The Structure of Emotion Evidence From Neuroimaging Studies

One common point of debate in the study of emotion is whether the basic, irreducible elements of emotional life are discrete emotion categories, such as anger, fear, sadness, and so on, or dimensions such as approach and avoidance. Resolving this debate will identify the basic building blocks of emotional life that are the most appropriate targets of scientific inquiry. In this paper, we briefly review meta-analytic work on the neuroimaging of emotion and examine its potential for identifying ‘‘natural kinds’’ of emotion in the brain. We outline criteria for identifying such natural kinds, summarize the evidence to date on category and dimensional approaches, and suggest ways in which neuroimaging studies could more directly address fundamental questions about the nature of emotion. KEYWORDS—emotion; affect; amygdala; neuroimaging ‘‘In nature’s infinite book of secrecy / A little I can read.’’ —Shakespeare, Antony & Cleopatra What are the basic building blocks of emotional life that a science of emotion should focus on? This question is almost as old as psychology itself, and it remains unanswered. The ‘‘basic emotion’’ approach argues that certain categories of emotion, described by such English words as anger, sadness, fear, happiness, and disgust, are biologically basic—inherited, reflex-like modules that cause a distinct and recognizable behavioral and physiological pattern (e.g., Ekman, 1972; Panksepp, 1998). The ‘‘dimensional’’ approach argues that anger, sadness, fear, and so on are categories that characterize more highly elaborated responses constructed from more fundamental, biological properties such as valence (pleasure/displeasure) and arousal (high activation/low activation; Russell & Barrett, 1999), positive and negative activation (e.g., Watson & Tellegen, 1985), or approach and withdrawal (e.g., Lang, Bradley, & Cuthbert, 1990). The pressing question is which typology is given by nature and consists of ‘‘natural kinds,’’ such that it is possible to make inductive discoveries about them? Natural kinds give psychobiological evidence of their existence. Neuroimaging techniques (functional magnetic resonance imaging, or fMRI, and positron emission tomography, or PET) have recently opened the door to searching directly for the circuitry that supports emotional processing in humans. To indicate a natural kind, patterns of neural activation must be consistent (i.e., show increased activation regardless of the induction method used) and specific (e.g., a fear circuit should be architecturally separable from an anger circuit even though the two may share some brain areas in common). To the degree that consistency and specificity criteria are satisfied, an emotion category or affect dimension can be said to have a ‘‘brain marker.’’ In principle, it should be possible to map patterns of activity within a connected set of brain areas, but in practice, most of the imaging research to date has searched only for the most salient or distinctive feature (e.g., brain area) in the circuitry for a given emotion construct. Meta-analytic summaries (statistical summaries of empirical findings across many studies) of the first 10 years or so of research are now available (Murphy, Nimmo-Smith, & Lawrence, 2003; Phan, Wager, Taylor, & Liberzon, 2002; Wager, Phan, Liberzon, & Taylor, 2003) and allow us to begin to search for evidence of natural kinds. The meta-analyses of Murphy et al. and Phan et al. focused on the neural activations for the five emotion categories—happiness, sadness, anger, fear, and disgust. Wager et al., along with Murphy et al., focused on the neural activations for two affective dimensional models (positive/negative affect and approach/withdrawal motivation). None of the meta-analyses assessed the valence/arousal affective model, in part because many neuroimaging studies fail to measure arousal separately. There were many potential methodological issues in conducting these meta-analyses (e.g., the sole reliance on reporting the peak activation within a broader area of activation, the inherent limitations in the signal-source resolution and spatiotemporal resolution of Address correspondence to Lisa Feldman Barrett, Department of Psychology, Boston College, Chestnut Hill, MA 02167; e-mail: [email protected]. CURRENT DIRECTIONS IN PSYCHOLOGICAL SCIENCE Volume 15—Number 2 79 Copyright r 2006 Association for Psychological Science current neuroimaging techniques, and the heterogeneity of the studies included). Nonetheless, these meta-analyses provide a starting point for evaluating whether categories or dimensions capture natural kinds of emotional phenomena in the brain. They also highlight issues that will help make the results of future studies more cohesive and interpretable. EVIDENCE FOR BASIC EMOTIONS IN THE BRAIN The main findings of the emotion-category–brain-location analyses from Murphy et al. (2003) and Phan et al. (2002) are summarized in Table 1, and the locations of some of the key brain regions are shown in Figure 1. Both meta-analyses agreed that right and left amygdalae were preferentially activated with fear, and that rostral (or forward) portions of the anterior cingulate cortex (ACC) were preferentially activated by sadness. These findings are in agreement with lesion and animal studies that have linked fear and depression to the amygdala and to the subcallosal portion of the ACC (i.e., the rostral portion of the ACC that is below the corpus callosum), respectively (e.g., LeDoux, 2000). Both analyses suggested that disgust produced frequent activation in the basal ganglia (particularly the globus TABLE 1 Summary of Emotion Category–Brain-Area Activation Correspondences Found in Two Studies (Phan, Wager, Taylor, & Liberzon, 2002; Murphy, Nimmo-Smith, & Lawrence, 2003) Phan et al. (2002) Murphy et al. (2003) N Brain activations N Brain activations Anger 5 None 8 Lateral orbitofrontal cortex Sadness 14 Subcallosal anterior cingulate cortex (ACC) 14 Rostral supracallosal anterior cingulate and dorsomedial prefrontal cortex Disgust 5 Basal ganglia 7 Insula/operculum and globus pallidus Fear 13 Amygdala 26 Amygdala Happiness 11 Basal ganglia 11 Rostral supracallosal anterior cingulate/ dorsomedial prefrontal cortex Note. Sample sizes from Murphy et al. (2003) were taken from Fig. 3, which reports the number of studies included in follow-up chi-square analyses (Murphy, personal communication, 2004). The subcallosal ACC is considered the ‘‘visceral’’ part of the ACC; it is connected to the medial orbital frontal cortex and is associated with autonomic control. The supracallosal ACC is considered to the ‘‘cognitive’’ aspect of the ACC; it is connected to the dorsomedial prefrontal cortex and dorsolateral prefrontal cortex and is associated with attention and working-memory functions. The globus pallidus is part of the basal ganglia (see Fig. 1). Fig. 1. Key brain regions implicated in emotion-related processing. Shown are the insular cortex (red outline); thalamus (red); amygdala (yellow); basal ganglia, composed of the caudate (blue), putamen (green), and globus pallidus (cyan); medial surfaces of the medial prefrontal cortex (MPFC) and anterior cingulate cortex (ACC) are not shown. 80 Volume 15—Number 2 Neuroimaging of Emotion