What emotions truly are?
There is a wide agreement among emotion researchers, both empirical and philosophical, that the function of emotions is to evaluate perceived changes in the environment for their significance to the subject's concerns and to motivate adaptive responding to the situation. Each emotion type serves this general function of emotions in its own specific domain. Thus, the function of fear is to detect threats and dangers to the subject and to motivate fight or flight in response to danger. Likewise, the function of anger is to detect transgressions, offences, or slights against the subject and to motivate revenge or retaliation - either actual or symbolic - against the perpetrator. Or, the function of guilt is to detect one's own transgressions against others and to motivate submissive and apologizing behaviour towards the victim or victims. Moreover, emotion types are largely individuated on the basis of those evaluations and behaviours - both expressive and purposive - that they involve either on the basis of evolutionary hard-wiring or cultural learning or, as is often the case, through a combination of the two, namely culturally influenced evolutionary action readiness. Special-purpose mechanisms that operate on logical processing of information can be built into robots as well. However, robots cannot respond by 'gut appraisals', nor can their evaluative states be inherently affective in the same way as emotions. Feelings of emotion face both inwards and outwards: they emerge from the body's changing action readiness but they can also infuse our intentional representations of particular objects in the world in typical emotional experiences.
In spite of rapprochement in the big picture, emotion theorists still disagree on the more precise nature of emotions. Paul Griffiths reinvigorated this debate with his provocatively titled book What Emotions Really Are (1997). There he claimed that emotions do not constitute a natural kind but fall into two or three importantly dissimilar subtypes - evolutionarily primitive affect programmes, cognitively complex emotions, and disclaimed actions (socially sustained pretences) - that have little in common with each other. The main problem is that the causal mechanisms of different types of emotion do not coincide, allowing reliable projections of properties, functions, and explanatory principles from individual instances to all responses that we call "emotions" in vernacular language or even in science.
To illustrate, fear of a terrorist attack against one's home town is significantly dissimilar from fear of falling from a steep cliff. The former fear normally responds to information about the extremely small likelihood of such an attack, whereas the information that one is safe may not help to alleviate the latter fear when one is standing on the edge of a cliff. Fear of falling from cliffs is a disposition that evolution has built into our biological constitution because this response helped our ancestors to survive and reproduce while their less fearful contemporaries perished. Even if all fears are sensitive to dangers, dangers may be dissimilar in different cases, indicating that evolutionarily early and late fears have somewhat dissimilar functions in our mental and behavioral economy. Moreover, the persistence of evolutionarily early fears in spite of contrary information suggests that these emotions operate on at least partially different neural pathways than culturally learned fears in the human brain.
However, instead of closing the debate on the nature of emotions, Griffiths' rejection of "emotion" as a plausible scientific category inspired a wave of new theories that sought to bring emotions back into a single explanatory frame. As the result of this development, disagreements between traditionally opposite approaches have become more sophisticated and subtle, which has contributed to a rapprochement in emotion theory as all theories now must accommodate - in one way or another - the intentional aboutness, affective phenomenology, evaluative function, motivational force, and rational justifiability of human emotions. Even so, it is important to emphasize that the rapprochement is limited to the set of those properties of emotion that any plausible theory must explain. Different theories of emotion still diverge from each other in their ways of accommodating these properties even if their disagreements are more difficult to detect behind a wide consensus on the properties that require explanation. One such disagreement is whether emotion involves cognition always, sometimes, or never. Cognitive theories of emotion are plausible only if human emotions always rather than merely sometimes involve cognitions; mere causation or regulation by cognition is insufficient unless cognitions are also constituents of emotion, as Clore and Ortony (2000) point out. Otherwise emotions are either noncognitive, or they divide between two main types, cognitive and noncognitive, as Griffiths argued.
Cognitive theories suggest that emotions are always cognitive because they can serve their evaluative function only if they involve appraisals or evaluative judgments of their particular objects in their content that can be explicated in terms of conceptually structured propositional attitudes. Cognitivists emphasize that the content of all emotions need not be constituted of concepts; otherwise the theory could not accommodate such emotions as fear of flying that we feel in spite of contrary evaluative judgments about the situation, or the emotions of pre-linguistic infants and higher animals who do not possess semantic concepts in the first place. However, cognitivists maintain that the content of emotion is nevertheless conceptually explicable even in those cases where it is not constituted of concepts (Nussbaum, 2001; Roberts, 2003).
In contrast, noncognitive theories deny that emotions must involve concepts in order to serve their evaluative function. Ascribing concepts to all human emotions is metaphorical at best and misleading at worst. Cognitive appraisals may contingently elicit and regulate emotions but they are never constituents of emotion. Moreover, the content of emotion can be explicated in functional terms that emphasize a reliable causal co-occurrence between emotions and those things -dangers, losses, offences, and so on - that emotions represent in their content (Prinz, 2004). Or the content of emotion is nonconceptual, constituted of fast, automatic, and highly modular appraisals that occur at a subpersonal and subdox-astic level of information processing which is distinct from cognition proper (Robinson, 2005).
In order to adjudicate the dispute about the role of cognition in emotion, we must clarify what we mean by cognition. A well-known debate in the 1980s between psychologists Robert Zajonc (1923-2008) and Richard Lazarus (1922-2002) addressed this question (for a review, see Schorr, 2001). In his studies on subliminal perception, Zajonc (1980) found that we can form likings and aversions to objects on the basis of mere exposure to them, without conscious awareness of exposure and recognition of the object. This phenomenon is known as "the mere exposure effect". From these empirical results Zajonc concluded that feeling precedes cognition, or, "preferences need no inferences". Moreover, Zajonc claimed that affect constitutes an information processing system that is functionally and neuroanatomically independent from conscious thinking as there are separate brain structures, neural pathways, and neurotransmitters for cognition and affect. This neurophysiological evidence explains why emotions can be recalcitrant and outlast changes of beliefs and appraisals.
Lazarus (1982, 1984) admitted Zajonc's empirical points but maintained that they do not undermine appraisal theory if we understand the notion of cognition in a broad sense. Thus he argued that in emotion, cognition always mediates the relationship between person and environment. Information processing cannot constitute relational meaning without some kind of evaluation of the information for its relevance to the subject's well-being. However, cognitive activity in appraisal does not imply deliberate reflection, rationality, or awareness. Moreover, Lazarus argued that mere feelings and preferences created through exposure effect are not yet emotions. Nevertheless, all processes of meaning generation are cognitive. An appraisal may operate at different levels of complexity, from the most primitive and inborn to the most symbolic and experience-based. Therefore, cognitive appraisal is always involved in emotion, even in creatures phylogenetically more primitive than humans.
The debate resolved in a consensus that the proper question concerns the correct definition of cognition, as Leventhal and Scherer (1987) pointed out. If cognition requires conscious reflection and thought, as Zajonc proposed, all emotions clearly do not require cognitive mental states as their necessary causes or constituents. Yet if every kind of receiving and processing information for its significance to the subject's concerns counts as cognition, as Lazarus suggested, all emotions come out as cognitive. Zajonc (1984) rejected the latter view of cognition as overly inclusive, because it blurs distinctions between cognition, perception, and sensation. However, no significant progress from these positions on cognition has taken place in emotion theory since the days of the Zajonc-Lazarus debate. Instead, a new dividing line between cognitive and noncognitive theories of emotion has emerged regarding their preferred views of cognition. Cognitive theorists of emotion typically favor broad functional accounts of cognition, whereas noncognitive theorists prefer narrow neuroanatomical, mechanism- or code-based accounts, or several of them (Moors, 2007). Thus, Martha Nussbaum (2001, p. 129) defends "a multifaceted notion of cognitive interpretation or seeing-as, accompanied by a flexible notion of intentionality that allows us to ascribe to a creature more or less precise, vaguer or more demarcated, ways of intending an object and marking it as salient." In contrast, Jesse Prinz (2004, p. 46) argues that "a cognitive act is an act of generating a thought under top-down control." These characterizations of cognition by leading representatives of cognitive and noncognitive theories of emotion in philosophy display no headway from the respective positions of Lazarus and Zajonc.
In this situation, it would be convenient to embrace and defend one account of cognition and examine whether or not emotions come out as cognitive or noncognitive from that perspective. Indeed, this straightforward strategy has been popular both among philosophers and affective scientists. Unfortunately, it is not a very informative strategy because the different concepts of cognition operate at different levels of analysis, as Agnes Moors (2007) remarks in her useful taxonomy and review of definitions of cognition. It shows that arguments for cognitive or non-cognitive views of cognition at one level do not solve the question once and for all at other levels of analysis. Individual authors typically downplay this problem by resorting to their favorite conception of cognition. Yet the fact that these conceptions operate at different levels of analysis reduces the force and appeal of this argumentative strategy.
Therefore, I suggest that we must confront and answer the question about the role of cognition in emotion at different levels of analysis. Following David Marr's (1982) classic analysis of vision that Moors utilizes in her taxonomy as well, the relevant levels of analysis are functional, algorithmic, and implementation. Of these three levels, the first and the second are relevant from a philosophical point of view. At the first level, the question is whether or not the process that leads from the intake of perceptual or conceptual information to the triggering of an emotional response always involves cognition. At the second level, the mechanisms and forms of representation involved in the processing of input into output come under focus. Examples of mechanisms are rule-based and associative processes, whereas forms of representation include conceptual and nonconceptual representations. The third level of analysis concerns the physiological and neural implementation of emotional processes in the brain. Accordingly, this level of analysis belongs to the domains of biological psychology and affective neuroscience.
In what follows, I shall focus on the role of cognition at the first two levels of analysis, functional and algorithmic, that already on disciplinary grounds belong to philosophical and psychological theorizing on emotions. Empirical evidence on the implementation of emotions in the human brain and elsewhere in the body is obviously important as well. Therefore, I include discussion on the neurophysiology of emotions in some sections to support my discussion at the other levels of analysis. Even so, questions about the role of cognition in emotion can never be solved by studying merely the brain. The basic reason is that we cannot identify the phenomena whose implementation in the brain we should study without reference to our experience. Indeed, an initial identification of "emotion" on the basis of phenomenological and conceptual investigation must precede any empirical study of emotion because otherwise we would not know what to study and how to demarcate "emotion" from other mental and bodily states with which it mixes and mingles all the time. Moreover, we cannot drop phenomenological and conceptual research even after emotion has been tentatively identified, because emotions are not purely physiological phenomena whose study could be handed over to natural science in the same way as the study of chemical substances. Instead, emotions have features such as experience and intentionality, a place in complex webs of meaning, and susceptibility to rational standards of warrant that resist scientific treatment (Roberts, 2003; Döring, 2007). For all these reasons, I discuss cognition in emotion at the level of implementation only in connection with the other two levels of analysis in Chapter 2.
However, this examination can only provide us with a typology of those representations and processes that figure in human emotions. Depending on one's preferred theory, emotions may last from seconds to minutes or even days. Few jokes amuse us longer than a few seconds, whereas grief at the loss of a significant other may last for weeks or months, even years. Time is an important factor because the longer emotions last, the more obviously they involve many types of representations and processes that interact with each other in producing and reinforcing emotions. Thus if emotions typically last from 0.5 to 4 seconds as affect programme theorists such as Richard Levenson (1988) suggest, appraisals are capable of serving as mere triggers of highly stereotypic emotional responses. Instead, if emotions are taken to last longer, from seconds to minutes or hours, appraisal can be understood as a process that constantly updates information about the emotion-eliciting situation in relation to one's active goals as well as to feedback from the body and the world, modifying the emotional response accordingly. The theoretical question is where to draw the line between an emotion and its regulation and where does cognition fit in this divide.
Emotion regulation is a process that extends beyond the generation of an emotional response and continues until its termination (Gross, 1998). This means that processes of emotion generation and emotion regulation overlap and intertwine during an emotion. Different theoretical approaches disagree on whether generation and regulation should be conceptualized as separate processes or as complementary aspects of a single process of emotion. Thus noncognitive theories prefer to define emotion in terms of processes and representations that figure in emotion generation, whereas cognitive theories maintain that generative and regulatory processes intertwine, giving all post-infantile human emotions a cognitive imprint (Gross & Barrett, 2011). Even so, the role of regulation in emotion has not been problematized in theoretical debates between noncognitive and cognitive theories on the nature of emotions (with the exception of Robinson, 2005). This is unfortunate because empirical research can offer evidence which helps us to adjudicate which approach - the narrow noncognitive or the broad cognitive - to the role of regulatory processes in emotion is correct. This question on the role of cognition in the dynamics of emotion will occupy me in Chapter 3.
-  Griffiths introduces disclaimed actions as the third main type of emotion in his book What Emotions Really Are. These are emotions experienced and enacted in transitory social roles that are functional either for the individual or for society or both. In order to serve their functions, these emotions are interpreted as passive and involuntary even if they did not exist without the relevant social roles as James Averill (1982) remarks. Griffiths later focuses on the distinction between affect programs and complex emotions to which he possibly includes disclaimed actions.
-  Here and in what follows, "concept" is understood as referring to semantic, linguistic concepts. I acknowledge that it is possible to ascribe psychological categories that functionally resemble concepts in mediating distinct behavioural responses to equally distinct perceived events to even nonlinguistic animals. In the Representational Theory of Mind, such categories are conventionally called "concepts". However, I believe that it is useful to distinguish concepts of this kind from semantic concepts as there are logical differences between the two. I will return to this topic in the next chapter when discussing differences between conceptual and non-conceptual content.