Edited by: Daniela P. Skwerer, Boston University, USA
Reviewed by: Katarzyna Chawarska, Yale University School of Medicine, USA; Brandon Keehn, Children's Hospital Boston, USA
*Correspondence: Robert T. Schultz, Center for Autism Research, Children's Hospital of Philadelphia, 3535 Market Street, 8th floor, Suite 860, Philadelphia, PA 19104, USA. e-mail:
This article was submitted to Frontiers in Developmental Psychology, a specialty of Frontiers in Psychology.
†These authors are co-first authors and contributed equally to this research.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in other forums, provided the original authors and source are credited and subject to any copyright notices concerning any third-party graphics etc.
Although the extant literature on face recognition skills in Autism Spectrum Disorder (ASD) shows clear impairments compared to typically developing controls (TDC) at the group level, the distribution of scores within ASD is broad. In the present research, we take a dimensional approach and explore how differences in social attention during an eye tracking experiment correlate with face recognition skills across ASD and TDC. Emotional discrimination and person identity perception face processing skills were assessed using the Let's Face It! Skills Battery in 110 children with and without ASD. Social attention was assessed using infrared eye gaze tracking during passive viewing of movies of facial expressions and objects displayed together on a computer screen. Face processing skills were significantly correlated with measures of attention to faces and with social skills as measured by the Social Communication Questionnaire (SCQ). Consistent with prior research, children with ASD scored significantly lower on face processing skills tests but, unexpectedly, group differences in amount of attention to faces (vs. objects) were not found. We discuss possible methodological contributions to this null finding. We also highlight the importance of a dimensional approach for understanding the developmental origins of reduced face perception skills, and emphasize the need for longitudinal research to truly understand how social motivation and social attention influence the development of social perceptual skills.
Face recognition is one of the more thoroughly studied skills in the field of autism research (Wolf et al.,
Face processing is believed to be a universal domain of expertise in humans and perhaps one of the earliest to develop (Gliga and Csibra,
Reduced attention to and motivation for engaging with face stimuli is a prominent hypothesis for why children with ASD might, on average, have reduced face perceptual skills (Schultz et al.,
Differences in social attention appear to be one of the earliest signs of autism. For example, a preference for non-social patterns (e.g., geometric shapes) in toddlers is a robust risk factor for developing the disorder (Pierce et al.,
The present research aims to provide a more direct test of the link between social attention and face perception by examining spontaneous attention to faces and objects in participants occupying the entire face expertise continuum. Prior research using ASD and typical participants has focused on group means, overlooking within-group variability. An alternative approach is to ignore diagnostic categories and boundaries and adopt a more dimensional approach (Insel et al.,
In the present study, participants' gaze was tracked as they watched movies of actors showing different facial expressions and videos of non-social moving objects (e.g., a bulldozer pushing earth, clothes on a line flapping in the wind) in the same display. The four videos composed a 2 by 2 design, faces vs. objects that were either of high vs. low salience (e.g., faces gazing directly at the camera vs. averted; bulldozers vs. clothing). This study tested the following hypotheses:
Attention to faces correlates with face perception accuracy as measured by two subtests of the Social skills (as measured by the SCQ) predict social attention and face perception skill. On average, the ASD group will score lower on face perception tests and will spend less time attending to social information.
We studied 110 children and adolescents, including 60 diagnosed with an ASD (7 female) and 50 typically developing controls (TDC; 12 female). ASD and TDC groups were matched on non-verbal cognitive ability as measured by the Differential Ability Scales, Second Edition (DAS-II, Elliot,
Mean age in years (SD) | 11.28 (2.89) | 11.34 (3.04) | 0.10 | 0.92 |
Age range | 6.17–17.92 | 6.33–17.92 | ||
Mean GCA (SD) | 111.63 (14.61) | 113.70 (14.58) | 0.74 | 0.46 |
GCA range | 88–158 | 87–150 | ||
Mean verbal (SD) | 110.12 (16.61) | 116.42 (16.70) | 1.98 | 0.05 |
Verbal score range | 77–161 | 89–165 | ||
Mean non-verbal (SD) | 111.07 (15.48) | 108.26 (13.71) | −1.00 | 0.32 |
Non-verbal range | 84–166 | 80–143 | ||
Mean LFI score (SD) | 78.83 (7.29) | 82.70 (7.78) | 2.69 | 0.008 |
LFI range | 61.67–96.66 | 65.00–96.66 | ||
Mean SCQ score (SD) | 20.67 (5.61) | 1.12 (1.29) | −24.11 | 0.000 |
SCQ range | 11–34 | 0–4 | ||
Chi-Square | ||||
Sex: Male | 53 of 60 | 38 of 50 | 2.90 | 0.09 |
Module 3 ( |
2.94 (1.29) | 7.08 (2.73) | 10.02 (3.69) |
Module 4 ( |
3.57 (1.72) | 7.13 (1.46) | 10.50 (2.83) |
The
a. The Matching Identity Across Expression subtest evaluates a child's ability to recognize facial identities across changes in expression (happy, angry, sad, disgusted, and frightened). A target face is shown alone for 500 ms, followed by three probe faces of different identities presented simultaneously with the target face. Children must select the face that matches the target's identity ignoring the fact that the expression is different.
b. The Matchmaker Expression subtest assesses the child's ability to match emotional expressions across different identities. Five basic emotions (sad, angry, happy, frightened, and disgusted) were tested. A target face depicting a basic emotion in frontal profile was shown alone for 1000 ms and then remained on the screen as three probe faces of different identities conveying different expressions were presented. Children must select the face with the expression that matches the target.
Participants were calibrated at the beginning of the experiment using a standard five-point calibration procedure. The experiment included twelve 15-s trials consisting of four silent videos playing concurrently, one in each quadrant of the screen (pseudo-randomized location). In order to minimize the predictability of the display, a jitter was introduced so that the videos were not consistently placed right in the center of each quadrant. The distance in pixels from the center of the screen to the mid-point of each image did not differ between conditions [Face clips
At the beginning of each study visit, parents provide informed consent for their child; participant assent was obtained when feasible. Next the DAS-II and the ADOS were administered to the child while parents completed the ADI-R. After a lunch break, children completed the eyetracking task and the LFI tasks. Eye tracking took place in a quiet room containing a chair and a 30-inch computer screen on an adjustable table. A Tobii X120 gaze tracker recorded participants' looking patterns at a rate of 60 Hz from a seated distance of approximately 60 cm. Above the computer monitor, a webcam simultaneously recorded a video of the participant. Participants were informed that they would see a few short videos, and were asked to watch the screen.
All participants and parents received oral feedback at the time of the visit, as well as a written report, and compensation for time and travel. The Institutional Review Board at The Children's Hospital of Philadelphia approved all procedures related to this project.
Accuracy scores from the two
Tobii software produces a variable called Total Fixation Duration, which is the sum total length of all fixations within a given AOI. It is often used as a measure of preference for looking at one stimulus type over another (Klin et al.,
Two types of analyses were performed. First, linear regressions were constructed to assess whether social attention predicts face processing skill and gaze to faces. Preliminary analyses revealed that age was significantly correlated with face processing skills (Pearson's
Forty-two different participants (23 with ASD, 19 TDCs, all male, average age = 15.03 years, average IQ = 105.45) were recruited using the criteria described in the participant section above. These participants were asked to complete the eye tracking experiment at two time points separated by a 9-week interval (±1 week). An intraclass correlation coefficient (ICC) was computed using a two-factor mixed-effects consistency model (Farzin et al.,
The Social Motivation theory of autism argues that varying levels of social motivation modulate experience with faces over the course of development, and ultimately impact children's face processing skills. A two-step multiple regression analysis was therefore used to discern whether visual attention to faces predicts face perception skill in the combined sample of ASD and TDC participants. Age was entered in Step 1, as preliminary analyses suggested that face processing skills are positively correlated with chronological age (Pearson's
Age | 0.52 | 6.20 | 0.000 | 0.23 | 0.23 |
Gaze to faces | 0.27 | 2.38 | 0.019 | 0.27 | 0.04 |
Next, we tested whether scores on the SCQ (a measure that evaluates autistic symptomatology, including social communication skills) predicted total fixation duration to faces and face perceptual skills. While the SCQ not a measure of social motivation
Age | 0.48 | 5.74 | 0.000 | 0.23 | 0.23 |
SCQ score | −0.27 | −3.41 | 0.000 | 0.31 | 0.08 |
To determine whether total fixation duration to faces differed by stimulus type, salience level, and diagnostic group, a 2 (Type: face/object) × 2 (Salience: high/low) × 2 (Diagnosis: ASD/TDC) repeated measures ANOVA was conducted. This analysis revealed a main effect of Type,
We began our analyses with very strong a priori hypotheses about gaze in ASD versus TDC participants, based on a significant body of research (Klin et al.,
Long segments of gaze data may obscure meaningful eye movements that occur in the first few seconds of an experiment (Swingley et al.,
After exhausting the possibilities, we determined that our original finding, while surprising given the broader literature, was undeniably accurate. As discussed below, we speculate that the object movies in our paradigm may have been too appealing to reveal group differences that other paradigms with more subtle manipulations were able to document.
We aimed to answer three questions with this study: First, does visual attention to faces predict face expertise? Confirming our hypothesis, we found that increased gaze to faces relative to objects was a significant positive predictor of children's scores on the
Our second hypothesis, that social skill as measured by the SCQ would predict visual attention and face expertise, was partially confirmed. Although children's scores on the SCQ did not predict eye gaze, they did predict face expertise. One obvious limitation of this measure is that the SCQ is not specifically designed to gauge social motivation, which may explain the lack of correlation with visual attention. Future research using an instrument that measures social motivation more directly (such as the Pleasure Scale, Kazdin,
Consistent with past work (Klin et al.,
In conclusion, our study treated face processing skills as a dimension that spanned both children with ASD and TDC and found that amount of time spent looking at faces during eye tracking predicts face processing skill on an independent measure. This process-based analysis is consistent with a growing emphasis on using dimensional approaches in other areas of mental health research, as captured by the NIMH's new focus on research domain criteria (Insel et al.,
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
This work was supported by grants from the Robert Wood Johnson Foundation #66727 (Robert T. Schultz), Pfizer (Robert T. Schultz) and by the Pennsylvania Department of Health SAP #4100042728 (Robert T. Schultz, Coralie Chevallier, Julia Parish-Morris) and 4100047863 (Robert T. Schultz, Janelle Letzen, Natasha Tonge).