Introduction
Impairments in social communication constitute a defining feature of autism spectrum disorder (ASD) that can seriously impact competence across social contexts (Tager-Flusberg
2000). For instance, a number of studies have documented impoverished narrative skills in ASD, particularly in unstructured and emotionally salient contexts (e.g., conversational narrative) that are strongly reliant on social cognitive abilities, such as reading thoughts and emotions of protagonists or conversational partners (e.g., Losh and Capps
2003,
2006). In line with this observation, a number of studies have documented links between narrative impairments in ASD and deficits in social cognitive skills, where difficulty reading others’ emotions and cognitive states can limit the ability to build meaningfully on narrative topics and evaluate a communicative partner’s understanding and engagement (Capps et al.
1998; Losh and Capps
2003; Ochs and Capps
2001; Tager-Flusberg and Sullivan
1995).
Subclinical differences in narrative (and broader pragmatic skills) have also been documented among parents of individuals with ASD, and are considered a core feature of the broad autism phenotype (BAP) (Landa et al.
1991,
1992; Piven et al.
1997; Losh et al.
2008,
2012). The BAP refers to subclinical personality and language traits observed at elevated rates among parents of individuals with ASD that are believed to reflect genetic liability to ASD (Piven et al.
1997; Bernier et al.
2012; Bolton et al.
1994; Losh et al.
2008; Virkud et al.
2009). In a landmark study aimed at further defining the language characteristics of the BAP, Landa and colleagues reported evidence of impoverished narrative skills in parents of individuals with ASD, relative to parents of individuals with Down syndrome included as a control for the influence of parenting a child with a developmental disability (Landa et al.
1991). Specifically, parents of individuals with ASD produced narratives that were lower in complexity and coherence than those of controls. These patterns have been mirrored in studies of conversational discourse of parents of individuals with ASD, which have noted increased tangential language and less contingent conversational contributions (Landa et al.
1992; Losh et al.
2008,
2012; Piven et al.
1997). Considered with the extensive literature documenting narrative impairments in ASD, these findings suggest qualitatively similar narrative differences in parents, implicating narrative as a skill potentially impacted by genetic liability to ASD.
An important next step in evaluating the significance of narrative as a trait influenced by genetic liability to ASD will be to understand whether there exist similar profiles of strength and weakness across structured and unstructured contexts in both ASD and among first-degree relatives. Furthermore, exploring potentially underpinning processes related to narrative deficits and differences in ASD and the BAP is critical for understanding whether common underlying mechanisms contribute to observed narrative profiles. As noted previously, social cognition appears to importantly relate to the narrative impairments in ASD, perhaps implicating differences in social attention and perception as important sources of narrative differences. Attention to less socially salient aspects of a scene, for instance, could impact the ability to formulate narratives around meaningful themes and infer motivations of protagonists to build coherent stories. Although this question has not been directly addressed in the BAP, differences in social cognition have been reported in parents (Adolphs et al.
2008; Losh et al.
2009; Losh and Piven
2007; Baron-Cohen and Hammer
1997), and in one study were linked to differences in parents’ pragmatic skills in conversation (Losh and Piven
2007). Identification of such features impacted in both ASD and the BAP, and linked with broader language and related phenotypes associated with ASD and the BAP, might provide a window into those core skills impacted by ASD genetic risk and their neuropsychological origins.
Analysis of eye gaze may provide such an intermediate link, with potential to reveal attentional and perceptual differences that stem from underlying neurobiological variation influenced by ASD genetic risk, and that impact clinical-behavioral phenomena such as narrative and social behavior (Klin et al.
2002). Differences in visual attention have been repeatedly documented in ASD (see Chita-Tegmark
2016; Frazier et al.
2017; Papagiannopoulou et al.
2014 for reviews), and attentional differences during dynamic social scenes have been found to predict greater language and social-communicative impairment in individuals with ASD (Flavell et al.
1981; Jones et al.
2008; Klin et al.
2002; Righi et al.
2018; Speer et al.
2007). Studies of parents of individuals with ASD have also shown differences in visual attention to social scenes. Groen et al. (
2012) reported reduced visual attention to socially relevant aspects of brief videos in both parents and their children with ASD. In a study of face processing in the BAP, Adolphs et al. (
2008) found that when determining affective expressions of faces, parents of individuals with ASD who displayed the BAP showed a marked reduction in reliance on the eye region, along with increased utilization of the mouth region, relative to controls and parents without the BAP (a pattern that paralleled patterns observed in ASD; Spezio et al.
2007). Considering findings that visual attention patterns appear highly heritable in twins (Constantino et al.
2017), studies of gaze in ASD and unaffected first-degree relatives may serve as a promising avenue for identifying neurocognitive features related to complex behavioral phenotypes, as well as a quantifiable target to utilize in studies of phenotypes linked to molecular-genetic variation in ASD.
In this study, we explored the relationship between visual attention and narrative ability in both ASD and in parents of individuals with ASD. Our objectives were twofold—first, we aimed to document the narrative profiles in ASD and among parents across discourse contexts that differed in structure and emotional content (shown to be critical in revealing broader pragmatic impairments in ASD). In line with prior studies of narrative in ASD (Losh and Capps
2003,
2006; Tager-Flusberg and Sullivan
1995; Losh and Gordon
2014) we predicted that both the ASD and ASD parent groups would show greater differences from controls in the less structured, more emotionally evocative context. Second, we explored potential links between narrative profiles and patterns of visual attention in these same participants. For this exploratory aim, we presented narrative stimuli on an eye tracker, and characterized gaze patterns in relationship to narrative quality. Although predictions were less clear for these data given the lack of prior research examining the relationship between narrative quality and gaze in general, and among individuals with ASD in particular, we hypothesized that previously reported differences in narrative quality in ASD and among parents might stem from underlying differences in visual attention. As such, we predicted that lower quality narrative would be related to decreased attention to social images during narration.
Discussion
This study investigated narrative skill across contexts in ASD and parents of individuals with ASD. Given well-documented differences in social communication, and narrative in particular, in ASD and evidence of similar, but more subtly expressed differences in parents (Landa et al.
1991,
1992; Losh and Capps
2003; Losh et al.
2008; Loveland et al.
1990; Piven et al.
1997), this study aimed to better characterize narrative across discourse contexts varying in structure and emotional complexity. A primary goal of the study was to determine whether parents might show similar patterns of narrative differences across structured and unstructured contexts that could constitute genetically meaningful phenotypic profiles and provide clues into core skills impacted by genetic liability to ASD. Additionally, we explored visual attention during narration as a potential source of differences in narrative competence, in line with prior work documenting visual attention biases in ASD and among first-degree relatives.
In line with hypotheses related to narrative production, individuals with ASD and parents of individuals with ASD showed parallel patterns of narrative performance, with narrative quality comparable to controls in the highly structured PB context, but with the ASD and ASD parent groups producing less coherent narratives than controls in the less structured, more emotionally evocative TAT narrative task. Gaze differences were also noted in this less structured TAT context, and some associations between gaze and narrative were detected. These findings implicate narrative ability as a complex communication skill that may be impacted by ASD genetic risk. Although differences in visual attention during narrative were subtle and will need to be replicated, evidence of difference in both the ASD and ASD parent groups support the need for further investigation of gaze and language links to understand the origins of the complex social-communicative features associated with ASD.
Consistent with prior literature documenting greater narrative impairments in ASD in less structured contexts (e.g., Diehl et al.
2006; Losh and Capps
2003), individuals with ASD and the ASD parent group showed reduced narrative quality in the less structured TAT narrative task, but did not differ from controls in the highly structured narrative PB task. This pattern could be due to the reduced cognitive and social-emotional demands in the PB task. For example, participants told the PB narrative while viewing a single page at a time, with a clear temporal unfolding of relatively unambiguous events, and characters showing obvious facial expressions of basic emotions. By contrast, the less structured TAT included more ambiguous and emotionally complex scenes, which require understanding of thoughts, emotions, contextual features related to different psychological states (indeed, the TAT was developed with the goal of tapping such complex social perceptual skills) (Murray
1943), and also required narrative generation
after viewing each image, placing greater cognitive demands (e.g., working memory) on participants. It is perhaps notable that such robust differences in the TAT were evident in spite of explicit instructions to focus on story structure, content, and cognitive/emotional states of characters (i.e., “tell a story with a beginning, middle, and end”, and “discuss thoughts, feelings, and actions of characters”), which would presumably steer individuals’ narratives along somewhat common paths. Although prior studies have not compared narrative ability across different contexts in parents, these findings are consistent with evidence of narrative differences among parents of individuals with ASD when presented with a general direction to tell a story and an initial introductory sentence as a prompt, without any supporting visual stimuli (Landa et al.
1991).
Together, results suggest a relatively specific pattern of narrative differences evident in ASD and among parents, that also showed evidence of familiality in the ASD and ASD parent groups, consistent with a large body of work highlighting subtle differences in social-communication and personality features in first-degree relatives of individuals with ASD thought to reflect genetic liability [i.e., Broad Autism Phenotype; (Piven et al.
1994,
1997; Losh et al.
2008,
2012)]. Familial aggregation of a trait does not necessarily imply a genetic influence [e.g., narrative styles are certainly learned within families, and during parent–child interactions in particular (Haden et al.
1997)]. However, as noted previously, narrative differences were among the first reported phenotypes in early studies documenting the presence of a broad autism phenotype among parents of individuals with ASD (Landa et al.
1991), and considered in this context, the current findings appear to highlight narrative as a fruitful focus for future investigations, such as twin studies, that might more definitively evaluate genetic influence, and the potential of narrative-related skills as ASD endophenotypes.
Evidence of this specific pattern of narrative differences in ASD and parents also builds on prior work applying a computational measure of narrative (i.e., Latent Semantic Analysis, or LSA), showing, importantly, that this method is not only sufficiently sensitive to capture context-dependent narrative deficits in ASD (Lee et al.
2017; Losh and Gordon
2014), but also the more subtle differences evident in clinically unaffected parents. Application of such efficient, automated, and objective computational measures to characterize complex language phenotypes in ASD and among unaffected relatives can provide distinct advantages over hand-coding methods, which, while providing deep characterizations of language samples, are highly labor intensive and difficult to apply to large samples or across different study samples and research groups. The quantitative, continuous index of complex language ability produced by computational methods may also be advantageous for studies of ASD-related endophenotypes, where continuous measures of complex traits, measurable in affected and unaffected individuals can optimize power to detect associations between phenotypes and underlying biological variation. One notable weakness of this computational approach, however, is that more specific aspects of narrative performance (e.g., discussion of character motivations, and mental states shown in prior work to be deficient in ASD) are not captured, which in this study may have impacted our ability to detect associations between gaze patterns and these more specific, and meaningful, aspects of narrative production.
Nonetheless, we did detect subtle differences in visual attention in both the ASD and ASD parent groups. Group comparisons of gaze patterns indicated that, consistent with narrative performance patterns, very few differences were observed during the PB context (individuals with ASD showed a small but statistically significant reduction (i.e., 2%) in fixations to setting elements of PB scenes relative to controls). More differences in gaze were observed during the open-ended TAT task, although the pattern of differences varied across images and were not consistently significant when correcting for multiple comparisons. For example, individuals with ASD and BAP(+) parents attended more to the setting in response to the image from the TAT with the most complexly depicted setting (“Farmland” image). In contrast, individuals with ASD and parents [the BAP(+) group in particular] attended more to faces in response to images from the TAT where facial expressions were prominently featured and the emotional content more ambiguous (e.g., Man, Woman Gaze image). These findings contrast with prior work with different paradigms showing more striking differences in visual attention to social scenes in ASD, including atypical face processing (e.g., see Chita-Tegmark
2016; Frazier et al.
2017; Papagiannopoulou et al.
2014 for reviews). However, the current paradigm was distinct from such prior work in that individuals were
explicitly instructed to narrate and to discuss the characters’ thoughts and feelings, which likely prompted more focused and directed attentional strategies (and potentially attenuated differences) in social attention than those studied in prior work. Given that successful narration requires attention to both the main characters and the setting that contextualizes characters’ thoughts and actions within a broader theme (Reese et al.
2011), increased allocation of visual attention to the most complex aspects of a given image in the ASD and BAP(+) groups may reflect greater effort to integrate and process visual information to construct meaningful narratives. Therefore, these results may inform future studies examining the role of context in shaping visual attentional differences in individuals with ASD and the BAP.
It is also important to consider that patterns of visual attention where groups differed were not the same aspects of visual attention associated with narrative in either context. Whereas individuals with ASD fixated more intensively on setting in the most visually complex TAT image (“Farmland”), it was an increased attention to bodies that was related to poorer quality narratives. It may be that individuals with ASD and the BAP differ not only in allocation of visual attention, but also in the ways in which they utilize visual information to inform social communication—e.g., even though individuals with ASD and parents with the BAP looked more to faces during emotionally ambiguous TAT images, perhaps they were less able or inclined to capitalize on that information to enrich their narratives. For parent controls, narrative quality increased with greater attention to faces, particularly during images where faces were more prominent, suggesting that they capitalized on this information to inform their narrations. Different associations between narrative quality and attention in the ASD and BAP groups may also indicate that the complexity and degree of ambiguity of a social stimulus impact the attentional strategies employed, with the potential to both miss key aspects of relevant non-social information or to become overly focused on less social stimuli. Additionally, as noted previously, LSA (a global measure of narrative) may not be a sufficiently sensitive index of the finer-grained aspects of narrative that relate to visual attention. Future studies might address these questions by examining gaze and language patterns across different and potentially more sensitive measures of language (ranging from basic language processing skills to more complex language use such as narrative and conversation) and gaze (including moment-to-moment visual attention sequences and synchronized language production), and in larger samples, that could more powerfully index important relationships between gaze and language in real time.
An additional important finding concerns the relative specificity (and subtlety) of gaze differences to the BAP(+) parent group, whereas differences in narrative quality were observed more broadly in the full ASD parent group. Global differences in social communication have been observed among first-degree relatives in other genetically-based disorders impacting language (e.g., specific language impairment; Ruser et al.
2007), raising the possibility that differences observed in the ASD parent group overall reflect more general genetic liability to language disorder, rather than ASD specifically. In line with this possibility, a number of prior family studies of ASD have noted broad-based differences from controls among ASD parent groups, with more specific patterns of differences observed among BAP(+) subgroups, including studies of social cognition (Losh et al.
2009), face processing (Adolphs et al.
2008; Yucel et al.
2015), and visual attention during a language processing task (Nayar et al.
2018). Future work including comparison groups of parents of children with other genetically-based language disorders will be informative in teasing out ASD-specific risk markers evident in parents. It could also be the case that more detailed characterization of narrative ability (rather than the global narrative analysis examined in this study) could reveal patterns of narrative differences more specific to parents with the BAP.
In summary, results from this study highlight a specific pattern of differences in narrative skill in individuals with ASD and among parents (particularly those with the BAP), that may be linked with visual attention patterns, where differences are most robustly observed in unstructured contexts involving emotionally evocative, ambiguous scenes. Such overlapping phenotypic patterns in ASD and among parents suggest that narrative ability and related visual attention patterns may be important phenotypes that could be used in future studies indexing genetic liability to ASD, which could help to inform the basis of the complex social-communicative impairments in ASD. Findings that many, but not all, differences among parents were driven by the BAP(+) subgroup (e.g., with all parents showing differences in narrative, yet most gaze differences were specific to the BAP) may also have important implications for understanding mechanistic differences relating to core language-related phenotypes in ASD and the BAP.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.