The relationship between cognition and emotion has fascinated important thinkers within the Western intellectual tradition. Historically, emotion and cognition have been viewed as largely separate. In the past two decades, however, a growing body of work has pointed to the interdependence between the two.
Cognition refers to processes such as memory, attention, language, problem solving, and planning. Many cognitive processes are thought to involve sophisticated functions that may be unique to primates. They often involve so-called controlled processes, such as when the pursuit of a goal (e.g., maintaining information in mind) needs to be protected from interference (e.g., a distracting stimulus). A prototypical example of a neural correlate of a cognitive process is the sustained firing of cells in dorsolateral prefrontal cortex as a monkey maintains information in mind for brief periods of time (Fuster and Alexander, 1971; Kubota and Niki, 1971). With the advent of functional MRI (fMRI), it appears that cognitive processes engage cortical regions of the brain (Gazzaniga et al., 2008).
Whereas there is relative agreement about what constitutes cognition, the same cannot be said about emotion. Some investigators use definitions that incorporate the concepts of drive and motivation: emotions are states elicited by rewards and punishers (Rolls, 2005). Others favor the view that emotions are involved in the conscious (or unconscious) evaluation of events (Arnold, 1960) (i.e., appraisals). Some approaches focus on basic emotions (Ekman, 1992) (e.g., fear, anger), others on an extended set of emotions, including moral ones (Haidt, 2003; Moll et al., 2005) (e.g., pride, envy). Strong evidence also links emotions to the body (Damasio, 1994). Brain structures linked to emotion are often subcortical, such as the amygdala, ventral striatum, and hypothalamus. These structures are often considered evolutionarily conserved, or primitive. They are also believed to operate fast and in an automatic fashion, such that certain trigger features (e.g., the white of the eyes in a fearful expression (Whalen et al., 2004)) are relatively unfiltered and always evoke responses that may be important for survival. Accordingly, an individual may not be necessarily conscious of a stimulus that may have triggered brain responses in an affective brain region, such as the amygdala. For discussion, see (Ohman, 2002; Pessoa, 2005).
Because of the inherent difficulty in providing clear definitions for both cognition and emotion, they will not be further defined here. We now turn to illustrating some of the interactions between emotion and cognition. Given the enormous scope of this topic, by necessity, this review will be relatively narrow in scope and will emphasize the brain systems involved in the interactions between emotion and i) perception and attention; ii) learning and memory; and iii) behavioral inhibition and working memory. Other valuable sources include (Damasio, 1994; LeDoux, 1996; Damasio, 1999; Dolan, 2003; Rolls, 2005; Phelps, 2006). A key conclusion from this review and from other current discussions of the relationship between cognition and emotion is that it is probably counterproductive to try to separate them. Instead, current thinking emphasizes their interdependence in ways that challenge a simple division of labor into separate cognitive and emotional domains. In particular, in the context of the brain, the general dichotomization alluded to above in terms of cortical-cognitive and subcortical-emotional brain areas is now viewed as largely simplified and breaks down rather quickly when more in-depth analyses are carried out; e.g., (Pessoa, 2008).
Before proceeding, however, a brief historical note is in order. The emotion/cognition debate came into sharp focus with the report of the mere-exposure effect (Kunst-Wilson & Zajonc, 1980), which led to a strong belief that affect was primary to and independent of cognition. It can be said that the mere-exposure effect and other behavioral findings shifted ongoing debates to focus on affect as being related to unconscious processing and subcortical activity, with cognition being related to conscious processing and cortical involvement. Interestingly, behavioral findings were interpreted in the context of the “low route” suggested by LeDoux (1996), which was purported to carry affective information subcortically. These early behavioral studies provided a strong impetus to the wave of neuroscience research in the late 1990s (and beyond) that investigated related phenomena. For some of the early theoretical arguments, see Fazio et al. (1986), Leventhal & Scherer (1987), Bornstein (1989), Lazarus (1994), Zajonc (1994), and Bargh (1997); also see Storbeck, Robinson, & McCourt (2006) and Storberk (2008).
Viewing emotion-laden visual stimuli is linked to heightened and more extensive visual system activation (Pessoa et al., 2002a; Vuilleumier, 2005). For instance, viewing faces with emotional expressions evokes increased responses relative to viewing neutral faces throughout ventral occipitotemporal visual cortex ( Figure 1).
Visual responses are also stronger when subjects view emotional scenes (e.g., a war scene) compared do neutral scenes (e.g., a lake scene). Increased visual activation is observed in both late visual areas, such as the fusiform gyrus and superior temporal sulcus, and early visual cortex in occipital cortex. Recent studies suggest that, in humans, even retinotopically organized visual cortex, including visual areas V1 and V2 along the calcarine fissure, are modulated by the affective significance of a stimulus (Padmala and Pessoa, 2008).
Enhanced visual activation when viewing emotional stimuli is consistent with the observed improvements in behavioral performance in several visual tasks. For instance, angry and happy faces are detected faster in visual search tasks (Eastwood et al., 2001), and possibly other emotional stimuli, too, such as a snake or spider (Ohman et al., 2001) compared to neutral stimuli. Stronger evidence comes from studies of the attentional blink paradigm, in which subjects are asked to report the occurrence of two targets (T1 and T2) among a rapid stream of visual stimuli. When T2 follows T1 by a brief delay, participants are more likely to miss it, as if they had blinked (hence the name). The attentional blink is believed to reflect a capacity-limited processing stage, possibly linked to a process of consolidation of the detected item for conscious reports. Interestingly, the attentional blink has been shown to be modulated by emotional stimuli, as subjects are significantly better at detecting T2 when it is an emotion-laden word (e.g., rape) than when it is a neutral word (Anderson, 2005). Converging evidence for a link between perception, attention, and emotion comes from additional studies. For example, patients with unilateral inattention due to spatial hemineglect (often as a result of right hemisphere parietal lesions) are better at detecting happy or angry faces compared to neutral ones (Vuilleumier and Schwartz, 2001). These findings are consistent with the notion that emotional faces may direct the allocation of attention. For instance, in one study, emotional faces were flashed at spatial locations that subsequently displayed low-contrast visual stimuli (Phelps et al., 2006). They found that detection of the target was strongest when the fear face served as the spatial cue, suggesting that emotional stimuli can provide additional attentional guidance above and beyond a generic spatial cue. How is the increase in perceptual processing and attentional capture that is observed during the perception of affective stimuli mediated in the brain? Growing evidence links the amygdala, a subcortical region, with these effects.
For instance, patients with amygdala lesions do not exhibit improved detection of T2 emotional targets during the attentional blink (i.e., a decrease in the magnitude of the blink effect) (Anderson and Phelps, 2001), and show less evidence of increased responses in visual cortex during the viewing of fearful faces (Vuilleumier et al., 2004); see Figure 2.Thus, it appears that the amygdala may underlie a form of emotional modulation of information that in many ways parallels attentional effects that are observed with non-emotional information (Pessoa et al., 2002a; Vuilleumier, 2005). There are several ways in which emotional modulation may be accomplished. First, it is possible that direct projections from the amygdala to visual processing regions enhance visual processing. The amygdala sends projections across all levels of the visual system, including anterior regions in temporal cortex and posterior regions in occipital cortex (including V1 and V2) (Amaral et al., 1992). Thus, the amygdala is well situated to modulate sensory processing according to the affective significance of a visual object. A second possibility is that the amygdala interacts with other brain regions that are important for the control of attention, such as frontal and parietal regions (Barbas, 1995), which, by their turn, modulate visual processing. In the latter scenario, the amygdala (possibly indirectly) would recruit attentional circuits so as to enhance the sensory processing of emotion-laden stimuli.
A final issue that should be addressed when considering interactions between emotion and perception/attention is whether the perception of emotion-laden stimuli is automatic, namely independent of attention and awareness. This question has received considerable attention because specific answers to this question (no or yes) suggest potentially different relationships between emotion and cognition (more or less independence between the two, respectively). Interestingly, evidence both for and against automaticity has been presented. For instance, emotional faces evoke responses in the amygdala even when attention is diverted to other stimuli (Vuilleumier et al., 2001; Anderson et al., 2003). Perhaps even more strikingly, amygdala responses are sometimes reported for emotional faces of which subjects are not conscious (Morris et al., 1998; Whalen et al., 1998; Etkin et al., 2004; Whalen et al., 2004). Furthermore, cases of affective blindsight have been reported. These and other related findings suggest that at least some types of emotional perception occur outside of cognitive processing – and may rely on direct subcortical pathways conveying visual information to the amygdala (LeDoux, 1996). At the same time, recent findings have suggested that the perception of emotion-laden items requires attention, as revealed by attentional manipulations that consume most processing resources, leaving relatively few resources for the processing of unattended emotional items (Pessoa et al., 2002b; Bishop et al., 2004; Pessoa et al., 2005; Bishop et al., 2007; Hsu and Pessoa, 2007; Lim et al., 2008). Furthermore, it also appears that amygdala responses evoked by unaware stimuli depend somewhat on the manner by which awareness is operationally defined (Merikle et al., 2001), such that no unaware responses are observed when awareness is defined, for instance, via signal detection theory methods (Pessoa et al., 2006). Overall, the automaticity debate remains unresolved and controversial (Pessoa, 2005; Wiens, 2006; Bishop, 2007).
Emotional content can change the formation and recollection of a memory event, consistent with findings in both human and animal studies. Compared to neutral items, humans remember better emotionally arousing information, including emotionally charged stories, film clips, pictures, and words. For instance, in one study participants viewed two videos, one composed of neutral film clips and another composed of emotional film clips (Cahill et al., 1996). Although the two types of clips were taken from the same source and were equated in terms of levels of understandability, subjects were better at remembering emotional relative to neutral clips when tested approximately 3 weeks following the initial viewing of the films. In another study (Bradley et al., 1992), subjects viewed a large array of emotional and neutral pictures from the International Affective Picture System, a stimulus set that has been normed in terms of the dimensions of valence (positive/negative) and arousal (calm/excited). Participants initially rated the pictures along the dimensions of valence and arousal. An incidental free-recall test was administered both immediately and at one year following the rating sessions. Pictures rated as highly arousing were remembered better than all other pictures, including those rated as moderately arousing. Interestingly, the pattern of results was very similar when the subjects were tested a year later, namely, highly arousing pictures were better remembered.
In humans, the amygdala is known to be a critical structure for the enhancement of memory by emotion, consistent with both lesion (Adolphs et al., 1997) and neuroimaging work (for a review, see Phelps, 2004). Recent studies have begun to delineate some of the specific functions of this structure. For instance, it appears that the right amygdala is more strongly involved in emotional memory formation, whereas the left amygdala is engaged by the retrieval of those memories (Sergerie et al., 2006), suggesting a potential hemispheric dissociation of amygdala involvement at different stages of emotional memory. In addition, amygdala responses are also linked to a novelty effect on memory tasks – i.e., the tendency to classify items as new as opposed to old (Sergerie et al., 2007).
In humans, there is some support for the notion that the enhancement of memory due to emotion is due mainly to the arousal dimension of emotional items and not valence (positive/negative) per se (Phelps, 2006), a notion that is more firmly established in nonhuman animal studies (McGaugh, 2004). In these studies, the effects of emotion on memory have been revealed by a vast array of experimental manipulations, including inhibitory avoidance training, contextual fear conditioning, cued fear conditioning, water-maze spatial and cued training, among others. Typically, the effects of emotion on memory are investigated via drug administration, including agonists and antagonists of specific brain receptors. For instance, in one experiment, rats were trained to swim to an escape platform after being placed in a water tank (Packard et al., 1994). To mimic the effects of arousal, a group of animals received an injection of d-amphetamine immediately after training; a control group received a saline injection. Behavioral testing revealed that d-amphetamine administration in the amygdala enhanced memory both on a spatial task and on a non-spatial cued task. A growing body of animal studies strongly supports a model in which emotion influences memory by modulating memory storage (McGaugh, 2004). In particular, the amygdala and the closely associated basal forebrain system involving the stria terminalis appear to play a major role in this modulatory process. These structures are thought to play a central role on memory consolidation by modulating activation in a network of brain regions, including the hippocampus, which is centrally involved in memory formation, but also additional brain structures, such as the nucleus accumbens, caudate nucleus, entorhinal cortex, in addition to other cortical regions (McGaugh, 2002) ( Figure 5).An important dimension of cognition involves behavioral inhibition. Response inhibition, namely the processes required to cancel an intended action, is believed to involve control regions in prefrontal cortex (e.g., dorsolateral prefrontal cortex, anterior cingulate cortex, and inferior frontal cortex) (Rubia et al., 2003; Aron et al., 2004). Response inhibition is often investigated by using so-called go/no-go tasks in which subjects are asked to execute a motor response when shown the go stimulus (e.g., press a key as fast as possible when you see a letter stimulus), but to withhold the response when shown the no-go stimulus (e.g., do not respond when you see the letter Y). Typically, the go and no-go stimuli are shown as part of a rapid stream of stimuli (e.g., a sequence of letters). A recent study investigated the interaction between the processing of emotional words and response inhibition (Goldstein et al., 2007). Response inhibition following negative words (e.g., worthless) engaged the dorsolateral prefrontal cortex. Interestingly, this region was not recruited by negative valence or inhibitory task demands per se; instead, the dorsolateral cortex was sensitive to the explicit interaction between behavioral inhibition and the processing of negatively valenced words.
Working memory, another important cognitive operation, involves the maintenance and updating of information in mind when the information is no longer available to sensory systems (e.g., when keeping a phone number in mind for a few seconds before dialing the number). Evidence for cognitive-emotional interaction comes from working memory studies, too. For instance, when participants were asked to keep in mind neutral or emotional pictures, maintenance-related activity in dorsolateral prefrontal cortex was modulated by the valence of the picture, with pleasant pictures enhancing activity and unpleasant pictures decreasing activity relative to neutral ones (Perlstein et al., 2002). Interestingly, emotional pictures did not affect dorsolateral responses during a second experimental condition during which participants were not required to keep information in mind, indicating that the modulation of sustained activity by emotional valence was particular to the experimental context requiring active maintenance. In another study, participants watched short videos intended to induce emotional states (e.g., clips from uplifting or sad movies), after which they performed challenging working memory tasks (Gray et al., 2002). Bilateral lateral prefrontal cortex activity reflected equally the emotional and working memory task components ( Figure 6). In other words, prefrontal activity did not stem from the working memory task alone or by the mood ensuing from the viewing of the video, but resulted from an interaction between cognition and emotion.Although this short review focuses on the impact of emotional content on cognitive functions, here we briefly discuss another important line of studies that has investigated cognitive-emotional interactions, namely, cognitive emotion regulation (Ochsner and Gross, 2005; Ochsner and Gross, 2008). A particularly informative regulation strategy is “cognitive reappraisal”, which involves rethinking the meaning of affectively charged stimuli or events in terms that alter their emotional impact. Reappraisal appears to depend upon interactions between prefrontal and cingulate regions that are frequently implicated in cognitive control and systems like the amygdala and insula that have been implicated in emotional responding. Interestingly, having the goal to think about stimuli in ways that maintain or increase emotion may boost amygdala activity whereas having the goal to decrease emotion may diminish it. Furthermore, changes in emotional experience and autonomic responding may correlate with the concomitant rise or fall of prefrontal and/or amygdala activity. Although much of the work on the cognitive regulation of emotion has relied on a relatively strict separation between cognition and emotion, which are in this context viewed as engaged in tug-of-war for the control of behavior, this framework is likely overly simplistic. As proposed by Ochsner and Gross (2008), a more fruitful approach will entail developing an integrated framework for specifying what combinations of interacting subsystems are involved in emotional responding, as individuals exert varying degrees and types of regulatory control over their emotions.
In attempting to understand the relationship between emotion and cognition, it is important to consider anatomical information. Advances in our understanding of brain connectivity suggest that a given brain region is only a few synapses away from every other brain region(Sporns et al., 2004; Sporns and Zwi, 2004). Indeed, it appears that the brain is configured according to a small-world topology in which the path length between nodes is small – typically, cortical areas are connected directly or via just one or two intermediate areas (Hilgetag et al., 2000; Sporns et al., 2000) – and nodes are highly clustered (Sporns, 2006). Thus, a careful consideration of brain connectivity is informative in understanding potential cognitive-emotional interactions.
In the past decade, several quantitative analyses of brain connectivity have been undertaken (Young et al., 1994; Stephan et al., 2000). Not surprisingly, prefrontal areas are among those most distant from the sensory periphery, suggesting that they receive highly-processed and integrated sensory information. Such potential insulation of the prefrontal cortex from the periphery is thought to be a key anatomical feature of this region and presumably confers the primate brain with a greater degree of flexibility (Mesulam, 2002). Highly processed information would also be able to support more abstract processing that is required for cognition. Interestingly, the amygdala, a region often linked to emotional processing, appears to be equally removed from the sensory periphery – although in some species, direct sensory thalamic projections may be present (LeDoux, 1996). In addition, the amygdala makes very widespread projections. Overall, it appears that the amygdala is very well situated to integrate and distribute information ( Figure 7).It is also instructive to consider the connectivity of the hypothalamus (Risold et al., 1997), as it has been long recognized for its importance in emotional behaviours (Swanson, 2000, 2003). In particular, via its descending connections that innervate brainstem motor systems, this structure is thought to play a key role in the implementation of goal-directed behaviors. Hypothalamic signals also can be conveyed to the cortex, mostly by way of the thalamus. Critically, prefrontal cortical territories project directly to the hypothalamus. Thus, the hypothalamus appears to be organized in such a way that it can generate both relatively reflexive behaviors and behaviors that are voluntarily triggered by inputs from the cerebral cortex (Swanson, 2000). Overall, this structure appears to be connected with all levels of the nervous system, including the neocortex (Swanson, 2000), enabling important hypothalamic regulatory signals to have widespread effects on the brain.
It is also important to consider the role of the ascending systems. For instance, the basal nucleus of Maynert is a major part of the so-called magnocellular basal forebrain system (Heimer and Van Hoesen, 2006). The projections from this system reach all parts of the cortical mantle (Heimer and Van Hoesen, 2006), and are involved in cortical plasticity in sensory cortex in the context of classical conditioning (Weinberger, 1995), in addition to arousal and attention mechanisms (see citations in (Sarter and Bruno, 2000; Heimer and Van Hoesen, 2006)). In particular, basal forebrain corticopetal cholinergic projections appear to be crucial for diverse attentional functions, including sustained, selective, and divided attention (Sarter and Bruno, 1999; Sarter et al., 1999; Sarter and Bruno, 2000). Of importance in the present context, the basal forebrain receives both cortical and amygdala inputs (for citations, see (Sarter and Bruno, 2000)). Notably, recent anatomical evidence suggests the existence of specific topographically organized prefrontal-basal forebrain-prefrontal loops (Zaborszky et al., 1999; Zaborszky, 2002; Zaborszky et al., 2005), so that specific prefrontal cortical targets of the basal forebrain connect back to sites from which the corticopetal fibers originate. Such loops provide a direct substrate for cognitive-emotional integration, for example by allowing amygdala signals to be broadcast widely, including to frontoparietal regions known to be important for the control of attention. More generally, the overall anatomical arrangement of the basal forebrain may involve multiple functional-anatomical macrosystems (Alheid and Heimer, 1988; Zahm, 2006) with wide-ranging effects on brain computations and important clinical implications (Alheid and Heimer, 1988; Sarter and Bruno, 1999). In summary, the picture that emerges from anatomical connectivity data suggests a remarkable potential for integration of information.
Historically, emotion and cognition have been viewed as separate entities. One factor that may have contributed to this separation in the past century is methodological. For instance, data arising from single-unit or lesion studies usually allow the researcher to only derive conclusions concerning the specific areas being targeted. Research in the past two decades suggests, however, that such a view is likely deficient and that, in order to understand how complex behaviors are carried out in the brain, an understanding of the interactions between the two may be indispensable. Indeed, some studies have suggested that it may be important to go beyond understanding interactions, some of which are suggested to be mutually antagonistic, to understanding how cognition and emotion are effectively integrated in the brain. As stated recently, at some point of processing functional specialization is lost, and emotion and cognition conjointly and equally contribute to the control of thought and behavior (Gray et al., 2002). While these statements were offered as a summary of specific findings concerning working memory performance following mood induction (see above), they may aptly characterize a vast array of real-world situations. In other words, whereas many behaviors may be reasonably well characterized in terms of cognitive-emotional interactions such that emotion and cognition are partly separable, in many situations, true integration of emotion and cognition may also take place ( Figure 8). The latter further blurs the distinction between cognition and emotion. See Duncan and Barrett (2007) for a similar view.