For the first time, we present an activation-likelihood meta-analysis of fMRI studies examining the effects of emotion recognition on brain responses in those with SAD. Our most robust finding was that limbic regions were consistently more active in SAD patients than compared with controls in response to emotional stimuli. Specifically, we found significant activation in the bilateral amygdala, left medial temporal lobe, encompassing the entorhinal cortex, left medial aspect of the inferior temporal lobe encompassing perirhinal cortex and parahippocampus, right anterior cingulate, right globus pallidus, and the distal tip of the right postcentral gyrus. It is of note that we found no significant activation of PFC regions, such as the mPFC, which perhaps indicates some competitive interference of arousal brain circuits on higher order functioning, in those with SAD (e.g., for reviews see Etkin, 2010
Activation of the bilateral amygdala in response to emotional stimuli was our strongest result. The amygdalae are subcortical gray matter nuclei involved in processing visual emotional cues, particularly signaling fear, aversion, or general salience (Phan et al., 2002
). Our data supports previous meta-analyses demonstrating the role of amygdala activation in response to explicit (Costafreda et al., 2008
) and subliminal (Brooks et al., 2012
) emotional stimuli, particularly in faces. Moreover, previous meta-analyses and systematic reviews of SAD indicate hyperactivity in the amygdalae in response to emotional stimuli (Etkin and Wager, 2007
; Freitas-Ferrari et al., 2010
). It is of note that the previous meta-analyses of SAD differ from ours in that they did not use the ALE approach, which is a well-validated, systematic, and computational method to objectively to meta-analyse neural activations (Radua et al., 2012
). For example, Etkin and Wager (2007
) fMRI and PET studies reporting on SAD, specific phobia, PTSD and fear conditioning, and found similar neural activation in these groups. Furthermore, Freitas-Ferrari et al. (2010
) conducted a systematic review of fMRI, PET, SPECT, MRS, and structural MRI in studies examining SAD, and reported amygdala and other limbic region activations. The added value of our meta-analysis is that we use ALE to focus only on fMRI studies during emotion recognition in SAD. Thus, our and previous meta-analytic and review data robustly implicate the role of the bilateral amygdala in emotional processing. Specifically, hyperactivation of the amygdalae likely contributes to the mediation of SAD, particularly when viewing emotional information from people's faces.
Animal literature is also concurrent with the observation in this meta-analysis that bilateral hyperactivity in the amygdala occurs when viewing stimuli perceived as emotionally threatening or aversive (de Carvalho et al., 2010
). For example, projections from the amygdala to the brainstem contribute to a “quick and dirty” response in animals and humans to potentially salient stimuli, and projections from the amygdala to the cortex simultaneously contribute to the experience of other cognitive aspects of emotional processing (LeDoux, 2003
). However, data suggests that the function of the amygdala extends beyond fear processing and is involved in processing of general salience and arousal to both positive and negative emotional processing (Siebert, 2003
). Thus, hyper-arousal in limbic brain circuits, regardless of the valence of the emotional stimulus, may underlie the mediation of SAD.
We also observed greater activation in the parahippocampal gyrus (BA34, 35) in those with SAD, consistent with the involvement of this structure in Pavlovian contextual fear conditioning in both animals and humans (Alvarez et al., 2008
). Previous systematic reviews and meta-analyses of functional neuroimaging in SAD also noted greater activation in this region (Etkin and Wager, 2007
; Freitas-Ferrari et al., 2010
). Parahippocampal hyperactivation in SAD has been interpreted as indicative of disruptions to contextual fear conditioning and an inability to assign accurate saliency value to a stimulus. Additionally, increases in gray matter volume in the parahippocampus have been demonstrated in SAD (Talati et al., 2013
), indicative of plasticity that may underlie hyperactivity in this region.
The ACC showed greater activation in SAD compared to healthy controls in this meta-analysis. Animal and human studies demonstrate that the ACC plays a key role in the regulation of cognitive and emotional processing (Whalen et al., 1998
; Bush et al., 2000
), particularly in relation to conflict monitoring/error detection of discrepancies between predicted and actual outcomes (Botvinick and Watanabe, 2007
; Yeung and Nieuwenhuis, 2009
; Kim et al., 2010
). Moreover, the ACC is highly activated during the anticipation of pain (Straube, 2009
). Activation of the ACC may therefore point to compensatory efforts to regulate high anxiety states in individuals with SAD, or the anticipation that a socially-painful experience may subsequently ensue. The findings here are again consistent with those of other systematic reviews and meta-analyses in SAD (Etkin and Wager, 2007
; Freitas-Ferrari et al., 2010
). Finally, alterations in gray matter volumes have recently been associated with anxiety disorders, especially in the ACC region (Radua et al., 2010
), however, it is still unclear how structural differences relate to functional aberrations in this region.
We are the first to demonstrate in a meta-analysis that increased activation of the globus pallidus is associated with SAD. The globus pallidus appears to play an important role in motor control (Salih et al., 2009
). One possibility therefore is that activation of this area reflects disruption of the motor system as a component of the emotional response in SAD. However, the globus pallidus has also been implicated in affect regulation (Murphy et al., 2003
), and in both the processing of information from complex stimuli and in aversive responses to fear and anxiety (Talalaenko et al., 2008
). Thus, activation in this region may reflect increased statistical power of this meta-analysis relative to those previously conducted.
Additionally, hyperactivity in the distal tip of the right postcentral gyrus was observed. This area has previously been implicated in conveying pure somatosensory information of an auditory nature (Job et al., 2011
). Literature on the functional significance of BA 43 in SAD is lacking, but it seems reasonable to suggest that hyperactivity in this region in people with SAD may be associated with a state of hypervigilance, often in concurrence with anxiety states.
Our data is in line with contemporary neural models of anxiety disorders (for reviews, see Etkin, 2010
). Specifically, a limbic-medial prefrontal circuit is implicated in an aberrant information processing circuit, whereby negative socio-emotional information is given greater salience, due to activation of limbic arousal circuits. Furthermore, the medial prefrontal, self-referential system is perhaps overloaded by bottom-up arousal information, biasing negative emotional stimuli, particularly in relation to the self, such that the value of the external stimulus is heightened, and the value of self-related goals, such as autobiographical memory, self-goals, and experience-based competence are reduced. However, there were a number of regions emphasized in the literature as important components of aberrant brain activity in SAD that were not evident in this meta-analysis. Of particular note, insula activation did not differentiate SAD from healthy controls, despite the recognition of the importance of this structure in mediating anxiety (Craig, 2009
; Holzschneider and Mulert, 2011
). This may reflect the emphasis on responding to external fearful stimuli rather than to interoceptive signals in the particular paradigms deployed in the included studies (Paulus and Stein, 2006
There are some limitations to our meta-analysis that must be considered and which may temper the strength of our findings. Firstly, only 7 fMRI studies, with small sample sizes were included. In addition, two of these studies employed a paradigm which did not rely on emotional recognition of facial expressions. Furthermore, due to the small samples, it is difficult to investigate how socio-demographic variables or neurobiological factors (e.g., gene variants) contribute to the brain activation we observe. Nevertheless, this study is the first to conduct an activation-likelihood meta-analysis of fMRI studies examining neural correlates of emotion recognition in SAD. Our findings support a growing body of work on development of a neurocircuitry model of SAD. Future work on larger samples, including perhaps collaborations across different sites, is needed to address more comprehensively some of the likely sources of inter-individual variance (Furmark et al., 2004
; Smoller et al., 2008
In conclusion, our meta-analysis confirms that increased activation of the bilateral amygdala is a prominent feature in those with SAD, particularly in response to socially emotive stimuli. The contribution of other brain regions implicated in neurobiological models of SAD, such as parahippocampus and anterior cingulate, are also highlighted (Morgan et al., 1993
). Thus, hyperactivation of limbic circuitry likely mediates the symptoms of SAD, and may be a target for clinical interventions and future research.
Conflict of interest statement
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.