Lindquist et al. address the long-standing debate concerning the representation of emotion at a neural level. They have used sophisticated meta-analytic techniques to determine the extent to which the brain honours locationist versus psychological constructionist accounts of emotion. Their analysis includes studies of emotion perception and experience that rely on a range of different stimuli, modalities, and cognitive processes, focusing on those studies that target discrete, or basic, emotions: anger, disgust, fear, sadness, and happiness. They conclude that the data show promise with respect to psychological construction accounts of emotion while providing little evidence for any consistent or specific correspondence between discrete emotion categories and distinct brain regions or networks.
The authors have contributed the largest meta-analysis of neuroimaging studies of emotion to date, and a psychological construction framework offers a valuable perspective on the affective neuroscience literature. In particular, it offers the potential for fruitful crossover with existing theoretical models in the cognition and emotion literature that emphasise the presence of multiple functional-level routes to the generation of emotion, such as schematic, propositional, analogical, and associative representational systems (SPAARS) (Power & Dalgleish Reference Power and Dalgleish2008) and interacting cognitive subsystems (ICS) (Teasdale & Barnard Reference Teasdale and Barnard1993).
Nonetheless, we feel that despite the wealth of neuroimaging studies dedicated to delineating the brain basis of emotion, it is premature to conclude that particular brain areas do not have a preferential role in coding particular emotional attributes. Before accepting this position, alternative explanations for the absence of locationist evidence must receive careful consideration. Lindquist et al. raise several possibilities themselves – that functional neuroimaging has spatial and temporal limitations, that discrete emotion categories may be represented in networks (rather than specific brain regions) not identified by the current analyses, or that the methods used to target emotion in individual studies may not capture the full complexity or phenomenology of emotional processes and experience in real life. It should also be considered that rather than reflecting the absolute response within a network of regions, different emotional processes may depend on interactions (or changes in connectivity) between these regions.
To this, we would like to add that while neuroimaging can establish associations between activation of certain brain regions or networks and particular emotional or cognitive functions, it cannot establish the necessity of those brain regions for particular functions. Demonstration of the latter requires, for example, targeted neuropsychological studies of patients with sustained brain injury to focal neural regions. This perspective has significant implications for the interpretation of the neuroimaging data. Even where neuroimaging shows little specificity of function for a given brain region or network, it could still be the case that lesion studies support a locationist view by demonstrating that a given region is particularly important for one emotion category. For this reason, neuroimaging evidence on its own cannot resolve this issue and must be considered in conjunction with corresponding human and animal lesion studies.
Finally, the vast majority of neuroimaging studies have studied neural responses to emotional stimuli in healthy individuals who are not preselected according to any specific criteria. However, there is tremendous variability in the extent to which healthy individuals vary in their experience of particular emotions or action tendencies, such as anxiety or behavioural activation. A wealth of cognitive and behavioural research has demonstrated that individual differences in relevant mood and personality dimensions influence the manner in which individuals process emotional stimuli. For example, behavioural studies in clinical and non-clinical populations show that individual differences in anxiety determine the behavioural response to emotional stimuli conveying threat (Bar-Haim et al. Reference Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg and van IJzendoorn2007).
Individual variation has received little consideration in the neuroimaging literature. However, it has been suggested that accounting for individual differences in relevant mood and personality dimensions may reveal aspects of neural function not readily apparent in studies that adopt standard group-based subtraction contrasts (Calder et al. Reference Calder, Ewbank and Passamonti2011). To illustrate, functional neuroimaging studies have shown that anxiety levels in non-clinical populations correlate positively with the amygdala response to fearful faces, particularly when these stimuli are unattended or presented outside conscious awareness (Bishop et al. Reference Bishop, Duncan and Lawrence2004). Moreover, although approximately 40% of studies in previous neuroimaging meta-analyses fail to find an amygdala response to fearful facial expressions (Murphy et al. Reference Murphy, Nimmo-Smith and Lawrence2003; Phan et al. Reference Phan, Wager, Taylor and Liberzon2002), a recent review showed that all studies investigating anxiety's influence on the brain response to fearful faces identified the amygdaloid area, regardless of whether the study showed an amygdala response (fear minus neutral contrast) across the entire group (Calder et al. Reference Calder, Ewbank and Passamonti2011). Anxiety also influences the amygdala response to viewing facial signals of anger (Ewbank et al. Reference Ewbank, Fox and Calder2010). Crucially, this effect is found only when angry faces are gazing towards, but not away from, the observer, suggesting that amygdala activity is not related to a specific facial expression but to the degree of perceived threat. A relationship has also been found between the ventral anterior insula response to pictures of disgusting foods and trait variation in disgust sensitivity (Calder et al. Reference Calder, Beaver, Davis, van Ditzhuijzen, Keane and Lawrence2007).
These results do not necessitate that the amygdala or insula or, indeed, other regions, are associated with threat and disgust specifically. Rather, these regions may be associated with more abstract emotional attributes on which these emotions weigh heavily but not exclusively (Calder et al. Reference Calder, Beaver, Davis, van Ditzhuijzen, Keane and Lawrence2007). To this extent, we do not disagree with Lindquist et al.'s conclusion that the brain may not respect discrete emotion categories per se. Nonetheless, we think it would be premature on the basis of neuroimaging data alone to say that the amygdala, for example, does not play a preferential role in coding a particular emotional attribute.
Studies such as those highlighted here suggest that the apparent inconsistencies observed in the literature may be attributable to variation in psychological dimensions that are meaningful and relevant to the particular emotion under consideration. The role played by these individual difference factors must therefore receive careful scrutiny and consideration before we can conclude that emotion categories are not natural kinds that are respected by the brain.
Lindquist et al. address the long-standing debate concerning the representation of emotion at a neural level. They have used sophisticated meta-analytic techniques to determine the extent to which the brain honours locationist versus psychological constructionist accounts of emotion. Their analysis includes studies of emotion perception and experience that rely on a range of different stimuli, modalities, and cognitive processes, focusing on those studies that target discrete, or basic, emotions: anger, disgust, fear, sadness, and happiness. They conclude that the data show promise with respect to psychological construction accounts of emotion while providing little evidence for any consistent or specific correspondence between discrete emotion categories and distinct brain regions or networks.
The authors have contributed the largest meta-analysis of neuroimaging studies of emotion to date, and a psychological construction framework offers a valuable perspective on the affective neuroscience literature. In particular, it offers the potential for fruitful crossover with existing theoretical models in the cognition and emotion literature that emphasise the presence of multiple functional-level routes to the generation of emotion, such as schematic, propositional, analogical, and associative representational systems (SPAARS) (Power & Dalgleish Reference Power and Dalgleish2008) and interacting cognitive subsystems (ICS) (Teasdale & Barnard Reference Teasdale and Barnard1993).
Nonetheless, we feel that despite the wealth of neuroimaging studies dedicated to delineating the brain basis of emotion, it is premature to conclude that particular brain areas do not have a preferential role in coding particular emotional attributes. Before accepting this position, alternative explanations for the absence of locationist evidence must receive careful consideration. Lindquist et al. raise several possibilities themselves – that functional neuroimaging has spatial and temporal limitations, that discrete emotion categories may be represented in networks (rather than specific brain regions) not identified by the current analyses, or that the methods used to target emotion in individual studies may not capture the full complexity or phenomenology of emotional processes and experience in real life. It should also be considered that rather than reflecting the absolute response within a network of regions, different emotional processes may depend on interactions (or changes in connectivity) between these regions.
To this, we would like to add that while neuroimaging can establish associations between activation of certain brain regions or networks and particular emotional or cognitive functions, it cannot establish the necessity of those brain regions for particular functions. Demonstration of the latter requires, for example, targeted neuropsychological studies of patients with sustained brain injury to focal neural regions. This perspective has significant implications for the interpretation of the neuroimaging data. Even where neuroimaging shows little specificity of function for a given brain region or network, it could still be the case that lesion studies support a locationist view by demonstrating that a given region is particularly important for one emotion category. For this reason, neuroimaging evidence on its own cannot resolve this issue and must be considered in conjunction with corresponding human and animal lesion studies.
Finally, the vast majority of neuroimaging studies have studied neural responses to emotional stimuli in healthy individuals who are not preselected according to any specific criteria. However, there is tremendous variability in the extent to which healthy individuals vary in their experience of particular emotions or action tendencies, such as anxiety or behavioural activation. A wealth of cognitive and behavioural research has demonstrated that individual differences in relevant mood and personality dimensions influence the manner in which individuals process emotional stimuli. For example, behavioural studies in clinical and non-clinical populations show that individual differences in anxiety determine the behavioural response to emotional stimuli conveying threat (Bar-Haim et al. Reference Bar-Haim, Lamy, Pergamin, Bakermans-Kranenburg and van IJzendoorn2007).
Individual variation has received little consideration in the neuroimaging literature. However, it has been suggested that accounting for individual differences in relevant mood and personality dimensions may reveal aspects of neural function not readily apparent in studies that adopt standard group-based subtraction contrasts (Calder et al. Reference Calder, Ewbank and Passamonti2011). To illustrate, functional neuroimaging studies have shown that anxiety levels in non-clinical populations correlate positively with the amygdala response to fearful faces, particularly when these stimuli are unattended or presented outside conscious awareness (Bishop et al. Reference Bishop, Duncan and Lawrence2004). Moreover, although approximately 40% of studies in previous neuroimaging meta-analyses fail to find an amygdala response to fearful facial expressions (Murphy et al. Reference Murphy, Nimmo-Smith and Lawrence2003; Phan et al. Reference Phan, Wager, Taylor and Liberzon2002), a recent review showed that all studies investigating anxiety's influence on the brain response to fearful faces identified the amygdaloid area, regardless of whether the study showed an amygdala response (fear minus neutral contrast) across the entire group (Calder et al. Reference Calder, Ewbank and Passamonti2011). Anxiety also influences the amygdala response to viewing facial signals of anger (Ewbank et al. Reference Ewbank, Fox and Calder2010). Crucially, this effect is found only when angry faces are gazing towards, but not away from, the observer, suggesting that amygdala activity is not related to a specific facial expression but to the degree of perceived threat. A relationship has also been found between the ventral anterior insula response to pictures of disgusting foods and trait variation in disgust sensitivity (Calder et al. Reference Calder, Beaver, Davis, van Ditzhuijzen, Keane and Lawrence2007).
These results do not necessitate that the amygdala or insula or, indeed, other regions, are associated with threat and disgust specifically. Rather, these regions may be associated with more abstract emotional attributes on which these emotions weigh heavily but not exclusively (Calder et al. Reference Calder, Beaver, Davis, van Ditzhuijzen, Keane and Lawrence2007). To this extent, we do not disagree with Lindquist et al.'s conclusion that the brain may not respect discrete emotion categories per se. Nonetheless, we think it would be premature on the basis of neuroimaging data alone to say that the amygdala, for example, does not play a preferential role in coding a particular emotional attribute.
Studies such as those highlighted here suggest that the apparent inconsistencies observed in the literature may be attributable to variation in psychological dimensions that are meaningful and relevant to the particular emotion under consideration. The role played by these individual difference factors must therefore receive careful scrutiny and consideration before we can conclude that emotion categories are not natural kinds that are respected by the brain.