Subliminal Perception of Emotional Faces in Autism

Background: Neuroimaging studies of face processing in autism have reported hypoactivation within the lateral fusiform gyrus (FFG), ventrolateral prefrontal cortex (vlPFC), and the amygdala (Schultz et al., 2000; Dalton et al., 2005; Pinkham, et al. 2008). Hadjikhani et al. (2006) have suggested that hypoactivation in these regions are driven by a lack of attention to faces as well as abnormal scanpaths. These findings suggest that the face processing system is present in individuals with autism, but is not automatically engaged during free viewing. The question thus arises: What is the neural basis of abnormal visual scanpaths to faces in autism? Work in typically developing individuals has reported that the amygdala responds to fearful faces even in the absence of conscious perception (i.e., when emotional faces are backwards masked), as well as the lateral FFG, amygdala, and vlPFC (Whalen et al., 1998; Morris, Pelphrey, McCarthy, 2007).

Objectives: We sought to evaluate the hypothesis that responses to subliminally presented, emotionally expressive faces in the amygdala, FFG, and vlPFC would be abnormally low in individuals with autism. Prior evidence indicates that brain activation observed in response to masked faces is more likely to reflect the initial activity of a face detector, similar to the process reflected by the N200 response recorded directly from the cortical surface in the FFG as compared to more traditional FFG activations measured via fMRI during longer, unmasked stimulus presentations that are strongly influenced by top–down processes including attention, familiarity, and visual scanpaths (Allison et al., 1994; Wojciulik, Kanwisher, & Driver, 1998; Henson, Shallice, & Dolan, R, 2000).

Methods: To date, 7 rigorously characterized adult participants with high-functioning autism and 6 age-, IQ-, and gender-matched neurotypical controls participated in an fMRI study. In this rapid event-related design, images of standardized emotional faces (40 trials each of angry, disgust, fear, happy, and neutral) were presented for 15 ms and immediately masked by black and white mosaic images. Stimulus and mask presentations were followed by a jittered intertrial interval lasting from 2-5 s. Participants were asked to fixate a crosshair placed in the middle of the screen at all times. To ensure they maintained attention, participants were asked to press a button when a green mosaic infrequently appeared.

Results: A random-effects comparison of activity in response to the fearful faces revealed greater activity for the neurotypical group compared to the autism group in the FFG, amygdala, and vlPFC (t=3.42, p<0.001133, q<0.05). The same comparison for the response to facial expressions of disgust revealed neurotypical > autism activity in the FFG and insula (t=3.27, p<0.001777, q<0.05).

Conclusions: Consistent with prior work, we observed hypoactivation of key components of the face processing system including the amygdala, FFG, vlPFC, and insula. This study extends prior work by more precisely characterizing the level of dysfunction in this brain system in demonstrating dysfunction in a signal that has been linked to face detection. We conclude that disruption in face detection (seeing a face as a “face”) might underlie atypical scanpaths in autism.