Concepts and terminology

Pessoa states that he will not define the terms “emotion” and “cognition,” but instead will use them “descriptively to refer to paradigms, task conditions, or ‘processes’ that are closer to the traditional, intended meanings of emotion and cognition” (p. 3). Although he observes that these meanings do not reflect real distinctions in kind but simply different lines of research, he often unwittingly perpetuates the emotion-cognition dichotomy by following the traditional usage. For example, in chapter 6 he describes “emotion” as referring to positive and negative affective states, and “motivation” as referring to motor actions of approach and avoidance, thereby following the standard division of the disciplinary territory into research on emotion and research on decision making. This division is perpetuated by discussing “motivation” and “value representation” with reference to decision making and by discussing “emotion” with reference to affective science, as well as by discussing these two bodies of literature in separate chapters. Yet, as Pessoa points out, these two lines of research actually investigate a single family of phenomena. Moreover, as he acknowledges, both lines of research manipulate both the positive and negative value (to use terminology from decision making) or valence (to use terminology from emotion literature) of a stimulus, which in turn influences both selective-attention processes that occur upstream of emotional state and action, and more downstream measures that include accuracy, choice, and eye movement. For example, as Pessoa describes in detail, both emotion and decision-making research employ aversive and appetitive conditioning to examine the influence of affective salience on attention, action, and control processes. The distinction between motivation and emotion also breaks down within affective science itself, as many emotion theorists include action tendencies as constitutive of emotion (e.g., Frijda Reference Frijda1986). Throughout the book there is also conflation between “motivation” and “emotion,” on the one hand, and the actual value or valence of a stimulus, on the other hand, where “emotion” sometimes refers to manipulations of negative value and “motivation” to positive value, in defiance of the previous distinctions between the terms.

For these reasons, we suggest that a consistent terminology would help to foster a common language between the parallel but mutually insulated areas of research in affective science and the science of decision making. Better still would be to treat these two areas as making up one single field of research. Hence, recent research coming from both traditions has benefited from distinguishing between “salience” (overall importance of a stimulus, negative or positive) and “value” in order to parse neural systems sensitive to salience, value, or both (Chikazoe et al. Reference Chikazoe, Lee, Kriegeskorte and Anderson2014; Kahnt et al. Reference Kahnt, Park, Haynes and Tobler2014).

This point is important for Pessoa's comprehensive review of amygdala function. He states, “though the precise contribution of the amygdala in these networks is still unknown, it does not map specifically onto emotion, but, instead, corresponds to broader and more abstract dimensions of information processing, including salience, ambiguity, unpredictability, and other aspects of biological value” (p. 78). We propose that positive and negative value (typically discussed in the emotion literature in terms of valence), ambiguity, and unpredictability are all aspects of overall salience. We have previously described “affective salience” as a quality that endows a stimulus with prioritized access to attention and action because it has been associated with pleasure/pain or approach/avoidance (e.g., novelty and surprise signal prediction error, which is arguably intrinsically affectively charged) (Markovic et al. Reference Markovic, Anderson and Todd2014; Todd et al. Reference Todd, Cunningham, Anderson and Thompson2012a). We describe this quality as “affective salience” in order to distinguish it from “objective salience,” which is related to physical features of an object as described in vision research (Todd et al. Reference Todd, Talmi, Schmitz, Susskind and Anderson2012b). The notion of “affective salience,” which incorporates features of ambiguity, novelty, surprise, and value, maps onto Pessoa's summary of amygdala function – determining what something is and whether it is important to attend to it and do something about it.

Along similar lines, it is also helpful to clarify the difference between emotional state and the influence of stimulus salience on behavior. “Emotional state” refers to a sustained state of arousal within the perceiver, whereas “stimulus salience” refers to the degree to which a discrete stimulus may bias attention because of its association with arousal, value, etc. This distinction is important but is often conflated in the literature. Pessoa discusses Mather's model of arousal-biased competition (Mather & Sutherland Reference Mather and Sutherland2011), which models the influence of affective state on attention and memory. We in turn have proposed the complementary notion of “affect-biased attention” (Todd et al. Reference Todd, Cunningham, Anderson and Thompson2012a), which is attention biased by the affective salience of the stimulus itself. Affect-biased attention is the predisposition to attend to certain categories of affectively salient stimuli over others; it tunes filters for initial attention and subsequent processing and thereby regulates subsequent emotional responses (Todd et al. Reference Todd, Cunningham, Anderson and Thompson2012a). Hence, it provides an important example of emotion-cognition integration in the brain and behavior.

Enlarging the dual competition model

Pessoa's dual competition model of cognition-emotion is a timely, sophisticated, and empirically grounded model that unifies disparate findings on the interrelations between affective salience, emotional state, and controlled cognitive processes into a coherent theoretical framework. In outlining this model, Pessoa proposes six key mechanisms for affective modulation of visual activity (pp. 161–66). The locus coeruleus/norepinephrine (LC/NE) system is not a component of this model, although it is discussed briefly in relation to its modulation of frontoparietal attentional networks via basal forebrain cholinergic and GABA projections. Beyond its interactions with other neurochemical systems, we propose that the LC/NE system in itself constitutes a seventh mechanism and is central to understanding the effects of salience on cognition and action.

The locus coeruleus is a brainstem nucleus that produces norepinephrine (NE) and sends projections to most regions of the brain, including the visual cortex, orbitofrontal cortex (OFC)/ventromedial prefrontal cortex (VMPFC), thalamus, and amygdala, as well as important nodes in networks mediating executive attention and motor response (Sara Reference Sara2009). Our own “biased attention via norepinephrine” (BANE) model outlines the role of this system, in interaction with the amygdala and OFC/VMPRC systems described in Pessoa's dual competition model, in biasing attention and memory to relevant aspects of the world (Markovic et al. Reference Markovic, Anderson and Todd2014). The LC/NE system plays an important role in the modulation of attention and control processes by affective salience (Sara Reference Sara2009; Sara & Bouret Reference Sara and Bouret2012). The LC is structurally well positioned to facilitate affect-biased attention. It receives inputs from the central nucleus of the amygdala (Berridge & Waterhouse Reference Berridge and Waterhouse2003), as well as from ventral prefrontal regions important for stimulus evaluation and decision making (for review, see Aston-Jones & Cohen Reference Aston-Jones and Cohen2005), facilitating tuning of LC activity to what is motivationally relevant. The LC itself also projects to regions of the thalamus and visual cortex (Jones & Moore Reference Jones and Moore1977), allowing for rapid tuning of sensory responses.

Nonhuman animal studies have found that motivationally relevant stimuli elicit LC response (for review, see Sara Reference Sara2009; Sara & Bouret Reference Sara and Bouret2012), and LC-NE activity has been shown to directly modulate visual cortex activation (Waterhouse et al. Reference Waterhouse, Azizi, Burne and Woodward1990). A wide body of evidence suggests that LC neurons facilitate responses to the overall salience of a stimulus (Sara Reference Sara2009), modulating activity for stimuli that are positive and negative in value, as well as those that are novel and surprising (Berridge & Waterhouse Reference Berridge and Waterhouse2003). Arousing stimuli elicit phasic LC activation, resulting in release of NE (Aston-Jones & Bloom Reference Aston-Jones and Bloom1981; Grant et al. Reference Grant, Aston-Jones and Redmond1988; Herve-Minvielle & Sara Reference Herve-Minvielle and Sara1995; Rasmussen & Jacobs Reference Rasmussen and Jacobs1986). Released NE may tune target neurons by improving their signal-to-noise ratio, inhibiting responses to neighbouring frequencies while sparing response to the best frequency (Manunta & Edeline Reference Manunta and Edeline2004). LC activity is also important in associative learning of what is salient. LC neurons fire in response to direct reward and punishment and subsequently to any stimuli associated with the salient event (Sara Reference Sara2009). Moreover, NE modulation of long-term changes in synaptic strength and gene transcription allow the LC/NE system to guide behavior based on stimulus salience within a given context (Berridge & Waterhouse Reference Berridge and Waterhouse2003).

NE also plays a crucial role in the emotional enhancement of memory via activity in the amygdala (Cahill et al. Reference Cahill, Haier, Fallon, Alkire, Tang, Keator and McGaugh1996; Reference Cahill, Gorski and Le2003; Roozendaal et al. Reference Roozendaal, McEwen and Chattarji2009). In humans, a deletion variant in the ADRA2b gene impairs alpha2b NE receptor function, putatively affecting tonic levels of NE availability (Small et al. Reference Small, Brown, Forbes and Liggett2001). The ADRA2b deletion variant has been linked to individual differences in tuning to affectively salient aspects of the world, including emotional enhancement of memory (de Quervain et al. Reference de Quervain, Kolassa, Ertl, Onyut, Neuner, Elbert and Papassotiropoulos2007). We have found that it also influences selective visual attention to affective salience. In an emotional version of the attentional blink task described by Pessoa (Anderson Reference Anderson2005; Anderson & Phelps Reference Anderson and Phelps2001), we found that carriers of the deletion variant show an “emotional sparing” or a reduced attentional blink for affectively salient stimuli (Todd et al. Reference Todd, Muller, Lee, Robertson, Eaton, Freeman, Palombo, Levine and Anderson2013). They also showed a stronger link between the perceived salience of stimuli at encoding and accuracy and confidence of subsequent recognition memory (Todd et al. Reference Todd, Muller, Palombo, Robertson, Eaton, Freeman and Anderson2014). Moreover, in deletion carriers we have found that enhanced visual processing of affectively salient stimuli is linked to greater activation in nodes of valuation networks (Rasch et al. Reference Rasch, Spalek, Buholzer, Luechinger, Boesiger, Papassotiropoulos and de Quervain2009). These findings support a role for the LC/NE system in affective biasing of visual attention in humans.

The BANE model makes a number of predictions that add to our understanding of processes outlined in Pessoa's model. These include the prediction that ADRA2b will mediate individual differences in the strength and duration of emotional learning. They also address the influence of NE activity on the relation between affective salience and other cognitive processes, as outlined in the dual competition model. One outstanding question concerns the relation between the notion of affective salience itself and that of prediction error: For example, can we say that things are by definition more salient if they elicit higher levels of prediction error? We suggest that a productive area for future research is to develop both the BANE and dual competition models in conversation with current views models of contextual influences on cognition, including predictive coding models (Clark Reference Clark2013; Grossberg & Seidman Reference Grossberg and Seidman2006; Summerfield & Egner Reference Summerfield and Egner2009; Summerfield et al. Reference Summerfield, Egner, Greene, Koechlin, Mangels and Hirsch2006).