This study examined visual attention to faces and a variety of control stimuli in individuals with ASD. Relations among fixations, face and pattern recognition, and social-emotional functioning were also explored. Visual fixations were measured while participants passively viewed five different homogenous classes of stimuli including upright human faces, inverted human faces, monkey faces, three dimensional curvilinear forms (Greebles), and two-dimensional geometric patterns. Typically developing individuals and those with ASD displayed comparable patterns of visual attention across stimulus categories. Both groups tended to focus attention to the upper regions of visual stimuli and less attention to the lower regions, particularly for upright face stimuli. The only between group difference to emerge was an interaction effect indicating that, when comparing upright faces, three-dimensional objects, and geometric patterns, individuals with ASD devoted proportionally greater attentional resources to the upper, relative to lower, portions of visual stimuli.
In contrast to our predictions, the sample of children with ASD in this study exhibited normative patterns of visual attention to human faces despite face recognition impairments and significant social deficits. These findings are consistent with other research investigating visual attention to static face stimuli in ASD (Sterling, et al., 2008
; van der Geest, et al., 2002
) and may reflect the diminished realism and consequently decreased ecological validity of grayscale, static stimuli (Speer, et al., 2007
). Dynamic in vivo social interactions may be better able to characterize the variability in eye gaze among individuals with ASD (Klin, Jones, Schultz, & Volkmar, 2003
Details of the experimental design may also have influenced results. Extended viewing times of eight seconds may have affected viewing patterns, as previous studies using shorter viewing times (e.g., 2 seconds; Pelphrey, et al., 2002
) have detected differences while those employing longer viewing times (e.g., 10 seconds; van der Geest, et al., 2002
) have not. The large stimulus display size in the current study may also have impacted scan patterns. There is evidence that, when individuals with ASD are provided with sufficient time and visual resolution to process high spatial frequency information, they display, not only more typical patterns of face recognition, but more typical patterns of brain activity associated with face processing (Hadjikhani, et al., 2004
). The absence of any competing stimuli on-screen stimuli may also have influenced results, as several studies documenting anomalous viewing patterns simultaneously displayed alternative sources of visual information (Klin et al., 2002b
; Vivanti, Nadig, Ozonoff & Rogers, 2008
). Face scanning atypicalities as a consequence of difficulties in attention shifting or disengaging might only emerge under such circumstances.
The variability among research studies investigating visual attention to static faces in ASD mirrors that observed in the clinical phenotype. Although social deficits represent a pervasive and unifying feature of ASD, there is much variation in the manifestation of this characteristic. Indeed, considering that diagnosis requires that only 6 of 12 diagnostic criteria be met, children may qualify for identical diagnostic labels despite distinct symptom profiles. The diversity observed in clinical manifestation and evident in the present study has also been noted in face processing research using other methods. For example, findings of electrophysiological anomalies in the processing of faces (Dawson, et al., 2002
; McPartland, et al., 2004
; O'Connor, Hamm, & Kirk, 2005
; Webb, Dawson, Bernier, & Panagiotides, 2006
) have failed to replicate in some samples (Kemner, Schuller, & van Engeland, 2006
; Webb, et al., 2009
). Hypoactivation of the fusiform gyrus during face perception (Schultz, 2005
; Schultz, et al., 2000
) has also failed to manifest in some samples (Hadjikhani, et al., 2004
; Hadjikhani, Joseph, Snyder, & Tager-Flusberg, 2007
; Kleinhans, et al., 2009
; Kleinhans, et al., 2008
; Pierce, Haist, Sedaghat, & Courchesne, 2004
; Pierce & Redcay, 2008
). The current results do not clarify whether individuals with preserved face perception represent a meaningful subgroup within the autism spectrum or simply variation in one aspect of social behavior. This heterogeneity represents informative variance, offering descriptive information about individual differences at a level not yet permitted by extant diagnostic categories.
Differences in visual attention to faces may also reflect developmental variation. We have suggested elsewhere (Dawson, Webb, & McPartland, 2005
) that atypical scan patterns to human faces may represent a developmental consequence of reduced attention to people in early childhood. It then follows that variability in social motivation would predict variability in scan patterns and that this developmentally-induced heterogeneity would be most evident in adolescents and adults. Thus, it is possible that preserved looking patterns in the current sample may reflect high levels of social motivation during development. Though formal quantification of social motivation was not obtained in the current study, clinical interviews suggested strong social interest despite underdeveloped social skills, corresponding to the “active but odd” classification ASD described by Wing and Gould (1979)
. Individuals with these characteristics might obtain higher levels of developmental exposure to other people and consequently increased experience with human faces and thus display more normative looking patterns. It will be important in future research to examine face processing in ASD in light of social motivation. Indeed, in typically developing populations; personality characteristics associated with social motivation, such as extroversion and introversion, modulate neural responses to faces (Cheung, Mayes, Rutherford, & McPartland, in press
). In considering the influence of development on face scanning, evidence for eye-biased scan patterns in neonates suggests that, to some extent, the typical pattern of upper face preference may be present from birth (Farroni, Csibra, Simion, & Johnson, 2002
; Farroni, et al., 2005
Our results may also reflect diagnostic distinctions within the autism spectrum. The symptom profile of all participants in the current study was consistent with DSM-IV-TR criteria for Asperger Syndrome. The absence of differential looking patterns between clinical and control groups suggests possible preservation of fixation patterns to faces in this diagnostic subgroup (but see Corden, et al., 2008
). The absence of clinical comparison groups in this study prevents analysis of fixation differences among Asperger Syndrome and other subtypes of ASD; however, the potential utility of fixation patterns as diagnostic differentiator merits further research.
It is also possible that cognitive or experiential characteristics of the children in the current study contributed to their performance. The sample in the current study was extremely high-functioning, with cognitive abilities a full standard deviation above average. Prior research suggests that face processing anomalies in ASD may be more likely in individuals with lower intelligence (Teunisse & de Gelder, 2003
). Though, in the current study, recognition ability did not correlate with IQ, the influence of cognitive ability cannot be ruled out. Intervention history may have also impacted performance. Given that eye contact is a common objective in social therapies, it is possible that (a) genuine improvement in eye gaze secondary to intervention was reflected in more normative gaze patterns or (b) acquiescent response bias skewed participants towards more typical viewing patterns under circumstances in which they knew their fixations were being monitored. Parent report and clinical observation suggest that the latter may be more likely, as all participants were impaired in eye gaze in naturalistic contexts.
Despite typical looking patterns and comparable performance to typical counterparts on a measure of pattern recognition, individuals with ASD demonstrated selective impairment on measures of face recognition. Their poor performance on the computerized test of facial recognition is most striking, as this measure was identical in procedure and administration to the pattern recognition subtest, on which they performed comparably to peers. These results suggest a potential disconnect between patterns of visual attention and face recognition skill; face recognition difficulties were evident even among individuals with normative patterns of visual attention. Shortened exposure times in the face computer task and on the face memory test may have also contributed to poorer performance. Performance on the CMS faces task was associated with stronger adaptive social function, as measured by the Vineland Social domain. This pattern of results is consistent with the prediction that individuals with stronger face recognition skills should show stronger social ability.
Despite difficulties on measures of identity recognition, children with ASD were comparable to typically developing children on measures of gender and emotion recognition, putatively the more challenging social-cognitive task. As described above, this finding may reflect “teaching to task” in the context of social skills training. This interpretation is also consistent with the observed marginal correlation in the ASD group between gender discrimination and increased attention to eyes. Because gender identification may be less focal in intervention programs for individuals with ASD, it may provide an “uncontaminated” measure of an individual’s ability to extract information from a human face; individuals may have thus been less likely to have had the opportunity or the inclination to develop and practice compensatory strategies for gender identification. When the likelihood of compensatory strategies was eliminated by a novel task, individuals with more typical patterns of attention (i.e., increased attention to eyes) were more proficient at interpreting the information contained in the human face.
One limitation of the current study is that measures of fixation were limited to theoretically-defined regions of interest. Though measures of looking time indicated that these regions captured a significant portion of visual attention in both groups, it is possible that between-group differences in visual attention might have been revealed with exploration of additional regions of interest (e.g., inner versus outer features or stimulus area versus non-stimulus on-screen area). An objective of ongoing work is to use empirically-derived regions of interest to compare viewers with ASD and typical counterparts (Jones & Klin, submitted