Current Emotion Research in Philosophy

What is Philosophy of Emotion?

Much philosophical work on the emotions is pursued by adherents of philosophical naturalism, the view that philosophy deals in knowledge of the natural world no different in principle from that revealed by the sciences. These philosophers look to the sciences of the mind to illuminate philosophical questions about emotion. In many areas of philosophy, naturalism is uncontroversial. The philosophy of physics proceeds by the analysis of historical and contemporary physics. The same can be said of the philosophy of biology, or the philosophy of economics. But it cannot be said of the philosophy of mind. Many philosophers still believe that important truths about the mind can be discovered using distinctively philosophical methods and working independently of the sciences of the mind. Some also believe that these truths are different in kind from those revealed by the sciences.

Two recent books on disgust exemplify these two approaches in the philosophy of emotion. Colin McGinn’s The Meaning of Disgust (2011) sets out to identify the essence of disgust—that which all disgusting things have in common—by reflecting on the author’s ideas about what is disgusting. McGinn concludes that disgust is produced by objects which mix death and life. This “death-in-life” theory is to be preferred to rival theories because it alone allows disgust to be defined with necessary and sufficient conditions. McGinn also offers an evolutionary explanation of disgust. It is a uniquely human emotion and developed because the insatiability of human desire required a special emotion to control our limitless appetites. McGinn imagines an evolutionary past in which, for example, humans lacking disgust engaged in necrophilia as an expression of their limitless sexual desire. Reviews of McGinn’s book have found it puzzling that he does not consider the most widely discussed theory of disgust in contemporary psychology, according to which disgust evolved as a response to potential sources of infection, and that he largely ignores the scientific literatures on disgust and its evolution. The second book, Daniel Kelly’s Yuck! The Nature and Moral Significance of Disgust (2011), could not be more different in this respect, being based on a thorough survey of those literatures. Kelly argues that disgust combines a taste-aversion mechanism with a mechanism for avoiding infection and parasitism. Kelly sees disgust as a uniquely human emotion, but his account of its uniqueness and its evolutionary origin is very different from McGinn’s. Disgust is a typical example of the co-option and developmental entanglement which occurs as an evolving lineage confronts new adaptive challenges and must solve them by modifying existing features. Both elements of the disgust response have homologs or analogs in other species—it is their entanglement that creates a uniquely human emotion. Moral disgust represents a further piece of evolutionary tinkering in which the emotion of disgust was co-opted to help enforce conformity with in-group norms.

Philosophers who work in these two modes can be very critical of one another. Kelly concludes his review of McGinn, whose theory he describes as “coherent, if incredible,” with the observation that:

Despite rather quaint use of the word “data” to describe the examples of what he deems to be disgusting … nowhere does he consult current empirical literature for evidence about disgust, cross-cultural or otherwise. The resulting book is as disappointing as it is irresponsible. (Kelly, 2012)

McGinn has not to my knowledge reviewed Kelly’s book, but naturalistic work on the emotions is regarded by its critics as unduly reductive, as a manifestation of scientific imperialism, and as unable or unwilling to address the real questions about emotion.

Amongst the questions which naturalistic approaches are supposedly unable to address, the role of the emotions in moral psychology is perhaps the most prominent. This critique is at the heart of Robert C. Roberts’ Emotions: An Essay in Aid of Moral Psychology (2003). Roberts rejects the conventional view of conceptual analysis as the study of our own concepts (Jackson, 1998). Instead, “conceptual analysis of the emotions [is] the investigation of them from the point of view of a human participant” (Roberts, 2003, p. 37). Its focus on rich human experience is similar to “literary and psychoanalytic examinations of emotion and a rather strong contrast with biological and neuroscientific examinations” (Roberts, 2003, p. 5). Roberts opens his book with an example of literary insight into moral character, namely Anthony Trollope’s 1876 description of his character Ferdinand Lopez, a London share trader. Roberts contrasts the answer to the question “what is an emotion” implicit in this passage with answers from the sciences, and suggests that the former is more suited to moral psychology (Roberts, 2003, p. 19). To be frank, I found little moral insight in this bit of Trollope, and a good deal of stereotyped Victorian anti-Semitism. I have no desire to decry the value of humanistic approaches to the emotions, but the special value of those approaches for moral psychology needs to be defended, not merely asserted. This is particularly true in the light of the recent flowering of naturalistic moral psychology, which looks to the sciences of the mind in exactly the same way as the naturalistic philosophy of emotion (Prinz, 2007; Sinnott-Armstrong, 2008).

About a decade ago there was an upsurge of naturalistic work in the philosophy of emotion. My own What Emotions Really Are: The Problem of Psychological Categories (Griffiths, 1997) pursued an ethological and evolutionary approach to the emotions. Its central thesis was that the psychological processes encompassed by the vernacular category of “emotion” are too diverse to make emotion per se a useful category for scientific investigation. Like memory, emotion is a collection of different psychological processes. Unsurprisingly, my book devoted a good deal of attention to the work of Paul Ekman and his collaborators. Ekman’s “basic emotions” were treated as an example of the value of carving out a subset of the vernacular category of emotion and analyzing them on their own terms. The basic emotions were also used as a case-study of psychobiological classification. The book argued that the basic emotions are homologs—categories defined by shared evolutionary origins. The basic emotion of fear, for example, exists in many mammalian species and it is the same thing in each just as the femur or the hypothalamus is the same thing—it has a shared evolutionary origin in the common ancestor of those species. This approach to categorization is very different from classification by shared function, which would place in the category of fear any psychological mechanism designed to detect and respond to danger. Biologists call such categories analogies. Homology and analogy are both useful, and indeed complementary, principles of classification, but it is essential to distinguish them and to use each appropriately.

Craig DeLancey’s Passionate Engines: What Emotions Reveal About Mind and Artificial Intelligence (2001) also focused on the basic emotions, and on Ekman’s claim that each of these corresponds to an “affect program,” a coordinated array of physiological and behavioral reactions. DeLancey used affect program theory to offer a new approach to the intentionality of emotion. This has been a central topic in the philosophy of emotion since the 1960s, when Anthony Kenny was the most prominent of several philosophers to argue that emotions are subject to normative standards of “fit” to the world—they can be appropriate or inappropriate, reasonable or unreasonable. This suggests that emotions have intentional objects—it is inappropriate and unreasonable to fear things that are not dangerous because fear represents the world as dangerous (Kenny, 1963). This in turn was widely held to refute theories which identified an emotion with a sensation or a physiological state per se (Kenny, 1963). Robert Solomon’s influential The Passions (Solomon, 1976), which argued that emotions are constituted by judgments about the significance for the subject of the situation confronting them, was the mature expression of a consensus that developed in philosophy in the intervening years. However, the view that emotions are judgments creates a dilemma. Commonly accepted statements of the content of emotional judgments involve sophisticated social and normative concepts. For example, anger is the judgment that the agent has suffered “a demeaning offense against me and mine” (Lazarus, 1991, p. 222; endorsed by Nussbaum, 2001; Prinz, 2004b). The intentionality of emotions is thus placed in tension with the plausible claim that adult human anger has something deeply in common with anger in frustrated infants, or the anger of a dominant monkey towards a subordinate. Either these agents possess the concepts of “demeaning” and “self,” or they are not angry in the true sense of the word. This problem is made worse by evidence that emotions in adult humans can be produced by low-level processes that seem equally implausible locales for such concepts.

DeLancey responds to this dilemma by arguing for the “heterogenous intentionality” of basic emotions (DeLancey, 2001, pp. 89–98). Basic emotions can be intentionally directed at a state of affairs, so that their content is a proposition. For example, I may be afraid that this dingo will bite me. But the very same emotion may be intentionally directed at what DeLancey calls a “concretum,” meaning an object as such, rather than as an element in a proposition. For example, I may be afraid of this dingo. This is where affect-program theory comes in. Because emotions are intrinsically action-directing, an emotion whose content is a concretum can nevertheless explain action. I flee the dingo because I am afraid of it. In order to flee I do not need a proposition about the dingo, such as that it is dangerous or that it will bite me, combined with a desire to avoid danger or not to be bitten. I just need to be afraid, and for the target of my fear to be the dingo. What we have in common with other animals, DeLancey argues, is the ability to have emotions that are intentionally directed at a concretum. What distinguishes us is the ability to have emotions intentionally directed at propositions. This ability introduces a far greater flexibility into our emotional responses. An instance of this flexibility which I think DeLancey could emphasize more strongly is that the ability to have proposition-directed emotions allows emotions to occur as the result of a chain of reasoning with the relevant proposition as its conclusion.

A third, influential naturalistic treatment of emotions was Jesse Prinz’s Gut Reactions: A Perceptual Theory of Emotion (2004b). Prinz’s theory builds on the “somatic marker hypothesis” (Damasio, 1994), which is itself a development of the late 19th-century James-Lange theory that emotions are perceptions of the body’s automatic reactions to stimuli. In this vein, Prinz argues that emotions are perceptions of aroused states of the body, or “gut reactions.” Prinz has a distinctive account of the intentionality of these gut reactions. Although emotions are not propositional representations, their content is correctly represented by the propositional descriptions with which we are familiar. The content of anger is that a demeaning offense to me or mine has been committed. To assign such complex intentional contents to gut reactions Prinz embraces a form of teleosemantics. This is the view that the content of a representation is the state of affairs which it is the function of that representation to detect. The actual “vehicle” of representation can be very simple, as long as that vehicle was designed to fulfill the right function. In an earlier book Prinz argued that all mental representations are perceptual images, and that these images can be used to represent the whole range of topics about which we are capable of having thoughts (Prinz, 2002). Applied to the emotions, this implies that they are perceptual images of the subject’s body which are used to represent the significance for the subject of the situation confronting them. Although anger itself is a perception of the body, its function is to detect demeaning offenses. What Prinz calls the nominal content of anger is an aroused body, but its real content is the proposition that a demeaning offense has been committed against me or mine.

At the heart of Prinz’s theory are a small range of basic emotions, but he builds on these to explain the whole range of human emotions. More complex emotions are simply basic emotions triggered by acquired sets of associations. Prinz calls these sets of associations “calibration files.” He argues that these files are not part of the emotion itself, properly understood, but merely its causes. The emotion in itself is only the representation of the bodily state, and hence all emotions are fundamentally “gut reactions.”

The methodological disagreement that I have described does not align neatly with positions on other issues. In the remainder of this article I will canvass three questions that have featured in the philosophical literature of the last decade. Naturalists and their critics will be found on both sides.

What is an Emotion?

According to Prinz, “All emotions are constituted by embodied appraisals alone . . . Emotions are a natural kind in a strong sense. They share a common essence. It is rare for nature (and folk psychology) to offer such a neat category” (Prinz, 2004b, . 102). However, having a neat definition, or even an essence, is not enough to make a category a natural kind. On the view of natural kinds that both Prinz and I accept, “natural kinds” are categories which admit reliable extrapolation from samples of the category to the whole category, or in other words, categories which are productive objects for scientific investigation. Aristotle thought that objects outside the orbit of the moon formed such a category. He was wrong, despite the neat definition of superlunary objects. I have continued to argue that the overall category of emotion is not a natural kind (Griffiths, 2004a, 2004b). It is instructive to compare emotion to memory. In light of distinctions like that between working memory and long-term memory, or episodic and procedural memory, a neat definition of memory would not show that these processes are essentially the same. All it would do is to neatly capture the similarities of function that explain why these processes are referred to by the same vernacular term. A definition of memory might be based on shared process or mechanism, but then it would probably not include all and only the things covered by the vernacular term.

The two main arguments for the view that emotion is a natural kind are what I will call the “entanglement argument” and the “argument from affective neuroscience.” The substantive argument which backs up Prinz’s assertion quoted before is an example of the argument from affective neuroscience. According to Prinz, affective neuroscience shows that at the core of every emotion is a perception of a bodily registration of the significance of a stimulus. The difference between more or less cognitively sophisticated emotional responses is only in the way they are caused, not in the emotion itself. Louis Charland’s version of the argument from affective neuroscience does not depend on a specific theory of emotion. Charland argues that neuroscience has identified a set of core processes that map reasonably well onto the vernacular category of emotion. These mechanisms capture what is common to a wide range of organisms and which leads us to recognize them as having an emotional life (Charland, 2001, 2002). It can be seen that Prinz, Charland, and I are asking the same fundamental question: what big-picture lessons about the nature of emotion can be derived from the current state of the sciences of the mind?

An important variant of the argument from affective neuroscience has been proposed by philosopher and cognitive scientist Jason A. Clark, who has argued that basic emotions and more cognitively sophisticated forms of the same emotion may be homologous to one another (Clark, 2010a, 2010b). The idea that emotions are homologous across species (taxic homology) is a familiar one. The facial expression of anger and the brain regions that produce anger have homologs in other primates, so why should anger itself not have psychological homologs? Clark’s innovation is to propose that simple and complex emotions in the same species may be homologous. The same evolved developmental patterns are used twice, just as the developmental patterns that produce skeletal elements are used twice in our arms and legs (serial homology). Clark proposes that a form of shame similar to that in other primates and a more cognitively sophisticated form unique to humans are both found in humans. We have two different kinds of shame in the same way that bats have two different kinds of phalanges, those in their feet which are similar to other mammals, and those in their hands which are dramatically specialized for their unique lifestyle. Despite their dramatic differences, these bones are fundamentally the same anatomical part—phalanges. Clark proposes that radically different forms of shame (and pride) may nonetheless be fundamentally the same emotion.

The entanglement argument is somewhat different from the argument from affective neuroscience, although the work of the neuroscientist Antonio Damasio is often cited to back it up. It proposes that the different processes involved in emotion are so bound up in natural episodes of human emotion that they cannot or should not be separated. The argument is presented at length by Robert C. Roberts, in “the most comprehensive defence to date of the coherence of the category ‘emotion’” (Döring, 2007, p. 372). His target is my (Griffiths, 1997) presentation of the view that experimental disassociations between cognition and emotion reveal the existence of separable, low-level processes which are capable of triggering an affect program response (Öhman, 1999; Zajonc, 1980, 2000). These phenomena are often discussed in psychology under the heading of “multilevel appraisal” (Scherer, 1999; Teasdale, 1999). I used such work to argue that there are two or three fundamentally different processes going on in “emotion” that, while these can occur together, can also occur independently of one another, and that the way forward for the study of emotion is to distinguish these processes, rather than studying “emotion” as a whole. In reply, Roberts describes intuitively plausible cases in which higher cognitive processes cause affect program responses. Such cases are so devastating that I was “shy to acknowledge” them and my book “seems to try to prevent the reader from looking at the obvious evidence” (Roberts, 2003, pp. 26–27). Roberts is unimpressed by my own discussion of cognition–emotion relations. I discussed Ekman’s well-known work on display rules, the use of cognitive management strategies to modulate emotional response, and various ways in which affect programs can be triggered by higher cognitive processes. My discussion, however, was “evasive and reducing” (Roberts, 2003, p. 25) because the word “triggering” glosses over cases where “the object of an affect program emotion is wholly and fundamentally presented by higher cognition and could be presented in no other way” (Roberts, 2003, p. 26). But this idea is ambiguous. On one interpretation, it could mean that there are some emotion episodes where an affect program response is triggered by a high-level appraisal process which makes use of concepts that could not plausibly be attributed to lower level processes (or to simpler organisms). On this first interpretation the idea is uncontroversial and is entirely consistent with the claim that on other occasions the same affect program is produced by low-level appraisal. Alternatively, Roberts might mean that these sophisticated concepts are present in the affect program itself, so that it could not occur without them, and no simpler organism without such concepts could have the same emotional response. This second, stronger interpretation is supported by the fact that Roberts also takes his cases to refute the idea that affect programs are informationally encapsulated (Roberts, 2003, p. 28). But so far as I can see Roberts has not argued for this stronger interpretation, or against the experimental evidence that the same affect program response can be triggered by different levels of appraisal.

Supporters of the entanglement thesis often cite the work of Damasio as evidence that no distinction can be drawn between higher and lower level emotional processes. The eminent hilosopher Simon Blackburn writes that:

Empirically, the suggestion that we split the operation of the affect program from “higher cognitive emotion” seems to ignore the most fascinating result of Damasio’s work, which is the extent to which “higher-order” decision making has to harness the limbic system in order to work at all. (Blackburn, 1998, p. 129)

But, empirically, the operation of higher and lower level processes in emotion can be split, in phobias and in the experiments reported in the literatures on multilevel appraisal (Öhman, 1999, 2002; Teasdale, 1999) and “affective primacy” (Zajonc, 1980, 2000). The interaction of affect and cognition is a flourishing field of research (e.g., Forgas, 2001), but this does not mean the two are so entangled that they should be treated as one.

There is a fundamental difference between the use of Damasio’s work as part of the “argument from affective neuroscience” and its use in the entanglement argument. The former seeks to identify a core of basic processes which play a role in all emotions. If that idea is correct it is in real tension with the claims of my 1997 book. The entanglement argument, however, seeks to show that lower- and higher-level emotional processing cannot be separated. Roberts thinks this is adequately demonstrated by Damasio’s work and that my failure to see this is “another example of [Griffiths’] evasiveness”:

Damasio’s basic thesis that organs like the amygdala can be triggered by higher cognitive processes is almost certainly correct and devastating for the basis thesis of Griffith’s book. One would expect Griffiths to aim all his firepower at that thesis, but instead he presents arguments against the weaker parts of Damasio’s picture, arguments that create a bit of smoke to keep the reader from seeing that Griffiths did not clearly answer the crucial argument against his view that the affect programs and the higher-cognitive emotions are two completely discrete classes of things. (Roberts, 2003, pp. 27–28)

Roberts apparently believes that the issue is whether emotional episodes come in two discrete and nonoverlapping kinds, those that involve higher cognition and those that do not. His version of the entanglement argument is like denouncing the distinction between working memory and long-term memory because items in working memory sometimes pass into long-term memory.

As we have just seen, philosophers who do not adopt a naturalistic approach to the philosophy of emotion are nevertheless enthusiastic about the work of Damasio. Martha Nussbaum is another eminent figure who praises his “nonreductionistic” approach to neuroscience (2001, p. 115). Opinion amongst naturalistic philosophers is more divided. Prinz has shown how the somatic marker theory can be expanded into a general account of the emotions (Prinz, 2004b) and of the role of emotion in moral psychology (Prinz, 2007). Others philosophers have emphasized the gap between the actual findings of Damasio and his collaborators and the broad picture of the role of emotion in mental life painted in his popular books and adopted wholesale by some philosophers. Giovanna Colombetti distinguishes two versions of the somatic marker hypothesis. In the first, somatic markers are the fundamental source of valence in decision making. In the second, somatic markers have a specific role in thinking about distant outcomes. Colombetti argues that neither hypothesis is supported by the data from Damasio’s key experimental paradigm, the Iowa gambling task (Colombetti, 2008). Philip Gerrans defends the second hypothesis, and argues that Damasio’s subjects had a specific deficit in the ability for “mental time-travel” (Gerrans, 2007, p. 459). There is nothing unusual about the existence of this kind of gap. Scientists need to go beyond what is immediately implied by their data if they are to develop theoretical frameworks that guide future experimentation. One aim of naturalistic philosophy is to ask to stand back from any particular research program and ask to what extent some broad conclusion is supported by the current state of knowledge. The work of Stefan Linquist and Jordan Bartol suggests that the broad conclusion drawn by Blackburn (see previous lines) is not well supported:

First, contrary to conventional wisdom there is no single somatic marker hypothesis. Even within the writings of its chief proponent, Antonio Damasio, somatic markers are assigned a variety of different functions … A second myth concerns the stages of practical decision making at which somatic markers are most likely engaged. The received view is that the available evidence suggests a role for somatic markers in the “core stages” of decision making, where a subject generates, evaluates and selects among alternative courses of action. We argue that Damasio’s own evidence suggests otherwise. Somatic markers appear to be involved (if at all) in the “peripheral” stages of decision making: one possibility is that they are engaged early-on, in notifying the subject that a decision-point has been reached; another possibility is that somatic markers are engaged at the terminal stage, in motivating the subject to execute a decision. (Linquist & Bartol, in press)

The pattern(s) of interaction between affect and cognition continue to be an important topic for philosophy. Some leading figures continue to argue that emotions are judgments (e.g., Nussbaum, 2001). Andrea Scarantino has forcefully restated the view that this is inconsistent with empirical evidence that emotional responses can be triggered by low-level appraisal (Scarantino, 2010). Scarantino argues that judgmentalists have responded to this evidence, and also to the existence of animal emotions, by stretching the concept of judgment until it becomes trivial: “all instances of emotions eventually manage to find room in the unboundedly malleable and arguably unprincipled notion of judgment” (Scarantino, 2010, p. 746).

According to one philosopher (Starkey, 2007), the experimental evidence Scarantino (2010) adduces is irrelevant to the issue of whether all emotions involve higher cognition, because the experiments involve placing the subject in abnormal conditions or using subjects with damaged brains. Charles Starkey argues that responses produced under these conditions do not count as emotion. Emotion should be restricted to states that occur when the system is intact and operating under the conditions for which it was designed (Starkey, 2007). Psychologists will be relieved to know that Starkey does not intend a wholesale rejection of the experimental method: his point is purely about the definition of emotion. However, I think it is fair to say that most researchers working in this area do not think of their work as merely the creation of abnormal phenomena in the laboratory, but as throwing light on the relative independence of different levels of emotional appraisal in normal life.

Are There Natural Kinds of Emotion?

A good deal of philosophical discussion has been provoked by the rise of core affect theory in psychology. Psychological constructionists like James A. Russell and Lisa Feldman Barrett deny that subordinate emotion categories like fear and anger are natural kinds (this is consistent with thinking that the superordinate category of emotion is a natural kind). Constructionists identify pleasure and arousal as affective primitives which combine into core affect, a blend of hedonic and arousal values that is “the most basic building block of emotional life” (Barrett, 2006, p. 48). All affective states are built out of core affect, perhaps with other ingredients. Mood is “prolonged core affect without an object” (Russell, 2003, p. 149). Discrete emotions emerge from a conceptual act of categorization of core affect (Barrett, 2006). A core affective state of high pleasantness and high arousal may be categorized as happiness, a state of low pleasantness and high activation as fear, a state of low pleasantness and low arousal as sadness, and so forth. “[C]ategorizing the ebb and flow of core affect into a discrete experience of emotion corresponds to the colloquial idea of ‘having an emotion’” (Barrett, 2006, p. 49). The key idea here is that discrete categories are imposed on underlying psychological processes that are continuous in nature. This general idea is not new, but by constructing a detailed model and presenting evidence in its support, core affect theory has become an important player in contemporary psychology of emotion.

There have been some productive interactions between psychologists and philosophers about core affect theory. Andrea Scarantino has argued that much of the evidence for the artificiality of discrete emotion categories targets the vernacular categories used in everyday life. The real target of core affect theory, however, is the scientific taxonomy of discrete emotions represented by Paul Ekman’s list of basic emotions (e.g., Ekman, 1999). Core affect theorists have evidence of the artificiality of that taxonomy too, but Scarantino’s point is that proper assessment of the relative merits of basic emotion theory and core affect theory requires that the disputants separate the question of whether our existing, vernacular categories of emotion are natural from the question of whether there are natural categories to be found (Scarantino, 2009; see also Scarantino & Griffiths, 2011).

Giovanna Colombetti has argued, using her dynamical systems account of emotion discussed in what follows, that discrete emotions are “softly assembled” as reasonably consistent atterns that emerge as the brain interacts with its context (Colombetti, 2009a). Colombetti argues that her proposal can do justice to both the evidence for discrete emotion types, and the evidence of the variability and context-dependence of emotion mustered by core affect theorists. In the exchange that followed, two things became clear (Barrett, Gendron, & Huang, 2009; Colombetti, 2009b). First, the disagreement between basic emotions theorists and core affect theorists is as much about the data itself as about which framework best explains that data. Barrett and collaborators do not regard it as an advantage that Colombetti’s framework can explain the evidence for discrete emotions: they do not think there is much to explain! Second, Barrett and collaborators emphasize that the categorization of core affect is not a superficial overlay on the underlying process. Categorization is at the heart of what brains do, and even highly culturally specific emotion categories may be implicit in the way the brains of enculturated individuals process emotional stimuli.

Is Emotion a Form of Perception?

Perceptual theories of emotion have become increasingly popular in philosophy (for a review, see Salmela, 2011). Perceptual theories seem to combine the best features of the traditional feeling theory of emotion and the judgmentalist theories that have dominated since the 1960s. Perceptual states have intrinsic phenomenology—the fact that it feels like something to experience emotion does not need to be added as an afterthought. But perceptual states are also intentional—they represent the world as being a certain way. As such they can be accurate or inaccurate, appropriate or inappropriate, the feature which attracts philosophers to judgmentalist theories of emotion. The idea that emotions are perceptions of evaluative properties has additional attractions for some moral psychologists. It suggests that emotions may be a source of evaluative knowledge. Michael Tye is one author who argues that evaluative qualities are directly given to us in our perceptual experience—we see that something is dangerous or disgusting (Tye, 2008).

The idea that emotions are perceptions of evaluative properties can be used to understand how emotions motivate actions. I described before DeLancey’s proposal that basic emotions can explain action directly, rather than by combining with a desire to obtain some goal. A basic emotion both represents the world as being a particular way and embodies an action tendency. Sabine Döring has developed a theory which is structurally similar, although in a very different philosophical tradition (Döring, 2003, 2007). Her concern is whether emotion can rationally explain action without the assistance of belief or desire. Can someone who has the emotion but does not perform the action be considered to have acted irrationally merely because of the intentional content of that emotion? Döring suggests that emotions can play this role, because as well as motivating a particular kind of action, the emotion represents the world as having a property in the light of which that action is rational.

In a later article Döring builds on these ideas to address a key debate in moral psychology. Externalist accounts of moral motivation hold that an agent who judges that an action is morally right will not be motivated to act unless they desire to do the right thing. Internalist accounts hold that someone who judges that an action is morally right thereby has a motive to act: moral judgments are intrinsically motivating. Döring uses a perceptual theory of emotion to uphold internalism:

Emotions are affective perceptions. They have motivational force so that they can contribute to the explanation of action. At the same time they can rationalise actions because they have an intentional content which resembles the content of sensory perception in being representational . . . Doing the right thing is much more a matter of seeing things right than of drawing the right inferences. Seeing things right, in its turn, is not only to justify an action: it necessarily implies to be motivated to act accordingly. (Döring, 2007, p. 363)

Emotions are being asked to do an extraordinary amount of work in this form of moral psychology. They are both a key element in moral epistemology—perceptions of evaluative properties, and a key element in moral motivation—linking evaluative properties to action. To my mind this makes it even more important to show that there actually are mental processes that play these roles, and that we are not merely projecting the desiderata of our ethical theories onto the emotions.

Colombetti has defended a very different approach to emotional perception (Colombetti, 2007). “Enactive” theories of perception argue that organisms perceive by acting in the world, and that perceptual skills are in part motor skills (Hurley, 1998; Noë, 2004). Applying this theory to emotional appraisal leads Colombetti to suggest that bodily arousal and action are means by which the organism appraises emotional significance. For example, becoming aroused and orienting to the stimulus are part of the process by which the subject determines that they are confronting danger. At the heart of Colombetti’s work is a dynamical systems approach to cognitive science: the mind is the emergent pattern of activity of the brain and the body. The best way to conceptualize emotional episodes is as self-organizing, dynamical patterns of the whole organism (Colombetti, 2009a, in press). There are some similarities between the enactive approach and Prinz’s theory of “gut reactions,” and both theorists talk of “embodied emotions” (Prinz, 2004a). But their understanding of embodiment is very different. For example, Prinz follows Damasio in allowing that the brain can substitute for the absence of an actual bodily state through a neural “as-if loop.” In Colombetti’s view it would be extremely surprising if the brain, uncoupled from the body, could recreate the dynamics of the whole system. Prinz’s proposal that appraisal is carried out by “calibration files” of learnt associations is also at odds with the enactive perspective, according to which appraisal is not a separate process which occurs prior to other aspects of becoming emotional.

As well as being embodied, emotion for Colombetti is “embedded” because the brain–body system is in continual commerce with the environment as an emotional episode unfolds. The course of the emotion depends on this exchange. I am not sure how far Colombetti would endorse the “situated emotion” perspective, according to which emotional competence—the ability to have the right emotion at the right time—depends on scaffolding by the physical and social environment (Griffiths & Scarantino, 2009; Parkinson, Fischer, & Manstead, 2005).

Colombetti’s work introduces to the philosophy of emotion the ideas and issues that have dominated the philosophy of cognitive science for the past decade—the so-called “three Es” of embodiment, embeddedness, and enaction. This is likely to set a new direction for the field (see also Greenwood, 2012). Philosophers have found much work unpacking what these ideas really mean and where they can plausibly be applied to cognition, and their application to emotion will raise the same questions.

Conclusion

Despite the at times heated methodological disagreement between naturalistic philosophers of emotion and other philosophers of emotion, the central concerns of their work remain quite similar. For example, the integration of the purely philosophical literature on emotion-as-perception with naturalistic approaches to emotional perception is the obvious way to meet the concern expressed above that there may be no actual psychological processes that play the role moral psychologists look to the emotions to play. Conversely, such integration will help ensure that naturalistic work continues to address important philosophical questions.