Discussion
First, this study examined differences between autistic and control young adults in social-pragmatic inferencing, visual social attention and physiological reactivity, and second, investigated how social-pragmatic inferencing, visual social attention, physiological reactivity and autistic traits were associated. Our findings, as predicted, confirm previous findings reporting that, at a group level, autistic young adults have social-pragmatic challenges in inferring others’ thoughts (see, e.g., Deliens et al.
2018; Heavey et al.
2000; Jolliffe and Baron-Cohen
1999,
2000; Loukusa, in press; Lönnqvist et al.
2017). Such challenges in context-sensitive inferencing of meaning and in interpreting others’ intentions can have a major impact on everyday interactions for these individuals. Our findings also show notable variation among the autistic group, suggesting that the identified challenges are distinctly evident in a subgroup of autistic young adults. Our study also expectedly found that higher autistic traits were associated with poorer performance in social-pragmatic inferencing (lending support for prior studies, e.g., Volden et al.
2009). However, interestingly, another subgroup of autistic individuals showed social-pragmatic inferencing skills comparable to those of the highest performing control participants, demonstrating the heterogeneity in the autism spectrum. It should be however noted that structured test situations can only ever measure some specific aspects of social-pragmatic inferencing, and therefore, do not directly tell how these individuals navigate social-pragmatic situations in their daily lives (see e.g., Loukusa et al.
2018).
In line with our predictions, the findings further show that differences between autistic and control young adults in visual social attention are related to how key social moments in interaction are attended to and thus, are evident on a moment-level rather than on an aggregated level (Falck-Ytter et al.
2013; Freeth et al.
2011; Lönnqvist et al.
2017; Nakano et al.
2010; Nyström et al.
2017). However, rather than concerning all the key social moments, we found between-group difference concerning only one of the investigated moment-level variables (i.e., percent looking time at Submissive Characters’ Turn Interruptions). One explanation for this could be that the control young adults were better at using social cues to predict how the interactions might unfold and thus, in the context of our stimuli, were quicker at attending to the Submissive Characters’ Turn Interruptions. Since these moments involved getting interrupted and/or being left without acknowledgement by the Dominant Characters who namely dominated the interactions, they could be considered as more difficult to predict than Dominant Characters’ Turn Interruptions and Submissive Characters’ Facial Emotion Expressions (latter of which were reactive in nature). This interpretation is supported by prior research suggesting that autistic and NT individuals differ in how they use social information to predict others’ actions (von der Lühe et al.
2016).
Our study finds that the differences in visual social attention between the autistic and control group are very subtle but social-pragmatically relevant given our finding showing that attention to interlocutors’ facial emotion expressions (i.e. percent looking time at Submissive Characters’ Facial Emotion Expressions) was positively associated with social-pragmatic inferencing in the autistic group. That is, it seems that looking longer at the faces during these key social moments was relevant for inferring social-pragmatic meaning for the autistic group, while this was not the case for the control group. This might reflect different kind of processing styles. Our results could be interpreted in the light of a more local processing style in autistic individuals that relies on focusing on specific local details whereas NT individuals might process social scenes more globally and could be quicker in taking advantage of a variety of social cues and their combinations (see also e.g., Grynzspan and Nadel
2015; Jolliffe and Baron-Cohen
2000; Lönnqvist et al.
2017; van der Hallen et al.
2015). Such differences in processing styles could result in autistic individuals focusing on some local details while missing out on others (such as facial emotion expressions that could give insights about an interlocutor’s thoughts), and perhaps explain some of the misunderstandings autistic individuals experience in social situations that unfold at fast pace. Investigating moment-level visual social attention therefore appears critical since not only between-group differences do exist but in addition, these differences can have practical significance and as predicted, associations between social-pragmatic inferencing and visual social attention appear more pronounced in the autistic group (supporting prior studies by Grynszpan and Nadel
2015; Hanley et al.
2015; Lönnqvist et al.
2017; Sasson et al.
2007). However, we did not find the predicted association between autistic traits and visual social attention.
The present study also contributes to the currently relatively scarce literature on physiological reactivity in autistic adults as measured via HRV reactivity, specifically as regard to social-pragmatic inferencing. We predicted that the autistic group would show less physiological reactivity in response to the social-pragmatic videos than the control group, which our findings provided support for. This indicates that at the group level, autistic individuals do not show typical HRV suppression, lending support for previous studies with similar results (e.g., Dijkhuis et al.
2019b; Toichi and Kamio
2003). Interestingly, a small subgroup of autistic young adults showed a clear increase in HRV during the task condition (instead of suppression or no reactivity). Previously, Toichi and Kamio (
2003) found a similar pattern in autistic adults, and Porges et al. (
2013) in children. Lack of HRV suppression, and especially the increase in HRV, could hinder the efficient processing of stimuli and have a negative impact on performance (Porges et al.
2013).
We further predicted that greater physiological reactivity would be associated with better social-pragmatic inferencing and with less autistic traits. Correlational analyses showed moderate associations between HRV reactivity and social-pragmatic inferencing in both groups (notably of different directions) but these were statistically nonsignificant. We did not find support for the predicted association between HRV reactivity and autistic traits. In the autistic group, however, anecdotal evidence suggests that an association between HRV reactivity and social-pragmatic inferencing could be present in the small subgroup of individuals who experienced distinct parasympathetic activation in response to the social-pragmatic videos: In responding to the inference questions, all four participants showing a clear increase in HRV scored below the autistic group mean (scores ranging between 0 and 6). On the other hand, two out of the four participants in the autistic group who performed well in responding to the inference questions, showed HRV suppression in response to the social-pragmatic video condition, lending support for prior research on physiological reactivity and task performance (Klusek et al.
2013; Porges et al.
2013).
Together with the fact that the autistic group showed difficulties with the inference questions, yet no HRV suppression was observed, our result may indicate that the autistic group engaged less with the inferential process overall, perhaps reflecting motivational issues with the task. Alternatively, instead of spotting the subtle social conflict in the social-pragmatic scenes, they may have treated the watched interactions untroubled, setting a different frame for the amount of mental effort the task would require. Considered the other way around, a capability of self-regulation in this kind of attention-demanding task may contribute to better performance in the control group, as compared to the autistic group. For the exploration of these hypotheses, a more detailed qualitative analysis of the responses to the inference questions would provide crucial insights on both similarities and differences in how the scenes were processed. In addition, more research is needed to clarify the amount of mental effort that social-pragmatic inferencing in different contexts requires from autistic and NT individuals. Toichi and Kamio (
2003) have pointed out another possible explanation for the increase in HRV during task condition, as compared to baseline: It may be that the individuals who showed increased HRV instead of HRV suppression, were not relaxed in the chosen baseline condition, thus, the baseline did not work for them as a condition requiring less mental effort when compared to the task condition. In our study, the participants with the clearest increase in HRV also had a relatively high HRV at baseline compared to other autistic participants, which does not provide support for the hypothesis on extensive anxiety during baseline. Importantly, there were no significant between-group differences in HRV at baseline, which indicates that our baseline condition was comparable for both groups. However, the possible differences in how the participants experienced the baseline situation should be kept in mind when making conclusions based on the findings.
We also explored associations between physiological reactivity and moment-level visual social attention which have received limited attention in previous research. Our findings indicate that the longer the autistic group looked at specific key social moments (Dominant Characters’ Turn Interruptions and Submissive Characters’ Facial Emotion Expressions specifically), the more their HRV was supressed in response to the social-pragmatic inferencing tasks. This suggests that perceptual processes could play a role in how some autistic individuals physiologically react to complex social scenes as they may miss out on crucial social cues that could elicit a physiological reaction. One explanation for the significance of these particular moments could be that these moments could be viewed as emotionally charged: attending to the Dominant Characters’ Turn Interruptions would show to a participant that the Dominant Characters were deliberately, not by accident, interrupting the Submissive Characters whereas attending to the Submissive Characters’ Facial Emotion Expressions would reveal to a participant the Submissive Characters’ negative stance toward the Dominant Characters. Relatedly, Lory et al. (
2020) have recently observed an association between overall HRV (indicating dysregulation of the autonomic nervous system) and parent-reported atypical social attention in children. As HRV reactivity is also considered to be associated with self-regulation (e.g., Porges et al.
2013), an alternative explanation could be that autistic individuals with better state regulation (i.e., a better so-called vagal brake, evident in HRV suppression from baseline to social-pragmatic video condition) could be better overall and/or quicker at orienting to social stimuli (albeit not necessarily better at social-pragmatic inferencing). Together, these findings encourage future research to investigate these associations in greater detail, particularly by looking at both direct and indirect effects.
Some limitations of the current study merit note. First, due to the limited amount of high-quality eye tracking data available from the study participants, our sample size was relatively small. It is probable that our experimental protocol that prioritised comfort and thus, did not require the participants to use a chin or head rest, resulted in the considerable amount of unsuccessfully recorded eye tracking data for the stimuli investigated here. It should be noted that the small sample size has had an impact on the statistical power of the analyses and therefore, our findings could be considered as preliminary and should be confirmed with larger data sets. Second, in assessing the generalizability of our findings, it should be kept in mind that the participants in our study did not have any observed intellectual disabilities and do not represent the entire heterogeneous autism spectrum. Additionally, since many participants were excluded based on inadequate data quality, it is possible that the findings particularly hold for autistic young adults with such cognitive and behavioural characteristics that allow the reliable recording of their eye movements in an unstrained set-up (e.g., the ability to sit rather still throughout a relatively long experiment), which is a common limitation for eye tracking studies with similar set-ups. Third, the participants were diagnosed with autism spectrum disorder in their childhood and since this study was part of a follow-up phase involving the same individuals, diagnoses were not re-assessed at adulthood. Albeit not a diagnostic tool, the between-group difference in the AQ scores provided evidence for the significantly higher number of autistic traits in the autistic group. Fourth, the transition period used as a baseline in this study differs from baseline situations used in some other studies. Previous studies have used variable situations as baseline, for example, from quietly looking at a wall (Toichi and Kamio
2003) to watching a neutral, non-social video (Dijkhuis et al.
2019a,
b), yet there is no clear consensus of what an optimal baseline situation would be (Laborde et al.
2017). In the present study, we chose to use a between-tasks transition period as a baseline, to have as natural a baseline situation as possible. In this situation, some structure was provided by the experimenter and some social elements were involved (e.g., there were minimal interactions with an experimenter) in order to help participants to be as relaxed as possible. Fifth, the stimuli used in the study involved dynamic, complex social situations, yet a passive third-person perspective typical of most structured test situations does not allow for the social participation inherent in real-life interactions. Examination of attention in real-life social interactions may therefore shed light on different aspects of visual social attention, in particular, how gaze is used in interaction (see, e.g., Dindar et al.
2017; Gobel et al.
2015; Hessels
2020), and may bring out perhaps different information on both competencies and challenges than found in the current study. In the future, such moment-level analyses of visual social attention in real-life interactions would be fruitful in increasing understanding of the role gaze plays in navigating pragmatically complex real-life interactions.
Given the between-group differences in social-pragmatic inferencing, visual moment-level social attention and physiological reactivity, and the observed associations between these, our study lends support for theoretical accounts that consider perceptual processes and their integration having a central role in autism spectrum (Frith and Happé
1994; Murray et al.
2005). It is possible that the challenges in self-regulation and in controlling the ‘vagal brake’ initially hinder the autistic individuals from efficiently processing social situations, having a potentially profound effect on how they navigate the social world (e.g., Porges et al.
2013). If this is the case, what follows then is, first, the need to understand in practice how to improve autistic individuals’ self-regulation to allow for more capacity to engage with the social world. Second, if visual moment-level social attention plays a role in social-pragmatic inferencing (and in the social domain more broadly) in autism spectrum, it could be useful to develop autistic individuals’ understanding of both where and when to look in their social interactions with neurotypical interlocutors so as not to miss out on key social cues. Third, since social interaction is inescapably a ‘two-way street’ (see e.g., Milton
2012), it would be valuable to assist neurotypical interlocutors to interact in a manner that is less likely to result in misunderstandings, such as carefully considering what kind of embodied social cues are used to communicate meaning and particularly, when in interaction these are used.
Acknowledgments
Funding for this research was awarded from the Academy of Finland (276578, 333672), Eudaimonia Institute of the University of Oulu, Finland, the Alma and K. A. Snellman Foundation, Finland, and the Finnish Brain Foundation, Finland (earlier the Rinnekoti Research Foundation, Finland, and the Child Psychiatric Research Foundation, Finland). We wish to thank all the participants for taking part in the study. We feel reverently grateful to Professor emerita Irma Moilanen for acting as the principal innovator of the Oulu Autism Research Group and to Sirkka-Liisa Linna, Ph.D., Marko Kielinen, Ph.D., and Katja Jussila, Ph.D., who gave their expertise to the diagnostic processes. We also wish to thank Linda Lönnqvist and Laura Mämmelä for participating in the data collection process and Sampsa Toivonen for scoring the participants’ responses to the inference questions for the interrater reliability analyses. We thank the rest of the Oulu Autism Research Group, including Aija Kotila and Veera Pirinen, for collaboration. We are also grateful for the anonymous reviewers for their valuable comments on earlier versions of this article.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.