User:Neuropsyc/sandbox

Meta-analyses of affective neuroscience:

A meta-analysis is a statistical approach to synthesizing results across multiple studies. Several meta-analyses examining the brain basis of emotion have been conducted. In each meta-analysis, studies were included that investigate healthy, unmedicated adults and that used subtraction analysis to examine the areas of the brain that were more active during emotional processing that during a neutral control condition. The meta-analyses to date predominantly focus on two theoretical approaches, locationist approaches and psychological construction approaches.

Locationist approaches

These approaches to emotion hypothesize that several emotion categories (including happiness, sadness, fear, anger, and disgust) are biologically basic. In this view, emotions are inherited biologically based modules that cannot be broken down into more basic psychological components. Models following a locationist approach to emotion hypothesize that all mental states belonging to a single emotional category can be consistently and specifically localized to either a distinct brain region or a defined networks of brain regions. Each basic emotion category also shares other universal characteristics: distinct facial behavior, physiology, subjective experience and accompanying thoughts and memories.

Psychological constructionist approaches

This approach to emotion hypothesizes that emotions like happiness, sadness, fear, anger and disgust (and many others) are constructed mental states that occur when many different systems in the brain work together. In this view, networks of brain regions underlie psychological operations (e.g., language, attention, etc.) that interact to produce many different kinds of emotion, perception, and cognition. One psychological operation critical for emotion is the network of brain regions that underlie valence (feeling pleasant/unpleasant) and arousal (feeling activated and energized). Emotions emerge when neural systems underlying different psychological operations interact (not just those involved in valence and arousal), producing distributed patterns of activation across the brain. Because emotions emerge from more basic components, there is heterogeneity within each emotion category; for example, a person can experience many different kinds of fear, which feel differently, and which correspond to different neural patterns in the brain. Thus, this view presents a different approach to understanding the neural bases of emotion than locationist approaches.

Phan et al. 2002

In the first neuroimaging meta-analysis of emotion, Phan et al. (2002) analyzed the results of 55 studies published in peer reviewed journal articles between January 1990 and December 2000 to determine if the emotions of fear, sadness, disgust, anger, and happiness were consistently associated with activity in specific brain regions. All studies used fMRI or PET techniques to investigate higher-order mental processing of emotion (studies of low-order sensory or motor processes were excluded). The authors’ analysis approach was to tabulate the number of studies that reported activation in specific brain regions during tasks inducing fear, sadness, disgust, anger, and happiness. For each brain region, statistical chi-squared analysis was conducted to determine if the proportion of studies reporting activation during one emotion was significantly higher than the proportion of studies reporting activation during the other emotions. Two regions showed this statistically significant pattern across studies. In the amygdala, 66% of studies inducing fear reported activity in this region, as compared to ~20% of studies inducing happiness, ~15% of studies inducing sadness (with no reported activations for anger or disgust). In the subcallosal cingulate, 46% of studies inducing sadness reported activity in this region, as compared to ~20% inducing happiness and ~20% inducing anger. This pattern of clear discriminability between emotion categories was in fact rare, with a number of other patterns occurring in limbic regions (including amydala, hippocampus, hypothalamus, and orbitofrontal cortex), paralimbic regions (including subcallosal cingulate, medial prefrontal cortex, anterior cingulate cortex, posterior cingulate cortex, insula, and temporal pole), and uni/heteromodal regions (including lateral prefrontal cortex, primary sensorimotor cortex, temporal cortex, cerebellum, and brainstem). Brain regions implicated across discrete emotion included the basal ganglia (~60% of studies inducing happiness and ~60% of studies inducing disgust reported activity in this region) and medial prefrontal cortex (happiness ~60%, anger ~55%, sadness ~40%, disgust ~40%, and fear ~30%).

Murphy et al. 2003

Murphy, et al. 2003 analyzed 106 peer reviewed journals published between January 1994 and December 2001 to examine the evidence for regional specialization of discrete emotions (fear, disgust, anger, happiness and sadness) across a larger set of studies that Phan et al. Studies included in the meta-analysis measured activity in the whole brain and regions of interest (activity in individual regions of particular interest to the study). 3-D Kolmogorov-Smirnov (KS3) statistics were used to compare rough spatial distributions of 3-D activation patterns to determine if statistically significant activations (consistently activated across studies) were specific to particular brain regions for all emotional categories. This pattern of consistently activated, regionally specific activations was identified in four brain regions: amygdala with fear, insula with disgust, globus pallidus with disgust, and lateral orbitofrontal cortex with anger. The amygdala was consistently activated in ~40% of studies inducing fear, as compared to less than 20% studies inducing happiness, sadness, or anger. The insula was consistently activated in ~ 70% of studies inducing disgust, as compared to sadness (~40%), anger (~20%), fear (~20%), and happiness (~10%). Similar to the insula, the globus pallidus was consistently activated in ~70% of studies inducing disgust, as compared to less than 25% of studies inducing sadness, fear, anger or happiness. The lateral orbitofrontal cortex was consistently activated in over 80% of studies inducing anger, as compared to fear (~30%), sadness (~20%), happiness (< 20%) and disgust (< 20%). Other regions showed different patterns of activation across categories. For example, both the dorsal medial prefrontal cortex and the rostral anterior cingulate cortex showed consistent activity across emotions (happiness ~50%, sadness ~50%, anger ~ 40%, fear ~30%, and disgust ~ 20%).

Barrett et al. 2006

Barrett, et al. 2006 examined 161 studies published between 1990-2001, subsets of which were analyzed in previous meta-analyses (Phan, et al. 2002 and Murphy et al. 2003). In this review, the authors examined the locationist hypothesis by comparing the consistency and specificity of prior meta-analytic findings specific to each hypothesized basic emotion (fear, anger, sadness, disgust, and happiness). Consistent neural patterns were defined by brain regions showing increased activity for a specific emotion (relative to a neutral control condition), regardless of the method of induction used (for example, visual vs. auditory cue). Specific neural patterns were defined as architecturally separate circuits for one emotion vs. the other emotions (for example, the fear circuit must be discriminable from the anger circuit, although both circuits may include common brain regions). In general, the results supported consistency among the findings of Phan et al. and Murphy et al., but not specificity. Consistency was determined through the comparison of chi-squared analyses that revealed whether the proportion of studies reporting activation during one emotion was significantly higher than the proportion of studies reporting activation during the other emotions. Specificity was determined through the comparison of emotion-category brain-localizations by contrasting activations in key regions that were specific to particular emotions. Increased amygdala activation during fear was the most consistently reported across induction methods (but not specific). Both meta-analyses also reported increased activations in regions of the anterior cingulate cortex during sadness, although this finding was less consistent (across induction methods) and was not specific to sadness. Both meta-analyses also found that disgust was associated with increased activity in the basal ganglia, but these findings were neither consistent nor specific. Neither consistent nor specific activity was observed across the meta-analyses for anger or for happiness. This meta-analysis additionally introduced the concept of the basic, irreducible elements of emotional life as dimensions such as approach and avoidance. This dimensional approach involved in psychological constructionist approaches is further examined in later meta-analyses of Kober et al. 2008 and Lindquist et al. 2012.

Kober et al. 2008

Instead of investigating specific emotions, Kober, et al. 2008 reviewed 162 neuroimaging studies published between 1990-2005 to determine if groups of brain regions show consistent patterns of activation during emotional experience (that is, actively experiencing an emotion first-hand) and during emotion perception (that is, perceiving a given emotion as experienced by another). This meta-analysis used multilevel kernal density analysis (MKDA) to examine fMRI and PET studies, a technique that prevents single studies from dominating the results (particularly if they report multiple nearby peaks) and that enables studies with large sample sizes (those involving more participants) to exert more influence upon the results. MKDA was used to establish a neural reference space that includes the set of regions showing consistent increases across all studies (for further discussion of MDKA see Wager et al. 2007). Next, this neural reference space was partitioned into functional groups of brain regions showing similar activation patterns across studies by first using multivariate techniques to determine co-activation patterns and then using data-reduction techniques to define the functional groupings (resulting in six groups). Consistent with a psychological construction approach to emotion, the authors discuss each functional group in terms more basic psychological operations. The first “Core Limbic” group included the left amygdala, hypothalamus, periaqueductal gray/thalamus regions, and amygdala/ventral striatum/ventral globus pallidus/thalamus regions, which the authors discuss as an integrative emotional center that plays a general role in evaluating affective significance. The second “Lateral Paralimbic” group included the ventral anterior insula/frontal operculum/right temporal pole/ posterior orbitofrontal cortex, the anterior insula/ posterior orbitofrontal cortex, the ventral anterior insula/ temporal cortex/ orbitofrontal cortex junction, the midinsula/ dorsal putamen, and the ventral striatum /mid insula/ left hippocampus, which the authors suggest plays a role in motivation, contributing to the general valuation of stimuli and particularly in reward. The third “Medial Prefrontal Cortex” group included the dorsal medial prefrontal cortex, pregenual anterior cingulate cortex, and rostral dorsal anterior cingulate cortex, which the authors discuss as playing a role in both the generation and regulation of emotion. The fourth “Cognitive/ Motor Network” group included right frontal operculum, the right interior frontal gyrus, and the pre-supplementray motor area/ left interior frontal gyrus, regions that are not specific to emotion, but instead appear to play a more general role in information processing and cognitive control. The fifth “Occipital/ Visual Association” group included areas V8 and V4 of the primary visual cortex, the medial temporal lobe, and the lateral occipital cortex, and the sixth “Medial Posterior” group included posterior cingulate cortex and area V1 of the primary visual cortex. The authors suggest that these regions play a joint role in visual processing and attention to emotional stimuli.

Vytal et al. 2010

Vytal, et al. 2010 examined 83 neuroimaging studies published between 1993-2008 to examine whether neuroimaging evidence supports the idea of biologically discrete, basic emotions (i.e. fear, anger, disgust, happiness, and sadness). Consistency analyses identified brain regions that were associated with a given emotion. Discriminability analyses identified brain regions that were significantly, differentially active when contrasting pairs of discrete emotions. This meta-analysis examined PET or fMRI studies that reported whole brain analyses identifying significant activations for at least one of the five emotions relative to a neutral or control condition. The authors used activation likelihood estimation (ALE) to perform spatially sensitive, voxel-wise (sensitive to the spatial properties of voxels) statistical comparisons across studies. This technique allows for direct statistical comparison between activation maps associated with each discrete emotion. Thus, discriminability between the five discrete emotion categories was assessed on a more precise spatial scale than what had been accomplished in prior meta-analyses. Consistency was first assessed by comparing the ALE map generated across studies for each emotion (for example, the ALE map identifying regions consistently activated by studies inducing fear) to ALE map generated by random permutations. Discriminability was then assessed by pair-wise contrasts of individual emotion ALE maps (for example, fear ALE map vs. anger ALE map; fear ALE map vs. disgust map) across all basic emotions pairings. Consistent and discriminable patterns of neural activation were observed for the five emotional categories. Happiness was consistently associated with activity in 9 regional brain clusters, the largest located in the right superior temporal gyrus. For the first time, happiness was discriminated from the other emotional categories, with the largest clusters of activity specific to happiness (vs. the other emotion categories) located in right superior temporal gyrus and left rostral anterior cingulate cortex. Sadness was consistently associated with 35 clusters (the largest activation cluster located in the left medial frontal gyrus) and was discriminated from the other emotion categories by significantly greater activity in left medial frontal gyrus, right middle temporal gyrus, and right inferior frontal gyrus. Anger was consistently associated with activity in 13 clusters (the largest of which was located in the left inferior frontal gyrus), and was discriminated from the other emotion categories by significantly greater activity in bilateral inferior frontal gyrus, and in right parahippocampal gyrus. Fear was consistently associated with 11 clusters (the largest activation cluster in the left amygdala) and was discriminated from the other emotion categories by significantly greater activity in the left amygdala and left putamen. Disgust was consistently activated with 16 clusters (the largest activation cluster in the right insula/ right inferior frontal gyrus) and was discriminated from the other emotion categories by significantly greater activity in the right putamen and the left insula.

Lindquist et al. 2012

Lindquist, et al. 2012 reviewed 91 PET and fMRI studies published between January 1990 and December 2007. The studies included in this meta-analysis used induction methods that elicit emotion experience or emotion perception of fear, sadness, disgust, anger, and happiness. The goal was to compare locationist approaches with psychological constructionist approaches to emotion. Similar to Kober et al. described above, a Multilevel Peak Kernel Density Analysis transformed the individual peak activations reported across study contrasts into a neural reference space (in other words, the set of brain regions consistently active across all study contrasts assessing emotion experience or perception). The density analysis was then used to identify regions (or voxels) within the neural reference space with more consistent activations for a specific emotion category (anger, fear, happiness, sadness, and disgust) than all other emotions. Chi-squared analysis was used to create statistical maps that indicated if each previously identified and consistently active regions (those identified during density analysis) were more frequently activated in studies of each emotion category versus the average of all other emotions, regardless of activations elsewhere in the brain. Chi-squared analysis and density analysis both defined functionally consistent and selective regions, or regions which showed a relatively more consistent increase in activity for the experience or perception of one emotion category across studies in the literature. Thus, a selective region could present increased activations relatively more so to one emotion category while also having a response to multiple other emotional categories. A series of logistic regressions were then preformed to identify if any of the regions that were identified as consistent and selective to an emotion category were additionally specific to a given category. Regions were defined as specific to a given emotion if they showed increased activations for only one emotional category, and never showed increased activity during instances of the other emotional categories. In other words, a region could be defined as consistent, selective and specific for e.g. fear perception if it only showed significantly greater increases in activation during the perception of fear and did not show increased activity during any other emotion categories. However, the same region would be defined as only consistent and selective (and not specific) to fear perception if it additionally displayed increased activations during anger perception. Strong support for the locationist approach was defined as evidence that basic emotion categories (anger, disgust, fear, happiness and sadness) consistently map onto areas of the brain that specifically activate in response to instances of only one emotional category. Strong support for the constructionist approach was defined as evidence that multiple psychological operations (some of which are not specific or selective to emotion) consistently occur across many brain regions and multiple emotional categories.

The results indicated that many brain regions demonstrated consistent and selective activations in the experience or perception of an emotion category (versus all the other emotion categories). Consistent with constructionist models, however, no region demonstrated functional specificity for the emotions of fear, disgust, happiness, sadness or anger. Based on the existing scientific literature, the authors proposed different roles for the brain regions that have traditionally been associated with only one emotion category. The authors propose that the amygdala, anterior insula, orbitofrontal cortex each contribute to “core affect,” which are basic feelings that are pleasant or unpleasant with some level of arousal. The amygdala, for example, appears to play a more general role in indicating if external sensory information is motivationally salient, and is particularly active when a stimulus is novel or evokes uncertainty. The anterior insula may represent core affective feelings in awareness across a number of emotion categories, driven largely by sensations originating in the body. The orbitofrontal cortex appears to function as a site for integrating sensory information from the body and sensory information from the world to guide behavior. Closely related to core affect, the authors propose that anterior cingulate and dorsolateral prefrontal cortex play vital roles in attention, with anterior cingulate supporting the use of sensory information for directing attention and motor responses during response selection and with dorsolateral prefrontal cortex supporting executive attention. In many psychological construction approaches, emotions also involve the act of interpreting one’s situation in the world relative to the internal state of the body, or what is referred to as “conceptualization.” In support of this idea, the dorsomedial prefrontal cortex and hippocampus were consistently active in this meta-analysis, regions that appear to play an important role conceptualizing during emotion, which are also involved in simulating previous experience (e.g. knowledge, memory). Language is also central to conceptualizing, and regions that support language, including ventrolateral prefrontal cortex, were also consistently active across studies of emotion experience and perception.