The present study confirms the previous results reported by Lio et al.1 and extend these findings by dissociating the time course of neural processes involved in face perception in WS and ASD patients.
Our analyses confirm that activity in neurotypical participants can be evoked by viewing the eye region of a natural face stimulus. Furthermore, the evoked activity originates bilaterally in the superior temporal regions and peaks at 260ms after stimulus onset. Notably, we show that the evoked STS eye sensitive response is also present in the WS group, contrary to what is observed in ASD participants by Lio et al1. This reveals a dissociation among these two patients’ groups and suggest that although both syndromes are associated with social disturbances, their impairment at the neural level may have a different origin.
When examining face processing related neural signals along the occipito-temporal stream, we found two peaks in neurotypical participants: the first at 170ms, an early signal known to be implicated in low-level face features, the second at 260ms, a late component implicated in decoding salient face social cues 1. Remarkably, both components were found distinctly impaired and preserved in WS and ASD. In WS, we could weakly decode the 170ms signal probably due to their relatively poor ability to process faces’ morphology while the late 260ms component shown to be eye sensitive was highly significant. The reverse pattern was observed in ASD participants who showed neurotypical like early 170ms evoked activity but impaired late evoked 260ms signal.
Hence, we report a functional double dissociation between early (170 ms) and late (260ms) neurophysiological responses to faces in two neurodevelopmental disorders known to offer dissociable and opposing behaviors in facial processing and social cognition. Face processing is thought to involve the ventral pathway (through the OFA and FFA) for low-level features that are generally invariant, such as gender, age and identity, while the dorsal pathway (through the STS) is thought to process facial movements related to emotional expressions, gaze cues or intentions 69–71. The latter is also considered to be the gateway to an extended system of social perception encompassing the orbitofrontal cortex and the amygdala 38. Indeed, the present finding supports the hypothesis that early (170ms) visual processing of faces in the temporo-occipital area (FFA) and late social processing of faces in the temporal sulcus (i.e. STS) are functionally dissociable (Figure 4).
The present data have theoretical implications for understanding WS and their characteristic atypical processing of social cues, particularly faces. Indeed, our findings suggest that patients with WS, when viewing the eyes of a face, lack specificity in their early neurophysiological response. Early neurophysiological responses to faces at 170ms are evoked by facial features 22,23. This is thought to allow a consistent description of facial parts to be computed and processed in an holistic manner in the fusiform face area at 170ms 24,25,28.
Previous fMRI studies have shown that, during face perception, less activity can be observed in patients with WS than in neurotypical participants in areas specialized in early processing of face parts 54. Our data are consistent with these results and suggest that the FFA-driven process of combining parts of the face into a holistic representation may be specifically impaired in WS.
We show that early responses in patients with WS are not modulated by the type of facial features on display (i.e., eyes, eyebrow, mouth compared to other face-parts). Such an absence of modulation may suggest that patients with WS make little distinction between facial features and their possible relevance as a means for facial identification and subsequent holistic processing. Indeed, WS patients show increased activity at N170 amplitudes when viewing trustworthy versus untrustworthy faces 72,73.
In the present study, we applied blind source separation techniques to analyze high-resolution EEG data. EEG data were spatially processed to ensure that the source occurring at 170ms did not originate in the STS. We also show that WS patients produce robust responses to late processing of facial cues at 260ms in the STS. This observation is consistent with previous fMRI studies reporting greater STS activity in WS patients compared to patients with anxiety disorders while performing a facial perception task 54.
Unlike the study by Shore et al. 72, the altered neural response we found cannot be attributed to different patterns of eye fixation when looking at the face stimulus 16,55,74. Indeed, our method assessed the neurophysiological response to specific regions of the face presented to the foveae of each participant. This experimental modulation is extremely relevant to WS, as it has been hypothesized that the facial scanning behaviour of these patients is influenced by a general difficulty in disengaging attention from salient targets, known as “sticky fixation“ 75,76.
Our findings support two alternative interpretations. First, the role of the STS, which is typically implicated in late social processing of faces in neurotypical participants 38, may be upregulated in patients with WS, thus leading to their characteristic highly social behavior. As a consequence, early analytical computations performed by the FFA 20 would be bypassed, thus contributing to reduced performance in holistic face processing. Such interpretation would be coherent with reports of increased cortical thickness in the superior temporal gyrus of patients with WS 53. Such interpretation would also fit well with the frontal dysfunction hypothesis, expressed as an inability to inhibit impulsive responses to social information 44 and the amygdala dysfunction hypothesis, expressed by atypical processing of emotional information relevant to social interactions 45–47. Children with WS are known to respond differently to social cues conveyed by the face 8,77–79, and to process emotional cues atypically resulting in poor processing of social information from faces 80. If the early analytical computation is by-passed it could explain why soft signs are not typically captured by patients with WS. Yet, in the present study, we do not evidence any atypical performance over the late component in patients with WS.
A second interpretation may be that early visual activity in temporo-occipital areas is dysfunctional in WS and are thus unable to discriminate specific facial features. This interpretation is supported by previous reports showing that the FFA is both enlarged 49 and structurally altered 45 in these patients. In our study, the poor decoding of the facial cue regressor may also be related to their poor visuospatial abilities. In fact, in neurotypical participants the onset of the maximal decoding of the facial cue map regressor was predicted by visuospatial reasoning abilities (in the matrix subtest). That is, participants who showed poor visuospatial abilities where more likely to exhibit a late peak than those with good performance. Therefore, it is possible that the first facial process in the ventral part of patients’ brain poorly decode facial features due to poor visuospatial abilities. This could also explain why WS patients showed a lower performance in our gender task given their impairments in the perceptual holistic processes of faces. Subsequent cascading developmental consequences or compensatory strategies may induce a secondary upregulation of facial processing in other functional structures, namely the STS in this case, thus, leading to a concomitant increase in the social processing of faces. The role of cascading developmental consequences has already been theorized to explain atypical behaviors in other domains associated with Williams’s syndrome (such as mathematical impairments explained by their poor visuospatial abilities, 81). Such hypothesis is consistent with the atypical visual processing of faces from early infancy in WS 72,82.
Overall, the notion of cascading pathways between early (170ms) and late (260ms) processes remains putative and could be coincidental. However, we argue that their development is closely intertwined. Still, whether the early neurophysiological face processing deficit in WS emerges from the neurodevelopmental consequence of the disorder or as the result of the poor social interaction exhibited by WS patients remains an unresolved question.
What are the interventional implications of these results? Targeting STS and FFA in neurofeedback for WS and ASD
Overall, our results raise the possibility of developing functional and behavioural rehabilitative procedures for both WS and ASD based on neurofeedback, a procedure in which self-regulation is stimulated by providing online feedback of neural activity to participants 83,84. By controlling specific neural substrates, it can be possible to pinpoint and modify specific behaviours. Due to its high spatial resolution, fMRI is usually used to target specific cerebral structures with neurofeedback 85,86. Most procedures using EEG trigger training in EEG signal coherence or frequency 87. Rather than being based on modulation of disorder-specific biomarkers, most current EEG neurofeedback protocols are based on the modulation of a few spontaneous brain rhythms, mainly defined by the frequency of their oscillation 83,84. This strategy is widespread because spontaneous brain rhythms demonstrate a high signal-to-noise ratio in EEG recordings and can also be disrupted in some mental disorders (e.g., ADHD 88). However, within the broad field of psychiatric disorders, such methods lack specificity.
Here, ERPs could be used to feeding back levels of local cortical arousal back to patients with the purpose of improving the self-regulation of cortical excitability in the specific structure related to the disorder 89. In order to respond to the specific needs of patients, training would be provided beforehand. Indeed, it has been demonstrated that patients with severe intellectual impairments, like some patients with ASD and WS, are able to follow training procedures to enhance the effectiveness of neurofeedback 90.
Our results provide unique and specific biomarkers associated with face processing deficits in ASD and WS. In light of these findings, it may be beneficial to explore whether neurofeedback can help children with ASD or WS to enhance STS or FFA activity during face processing. Here, the goal would simply be to manipulate neural activity in the STS or FFA in order to rebalance social processing of face related to this structure, i.e., the identification of facial features. Neural activity in the STS or FFA could be measured by high-density EEG which could then be represented in real time, through visual, audio or other means, back to participants while the focused area on the face different facial areas (either social, eyes, nose, brows, mouths or not) to facilitate self-regulation of cortical excitability underlying the specific behaviour. This might be achieved with the use of one signal for the FFA (e.g., visual) and another signal for the STS (e.g., auditory) to assist patients in reassigning their attention to facial features for the purposes of identity recognition or social cognition 17.
One would expect such training to facilitate the identification of a neural strategy that can be employed by patients when viewing faces. Indeed, the idea would be to train patients both at the behavioural level to pay attention to socially relevant facial cues but to also to train them to engage their FFA and STS when viewing specific facial features (e.g., eye, eyebrows, mouth). The long-term goal could be that both groups of patients would engage the ventral and-dorsal areas in a more balanced way during social interactions.
More generally, facial and social skills are essential abilities for navigating in our modern society. With the ongoing pandemic, the ability to decode social information from faces on a screen or from the eye region (as faces are partially occluded with masks) has recently become one of the main tools in social interactions 91. This ability relies on a complex cerebral network 28. Double dissociations using fine spatiotemporal analysis of this network provide essential evidence for understanding what, where and when neurocomputations are performed in our brain. In patients with Williams syndrome, poor decoding of facial features by low-level visual processing can result in some of the observed social symptoms that can have catastrophic influences on their lives. Future works may examine whether, in turn subtle changes in the balance of this network, through self-regulated neurofeedback training, can attenuate their characteristic social behaviour and related symptoms of social dysfunctions, such as anxiety.