The theoretical landscape in emotion theory has not moved much in recent times. There is thus particular reason to welcome the new ‘attitudinal theory’ defended by Julien Deonna and Fabrice Teroni (Reference Deonna and Teroni2012; Reference Deonna, Teroni, Roeser and Todd2014; Reference Deonna and Teroni2015), a philosophically sophisticated account of emotions, which promises to improve on all the existing accounts. In this paper, however, we shall argue that Deonna and Teroni’s theory faces two serious problems, which significantly reduce its appeal. The first is that the argument they offer to establish the central tenet of their theory, namely, that the formal objects of emotions do not feature in the content of emotions, is flawed. The second is that the attitudinal theory itself is vulnerable to a dilemma. In a nutshell, either the notion of correctness at work in the theory is understood in terms of correspondence, and the attitudinal account is likely to collapse into a perceptual theory; or the notion of correctness is understood in normative terms such as ‘reasons’ or ‘ought’, and the attitudinal theory cannot motivate another of its main tenets, namely, that the formal object determines a mental state’s correctness conditions. As the two main contenders are arguably the perceptual theory and the attitudinal theory, this is good news for the perceptual theory of emotions.
Our plan is as follows. In Section 1, we start with a presentation of the contemporary landscape of emotion theories. In Section 2, we expound the main features of the attitudinal theory. In Section 3, we debunk their main argument in its favor, while in Section 4 we spell out a dilemma that the theory faces.
1. Sketching the landscape
Theories of emotions generally focus on occurrent episodes of emotions, such as the fear you experience when a bear attacks you, as opposed to the disposition to feel fear when confronted with bears. Such theories can be divided into three broad families: feeling theories, conative theories, and cognitive theories.Footnote 1 According to feeling theories, emotions are states that are characterized by their phenomenal properties, but which lack representational content (James Reference James1884; Lange Reference Lange, Lange and James1885; Whiting Reference Whiting2011). Conative theories focus on the motivational power of emotions and claim that emotions are, in essence, motivational states, such as action-tendencies or desires (Frijda Reference Frijda1986; Scarantino Reference Scarantino, Jacobson and D’Arms2014). Finally, cognitive theories hold that emotions are partly or wholly constituted by cognitive states. There are different ways to spell out this idea, depending on the kinds of cognitive states that are considered. For instance, judgmental theories hold that emotions are, or necessarily involve, kinds of judgments, namely evaluative judgments (Solomon Reference Solomon1976; Nussbaum Reference Nussbaum2001), while so-called quasi-judgmental theories focus on states such as evaluative thoughts (Greenspan Reference Greenspan1988) or construals (Roberts Reference Roberts2003; Roberts Reference Roberts2013). According to these different theories, emotions require the possession of evaluative concepts. By contrast, the perceptual theory argues that emotions are a kind of perceptual experiences, i.e. perceptual experiences of evaluative properties, such as the fearsome or the admirable, where it is understood that there is no need for an individual to possess the concepts of the fearsome or the admirable to be able to feel the emotions of fear or admiration (de Sousa Reference De Sousa1987; Tappolet Reference Tappolet2000; Tappolet Reference Tappolet2016; Döring Reference Döring2003; inter alia).
Over the last few years, philosophers working on emotions have paid special attention to cognitive theories. These theories share four important elements. The first is that they take emotions to have intentional objects, i.e. to be about some items or states of affairs. We are indeed afraid of a bear, angry at our neighbors, and hope that it snows, and so on – something which the feeling theory has trouble acknowledging. The second is that they conceive of emotions as evaluative states. Emotions are not neutral representations of their objects, but they ‘evaluate’ them. For instance, an emotion such as the fear you feel toward a bear evaluates the bear, in one way or another, as being fearsome or dangerous. The third is that they take emotions to have evaluative contents: emotions represent their objects (either conceptually or non-conceptually) as possessing certain evaluative properties. Thus, an emotion like your fear toward a bear represents the bear (either conceptually or non-conceptually) as being fearsome. The fourth element is that emotions have correctness conditions.Footnote 2 More specifically, and by contrast with conative theories, cognitive theories recognize that emotions can be assessed as either correct or incorrect, depending on whether or not their objects possess the evaluative properties that they represent them to have. To illustrate this point, consider again fear: your fear toward a bear is correct if and only if the bear is really fearsome.
The perceptual theory has arguably become the most popular cognitive theory. This is so for a number of reasons. For a start, the perceptual theory improves on judgmental and quasi-judgmental theories in that it allows for the possibility that one can experience emotions without possessing evaluative concepts. By so doing, it avoids the counter-intuitive implication that young children and non-human animals cannot feel emotions. Furthermore, the perceptual theory explains some aspects of emotions that judgmental and quasi-judgmental theories have trouble explaining, for example the phenomenon of ‘recalcitrant’ emotions. Emotions sometimes conflict with evaluative judgments, in the same way as perceptual experiences sometimes conflict with perceptual judgments, as it happens, for instance, in the case of perceptual illusions such as the Müller-Lyer illusion. Yet, judgmental and quasi-judgmental theories seem to have no logical space to account for this phenomenon.
This brings us to the main argument in favor of the perceptual theory, namely, the argument that emotions are relevantly analogous to perceptual experiences.Footnote 3 First, like visual experiences, emotions may both be conscious as well as characterized by a rich phenomenology. Additionally, emotions and perceptual experiences are typically caused by things in our environment. A third characteristic shared by emotions and perceptual experiences is their relation to the will: neither is under its direct control. Even if it is possible to position oneself to see things in a certain way, one can hardly decide not to have the visual experience once one has opened one’s eyes. Emotions are not under the direct control of our will in much the same way. We cannot simply decide to feel angry – even though we doubtless have a certain indirect control over our emotions, as when we breathe through our nose to stave off a bout of anger.Footnote 4
The perceptual theory of emotions is plausible, but by no means uncontroversial. In fact, several authors have pointed out that there are also important differences between emotions and perceptual experiences.Footnote 5 Unlike paradigmatic perceptual experiences, e.g. color experiences, emotions do not depend on sensory organs. Instead, they depend on the information coming from other mental states, including sensory perceptions, beliefs, memories, and imaginings. This point is often expressed by saying that, contrary to sensory experiences, emotions have cognitive bases. Another difference is that emotions seem to admit of reasons, in the sense that one can have reasons to fear something or admire someone, while one cannot have reasons to have visual experiences. One would not, for example, ask of someone who sees a lemon as yellow, what her reasons are for seeing the lemon as yellow. Generally speaking, the question at issue is whether these differences imply that emotions cannot be understood as perceptual experiences. This is a question that depends, in the end, on what perceptions are taken to consist in. Arguably, on a sufficiently liberal conception of perception, which does not require inputs from a dedicated sensory organ, nothing prevents us from thinking that emotion is a kind of perceptual experience.Footnote 6
In light of the previous issues, however, some authors have chosen to explore alternative accounts of emotions. Deonna and Teroni’s attitudinal theory is perhaps the most promising of these accounts. In the next section, we illustrate the main characteristics of their approach.
2. The attitudinal theory
The attitudinal theory is an attempt to offer a fourth kind of theory, distinct from all those offered in the literature. The basic idea is to apply to emotions the distinction between attitude and content, as exemplified in the claim that believing that p, supposing that p and desiring that p are different attitudes toward the same content, p. Deonna and Teroni’s suggestion is that emotions form a specific family of attitudes, namely, evaluative attitudes, whose tokens may sometimes have the same contents. Deonna and Teroni agree, then, that emotions are evaluative states. However, by contrast with cognitive theories, they locate the evaluation at the level of the attitude, rather than at the level of the content, for in their view emotions do not represent their objects (either conceptually or non-conceptually) as possessing certain evaluative properties.Footnote 7 Thus, for example, to fear that it will snow is to have a particular evaluative attitude toward the content that it will snow, while to hope or to feel outraged that it will snow consists in different evaluative attitudes toward the same content, and so forth for each and every type of emotion.
Deonna and Teroni’s account is motivated by a particular diagnosis of the alleged failure of cognitive theories. As we have seen, one aspect that cognitive theories have in common is that they conceive of emotions as having evaluative contents. According to Deonna and Teroni, this is the main source of the problems that these theories face. For instance, the perceptual theory’s difficulty in accounting for the fact that emotions can be based on reasons derives, in their view, from the construction of such reasons as reasons for having a particular evaluative content. However, where ordinary perceptions are concerned, there exists no equivalent type of reason, i.e. no reason for having a particular perceptual content. If this is the case, then it appears that the way out of problems such as this must come from construing emotions not as attitudes toward an evaluative content, but as evaluative attitudes themselves, something which undoubtedly allows for reasons. This is precisely what their account sets to do.
According to Deonna and Teroni, ‘there are as many emotional attitudes as there are types of emotions.’ (Deonna and Teroni Reference Deonna and Teroni2015, 294; see also Deonna and Teroni Reference Deonna and Teroni2012, 78) The difference between types of emotional attitudes is grounded in differences in the evaluative properties in terms of which emotions ‘evaluate’ their objects. Evaluative properties still play a role in determining the correctness conditions of emotions. However, they do so via their connection to the attitude that each emotion consists in, rather than as part of the content of emotions. For example, an emotion like fear toward a bear, while not representing the bear as fearsome, has nonetheless fearsomeness in its correctness conditions because of the kind of attitude that fear consists in. Symmetrically, the evaluative attitudes that emotions consist in are individuated in terms of the evaluative properties to which one appeals when one assesses the correctness of the emotion.
To give some meat to their account, Deonna and Teroni make the further claim that emotions are ‘felt bodily attitudes’, i.e. ‘bodily experiences of being disposed or tending to act in differentiated ways vis-à-vis a given object or event’ (Deonna and Teroni Reference Deonna and Teroni2015, 302; see also 2012, chap. 7). Indeed, they claim ‘that emotions are evaluative attitudes because they are’ such bodily experiences. Even if the claim that emotions such as sadness, hope, relief, joy or admiration always involve the experience of a specific motivational tendency is far from obvious, it is clearly plausible in the case of many other types of emotions. Consider fear, for instance. According to Deonna and Teroni, if it is true that fear is correct on condition that its object is fearsome or dangerous, this is because to experience fear at something is to experience one’s body as being prepared to forestall the danger, be it by flight, attack, immobility, or something else. Note, however, that since bodily experiences concern bodily changes, and not objects that we confront in the world, the claim that emotions are felt bodily attitudes seems best understood as a claim about the phenomenology of emotions, as well as their motivational power, rather than as a claim regarding the intentionality of emotions.
The main differences and similarities between the perceptual theory and the attitudinal theory can be summarized as follows. Both kinds of theories agree that emotions are intentional states, in that they are directed toward (i.e. they are about) particular objects. They also agree that the intentional object of emotions is part of their content, in the sense that emotions represent their particular objects. Both the perceptual and the attitudinal theories take emotions to be evaluative states, in that they ‘evaluate’ their particular objects as possessing certain evaluative properties. The latter properties are typically said to constitute the formal objects of emotions, Footnote 8 where formal objects have two main features.Footnote 9 First, they determine the correctness conditions of emotions. Second, they individuate the type of emotion to which an emotion token belongs. Both the perceptual and the attitudinal theories take emotions to have formal objects in this sense. The main difference between the two, then, is that by contrast to the attitudinal theory, the perceptual theory holds that emotions represent their formal objects (i.e. that the formal object is part of the content of emotions). Thus, while both kinds of theories recognize that, e.g. to fear that it will snow is correct just if this event is fearsome, the perceptual theory claims this is the case in virtue of the evaluative content of that emotion, whereas the attitudinal theory claims that this is so in virtue of the evaluative attitude plus a content that is not evaluative, namely, that it will snow (see Deonna and Teroni Reference Deonna and Teroni2012, 77).Footnote 10
It is easy to see that the attitudinal theory shares the same advantages as the perceptual theory over feeling theories, conative theories, as well as judgmental and quasi-judgmental theories. Moreover, to the extent that it conceives of emotions as attitudes, rather than perceptual experiences, and that it denies that emotions have evaluative contents, the attitudinal theory also appears immune from the objections raised against the perceptual theory.Footnote 11 In particular, it seems to face no obstacles in recognizing that emotions, qua attitudes, have cognitive bases and can be grounded in reasons.Footnote 12
As we see things, the contest is pretty much one between the perceptual theory and the attitudinal theory, and to see who wins that contest a great many considerations need to be assessed. However, Deonna and Teroni offer a distinctive argument to show that their account is superior to the perceptual theory. In the next section, we shall argue that what we call their ‘master argument’ is unsuccessful.
3. Troubles with the master argument
The main claim to which Deonna and Teroni appeal in order to support their attitudinal theory is that, unlike the perceptual theory, it allows us to make sense of the intuitive idea that different emotions, as well as other mental states, can have the same content (Deonna and Teroni Reference Deonna and Teroni2012, 77; Deonna and Teroni Reference Deonna and Teroni2015, 297). As we have seen, the idea is that you can fear, hope, be outraged, etc. that it snows, but you can also believe, suppose, or desire that it snows, to mention only a few possibilities. Deonna and Teroni’s ‘master argument’ can be reconstructed as follows:
(1) For any two non-emotional mental state types X and Y that have a formal object, it is possible for two tokens x and y, where x is a token of X and y a token of Y, to have the same content.
(2) For any two non-emotional mental state types X and Y that have a formal object, if their formal objects featured in the content of their tokens, then it would not be possible for such tokens to have the same content.
(3) Therefore, for any two non-emotional mental state types X and Y that have a formal object, it is not the case that their formal objects feature in the content of their tokens.
(4) If (3) is true, then it is plausible to maintain that the formal objects of emotions do not feature in the content of their tokens either.
(5) Therefore, the formal objects of emotions do not feature in the content of their tokens.
If (5) is true, then emotions cannot be conceived of as having an evaluative content, as all cognitive theories maintain. One needs, therefore, to look to alternative theories. We can agree that the most promising non-cognitive theory currently on offer is the attitudinal theory. If so, Deonna and Teroni’s master argument provides support in favor of this view.
Deonna and Teroni’s master argument appeals to the alleged similarities between emotions and mental states such as beliefs, suppositions, desires, wishes, etc.; most notably, to the fact that they all have contents, which they seem able to share, and specific formal objects. These similarities make premise (4) compelling. It appears indeed that whatever is true about the formal objects of all non-emotional mental states must also be true of the formal objects of emotions. One might, of course, reject premise (4) by stipulating that emotions constitute a special class of mental states. However, this line of thought does not seem very plausible. So, in what follows we shall take premise (4) for granted. Even so the previous argument fails, or so we argue.
The main problem for Deonna and Teroni is that the conclusion stated in (3) is false. Consider chromatic perceptions. According to a plausible view, which can be traced back to Aquinas, color is the formal object of such perceptions.Footnote 13 However, color does feature in the content of chromatic perceptions. If so, we have a counter-example to the conclusion stated in (3).
This argument can be refined. Color is a determinable property, which admits of many determinate specifications. There are indeed various ways of being colored, e.g. being blue, being red, being yellow, etc.Footnote 14 If so, color figures in the content of chromatic perceptions only in the sense that one of its determinate specifications does. With that in mind, we can say that each of the determinate specifications of color, i.e. redness, yellowness, blueness, etc., constitutes the formal object of a specific type of visual perceptions. For instance, redness constitutes the formal object of perceptions of red; yellowness constitutes the formal object of perceptions of yellow, and so on. What is important for the present purpose is that these specific chromatic properties do feature in the content of their corresponding perceptions. For instance, redness features in the content of a perception of an object as red. But then, if each specific chromatic property is the formal object of the corresponding chromatic perception, it follows that the formal objects of all chromatic perceptions do feature in the content of their token instances.
The view that chromatic perceptions have formal objects is, of course, not uncontroversial.Footnote 15 Crucially, however, it is the view to which Deonna and Teroni’s own conception of formal object directly leads.Footnote 16 As we have seen with respect to emotions, Deonna and Teroni agree that the formal object of a mental state token is that which, in conjunction with its intentional object, determines whether that mental state is correct or not. Moreover, they agree that the formal object individuates the type of mental state to which that token belongs. These conditions are satisfied in the case of chromatic perceptions. Consider the claim that redness is the formal object of a perception of an object as red. First, redness is that which, in conjunction with the intentional object of a perception of red, determines whether that perception is correct or not. Indeed, the perception of an object as red is correct if and only if that object is really red. Second, redness is the property that individuates the type of perception in question, namely, a perception of red.
In light of this, we can in fact vindicate the initial claim that the determinable property of being colored is the formal object of the class of chromatic perceptions. Color is indeed the property that identifies certain perceptions as chromatic perceptions. Moreover, it figures in the correctness conditions of such perceptions via its determinate specifications.
It is worth noticing that the problem for Deonna and Teroni generalizes beyond chromatic perceptions. Consider auditory perceptions. In the same (qualified) way in which color can be considered the formal object of chromatic perceptions, so sound can be considered the formal object of auditory perceptions. But sound does feature in the content of such perceptions. The list may continue and include other kinds of perceptual experiences together with their formal objects, e.g. olfactory perceptions and smell. This shows that premise (3) is false: Deonna and Teroni’s universal claim about the formal objects of non-emotional mental states needs to be rejected. At least in the perceptual domain, being the formal object of a perceptual state does not exclude being part of the content of such state.Footnote 17
If these considerations are on the right track, then Deonna and Teroni’s master argument fails. Generally speaking, the problem with their argument is that it attempts to derive a conclusion about the formal object of all mental state types from premises that are, at most, sound only with respect to some mental state types. Perhaps, then, we should reformulate premise (1) as saying that the token instances of some mental state types (other than emotions) can have the same content. From this starting point, however, we can only derive the conclusion that the formal objects of some mental state types (i.e. those whose token instances can share the same content) do not feature in the content of their token instances. Clearly, this is a more modest conclusion than the one stated in (3). In particular, it leaves entirely open the question of whether emotions belong to the set of mental state types whose token instances can share the same content.
For their part, however, Deonna and Teroni explicitly maintain that the content of the token instances of different emotions can, sometimes, be the same. In ‘Emotions as Attitudes’, for instance, they claim that somebody’s aloofness can make one person angry but amuse another. According to them, ‘it is quite reasonable to say that [these two persons’ emotions] relate in different ways to one and the same thing’. Most importantly, Deonna and Teroni take this to be evidence that ‘different attitudes can have exactly the same content’ (Deonna and Teroni Reference Deonna and Teroni2015, 297; our italics).
This claim provides the basis for an alternative version of the master argument:
(1*) For any two emotional mental state types X and Y, it is possible for two tokens x and y, where x is a token of X and y a token of Y, to have the exact same content.Footnote 18
(2*) If the formal objects of X and Y featured in the content of their tokens, then it would not be possible for their tokens to have the exact same content.
(3*) Therefore, the formal objects of X and Y do not feature in the content of their tokens.
Unfortunately, this version of the master argument is no more successful than the previous. The main problem is that (1*) completely begs the question against the perceptual theory. Indeed, (1*) simply assumes that different emotions can share the exact same content with each other. However, this is precisely what the perceptual theory denies. The latter holds that emotions represent their formal objects. Hence, different emotions cannot have the exact same content. According to the perceptual theory, Deonna and Teroni’s remark that different emotions can relate to the same thing, e.g. somebody’s aloofness, shows, at most, that they can have the same intentional object, but not that they can have the exact same content. In other words, it shows only that different emotions can share part of their content, but not their entire content. Assuming otherwise from the start, without an argument, constitutes a petitio principii against the perceptual theory.
If this is correct, then Deonna and Teroni’s argument, as we understand it, is unsuccessful. There may be other ways of spelling out their argument. However, the ones we discussed seem to us the most plausible. The upshot is that in all likelihood Deonna and Teroni’s master argument fails to provide adequate support to their attitudinal theory.
4. The dilemma
At this stage, one might claim that, rather than refuting the master argument in favor of the attitudinal theory, the previous section has simply led us to an impasse. Surely – as we have shown – Deonna and Teroni cannot simply assume that different emotions can have the exact same content, on pain on begging the question against the perceptual theory. But one cannot presuppose the opposite either, on pain of begging the question against the attitudinal theory.
However, the case against Deonna and Teroni’s attitudinal theory can be strengthened further. As a starting point, note that Deonna and Teroni provide a second argument in favor of their theory. This is that the attitudinal theory offers a more convincing account of the correctness conditions of emotions than its competitors – one that highlights the respective contribution of the attitude and the content of emotions (see, e.g. Deonna and Teroni Reference Deonna and Teroni2015, 298–299).Footnote 19 Contrary to what Deonna and Teroni claim, however, we believe that their account of the correctness conditions of emotions is far from convincing, and that their theory performs significantly worse in this respect than the perceptual theory. In fact, we will now argue that, when it comes to the correctness conditions of emotions, Deonna and Teroni’s attitudinal theory faces a dilemma.
To begin with, let us grant that emotions are attitudes. By itself, this claim does not mark a fundamental difference between the attitudinal and the perceptual theory. The latter theory can maintain that emotions are sui generis attitudes, while simply adding that what makes them sui generis is that they are affective perceptual experiences concerned with evaluative properties. The main difference between the attitudinal and the perceptual theory concerns the relation between the formal object and the content of emotions. In light of what we have seen so far, we can say that Deonna and Teroni are committed to the following theses:
(i) For all attitude types having a formal object, the formal object is that by reference to which the token instances of that attitude type are assessed as correct or incorrect.
(ii) For all attitude types having a formal object, the formal object does not feature in the content of the token instances of that attitude type.
As we have seen, thesis (i) is commonly held by both attitudinal and perceptual theorists (as well as by all cognitivist theorists). By contrast, thesis (ii) is the thesis that demarcates the two theories. In this section, though, we argue that Deonna and Teroni cannot hold (i) and (ii) together, since they face a dilemma that obliges them to reject at least one of these two theses. If we are right, Deonna and Teroni must reject either a generally accepted thesis about formal objects or the very thesis that defines their attitudinal approach.
Our argument can be summarized as follows:
(1) Correctness is either conceived of as correspondence or in normative terms such as ‘reasons’ or ‘ought’.
(2) If correctness is conceived of as correspondence and if (i) is true, then the formal object of at least some attitude types (i.e. emotions) features in the content of their token instances. If so, (ii) is false.
(3) If correctness is conceived of in normative terms such as ‘reasons’ or ‘ought’, then either, whether or not (ii) is true, the formal object of at least some attitude types (i.e. beliefs and other cognitive states) is not that by reference to which their token instances are assessed as correct or incorrect, or the normative understanding of correctness reduces to the understanding in terms of correspondence. If so, either (i) or (ii) is false.
(4) Therefore, either (i) or (ii) is false.
Let us elaborate on each of these premises. Premise (1) states that there are two plausible ways of understanding an attitude’s correctness conditions. According to the first, to be correct is to correspond to what is the case.Footnote 20 According to the second, to be correct is to be such that there is sufficient reason to hold that attitude, or for that attitude to be required, or for it to be the attitude that one ought to have, etc.
4.1 The first horn of the dilemma
Let us consider the first understanding of correctness. If correctness is conceived of as correspondence, then an attitude’s correctness conditions are simply its correspondence conditions. This raises the question of what features of an attitude determine its correspondence conditions. One answer is that an attitude’s correspondence conditions are determined only by its content.Footnote 21 Accordingly, an attitude corresponds to what is the case if and only if its content corresponds to what is the case. To this, we can intuitively add that the content of an attitude corresponds to what is the case if and only if the world is as the content says it is. This leads us to the following statement of an attitude’s correctness conditions:
(C) An attitude is correct if and only if the world is as the content of that attitude says it is.
Let us combine this account of correctness with thesis (i). Given (C), (i) entails that the formal object of an attitude type can be seen as that by reference to which we can assess whether or not the world is as the content of an instance of that attitude type says it is. How exactly does the formal object play this role? In particular, does the formal object feature in the content of that attitude token? The answer to the latter question depends on the specifics of each attitude type. Consider emotions. As we have seen, generally speaking we can say that an emotion token is correct if and only if its object possesses the evaluative property that constitutes the formal object of the emotion type to which that token belongs. From (C), it then follows that the world is as the content of an emotion token says it is if and only if its object possesses the evaluative property that constitutes the formal object of the emotion type to which that token belongs. Once things are put this way, however, it appears evident that the formal object of an emotion features in the content of its token instances. Indeed, the most plausible way to make sense of the previous bi-conditional consists in saying that the world is as the content of an emotion token says it is if and only if (a) that emotion token represents its object as possessing the evaluative property that constitutes its formal object, and (b) its object does possess such an evaluative property.
Let us contrast emotions with beliefs. Generally speaking, we can say that a belief token is correct if and only if its propositional object is true, where truth is understood as the formal object of beliefs. From (C), it follows that the world is as the content of a belief token says it is if and only if its propositional object is true. Notice, however, that the following bi-conditional holds as well: the propositional object p of a belief token is true if and only if p (is the case). If so, it follows that the world is as the content of a belief token says it is if and only if p (is the case); or, more precisely, that the world is as the content of a belief token says it is if and only if (a) that belief represents its propositional object p and (b) p (is the case). This shows that the formal object of beliefs does not feature in the content of belief instances.
Let us take stock. We have argued that, if correctness is conceived of as correspondence, then the formal objects of at least some attitude types, i.e. emotions, feature in the content of their token instances. This is not the case for other attitude types, e.g. beliefs. This conclusion is precisely the perceptual theorist’s view of the matter. Thus, if the preceding argument is correct, it seems we have an argument that favors the perceptual theory over the attitudinal theory.
Deonna and Teroni might object that the line of thought explored so far relies on a mistaken assumption, namely, that the only feature of an attitude that is relevant for determining its correctness conditions is its content. According to them, however, the correctness conditions of an attitude token are jointly determined by the content of that attitude and by the attitude type to which that token belongs. How can their account be spelt out more precisely when correctness is conceived of as correspondence? We can begin by saying that, on this understanding, an attitude token is correct if and only if both its content (e.g. that it will snow) and the attitude itself (e.g. fear) correspond to what is the case. Let us combine this with thesis (i). The role of the formal object is now different from what we have seen before. On the current understanding of correctness, the formal object is that by reference to which we can assess whether or not the attitude itself, rather than the content of the attitude, corresponds to what is the case. In particular, Deonna and Teroni are committed to saying, on the one hand, that the content of an emotion token corresponds to what is the case if and only if (a) the emotion represents its object as possessing certain properties, and (b) the object does possess such properties; and, on the other hand, that the emotional attitude itself corresponds to what is the case if and only if (c) its object possesses the evaluative property that constitutes the formal object of the emotion type to which that token belongs. For example, the content of an episode of fear toward a bear attacking you corresponds to what is the case if and only if (a) the fear represents the bear as attacking you, and (b) the bear is indeed attacking you, whereas the attitude of fear itself corresponds to what is the case if and only if (c) the bear is fearsome. According to the resulting account, then, the formal object plays a role in the assessment of whether the correspondence conditions of an emotion token are met, but does not feature in its content.
We believe, however, that there is a strong reason to reject this account. The main problem is that the meaning of ‘correspondence’ involved in stating the content’s correspondence conditions is different from the meaning of ‘correspondence’ involved in stating the emotional attitude’s correspondence conditions. To illustrate this point, suppose we adopt a conception of correspondence as isomorphism. This conception can be straightforwardly applied to content correspondence. Within Deonna and Teroni’s account, the content of an emotion corresponds to what is the case if and only if there is an isomorphism between the object represented by the emotion and the object in the world. By contrast, this conception of correspondence is hardly applicable to emotional attitude correspondence. An emotional attitude token is indeed not isomorphic to an evaluative property, even if the latter is the property that defines the emotion type to which that token belongs. Thus, for instance, fear is not isomorphic to the fearsome. That its object is fearsome is, perhaps, what makes fear an emotion of the right kind in the circumstances, or the emotion that one ought to have in the circumstances. One prominent way to understand such claims that seems congenial to Deonna and Teroni’s account appeals to the notion of pragmatic reasons. The idea is that since the action-tendencies of fear have the function to help one deal with danger, this is the emotion that one ought to have when there is danger. In this, but only in this, sense, fear ‘corresponds’ to the fearsome. Importantly, this is clearly a normative sense of ‘correspondence’ that has nothing to do with the sense involved in describing content correspondence.
One may, of course, take issue with the conception of correspondence-as-isomorphism. But we have used this conception of correspondence only for illustrative purposes.Footnote 22 The previous considerations are simply meant to convey the intuition that the sense in which the content of an emotion ‘corresponds’ to the world is different from the sense in which the emotional attitude does. The problem is that the meanings of ‘content correspondence’ and ‘attitude correspondence’ must be the same, in order for them to be part of an unequivocal account of the emotions’ correspondence conditions. Put differently, Deonna and Teroni owe us an account of ‘correspondence’ that can be equally used to characterize the correspondence conditions of the emotional attitude and of its content. Moreover, they have to do so in a way that does not reduce the notion of ‘correspondence’ to one of the normative notions explored in the next section. Nothing said so far shows that it is impossible for Deonna and Teroni to provide such an account. However, the burden of the proof lies with them. Unless Deonna and Teroni take on this burden, we have prima facie reason to think that, if correctness is conceived of as correspondence and (i) is true, then the best account of emotions’ correspondence conditions is one in which the formal objects of emotions feature in the content of their token instances. Absent any better account, premise (2) of our argument is thus vindicated.
4.2 The second horn of the dilemma
Let us move to premise (3) of our argument and consider the normative understanding of correctness. Our claim is that, in this case, either, whether or not we endorse thesis (ii), thesis (i) does not hold or the normative understanding of correctness reduces to the understanding in terms of correspondence. To see why, consider an attitude type such as belief. As we have seen, it is common to think that the formal object of belief is truth. Thesis (i) implies, then, that a belief token is correct if and only if that belief is true. Suppose, now, that we conceive of correctness in terms of ‘sufficient reason’. The following claim should hold: there is sufficient reason to have a belief token if and only if that belief token is true. It is easy to see, however, that this claim does not hold – and this is the case for whatever kind of reason one considers. On the one hand, there may be sufficient epistemic reason to have a particular belief, which is actually false. For instance, this may happen when there is strong evidence in favor of a particular belief and, yet, this evidence is misleading, since this belief is false. On the other hand, and even more strikingly, there may be sufficient practical reason to believe something even if that belief is false, as evil demon scenarios amply show (see Rabinowicz and Rønnow-Rasmussen Reference Rabinowicz and Rønnow-Rasmussen2004, for instance).
One may suggest that this objection can be avoided if correctness is conceived of in terms of ‘ought’, rather than ‘sufficient reason’. According to this suggestion, an attitude token is correct if and only if one ought to have that attitude. In the case of belief, the following claim should then hold: one ought to have a belief token if and only if that belief token is true. Once again, however, this claim does not hold; and this is so whether we consider the epistemic or the practical ought. If the epistemic reasons are particularly strong, it may indeed be the case that we epistemically ought to have a false belief. Likewise, if the pragmatic reasons for a belief are particularly strong, it may be the case that we ought to have a belief that is not true.
As a final attempt, one might counter that correctness should be formulated in terms of an ‘ought’ that is both epistemic and objective. According to this view, what one ought to believe depends entirely on the facts that make one’s belief true, rather than on one’s subjective evidence or pragmatic interests. Once this formulation is adopted, the claim that one ought to have a belief token if and only if that belief token is true follows as a tautology. The relevant ‘ought’ in terms of which beliefs’ correctness is understood is defined in terms of the formal object of beliefs.
The problem with this solution is that it renders the normative understanding of correctness indistinguishable from, and in fact reducible to, the understanding of correctness in terms of correspondence. What one epistemically ought, objectively, to believe is no longer determined by one’s (epistemic) reasons, or how these (epistemic) reasons weigh against each other, but by a sort of direct ‘match’ between one’s belief and what is the case.Footnote 23 In the case of belief, such a ‘match’ is best conceived of in terms of the correspondence between the content of one’s belief and the world. What about other mental states, such as emotions? What does it mean to say that one epistemically ought, objectively, to have a particular emotion token, when the objective epistemic ought is completely detached from someone’s (epistemic) reasons to have that emotion? One possibility consists in conceiving the objective epistemic ought of emotions in terms of (content) correspondence, as we do for beliefs. In this case, however, we run into the problems explored in sub-section 4.1. If we adopt this option, it is indeed correct to say (e.g.) that the fearsome is what makes fear the emotion that one epistemically ought, objectively, to have. However, this is so because the emotion that one epistemically ought, objectively, to have is the emotion whose content corresponds to what is the case, i.e. in the present example, that its object is fearsome. In order for Deonna and Teroni’s theory to be viable, then, it must be the case that there exists another account of the objective epistemic ought of emotions, which neither reduces to (content) correspondence nor falls prey to the objections presented above. So far, Deonna and Teroni have not offered any such account. In fact, we are skeptical about the prospects of doing so. Be that as it may, the burden of the proof rests squarely with them.Footnote 24
The upshot is the following: if truth is the formal object of beliefs (as it is commonly maintained), then either (i) is false or the normative understanding of correctness reduces to the understanding in terms of (content) correspondence. Premise (3) is also vindicated.
The overall conclusion of our argument can be summarized as follows. On the one hand, if correctness is defined as correspondence, then (ii) is false. On the other hand, if correctness is defined in normative terms, then either (i) or (ii) is false. Therefore, either (i) or (ii) is false. The only way out is to argue that there is another concept of correctness at play. Once again, nothing said here shows that this cannot be done. Though given the very general specification of the two options we gave, the odds are against the attitudinal theory.
5. Conclusion
In this paper, we have argued that the master argument offered by Deonna and Teroni in favor of their attitudinal theory is flawed. We have also argued that the attitudinal theory itself faces what is likely to be a serious dilemma. Both lines of argument show that, despite its promise, the attitudinal theory is in trouble. This is good news for the other cognitive theories of emotions on the market. In particular, since the attitudinal theory is specifically presented as an improvement on the perceptual theory, and since the master argument for the attitudinal theory is in fact an objection against the perceptual theory, this is good news for the perceptual theory.
Acknowledgments
We are very grateful to Paul Boswell, Julien Deonna, Kevin Mulligan, Hichem Naar, Michele Palmira, Angie Pepper, Sarah Stroud, Fabrice Teroni, the participants to the 4th Conference of the European Philosophical Society for the Study of Emotions, and two anonymous referees for their useful comments on previous drafts or presentations of the paper.
Disclosure statement
No potential conflict of interest was reported by the authors.
Funding
This work was supported by the Social Sciences and Humanities Research Council of Canada [435-2016-0711];
