Background
22q11.2 deletion syndrome (22q11.DS), also known as DiGeorge or velocardiofacial syndrome, is one of the most frequent microdeletions, with an incidence of approximately 1:2000 in pregnancies and 1:4000 births [
1‐
3]. The syndrome has been associated with a characteristic facial appearance, hypernasal speech, cardiac anomalies, learning disabilities, attention deficits, and social impairments [
4,
5]. Specifically, the 22q11.2DS social phenotype is characterized by social and emotional withdrawal, high rates of shyness and anxiety disorders, and difficulties initiating and maintaining social interactions [
4‐
9]. Studies on other clinical populations with social dysfunction such as Down and Williams syndromes, autism spectrum disorder (ASD), developmental prosopagnosia, and schizophrenia have highlighted the implications of face processing skills for social interaction [
10‐
12]. The investigation of face processing was also found to be highly relevant to gain a better insight into the mechanisms associated with social difficulties in 22q11.2DS [
13,
14]. Several studies using eye-tracking technology have reported alterations of visual scanpaths during face and emotion processing in small samples of children and adolescents with 22q11.2DS compared to control groups [
13‐
16]. The mean age of individuals participating in these studies ranged from 12.36 to 17.4 years old. When looking at non-emotional/neutral or emotional faces, children and adolescents with 22q11.2DS fixated more on the mouth and less on the eyes than typically developed (TD) and developmentally delayed groups and tended to have fewer fixations and shorter scanpath lengths than controls [
13,
14]. Other studies on emotional face processing showed that adolescents with 22q11.2DS fixated less on face stimuli than controls, which is reminiscent of results in individuals with ASD [
15,
17]. Finally, McCabe et al. [
17] argue that a failure to change exploration patterns according to the content of the visual information (i.e., perseverative and inflexible behavior) may play a role in the aberrant pattern of fixations on faces in 22q11.2DS. These eye-tracking findings were examined in relation to several aspects of the 22q11.2DS phenotype in order to understand their relationship with social difficulties. In particular, Glaser et al. [
13] observed a significant negative association between time spent on the eyes and higher rates of anxiety, which could suggest a link between impairments related to the processing of socially relevant stimuli and the socio-emotional dysfunctions found in 22q11.2DS.
The ability to process faces improves with age and is associated to the comprehension of emotional and mental states and to an adequate communication and behavior during social interactions [
12,
18,
19]. Previous studies have differentiated two core mechanisms of visual scanning of faces: configural face processing (CFP) and featural face processing (FFP) [
20,
21]. FFP (or component processing) refers to the exploration of individual parts of a face, such as contour, color, and shape of the facial features (e.g., nose, eyes, mouth). On the other hand, CFP refers to the analysis of spatial distances between the features. CFP contributes to the achievement of a high level of expertise in face recognition and in emotion recognition [
22‐
24]. From a developmental perspective, CFP develops significantly later than FFP [
25,
26]. Whereas some studies indicate that CFP is already adult-like by the age of 10 [
25,
26], others suggest that we reach proficiency in CFP in adulthood only [
11,
19]. Although CFP and FFP show different developmental trajectories, together they account for the expert skills observed in adulthood [
19,
27].
Evidence for atypical development of CFP has been reported in several populations with social impairments. In order to measure CFP and FFP, face discrimination tasks were administrated to participants [
26]. In these tasks, faces were modified according to the specific type of face processing: configural changes affected the distances between features and featural changes included differences in features without manipulating the distances between them (e.g., replacing the eyes with the ones of another person). Lower accuracy in discriminating faces with configural changes has been reported in children with neurodevelopmental disorders (e.g., ASD, Williams and Down syndromes) compared to TD participants [
11]. This observation suggests that the development of CFP is fragile and easily altered in neurodevelopmental disorders. Nonetheless, high-functioning adults with ASD were shown to have similar CFP compared to controls, which reinforces the usefulness of examining CFP from a developmental perspective [
28]. To our knowledge, only one study explored FFP and CFP in 22q11.2DS and demonstrated impairments and lack of improvement with age in both types of processing [
13]. To gain further insight into the development of CFP and FFP, it is necessary to investigate these difficulties in a broader age range and verify whether they expand through adulthood by using longitudinal designs.
Eye-tracking technology is a promising tool used to gather valuable information, such as eye gaze, which is not easily observable by experimenters. Hence, eye-tracking collects data regarding “when” (temporal) and “where” (spatial) the attention was allocated to a stimulus, such as the number of fixations, percentage of time spent, fixation duration, location of first fixations, and number of transitions between areas of interest [
29]. Previous studies [
30‐
32] also recommended examining biases that could occur during the early phase of information processing (e.g., the landing positions of first and second fixations, the sequence of first fixations). Accordingly, the initial fixation positions could influence the pattern of fixations that will follow during visual scanning and are important for an optimal information extraction. For instance, it was proven that the first two fixations suffice to achieve performance in a face-recognition task [
32]. Furthermore, recent eye-tracking studies revealed that individuals with disorders were less likely to return to the eyes region during face scanning and more likely to hyperscan the remaining facial features [
33,
34].
To date, no study has used combined measures of accuracy in discriminating faces with configural and featural changes and eye-tracking to explore CFP and FFP, along with their development over time in 22q11.2DS. In the present study, a large sample of participants with 22q11.2DS and TD with a wide age range performed a face discrimination task (“Jane task”), in which participants are presented with two portraits with variations in the individual features (FFP) or spacing of the features (CFP) and asked to judge whether they are identical or different [
13,
26]. Our first aim was to extend the results previously described by Glaser et al. [
13] on an overlapping sample and to further characterize face processing alterations in 22q11.2DS. Taking into consideration the meaningfulness of first and second fixations for information processing [
30‐
32], we specifically examined the first two landing positions (taken separately) in both groups. We expected to observe the presence of markers indicating an atypical face processing that could occur at the beginning of visual exploration. We therefore hypothesized that participants with 22q11.2DS would show a reduced tendency to look at the eyes during both first and second fixations on faces compared to TD participants. Based on previous evidence on perseverative visual exploration [
13,
17], we also expected to observe fewer transitions between the faces and longer fixations in participants with 22q11.2DS. Our second aim was to compare visual scanning patterns during FFP and CFP. Given that difficulties found in clinical populations depend on the type of face processing (CFP or FFP) [
11,
17], we hypothesized that participants from both groups would show different patterns during the exploration of faces modified on a configural or featural level. Therefore, we expected that eye movements would adapt to the type of face processing induced by the stimuli only in TD group during face exploration. Thirdly, we investigated changes in CFP and FFP occurring from childhood through adolescence using longitudinal data available in a subsample of participants. Based on the atypical development of face processing found in individuals with neurodevelopmental disorders [
11,
13], we expected to observe a lack of improvement with age in the ability to discriminate configural differences in participants with 22q11.2DS. Finally, we examined whether alterations in discriminating configural and featural differences would relate to the presence of social difficulties in participants with 22q11.2DS. Specifically, we expected to observe significant associations between CFP abilities and clinical measures of social impairment (anxiety, emotional and social withdrawal, poor socialization).
Discussion
The current study replicated previous findings [
13] and covered new aspects regarding the face processing in 22q11.2DS in a longitudinal sample. Hence, we showed that difficulties in face processing persist throughout a broad age range of individuals with 22q11.2DS, from childhood to adulthood. Particularly, participants with 22q11.2DS showed a bias toward the mouth region during the first two fixations on faces and an increased number of fixations toward this specific region from the first to the second fixation. They also had longer fixations and spent more time looking at less relevant regions. This pattern is indicative of a perseverative type of scanning which could impair flexibility and the identification of relevant information during visual exploration of human faces. Participants with 22q11.2DS did not show distinct visual scanpaths between CFP and FFP, whereas the TD group used different eye scanning strategies depending on the type of processing. We also showed that CFP accuracy improved less drastically with age in the 22q11.2DS group, compared to the TD group.
The first aim of our study was to delineate face scanning in 22q11.2DS using various eye-tracking parameters: first and second fixation locations, average fixation duration, number of transitions during face exploration, and scanpath patterns during CFP versus FFP. First, we found a higher proportion of first fixations on the mouth in the 22q11.2DS group compared to the TD group and a higher proportion on the eyes in the TD group compared to the 22q11DS group. Unlike the comparison group, we did not observe an increase in the number of fixations on the eye region from the first to the second fixation in participants with 22q11DS. Rather, the 22q11.2DS group increased their fixations on the mouth between the first and second fixation. This indicates the presence of an early bias during face processing, leading individuals with 22q11.2DS to direct their attention toward the mouth and to gradually increase their fixations on this specific feature. Similarly, several studies have already reported that participants with social anxiety disorder, social phobia, and schizophrenia are less likely to look at salient features when exploring faces [
33,
34,
50‐
52]. Future research should further investigate the dynamic and the temporal evolution of the initial fixation positions to better understand the mechanisms underlying these particularities. Although individuals with 22q11.2DS tend to look more at the eyes compared to the mouth or nose, their scanpath remains atypical when compared to the TD group. Indeed, the proportion of time spent on mouth and features outside of the face region (shower cap) was greater than what was observed in the TD group. These results could indicate that participants with 22q11.2DS have difficulties to identify and maintain attention on socially relevant features (e.g., eyes) during face processing, which is line with previous research [
13,
17]. Another possible interpretation for these findings is that an early bias toward the mouth and a return to the scanning of non-salient features occurring during the initial phase of face processing impair an optimal information extraction in participants with 22q11.2DS. These results have implications for the design of socio-emotional intervention programs aimed at improving efficiency in visual exploration and correcting face scanning from the very first fixations in individuals with developmental delay [
53,
54]. Using cueing techniques, either explicit (e.g., a verbal cue) or implicit (e.g., a cross on the screen), may be one way to correct the bias to the mouth and improve sensitivity to configural changes in salient facial features. Further, targeting CFP may allow us to indirectly improve emotion and face recognition [
23,
24]. For example, Russell et al. [
55] found that verbal instructions during an emotional training task with patients with schizophrenia improved their emotion recognition by re-directing their attention to relevant facial features. Hadjikhani et al. [
56] showed that visual cueing guiding the attention of individuals with ASD to the faces activates fusiform face area (FFA), a brain region specialized for face perception. However, these studies did not investigate whether the observed improvements in face processing after intervention are associated with changes in social functioning. Further work is therefore required to address this question more in depth.
Furthermore, participants with 22q11.2DS demonstrated longer fixation durations than the comparison group, confirming results from previous research [
17]. Longer fixation duration is usually described as a marker of higher cognitive load and may thus indicate a more effortful information processing [
29]. Our results also showed a reduced number of transitions between the two faces in the 22q11.2DS group. Consequently, we can assume that simultaneously exploring and comparing two faces was more difficult for participants with 22q11.2DS and involved more cognitive resources [
17], which may lead to a perseverative and a less organized exploration (see also [
53]). McCabe et al. [
17] also reported poor performances on a task depicting faces in participants with 22q11.2DS and suggested that cognitive inflexibility could explain failure in meeting task demands when looking at more complex stimuli, such as human faces. Given that no significant association between IQ scores and fixation durations or number of transitions was found, the obtained results could be specific to the processing of social stimuli.
Finally, eye-tracking data highlighted distinct scanning paths during the CD vs. FD trials in the TD group only. A possible interpretation for this finding is that individuals with 22q11.2DS failed to modulate their visual scanpaths in order to adapt to stimuli and task demands. Accordingly, McCabe et al. [
17] also found evidence of maladaptive visual scanning when participants with 22q11.2DS were looking at faces vs. weather scenes. This finding may indicate a lack of consistency when scanning faces (more variability in the way faces are explored) due to inattention and/or a difficulty adapting the scanning strategy to the context [
13,
17,
57].
The second aim of our study was to examine FFP and CFP between groups and the developmental trajectory of CFP. Even though identifying configural differences was more difficult for both groups than identifying featural differences, CFP was proportionally more difficult than FFP for individuals with 22q11.2DS. They showed a larger gap in accuracy between CFP and FFP performances relative to TD individuals. As expected, these results extended the findings of a previous report [
13] to a broader age range. Moreover, in accordance with our hypothesis, the longitudinal analyses revealed that individuals with 22q11.2DS improved to a lesser extent than TD participants in configural processing over time. This finding confirms in a broad age range the results previously reported in a cross-sectional sample [
13]. A pronounced gap between featural and configural face processing in our 22q11.2DS sample could commensurate with structural or functional alterations in related brain regions. To date, few neuroimaging studies have provided evidence regarding the presence of separate cerebral pathways for the processing of featural and configural information [
58,
59]. For example, Renzi et al. [
58] conducted a Transcranial Magnetic Stimulation (TMS) study using configural and featural stimuli from the Jane task in a sample of healthy young adults. They found that the right inferior frontal cortex was responsible for the configural processing of faces, whereas the left middle frontal gyrus played a role for featural processing of faces. Furthermore, fMRI studies showed that individuals with 22q11.2DS have reduced activity compared to controls in cerebral regions involved in social cognition while looking at emotional faces [
60] and show alterations in frontal brain regions [
61,
62]. Combining fMRI and eye-tracking would enrich our knowledge regarding the cerebral regions contributing to configural face processing deficits in 22q11.2DS. The correlations between eye-tracking measures and behavioral results (task performance) were also investigated, but no significant association between these variables was observed. Contrary to what could have been expected, an increased time spent and number of fixations on the mouth did not help individuals with 22q11.2DS to distinguish the differences between the portraits and obtain a similar performance to TD in FD trials. These observations raise important questions regarding the link between eye gaze and performance on tasks using face stimuli, and this topic was already debated in previous studies [
63,
64]. Our results might suggest that the eye-tracking measures used in the current study are insufficient for understanding task performance. Other mechanisms such as idiosyncratic scanning strategies and encoding process of visual stimuli and visual memory of faces could also contribute to their performance [
63,
65,
66]. Future research should consider the use of the Dynamic Scanning Index (reflecting the number of times that the eye gaze goes in and out of a core feature) that might better reflect task performance [
64].
Finally, we observed that individuals with more negative symptoms were characterized by lower performance on CD trials. As CFP has been shown to be important not only for face recognition but also for emotion recognition [
23,
24], it is likely that difficulty perceiving configural changes may alter the ability to perceive emotional face changes during social interactions. However, the association between CFP and negative symptoms is rather modest and should be interpreted with caution. Contrary to our expectations, none of the eye-tracking or behavioral measures correlated with the anxiety measures. This result can be explained by the choice of our measures. Due to the wide age range of our sample, the anxiety measures used in the current study were different and less exhaustive than the ones used in Glaser et al. [
13]. Further work is necessary to better understand the link between face processing, psychopathology, and social problems in 22q11.2DS.
The present study has several limitations that should be considered when interpreting the results. First of all, we cannot conclude from this study whether the atypical developmental trajectory of configural processing is face-specific. To answer that question, the development of configural and featural processing should also be explored in non-face stimuli (see [
67]). For example, Giersch et al. [
68] found evidence for spatial processing impairments during a discrimination task involving geometric forms, which may be suggestive of a global deficit in the configural processing in 22q11.2DS. A previous fMRI study [
69] found that participants with 22q11.2DS did not show face-specific responses in the fusiform gyrus, while responses related to objects (houses) were intact in parahippocampal gyrus and similar to TD. This finding suggested that participants with 22q11.2DS are characterized by face-specific processing difficulties, rather than deficits in basic visual processing. Although a recent study showed that visual perception and processing, particularly form perception, predict performance on facial identity recognition in 22q11.2DS [
70], no study to date has explored the link between basic perceptual and spatial processing and CFP and FFP. In the present study, we examined correlations between CFP and FFP performance and IQ measures related to full-scale score and visual processing (POI and PIQ) but did not find any significant association. This could indicate that the mechanisms underlying CFP and FFP are specific to the processing of social stimuli, but this interpretation needs to be considered with caution. Given that low IQ is an intrinsic characteristic of individuals with 22q11.2DS, the group (22q11.2DS vs. TD) and IQ variables are confounded. Hence, the observed differences could also be the result of intellectual impairments. The inclusion of IQ as a covariate in the analyses could be misleading and conduct to a bias caused by overadjustment [
71‐
73]. Further work is required to better address this topic. For instance, the inclusion of a group with non-syndromic intellectual disability in future studies could help to better understand the impact of IQ on the different types of face processing. Another limitation concerns the choice of stimuli. In daily life, variations in facial features are more complex and social interactions are based on dynamic changes in facial expressions, whereas the stimuli used in this study are limited to static portraits of one individual. Another observation regarding the choice of stimuli is that some of the modified faces have an unnatural appearance, despite the fact that they were created based on anthropomorphic norms to maintain a natural quality [
26]. Finally, eye-tracking has several limitations that have been described by Bojko [
29]. Briefly, eye-tracking technology detects the foveal vision (center of the retina) and does not give information about the periphery of the visual field. For instance, upright faces can be processed by extracting information out of the foveal area only [
20]. Furthermore, eye-tracking does not provide information about why a person looks at a stimulus of interest, or whether they understand what they see. However, reliable reporting about scanning behavior after completing a task is difficult to obtain in youngsters with developmental disabilities. Despite these drawbacks, eye tracking remains a useful technique that allows the investigation of scanning patterns and perception biases in clinical populations [
74].