Effects of sensory distraction and salience priming on emotion identification in autism: an fMRI study

Autism spectrum disorders (ASD) are characterized by difficulties with social interaction, which relies heavily on the ability to process emotional expressions. Individuals with ASD consistently show altered face processing [1, 2]; yet, why this occurs is still not well understood. One area that is understudied is the effect of sensory distraction on face processing. Sensory over-responsivity (SOR), defined as an extreme sensitivity to sensory stimuli such as touch or sound, is very common in ASD, occurring at rates of 56–78% [3‐8]. SOR is also associated with greater social difficulties [9, 10]. Given the essential role that face processing plays in social communication, understanding the impact of sensory distraction upon emotional face processing is highly relevant for developing more effective interventions for ASD that take into account real-world social settings in which there is substantial competing sensory information.

Neuroimaging research can provide insight into the mechanisms underlying these processes, which can then inform intervention. Although behavioral studies of emotional face processing in ASD have generated mixed results, fMRI and EEG studies have consistently revealed atypical neural activation during emotional face processing in individuals with ASD. However, the direction of the effects and the regions affected have varied depending on both the context of the task and participant characteristics (for reviews, see [1, 11, 12]). For example, multiple studies have reported lower activation in the fusiform face area during face processing tasks in individuals with ASD [13‐19], but in studies where participants were explicitly instructed to look at a fixation cross in the center of the presented faces or were directed to look at the eyes, participants with ASD displayed typical fusiform activity [20, 21]. Differences have also been found between implicit and explicit facial emotion recognition tasks, with participants with ASD displaying lower fusiform activity during implicit but not explicit facial emotion recognition [22]. In addition, activity in the fusiform gyrus and amygdala during face processing has been reported to correlate with time spent fixating on the eyes for individuals with ASD, increasing to similar levels as control subjects in the case of the fusiform and increasing to higher levels than control subjects in the case of the amygdala [13]. Therefore, the brain mechanisms underlying face processing in ASD may depend greatly on where attention is directed, which in turn may depend on the extent to which individuals are distracted by extraneous sensory stimuli.

SOR has generally been associated with greater social challenges within individuals with ASD at a behavioral, physiological, and neural level. Multiple studies have shown associations between parent-reported SOR and poorer social functioning in children with ASD on measures including the social awareness subscale of the Social Responsiveness Scale (SRS), the socialization subscale of the Vineland Adaptive Behavior Scales (VABS), and the DSM-IV social scales [9]. In children with ASD, higher SOR was related to higher cortisol levels during a peer-interaction paradigm, indicating higher levels of physiological stress during social interaction [23]. At a neural level, previous work from our lab demonstrated that youth with ASD showed reduced activity in auditory language areas and mentalizing frontal regions when completing a social task (interpreting communicative intent) in the presence of simultaneous mildly aversive tactile stimuli, whereas typically developing youth demonstrated increased activity in these same regions [24].

To date, very few studies have looked specifically at the relationship between face processing and sensory processing atypicalities. One recent study found that atypical event-related potential (ERP) correlates of face processing in siblings of children with ASD were associated with greater parent-reported social communication and sensory difficulties [25]. Another study found that self-reported sensation avoiding was correlated with greater right fusiform activity while looking at faces in adults with ASD but not in controls [26]. However, to our knowledge, the neural basis of facial emotion identification in ASD in the presence of competing aversive sensory stimuli has not yet been examined.

One possible explanation for the relationship between sensory and social challenges in ASD is an over-attribution of salience to extraneous sensory information at the expense of social cues. This is supported by findings that SOR in youth with ASD is associated with reduced neural habituation to mildly aversive sensory stimuli in primary sensory cortices as well as in regions related to salience and affective valence such as the amygdala [27‐29]. Sustained responses to sensory information in the amygdala suggest sustained attention and salience attribution. Furthermore, SOR severity in youth with ASD is related to atypical resting-state connectivity in the salience network, an intrinsic brain network thought to be involved in selecting which of many competing external and internal stimuli an individual attends to [30, 31]. Specifically, youth with higher SOR showed increased resting-state functional connectivity between salience network hubs and primary sensorimotor regions but decreased connectivity between salience network and visual association areas involved in social cognition [32]. According to this framework, interventions could reduce the impact of sensory distractions on social functioning either by reducing the salience of the sensory information or increasing the salience of the social information.

There has been some evidence showing that directing attention to relevant cues in social tasks can improve behavioral performance and activate normative or compensatory neural circuitry. For example, Li and Tottenham [33] showed that neurotypical adults with higher autism traits were initially slower to identify emotional expressions. However, this difference was no longer observed when participants were primed with videos of their own faces making emotional expressions. In addition, our prior work showed that directing attention to key social cues led to children with ASD engaging medial prefrontal cortex (mPFC) and maintaining activation in social processing regions during a social inference task [34] even with sensory distraction [24]. Increased activation in mPFC was associated with lower levels of social difficulties [34] and sensory over-responsivity [24]. Taken together, these studies suggest that increasing the salience of as well as attention to social cues can improve social functioning even in the context of extraneous sensory distracters, likely through engagement of the prefrontal cortex.

The role of mPFC as a potential compensatory mechanism and target for social intervention is consistent with its known role in mentalizing, social cognition, and emotional processing. Dorsal medial PFC (dmPFC) has been shown to be involved in appraisal and expression of emotion while ventromedial PFC (vmPFC) regulates limbic regions involved in generating emotion [35]. The dmPFC has been considered a part of a “mentalizing network” [36, 37] and plays a critical role in self and other referential processing [38, 39]. The dmPFC is also known for playing a role in social cognition and has been found to preferentially respond to scenes of social interaction during viewing of a naturalistic movie [40]. The extent to which salience-enhancing intervention strategies are successful at engaging mPFC may vary depending on individual differences in social ability and sensory processing. For example, Green et al. [24] found that attentional direction engaged mPFC more for ASD youth with lower levels of SOR. In contrast, after attentional direction, higher SOR was associated with greater activation of disparate sensory cortical regions, suggesting a potentially less efficient strategy involving separately processing each individual social cue rather than integrating the information. Therefore, it is possible that explicit attentional direction may result in a more effortful, less efficient processing experience for youth with SOR who are already overwhelmed with interpreting multiple competing types of incoming information. To address this concern, we tested an implicit priming strategy—viewing one’s own emotion faces—to examine whether this could lead to more integrative, less effortful processing strategies.

To our knowledge, this is the first study to examine the neural effects of distracting sensory stimuli on emotion identification in youth with ASD. We additionally investigated the impact of a prime designed to increase the salience of the emotional cues by showing participants videos of their own faces making emotional expressions.

First, we found that there were no behavioral differences in accuracy or response time between ASD and TD youth during the emotion identification task, which indicates that both groups were able to understand and complete the task at a similar level of performance. Similarly, there were no group differences in the conditions where participants were identifying faces without background noise before seeing their own faces, potentially indicating that this is a relatively simple task that, when presented without distraction, can be easily accomplished by youth with ASD.

Secondly, we aimed to explore how distracting aversive background noises would affect brain responses while processing emotional faces in youth with ASD. With the addition of the auditory distracting stimuli, both groups showed similar increases in the auditory cortex and right IFG, a region that has been implicated in inhibition and attentional control [49]. This upregulation of IFG is also consistent with the pattern of results observed within the TD group in previous work investigating the effects of a simultaneous aversive tactile distracter during a social cognition task [24].

Group differences were seen in a number of regions when completing the task in the presence of aversive sounds: the ASD group showed greater increased activity compared to the TD group in limbic and salience regions, including the amygdala and insula, as well as primary sensory regions and fusiform. In contrast, the TD group showed decreased activity compared to the ASD group in the fusiform and primary visual cortex. It is possible that the ASD group was upregulating amygdala and fusiform activity, in order to complete the face processing task despite distraction, given the role of these structures in emotional face processing [50]. However, it is more likely that the increase in amygdala and insula activity reflects greater aversiveness and/or salience of the auditory stimuli in the ASD group given that these regions have previously been found to show over-reactivity in youth with ASD during aversive sensory stimulation even in the absence of a social task [27‐29]. Additionally, the decrease in fusiform gyrus activity in the TD group is unexpected based on previous work showing that TD youth upregulate areas involved in language and social processing while completing a social cognition task in the presence of tactile distraction [24]. These different findings may be due to the fact that the emotional face processing task used in this study is a less complex social task that, unlike the previously used social inference task, did not require the integration of visual and auditory cues and therefore may be more automatic for the TD group. Our results showing that the ASD group displayed increased activity in lateral occipital cortex and primary visual regions when sound was added while the TD group showed decreased activity in these regions is consistent with previous work suggesting reduced differentiation and segregation of sensory systems in ASD [27, 51]. Overall, these group differences seen in neural face processing accompanied by auditory distraction are relevant for understanding how youth with ASD process noisy real-world environments that require multimodal sensory information processing.

Halfway through the task, participants were shown videos of their own faces making emotional expressions in an attempt to direct their attention towards relevant emotional cues expressed by others. After watching the self-video prime, both groups increased activation in visual processing areas when observing others’ emotional expressions, suggesting increased visual attention directed towards the faces, as well as in the hippocampus. The hippocampus plays a role in memory, including the recall of faces [52‐55], suggesting that participants may be retrieving memories of their own emotion expressions while processing others’ emotions.

Significant group differences were also seen in changes in response after the self-video prime. The ASD group showed greater signal increases in dmPFC and middle temporal gyrus when completing the task compared to the TD group. These results indicate that the self-video prime was engaging prefrontal cortex regions involved in mentalizing and social cognition [36, 37] for the ASD group, potentially by increasing the salience of facial emotional expressions. This increase in mPFC is consistent with previous work demonstrating that directing attention to relevant social cues increased activity in mPFC for youth with ASD during a social cognition task [34] and during the same social cognition task in the presence of aversive tactile sensory distraction [24]. This similar pattern of results across studies using different social tasks, different modalities of sensory distraction, and different methods of directing attention, indicates that more general mechanisms may be involved when directing attention to relevant social cues and that this mechanism has the potential to be broadly targeted in the development of social skills interventions.

Finally, we investigated the relationship between auditory SOR within the ASD group and changes in brain activity during the task with the addition of sound and after viewing the self-video prime. While on average, all youth with ASD showed increased vmPFC activity following priming, youth with lower SOR specifically showed greater increases in vmPFC activation in post-prime trials with sound compared to no sound. The vmPFC plays a role in modulating emotional responses by regulating limbic structures such as the amygdala [35]. Therefore, while greater vmPFC involvement following the prime may help emotion regulation in the ASD group in general, regardless of SOR, youth with lower auditory SOR may be better able to sustain this vmPFC engagement in the context of distracting sound. These findings are consistent with other research suggesting that youth with lower SOR are better able to regulate their amygdala response to aversive sensory stimuli [27, 28]. ASD youth with lower SOR also showed greater activation of brain regions involved in emotion regulation and higher-level social cognition after their attention was directed to social cues while completing a social inference task with simultaneous tactile distraction [24]. Here we also found that higher SOR correlated with greater increased activity after the prime in brain regions related to salience detection and sensorimotor processing, including the insula, cerebellum, thalamus, and basal ganglia. As the basal ganglia are thought to be involved in the non-conscious perception of emotional signals [56], the prime may be recruiting these alternative pathways for processing emotional facial stimuli in youth with higher SOR. Higher SOR also correlated with greater post-prime signal increases in the thalamus, which is known for its role in modulating sensory input [57]. Previous research has shown that youth with ASD have reduced thalamocortical modulation during aversive sensory exposure [58] so it is possible that the prime may be helping high-SOR ASD youth filter the background noise and direct attention back towards the social stimuli through increased thalamic modulation. Importantly, while overall there were no overall diagnostic group differences in brain responses to faces prior to the prime, within the ASD group, youth with higher SOR showed reduced activation in a number of regions including amygdala and basal ganglia regions involved in salience detection and reward. In contrast, higher-SOR youth did not show reduced activation in these regions after the prime, which is consistent with the idea that the prime may have helped to increase salience and attention to others’ facial expressions. This finding that SOR moderates the effect of the prime on brain activation during a social task is consistent with our previous work showing the neural effects of explicit attentional direction to social cues differ based on SOR severity [24], and likely indicates that despite similar behavioral performance, high and low-SOR ASD youth maybe be using different neural strategies to complete the same task. However, in contrast to the Green et al. [24] study, in which high-SOR youth used a likely low-efficiency, high-effort strategy of processing multiple social cues separately (e.g., tone of voice, facial expression, visual background cues), here we found that SOR was associated with post-Prime increases in brain regions related to salience detection, non-conscious perception of emotional signals, and modulation of sensory inputs and attention. It is possible that the implicit nature of viewing one’s own emotional faces was better able to elicit subconscious, lower-effort modulation processes that helped to increase the relative salience of the social cues compared to the extraneous background noise in contrast with the explicit attentional direction used in the Green et al. [24] study. Our finding that high SOR was associated with reduced activation in salience detection and social reward regions before but not after the prime, further supports the idea that for this subset of youth with ASD, viewing one’s own emotional faces might increase the salience and intrinsic reward value of viewing others’ faces. Notably, both the explicit and implicit cues increased mPFC activation for lower-SOR youth, suggesting that the type of intervention may be less important for low-SOR youth who are less distracted by multiple, competing incoming stimuli.

Overall, the emotion identification task used in this study was designed to approximate a real-world multi-sensory environment by asking participants to complete a social task in the presence of common, mildly aversive environmental noises. However, the task was still over-simplified compared to real social interactions when face processing and emotion identification are more subtle and complex, which may explain why both groups showed similar, high levels of behavioral task performance in all conditions and did not show overall diagnostic group differences in neural activation before the prime. In addition, particularly given the heterogeneity within the ASD group, the task had reduced power to look at diagnostic group differences within each condition independently. The simplicity of the task limits our ability to fully interpret the success of the prime, which needs further study to examine its effectiveness (a) in more complex, real-world social scenarios, (b) when integrated into a larger-scale social skills intervention, and (c) across a longer period of time. For example, if the implicit priming strategy does indeed reduce the effort needed to process social stimuli, this would become more clear with more complex social situations, where task performance might improve, and over a longer period of time, where an extended, implicit, social-salience-enhancing intervention might reduce SOR-related challenges such as emotional and behavioral dysregulation or social anxiety. Despite these limitations, this study is an important preliminary proof-of-mechanism step showing that an implicit social prime can indeed target brain regions important to salience detection, enhancing social-emotional processing, and modulating sensory inputs. Taken together with prior research (e.g., [24]), this suggests that even very basic attentional direction strategies have an effect on neural responses to social cues for youth with ASD, particularly in the presence of distracting extraneous sensory stimuli.

An additional limitation is the challenge of examining responses to auditory stimuli in the noisy MRI environment. Pilot testing was conducted to ensure that the sounds were audible and aversive over the sounds of the scanner, and noise-canceling headphones helped reduce the sound of the scanner, but comparing one noisy condition to another (e.g., auditory environmental sounds with MRI sounds to MRI sounds only) is likely to reduce group differences in neural responses. Consistent with this, our previous studies have found fewer group differences in fMRI responses to auditory stimuli alone vs. tactile stimuli alone or simultaneous auditory and tactile stimuli [27‐29]. Therefore, future studies might see greater group differences using a similar face processing task with a non-auditory modality of a sensory distractor.

Our task was also limited in that we were unable to conduct an additional control for potential order effects whereby neither group received the self-face prime. However, the most likely change we would expect due to order effects and time itself would be habituation and/or fatigue. The fact that we see greater activation in the Prime > Pre-Prime condition suggests that the prime is having an effect over and above habituation. In addition, an earlier behavioral study using this same task without auditory distractors tested for between-group effects of a self-face prime compared to an other-face prime, finding that only the self-face prime normalized reaction time for TD participants with high ASD traits [33]. This also suggests that the self-face prime is having an effect that is greater than pure timing effects. However, future studies with additional control groups who do not receive the self-face prime are needed to fully investigate this effect.

In addition, future research should investigate the effects of age upon both face processing and sensory over-responsivity in ASD. While our study took into account possible developmental effects by ensuring that groups were age-matched and covarying for age in all of our analyses, there is potential for future research to specifically examine these effects, especially given that younger participants were more expressive in their emotion videos used as the prime. Some of the differences seen in the face processing literature in ASD may be partially accounted for differing age of participants [11, 12], and one cross-sectional study indicated that functional connectivity of the fusiform face area correlates differently with age in ASD and TD groups [59]. Though not significant, the ASD group also trended towards being more expressive, which could relate to differences in developmental levels between the two groups. In addition, SOR has been found to decrease with age [60]. Therefore, future longitudinal research looking at the development of face processing, emotion expressiveness, and sensory processing in ASD and TD samples is needed to better understand the optimal target window for interventions.

Taken together, our findings have implications for designing interventions for youth with ASD. First, we demonstrated that a simple prime designed to direct attention to appropriate social cues led to youth with ASD recruiting underlying neural circuitry involved in social cognition, even in the presence of aversive environmental noises. This indicates that more complex social interventions focused upon explicitly increasing salience of important social cues compared to extraneous sensory distracters might be able to tap into these same neural mechanisms. Second, we demonstrated that youth with ASD increased activity in limbic and salience regions when completing the task in the presence of aversive environmental noises. This suggests that the sensory environment should be taken into account when delivering a social intervention in order to maximize its effectiveness and translation into real-world settings, which are often noisy and filled with competing sensory information. Finally, our results showed that for youth with ASD, the effect of viewing their own emotional expressions on neural processing of others’ expressions varied depending on their auditory SOR levels. Based on this finding, the heterogeneity of sensory challenges in ASD should be taken into account as a part of treatment, even when designing interventions that target another domain such as social skills. For example, youth with ASD entering a social skills program could be screened for SOR to specifically practice skills with differing levels of sensory distraction, as well as coping strategies for managing SOR, when indicated. Overall, this research highlights the importance of taking a holistic approach to treating ASD symptomatology including both sensory and social components.