Successful navigation of a complex and dynamic social world requires the ability to perceive, interpret, and respond appropriately to social cues. As such, human perceptual systems are specialized to preferentially detect and attend to social information including the movements of animate beings, known as biological motion (Johnson, 2006). Research using point-light displays, which capture movement of the major joints, indicates that the dynamics of biological motion alone are sufficient for the automatic perception of an animate being (Johansson, 1973). Rather than requiring other perceptual cues, humans are able to recognize not just animacy, but even emotions (Atkinson et al., 2004) and intentions of others (Manera et al., 2010) based solely on the dynamics of these impoverished stimuli.

Biological motion perception can be differentiated from perception of other forms of coherent motion and mapped to specialized neural systems. For example, posterior superior temporal sulcus (pSTS) is an important region for perception of biological motion and plays a critical role in understanding motion within a social context (Allison et al., 2000). Sensitivity to biological motion emerges early in development (Bardi et al., 2011) and may serve as a precursor to subsequent social-cognitive abilities (Pavlova, 2012).

Autism spectrum disorder (ASD), a neurodevelopmental disorder defined by deficits in social and communication abilities and restricted, repetitive behaviors and interests, is often associated with differences in biological motion perception. When compared to control groups, individuals with ASD fail to orient to (Klin et al., 2009; Annaz et al., 2012) or show decreased visual sensitivity for biological motion stimuli (Blake et al., 2003; Kaiser et al., 2010a). During biological motion perception, autistic individuals often show hypoactivation of pSTS (Herrington et al., 2007; Freitag et al., 2008; Kaiser et al., 2010b; Ahmed and Vander Wyk, 2013) and regions of prefrontal cortex (PFC; Kaiser et al., 2010b; Koldewyn et al., 2011). However, group-level differences are not always consistent (e.g., Murphy et al., 2009; Saygin et al., 2010), which might reflect the vast heterogeneity in social cognitive abilities within both clinical and non-clinical populations. For example, it has been shown that even within autistic groups, visual sensitivity to biological motion varies as a function of severity of the disorder (Blake et al., 2003). Therefore, considering the autism spectrum as a continuous measure may better enable the detection of meaningful biomarkers associated with this important social perceptual ability.

It is increasingly understood that the traits associated with ASD are not limited to diagnostic categories, but are normally distributed within the general population (Baron-Cohen et al., 2001). Unaffected individuals with a high number of autistic-like traits also fail to automatically attend to (van Boxtel and Lu, 2013) or show decreased visual sensitivity for human biological motion stimuli (Kaiser and Shiffrar, 2012; Miller and Saygin, 2013). Converging evidence suggests that autistic traits are specifically associated with difficulties with global, holistic interpretations of human motion, rather than perceiving coherent motion cues per-se (van Boxtel et al., 2016). For example, higher autistic-like traits are associated with difficulties detecting human but not object or animal motion (Kaiser and Shiffrar, 2012), the direction that a point-light walker is facing, but not walking (Miller and Saygin, 2013), and whether dynamic stimuli depict humans interacting (van Boxtel et al., 2016).

Non-clinical autistic-like traits are also associated with differences in brain structure and function in regions important for perceiving and understanding human motion (von dem Hagen et al., 2011; Nummenmaa et al., 2012; Anderson et al., 2013; Thurman et al., 2016). For example, Anderson et al. (2013) found a negative association between a measure of social impairment associated with autism and activation to biological motion in frontal and partial regions in a non-clinical sample. However, these authors only considered traits in the social domain, and did not report descriptive statistics for this measure making it difficult to assess what range of the spectrum was captured in this study.

The current study adds to this growing body of literature suggesting neural differences in global biological motion processing across the autistic-trait continuum by adopting several important methodological considerations: (1) we include a large sample that displays high variability in autistic-like traits to identify how neural response to biological motion varies across the non-clinical autism spectrum; (2) we employ a whole-brain exploratory approach to avoid an overly modular view of the brain and examine whether autistic-like traits are associated with recruitment of differential neural networks beyond those canonically associated with biological motion perception; (3) we use a measure of autistic traits that considers both social and non-social behaviors to explore whether specific aspects of the autistic phenotype are differentially associated with neural response; and (4) we utilize a passive viewing paradigm. While previous studies used forced-choice discrimination paradigms (e.g., Thurman et al., 2016), it is well-known that instructing participants to identify an animate entity within a dynamic display biases viewing behavior and neural response toward this end (Stanley et al., 2010; Zwickel et al., 2012; Lee et al., 2014; Gowen et al., 2016). Therefore, we expect a passive viewing paradigm to evoke differences in participants' spontaneous tendency to globally perceive or interpret stimuli as biological, thus optimizing our ability to detect individual differences in neural response.

This study shows that neural response during passive viewing of point-light displays of biological motion varies as a function of autistic-like traits assessed via the AQ. Individuals with more autistic-like traits show increased neural response to biological motion in prefrontal and left temporal cortices. By using a continuous measure of autistic-like traits coupled with an exploratory, whole-brain approach, we were able to assess the role of a network of brain regions not traditionally associated with biological motion perception in healthy adults. For example, right pSTS is often discussed as the area of functional specialization for biological motion stimuli (Allison et al., 2000), and is found in our examination of group activation as a whole. However, when including AQ score as a covariate, we find differential activation within pSTS localized to the left hemisphere.

As a whole, the clusters that emerge from the analysis of the main effect of viewing biological vs. scrambled motion are different from those that emerge when including AQ as a covariate in the analysis. Only 0.01 and 0.03% of voxels identified in the AQ covariate analysis overlap with the biological>scrambled and the biological<scrambled contrasts of the task main effect analysis, respectively. A targeted region of interest approach testing for relationships with AQ only within regions identified in the main effect analysis would have actually greatly reduced our ability to identify meaningful individual differences because this analysis specifically identifies those regions that show similarities across subjects. Rather, by including AQ as a covariate in the model of whole-brain activation, we are able to identify regions that show differential activation as a function of autistic-like traits. Furthermore, mean is included as an explanatory variable in the covariate model because we expect activation to vary as a function of task condition. Therefore, only regions in which the covariate accounts for significant additional variance in activation above and beyond that which is accounted for by the task contrast will emerge. While a covariate certainly can account for additional variance in regions identified in the main contrast, it is not surprising that regions not identified in the main contrast emerge in this analysis.

The regions that emerge in the AQ covariate analysis largely overlap with regions involved in understanding others' intentions and behaviors (Carrington and Bailey, 2009; Ames et al., 2011). For example, while right pSTS is robustly activated during biological motion perception (Allison et al., 2000), left pSTS is more sensitive to the intent of biological motion (Morris et al., 2008). Similarly, ventromedial PFC is implicated in understanding others' intentions and behaviors (Amodio and Frith, 2006; Gilbert et al., 2006; Frith, 2007), but is not typically discussed with biological motion perception in adults. The increased activation in these regions among individuals with more autistic-like traits might reflect greater inferential effort to understand the biological motion.

This interpretation is in line with a recent study that examined neural response to eye gaze as a function of autistic-like traits (Nummenmaa et al., 2012). Biological motion perception is not limited to the dynamics of whole body, but can include other forms of human motion that convey meaningful social information, such as eye movements (Allison et al., 2000). Neumann and colleagues presented participants with images of faces in which the eye gaze either remained constant or shifted from side, to direct, to side. They found that increased autistic-like traits were associated with increased neural response to shifted vs. constant eye gaze in regions that also emerge in the present study, including left temporal cortex and regions of prefrontal cortex. They interpret this increased activation within regions involved in inferring others' mental states to potentially reflect individual differences in the spontaneous tendency to draw inferences from eyes (Nummenmaa et al., 2012).

A similar hypothesis is further supported in the present study by the fact that the communication subscore of the AQ largely drives these effects, and also identifies an additional region associated with interpreting others' intention, the posterior cingulate cortex (Cavanna and Trimble, 2006; Carrington and Bailey, 2009). While defined as a communication domain, this subset of questions (e.g., “I find it easy to ‘read between the lines’ when someone is talking to me;” “I know how to tell if someone listening to me is getting bored”) largely touches upon one's ability to understand and appropriately interpret the mental states of their communication partner.

It has been hypothesized that the overarching function of the brain is to compute predictions based on prior experiences (Friston et al., 2009; Friston, 2010). Discounting information that is consistent with predictions reduces redundancy and results in lower activation (van Boxtel and Lu, 2013). Specifically, pSTS and regions of the action observation network, which includes temporal and prefrontal regions, are thought to play an anticipatory or predictive role in action perception (Rizzolatti and Sinigaglia, 2010; Vander Wyk et al., 2012). Consistent with this hypothesis, it has been shown that viewing highly unfamiliar actions (Cross et al., 2012) or motions that violate expectations (Brass et al., 2007; Jastorff et al., 2011) is associated with increased activity in prefrontal and temporal regions in adults.

Increased activity in similar regions in the present study might reflect a diminished ability to use prior experience with human action to accurately predict and interpret the current stimuli. Evidence from a recent biological motion adaptation study supports this hypothesis. Thurman et al. (2016) found increased autistic traits were associated with reduced perceptual aftereffects and diminished neural adaptation to biological motion in pSTS. A failure to adapt and desensitize to biological motion stimuli might reflect an inability to integrate prior perceptual experience with biological motion into the stored motion concept for use in subsequent predictions.

Sensitivity for biological motion increases throughout development—children reach adult levels of discrimination around middle adolescence (Hadad et al., 2011). At a neural level, young children automatically recruit PFC to a greater extent while simultaneously relying less on posterior regions during biological motion perception (Carter and Pelphrey, 2006) and understanding the intention of motion (Vander Wyk et al., 2012). With age, a shift in neural reorganization from frontal to posterior regions co-occurs with skill acquisition and is indicative of reduced effort and increased automatization (Johnson et al., 2009). The pattern of prefrontal neural activation associated with autistic-like traits in the present sample may therefore reflect a differential neurodevelopmental trajectory for individuals across the autistic-trait spectrum.

Anderson et al. (2013) specifically considered age-related differences in biological motion perception and its relation to non-clinical social deficits. They found that children and adolescents show a negative relationship between autistic traits and neural response in bilateral posterior and right frontal regions, and a positive relationship within left temporal regions. However, adults only showed a negative association between social deficits and neural response to biological motion perception. The present results in adults using a wider measure of autistic-like traits conflict with the adult data in this previous study. We see a positive relationship between autistic-like traits and neural activation in left temporal and prefrontal regions, which more closely aligns to their child population. One potential explanation for this discrepancy is that our sample is younger (18–25, M = 20.7 years) than the adults in this prior study (20–35, M = 24.7 years).

It is noteworthy that many studies that failed to find differences in biological motion perception between autistics and controls used older populations. This suggests that older ASD individuals may have adopted alternative strategies to discriminate biological motion (van Boxtel and Lu, 2013) that would be reflected in the recruitment of different neural networks resulting from divergent neurodevelopmental trajectories across the autism continuum.

Such altered neurodevelopmental trajectories may be established by variation at the genetic level. Converging research suggests variable genetic load for ASD may differentially impact neural systems specialized for biological motion perception. For example, while diagnosed individuals show hypoactivation of pSTS and PFC, their siblings, who share similar genetics but do not display autistic traits themselves, show increased activation in these regions during biological motion perception relative to controls (Kaiser et al., 2010b). The authors interpret this distinct neural response to reflect a compensatory mechanism by which siblings over-recruit these regions to overcome their high genetic risk. Similar etiologic mechanisms are thought to contribute to ASD traits across the continuum (Robinson et al., 2011), and thus might account for the positive association between autistic traits and neural response in these regions in our non-clinical sample.

Our study reveals a relationship between autistic-like traits in healthy adults and individual differences in neural response while passively viewing biological motion. These results emphasize the need for continuous phenotypic measures and the identification of neural endophenotypes, and are directly in line with the research domain criteria (RDoC) identified as a priority by the National Institute of Mental Health (Insel et al., 2010; Cuthbert and Insel, 2013). Our quantitative pairing of brain and behavioral assessments in healthy adults has revealed a potential mechanistic target for future research into typical and atypical developmental processes. Future studies investigating both typical and disordered perceptual and cognitive abilities, particularly those investigating underlying biological mechanisms, should carefully consider individual variability across the full trait continuum.

MP and JM designed the research. MP acquired and analyzed the data. MP wrote the paper with contributions and comments from JM.