People with autistic-spectrum disorder (ASD) experience difficulties in comprehension of figurative language, and tend to give literal interpretation to ambiguous syntactic expressions, to phrases that convey irony, humor, or sarcasm, to idioms, or to metaphors (Happé, 1993, 1995; Kerbel and Grunwell, 1998; Rapin and Dunn, 2003; Adachi et al., 2004; Rundblad and Annaz, 2010; Mashal and Kasirer, 2011). Despite extensive research on comprehension of metaphoric language in autism, very little is known about how people with ASD generate metaphors. In fact, earlier research on creativity in ASD has focused primarily on non-verbal rather than on verbal abilities. Findings have mostly documented deficient performance in ASD, especially when using tasks that are imagination-oriented (e.g., Frith, 1972; Lewis and Boucher, 1991; Craig and Baron-Cohen, 1999). The present study investigates comprehension and generation of novel metaphors in ASD.

In a recent study we found that children with ASD (aged 12–15) differ from typically developing (TD) children in comprehension of conventional metaphors but not in comprehension of novel metaphors (Mashal and Kasirer, 2011). In this study we used a multiple-choice task that included conventional metaphors (e.g., sharp tongue), novel metaphors (e.g., joy bits), and unrelated word-pairs (e.g., laundry rabbit). Unlike conventional metaphors, which are coded in the mental lexicon, novel metaphors require the computation of on-line interpretation. It might be the case that the ability to create novel semantic connections between apparently unrelated concepts is not impaired in ASD. Consistent with our results, Hermann et al. (2013) found that processing of novel metaphors in 20 adults with Asperger syndrome (aged 22–68) was similar to performance of age-matched TD individuals. The experiment included a semantic judgment task, in which participants were asked to make rapid decisions about the literal truth of a given sentence. The task consisted of four types of sentences: literally true high-typical sentences (e.g., Some lawyers are juvenile magistrates); literally true low-typical sentences (e.g., Some lawyers are tax advisors); metaphors (e.g., Some lawyers are scale rulers); and scrambled metaphors (e.g., Some lawyers are days off spring). Responses to metaphors were significantly slower than were responses to the other types of expressions, but this pattern of results characterized both groups, with no impairment in performance seen in the ASD group. Thus, there is evidence that people with ASD demonstrate no difficulties with figurative language, contrary to previous arguments in the literature.

Note that some earlier work has suggested that individuals with ASD do not comprehend figurative language due to difficulties in executive functions (Russell, 1997). According to this account, ASD impairs planning, set-shifting (flexibility), working memory, and inhibition (Ozonoff and Strayer, 1997). It has been suggested that executive dysfunction contributes to the typical communication and social impairments in ASD (Landa and Goldberg, 2005). However, evidence for executive dysfunction in ASD remains equivocal. For example, Robinson et al. (2009) found differences in inhibition, planning, and self-monitoring between individuals with ASD and TD participants, but no group differences in cognitive flexibility and verbal fluency were noted. Other studies have recorded difficulties in linguistic and categorical (semantic) fluency in ASD (Rumsey and Hamburger, 1988). Executive dysfunction has a considerable adverse impact on the ability to plan well formed novel linguistic constructs, and to shift from one meaning of a word or phrase to another, especially in the case of metaphoric language (Landa and Goldberg, 2005). The ability to combine the two words that form a metaphor to a meaningful expression might thus be more affected by deficits in executive functions than the ability to interpret literal expressions (Amanzio et al., 2008). Mashal and Kasirer () have indeed demonstrated that executive dysfunction contributed to comprehension of novel metaphors among children with ASD.

Executive functions are important not only for comprehension but also for the generation of metaphors (Dietrich, 2004; Chiappe and Chiappe, 2007). To generate a metaphor that contains a topic and a vehicle, a speaker must first select the property that will be attributed to the topic, and then scan semantic knowledge for suitable vehicles that exemplify the abstract, higher-order attributive category. This process requires that one maintains access to the category while inhibiting features of both topic and possible vehicles that are irrelevant to the higher-order category. That is, highly accessible but irrelevant semantic knowledge (e.g., adjectival descriptions of the topic) as well as many accessible but trite possibilities, such as idioms, clichés, and conventional metaphors must all be inhibited. Finally, the speaker must evaluate whether the metaphor conveys the desired meaning and emotional tone, and then revise it if necessary (Silvia and Beaty, 2012).

There is evidence that the generation of conventional and novel metaphors relies on different cognitive abilities (Bowdle and Gentner, 2005; Mashal, 2013). Silvia and Beaty (2013) suggested that generation of conventional metaphors is influenced by general knowledge or vocabulary, whereas generation of novel metaphors is affected by fluid intelligence and executive processes. Silvia and Beaty (2012) gave their participants two different prompts and instructed them to describe past emotional experiences through a metaphor. The first prompt asked people to “think of the most boring high-school or college class that you have ever had. What was it like to sit through?” For the next prompt, participants were asked to “think about the most disgusting thing you ever ate or drank. What was it like to eat or drink it?” Six measures of inductive reasoning, primarily non-verbal and visuo-spatial, were used to assess fluid intelligence. Results showed that fluid intelligence predicted the creative quality of novel metaphors. This suggests that people with high non-verbal intelligence can produce more creative metaphors than do people with lower intelligence.

Within the ASD literature, creativity has been measured primarily by non-verbal tasks that involved the production of spontaneous sequences of colors, tones, or drawings (Frith, 1972; Lewis and Boucher, 1991; Liu et al., 2011) or by verbal fluency tasks (Boucher, 1988). No extensive investigation of generation of novel verbal concepts has been conducted. Craig and Baron-Cohen (1999) examined creative thinking in children with ASD by using the Torrance Creativity Test (Torrance, 1974). Their participants could generate novel uses to an object (e.g., “what else can a pencil be?”), but they generated fewer such novel uses overall compared to TD children. Moreover, their ideas tended to be quite concrete rather than imaginary. Liu et al. (2011) emphasize that imagination is not necessarily important for successful performance on tasks that involve creativity and originality. They used an exercise in divergent thinking which included 12 incomplete figures, and asked their participants to complete the figures in an original way as well as to title them. The completed drawings were assessed on domains of fluency, openness, flexibility, originality, and elaboration. According to Liu et al., children with Asperger’s syndrome demonstrated better elaboration and originality than did age-matched TD peers.

The main goal of the current study is to examine metaphor comprehension and generation in adults with high-functioning ASD. Generation of novel and original metaphors is assumed to measure verbal creativity. In addition, we examine executive functions and analyze the contribution of these functions to performance on the metaphor tasks. We hypothesize that adults with ASD will show difficulty in comprehension of conventional metaphors but not in comprehension of novel metaphors. We further expect participants with ASD to be able to generate novel metaphors but to a lesser extent than TD participants. Finally, we assume that executive functions will contribute to metaphor comprehension and generation in both the ASD and the TD groups.

The present study examined metaphor processing in adults with ASD, differentiating between conventional and novel metaphors on tests of comprehension and generation. Our results show that adults with ASD demonstrate no difficulties in comprehension of conventional and novel metaphors. Furthermore, adults with ASD outperformed age-matched TD peers in metaphor generation. An inspection of the type of metaphors generated indicated that the ASD group produced more original and creative metaphors than did the TD group.

Whereas previous studies highlighted difficulties with metaphoric language comprehension in ASD (e.g., Happé, 1993, 1995; Ozonoff and Miller, 1996; Rundblad and Annaz, 2010; Kaland et al., 2011; Mashal and Kasirer, 2011), the current research shows similar comprehension of metaphors among ASD and age-matched TD peers. One of the differences between our study and earlier studies might be that we examined adults, unlike previous studies that focused primarily on children and adolescents. It is possible that the accumulated verbal knowledge that comes with age leads to greater familiarity with conventional metaphors within the ASD group. Indeed, previous studies have emphasized the contribution of vocabulary to the comprehension of metaphors (Evans and Gamble, 1988; Nippold, 1998; Chiappe and Chiappe, 2007; Silvia and Beaty, 2013). Consistent with this observation, we found that higher vocabulary scores were correlated with better comprehension of conventional metaphors (but not with comprehension of novel metaphors). The fact that we found equivalent comprehension across groups might also be related to the task that we used. It is possible that presenting metaphors in a multiple-choice questionnaire facilitated comprehension and was easier than were previously used tasks, such as asking comprehension questions about metaphoric stories (Rundblad and Annaz, 2010), requiring sentence completion (Norbury, 2005), or having participants perform a semantic judgment task (Hermann et al., 2013).

As expected, and in line with previous studies (Mashal and Kasirer, 2011; Melogno et al., 2012; Hermann et al., 2013), no difference was found between participants with ASD and TD age-matched peers in comprehension of novel metaphors. Novel metaphor interpretation is not coded in the mental lexicon and hence is not dependent on previous knowledge. Indeed, unlike the correlation between vocabulary and comprehension of conventional metaphors, no association was found between vocabulary and comprehension of novel metaphors. The ability to understand novel semantic connections between seemingly unrelated concepts appears to be intact in ASD, probably because it does not rely on activation of lexicalized expressions. Consistent with this finding, Melogno et al. (2012) emphasized that individuals with ASD can explain novel metaphoric phrases in unique ways that rely on phonological or semantic association. Thus, comprehension of novel metaphor is intact in ASD because it relies on good associative abilities rather than on lexicalized verbal knowledge.

Another finding of the current study is that contrary to our hypothesis, participants with ASD generated more metaphors than did age-matched peers. We were specifically interested in testing whether adults with ASD relied on previous knowledge, and were thus using familiar metaphors or idioms, or alternatively, whether they generated their own novel and original metaphors. We found that adults with ASD demonstrated greater verbal creativity than did TD individuals. Examples of creative sentence completions included phrases such as “Feeling successful is like seeing the view from the mountaintop” and “Feeling worthless is like offering a salad to South Americans.” These examples contrasted with more conventional figurative expressions provided by TD adults, such as “Feeling sad is to get the blues.” Our results suggest that adults with ASD can create unique verbal associations that are not restricted to previous knowledge, thus pointing to unique verbal creativity in ASD. It has been reported that one in ten people with autism shows some savant skills in such categories as music, art, calendar calculations, or mathematics (Treffert, 2009). Yet, most studies have indicated impoverished creativity in autism. For example, in standardized tests such as the Torrance Creativity Test (Torrance, 1974), people with ASD demonstrate difficulties in cognitive flexibility and imaginative fluency, as well as lack of imagination and originality compared to TD participants (Craig and Baron-Cohen, 1999).

Why, then, do adults with ASD demonstrate greater creativity in metaphor generation when previous studies found that they were lacking in imagination? Baron-Cohen et al. (2009) suggest that the hypersensitivity that characterizes autism gives rise to excellent attention to details. With respect to verbal creativity, this attention to details is associated with weak central coherence, leading to greater appreciation of local features over global ones. Lyons and Fitzgerald (2005) emphasized that in contrast to their social impairments individuals with Asperger syndrome are gifted with creativity and originality, excellent memory, strong focus of attention and specific cognitive styles. The unique verbal associations generated by the ASD group in the current study could thus reflect memory for details and weak central coherence. Another reason why adults with ASD generated more original metaphors might relate to difficulty in theory of mind. Mind-blindness makes one focus on one’s own thoughts, ignoring the addressee (Happé and Vital, 2009), possibly leading to production of expressions that are less conventional (Liu et al., 2011). Kanner (1946) was the first to note that the unique phrases produced by the children he studied could be interpreted as metaphoric language. Asperger (1941) also identified certain expressions in the speech of his patients that resembled the novel linguistic forms produced by young TD children. Asperger proposed that characteristics such as concrete intelligence and disregard of social conventions might be prerequisites for certain forms of new thinking and creativity (Gillberg, 2002). Fitzgerald (2004) also noted that individuals with Asperger syndrome have remarkable capacities for persistence and observation, high levels of energy and motivation, and abilities to focus intensely on a single topic. It appears that some individuals with autism and Asperger syndrome are highly creative, imaginative and original and even their humor can range from word-play and sound associations to precisely formulated, truly witty comments. According to Lyons and Fitzgerald (2004) some individuals with autism and Asperger syndrome also seem to master the cognitive processing of humor, i.e., incongruity and its resolution and switching of meanings as portrayed by the production of relatively sophisticated puns and word games.

In our final analysis we sought to identify the extent to which executive functions as well as verbal and non-verbal skills contribute to the prediction of metaphor comprehension and generation. We found that while verbal abilities, i.e., vocabulary and picture naming, contribute to the comprehension of conventional metaphors, this is not the case with regard to novel metaphors, in line with previous studies (e.g., Nippold, 1998; Chiappe and Chiappe, 2007; Silvia and Beaty, 2013). Furthermore, executive functions were shown to predict novel metaphor comprehension, as has been previously reported (Pennington and Ozonoff, 1996; Amanzio et al., 2008; Mashal and Kasirer, 2011; Mashal, 2013). More specifically, the TMT-B, which demands mental flexibility, contributed to the prediction of novel metaphor comprehension. We believe that comprehension of novel metaphors relies on flexibility because it is based on a shift between the literal and the metaphoric meanings of the words that appear in the new expression. Champagne-Lavau and Stip (2010) also found this correlation between the TMT-B and metaphor comprehension in schizophrenic and TD adults.

Finally, metaphor generation was predicted by non-verbal cognitive ability. Indeed, Silvia and Beaty (2012) have also emphasized the contribution of fluid intelligence to the generation of creative metaphors. It is possible that generation of metaphors, specifically novel ones, relies on the ability to come up with new solutions, assessed by tests of fluid intelligence. We note that a similar connection was documented in our study between non-verbal intelligence and simile generation, although no group differences in simile generation were found.

Some limitations of the study should be taken into account. Our measures of executive functions were rather limited and may need to be expanded to include the assessment of working memory and inhibition. Studies have shown that people with ASD may present difficulties in tasks that require response inhibition (e.g., Hughes and Russell, 1993; Hughes, 1996; Russell, 1997; Minshew et al., 1999; Hill, 2004; Robinson et al., 2009) or when they are required to shift from one response set to another (Ozonoff and Strayer, 1997; Ozonoff et al., 1998). It is possible that low response inhibition led to generation of unique verbal expressions by our ASD participants. Future studies will have to examine whether inhibition is associated with metaphor generation in ASD.

In sum, our findings indicate that adults with ASD are not impaired on comprehension of conventional and novel metaphors. We also found that adults with ASD generate more original metaphors relative to age-matched peers and that non-verbal skills contribute to this ability. The study points to unique verbal creativity in ASD which has not been studied in this way before.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.