From affect to control: Functional specialization of the insula in motivation and regulation

The insula plays a key role in a wide range of brain processes, from viscerosensation and pain to motivation, emotion, and cognitive control. While human neuroimaging studies in all these domains report activations in the insula, little systematic attention is paid to anatomical subdivisions that may provide the basis for functional sub-regions. We conducted a meta-analysis of insular tasks across studies in four domains: emotion, pain, attention switching, and working memory. Using a priori subdivision of the insula based on anatomical studies, we provide evidence that different sub-regions are preferentially activated in different tasks. We suggest that the ventral anterior insula is most important for core affect, a term that describes broadly-tuned motivational states (e.g., excitement) with associated subjective feelings. The dorsal anterior insula, by contrast, may be critical for developing and updating motivational states with specific associated actions (i.e., goals). This region is activated by cognitive control tasks, pain, and some tasks that elicit affective processing. The posterior insula, including SII and portions of parietal operculum, is distinctly activated by pain, providing a double dissociation between pain and tasks that elicit emotions. In 1953, scientists found a new way to think about the human mind. The computer metaphor cast the brain as the ultimate machine, the deus ex machina of conscious will reduced to cogs and wheels cranking away behind the curtain, and a new discipline—cognitive science—was born. The computer metaphor continues to be pervasive: information is maintained online in short term memory ‘buffers,’ addressed and tagged, and written to permanent long-term memory storage. Attention works to enhance ‘information processing.’ Reasoning and even some forms of mental imagery proceed through propositional, symbolic representation. In the heyday of the cognitive science revolution, emotion was regarded as a volume knob, an orphaned sideshow in the great circus of the machine. New developments in recent years are beginning to challenge this view of the mind. The cognitive and neural sciences have increasingly merged, and researchers studying the neural bases of thought and behavior have found a surprising degree of overlap between the brain regions engaged in cognitive and affective processes. The anterior cingulate and insular cortices, among other regions, are engaged reliably in animal and human studies of both central topics in cognitive science—working memory, long-term memory, control of attention—as well as tasks designed to isolate emotional processes. Understanding behaviors or feelings at a given instant in time is a task that requires consideration of the whole organism, and this challenge has forced us to reexamine the ways in which we think about cognitive and emotional processes. There are two main views on the relationship between cognition and emotion. One view, which has existed in various incarnations since the time of the ancient Greeks, is that cognition and emotion are embodied in two separable, opposing systems. Feeling emotion turns off cognition, and thinking dampens emotional impact (Drevets & Raichle, 1998; Mayberg et al., 1999; Metcalfe & Mischel, 1999; Mischel, Shoday, & Peake, 1989). Another view seems initially to stand in opposition to the first: Emotion is critical to motivating cognition and behavior, and the roots of attentional control lie in affect. According to this view, emotions arise from cognitive appraisals of situations (R. S. Lazarus, 1991; Richard S. Lazarus, 1991b; Scherer, Schorr, Ed, & Johnstone, 2001; C. A. Smith & Ellsworth, 1985; Craig A. Smith & Lazarus, 2001), which particularly involve evaluations of how objects and events affect the self. The basic affective experience that arises when a self-relevant event occurs has been labeled “core affect” (J. A. Russell & Barrett, 1999). Core affect is the seed of full-blown emotion, and from affective responses arise the physiological and motivated response tendencies that have been shaped over the course of our evolution to promote adaptive cognitions and behaviors. Thus, in this view, emotion and cognition are not opponents in a zero-sum tug of war. Rather, they are synergistic partners in the game of adaptive self-regulation, each shaping the direction of the other. The answer may be that neither of the metaphors of opposition and synergy is adequate. Emotion does not stop cognition; rather, it directs cognition into channels most appropriate for the situation. In threatening situations, attention is focused on the perceived threat, and extraneous thought stops. In safe situations dominated by positive affect, the impulse to explore and build new skills may prime a broad repertoire of thoughts and behaviors (Fredrickson, 2001). We are at the frontier of exploring the physical brain systems that give rise to thoughts and feelings, and many of these questions may be addressed empirically. Understanding the roles of certain key brain structures may provide critical information on how affective information shapes attention, and how thoughts direct, or in some cases stem, the flow of affective signals in the brain. Some of the most important such regions are likely to be those that lie at the physical junctions between neocortical and evolutionarily older subcortical nuclei—the limbic and paralimbic regions, so named because they form a limbus, or border, around the oldest parts of the brain (Maclean, 1955, 1958; Papez, 1995). Cortical limbic and paralimbic areas are generally thought to include the cingulate cortex, parahippocampal gyrus and entorhinal cortex, orbitofrontal cortex, and the insula. The insula: A key link between cognition and affect In this paper, we focus on the insula as a potential nexus for motivated cognition and emotional behavior. The insula has long been considered part of the emotional and viscerosensory brain (Janig & Habler, 2002; Maclean, 1955), with multiple roles in regulating physiological and psychological homeostasis (Flynn, Benson, & Ardila, 1999). The insula and surrounding operculum contain primary cortical representation of smell and taste (Francis et al., 1999; Rolls, 1996), viscerosensation (Craig, 2002), and pain perception (Coghill, Sang, Maisog, & Iadarola, 1999; Davis, Kwan, Crawley, & Mikulis, 1998). For this reason, it has been termed "limbic sensory cortex" and associated with the subjective feeling of emotional states, or the "feeling self" (Craig, 2002, 2003). Recent evidence from neuroimaging studies corroborate this view. The insula is commonly activated in emotion tasks, predominantly those associated with negative or withdrawal-related emotions (Phan, Wager, Taylor, & Liberzon, 2002). Several individual examples illustrate that the insula plays a broad role in the development of subjective, self-relevant feelings. In a recent study, Sanfey and colleagues (Sanfey, Rilling, Aronson, Nystrom, & Cohen, 2003) found insular activity when participants received an unfair monetary offer from what they believed was another human—but not when the offer came from a computer. Why does the same loss sting more if it results from another's intention? It could be because the loss also signals social rejection, "unfairness," and/or the motivating possibility of reprisal or control of the situation. Another study found that pain-responsive portions of the anterior insula showed decreased activity when a placebo—a medication with no real effect, but believed by participants to be a potent analgesic—was administered prior to pain (Wager, Rilling et al., 2004). A third study showed common activations in the insula when feeling pain and when observing a loved one experience pain (Singer et al., 2004). However, the story suggested by these studies, that of the insula as a structure for feeling, is incomplete. It omits a host of studies that suggest alternative roles for the insula—in particular, a role in cognitive control. If the anterior insula is the seat of emotional awareness, why is the anterior insula (as we show here) activated in so many cognitive tasks ostensibly devoid of emotion (Wager & Smith, 2003)? One possibility is that cognitive tasks simply activate a separate portion of the insula and frontal operculum devoted to attentional control. Without detailed comparison of results from cognitive and affective tasks, we cannot tell. Another possibility is that these cognitive tasks, which share as a common feature the requirement for executive control of attention, share common psychological processes with affective tasks. These could include a role for motivated decision making in goal formation, updating of task relevance based on affective information, and affective error-detection processes in cognitive tasks. A second unresolved issue is the relationship between pain and affect, and how that relationship is informed by insular participation in both. Does pain involve affective representations, as is commonly believed (Melzack & Casey, 1968; Rainville, Duncan, Price, Carrier, & Bushnell, 1997), and do the insular regions involved in "pain affect" form the basis for the "feeling self"? Again, we can ask the same questions about whether pain and emotion activate common parts of the insula, or whether common involvement of the "anterior insula" in pain and emotion is in fact an overgeneralization. These questions are approached using meta-analysis of human neuroimaging studies (Fox, Parsons, & Lancaster, 1998) across four domains: emotion, pain, attention shifting, and working memory. We begin with functional subdivisions suggested by cytoarchitectural and functional studies in animals (Figure 1), and ask whether these subregions produce meaningful dissociations between tasks in human neuroimaging studies. We then ask whether activation of particular insular subregions, or the pattern of activation across subregions, provides useful information about which psychological task or process elicited the activations. This is a fundamentally different question that is typically investigated in neuroimaging studies to date. Rather than asking, "What brain areas implement attention shifting," we ask, "Is there a pattern of activations that uniquely identifies attention shifting, and separates it from other processes?" We present evidence for distinct insular sub-regions, and we use an inductive approach to develop hypotheses about the processes they may implement. In this paper, we argue for a distinction between pain and emotional feelings, which activate the dorsal and ventral portions of the anterior insula, respectively. The pattern of activations we observe also suggests that attentional control tasks activate a specific part (dorsal anterior) of the insula, in common with pain. We frame our conclusions in terms of hypotheses to be tested. One hypothesis is that dorsal anterior insula is directly involved in attentional control, and pain activates this region because it recruits mechanisms of executive attention (Eccleston & Crombez, 1999). Alternatively, we present the hypothesis that executive attention recruits anterior insula because this region links general motivational tendencies with specific action plans—i.e., it is involved in goal formation and re-formation. The ventral anterior insula may represent motivational states with very general action tendencies (e.g., affiliate, protect), and the dorsal aspect represents motivational states associated with specific action plans. This distinction is consistent with patterns of results across all the task domains we studied.