The present proposal raises a number of additional questions and objections. Let us start with the most radical objection.
Granting that feelings, affective or not, can be pure “physical effects of objects on the nerves”, in William James’ terms (1890, vol. 2, p. 458), they do not need to have any genuine representational value. James invites us to take the case either of purely somatic feelings or of objectless emotions when they are generated by a pathological condition—such as the precordial catch syndrome (PCS) which is a feeling of pain in the chest that usually goes away without treatment, but can lead the victim to think he or she is suffering a heart attack. In this case, the emotional experience of dread, James says, is “nothing but the feeling of a bodily state, and it has a purely bodily cause” (1890, vol. 2, p. 459). From this, one might conclude that a feeling is a merely peripheral phenomenon: it does not have a function to represent, nor does it express anything in particular. What can be said, in response, is, first, that feelings have a crucial evaluative function, which they perform thanks to their expressive structure. In PCS, the patient’s experience of dread has valence and intensity, expressed through sudden breathlessness, chest constriction, blurred vision, tingling sensations in the skin, an elevated heart beat, and a disposition to crouch. These feelings are not only a matter of sensory “peripheral” experience: they are also used by the patient to collect her existing Bayesian correlations, and to monitor with their help the present affordance expressed. A second illustration of the representational nature of feelings is that they can arise in the absence of the sensory basis they seem to have. For example, illusory feelings of being touched—a reactive somatosensory feeling about a change occurring on one’s body surface—can be created by manipulating the coherence of the intermodal inputs from vision, touch, and proprioception. In the so-called “rubber-hand illusion”, participants feel that their hand is being touched with a paint brush, when in fact it is an artificial hand, not theirs, that they see being touched. They also, after a while, “feel as if their (real) hand is turning ‘rubbery’” (see Botvinick & Cohen 1998). This experiment is evidence that feelings are informational states, which monitor inputs, and, in extreme cases like this, cause the brain to try to reconcile contradictory multimodal input. In the proposed interpretation, however, seeing one’s hand being touched is a reactive feeling, while actively touching an object generates a percept—which plays quite a different role in cognition.
Second, speaking of “subjective resonance” to an affordance (see the discussion of how a feeling “resonates” to an affordance in section 2 above) may look improperly metaphorical.[20] This is meant, however, to mark the difference between feeling and perceiving. While percepts allow recognition and identification of external objects and properties, feelings express specific affordances in a perceived, imagined, or remembered situation. For example, one can feel cold right now, or simulate being cold when planning a polar trip; one can remember how angry, or bored one was in a given episode and context. Feelings give agents prompt access to the relevant features of a new situation through sensed changes in their experience. Importantly, resonance is also an apt term for empathy, i.e., for the propagation of feelings from an agent to an onlooker, based on expressive behavior (Decety & Meyer 2008; Dezecache et al. 2013). Brain imagery suggests that the perception of pain in another individual largely overlaps with the regions activated when experiencing pain oneself (Jackson et al. 2005). Such empathy, in the present proposal, exemplifies how a feeling structure can be communicated through a set of congruent behavioral cues associated with a given affordance (here a painful stimulus), with a valence and intensity that are bodily conveyed.
Third, one might object that a common feature of feelings and percepts is that they include non-conceptual contents. This is true; but notice the difference between the two types of non-conceptual content: while non-conceptual ingredients in perception are related to objective, external contrastive cues such as shapes, edges, colors, volumes, and auditory patterns, which can be static or dynamic, but are always purely descriptive, non-conceptual contents in feelings only include evaluative states, which combine the general type of the affordance, its valence, its intensity, the proper action program, where all constituents are “bodily marked”, i.e., expressed through somatic markers. Therefore we cannot say that feelings “perceive” affordances, for this would suppose either that feelings have direct sensory access to the world—which they don’t, for they extract their inputs from sensory perception—or that they have direct sensory access to the body, which they don’t have either—feelings are the subjective counterpart of bodily changes. Therefore we cannot say that agents "perceive affordances" when they experience a feeling, for this would suppose either that feelings have a direct sensory access to the world, which they don’t, for they extract their inputs from sensory perception, or that they have direct sensory access to the body, which they don’t have either. Feelings are the subjective counterpart of bodily changes.
Neuroscientific research about the role of emotion in perception offers evidence in favor of this view. An affordance is made immediately salient by the system’s ability to sensitively react to a (half-)perceived element in a given known context.[21] We speak of “half-perception” on the basis of what is known about the timing of object perception. Affordance predictions are made only milliseconds after visual sensations register on the retina, i.e., before the categorisation of perceived objects is completed (Barrett & Bar 2009). The orbitofrontal cortex (OFC; involved in emotion and reward in decision making, thanks to projections from the thalamus) is able to extract an affordance in the first 80ms of the visual process, merely on the basis of low spatial frequency and magnocellular visual input (Lamme & Roelfsema 2000). What happens to perceptual access when a perceiver cannot extract affordances? Barrett & Bar (2009) have shown that the lack of emotional reactivity in early perception impairs object categorization. A patient who accidentally lost his visual ability when three years old received in adulthood a corneal transplant. In spite of his recovered ability to extract visual information from the world, this perceiver had trouble categorizing what he saw. The authors’ suggestion is that reconstituting the internal affective context associated with past exposures to an object (which was lacking in this particular case) is “one component of the prediction that helps a person see the object in the first place” (Barrett & Bar 2009, p. 1325).
In summary: the medial OFC uses early low-level visual output to match the affordance associated with it in past experience of the object: somatic markers are thereby activated, and the appropriate action is prepared. A FS enables an object to be more swiftly categorized at higher perceptual levels. This evidence suggests that affordances are extracted from perception, but that feelings are not themselves perceived.[22] On the contrary, they offer a separate kind of feedback to cognitive perceptual processes.
Let us turn now to one of the most central questions that our proposal raises. How does it explain the respective roles, in expressive intentional content, of somatic markers, on the one hand, and of the represented formal objects on the other? Cognitive theorists take emotions to represent both salient aspects of the agents’ own bodily changes and an evaluative belief about an external fact, with, possibly, a causal relation between this fact and the experienced bodily change (see Gordon 1987; Tye 2008 and Solomon 2007). For example, when perceiving a bear in the near vicinity, one’s experience is taken to be about a complex of subjective bodily impressions (a pounding heart, trembling legs, etc.) and about the perception of a bear as being the cause of these changes. Such a construal of the intentional content of feelings only makes sense within a propositional mode of thought. Can our expressive mode reflect or approximate the information contained in this complex causal structure?
Clearly, FS does not explicitly convey a causal relation between situation, somatic markers and subjective feeling. It carries this causal relation implicitly, however, as a consequence of the control architecture that produces feelings. In an emotional control loop, a perceived affordance causes (rather than being represented as causing) its expressive evaluation through its specialized sensory feedback. Emotional awareness expresses this functional relation. An external event (made accessible through a perceived affordance, as detailed above) is immediately followed by subjectively experienced somatic cues of a given intensity and valence. In functional terms, this sequence makes sense in the following way. When an associated forward model has been selected (often automatically, on the basis of an environmental, somatic, or cognitive affordance), the associated sensory cues (the somatic markers in this particular episode) are automatically activated in order to monitor how this affordance is to be processed and reacted to. As has been shown elsewhere, monitoring implicitly carries information about the command (or the affordance) that is being monitored (see Proust 2013). This explanation is particularly detailed and convincing in the case of motor representations of action; the feelings of agency that result from the comparators associated with a given feedforward model express (among others) whether the emoter is, or is not, the author of the action currently attended to (see Wolpert et al. 2001 and Pacherie 2008). The present proposal generalizes the functional significance of feelings throughout their diverse types (reviewed in section 1). As the outcome of sensory comparators, feelings always carry a structured information set about the type of affordance they contribute to regulating, about its amount, and about which actions are appropriate. This information, in its own expressive mode, functionally approximates a causal relation that is, when propositionally expressed, represented as a relation between an internal state, an external cause, and a disposition to act.
In summary: Feelings do not gain their aboutness through a propositional thought where the contrast between object and property is semantically marked; they gain their functional (rather than propositional) aboutness (f-aboutness) through the respective roles, in adaptive control, of the selection of an affordance-dependent control model and of the markers that allow comparisons of valence and intensity to be expressed.
This account, however, fails to explain observed variability in the production of feelings and the interpretation of what feelings are “about”. There are cases where agents misattribute their sadness, their anger, or their happiness to an event that is either not real, or that actually played no role in feeling production. How can such a misattribution be explained on the present proposal? Our first attempt to address this question is based on the subjective grounding of affordances. “Feeling f” normally means that an affordance is sensed, expressed, and subjectively represented as present. This does not mean that the affordance has an objective counterpart. Thus a thirsty traveller can be delighted or relieved when subjected to a water mirage. It is no problem for this view, then, if an event does not have the action potential for a given affordance it is expressed as having.
A trickier problem for the proposal is that a person might feel an f-feeling while she thinks that she has a g-feeling. Is such a situation even possible? To deal with this question, we must first clarify what “transparency” means when applied to feelings. A mental state is transparent if, when it is activated, its intentional content is accessible to the subject who entertains it, while its vehicle properties are not. On the view defended above, feelings are transparent, because their somatic markers are felt in connection with a certain affordance, and because their valence and intensity directly influence the emoter’s motivation to act in a given way. Such transparency, however, does not need to entail the subject’s ability to verbally report the content of her feeling. First, as seen above, a feeling can be felt by a nonhuman or by a child, both of whom lack the requisite verbal and conceptual capacities. Second, even an agent endowed with language can express through somatic markers a feeling with a distinctive FS content while failing to accurately report, in conceptual terms, what her feeling is “about”. We saw that [aboutness], i.e., reference to an independent event or object, is not a concept that belongs to FS. When subjects try to infer [aboutness] from their experience, their propositional system of representation (PS) is solicited. Because the latter has an analytic rather than an evaluative function, additional constraints step in. While nonconceptual, intensive (analog) and value considerations and norms regulate FS, conceptual, digital, and instrumental considerations and norms regulate PS.[23]
Hence, when having to report about her feelings, a subject needs to translate one mode of representation into another, with no guarantee that this translation will not enrich or modify FS intentional content. First, she may no longer have access to the rich diversity of her FS experience, because her attention is no longer directed toward the relevant contextually-activated affordance. Second, she has to monitor other goals and their corresponding (social, instrumental, or epistemic) norms. For example, she needs to present her feelings to herself and to others in a socially acceptable way, and to try to justify them rationally. This in turn will depend on her existing background beliefs, on her self-concept, on her capacity for making self-attributions of this particular kind, and on her willingness to perform this kind of introspective report. A number of experiments and novels have documented the wide gap between people’s feeling experiences and the verbal report they provide, or the reasons they offer, for having this or that feeling. These considerations suggest, then, that the issue of transparency cannot be adjudicated independently of one’s viewpoint about mental architecture.[24] According to the present proposal, an affordance is first subjectively recognized through the resonance it produces—through its specific feeling, rather than through a concept-based interpretation.
Let us now return to our earlier question. Can a person actually feel an f-like feeling, and mistake this f-feeling for a g-feeling? According to the present account, this situation would presuppose that an f-feeling, as it occurs in the expressive mode, is misdescribed in a verbal report as a g-feeling, to finally be genuinely felt to be g. On this view, a change in representational form would not only make it possible to reinterpret the initial experience in terms of a different one, but also to feel differently. To see whether this case is plausible, it is worth discussing Schachter and Singer’s (1962) adrenaline experiment.