In order to more thoroughly understand the role of the amygdala and fusiform in autism, we conducted a comprehensive investigation using functional and structural MRI in a group of well characterized children with high functioning autism compared to typically developing peers of the same age.
For the fMRI study, we hypothesized decreased amygdala activation in the autism group. Consistent with this hypothesis, the behavioral results showed that the children with autism accurately matched the emotions, but revealed limited engagement of the amygdala while performing the task. This was in contrast to the typical group which demonstrated recruitment of the amygdala during both the emotion (bilateral amygdala) and person (left amygdala) matching conditions. Ostensibly, the children with autism utilize different brain regions such as frontal and parietal regions as suggested by the pattern of results seen in the whole brain analysis for the appraisal of affective stimuli. A recent fMRI study also showed that ASD children utilized different strategies and recruited atypical neural networks during automatic emotion processing amidst intact emotion identification (Wang et al., 2004
). The limited amygdala activation in the current investigation is consistent with previous fMRI studies in autism during the perception of fear faces (Ashwin et al., 2007
), the implicit processing of emotions (Critchley et al., 2000
) and during emotion judgment (Baron-Cohen et al., 1999
In regards to face processing (person matching condition) a different pattern emerged. The children with autism showed impairment in matching facial stimuli while showing some FG activation albeit less than typical children. This supports the idea that the autism group were processing the face stimuli, but less efficiently and effectively than that seen in the typical group, as shown in some previous reports (Critchley et al., 2000
; Pierce et al., 2001
; Schultz, 2005
; Schultz et al., 2000
). The presence of some fusiform activation suggests that the children with autism are encoding the facial stimuli albeit to a limited degree. Therefore, to some extent, the ASD group is recruiting brain regions important in the interpretation of facial information. However, activation of fusiform alone is not sufficient for the processing of faces since individuals with prosopagnosia, a severe inability to recognize faces, recruit the FG (Avidan et al., 2005
) despite their behavioral deficits. Thus, the atypical activation of the fusiform and amygdala provide support for a dysfunctional neural network or pathway related to deficits in socioemotional processing in autism, especially identifying faces.
It was recently reported that activation in the amygdala and FG were positively associated with the time spent fixating on the eyes, which may explain the observed hypoactivation often reported in these brain structures in autism (Dalton et al., 2005
). Although we did not utilize an eye tracker in the current investigation to ascertain eye fixation or gaze duration, the adequate performance on the emotion matching task suggests that the children with autism were attending to the facial stimuli. The determination that subjects who look into the eyes activate the amygdala is provocative (Dalton et al., 2007
; Dalton et al., 2005
) and the lack of eye tracking data could be seen as a weakness of the current study. It was recently shown that amygdala damage impairs eye-to-eye contact with an increase in gaze-to-mouth looking (Spezio et al., 2007
), a finding previously reported in autism (e.g., Klin et al., 2002
). Thus, it seems likely that the present results, with reductions in amygdala and FG activation during emotion and face processing reflect a true neural correlate of an inherent deficit in ASD children.
The finding of relatively intact emotion matching abilities is consistent with some studies (Castelli, 2005
), but not others (Adolphs, 2001
; Baron-Cohen et al., 1999
; Hobson, 1986
; Hobson et al., 1988
; Howard et al., 2000
), which likely pertains to the complexity and diversity of the tasks across the studies (Castelli, 2005
). The accurate emotion matching, despite altered amygdala and fusiform function, suggests that children with autism can distinguish basic emotions, but they may not be assigning the emotional and social significance. The failure to provide emotional salience may reveal a lack of automatic emotional processing provided by the amygdala (Dolan and Vuilleumier, 2003
). A recent report showed a lack of affective priming for emotional faces in autism (Kamio et al., 2006
). A fundamental role of the amygdala is to rapidly provide the emotional salience of incoming sensory stimuli even preceding conscious awareness (Halgren, 1992
; Vuilleumier et al., 2003
). Although children in our investigation were able to identify emotions at a conscious level they showed reduced amygdala recruitment during the detection of socioemotional facial stimuli (Kamio et al., 2006
; Wang et al., 2004
Alternatively, the diminished activation of the amygdala and fusiform on the part of the children with autism across the fMRI contrasts may indicate limited functional connectivity between these interconnected brain regions. It has been shown that emotional faces increase effective connectivity between the amygdala and FG through an enhanced, dynamic coupling resulting in increased brain activation and attentional resources (Fairhall and Ishai, 2006
). Based on dynamic causal modeling the authors proposed that the FG provides the primary causal input into extended brain systems, which in turn process the subsequent emotional and social characteristics of face stimuli.
These aforementioned interpretations are not mutually exclusive. Based on the pattern of activation in the amygdala and fusiform in the children with autism, we hypothesize a dysfunctional connection between these structures; resulting in a failure to provide the emotional salience and social relevance, respectively. Schultz (2005)
convincingly argued that deficits in the amygdala-fusiform network point to a fundamental causal mechanism in the social perception deficits in autism. Recently, Bachevalier and Loveland (2006)
presented a model compatible with our interpretation that dysfunction of the orbitofrontal-amygdala circuit results in deficits in social and emotional cognition in autism. Additionally, it has been noted that deficits in face processing in autism are part of a dysfunctional distributed network (Hadjikhani et al., 2004
), and thus extend beyond the dysfunction of any single brain region such as the amygdala (e.g., Baron-Cohen et al., 2000
) or the fusiform (e.g., Schultz et al., 2000
The typically developing children showed relative deactivation of the FG on the emotion matching condition. Although unexpected, a similar finding has been reported in another study of individuals with typical development compared to those with ASD, and interpreted to reflect modulations in FG activation based on the varying intensities of the emotion stimuli presented in the task (Ashwin et al., 2007
). Our task contained variable emotion stimuli, which may similarly have dampened the recruitment of the FG and contributed to the deactivation findings in the typical group.
Regarding the volumetric analyses, we did not observe structural differences in the right or left amygdala or total cerebral volume across the groups. Additional within-group analyses revealed that amygdala volume appeared smaller as a function of younger age. We also observed an association between reported social anxiety and the right amygdala in that greater anxiety was related to smaller right amygdala volume. Although the results may seem counterintuitive since larger amygdala volume in children and adolescents with generalized anxiety disorder have been reported (De Bellis et al., 2000
). Recently, smaller amygdala volumes have been shown in children with pediatric anxiety using voxel-based morphometry (Milham et al., 2005
). The association between anxiety symptom severity, the size of the amygdala and age may support a complex interplay between these covariates (Juranek et al., 2006
; Nacewicz et al., 2006
; Schumann et al., 2004
). Still, notable variability in amygdala volumes were observed as found in other investigations of amygdala volumes in autism (Salmond et al., 2003
), highlighting the heterogeneity and complexity of the disorder.
Despite our careful recruitment of a homogeneous sample within a narrow age range and strictly defined autism, significant variability of biological profiles remains and challenges the study of autism. Nevertheless, it has been shown in our studies and others that variability in and of itself remains one of the more consistent findings in autism (e.g., Corbett et al., 2006
; Corbett et al., 2008
)). As such, examination of individual differences or subtypes may ultimately be a more meaningful exploration of brain and behavior relationships.
A potential limitation of the study is the significant discrepancy in IQ between the groups. We did not attempt to match on current IQ, since equating groups on inherent factors related to the diagnosis of autism, may introduce systematic bias referred to as matching fallacy (Meehl, 1970
). Instead, we selected subjects based on an IQ ≥ 80. Even though the mean IQ for the autism group was solidly in the average range, the mean of the control group was very high. It is possible that the IQ difference may contribute to differences in other cognitive processes beyond face and emotion perception. Another limitation of the fMRI portion of the study is the short duration of the scan which resulted in fewer time points and reduced power. Although this allowed the majority of our subjects to complete the scan, there is the possibility that some of our findings are the result of chance.
In summary, to the best of our knowledge we conducted one of the first investigations employing both functional and structural MRI in 8 to 12 year old children with and without autism evaluating socioemotional processing. Based on an explicit emotion and face matching fMRI paradigm, children with autism showed a lack of amygdala activation and reduced activation of the FG. We hypothesize that the amygdala fails to provide the socioemotional relevance and context to other brain regions, such as the FG, during basic and complex social situations that require higher-level cognitive interpretation. Further, whether attributed to an innate lack of social preference or entrainment to look in the eyes, we speculate that the limited engagement of the amygdala results in a dysfunctional connection or poor effective coupling between the amygdala and the FG in many children with autism. Despite the observed disruption in the amygdala and FG, it is likely that the neuropathology extends well beyond these brain regions. Thus, it will be important for future studies to adopt a systems-based approach to better understand the complexity in regards to both the neural mechanisms and symptoms of autism.