Nonconceptual content: Cognitive or noncognitive?
The previous discussion has shown that neither account on the nature of emotional content, conceptual or nonconceptual, is capable of accommodating all human emotions at the algorithmic level of analysis. Some emotions have conceptual content whereas others have nonconceptual content, but the divide is not sharp because most actual emotions incorporate representations of more than one type. This result brings us back to the first question on whether particular types of non-conceptual content are cognitive or noncognitive. Everyone agrees that subdoxastic and subpersonal nonconceptual content is noncognitive. Therefore, the question is relevant only for the inferential and phenomenological types of nonconceptual content. Advocates of these types of content usually regard noninferentially structured or phenomenologically fine-grained content as noncognitive (see e.g. Griffiths, 2004; Gunther, 2003; Deigh, 2008). An exception is Döring (2004), who suggests that affective perceptions qualify as cognitive by virtue of having the hallmarks of cognitive content: propositional structure, the mind-to-world direction of fit, and a correctness condition. This proposal calls for more elaborate examination.
Dörings argument for the cognitive character of nonconceptual content is founded on the idea that emotional content is structured by nonconceptual gestalt qualities such as fearsome (fear) or outrageous (anger). She invokes a distinction between two kinds of propositional content: propositional in the Fregean sense, or propositional F' in short, and 'propositional S' in which the subscripted 'S' stands for sentential or conceptual content. For the former kind of propositional content, a mental state must only ascribe a property to the particular object that it represents. Both perceptions and emotions qualify as propositional F as they ascribe certain sensory properties such as colours, shapes, and affective gestalt qualities to their particular objects. The content of perceptions and emotions need not be conceptually available to the subject unlike the content of beliefs and evaluative judgments that is propositional in both senses of the term. In essence, propositional content in the Fregean sense boils down to a predicative structure that Döring associates with the mind-to-world direction of fit and a correctness condition. However, these epistemic notions are problematic in the context of perceptions of affective gestalt qualities.
It is true that emotions ascribe such properties as fearsome, disgusting, and outrageous to their particular objects. Yet in order to have the mind-to-world direction of fit, these ascriptions should not be fitting in every token emotion. Unfortunately, this seems to be the case because there is a difference between affective gestalt properties and ordinary perceptual properties as the former properties are projective rather than merely response-dependent. Typical perceptual properties such as colours and sounds are secondary properties that can be defined through their appearance to normal subjects in normal conditions of perception. To be, for instance, green is to appear green to normal subjects in normal conditions, where normality refers to capacities of visual perception, health, freedom from the influence of substances, and so on, while normal conditions include bright daylight, short distance, and so on. In contrast, affective properties are "tertiary" as their perception depends on the perception of primary and secondary properties such as a pit-bull terrier's fearsome-making properties. More importantly, affective properties are response-dependent in a deeper sense than secondary properties because any object that gives rise to an emotional response receives the relevant tertiary property in emotional experience. Thus, everything that appears fearsome is fearsome, whereas everything that appears green is not green, because green is not an entirely phenomenal property but depends both on the subject and her conditions of perception. However, without a meaningful is/seems distinction there is no correctness condition either because this condition is such a distinction.
Emotion theorists have recognized this problem in discussing the formal objects of emotion. In order to qualify as a standard of fittingness, a formal object cannot be a property that every token emotion of the same type ascribes to its particular object (McDowell, 1997b, p. 207; Prinz, 2004, pp. 60-64; de Sousa, 2011, p. 64). Fear, for instance, is fitting only if its object merits fear by being dangerous; not merely frightening or fearsome or scary. Therefore, the formal object of fear is the property of being dangerous rather than the property of being frightening. Unfortunately, language does not allow us to make the same kind of distinction between evaluative and phenomenal properties in the context of all emotions. The concepts of being enviable and being shameful, for instance, can be used in both an evaluative and a phenomenal sense. Thus, we describe as shameful both acts that we actually feel ashamed of and acts that we or other people should feel ashamed of, whether or not we do so. Yet these properties qualify as formal objects of envy and shame only if they are interpreted in the evaluative sense that refers to fitting envy and shame, respectively. Döring's affective gestalt properties, in contrast, are those phenomenal properties that figure in all emotional experiences of the same type - otherwise they could not phenomenally structure those experiences. Yet the phenomenal role of affective gestalt properties undermines their role as formal objects because no plausible is/seems distinction can be grounded on such properties that trivially fit their particular objects.
Therefore, I conclude that nonconceptual content is noncognitive. Attempts to expand cognition to cover both conceptual and nonconceptual representations blur important distinctions between different kinds of representations involved in emotion. Delicate distinctions such as Deigh's divide between primitive and tutored emotions rather than sweeping generalizations are needed in order to understand how the emotions of pre-lingual infants differ from those of adult humans. Even so, a structural approach to emotional content can only lay out the elements of which human emotions are composed. It is another thing how these elements enmesh and interact in actual human emotions. Therefore, a structural approach must be complemented with a dynamic approach to the content of emotions.
In this chapter I have discussed the role of cognition in the content of human emotions from a structural point of view. I have argued that cognition is always involved in emotional content at the functional level of analysis, whereas emotional content can be divided schematically into two kinds, cognitive and noncognitive, at the algorithmic level of processes and types of representation. My discussion focused on the analysis of cognitive and noncognitive content in terms of conceptual and nonconceptual content; on subdoxastic, logical, and phenomenological subtypes of the latter; and on the possibility of analysing the content of all or most human emotions in terms of any single type of representation. Unsurprisingly, no single type of content was found to fit the bill: some human emotions involve representations with conceptual content whatever other components they involve; while others involve representations with nonconceptual content in the subdoxastic, noninferential, or phenomenological sense, or several of them. Nevertheless, instances of pure types were found to be rare as most human emotions have contents that mix properties of conceptual and nonconceptual content, being more or less conceptual or nonconceptual on a continuum from strongly conceptual to nonconceptual.
This result shows first of all that a quest for a single type of emotional content is misdirected. Human emotions are neither cognitive nor noncognitive, categorically, but typically have aspects of both to varying degrees. This conclusion also speaks against a sharp dichotomy between radically dissimilar types of emotion such as "pure" affect programme emotions and cognitively complex emotions. These types are idealizations that actual human emotions resemble to a greater and lesser extent. Secondly, the blended or mixed character of emotional content speaks against parallel-competitive views of dual processing in the emotional domain. Emotional processing is parallel but the two types of processing, implicit and explicit, typically do not compete with each other - rationally recalcitrant emotions are an exception - leaving the determination of response to just one type of processing. Instead, emotional processing is sequential, especially when the emerging emotion is about a present state of affairs whose unfolding the subject perceives with all her senses. Implicit nonconceptual processing launches the initial stages of responding on the basis of immediate perceptual impressions, but explicit conceptual processing immediately joins in to refine the appraisal of the situation and to monitor the adaptiveness or fittingness of the emerging response. The question is whether such a role of explicit processing renders even implicitly elicited human emotions cognitive in a robust sense at the algorithmic level of analysis. The following chapter about the role of cognition in dynamics of emotion is organized around this question.
-  Actually this is not the case because the content of an emotional experience depends on the subject's focus of attention during the experience. An intentional affective perception emerges only if the subject's attention is focused on the particular object of emotion. Focus on the unfolding bodily response gives rise to a different kind of emotional experience. See e.g. Lambie & Marcel, 2002; Salmela, 2002, 2005; Deonna & Teroni, 2012, Ch. 7.