Recent Advances and New Directions in Measuring Theory of Mind in Autistic Adults

Conducting ToM research in autistic adults initially required changes to measures given to children that produced ceiling effects in adults (e.g., false belief tasks; Frith 1994). Therefore, ‘advanced’ measures were created to measure adults’ ability to reason about mental states in more complex contexts. The widely-used Strange Stories task (Happé 1994), for example, requires participants to infer false beliefs of characters in verbal vignettes, while the Reading the Mind in the Eyes Task (RMET; Baron-Cohen et al. 2001) requires matching static images of the eye with mental state words. These are considered ‘explicit’ measures as ToM is prompted by direct mental state questions. Despite their popularity and usefulness in advancing autism research (e.g., genetic aetiology; Warrier et al. 2017), there is growing awareness of the limitations of these ‘classical’ ToM tasks. IQ, for example, is highly correlated with performance on the Strange Stories task and the RMET (e.g., Baker et al. 2014). Accordingly, some individuals’ ToM ability is under-estimated due to verbal impairments, whereas individuals with good verbal skills may compensate for ToM impairments and perform well on these tasks (Livingston and Happé 2017). In addition, recent work challenges whether these tasks measure ToM or emotion processing, leading to caution against their use in autistic adults with emotional difficulties (Oakley et al. 2016; Olderbak et al. 2015). In attempts to address some of these issues, ‘implicit’ ToM tasks have been used to probe automatic behavioural responses to scenarios that induce spontaneous ToM (see Apperly and Butterfill 2009 for overview). These include measuring participants’ eye movements while they view scenarios where agents hold false beliefs (e.g., Senju et al. 2009) or the extent to which participants spontaneously attribute mental states to animated triangles (e.g., White et al. 2011). The theory and methodology surrounding ‘implicit’ ToM is, however, also not without controversy. Its existence and distinction from ‘explicit’ ToM is currently debated (Carruthers 2017), particularly given claims that many ‘implicit’ ToM tasks measure general attentional processing, not ToM (e.g., Heyes 2014; Santiesteban et al. 2015). Overall, therefore, studies using classical measures of ToM to measure ToM in autistic adults are inconsistent (see Chung et al. 2014, for meta-analysis).

We will not dwell on conceptual and theoretical debates about classical ToM tasks, which can be found elsewhere in the literature (see Brewer et al. 2017, for recent discussion). Indeed, we suggest that a focus on classical tasks, and resulting debate, is stymieing development of novel methods, thereby constraining the clinical relevance of ToM research in autistic adults. This is highlighted by several findings that individuals with ASD show relatively minor impairments on classical tasks compared with their difficulties observed in clinical settings and everyday social situations (e.g., Begeer et al. 2010; Couture et al. 2010; Lever and Geurts 2016; Scheeren et al. 2013; Schneider et al. 2013; Spek et al. 2010; Wilson et al. 2014). This finding is often attributed to differences between laboratory and naturalistic settings, yet little is done to understand this discrepancy. Addressing this issue offers a good opportunity to study ToM using behavioural tasks with the aim of supporting autistic individuals with everyday social difficulties (see also, Fletcher-Watson et al. 2014). To this end, we direct the readers’ attention to recent developments in measuring ToM that have particularly clear potential for improving understanding and management of ASD.

First, a computerised paradigm has been developed by Deschrijver et al. (2016) in which adult participants are required to track an agent’s belief about the location of a ball to measure ToM ability. Using a keypad, they are tasked with responding when the ball is revealed from behind an occluder. Critically, using the occluder, the ball’s actual location and the agent’s belief about its location are manipulated, so that the agent may hold a false belief about its location. Autistic adults, compared to matched controls, were slower to detect the ball when the agent falsely believed the ball was behind the occluder and the participant knew it was not. The group difference was not statistically significant, however performance on the task was closely associated with self-reported autistic traits and observational measures of autistic behaviour. Deschrijver et al.’s task is an ‘implicit’ measure of ToM, therefore it may be subject to the ongoing limitations and debate on the existence and measurement of ‘implicit’ ToM (see Nijhof et al. 2016). Nonetheless, this recent development indicates that ToM atypicalities, when measured by response time (RT) on appropriate cognitive tasks, may help to understand autistic symptoms in adulthood. More broadly, this suggests that measuring RTs—a relatively underused technique within classical ToM research—may be used to refine older ToM tasks, where RTs were previously not measured (see below).

Second, recently developed tasks have sought to measure ToM by examining accuracy, via open-ended or multiple-choice questions, of mental state attribution after watching video-clips of social situations (Brewer et al. 2017; Dziobek et al. 2006; Murray et al. 2017). Video-based tasks require more than simple inference from images or vignettes, and instead require inference of multiple socially relevant cues. This is designed to reflect mental state inference in ‘real’ social situations. Importantly, these tasks include non-ToM questions, allowing for alternative explanations of ToM impairment to be explored (e.g., poor memory). Although development of ToM videos may seem relatively straightforward, this process has been informed by extensive theoretical research and technological advancements. The advantage of these ecologically valid video-based tasks is demonstrated by the clarity of findings from recent studies (e.g., Murray et al. 2017), which report that autistic adults show impaired accuracy compared to neurotypical controls. Such clear difficulties on video-based tasks contrast with findings that autistic individuals perform well on classical ToM tasks (e.g., Chung et al. 2014).

Together, recent research demonstrates that new RT and video-based tasks can successfully measure atypical ToM. Large group differences between adults with and without ASD are being observed, far exceeding those found in previous research. Importantly, these results hold even after controlling for other cognitive abilities and appear to be consistent across studies. Furthermore, and importantly, recently developed ToM measures are sensitive to individual differences within samples of (a)typical adults, enabling their use in correlational analyses with other cognitive/behavioural processes associated with ToM (e.g., Shah et al. 2017). Most of these tasks are freely available, generating new opportunities and challenges to be addressed in future. Building on these advances, we propose a series of novel and previously overlooked approaches could further improve the measurement of ToM in research and clinical settings.

Atypical ToM is a clinically relevant feature of ASD, however difficulties in measuring ToM in ASD has constrained understanding of, and support for, social-communication difficulties in autistic adults. Recent advances in the field have improved the measurement of ToM in adults with ASD, as evidenced by their atypical performance on RT- and video-based ToM tasks that is in line with their symptom severity. Moving forward, we suggest that these tasks are abbreviated, so that they are suitable for collecting large datasets on web-based platforms and useful in clinical settings. Future use of RT measurement should also improve task sensitivity and overcome issues of autistic adults using compensatory strategies to ‘solve’ ToM tasks. Other novel approaches, including investigation into neurotypical ToM of autistic minds and autistic ToM of other autistic minds, should facilitate the move toward a suitably nuanced understanding of ToM in ASD in the context of ‘real-world’ interactions. Together, these developments will enable researchers and clinicians to better understand why most autistic adults, despite performing well on some ToM tasks, experience profound social-communication difficulties. It is hoped this will inform future clinical practice to assist autistic adults to manage their condition.