Figurative language processing in autism spectrum disorders: A review

Impairments in the broader domain of pragmatics are considered to be a defining feature of Autism Spectrum Disorders (ASD). A challenging aspect of pragmatic competence is the ability to process nonliteral language. Interestingly, previous studies in figurative language comprehension in ASD have demonstrated conflicting results regarding participants' performance. The main scientific debate focuses on the underlying skills which facilitate processing of nonliteral speech in ASD. Namely, Theory of Mind (ToM), language abilities and Executive functions (EFs) are regarded as factors affecting autistic individuals' performance. This review addresses figurative language comprehension in ASD in light of the above three interpretive accounts. We reviewed data from recent studies in this field concluding that autistic children indeed encounter systematic difficulties in the processing of non‐literal language. Moreover, only ToM and verbal skills were found to correlate the most with figurative language comprehension in ASD. Notably, we found that differences related to research methodology and tasks' properties may have led to discrepancies between studies' results. Finally, we argue that future studies should encompass in their experimental design figurative comprehension tasks with minimal linguistic demands and also measures of ToM, verbal ability and EFs in order to shed more light in the independent contribution of those skills to the processing of nonliteral language in ASD.


INTRODUCTION
Autism is a neurodevelopmental disorder characterized by deficits in social interaction, verbal and non-verbal communication accompanied by limited interests and repetitive stereotypical behaviors (American Psychiatric Publishing, Inc., 2013).Although autism spectrum disorders (ASD) exhibits great heterogeneity in individuals' language abilities (Tager-Flusberg & Joseph, 2003), a common feature within the disorder is deficits in pragmatics (Järvinen-Pasley et al., 2008;Kalandadze et al., 2022).Even autistic individuals with high verbal or/and IQ skills appear to have difficulty in social use and understanding of language in context (Volden et al., 2009;Vulchanova et al., 2015).
Among the main deficits in pragmatics, the difficulty of processing figurative language stands out, emerging through scientific research as one of the defining features of ASD (Morsanyi et al., 2020).By the term figurative language we define a set of linguistic expressions whose interpretation is always nonliteral and serve the intention of the speaker to communicate something different from what he actually utters (Gibbs & Colston, 2012).Metaphors, similes, idioms, proverbs, humor and irony are some of the characteristic types of figurative language.In typically developing (TD) children the ability to understand nonliteral aspects of language appears in early childhood and is a skill that gradually develops through time to adulthood (Cacciari, 2014;Falkum et al., 2017;Rundblad & Annaz, 2010).In contrast, autistic children develop their pragmatic skills, including the understanding of figurative speech at a slower developmental rate (Melogno et al., 2012;Saban-Bezalel & Mashal, 2018;Whyte & Nelson, 2015).
A growing body of research (Adachi et al., 2004;Gold & Faust, 2010;Happé, 1993;MacKay & Shaw, 2004;Rundblad & Annaz, 2010) has documented the systematic difficulties of autistic individuals in figurative language processing, often exhibiting tendencies to interpretate figurative expressions in a literal manner.However, previous studies have yielded contradictory results regarding the source of these difficulties.Some researchers (Happé, 1993;Huang et al., 2015;Tager-Flusberg, 2000) have suggested that mentalizing difficulties which are attributed to deficits in Theory of Mind (ToM) explain the pragmatic language impairment in ASD, whereas others have challenged this association between ToM and pragmatic competence arguing that deficits in figurative language processing in ASD are neither specific nor universal and derive from inadequate skills in the key areas of language such as morphosyntax and vocabulary (Brynskov et al., 2016;Gernsbacher & Pripas-Kapit, 2012;Geurts et al., 2019;Kalandadze et al., 2018;Norbury, 2005;Whyte et al., 2014).Furthermore, several neuroimaging studies (Colich et al., 2012;Wang et al., 2006;Williams et al., 2013), attribute the causes of these difficulties to executive dysfunction in ASD which results in deficient semantic integration ability (Eigsti, 2013;Saban-Bezalel & Mashal, 2018;Vulchanova et al., 2015).Semantic integration skills allow us to obtain information from multiple sources, create connections and judge their plausibility and relevance (Gold et al., 2010).From the above, it becomes clear that many different factors may influence the processing of figurative language in ASD.The studies that have been conducted so far are on the one hand limited in number (Melogno et al., 2012) and on the other hand are not directly comparable as they either examine different types of figurative language or exhibit many differences in the research methodology they follow (Vulchanova et al., 2015).The exact source of the welldocumented difficulties in the processing of figurative language in ASD remains a puzzle.Only few studies (Saban-Bezalel et al., 2019;Saban-Bezalel & Mashal, 2019;Tzuriel & Groman, 2017) have addressed the comprehension and production of figurative language in relation to ToM, language, and cognitive abilities of autistic individuals.
The aim of this review is to address figurative language comprehension in ASD in light of the three main interpretative theories (ToM, executive dysfunction and impaired language skills) that have been put forward to explain these deficits.In order to do so, we will critically review evidence obtained by recent studies in this domain.

ACCOUNTS OF FIGURATIVE LANGUAGE DEFICITS IN ASD Theory of mind and figurative language comprehension in ASD
The ToM refers to a set of mentalizing skills that allow individuals to attribute, recognize, understand and interpret the thoughts, intentions, feelings and mental states of others (Nguyen & Astington, 2014), which allows one to predict behaviors and adjust their own accordingly (Hoogenhout & Malcolm-Smith, 2017).Autistic children exhibit individual differences with regards to the acquisition of ToM (Andreou & Skrimpa, 2020).Deficient ToM in autistic individuals is associated with social, behavioral and communication impairments (Begeer et al., 2010).
Mentalizing ability is considered an important skill for the processing of metaphorical speech, as decoding a figurative expression requires the understanding of mental state (Saban-Bezalel & Mashal, 2019) and the intention of the speaker (Rapp & Wild, 2011;Saban-Bezalel & Mashal, 2018).Therefore, impairments in ToM could lead to difficulties in processing figurative language in nontypical populations.This hypothesis has been supported by many researchers (Happé, 1995;Whyte et al., 2014) investigating the link between ToM and pragmatic skills in autism, pointing to the deficit in ToM as the root cause of the failure of autistic individuals to grasp the figurative meaning of speech.Happé (1993), in her pioneering study, was the first to address the association between ToM and figurative language comprehension in ASD, concluding that first-order ToM ability was sufficient for autistic children to comprehend metaphors, while processing irony required second-order ToM skills.First-order ToM is the ability to reason about another person's beliefs, while secondorder ToM refers to a more complex ability that allows us to infer what a person believes about another person's thoughts (Wimmer & Perner, 1983).The prevalence of ToM as a predictor of figurative language comprehension is further supported by recent studies that examined irony (Saban-Bezalel et al., 2019) and idiom comprehension in autistic children (Whyte et al., 2014).

Language skills and figurative language comprehension in ASD
Unlike researchers who suggest that the well-attested problems in figurative language comprehension in ASD are related to deficient ToM and are common among autistic individuals, others (Gernsbacher & Pripas-Kapit, 2012) argued that these deficits should not be considered as a defining feature of ASDs, since they are not observed in all individuals in ASD and actually reflect low levels of language competence.
More specifically, according to this account, the literalism trend which characterizes some autistic individuals arises from impairments in structural language abilities such as vocabulary (Norbury, 2005) and syntax (Whyte et al., 2014).In other words, autistic individuals who display poor language skills are expected to have difficulty in processing both literal and figurative speech, with the latter being considered as one of the most complex and demanding aspects of language (Martin & McDonald, 2003).The main argument of the researchers who adopt this account is that in studies in which autistic individuals were compared to TD controls with similar language skills, no statistically significant differences were found in the ability to understand figurative utterances (Norbury, 2005;Saban-Bezalel & Mashal, 2015;Whyte et al., 2014).Norbury (2005), for example, found that the most important predictor of metaphor comprehension was semantic competence and inferred that ToM abilities are not sufficient to facilitate the comprehension of idioms in autistic individuals.Of note, this theoretical framework is further supported by Kalandadze et al. (2018), who recently performed a meta-analytic review of figurative language comprehension in ASD, highlighting the crucial role of individual language abilities in processing different types of figurative speech, including idioms, irony, sarcasm, metaphors, metonymies and similes.

Executive functions and figurative language comprehension in ASD
Another fairly popular account that seeks to explain autistic individuals' deficits in processing figurative language is the executive dysfunction hypothesis.EFs are an umbrella term that encompasses a wide range of higher cognitive abilities such as cognitive flexibility, inhibition, working memory, updating, reasoning, planning and problem solving which are essential for cognitive control (Andreou et al., 2020).These complex skills play an important role in everyday life, as they allow us, on the one hand to plan, organize and control cognitive activities when solving problems (Anderson, 1998;Kouklari et al., 2018) and on the other hand to adjust our behavior and evaluate the results of our actions (Sun et al., 2017).
Executive dysfunction is a key feature of ASD (Hill, 2004) and is present in both children and adolescents with high-functioning autism (see Lai et al., 2017 for a related recent meta-analysis).Executive dysfunction has been associated with pragmatic difficulties and more specifically deficits in figurative language processing in atypical populations (Cummings, 2009).Indeed, research in TD individuals has revealed that EFs are significantly related to the ability to process metaphors (Benedek et al., 2014;Dietrich, 2004;Iskandar & Baird, 2014).In particular, EFs such as working memory and inhibition predict the comprehension of metaphors (Chiappe & Chiappe, 2007).
According to Landa and Goldberg (2005) cognitive flexibility, that is, the process of switching between the two different meanings (literal and conveyed meaning) of a figurative expression can be extremely challenging for autistic individuals.This was confirmed by Mashal and Kasirer (2011), who found that cognitive flexibility significantly predicts the ability to comprehend and generate metaphors in autistic children.In line with previous findings a recent study (Kasirer & Mashal, 2016) argued that EFs such as strategic search, response initiation, monitoring, shifting, and cognitive flexibility as measured through a phonemic fluency task contributed to the generation of novel metaphors in autistic participants.
Overall, the exact source of figurative language processing impairments in ASD remains controversial.While many scholars attribute these deficits mostly to cognitive traits or language skills of ASD individuals, further research is warranted to support a single prevailing account.

FIGURATIVE LANGUAGE PROCESSING IN ASD
The ability to comprehend and generate nonliteral (or nonexplicit) language is one of the most complex and demanding stages of individual's language development (Glucksberg, 2001).The acquisition of this ability relies on both linguistic and cognitive abilities (Dietrich, 2004), in addition to meta-pragmatic competence (Bernicot et al., 2007).Typically developing individuals begin to understand nonliteral forms of speech at preschool age and gradually develop this ability, culminating in adulthood (Selimis & Katis, 2010;Semrud-Clikeman & Glass, 2010).In contrast, autistic individuals appear to have a deviant developmental trajectory, concerning the acquisition of figurative language processing competence.Specifically, they demonstrate the ability to understand different kinds of nonliteral speech at an older age compared to neurotypical individuals.Nevertheless, they manage to further develop their corresponding skills as they get older (Melogno et al., 2012;Vulchanova et al., 2012;Whyte & Nelson, 2015).Van Herwegen and Rundblad (2018) in a cross-sectional and longitudinal study found that autistic participants showed poorer comprehension of novel metaphors and metonymies than their TD controls, across all ages.However, it is worth mentioning that despite the wealth of developmental studies available in the field of figurative language comprehension in TD populations, limited relevant longitudinal studies have been conducted in autistic individuals (Van Herwegen & Rundblad, 2018).
Thus, in order to draw firm conclusions about the development of figurative language comprehension in ASD, it is imperative that further longitudinal studies are conducted.According to Vulchanova and Vulchanov (2022), a vast majority of studies have focused on the comprehension of nonliteral language in autistic individuals (Melogno et al., 2017;Pexman et al., 2011;Zheng et al., 2015), while only a few studies in figurative language production have been undertaken (Kasirer & Mashal, 2014, 2016).
As the interpretation of different types of figurative speech assigned by autistic individuals is nonliteral and these types differ significantly in their structural characteristics (Vulchanova et al., 2015), grades of familiarity, decomposability, and transparency (Lada et al., 2023) and as to the mechanisms required for their decoding (Wilson & Sperber, 2012), it becomes clear that processing nonliteral linguistic expressions can be particularly demanding for atypical populations, such as autistic individuals (Vulchanova & Vulchanov, 2022).This claim was confirmed by findings in recent studies (Baixauli-Fortea et al., 2019;Gold et al., 2010), as well as in narrative and meta-analytic reviews in this domain (Kalandadze et al., 2018;Melogno et al., 2012;Morsanyi & Stamenkovi c, 2021), which have highlighted the deficits of autistic individuals in the processing of metaphorical speech.Notably, impairments in pragmatics, including the comprehension of implicit meaning are also observed in high-functioning autistic (HFA) individuals, despite having high verbal skills (Landa, 2000).Conflicting results have been reported in the literature (Chouinard & Cummine, 2016;Norbury, 2004;Pexman et al., 2011) showing that HFA individuals decode figurative expressions similarly to their TD controls.Nonetheless, autistic participants consistently exhibit longer reaction times in processing metaphors (Chahboun et al., 2017;Giora et al., 2012;Gold et al., 2010) and detecting ironic remarks (Colich et al., 2012), regardless of their performance in figurative language comprehension tasks.
We devote the rest of this chapter reviewing evidence from recent studies in figurative language processing in ASD.Specifically, we will analyze data from metaphor, idiom and irony comprehension studies in ASD, which have been conducted in the last decade.These three forms of nonliteral language have garnered the interest of researchers in this scientific field.The review encompasses a total of nine studies that fulfilled our set of inclusion criteria: (1) studies that examined participants with no intellectual disabilities (2) autistic and TD groups matched based on chronological age or verbal ability and (3) studies that examined participants' performance on at least one of the following domains: language ability, EFs, ToM.

Metaphors
Metaphors are one of the most representative and commonly used types of non-literal language in everyday communication (Steen et al., 2010).They are characterized by a dissociation between their literal interpretation and their implicit-metaphorical meaning (Carston, 2010).They are verbal formulations (e.g., My life is a shipwreck) which introduce an implicit comparison between two seemingly different concepts, the concept-target (i.e., life) and the concept-source (i.e., shipwreck) (Berber Sardinha, 2008;Mashal & Kasirer, 2011).Metaphors are categorized in terms of how they are perceived into sensory (introduce a comparison which is perceptible through our senses) and nonsensory (introduce a comparison which is not perceptible through our senses) (Winner, 1988) and in terms of their degree of familiarity are classified to novel metaphors (innovative, prototype) and conventional metaphors (well-known, familiar) (Bowdle & Gentner, 2005).The different characteristics of each type of metaphor imply different processing modes (Chahboun et al., 2016).Deficits in comprehension of metaphors can pose serious barriers to language understanding and effective social communication (Kalandadze et al., 2019).According to Saban-Bezalel and Mashal (2018) task properties such as response format, type of stimuli and linguistic features (i.e., type of metaphor, existence of context and stimulus modality) can have a significant impact on the participants' performance and on the configuration of differences between groups.Previous research has established that autistic individuals experience significant difficulties compared to TD controls in the comprehension of metaphors (Rundblad & Annaz, 2010;Van Herwegen & Rundblad, 2018).Whyte and Nelson (2015) in a cross-sectional trajectory study examined the development of nonliteral language in 27 HFA and 69 TD children (ages 5-12 years).To assess figurative language processing, they used a comprehensive assessment of spoken language (CASL) subtest that included metaphors, indirect requests, and sarcasm.The participants also underwent vocabulary, syntax, and ToM measures in order to be determined whether these factors alongside to age could predict performance on a nonliteral language comprehension test.The results demonstrated that autistic children exhibited a slower developmental rate in terms of age compared to TD children in metaphor comprehension.However, when syntax, vocabulary and ToM skills were considered no statistically significant differences were observed in the nonliteral trajectories of the two groups.The study found that both ToM and structural language abilities significantly predicted metaphor comprehension in the autistic and TD group.In support of this, Huang et al. (2015) provided strong evidence that ToM competence and verbal ability (receptive vocabulary scores) of autistic participants correlated with metaphor comprehension performance.
The results of this research are quite contrary to those shown in a previous cross-sectional study by Rundblad and Annaz (2010) which used a similar developmental trajectory analysis to investigate the comprehension of metaphors and metonyms in autistic children.The findings indicated that neither age no vocabulary were significantly associated with the ability to process metaphors in autistic children.
Further support in the role of vocabulary as a significant predictor of metaphor processing is given by Kasirer and Mashal (2016), who examined the comprehension and generation of conventional and novel metaphors in 34 autistic children and 39 TD age-matched controls (aged 9-16).In this study, participants were administered a number of tests that measured vocabulary and EFs (phonemic and semantic fluency, Trail Making Test), to determine which of these factors affect the comprehension and production of metaphors in children across both groups.The results showed that vocabulary was the only factor that significantly predicted metaphor comprehension performance in autistic participants.Moreover, although EFs did not correlate with novel or conventional metaphor comprehension, they proved to be an important indicator of novel metaphor production in autistic children.Concerning the participants' performance in the metaphor comprehension test, while autistic children seemed to understand novel metaphors similarly to their TD counterparts, conventional metaphors were quite difficult to process.This finding is inconsistent with previous studies which have suggested that conventional metaphors might be easier to comprehend due to their familiarity (Giora, 1997;Vulchanova et al., 2012).Remarkably, while autistic children were not able to produce enough conventional metaphors in the metaphor production test, they managed to generate more novel metaphors compared to TD children.These results reveal that autistic children appear to face more difficulties in the comprehension of nonliteral language compared to production, which seems to be intact (Vulchanova & Vulchanov, 2022).However, given the paucity of figurative language production studies in ASD, such results cannot be easily generalized in order to draw firm conclusions.Furthermore, such findings can be perceived as evidence of a unique creativity that characterizes autistic individuals compared to those of typical development (Kasirer & Mashal, 2014;Liu et al., 2011).Kasirer and Mashal (2016) in their study suggest a rather interesting explanation for the performance pattern of autistic participants in the metaphor generation task, reinforcing the broader discussion around creativity in ASD.Namely, the authors attribute the creativity in metaphor production to the deficient ToM that leads autistic individuals to adopt a unique and at the same time original way of thinking, as they are unable due to their mind-blindness to be influenced by the thoughts of others and tend to think unconventionally (Happé, 1999).
A recent study by Cardillo et al. (2021) aimed at investigating the underlying roles of ToM and EFs in the pragmatic language abilities of 73 autistic children and adolescents and 70 TD controls.Participant's ability to comprehend nonliteral language was evaluated through a metaphor task.The task comprised of two subtests (verbal and pictorial metaphors).Competence in ToM was measured by a verbal (Nepsy-II) (Korkman et al., 2007) and a nonverbal test (Reading the Mind in the Eyes Test-Children's Version) (Baron-Cohen et al., 2001).The study also enclosed measures for EFs such as updating (assessed through a computer-presented task), in addition to inhibition and switching (assessed through two subtests of the Nepsy-II).The sample of autistic children and adolescents lagged behind TD peers in the comprehension of metaphors, as well as in the ToM and EFs scores.These results reinforce the general belief that deficits in pragmatics and particularly in figurative language comprehension is a pervasive characteristic of autistic individuals (Baixauli-Fortea et al., 2019).One of the key findings of this study was that among ToM and EFs it was only ToM performance that contributed to the ability of autistic participants to comprehend metaphors.This finding is in line with the results of prior studies which have highlighted the critical role of mind-reading skills in the processing of figurative language (Huang et al., 2015;Whyte & Nelson, 2015), as well as in the general pragmatic competence of autistic individuals (Rosello et al., 2020).
The overall results of studies investigating metaphor processing in autistic children indicate that both ToM and language ability (i.e., vocabulary) are the factors that have a decisive impact on the ability of autistic children to understand metaphors (Table 1).On the contrary, studies have shown that EFs contribute mainly to the metaphor production abilities of autistic children.However, it should be noted that the above studies did not include measures for all these variables.Therefore, it is necessary for future studies in the field to incorporate in their experimental design tests for ToM, EFs, and verbal ability in order to be able to draw robust conclusions about the independent effect of these factors on the comprehension and production of metaphorical language in autistic children.

Idioms
Idioms are complex multiword linguistic expressions that convey a nonliteral meaning that is noncompositional, it does not derives from the combination of the individual literal meanings of its consisting words (Glucksberg, 1991;Titone & Connine, 1994).For example, the lexical units in the popular idiomatic expression "He is crying over spilt milk" have a nonliteral interpretation (i.e., "He is complaining about a mistake/ failure/loss"), that cannot be easily deciphered through the semantic analysis of the expression.According to Lada et al. (2023), idiomatic expressions vary by degree of ambiguity (potential overlap between lexical semantics and metaphorical concept), familiarity (frequency of use), transparency (proximity between the literal and nonliteral meaning) and decomposability (easy to decode).Idioms are fixed expressions whose nonliteral meaning is stored in semantic memory and their processing requires two parallel operations, namely, their direct retrieval from Mental Lexicon (internalized knowledge of lexical items' properties) (Cacciari, 2014) and the synthesis of their structural components in order to decode their nonliteral meaning (Vulchanova et al., 2011).Idioms are syntactically and semantically fixed expressions that operate as one unit (Schapira, 1999;Swinney & Cutler, 1979).According to Holyoak (2019), the processing of idioms relies more on language skills compared to metaphors.Given the semantic opacity of idiomatic expressions and their non-compositional nature, they are regarded as being a complex and demanding type of nonliteral language for both TD and autistic individuals (Morsanyi & Stamenkovi c, 2021).It is rather surprising the fact that while the acquisition and processing of this particular type of figurative language has been studied extensively in typical populations, research into idiom comprehension among autistic participants is relatively scarce.Whyte et al. (2014) studied idiom processing in HFA children aged 5-12 years who were compared with two control groups consisting of TD children matched in chronological age and syntactic ability.The authors focused on the factors that predict performance in the comprehension of idiomatic expressions.Thus, in addition to the idiom comprehension task, they administered tests that measured syntax, vocabulary, and higher-order ToM ability (i.e., the "strange stories" (Happé, 1994) and "Reading the Mind in the Eyes" (Baron-Cohen et al., 2001) task).Higher-order ToM abilities refer to advanced age-related mindreading skills which onset is placed approximately at the age of 5 years (Osterhaus & Bosacki, 2022).
In this study, the researchers used a language demanding idiom comprehension task which presented idioms in the context of a short text and required participants to provide a verbal explanation of idioms.The results showed that although the age-matched TD group outperformed the autistic group in idiom comprehension, this was not the case for the syntax-matched group which performed similarly to autistic peers.The findings also demonstrated that both higher-order ToM and syntax skills have a close correlation with idiom comprehension in autistic children.Furthermore, the authors argued that, besides age and mental ability, verbal skills and particularly syntax should also be a criterion for matching groups in studies examining figurative language processing in ASD.These findings are consistent with another study (Saban-Bezalel & Mashal, 2015) which also found no statistically significant differences in idiom comprehension between autistic and TD adults with similar language abilities.Results of the Whyte et al.'s (2014) study should be interpreted with caution in terms of the close association which was observed between ToM competence and idiom comprehension performance mainly due to the fact that the use of highly verbal measures of ToM could have affected findings.Namely, when linguistically demanding ToM tasks are used we cannot draw safe conclusions about whether the participants' performance is a result of their level of mentalizing abilities or their general language skills (see Marinis et al., 2023 for a relevant discussion).
The above results differ significantly from those obtained by a recent study of Chahboun et al. (2016).This study dealt with the ability of autistic children to comprehend idioms and with the effect of stimulus modality and properties of idiomatic expressions (e.g., the degree of transparency) to idioms processing.The participants were 25 high-verbal autistic children, who were matched with 19 TD children on age, IQ scores and verbal ability (reading comprehension, receptive grammar and semantic skills).The results revealed that although the autistic group possessed similar language skills as the comparison group of TD children, they showed poorer performance in idiom comprehension.These results challenge the accounts which correlate the ability of autistic individuals to process nonliteral forms of language to their general verbal ability (Gernsbacher & Pripas-Kapit, 2012; Kalandadze et al., 2018).In addition, such findings are inconsistent with the results of previous studies that matched participants based on their structural language skills (Norbury, 2004;Saban-Bezalel & Mashal, 2015;Whyte et al., 2014).We speculate that these discrepancies in the results across studies may arise from task effects.Specifically, Chahboun et al. (2016) used a low-verbal idiom comprehension task with visuals as a supportive context, in contrast to prior studies (Norbury, 2004;Whyte et al., 2014) which utilized more verbally-loaded measures of idiom comprehension.The difficulty of the autistic group to understand idiomatic expressions in the low-verbal idiom comprehension task, indicates that this process probably required the use of alternative underlying skills or cognitive mechanisms beyond language ability.However, this study did not include measures of ToM ability or executive functioning.Another interesting result in Chahboun's et al. (2016) research was that idiomatic expressions, which are more transparent and whose literal meaning can therefore be more easily deduced from the context, are easier to process by both autistic and TD participants.This finding is in agreement with a previous study (Giora et al., 2012), which called attention to the important role of the expression's degree of familiarity and transparency as significant predictors of performance in the comprehension of nonliteral language in ASD.
A recent study by Saban-Bezalel and Mashal (2019) shed light on the independent effect of ToM, vocabulary and EFs in the ability of autistic children and adolescents to process idioms.The sample comprised 47 participants (23 ASD and 24 TD) who were matched on age, vocabulary, and EFs ability.The results demonstrated that the autistic group encountered more problems in the comprehension of idioms compared to the TD group.This finding is in accordance with the results of Chahboun's et al. (2016) and challenges the argument that when autistic and TD individuals with similar language skills are compared, they performed equally in the comprehension of nonliteral language.Moreover this study provided evidence that the two groups of participants relied on different underlying mechanisms during idioms processing.Specifically, it was found that for the autistic group, vocabulary was the sole significant predictor of performance in the idiom comprehension task, while for the TD group, cognitive flexibility was the variable that significantly predicted participants' performance.In contrast, ToM ability was not found to be significantly correlated with idiom comprehension in either of the two groups.Although this finding at a first glance confirms the accounts that link the ability of autistic individuals to comprehend nonliteral forms of language to their broader language skills, it was unexpected that autistic participants in Saban-Bezalel and Mashal (2019) exhibited greater difficulty in the processing of idioms, while they had similar verbal abilities to their TD peers.The authors argued that this result may be assigned to some of the properties of the idiom comprehension task.Notably, in Saban-Bezalel and Mashal (2019), the fact that the idiom comprehension test lacked a supporting context, may have increased the level of difficulty for autistic participants, as they had to decipher the figurative meaning of idioms based solely on retrieval from the Mental Lexicon or alternatively on the semantic processing of the idiomatic expression.In a previous study (Vogindroukas & Zikopoulou, 2011) where idioms were presented in isolation the autistic children were also less accurate compared to the control group.Yet, in Whyte's et al. (2014) study where the autistic group demonstrated comparable performance to the TD group, the idiom comprehension task that was used included a supporting context which probably facilitated the participants through idiom processing.Several studies have highlighted the decisive and supportive role of the context in the process of idioms' semantic analysis (Le Sourn-Bissaoui et al., 2009;Levorato, 1999;Norbury, 2004).On the other hand, it is important to mention that other studies (Greimel et al., 2012;L opez & Leekam, 2003) have pointed out that autistic individuals face significant difficulties in processing and synthesizing semantic elements enclosed in context when performing in verbal tasks.
T A B L E 2 An overview of the selected research studies in idiom comprehension in ASD for this review.To summarize, the results of studies investigating the comprehension of idioms in ASD are far from conclusive (Table 2).There have been studies where autistic children showed similar performance to TD children in the comprehension of idiomatic expressions and other studies in which participants of typical development outperformed their autistic counterparts.Regarding the factors affecting the performance of autistic participants on idiom processing tasks, discrepancies in the results were also reported.In particular, structural language skills (i.e., syntax and vocabulary) were proved to be significantly related to idiom comprehension ability, but there were conflicting findings for the contribution of ToM.Of note, EFs seem to predict the comprehension of idioms only in TD individuals.However, the observed discrepancies in the results of the studies discussed above could be due to factors such as the experimental design and task characteristics.

Irony
Verbal irony is perhaps the most characteristic type of nonliteral speech in which the speaker's communicative intent directly contradicts the literal meaning of his words (Gibbs et al., 2014;Wilson & Sperber, 2012).Thus, if during a rainy day someone says "Great weather", he uses verbal irony in order to communicate his disappointment about the weather conditions.In the example discussed above, the choice of words does not correspond to the critical attitude of the speaker, as he is actually uttering something completely opposite to what he believes.This specific type of figurative language is present in dayto-day communication as a verbal mean of expressing humor, criticism, or even endorsement (Roberts & Kreuz, 1994).Therefore, its use serves multiple communicative purposes during social communication and interaction (Dews & Winner, 1995).With respect to the content and the valence of the ironic message we discern two subtypes of verbal irony, namely ironic criticism and ironic praise (Filippova, 2014).In ironic criticism, a word with a positive meaning is used to criticize (e.g., someone just spilled a cup of coffee on Mary and she says: "Good job"), while in ironic praise a word with a negative connotation is used as a mean of approval (e.g., Mary says "bad luck" after winning the lottery).
Research findings suggest that ironic criticism is more easily perceived by TD individuals than ironic praise (Hancock et al., 2000;Whalen & Pexman, 2010).The mastery of verbal irony requires the individual to possess both developed pragmatic skills and sophisticated mindreading abilities (Panzeri et al., 2022).Several studies have provided evidence of significant correlation of first and second-order ToM abilities with irony comprehension (Nilsen et al., 2011;Sullivan et al., 1995).Moreover, in order to decode the speaker's ironic intent, the listener first needs to recognize and then interpret important cues of irony such as context, prosody and gestures (Burgers & van Mulken, 2017).According to Rivière et al. (2018) the key element in detecting verbal irony is the excessive antithesis between the ironic statement and reality (i.e., context).Understanding the different types of verbal irony has proven to be an extremely challenging task for both typical and atypical populations (Pexman et al., 2019).The multitude and complexity of the underlying skills required to process verbal irony might be the reason for its late acquisition by TD children (Filippova & Astington, 2010).Although some scholars have argued that early signs of irony appreciation may be present in early childhood, around 3-4 years of age (Loukusa & Leinonen, 2008;Recchia et al., 2010), a growing body of literature indicates that TD children begin to appreciate the ironic intent of a speaker around the age of 5 (Angeleri & Airenti, 2014) developing this ability throughout childhood (Aguert et al., 2016;Filippova & Astington, 2008).In the case of autistic children, even though it is widely accepted that they exhibit a delayed developmental course in irony acquisition compared to TD children, the results of studies that have focused on the comprehension of verbal irony are quite inconclusive.There are findings which demonstrate that autistic participants are distinctively deficient at processing verbal irony compared to their TD peers (Adachi et al., 2004;MacKay & Shaw, 2004;Wang et al., 2006), however results exist that indicate that autistic participants are able to perform similarly to TD groups in irony comprehension tasks (Colich et al., 2012;Glenwright & Agbayewa, 2012;Pexman et al., 2011;Williams et al., 2013).Huang et al. (2015) examined the effect of ToM and language skills in the comprehension of five types of nonliteral language, including irony.The study sample consisted of 50 HFA and 50 TD children matched by chronological age, gender and receptive vocabulary ability.The results showed that autistic participants were less accurate in irony comprehension compared to their TD peers.Additionally, statistical analysis demonstrated a correlation between ToM competence and the comprehension of ironic utterances, but only in the case of autistic participants.Nevertheless, it should be noted that the study showed that the performance of the autistic group in the comprehension of verbal irony, did not correlated with a specific level of ToM ability.Indeed, according to the data, no differences were found in the comprehension of verbal irony between autistic participants who had successfully completed first and second-order ToM tasks.This finding is opposite to Happe's (1993) results according to which the ability to process verbal irony requires second-order ToM skills.With respect to the role of receptive vocabulary in the ability of autistic children to understand verbal irony, the results showed no correlation.
The prevalence of ToM as a predictor of irony is further supported by Saban-Bezalel et al. ( 2019), who addressed irony comprehension in autistic individuals in association with ToM, vocabulary, and EFs (i.e., mental flexibility) as mediating factors.The study included a total of 40 participants, among which 20 ASD and 20 TD children and adolescents matched on age, language ability, mental flexibility, and second-order ToM competence.The results revealed that whereas the autistic group performed well above chance in the irony comprehension task, their scores fell behind those of their TD controls.Quite interestingly, the between group differences were no longer present, when controlling for participants' scores in Hinting Test which is a measure of ToM competence focusing on the detection of intentions.This finding reinforces the notion that ToM abilities serve a facilitating role in the comprehension of nonliteral language.Instead, the study found no correlation between expressive vocabulary and irony comprehension ability in neither autistic nor TD group.Regarding the relationship between EFs and irony comprehension, the results indicated that mental flexibility predicted irony comprehension only in TD participants.This corroborates the findings of previous research (Landa & Goldberg, 2005) which examined the underlying role of EFs in the processing of nonliteral language in ASD and found no significant correlations.A closer examination of the results reveals that the lower performance of autistic participants compared to their TD peers may be attributed to some of the irony comprehension task's features.Namely, in the present study the utilized task besides providing visuals to support irony comprehension, was linguistically demanding as it required verbal answers in open questions.Also, the comic strips in the irony comprehension task, as they were not presented auditorily to the participants, did not encompassed the clue of prosody (i.e., ironic intonation) which is thought to be an important contextual cue in the processing of verbal irony (Milosky & Ford, 1997).However, prior studies (Rutherford et al., 2002;Wang et al., 2006) have reported that autistic participants face difficulties in the processing of prosodic cues.MacKay and Shaw (2004), who also used a language demanding task with less contextual cues to test irony comprehension, found that autistic children lagged behind TD peers in the comprehension of ironic statements.On the contrary, Glenwright and Agbayewa (2012), using a computer-based task with less pragmatic demands and forced-choice questions, showed that autistic participants were able to perform similarly to their TD counterparts in the comprehension of verbal irony.
In a recent study conducted by Panzeri et al. (2022) the authors compared the performance in irony processing between 26 autistic children and two comparison groups of TD children matched on age and language ability.The study placed emphasis on the identification of the factors that predict irony comprehension performance.Hence, the study design encompassed ToM measures and the grammar comprehension task of the BVL (Marini et al., 2015) which assesses the comprehension of several morphosyntactic features.The irony comprehension task which was computerized, included ironic criticisms and ironic compliments providing both visual and prosodic cues to facilitate irony comprehension.The results demonstrated that the autistic participants did not exhibit adequate performance in the comprehension of both subtypes of verbal irony, since they significantly lagged behind both groups of TD participants.This result is in contrast with the findings from prior studies which argued that autistic children performed at the same level to TD peers in terms of irony comprehension (Colich et al., 2012;Pexman et al., 2011).In addition Panzeri et al. (2022) found close correlation between mentalizing abilities and irony comprehension in autistic participants.In particular, the results showed that both first-order and second-order ToM skills predicted comprehension of ironic criticisms, while the processing of ironic compliments required second-order ToM competence.The data also indicated that performance in the grammatical comprehension task related only to the processing of ironic criticisms.It should be noticed that the most significant finding of this study was the detection of two completely different profiles of autistic participants with respect to the comprehension of verbal irony.The at-ceiling achievers (six out of 26 autistic children had perfect scores in the irony comprehension task) and the at-bottom performers (18 out of 26 autistic participants showed poor performance in the irony comprehension task).Regarding the factors explaining this bimodal distribution, the authors found that while the at-ceiling achievers performed better than the at-bottom performers on all the included language and cognitive tasks and were also marginally older than them, only ToM performance was significantly related to between-group differences.Nevertheless, this result does not necessarily proves that ToM competence is the key factor that distinguished the two groups, as among the at-ceiling achievers some (two out of six) failed to pass both first and secondorder ToM tasks.We speculate that beyond mindreading skills, the source of this bimodal distribution may lies in other cognitive traits (i.e., EFs) which were not examined in this study or alternatively in external factors such as the participants' degree of exposure in verbal irony cues in their broader family and social environment.Indeed, previous studies (Pexman et al., 2009;Recchia et al., 2010) found that social context is a variable that can significantly enhance the production and comprehension of ironic language in TD children.The extent to which this is the case also for autistic children remains to be investigated.Such findings raise the issue of the great heterogeneity in ASD, emphasizing on the necessity of a more in-depth investigation of individual profiling among autistic participants (see Silleresi et al., 2020 for similar results and Andreou & Skrimpa, 2022 for a relevant review).
Taken together, the findings of the studies reviewed above suggest that irony is a quite challenging phenomenon of nonliteral language for autistic participants (Table 3).ToM seems to be one of the most significant factors underlying the ability of autistic individuals to comprehend different forms of verbal irony.Instead the mediating role of structural language abilities (i.e., vocabulary and grammar) is still a matter of debate, due to conflicting data.Existing studies in the comprehension of verbal irony in ASD have not yet provided evidence of relations between EFs and the ability to process irony.However, only few studies on irony have included measures of EFs in their experimental design (Deliens et al., 2018;Saban-Bezalel et al., 2019).

CONCLUSIONS
The acquisition of figurative language is an important milestone in the language development of individuals, playing a decisive role in the configuration of their general pragmatic competence.However, processing different subtypes of nonliteral speech has proven to be an extremely difficult task for both typical and atypical populations, including ASD.This is probably due to the fact that such abilities require from individuals to possess and coordinate complex cognitive and linguistic abilities.The present review attempted to explore the empirical landscape in the processing of figurative language in ASD with a focus on the main underlying skills that have been suggested to influence the development of nonliteral speech in autistic individuals.As we can conclude from the studies reviewed above, autistic children exhibit substantial difficulties in the processing of figurative speech compared to TD controls.In addition, based on the results of the included studies, the factors that affect the mastery of different types of figurative language across studies seem to be divergent.Thus, while the ToM and language ability seem to affect the comprehension of metaphors, in the case of idiom processing, structural language abilities (i.e., syntax and vocabulary) seem to play a decisive role.Finally, in the comprehension of verbal irony, mentalizing abilities play a key role in the performance of autistic participants.It thus seems that each type of nonliteral speech has it owns unique features, requires different processing mechanisms.Remarkably, EFs and specifically cognitive flexibility were found to correlate only with the production of metaphors in ASD.Nevertheless, as already mentioned the majority of studies in this domain, including those selected for this review, did not examined the relation between EFs and the processing of figurative language.This gap does not allow us to draw firm conclusions about the facilitating role of EFs in the comprehension of nonliteral language in ASD.An important issue that has emerged through this review is the fact that the conflicting data across studies may have stemmed from differences in research methodology and the properties of figurative language comprehension tasks (see Kalandadze et al., 2019 for a similar conclusion).Namely, using linguistically demanding tasks raises the question of whether the results reflect the participants abilities in nonliteral language or their verbal competence per se.Future studies should adopt measures of figurative speech comprehension that set less language demands on the participants.For example, lowverbal tasks with supportive contextual cues could reduce the difficulty of the tests pinpointing the exact roots of figurative language deficits in ASD.Finally, contrary to the conclusion of the meta-analytic review of Kalandadze et al. (2018), which attributed the commonly attested figurative language impairments of autistic individuals to their broader language deficits, we believe that decoding the source of these difficulties is not a one-dimensional issue.With respect to the heterogeneity that characterizes ASD, we argue that ToM, core language skills and EFs are factors that may have different loadings on the ability of autistic individuals to understand and generate different types of nonliteral language.It is therefore important for future studies to include measures of all these variables in their experimental design.The thorough mapping of the factors that influence the processing of nonliteral language in ASD will contribute decisively to the future design of educational programs and interventions aiming at enhancing the pragmatic skills of autistic children.More specifically, scientific research in this field could assist educational programs to target the training and development of specific skills (e.g., EFs, language, or mind-reading skills) underlying the processing of figurative language in autistic children in order to attain the further growth of their corresponding skills.

AUTHOR CONTRIBUTIONS
Overview of the selected research studies in metaphor comprehension in ASD for this review.
T A B L E 1Abbreviations: ASD, Autism Spectrum Disorders; CASL, Comprehensive Assessment of Spoken Language; EFs, executive functions; TD, typically developing; ToM, Theory of Mind.
Overview of the selected research studies in irony comprehension in ASD for this review.