover, the required answer (selection of the image the speaker is referring to) does not involve a meta-linguistic judgment (and can thus be performed even by young children) and is meant to capture the real grasping of ironic intent and not simply the detection of a non-canonical intonation.

With this task, we identified a clear developmental trend, demonstrating that the detection of ironic prosody is a complex ability that develops over time. Even if children reached an overall good accuracy (almost 70%) in the Ironic Prosody Recognition task already since age four, if we consider children who respond above the chance level, a full grasp of irony based on acoustic cues is reached only in the second year of primary school, after age seven, and this result is also confirmed by d’ scores, which are unaffected by potential response biases. These results cannot be directly compared to previous studies since they manipulated the presence of ironic or sincere prosody of remarks uttered after contexts that were biasing specific interpretations. In our task, on the other hand, remarks were presented in isolation, with prosody constituting the only cue for attributing a communicative intent to the speaker. Even if we used a completely different paradigm, our results appear to be in line with the findings of Ackerman (1983), Winner and Leekam (1991), Demorest et al. (1984), Sullivan et al. (1995), and Köder and Falkum (2021), who found that six-year-olds could not yet fully appreciate ironic prosody.

As attested in the literature, younger children tend to stick to the literal interpretation of statements, and, in our task, we could have expected a tendency to interpret the statements literally, selecting the image that corresponds to the literal interpretation of the remarks, independently of the prosodic contour with which they are pronounced. This would have resulted in a higher accuracy for sincere statements compared to ironic ones. As is evident from a visual inspection in Figure 3, the situation is not clear. Younger (three- and four-year-old) children perform better for sincere items than ironic ones, but this trend appears to be reversed for slightly older children, at five and six years of age. From age seven, the errors are much fewer and are equally distributed across sincere and ironic items. The data of three- and four-year-old children are consistent with the hypothesis that younger children tend to stick to a literal interpretation of statements (Winner, 1988/1997; Levorato & Cacciari, 2002; Köder & Falkum, 2020), even in the presence of a warm up introduction that alerted them to expect figurative expressions. As suggested by one of our reviewers, the training phase might have caused a sort of “non-literalist bias” for slightly older (five- and six-year-old) children, whose performance still appears to be vulnerable. As already highlighted in the discussion of d’ scores and above chance-level performance, it is only after age seven that children master ironic prosody.

The second goal of the study was to verify whether the detection of ironic prosody was related to the identification of emotions acoustically transmitted. In the logistic regression analysis we ran, the interaction between the type of prosody and emotion recognition score was not significant. This result could corroborate Filippova and Astington’s (2008) conclusion that the recognition of ironic intonation does not go hand in hand with the ability to decode emotions. Notice that in Filippova and Astington (2008), this conclusion was only indirectly obtained since they looked for possible correlations between the ability to recognize emotions and to attribute an ironic intent to a remark that was incongruent with the context. In our study, though, we used two prosody recognition tasks with a similar design since children had to associate an acoustic stimulus to an image (either an emoji depicting emotions or a picture corresponding to the literal or the reverse interpretation of the remark) based on the way it was pronounced. These data could then indicate that so-called affective prosody, which encompasses the acoustically realized manifestations of emotions and of the speaker’s attitude (Ross, 2000), is not acquired in a uniform way.

It has, nevertheless, to be acknowledged that the study presents some limitations. First of all, the sample of children is rather small, and more data should be collected to strengthen the conclusion. Moreover, we did not use counterbalanced lists, so each pair of images was associated with just one prosodic contour, and we cannot a priori exclude potential preferences for certain pictures. Another possible limitation of the study is the warm up. To familiarize children with the task, we presented instructions that explicitly alerted children that they should expect non-literal interpretations of the remarks; moreover, we provided two practice trials in which children were corrected in case they gave the wrong answer. It must be acknowledged that, as argued by Diehl and Paul (2009), these instructions invite children to draw attention to prosody, whereas in everyday communication, prosody is not highlighted, and thus this choice might have an impact on the ecological validity of the task; moreover, it might be difficult to determine whether children who respond correctly are indeed mastering prosody or have simply learned the demands of the task, given the initial feedback provided for the practice items might. Nevertheless, we believe that there are reasons to believe that these concerns do not impact the validity of the results. First of all, we obtained a clear developmental trend, with children improving their performance with age. Moreover, other tasks that evaluate prosody recognition include warm up items as well (Peppé & McCann, 2003; Van Lancker et al., 1989; a.o.), and Hoekert et al. (2007) explicitly suggest using practice items in these types of tasks. Finally, as highlighted in the introduction, we thought that children had to be alerted about the presence of non-literal interpretations to override their literalist bias (Winner, 1988/1997; Levorato & Cacciari, 2002; Köder & Falkum, 2020; Falkum & Köder, 2024).

Our material consisted of remarks uttered in Italian, and since it has been found that the acoustic realization of the ironic tone of voice is language dependent (see the discussion in Bettelli et al., 2024), testing children’s detection of irony through prosodic cues in other languages might provide a different picture. Even within the same language, it might be interesting to test different realizations of the ironic intonation, as in Glenwright et al. (2014) and Köder and Falkum (2021), to verify whether there are prosodic contours that are more easily detectable by children, independently of the presence of a previous context.

Moreover, it would be interesting to verify whether the ability to recognize the ironic tone of voice is linked to other linguistic or cognitive abilities. In particular, future studies might also test children’s ability to recognize linguistic prosody (such as the intonational difference between a statement and a question or the contrastive use of focus). If competence in ironic prosody correlates with this ability, we could hypothesize that attunement to prosody constitutes a unitary phenomenon. Moreover, since many studies found a correlation between sophisticated mindreading abilities (second-order Theory of Mind) and irony comprehension in typically and atypically developing children, it would be interesting to verify whether accuracy in the Ironic Prosody Recognition task correlates with these cognitive skills.

6. Conclusions

We proposed a new task to evaluate the detection of ironic prosody in children, testing for the first time their sensitivity to sincere and ironic intonation directly, without presenting a preceding context that biases the interpretation of the speaker’s communicative intent. The task requires the selection of the image that corresponds to what the speaker intends to refer to, and children’s responses are meant to capture the real understanding of sincere and ironic remarks, without requiring meta-linguistic judgments that might obscure children’s actual competence. We believe that this task might be profitably utilized with atypical populations to test their sensitivity to ironic prosody. In particular, there is debate in the literature about autistic individuals’ ability to recognize irony and ironic intonation. In Chevallier et al.’s (2011) study, autistic participants were not impaired in recognizing irony from acoustic cues; nevertheless, as highlighted by Deliens et al. (2018b), autistic participants might have resorted to intellectually based compensatory strategies to solve the task (see also Panzeri et al., 2022). The task we are proposing requires the selection of the image that the (sincere or ironic) speaker is referring to, without the need to understand complex instructions or provide verbal justification. We believe that this methodology could then investigate autistic participants’ ability to detect ironic intent relying on prosodic cues.

Participants with cochlear implants would be an interesting population to test as well. They are impaired in the recognition of emotions transmitted through the vocal channel (see Jiam et al., 2017, a.o.) and in irony comprehension (Panzeri et al., 2021), and assessing their ability to distinguish ironic intonation could clarify whether their problems in irony comprehension are linked to difficulties in attunement to expressive prosody or whether they are imputable to other cognitive, linguistic, or social difficulties.