Proust 2014 The Representational Structure of Feelings

The Representational Structure of
Feelings
Joëlle Proust
The word “feeling” denotes a reactive, subjective experience with a distinctive em- Author
bodied phenomenal quality. Several types of feelings are usually distinguished,
such as bodily, agentive, affective, and metacognitive feelings. The hypothesis de-
Joëlle Proust
veloped in this article is that all feelings are represented in a specialized, non-
joelle.proust @ ehess.fr
conceptual “expressive” mode, whose function is evaluative and action-guiding.
Feelings, it is claimed, are conceptually impenetrable. Against a two-factor theory Ecole Normale Supérieure
of feelings, it is argued, in the cases of affective and metacognitive feelings, that Paris, France
background beliefs can circumvent feelings in gaining the control of action, but
cannot fully suppress them or their motivational potential. Commentator
Keywords Iuliia Pliushch

Affective feelings | Affordance | Agentive feelings | Appraisal | Arousal | Bodily pliushi @ students.uni-mainz.de
feelings | Comparator | Control | Cues | Evaluative | Expressive | Familiarity | Johannes Gutenberg-Unversität
Fluency | Formal object | Illusory feeling | Incidental feelings | Integral feelings | Mainz, Germany
Intensity | Metacognitive or noetic feelings | Monitoring | Nonconceptual con-
tent | Predictive | Reactive | Resonance | Retrospective | Somatic marker | Trans- Editors
parency | Two-factor account | Valence
Thomas Metzinger
metzinger @ uni-mainz.de
Johannes Gutenberg-Universität
Mainz, Germany
Jennifer M. Windt
jennifer.windt @ monash.edu
Monash University
Melbourne, Australia
1 Introduction
“Feeling” denotes a reactive, subjective experi- It rather refers to the feedback of one’s own
ence with a distinctive embodied phenomenal key-touching activity. This type of perceptual
quality and a formal object, which may or may feeling is expected to result from one’s action
not coincide with embodied experience. Feelings and, hence, does not belong to the domain of
typically express affect and valence in sensation. reactive feelings. What is called the “formal ob-
“Reactive” means that feelings are closely asso- ject” (see Kenny 1963) of a feeling is the prop-
ciated with an appraisal of a present property erty in the triggering event that elicits the re-
or event. The term “reactive” is crucial. The active feeling. For example, the formal object of
term “feeling” is sometimes used to refer to a fear is some threatening property detected in
non-reactive, perceptual experience. For ex- the perceptual field.
ample, when one perceives an object through Feelings can be pleasant or aversive,
touch, it is common to say that “one feels one’s strong or weak, short-lived or long-lasting, or
key in one’s pocket”. But “feeling”, in this con- have an arousing or depressing character. They
text, does not refer to a reactive phenomenon. motivate distinctive dispositions to act, whose
Proust, J. (2015). The Representational Structure of Feelings.
In T. Metzinger & J. M. Windt (Eds). Open MIND: 31(T). Frankfurt am Main: MIND Group. doi: 10.15502/9783958570047 1 | 25
www.open-mind.net
Glossary
Feeling “Feeling” denotes a reactive, subjective experience with a distinctive embodied

phenomenal quality and a formal object, which may or may not coincide with the
embodied experience. Feelings typically express affect and valence in sensation.
Reactive “Reactive” means that feelings are closely associated with an appraisal of a
present property or event.
Formal object “Formal object” of a feeling is the property in the triggering event that elicits the
reactive feeling.
Metacognitive feelings Metacognitive feeling are experienced while conducting a cognitive task: the agent
may find the task easy or difficult, anticipate her ability or inability to conduct it.
Once the task is completed, the agent may have the feeling of being right, or have
a feeling of uncertainty about the outcome of her endeavour.
Affordance Affordances are positive or negative opportunities, expressed in feelings: an af-
fordance-sensing swiftly and non-reflectively motivates the agent to act in a par-
ticular way.
FS Affordance FS Affordancea [Placea=here], [Timea=Now/soon], [Valencea=+], [Intensitya=.8
(comparatively specified on a scale 0 to 1)], [motivation to act of degreed accord-
ing to action programa].
Transparency A mental state is transparent if, when it is activated, its intentional content is ac-
cessible to the subject who entertains it.
Incidental and integral feeling Metacognive feelings are called “incidental” when they are not based on valid cues
about the cognitive task at hand, and hence, have no predictive value. They are
called “integral” when they actually carry information about cognitive outcome.
urgency is entailed both by the feeling experi- they are not affective. A feeling tends to be
ence and the context in which it is experienced: more explicitly felt as bodily when it has a
feeling an intense pain disposes the person to body-related function; that is, the phenomeno-
promptly locate and remove the cause of the logy makes the need to be served salient (feeling
pain; except, for example, when it is self-inflic- tired, feeling a pain in the joints) in order to
ted, or when it is part of a ritual. motivate action. In affective feelings, in con-
Most theorists of feelings agree that they trast, the bodily phenomenology tends to recede
are associated with—or, for those who identify to the fringe of consciousness (feeling in love
emotions with conscious experiences1 consist of with A, feeling angry with B).2 From this obser-
—specialized, internally generated bodily sensa- vation, it is easy to infer that types of feelings
tions, such as an increase in heart rate, contrac- differ in their respective meanings: they in some
tions or relaxations of the facial muscles, vis- sense express what they are about. In affective
ceral impressions, tremors or tears, impulses to feelings, an experience of “feeling toward” is
run away, etc. As will be seen below, some feel- supposedly present: the emotion is felt as being
ings, however, do not express emotions., i.e., about an object, a person, or a situation—the
objects, rather than bodily sensations, are the
1 From the viewpoint of the somatic feeling theory of emotions, emo-
tions can be explained as a somatic change caused by the perception,
focus of one’s emotional attention. Affective
real or simulated, of a particular object. See James (1884, p. 190), feelings also include mixed cases where one
and Damasio (1994, 2003). Other theorists of emotion, however, con- seems to both experience a strong bodily feeling
sider that the conscious experience of having an emotion includes
propositional attitudes, and not only feelings. See sections 4 and 5 at the same time as the intentional content that
below. Moods are long-term affective states, and will not concern us
here. 2 On this concept, see Mangan (1993, 2000) and Reber et al. (2002).

www.open-mind.net
this feeling seems to refer to, as when Marcel 2 Are feelings a natural kind?
Proust’s narrator reports experiencing an acute
pain in the chest when thinking about his be- Paul Griffiths has claimed that emotions do
loved deceased friend, Madame de Guermantes.3 not constitute a natural kind, in the sense that
It is unclear whether metacognitive (also called they do not form “a category about which we
noetic, or epistemic) feelings are affective or can make inductive scientific discoveries”
non-affective (see section 7 below). They are ex- (2004, pp. 901–911). One can agree with latter
perienced while conducting a cognitive task: the claim, however, without concluding that feel-
agent may find the task easy or difficult, and ings do not constitute a natural kind. First,
may anticipate her ability or inability to con- feelings are not only affective ingredients in
duct it. Once the task is completed, the agent emotional awareness. Some feelings, such as
may have the feeling of being right, or may have feeling cold or sick, or feeling that one is act-
a feeling of uncertainty about the outcome of ing, have nothing to do with affective episodes.
her endeavour. Take the case of a person who Second, there are evolutionary reasons to dis-
feels unable, presently, to remember what she tinguish, within emotions, two classes of sub-
had for dinner last night. Her feeling of not re- jective appraisals. Emotion theorists usually
membering is correlated with activity in a facial contrast feelings expressed in primary emotions
muscle, the corrugator supercilii (Stepper & —fear, anger, happiness, sadness, surprise, and
Strack 1993). Her feeling, however, is not about disgust—with various appraisals cum conative
her disposition to contract or relax this or that dispositions associated with higher cognitive
muscle, of which she is certainly unaware. It is, emotions, such as envy, guilt, pride, shame,
rather, about her present disposition to remem- loyalty, vengefulness, and regret. The first are
ber what she had for dinner. Epistemic feelings phylogenetically and ontogenetically prior to
seem to be “feeling-toward” experiences, and cognitions. They belong to the ancient limbic
have cognitive dispositions or contents as their system, which is present in some form in most
object. animals. A quick route from the retinal image
Descriptive phenomenology, however, does to the amygdala through the thalamus allows
not offer in itself an account of the intentional affective information to control behavior (see
structure of feelings. We need to understand LeDoux 1996). Primary feelings are thus
how feelings in general gain their real or sup- triggered independently of concept possession
posed aboutness, and how they relate to action- and motivate specific responses. Secondary af-
guidance as a function of context; i.e., we need fective experiences, in contrast, might have
to provide a functional analysis of feelings. Sec- evolved on the basis of social constraints in re-
tion 2 will begin to provide such an analysis, lation to cooperative action among humans. In-
and will address a preliminary issue—namely, deed (with the possible exception of pride and
Do the phenomena that are usually called “feel- shame) they are not present in nonhuman
ings” share a property that makes them a nat- primates.4 They activate newer brain struc-
ural kind? In section 3, the specific informa- tures; they require concept possession, depend
tional structure of feelings will be seen to ac- on background beliefs, and do not generate
count for their generic characteristics. Section 4 characteristic behaviors. Finally, primary feel-
will clarify the account by way of addressing ings are clearly embodied, while secondary
various objections. Section 5 will attempt to emotions seem to have no proprietary somatic
show that the proposed account fares better markers. An interesting idea, suggested by
with experimental evidence than a cognitivist Jesse Prinz (2004, p. 95), is that the facial or
account of affective and metacognitive feelings. somatic correlate of secondary emotions, when
Section 6 will examine whether or not metacog- they have one, involves a blend of the somatic
nitive feelings have an affective valence. markers for primary feelings.
4 On this contrast, see Frank (1988), Griffiths (1997), and Prinz (2004,
3 See the analysis of this example in Goldie (2002), p. 56. pp. 82-83). On whether they qualify as emotions, see Ekman (1992).

www.open-mind.net
In summary: emotions differ, among other trol loop; they carry non-conceptual information
things, because of the unequal role that feelings about how much one’s present condition devi-
have in the two classes of emotions just dis- ates from the expected condition. From a func-
cussed. The wider scope of feelings, when un- tional viewpoint, they form a natural kind inso-
derstood as “reactive, subjective experiences far as their function is to indicate a comparative
with a distinctive embodied phenomenal qual- outcome through a dedicated embodied experi-
ity”, seems to be more unified than emotions, ence.
and making feelings seem like plausible candid- Note, however, that there are comparators
ates for a natural kind. that trigger no feelings at all: these non-sensit-
We need, however, to turn this tentative ive comparators may either work outside con-
definition into a general functional characteriza- sciousness (for example, error signals driving
tion that presumably holds for all feelings (bey- immediate correction7, not to mention compar-
ond affective ones) and only for them. Here is a ators that work at the cell level), or they can
proposal: feelings constitute the sensitive part of take concepts as their input, rather than react-
predictive and retrospective processes of non- ing to percepts or situations (for example com-
conceptual evaluation of one’s own and others’ parators of currency or of educational value).
well-being and actions. Being essentially evalu- As far as feelings are concerned, they are
ative, feelings are always the output of a com- directly related to a presently-perceived context
parator: in other terms, they are crucial monit- (or an imagined or remembered context, but in
oring ingredients in self-regulated adaptive con- a “present-like”, indexical mode): one can feel
trol systems. In such systems, the specific func- too hot, too cold, or too tired (or feel “OK”,
tion of a feeling consists in detecting how much which usually means a tolerable deviation from
a current observed value of a parameter deviates the expected value). One can feel the fright one
from its expected value, on one or several dimen- has had, even after the frightening event has
sions relevant to survival (see Carver & Scheier ended. The outcome of a feeling-based ap-
2001). Their formal object, when they have praisal, from a functional viewpoint, has to con-
one,5 (such as being afraid of the bear in front sist in some disposition to act that is adaptive,
of me) cannot be analyzed independently of the relative to the input to which the feeling is a re-
monitoring function they serve within a special- action. Granting that feelings, as sensitive com-
ized control loop.6 Relevance to well-being, how- parators in a control system, form a natural
ever, extends to bodily condition, goal achieve- kind, there should be common properties cut-
ment, and availability of preferred goods of all ting across the various types listed above. In
kinds (food, partner, social status). The relev- fact we find three types of functional relations
ant dimensions of variation that feelings track between feelings of a given kind and the associ-
may accordingly be of a sensory, proprioceptive ated disposition to act. First, feelings, according
kind (feeling thirsty, cold, etc.), social-affective to their embodied valence, typically determine
(feeling angry), or agentive (goal-related). Goal actions of approach or of avoidance. Some dic-
achievement, however, involves either epistemic tate caution, others boldness. Some encourage
or instrumental success, respectively generating self-restraint, others self-assertion. Fear pro-
epistemic feelings (feeling interested, bored, epi- motes a flight tendency, hunger a tendency to
stemically uncertain) and agentive feelings (feel- approach food. Second, they have a specific ori-
ing of happiness, of agentive confidence, of own- entation in time: some feelings have a predictive
ership of one’s action, etc.). Feelings, in sum- function, and thus induce a behavior that is
mary, are the outcomes of comparators in a con- based on contingencies to be further displayed
in the present context. For example, fear, when
5 As observed by Goldie (2009), some feelings, for example, [feeling
anxious] or [feeling depressed], seem to lack a formal object, which is directed at a possible danger, increases the
typically the case with moods. As indicated above, moods will not be readiness to flee in case the danger concretizes.
discussed in this article.
6 Bechara et al. (2000) make it clear that the somatic marker theory
applies to action, whether it engages affects or not. 7 see Logan & Crump 2010 and Nieuwenhuis et al. 2001

www.open-mind.net
Others have a retrospective function, and induce Automaticity and informational encapsu-
corrections to the commands one has previously lation seem also to characterize agentive feel-
used, or to one’s previous preferences. For ex- ings (see Pacherie 2008). Feelings generated in
ample, feeling nauseous after food ingestion in- the course of a physical action come in two
duces food avoidance, i.e., a change in the varieties: generalized or specialized. Some, such
agent’s preferences. In contrast, feeling disgust as feelings of agency, of initiation of action, of
at the sight of some food may prevent the agent ownership and of motor control, are indicators
from approaching it. A subset of feelings, such monitoring action in progress: they concern
as feeling happy, have both temporal orienta- “who” is performing the action, and “how” the
tions. Third, according to their embodied dy- action is being conducted (see Proust 2000).
namics and intensity (which is called their Others concern the evaluation of an action in
“level of arousal”), feelings can provoke an elev- one’s own repertoire: a professional carpenter
ation in the energy available to the system: they or an experienced musician, for example, have
provoke excitement, agitation, power in the feelings telling them if an action sequence
coming response; or, on the contrary, they may (whether their own or another agent’s) in this
have a soothing effect and diminish the tend- repertoire sounds or looks right, even before
ency to act. they identify why they have this feeling. These
One major functional property of feelings, feelings are also the outputs of a comparison
from the viewpoint of information extraction between motor anticipations and observed
and use, is that they can very rapidly extract properties of the action (a “forward model of
and synthesize multiple cues from perception. action” supposedly stores the expected values
This rapidity is a consequence of the automatic of crucial parameters; Wolpert et al. 2001).
and encapsulated character of the control mech- They can predict the likelihood of completing
anism whose output they express. Feelings are an action (when the question arises, in difficult
automatically triggered by a specific type of in- or non-routine cases), or evaluate—on-line or
put (which is the definition of informational en- in retrospect—how swiftly, effortlessly, or un-
capsulation).8 Automaticity is associated with hesitatingly an action was performed. Agentive
feelings being inescapable, at least for those feelings thus have an essential role in regulat-
feelings that have been allowed to develop ing the fundamental properties of physical ac-
within a culture, granting normal development.9 tions, such as the quality of the outcome, 11 and
The mechanism that generates somatic, noetic, the ownership of the action. 12
or affective feelings from inputs (perceptual, Noetic feelings, finally, are functionally
imaginative, or memorial) does not require one similar to somatic, affective, and agentive feel-
to have specific beliefs or intentions. 10 Informa- ings—although their evolutionary pattern seems
tional encapsulation explains why transitive to be different from the other three kinds.
feelings persist when the agent finds out that While most organisms have proprioceptive, af-
the situation is different from what she thought fective, and motor control, and hence, presum-
to be the case. Just as an optical illusion such ably, somatic, affective, and agentive feelings,
as the Müller-Lyer effect does not immediately few are able to control their cognitive decisions
dissipate when it turns out that the segments through metacognitive feelings (see Beran et al.
are equal, a feeling of anger does not disappear 2012 and Proust 2013). The latter are generated
as soon as the agent realizes that its formal ob- when trying to perceive, to remember, or to
ject is not exemplified. plan a cognitive task (in particular, when trying
8 Automaticity in appraisal is central to Ekman’s analysis of to plan how long to study material in order to
primary emotions (1992). See also Griffiths (1997), Prinz (2004),
and Zajonc (1980). On informational encapsulation, see Fodor 11 Non-conscious error signals can also guide corrective steps, without
(1983). the agent noticing them.
9 For example, fearlessness in the presence of danger may result from a 12 Pat Haggard et al. (2002) have demonstrated the crucial role of the
disturbed childhood. temporal binding between felt initiation of action and output in the
10 Some affective feelings, however, can be intentionally controlled in sense of being the agent of an action. See, among other articles, Hag-
the long run, through cultural learning. See Murata et al. (2013). gard et al. (2002).

www.open-mind.net
master it).13 They are also relied upon when Taken together, these considerations are
trying to reason or to solve a problem; when compatible with the view that somatic, agent-
conversing, feelings of effort, and of informative- ive, metacognitive, and “primary” affective feel-
ness, are monitored by speakers and hearers in ings, even if they differ in their formal objects,
order to maintain a common level of relevance. form a natural kind. Our attempt above at a
Like other feelings, they have two distinctive functional characterization focused on the gen-
temporal orientations. Some have a predictive eral relations of feelings to inputs, outputs, and
function. A feeling of knowing (FOK) may arise mediating evaluative mechanisms. From this
when trying to remember an item—for example characterization, it emerges that feelings are
a proper name—that one has not yet retrieved: gradients in comparators that are felt subject-
having a strong FOK reliably predicts that one ively, rather than being propositional states de-
will finally retrieve the searched content (Koriat scribable in analytic, objective terms. These ob-
& Levy-Sadot 2001). A feeling of having a name servations, however, suggest that, in order to
on the tip of one’s tongue (TOT) both signals express a specialized and fine-tuned reactivity
the fact that a word is not presently available, to one or several formal objects, and to motiv-
and, according to its onset, valence, and intens- ate adapted behaviors, in order to be re-
ity, whether it is worth or not worth pursuing membered and conveyed to others feelings must
one’s effort to retrieve it (see Brown 1991 and have their own representational format. We now
Schwartz et al. 2000). Feelings of fluency are the turn to the following question: What is the
sense of ease of processing one may feel or fail structure of the information that is extracted
to feel when attempting to perceptually discrim- and expressed in a feeling?
inate objects with a given property, or to re-
trieve items from episodic or semantic memory. 3 What kind of information do feelings
A feeling of familiarity is particularly salient, in express?
human adults, when no further fact about the
target can be retrieved. It offers useful informa- The above question is important for clarifying
tion about the epistemic status of the target: the relation of feelings both to their formal ob-
that it is not new, but nevertheless not fully reject, when they have one, and to the action that
cognized. A feeling of familiarity, then, motiv- they motivate. In the case of metacognitive feel-
ates, among others, an attempt to recognize ings (M-feelings), the difficulty is particularly
what or who a target is. Other metacognitive pregnant: it stems from the fact that, if we
feelings have a retrospective function. When a grant that M-feelings do not require concept
name is retrieved, a feeling of rightness (FOR) possession to be felt, then it is unclear how
motivates the agent to consider her response the their formal object should be construed: What
expected one.14 Various feelings of uncertainty, are they about? Let us take a feeling of uncer-
based on fluency, coherence, plausibility, inform- tainty, felt while trying to remember a proper
ativeness, or relevance, also have retrospective name. Is this feeling about a memory state, or
functions: their valence and intensity tell the about a disposition to retrieve a proper name?
agent whether she should accept or reject a cog- If a feeling is about a memorial state or a dis-
nitive outcome. These parameters are expressed position, its intentional content needs to include
through specialized somatic markers, such as in- concepts of memory, of correctness, and of un-
creased activity in the facial muscle involved in certainty. Empirical evidence, however, demon-
smiling, the zygomaticus major—for positive strates that animals with no mindreading abil-
valence—or the corrugator supercilli (involved ity, and hence that are deprived of concepts of
in frowning)—for negative valence (Winkielman perception or of memory, are able to monitor
& Cacioppo 2001). their perception and memory as reliably as hu-
mans do.15 Furthermore, human children, from
13 This prediction involves judgments of learning (JOL). See Koriat &
Ackerman (2010). 15 Rhesus monkeys have been found to opt out of more or less challen-
14 On FORs, see Thompson et al. (2011). ging perceptual or memory trials as a result of trial difficulty. For a

www.open-mind.net
early on, are sensitive to the contrast between motivates the agent to act in a particular way.
familiar and unfamiliar faces and environments. Departing somewhat from Gibson’s use of this
This supports our claim above: one can feel cold term within his ecological theory of perception,
or anxious or uncertain without having the cor- “affordance” is used here to refer to a non-con-
responding concepts of those feelings. A propos- ceptual and entirely subjective appraisal of the
itional format does not seem to apply to feelings environment by the agent: an affordance is a
in general.16 perceived utility, which can be positive (some-
How do feelings fulfill their particular em- thing to approach and grasp) or negative (some-
bodied, subjective way of representing—a mode thing to avoid and from which to flee).17
we will call the “expressive mode”? The broadly The corresponding representation has an
functional characterization given above provides indexical structure, because it has an essential
useful clues. Expressive representations com- relation to an occurrent represented property.
prise exclusively non-conceptual, perceptual, Indexicality, however, has to be understood here
and evaluative (gradient- and valence-based) in a non-referential sense. What is indexed is an
elements, which taken together express a sub- occurrent (relational) affordance, rather than an
jective relation to the environment (internal or individual event or object. Here is our proposal
external) and a given tendency to act. It should for what a given feeling structure (FS) looks
be emphasized, however, that adult humans can like:
obviously entertain simultaneously expressive
and conceptual representations. The present hy- • FS Affordancea [Placea=here],
pothesis, in conformity with the literature on [Timea=now/soon], [Valencea=+], [Intensitya=.8
dual-processing, is that the expressive system (comparatively specified on a scale 0 to 1)],
processes information and influences decisions [motivation to act of degreed according to ac-
on the basis of its own narrow range of associ- tion programa].
ations and norms; while the conceptual system
takes advantage of background beliefs and infer- The subscript “ a” is meant to indicate that all
ential reasoning to make decisions in light of a the elements that have this subscript are rep-
broader set of norms. Let us take the case of an resentational cues, i.e., ingredients, in present
agent feeling joy after having won the lottery. A affordance-sensing a. Note that the strength
human adult normally has [lottery] in her con- (or degree) of the motivation to act does not
ceptual repertoire, along with some of the infer- depend only on the fitness significance, i.e., on
ences that can be made on its basis. However, the valence and intensity of the affordance.
the agent’s reactivity to the winning event falls Other factors, such as the physical condition
under the expressive mode of representation, of the agent and her prior arousal level (her
because this is the mode in which evaluation of mood) also modulate her motivational level
the opportunities is conducted. This feeling rep- (Schwarz & Clore 2007). The specification of
resentation presumably enlightens and orients the location of the affordance may vary, de-
the concept-based reasoning that can be con- pending on the way the feeling was generated,
ducted concerning the same event, such as won- but indexicality and reactivity suggest that
dering how to spend the money, or whether the relevant affordance is often sensed to oc-
quitting her job is a good idea. We propose to cur where the feeling is experienced. As will
call “affordance-sensing” the information that a be seen later, however, M-feelings do not in-
feeling expresses. Affordances are positive or volve a specification of place.
negative opportunities, expressed in feelings: an The feeling structure proposed above in-
affordance-sensing swiftly and non-reflectively cludes somatic markers, even if they are not
summary of the results and a methodological discussion of their sig- 17 See Proust (2009, 2013). Prinz (2004) briefly discusses this idea in
nificance, see Beran et al. (2012), Chapter 1, and Proust (2013), connection with the intentional content of emotions (p. 228). See also
Chapter 5. Griffiths & Scarantino (2009): in emotion, “the environment is rep-
16 For a defense of emotional representations as nonconceptual and ac- resented in terms of what it affords to the emoter in the way of skill -
tion oriented, see Griffiths & Scarantino (2009). ful engagement with it.”

www.open-mind.net
made explicit: these markers are the sub- evaluative informational system screens the in-
strates for the information of valence and input with its own independent memorial struc-
tensity. This information is carried by neural tures.19
activations and associated bodily changes, An affordance does not need to have an
such as a sudden sensation of pleasant muscle objective counterpart to be sensed, i.e., for a
relaxation, or of unpleasant muscle contrac- feeling to arise: it is enough that the agent anti-
tion, or of visceral contractions associated cipates it (even wrongly), imagines it, or re-
with fear. Intensity of affordance, i.e., the members it, for the corresponding feeling to be
arousal produced by a feeling, is also felt expressed. A feeling, thus, does not presuppose
through the comparative amount of bodily re- a conceptual appraisal of the context, but
activity to the affordance. These somatic rather it indexes in an embodied way a direct
markers, as emphasized above, are themselves evaluative registration. Given that an afford-
part of a monitoring system designed to pre- ance does not aim at characterizing the world,
dict and assess one’s relations to the environ- one cannot say, when the expressed affordance
ment along the relevant dimensions listed has no objective counterpart, that a feeling
above (agency, individual and social well-be- “misrepresents” the world as having a given af-
ing, preferences, and metacognition). fordance, or reciprocally that an existing afford-
Let us consider further how to read the ance was “missed” by the agent when the latter
feeling structure given above. It is meant to re- failed to detect it. For misrepresentation to oc-
flect not only what is presently felt, but also cur, a system must be equipped to attribute
what is stored in memory when a feeling is ex- properties to individual objects, that is, it must
perienced, what can be imagined, and what can be able to apply concepts. The expressive sys-
be conveyed to others in expressive behavior. tem, however, does not refer to objects as inde-
The central idea is that feelings sensitively ex- pendent entities. Hence, affordance is not liter-
press a subjective, embodied relation to an op- ally what a feeling is about, because aboutness
portunity in an input from the environment (un- presupposes that what is represented is inde-
derstood in a broad sense as including external pendent from the representational system. Be-
and bodily properties relevant to well-being). ing relational, affordances cannot be grasped in-
This primitive intentional relation is best cap- dependently of the experience of a sensitive
tured by the term affordance-sensing. Feelings agent. When saying that a feeling “expresses”
express this affordance as their focus (or formal an affordance, we mean that it “resonates” to it
object), along with its graded valence—ranging (or that it monitors it). Resonance is a neural-
from very unpleasant to very pleasant—and somatic reactivity: it carries indexical and eval-
with its intensity gradient, which ranges from uative information, but it does not refer to the
small to large.18 world or attempt to describe it.
As often emphasized, reactivity to an af- It is possible, however, to objectively char-
fordance occurs very rapidly in a processing se- acterize what a feeling functionally refers to,
quence—even before the perceptual processing and to pinpoint cases of misrepresentation, by
has been completed—and well before a concept- re-describing the feeling structure above in non-
based judgment can be made (see Dolan 2002, subjective, non-evaluative propositional terms.
p. 1191; Griffiths 1997, pp. 77; LeDoux 1996, Taking advantage of her perceptual and back-
pp. 174; Prinz 2004, pp. 60, and Zajonc 1980, ground beliefs, the agent can claim to have mis-
pp. 153). This suggests that an alternative, taken a piece of wood for a snake, for example,
18 For a review of the theories of valence, see Prinz (2004), Ch. 7. Prinz 19 These expressive representations do not require a system to have
takes valence to be a different determinate experience in each feeling. the capacity to form propositional representations. They are
On valence as determined by overall value, from a consumer se- close to what Strawson called “a featural representational sys-
mantics viewpoint, rather than as an experience of pleasure/displeas- tem”, allowing an animal to navigate with no propositional
ure, see Carruthers (2011), pp. 127–130. This view, however, does thinking (1959). On the comparison between the two representa-
not build on the nonconceptual information being felt, but rather on tional modes, see Proust (2013). The question of the penetrabil-
its being represented “in an abstract and amodal way”, which, never- ity of feelings by propositional thought is explored below, in sec -
theless, is motivating. tion 5.

www.open-mind.net
and to make explicit that there is no reason to tient’s experience of dread has valence and in-
be afraid of a piece of wood. tensity, expressed through sudden breathless-
Our analysis of FS helps us to clarify why ness, chest constriction, blurred vision, tingling
“feeling one’s keys in one’s pocket” does not be- sensations in the skin, an elevated heart beat,
long to reactive feelings. Recognizing through and a disposition to crouch. These feelings are
touch the object in one’s pocket as being one’s not only a matter of sensory “peripheral” exper-
keys, or merely having a proprioceptive experi- ience: they are also used by the patient to col-
ence in fact caused by one’s keys, are two ways lect her existing Bayesian correlations, and to
of perceiving one’s keys, involving respectively monitor with their help the present affordance
cognitive and sensory proprioception. But expressed. A second illustration of the repres-
neither needs as such to involve an affordance of entational nature of feelings is that they can
a given intensity and valence. In contrast, let us arise in the absence of the sensory basis they
suppose that the perceiver believes wrongly that seem to have. For example, illusory feelings of
she has forgotten her home keys, which are in being touched—a reactive somatosensory feeling
some distant location, and will not be able to about a change occurring on one’s body surface
get back home. Feeling her keys in her pocket —can be created by manipulating the coherence
immediately triggers a positive affordance, of the intermodal inputs from vision, touch, and
opening up the field of possible actions. proprioception. In the so-called “rubber-hand il-
lusion”, participants feel that their hand is be-
4 Questions and objections ing touched with a paint brush, when in fact it
is an artificial hand, not theirs, that they see
The present proposal raises a number of addi- being touched. They also, after a while, “feel as
tional questions and objections. Let us start if their (real) hand is turning ‘rubbery’” (see
with the most radical objection. Botvinick & Cohen 1998). This experiment is
evidence that feelings are informational states,
4.1 Are feelings representations? which monitor inputs, and, in extreme cases like
this, cause the brain to try to reconcile contra-
Granting that feelings, affective or not, can be dictory multimodal input. In the proposed in-
pure “physical effects of objects on the nerves”, terpretation, however, seeing one’s hand being
in William James’ terms (1890, vol. 2, p. 458), touched is a reactive feeling, while actively
they do not need to have any genuine represent- touching an object generates a percept—which
ational value. James invites us to take the case plays quite a different role in cognition.
either of purely somatic feelings or of objectless
emotions when they are generated by a patholo- 4.2 What does “resonating to an
gical condition—such as the precordial catch affordance” mean?
syndrome (PCS) which is a feeling of pain in
the chest that usually goes away without treat- Second, speaking of “subjective resonance” to
ment, but can lead the victim to think he or she an affordance (see the discussion of how a feel-
is suffering a heart attack. In this case, the ing “resonates” to an affordance in section 3
emotional experience of dread, James says, is above) may look improperly metaphorical.20
“nothing but the feeling of a bodily state, and it This is meant, however, to mark the difference
has a purely bodily cause” (1890, vol. 2, p. between feeling and perceiving. While percepts
459). From this, one might conclude that a feel- allow recognition and identification of external
ing is a merely peripheral phenomenon: it does objects and properties, feelings express specific
not have a function to represent, nor does it ex- affordances in a perceived, imagined, or re-
press anything in particular. What can be said, membered situation. For example, one can feel
in response, is, first, that feelings have a crucial cold right now, or simulate being cold when
evaluative function, which they perform thanks 20 In a similar vein, William James writes that, in emotions, “the whole
to their expressive structure. In PCS, the pa- organism is a sounding board” (1890, vol. 2, p. 450).

www.open-mind.net
planning a polar trip; one can remember how ory access to the world, which they don’t, for
angry, or bored one was in a given episode and they extract their inputs from sensory percep-
context. Feelings give agents prompt access to tion, or that they have direct sensory access to
the relevant features of a new situation through the body, which they don’t have either. Feelings
sensed changes in their experience. Importantly, are the subjective counterpart of bodily
resonance is also an apt term for empathy, i.e., changes.
for the propagation of feelings from an agent to Neuroscientific research about the role of
an onlooker, based on expressive behavior (De- emotion in perception offers evidence in favor of
cety & Meyer 2008; Dezecache et al. 2013). this view. An affordance is made immediately
Brain imagery suggests that the perception of salient by the system’s ability to sensitively re-
pain in another individual largely overlaps with act to a (half-)perceived element in a given
the regions activated when experiencing pain known context.21 We speak of “half-perception”
oneself (Jackson et al. 2005). Such empathy, in on the basis of what is known about the timing
the present proposal, exemplifies how a feeling of object perception. Affordance predictions are
structure can be communicated through a set of made only milliseconds after visual sensations
congruent behavioral cues associated with a register on the retina, i.e., before the categorisa-
given affordance (here a painful stimulus), with tion of perceived objects is completed (Barrett
a valence and intensity that are bodily con- & Bar 2009). The orbitofrontal cortex (OFC; in-
veyed. volved in emotion and reward in decision mak-
ing, thanks to projections from the thalamus) is
4.3 Non-conceptual content as a common able to extract an affordance in the first 80ms
feature of feelings and percepts of the visual process, merely on the basis of low
spatial frequency and magnocellular visual in-
Third, one might object that a common feature put (Lamme & Roelfsema 2000). What happens
of feelings and percepts is that they include to perceptual access when a perceiver cannot
non-conceptual contents. This is true; but no- extract affordances? Barrett & Bar (2009) have
tice the difference between the two types of shown that the lack of emotional reactivity in
non-conceptual content: while non-conceptual early perception impairs object categorization.
ingredients in perception are related to object- A patient who accidentally lost his visual ability
ive, external contrastive cues such as shapes, when three years old received in adulthood a
edges, colors, volumes, and auditory patterns, corneal transplant. In spite of his recovered abil-
which can be static or dynamic, but are always ity to extract visual information from the world,
purely descriptive, non-conceptual contents in this perceiver had trouble categorizing what he
feelings only include evaluative states, which saw. The authors’ suggestion is that reconstitut-
combine the general type of the affordance, its ing the internal affective context associated
valence, its intensity, the proper action pro- with past exposures to an object (which was
gram, where all constituents are “bodily lacking in this particular case) is “one compon-
marked”, i.e., expressed through somatic mark- ent of the prediction that helps a person see the
ers. Therefore we cannot say that feelings “per- object in the first place” (Barrett & Bar 2009,
ceive” affordances, for this would suppose either p. 1325).
that feelings have direct sensory access to the In summary: the medial OFC uses early
world—which they don’t, for they extract their low-level visual output to match the affordance
inputs from sensory perception—or that they associated with it in past experience of the ob-
have direct sensory access to the body, which ject: somatic markers are thereby activated, and
they don’t have either—feelings are the subject- the appropriate action is prepared. A FS en-
ive counterpart of bodily changes. Therefore we ables an object to be more swiftly categorized
cannot say that agents “perceive affordances”
when they experience a feeling, for this would 21 For a defence of this view in terms of situated cognition, see Griffiths
& Scarantino (2009). The authors emphasise the environmental scaf-
suppose either that feelings have a direct sens- folding that makes possible affordance detection in emoters.

www.open-mind.net
at higher perceptual levels. This evidence sug- the following way. When an associated forward
gests that affordances are extracted from per- model has been selected (often automatically,
ception, but that feelings are not themselves on the basis of an environmental, somatic, or
perceived.22 On the contrary, they offer a separ- cognitive affordance), the associated sensory
ate kind of feedback to cognitive perceptual cues (the somatic markers in this particular
processes. episode) are automatically activated in order to
monitor how this affordance is to be processed
4.4 Respective role of somatic markers and reacted to. As has been shown elsewhere,
and formal content monitoring implicitly carries information about
the command (or the affordance) that is being
Let us turn now to one of the most central monitored (see Proust 2013). This explanation
questions that our proposal raises. How does it is particularly detailed and convincing in the
explain the respective roles, in expressive inten- case of motor representations of action; the feel-
tional content, of somatic markers, on the one ings of agency that result from the comparators
hand, and of the represented formal objects on associated with a given feedforward model ex-
the other? Cognitive theorists take emotions to press (among others) whether the emoter is, or
represent both salient aspects of the agents’ is not, the author of the action currently atten-
own bodily changes and an evaluative belief ded to (see Wolpert et al. 2001 and Pacherie
about an external fact, with, possibly, a causal 2008). The present proposal generalizes the
relation between this fact and the experienced functional significance of feelings throughout
bodily change (see Gordon 1987; Tye 2008 and their diverse types (reviewed in section 2). As
Solomon 2007). For example, when perceiving a the outcome of sensory comparators, feelings al-
bear in the near vicinity, one’s experience is ways carry a structured information set about
taken to be about a complex of subjective bod- the type of affordance they contribute to regu-
ily impressions (a pounding heart, trembling lating, about its amount, and about which ac-
legs, etc.) and about the perception of a bear as tions are appropriate. This information, in its
being the cause of these changes. Such a con- own expressive mode, functionally approximates
strual of the intentional content of feelings only a causal relation that is, when propositionally
makes sense within a propositional mode of expressed, represented as a relation between an
thought. Can our expressive mode reflect or ap- internal state, an external cause, and a disposi-
proximate the information contained in this tion to act.
complex causal structure? In summary: Feelings do not gain their
Clearly, FS does not explicitly convey a aboutness through a propositional thought
causal relation between situation, somatic mark- where the contrast between object and property
ers and subjective feeling. It carries this causal is semantically marked; they gain their func-
relation implicitly, however, as a consequence of tional (rather than propositional) aboutness (f-
the control architecture that produces feelings. aboutness) through the respective roles, in ad-
In an emotional control loop, a perceived afford- aptive control, of the selection of an affordance-
ance causes (rather than being represented as dependent control model and of the markers
causing) its expressive evaluation through its that allow comparisons of valence and intensity
specialized sensory feedback. Emotional aware- to be expressed.
ness expresses this functional relation. An ex-
ternal event (made accessible through a per- 4.5 The attribution problem
ceived affordance, as detailed above) is immedi-
ately followed by subjectively experienced so- This account, however, fails to explain observed
matic cues of a given intensity and valence. In variability in the production of feelings and the
functional terms, this sequence makes sense in interpretation of what feelings are “about”.
22 When we say that a feeling is felt, “felt” is not intended to mean
There are cases where agents misattribute their
“perceived”, but, rather, “entertained”. sadness, their anger, or their happiness to an
www.open-mind.net
event that is either not real, or that actually strumental considerations and norms regulate
played no role in feeling production. How can PS.23
such a misattribution be explained on the Hence, when having to report about her
present proposal? Our first attempt to address feelings, a subject needs to translate one mode
this question is based on the subjective ground- of representation into another, with no guaran-
ing of affordances. “Feeling f” normally means tee that this translation will not enrich or
that an affordance is sensed, expressed, and modify FS intentional content. First, she may
subjectively represented as present. This does no longer have access to the rich diversity of her
not mean that the affordance has an objective FS experience, because her attention is no
counterpart. Thus a thirsty traveller can be de- longer directed toward the relevant contextu-
lighted or relieved when subjected to a water ally-activated affordance. Second, she has to
mirage. It is no problem for this view, then, if monitor other goals and their corresponding
an event does not have the action potential for (social, instrumental, or epistemic) norms. For
a given affordance it is expressed as having. example, she needs to present her feelings to
A trickier problem for the proposal is herself and to others in a socially acceptable
that a person might feel an f-feeling while she way, and to try to justify them rationally. This
thinks that she has a g-feeling. Is such a situ- in turn will depend on her existing background
ation even possible? To deal with this ques- beliefs, on her self-concept, on her capacity for
tion, we must first clarify what “transpar- making self-attributions of this particular kind,
ency” means when applied to feelings. A men- and on her willingness to perform this kind of
tal state is transparent if, when it is activated, introspective report. A number of experiments
its intentional content is accessible to the sub - and novels have documented the wide gap
ject who entertains it, while its vehicle proper- between people’s feeling experiences and the
ties are not. On the view defended above, feel- verbal report they provide, or the reasons they
ings are transparent, because their somatic offer, for having this or that feeling. These con-
markers are felt in connection with a certain siderations suggest, then, that the issue of
affordance, and because their valence and in- transparency cannot be adjudicated independ-
tensity directly influence the emoter’s motiva- ently of one’s viewpoint about mental architec-
tion to act in a given way. Such transparency, ture.24 According to the present proposal, an af-
however, does not need to entail the subject’s fordance is first subjectively recognized through
ability to verbally report the content of her the resonance it produces—through its specific
feeling. First, as seen above, a feeling can be feeling, rather than through a concept-based in-
felt by a nonhuman or by a child, both of terpretation.
whom lack the requisite verbal and conceptual Let us now return to our earlier question.
capacities. Second, even an agent endowed Can a person actually feel an f-like feeling, and
with language can express through somatic mistake this f-feeling for a g-feeling? According
markers a feeling with a distinctive FS content 23 About the nature and role of nonconceptual norms, see Proust
while failing to accurately report, in concep- (2013).
24 An alternative proposal by Carruthers (2011) sees as a condition of
tual terms, what her feeling is “about”. We transparency of an affective feeling, rather, that the corresponding
saw that [aboutness], i.e., reference to an inde- appraisal include the detection of the details of the associated non-
pendent event or object, is not a concept that conceptual somatic markers, which makes the recognition of a spe-
cific emotion possible, as well as its subsequent global broadcast—
belongs to FS. When subjects try to infer hence making this information available to the mindreading system.
[aboutness] from their experience, their pro- This analytic view of feelings, however, makes it utterly mysterious
how a given pattern of autonomic measures is ever recognized,
positional system of representation (PS) is so- among thousands of similar patterns, as distinctive of an emotion.
licited. Because the latter has an analytic On the present view, a feeling is produced within a given forward
model, which automatically activates the comparator for this afford-
rather than an evaluative function, additional ance. Transparency, then, is effective only when a given forward
constraints step in. While nonconceptual, in- model is activated, and does not need to transfer to a verbal modal -
tensive (analog) and value considerations and ity. This seems to be recognized in part by Peter Carruthers, when
he concludes that “we can have transparent access to the strength of
norms regulate FS, conceptual, digital, and in- only our occurrent context-bound affective attitudes” (2011, p. 146).

www.open-mind.net
to the present account, this situation would pre- disturbingly close to influencing participants’
suppose that an f-feeling, as it occurs in the ex- responses (see Plutchik & Ax 1967 and Gordon
pressive mode, is misdescribed in a verbal re- 1987, p. 100). Furthermore, as noted above, ex
port as a g-feeling, to finally be genuinely felt to post-facto reflective labeling of one’s emotion
be g. On this view, a change in representational does not need to express one’s original feelings.
form would not only make it possible to reinter- As shown by Nisbett & Wilson (1977), self-re-
pret the initial experience in terms of a different porting is highly sensitive to rationalizations
one, but also to feel differently. To see whether from context. A second problem, mentioned by
this case is plausible, it is worth discussing the authors in the discussion, is that the sub-
Schachter and Singer’s (1962) adrenaline experi- jects’ verbal reports and emotional behavior
ment. failed to confirm expectations in the euphoric
condition. A third methodological problem, also
5 Do beliefs influence affective report? recognized by the authors, is that the student
participants had their own independent reasons
Schachter and Singer’s famous adrenaline study for feeling anger in passing this longish test,
aimed to collect evidence in favor of a two- which predisposed them to feel anger. There
factor theory of emotion, according to which a are, however, more theoretical objections.
changed state of arousal leads agents to form On Schachter and Singer’s view, the core
feelings with a given valence that depends only feeling of an emotion is an arousal change,
on the epistemic/motivational context. Parti- which can be artificially induced by drugs.
cipants’ arousal was manipulated by injecting Valence is supposedly gained through contex-
them, under pretext, with adrenaline or a tual beliefs and motives. If this view is accep-
placebo. Only a subgroup of the adrenaline par- ted, why should we expect that contextually rel-
ticipants were informed that they had received evant beliefs specify the feeling itself (e.g., the
a drug that would modify their arousal level. anger experience)? Participants may indeed
Participants were subsequently invited to stay have been led to believe that they were angry
in a waiting room where a confederate was when they were actually merely aroused. This
either pretending to be euphoric or angry. Parti- does not show, however, that they ever felt any-
cipants’ emotional responses, observed in their thing else than an arousal change (Gordon
behavior and subsequent self-report, differed in 1987, pp. 100–101). Schachter and Singer may
the various conditions: those unaware of having have only biased self-attributions and self-report
been injected with adrenaline, and placed in the toward target emotions. The behavioral changes
anger condition, felt angriest, followed by the that were observed and attributed to felt emo-
placebo + anger subjects. The least angry were tion, in addition, can be imputed to social influ-
the adrenaline informed participants. In the eu- ence, rather than to intrinsic changes.
phoria condition, misinformed adrenaline parti- A final worry is that inducing in a parti-
cipants were “somewhat” happier, adrenaline in- cipant a somatic marker normally associated
formed ones somewhat less happy (in the eu- with a given feeling (e. g., increased heart rate),
phoria condition, the results failed to reach sig- and providing the person with a context ration-
nificance both for behavior and self-report). alizing this somatic change, does not amount to
Were Schachter and Singer successful in an ecological way of producing a feeling. A cog-
making the point that valence of a feeling is a nitivist theorist of emotion will insist that the
matter of attribution of the source of an experi- mere association between a physiological cue of
enced arousal? Several powerful objections have the feeling f and a context does not amount to
been raised against this claim. Recall that sub- the realization, by a participant, that she feels f
jects were asked to what degree they would de- because she is in such and such a context (Gor-
scribe themselves as happy or angry. A first don 1987, pp. 98–99).25 As discussed in section
problem is that the questionnaire suggested the 25 As Gordon observes, “one will not experience fear unless one con-
relevant target categories of emotions, which is nects up that cognition with the arousal one feels. To do this re-

www.open-mind.net
4, the expressive mode has a nonconceptual rep- ise that “the vivacity of the idea gives pleas-
resentation of this causal connection. The archi- ure”, and that “its certainty prevents uneasiness
tectural relation between feelings and afford- by fixing one particular idea in the mind, and
ances explains why subjects experience a sys- keeping it from wavering in the mind of its ob-
tematic connection between their feeling and jects” (Hume 1739/40, 2007, p. 289). Thus
what it is “about”, much in the same way that Hume was glad to accept that epistemic feelings
an agent experiences a systematic connection exist, and that they vary in their vivacity and
between an intention to move and the goal that in their pleasantness, i.e., in their intensity and
is aimed at—that is, without needing to repres- in their valence. Following Hume’s lead, let us
ent conceptually the causal connection between test how our analysis of FS above fares with the
the two. Nothing prevents the emoter, however, case of noetic feelings. Here, again, is our pro-
from forming a secondary conceptual represent- posal about the general structure of feelings.
ation of the emotional experience she has had,
and reappraising the context on the basis of her • FS Affordancea [Placea=here],
background beliefs. As a consequence of this [Timea=now/soon], [Valencea=+/-], [Intens-
concept-based reapparaisal, the agent may itya=.n(comparatively specified on a scale 0 to
either discount the relevance of her initial feel- 1)], [motivation to acta of degreed according
ing (as in the fear-of-snake case), or redescribe to action programa].
it in the richer terms that she now has available
(as was done, presumably, by the Schachter and What is specific to noetic feelings is that the af-
Singer participants). fordances to which the system resonates are “in-
Taken together, these objections have led formational” or “metacognitive” rather than en-
most theorists to reject Schachter and Singer’s vironmental. Hence, the affordance does not re-
two-factor theory of emotion, and to look for al- late to the external environment (the “here”
ternative accounts of the role of inferences in slot is often irrelevant, except for perceptual af-
self-attribution of feelings. It is interesting to fordances, or place-dependent metacognitive af-
see, however, that a two-factor theory has also fordances, such as concentrating in a noisy
been applied to the case of M-feelings. spot). Although a cognitive action does not, in
general, consist in physical moves towards or
6 Are metacognitive feelings sensitive to away from an affordance, similar decisions are
beliefs and inferences? motivated or inhibited in the domain of mental
agency: a high retrieval affordance motivates
What are metacognitive (also called noetic, or pursuing the memory search, a low one to quit,
epistemic) feelings? Juxtaposing [being meta- etc. Hence our FS analysis also applies to noetic
cognitive] and [being a feeling] sounds, at least feelings.
prima facie, dangerously close to an oxymoron. As already emphasized, the affordances ex-
When Descartes, Locke, and other 17th-century pressed in feelings do not need to be construed
philosophers explored the properties of ideas as conceptually in order to be detected and as-
being “clear”, “distinct”, “evident”, and “cer- sessed through their associated somatic mark-
tain” they certainly never took them to be feel- ers. A conceptual construal, however, is sugges-
ings. These notions were taken, rather, to be ted by the names given, in the literature and in
objective representational properties that the ordinary language, to M-feelings. The term
mind, unaided by imagination, is able to detect. “feeling of knowing” (in response, for example,
David Hume, in contrast, observed in his Treat- to the question: “what is the capital of Aus-
tralia?”) implicitly presupposes that the emoter
quires, according to him, a second cognition: a recognition or belief
that is one’s being (or taking oneself to be) in a situation of danger has access to the concept of knowledge. Ex-
that is causing the arousal one feels. This “cognitivist” objection is pressing her feeling verbally, indeed, an emoter
correct when targeting S and S’s theory, who also defend a cognitiv-
ist view of feelings. The present view, however, proposes a non-
might say: “I feel that I know the response to
doxastic account of feelings, and is thus immune to this objection.” this question”. In this sentence, she indeed refers
www.open-mind.net
to her disposition to retrieve knowledge and, way, what there is to learn, to retrieve from
hence, metarepresents her knowledge disposi- memory, to extract from perception, or what is
tion.26 The affordance theory of noetic feelings worth storing in memory. These are all to do
suggests a different picture. When trying to re- with the dynamics of information processing:
member a proper name, a feeling of knowing is with its onset, with the comparative amount of
a specific experience of having the ability to de- activity in incompatible neural responses, and
tect the target, and of predicting its imminent with the time needed to converge on a threshold
recall. It can be associated with a feeling of ten- value. Indeed, the neural activity recorded in
sion (Koriat & Levy-Sadot 1999, p. 486). This rats’ OFC when attempting to categorize olfact-
experience is associated, then, with a graded, ory stimuli was found to correlate with their
intuitive, and affect-like appraisal of a [remem- predictive behavior (consisting in accepting or
bering] affordance. Rhesus monkeys working in rejecting a task trial); similar patterns have
experimental labs in comparative psychology been found in other species.29
show that they can assess their memory afford- On the FS model, somatic markers have
ances (see Beran et al. 2012, Chapter 1).27 What the function of expressing the intensity and
kind of feedback, then, do monkeys use? A sur- valence of the noetic predictions generated from
prising and substantive fact about metacognit- feedback at the neural level. As indicated in sec-
ive control, first revealed through the pioneering tion 2, psychophysiological measures (elec-
research of Asher Koriat, is that the comparator tromyography) provide evidence for the exist-
generating metacognitive feelings (such as a ence of facial markers associated with feelings of
feeling of knowing in a memory task, or a feel- fluency and of disfluency (Winkielman & Ca-
ing of clearly discriminating in a discrimination cioppo 2001). Increased activity in the smile
task) has no access to the semantic contents muscle, the zygomaticus major, produces feel-
stored in memory or made available through ings with a positive valence. A reduction of flu-
perception. In Koriat’s words, M-feelings “are ency is correlated with activity in the corrug-
mediated by the implicit application of non-ana- ator supercilii (involved in frowning), which
lytic heuristics, relying on a variety of cues.” suggests that this additional effort is felt as un-
These cues “pertain to global, structural aspects pleasant. Intensity of positive or negative con-
of the processing of information”, such as ease fidence, computed implicitly, is expressed by the
of processing, time devoted to a task, familiar- corresponding intensity of the noetic feeling. A
ity, and accessibility (Koriat 2000; Koriat & different somatic marker of memory appraisal is
Levy-Sadot 1999).28 Therefore, contrary to what the TOT phenomenon. This often occurs when
epistemologists have always believed, the most a search in memory for a specific word fails to
common type of epistemic appraisal is not dir- retrieve that word within the usual time inter-
ectly based on the content of the thoughts to be val. The informational ingredients of FS are
evaluated, but on the properties of the underly- conveyed by the intensity of the activity in the
ing informational process. tongue muscle, and by the affective quality of
Neuroscientific research confirms Koriat’s TOT. Taken together, these predict the likeli-
claim. Implicit, associative cues are extracted 29 See Kepecs et al. (2008). An interesting account of the predictive
by the working brain to select, in a cost-efficient activity reflected in noetic feelings is that the dynamic activity in the
neurons activated by a given task correlates with the so-called “accu-
26 Arango-Muñoz (2012) claims that feelings of forgetting and feelings mulation of evidence” that is diagnostic of success or failure in that
of knowing are cases of “conceptual experiences”. According to the task. For example, in a perceptual discrimination task, where a tar-
present view, following the lead of Koriat and colleagues, M-feelings get might be categorized as an X or as a Y, evidence for each altern-
can overlap with judgments, and be redescribed in conceptual terms; ative is accumulated in parallel, until the difference exceeds a
they pertain, however, to different representational levels. There are threshold, which triggers the perceptual decision. The information
no “conceptual experiences”, except in the sense of experiencing the that will generate a feeling consists, first, in the differential rate of
comparative fluency of concepts. accumulation of evidence for the two (or more) possible responses,
27 As indicated above, rhesus monkeys are able, in a perceptual or and second, in stored information about the threshold value, com-
memory task, to opt out of more or less challenging trials as a result puted from prior trials, which the rate of accumulation should reach
of trial difficulty. in order to make a cognitive decision likely to be correct. For a dis-
28 As will transpire below, all these cues are, as far as we know, dimen- cussion and review of the literature, see Fleming & Dolan (2012),
sions or effects of fluency, i.e., of ease of processing. and Proust (2013, pp. 99).

www.open-mind.net
hood of successful retrieval. An implicit cue- information made consciously available to a par-
based heuristic might thus explain why TOTs ticipant, be experienced as a feeling of familiar-
have the valid predictive value they do ity, or as a feeling of recognition of that name
(Schwartz et al. 2000). as “old” (i.e., presented in a former list). They
conclude that a feeling of fluency (generated by
6.1 Two-factor theories of M-feelings a perceived name) will be experienced as a func-
tion of the alternative ways of interpreting this
In our FS single-factor model, M-feelings have feeling, on the basis of the agent’s goals and the
an intrinsic intensity and an intrinsic valence. additional cues available.30
Two-factor theories make a different claim, in A similar two-factor theory has been de-
ways analogous to Schachter and Singer’s the- fended in the (Whittlesea & Williams 2000;
ory of aboutness in affects: M-feelings have an Whittlesea & Williams 2001) model of M-feel-
intrinsic arousal level, but their valence depends ings. According to this model, feelings of famili-
on the environment. Jacoby and his colleagues arity result from the perception of a nonspecific
were the first to embrace a two-factor view discrepancy between the expected and the ob-
about feelings of fluency. They manipulated served rate of processing of elements in a given
participant’s exposure to an item in order to context. Valence and the associated action guid-
show that enhanced fluency generates an illus- ance, on the other hand, are based on a concep-
ory feeling of familiarity. Under conditions of di- tual interpretation of what this discrepancy
vided attention, reading a list containing both means. For example, you find yourself waiting
famous and not famous names raised parti- for the bus next to people you expect to be
cipants’ disposition to wrongly judge as famous total strangers. Suddenly, you have an unexpec-
some names presented in a second list, merely tedly high fluency experience when looking at
because these names had already been read in the face of someone you have already en-
the first list. Schachter and Singer’s idea was countered several times—a clerk from the local
that fluency is a generic feeling, that needs to grocery shop. This unexpectedly high rate of
be interpreted on the basis of goals and current discrepancy-reduction determines an intense
cues, in order to deliver a qualitatively different feeling of familiarity with a strong motivation to
specific feeling: identify the familiar face (see Whittlesea & Wil-
liams 2001). Had you seen the clerk in the local
Inherent in the idea that the subjective ex- grocery store instead, you would have merely
perience of familiarity arises from an inter- had a feeling of recognition when seeing the
pretation of cues is the notion that cues clerk.
can be interpreted in a variety of ways. As To summarize: the core idea in two-factor
noted above, if ease of identifying an item accounts is that participants have a primary
is obviously being manipulated by the ex- feeling of fluency, which they interpret in more
perimenter, the resulting perceptual flu- specific terms as a function of their goals and of
ency does not give rise to a feeling of fa- the context as they consciously represent it to
miliarity. Attributions are also affected by be. Thus, on this view, a feeling partly relies on
one’s goals. In the context of attempts to background knowledge, and partly on a naïve
remember, people may be more likely to theory concerning the relation between feelings
interpret ease of generating an item or and mental activity (Schwarz & Clore 2007).
perceiving it as familiarity. In the context The naïve theory is as follows: feelings are
of other tasks, the same cues may be inter- about what one is doing, so this feeling must be
preted in other ways. (Kelley & Jacoby about this event of trying to perceive, or this
1998, p. 129) attempt at retrieving, etc.
From their viewpoint, the fluency generated by 30 Jacoby & Whitehouse (1989) similarly argue that a feeling of fluency
can be experienced as familiarity in a memory task, and as confid-
a given name can, according to the task and the ence in a problem-solving task.

www.open-mind.net
As already observed above, a naïve-theory the authors are referring to unconscious cues
view is incompatible with monkeys’ and young being recruited for a task: they are thus refer-
children’s epistemic evaluations based on flu- ring to unconscious associative heuristics rather
ency. Our FS structure offers an alternative ac- than to explicit concept-based reasoning. The
count: cues (associative heuristics) dictate how memory interactions they are exploring, how-
an affordance is detected, assessed, and ex- ever, typically involve both automatic and con-
ploited in a context, but these cues are not controlled processes, which is a source of confusion.
sciously available, and hence do not depend on As Jacoby and Kelley are eager to show, impli-
a naïve theory of the task. The Jacoby and cit associations and explicit reasoning lead to
Whitehouse evidence is compatible with a pro- different, incompatible predictions. As a result,
cedural view of engagement in a task through the evidence they present shows how automatic-
automatic memory processes, and of the feelings ally-generated feelings can be theorized about in
of familiarity they generate. A comparator is al- controlled processes. It does not demonstrate,
ways activated as a function of a subject having however, that feelings depend upon theoriza-
been highly trained in the corresponding first- tion. A theory of the task, in contrast with
level cognitive task. Monkeys and humans feel automatically generated feelings, offers reasons
that a memorial or perceptual affordance is to attribute to oneself beliefs and motivations to
present because, if they need to assess whether, act, and, possibly, to reject the relevance of feel-
for example, an item was seen earlier, the asso- ings for any particular task.
ciated comparator produces a feeling of a given Our proposal, then, has several advantages
intensity and valence indexing the remembering over inferential or theory-based accounts of f-
affordance. Thus, it is uncontroversial that a aboutness. First, it explains why a feeling of flu-
context-dependent factor determines both the ency can be experienced, and why it can motiv-
task to be performed and the reactive metacog- ate agents’ metacognitive responses in species or
nitive feeling about this task. individuals with no concept-based attributive
It does not follow from the context-de- capacity (i.e., with no capacity for mindread-
pendence of a cognitive task, however, that a ing). Second, our proposal accounts for the dif-
concept-based interpretation will affect the ex- ference between a type of M-feeling (a feeling of
perienced feeling itself, as maintained by the fluency) and the various ways in which it is ex-
two-factor theorist. A cue-based, non-analytic perienced across cognitive tasks. Granting that
heuristic is not inferential in the interpretive, comparative ease of processing can always be
first-person sense. Regrettably, the word “infer- computed, and can be used as a reliable indic-
ence” has been loosely used in affective and in ator of the likelihood of success across a wide
metacognitive studies, to refer both to “auto- range of cognitive activities, it is not surprising
matic, non-analytic, largely unconscious and that there is a type of feeling based upon it.
fast associative processes” (Nussinson & Koriat Fluency can be perceptual, memorial (“retrieval
2008) and to conscious reasoning and theory- fluency”), or conceptual. It can be used in pre-
building (Schwarz & Clore 2007). These two dictive or retrospective evaluations. If agents
types of processes (respectively called “auto- are asked to determine which statements are
matic” and “controlled”), are now held by many likely to be true or false (presumably a question
authors to operate independently.31 While un- that only—but not all—humans can under-
conscious heuristics rely on implicit associations stand), felt perceptual fluency will induce a
between cues, inferences comprise deductions “truth effect”. Agents will evaluate a statement
from premises to conclusions. Looking back at as more likely to be true than another merely
Jacoby and Kelley’s point above, we see that because it is easier to read.32 If agents are asked
31 For a defence of the distinction see Jacoby & Brooks (1984), 32 There is abundant evidence, however, that M-feelings uncritically guide
Koriat & Levy-Sadot (1999), Recanati (2002) and Smith & epistemic decision (i.e., are unopposed by concept-based processes)
DeCoster (1999). Koriat & Levy-Sadot (1999) both emphas- mostly when the cognitive task is unimportant, when cognitive resources
ize the distinction and use the term “inference” in both are limited (under time pressure or divided attention), and when agents
cases. are in a good mood (Nussinson & Koriat 2008; Schwarz 2004).

www.open-mind.net
to detect faces of known people (or of stimuli 6.2 Incidental versus integral feelings
previously shown), felt fluency will generate a
sense of familiarity, which motivates agents to Our proposal also allows us to address in af-
try to identify the target. If people are asked to fective terms the issue of incidental versus in-
assess the frequency of a given phenomenon, felt tegral feelings, which, in the literature, is in-
retrieval fluency—that is, what comes immedi- variably framed in inferential terms (with all
ately to mind—will be used to judge what is the ambiguity relating to this expression).
more frequent. Felt fluency will also have effects Metacognive feelings are called “incidental”
outside of metacognition: if agents are asked when they are not based on valid cues for the
which particular face, landscape, or picture they cognitive task at hand, and hence, have no pre-
prefer, felt fluency will influence their decision. dictive value. They are called “integral” when
Several affordances, then, may be associated they actually carry information about cognitive
with the same globally expressive type of feeling outcome. Granting the universal role of fluency
(constructed as the set of feelings with the same in metacognition, how do people know when a
type of facial markers for ease of processing, for feeling of fluency is relevant to a given task,
example). The notion of type of feeling is a tech- and which sequence of their cognitive activity
nical term, which is useful to distinguish the di- needs to be monitored? A frequent answer, in
verse ways in which fluency is used by the brain. the literature, is that agents believe that flu-
But a type of feeling is never experienced; only ency applies by default to the present domain
tokens of the type are. Tokens of feelings of the of judgment. When, however, agents are led to
same type will differ in the specific affordances believe that a feeling of fluency is purely incid-
that are detected, and in the tendencies to act ental to the task at hand, they will discount it
that the feeling motivates. As a consequence, one in their decision, on the basis of a theory of the
cannot say that feelings of fluency “feel the domain of interest (see Schwarz & Clore 2007
same” to an emoter: fluency experienced in an and Whittlesea & Williams 2000, 2001). Let us
FOK and in an FOR, for example, apply to dif- suppose, in what we shall call case (a), that an
ferent segments of processing, assess different agent is explicitly told that a given cue, such
things, and motivate a different action program. as the ease of reading a given sentence, is irrel-
You may first have an FOK after a question is evant to a given task—such as assessing the
addressed to you, and then fail to have the asso- truth value of the written statement. Or, al-
ciated FOR after having come up with a re- ternatively, let us suppose—case (b)—that the
sponse. These differences have nothing to do agent discovers by himself that there is a con-
with an interpretation: they are constitutive of nection, but with reverse relevance. Perhaps he
what sensitivity to a given affordance amounts finds that badly written sentences, involving
to. Take the case of feelings of familiarity. As added processing effort—in a given context—
summarized above, Whittlesea and Williams are likely to be true (see Unkelbach 2007 and
claim that fluency is the core of the experience, Unkelbach & Greifeneder 2013). A popular ac-
while familiarity is a conceptual interpretation of count of these cases is that people will infer re-
this core feeling. It is more economical, however, spectively, for (a): that the feeling of fluent
to suppose that familiarity is a different feeling reading they have had is not about the target
within the general fluency type, and that it is as- task, which entails that reading fluency does
sociated with a different affordance. not predict truth, or, for (b): that what pre-
In summary: engaging in a particular cog- dicts the truth of a written utterance, in this
nitive task (e.g., trying to remember, evaluating particular context, is disfluent reading (see
retrieval, assessing frequency) does not need, Schwarz & Clore 2007, p. 394).
per se, to involve a naïve theory of the task. It According to this two-factor account, M-
only requires having a salient affordance, and feelings are cognitively penetrable. They can be
an implicit heuristic for metacognitive predic- suppressed at will, on the basis of a reinterpret-
tions in that task. ation of their being experienced, or can even be
www.open-mind.net
used to predict falsity instead of truth.33 On the well, which weakens the case for a theory-laden
account proposed here, in contrast, M-feelings view of feelings.
are never cognitively penetrable. Why, then, do These observations suggest that feeling-
subjects stop trusting their feeling of fluency? based and analytic appraisal, as hypothesized in
Our answer is the following. In the first type of this proposal, “tap separate databases represent-
case, subjects do not allow their feelings of flu- ing knowledge in different formats.”34 A feeling of
ency to guide their decision because they have fluency, as a result, can survive being discounted
received verbal instructions to this effect. In the in decision-making. Another finding points in the
second type of case, subjects no longer use their same direction. There is evidence that, even when
feelings of fluency to form an epistemic decision an M-feeling has been explicitly discounted (i.e.,
in the proposed task, because they have shown to agents to unduly bias their epistemic as-
learned, over time, that these feelings do not sessment), the initial feeling remains unaffected,
predict truth in this task. and is able to promote further epistemic de-
In case (a), then, subjects are confronted cisions. In Nussinson & Koriat’s (2008) study,
with a different task. They are no longer asked agents exposed to unsolved anagrams and to ana-
to express their confidence in the truth of a grams accompanied by their solution, were asked
given sentence (an intuitive, associative task); to rate the difficulty of these anagrams for naïve
they are asked to assess the truth of sentences participants with no prior access to the solution.
by taking into account the fact that their feel- The participants’ ratings were influenced by the
ings of fluency are irrelevant. This new task re- differential fluency that the anagrams presented
quires the participants to form appraisals based for them: the higher fluency of solved anagrams
on analytic reasoning. Feelings no longer drive biased their attributions of difficulty. After being
their evaluation and epistemic decision. informed of the contaminating effect of knowing
In case (b), where bad writing is associ- the solutions, the participants were invited to cor-
ated with likely truth, no “theory of the task” rect their attributions by re-rating the difficulty
needs to be formed, on top of the first-order of the anagrams, which they did. However, the
task, which consists in judging whether a writ- participants were subjected to a subsequent test,
ten statement is true or not. A mere change in where, under time pressure, they had to predict
cue validity can produce, over time, a change in which of two anagrams would be harder for oth-
associative heuristics, and, hence, in feelings ers to solve. These other-attributions of difficulty
and in decisions to act. For example, just as our presented, again, the same bias for known ana-
thirsty traveller will eventually learn not to grams. Being under pressure allowed participants’
trust an apparent “drinking affordance”, an M-feelings to guide decision. The verbal instruc-
agent will learn, in certain recurrent contexts, tion could shift their controlled responses when
not to trust an apparent “fluency affordance”. re-rating the anagrams, but did not lead the par-
Obviously, cue validity can, in humans, be con- ticipants to recompute them, as should have been
veyed verbally; this will considerably abridge the case if feelings are cognitively penetrable.
the revision process of the associated program In summary: what participants learned
of action. We then return to case (a): parti- (that solved anagrams only look easier to pro-
cipants will be able to immediately discount an cess) did not influence what they felt later
apparently valid cue, to turn to analytic ap- (higher fluency is diagnostic of ease of solving).
praisals, and to refrain from acting on their flu-
ent feeling (which, however, is still there). Cue 7 Are all feelings affective?
validity, however, can be learnt implicitly as
It is often noticed that a phenomenological con-
33 This two-factor account is endorsed by Unkelbach (2007): “the feel-
ing resulting from the discrepancy is non specific, and the discrep- trast seems to exist between feelings—that is,
ancy triggers a search for an explanation […]. The experienced vari- they are not equally emotional. Are not M-feel-
ations are not attributed to prior exposure, resulting in a feeling of
familiarity, but to some other quality of the statement, namely, that 34 A quote from Smith & DeCoster (1999), p. 329, who offer a strong
a statement is true.” defence of this view.

www.open-mind.net
ings in general as “cold” as the proprioceptive when certain dynamic conditions for affective
feeling that my right arm is being extended? Or reactions are present. But what are these condi-
can they also be “hot”—that is, involve valence, tions?
i.e., be pleasant or unpleasant? Our proposal of Let us first examine an area where these
a common expressive evaluative format suggests dynamic conditions seem to have a minimal
that all the feelings vary in affect in roughly the role. This is the area of first-order motor control
same way, because they all include valence in (including the initiation of an action, the monit-
their informational structure. Stepper & Strack oring of its development, and of goal comple-
(1993), however, have emphasized that epi- tion). As with any other form of control, motor
stemic feelings are “cold”. Feelings like effort, fa- control involves specialized feelings, in the
miliarity, surprise, or feeling of knowing “have above sense of subjective experiences with a dis-
no fixed valence”, in the sense that they don’t tinctive embodied phenomenal quality (see
feel particularly good or bad. Linguistic re- Pacherie 2008). At first glance, these feelings do
search on the emotional lexicon is invoked as not typically seem to be affective.35 Why is this
congruent evidence: for words referring to readi- so? According to Carver and Scheier, this can
ness, success, and a desire to deal with new in- be explained by the dynamics of a monitored
formation (like “alert” “confused”), i.e., terms activity that generates feelings. Affective feel-
expressing metacognition, affects are not “fo- ings are part of a second-order type of feedback,
cal”, which implies that they are not centrally having, in their terms, “the meta-monitoring
emotional (Ortony et al. 1987). function” of “checking on how well the action
There is abundant evidence, however, that loop is doing at reducing the behavioral discrep-
feelings of fluency increase perceivers’ liking of ancy that the action loop is monitoring”. This
the objects perceived. Familiar items (other meta-loop, then, monitors a particular aspect of
things being equal) are found to be more pleas- one’s progress in relation to one’s distal goal: it
ant than new ones. An initially neutral stimulus represents “the rate of discrepancy reduction in
is felt to be pleasant after repeated exposure. the behavioral (monitoring) system over time”.
This “exposure effect”, first demonstrated by This dynamic representation is what a feeling is
Zajonc, has been attributed to increased percep- equipped to offer: the intensity and quality of a
tual fluency (Zajonc 1968). This affective effect positive, or a negative, feeling express how far
of fluency has since been found to apply to any above, or how far below, the observed rate of
dimension of a perceptual input. The sense of discrepancy reduction is, with respect to some
beauty in a symmetrical face or in a landscape, reference value. One consequence of this view, if
or the pleasure felt in contemplating a picture it turns out to be experimentally validated, is
seem to be inherent to the feeling of fluency fascinating and deep: affect in action does not
generated in the perception. As noted above, depend merely on the amount of discrepancy
psychophysiological measures in the facial being reduced. An agent may be an inexperi-
muscles provide additional evidence for the af- enced performer in a task; if the velocity of her
fective character of the feeling of fluency (Reber progress to the goal is higher than expected, she
et al. 2004; Winkielman & Cacioppo 2001; for a will feel more confident, and have retrospect-
review see Oppenheimer 2008). ively more positive feelings when reaching her
An interesting, untested, speculation in- goal than a competent performer whose pro-
tended to explain the presence of cold and hot gress to the goal is as steady as predicted.
versions of feelings is that valence, although There is a second type of affect, according
never fully absent from monitoring, is modu- to Carver and Scheier, that the dynamics of
lated by dynamic aspects of the task under prediction can generate. Acceleration is the rate
evaluation (Carver & Scheier 1990; Carver & of change of velocity. Feelings express such ac-
Scheier 2001). On this view, affective feelings celeration when the rate of discrepancy reduc-
can appear in physical and cognitive action, and 35 Even in this domain, however, an error signal, when conscious, is as-
probably also in somatosensory experience, sociated with an unpleasant feeling.

www.open-mind.net
tion increases beyond expectancy—a sense of experiences is modulated by an intense, highly

exhilaration then occurs. Lucky athletes, who motivating affect of joy and renewed passion for
break several records within days, experience the associated cognitive activity; while the un-
this. Symmetrical feelings of sinking, or despair, certainty of the negatively surprised agent is as-
arise when the rate of discrepancy reduction de- sociated with an intense, highly demotivating
celerates unexpectedly and falls below the ex- affect of discouragement, or loss of interest.
pected threshold more quickly than anticipated. Note how crucial an intense feeling of this kind
In summary, cold motor feelings are generated can be, especially with regard to future motiva-
when one is routinely acting on the world, when tion. It can precipitate in children a passion for
things develop as expected, except for small mo- learning; or it can lead them to reject an activ-
tor adjustments. Hot action feelings are gener- ity, or even a whole group of similar activities,
ated when action monitoring involves unexpec- because of the threatening affect associated
ted dynamics of reduction or increment of likely with the activity, often combined with a still
success or failure. more threatening social affect (the sense of be-
How does this theory apply to M-feelings? ing an inferior, incompetent performer, or of be-
A similar contrast may exist in M-feelings. ing stupid). This kind of meta-monitoring cog-
Carver and Scheier’s model allows us to predict nitive affect, important as it is in predicting and
that M-feelings can have colder and warmer fuelling epistemic motivation, is not easily ob-
varieties, depending on the dynamics of the dis- servable in experimental settings, because it is
crepancy reduction that they express. As seen elicited in middle or long-term forms of cognit-
above, there are two varieties of M-feelings, dis- ive tasks, such as studying at school in a given
tinguished by their function. Some, like FOKs, grade, learning algebra, etc. This may in part
have a predictive function. Others, like FORs, explain why Stepper and Strack have failed to
perform retrospective evaluation. Neuroscient- encounter it.
ists explain these feelings through the rate of To summarize: noetic feelings, like all
the accumulation of evidence, measured through feelings, have an evaluative function. They are
the comparative activity of the neural assem- the output of a monitoring process, which ex-
blies involved in cognitive decision. (This rate of presses how likely it is that an agent’s cognit-
accumulation has to be compared with a stored ive preferences or goals will be (or have been)
standard in order to produce a reliable feeling fulfilled in a given task and context. They all
of confidence.) From this widely accepted have a valence, but their affective tonality is
model, it follows that the rate of reduction of more intensely felt in special cases that arise
discrepancy toward a confidence threshold is when meta-monitoring makes “intensively new”
automatically computed, and plausibly ex- affordances salient. The rate or the accelera-
pressed through somatic markers that them- tion with which an observed initial discrepancy
selves have a varying intensity. differs from a predicted standard value may
If this reasoning is correct, then although either exceed the expected value, thereby pro-
all M-feelings do not often have a definite “hot” ducing positive feelings of confidence or feel-
quality comparable to fear and love, they always ings of knowing, or be insufficient to reach this
have a valence, according to whether they pre- value, producing negative feelings of uncer-
dict an agent’s progress towards or away from tainty. The intensity of positive or negative af-
her cognitive goal. To find more intense M-feel- fect in M-feelings thus depends on particularly
ings, however, one needs to look at the dynam- unexpected properties of the underlying cognit-
ics of meta-monitoring, which is when an agent ive activity.
expects a given rate of reduction of the discrep-
ancies toward her cognitive goal, and either ob- 8 Conclusion
serves a rate that is well above the expected
rate or well below it. In these cases, the sense of On the present proposal, “feelings” are not
confidence that the positively surprised agent isolated sensory events. They are, rather, the
www.open-mind.net
ingredients of a nonlinguistic expressive mode Acknowledgement

that allows organisms to evaluate and predict
environmental changes and affordances. This I am grateful to Dick Carter, Laurence Conty,
expressive mode is of a relational, intensive Terry Eskenazy, Martin Fortier, Jonathan
kind that is not suitable for a predicative, Frome, Thomas Metzinger and Jennifer Windt
concept-based representation of the world. As for their critical comments. Special thanks to
a consequence, feelings are not themselves Dick Carter for his linguistic advice, to Tony
judgments about the world or about one’s own Marcel, whose critical questions about the com-
thoughts. They are not “about” anything in monality between M-feelings and affective feel-
the objective, referring sense of the term. Feel- ings inspired the present article, and to Robert
ings are able to approximate (in their own Gordon, for giving me access to some of his un-
mode) the guidance offered by full-blown published writings. This research has been sup-
judgments, and hence can be re-described in ported by an ERC Senior Grant “Dividnorm” #
conceptual terms when the latter are available 269616, and by two institutional grants: ANR-
to the emoter. 11-LABX-0087 IEC and ANR-11-IDEX-0001-02
The importance of the duality between PSL.
an expressive and a propositional system of
representation has generally been overlooked.
Even dual-processing theorists rarely appreci-
ate that the two systems involved in cognitive
evaluation and in reasoning have their own in-
dependent, although asymmetrical, role to
play. A purely automatic, reactive type of
evaluation is possible, and is present in nonhu-
mans and young children. It is prone, how-
ever, to generating throughout life illusions of
competence and reasoning errors. A conceptu-
ally-controlled type of evaluation, on the other
hand, can partially inhibit the influence of the
expressive system, but it still depends on the
latter to weigh the impact of context on abil-
ity, and to assess the trade-off between ease of
processing and informativeness—that is, relev-
ance—that is crucial in communication and in
problem solving.
A major practical consequence of the du-
ality between the two target representational
modes concerns pedagogy. Children cannot
learn what they are not motivated to learn.
Their motivation heavily depends on their
subjective experience of what a school context
affords them. Their feelings of confidence, i.e.,
the feedback from the cognitive tasks they en-
gage in, have to be sufficiently positive and
appropriately calibrated in order for them to
form their own realistic and motivating cog-
nitive goals. No amount of analytic reasoning
can replace a positive experience when learn-
ing.
www.open-mind.net
References Ekman, P. (1992). An argument for basic emotions. Cog-

nition & Emotion, 6 (3-4), 45-60.
Arango-Muñoz, S. (2012). The nature of epistemic feel- 10.1080/02699939208411068
ings. Philosophical Psychology, 27 (2), 193-211. Fleming, S. M. & Dolan, R. J. (2012). The neural basis of
10.1080/09515089.2012.732002 metacognitive ability. Philosophical Transactions of the
Barrett, L. F. & Bar, M. (2009). See it with feeling: Af- Royal Society B: Biological Sciences, 367 (1594), 1338-
fective predictions during object perception. Philosoph- 1349. 10.1098/rstb.2011.0417
ical Transactions of the Royal Society B: Biological Fodor, J. (1983). The modularity of mind. Cambridge,
Sciences, 364 (1521), 1325-1334. MA: MIT Press.
10.1098/rstb.2008.0312. Frank, R. H. (1988). Passions within reasons: The stra-
Bechara, A., Damasio, H. & Damasio, A. R. (2000). Emo- tegic role of the emotions. New York, NY: Norton.
tion, decision making and the orbitofrontal cortex. Goldie, P. (2002). The emotions: A philosophical explora-
Cerebral Cortex, 10 (3), 295-307. tion. Oxford, UK: Oxford University Press.
10.1093/cercor/10.3.295. (2009). Getting feelings into emotional experience
Beran, M. J., Brandl, J., Perner, J. & Proust, J. (Eds.) in the right way. Emotion Review, 1 (3), 232-239.
(2012). The foundations of metacognition. Oxford, UK: 10.1177/1754073909103591
Oxford University Press. Gordon, R. (1987). The structure of emotions. Cam-
Botvinick, M. & Cohen, J. (1998). Rubber hands ‘feel’ bridge, UK: Cambridge University Press.
touch that eyes see. Nature, 391 (6669), 756-756. Griffiths, P. E. (1997). What emotions really are.
10.1038/35784. Chicago, IL: The University of Chicago Press.
Brown, A. S. (1991). A review of the tip-of-the-tongue ex- (2004). Emotions as natural and normative kinds.
perience. Psychological Bulletin, 109 (2), 204-223. Philosophy of Science, 71 (5), 901-911. 10.1086/425944
10.1037/0033-2909.109.2.204 Griffiths, P. E. & Scarantino, A. (2009). Emotions in the
Carruthers, P. (2011). The opacity of mind. An integrat- wild: The situated perspective on emotion. In M. Ay-
ive theory of self-knowledge. Oxford, UK: Oxford Uni- dede & P. Robbins (Eds.) The Cambridge handbook of
versity Press. situated cognition (pp. 437-453). New York, NY: Cam-
Carver, C. S. & Scheier, M. F. (1990). Origins and func- bridge University Press.
tions of positive and negative affect: A control-process Haggard, P., Clark, S. & Kalogeras, J. (2002). Voluntary
view. Psychological Review, 97 (1), 19-35. action and conscious awareness. Nature Neuroscience,
10.1037/0033-295X.97.1.19 5, 282-285. 10.1038/nn827
(2001). On the self-regulation of behavior. Cam- Hume, D. (2007). Treatise of human nature. Oxford, UK:
bridge, UK: Cambridge University Press. Clarendon Press.
Damasio, A. (1994). Descartes’ error: Emotion, reason Jackson, P. L., Meltzoff, A. N. & Decety, J. (2005). How
and the human brain. New York, NY: Putnam. do we perceive the pain of others? A window into the
(2003). Looking for Spinoza: Joy, sorrow and the neural processes involved in empathy. NeuroImage, 24
feeling brain. Orlando, FL: Harvest Book Harcourt, (3), 771-779. 10.1016/j.neuroimage.2004.09.006
Inc. Jacoby, L. L. & Brooks, L. R. (1984). Nonanalytic cogni-
Decety, J. & Meyer, M. (2008). From emotion resonance tion: Memory, perception, and concept learning. The
to empathic understanding: A social developmental Psychology of Learning and Motivation, 18, 1-47.
neuroscience account. Development and Psychopatho- 10.1016/S0079-7421(08)60358-8
logy, 20 (04), 1053-1080. 10.1017/S0954579408000503 Jacoby, L. L. & Whitehouse, K. (1989). An illusion of
Dezecache, G., Conty, L., Chadwick, M., Philip, L., memory: False recognition infuenced by unconscious
Soussignan, R., Sperber, D. & Grèzes, J. (2013). Evid- perception. Journal of Experimental Psychology: Gen-
ence for unintentional emotional contagion beyond dy- eral, 118 (2), 126-135. 10.1037/0096-3445.118.2.126
ads. PLoS One, 8 (6), e67371-e67371. James, W. (1884/1890). The principles of psychology.
10.1371/journal.pone.0067371 New York, NY: Dover.
Dolan, R. J. (2002). Emotion, cognition, and behavior. Kelley, C. M. & Jacoby, L. L. (1998). Subjective reports and
Science, 298 (5596), 1191-1194. 10.1126/sci- process dissociation: Fluency, knowing, and feeling. Acta Psy-
ence.1076358 chologica, 98 (2), 127-140. 10.1016/S0001-6918(97)00039-5

www.open-mind.net
Kenny, A. (1963). Action, emotion and will. London, UK: Nisbett, R. E. & Wilson, T. D. (1977). Telling more than
Routledge & Kegan Paul. we can know: Verbal reports on mental processes. Psy-
Kepecs, A., Naoshige, U., Zariwala, H. & Mainen, Z. F. chological review, 84 (3), 231-259.
(2008). Neural correlates, computation and behavioural Nussinson, R. & Koriat, A. (2008). Correcting experience-
impact of decision confidence. Nature, 455, 227-231. based judgments: The perseverance of subjective exper-
10.1038/nature07200 ience in the face of the correction of judgment. Meta-
Koriat, A. (2000). The feeling of knowing: Some metathe- cognition and Learning, 3 (2), 159-174.
oretical implications for consciousness and control. 10.1007/s11409-008-90
Consciousness and Cognition, 9 (2), 149-171. Oppenheimer, D. M. (2008). The secret life of fluency.
10.1006/ccog.2000.0433 Trends in Cognitive Sciences, 12 (6), 237-241.
Koriat, A. & Ackerman, R. (2010). Metacognition and 10.1016/j.tics.2008.02.014
mindreading: Judgments of learning for self and other Ortony, A., Clore, G. L. & Foss, M. A. (1987). The refer-
during self-paced study. Consciousness and Cognition, ential structure of the affective lexicon. Cognitive Sci-
19 (1), 251-264. 10.1016/j.concog.2009.12.010 ence, 11 (3), 341-364. 10.1207/s15516709cog1103_4
Koriat, A. & Levy-Sadot, R. (1999). Processes underlying Pacherie, E. (2008). The phenomenology of action: A con-
metacognitive judgments: Information-based and experi- ceptual framework. Cognition, 107 (1), 179-217.
ence-based monitoring of one’s own knowledge. In S. 10.1016/j.cognition.2007.09.003
Chaiken & Y. Trope (Eds.) Dual-process theories in social Plutchik, R. & Ax, A. F. (1967). A critique of determin-
psychology (pp. 483-502). London, UK: The Guilford Press. ants of emotional state by Schachter and Singer (1962).
(2001). The combined contributions of the cue-famili- Psychophysiology, 4 (1), 79-82.
arity and accessibility heuristics to feelings of knowing. 10.1111/j.1469-8986.1967.tb02740.x
Journal of Experimental Psychology: Learning, Memory, Prinz, J. (2004). Gut reactions. A perceptual theory of
and Cognition, 27 (1), 34-34. 10.1037/0278-7393.27.1.34. emotions. Oxford, UK: Oxford University Press.
Lamme, V. A. & Roelfsema, P. R. (2000). The distinct Proust, J. (2000). Awareness of agency: Three levels of
modes of vision offered by feedforward and recurrent analysis. In T. Metzinger (Ed.) The neural correlates
processing. Trends in Neurosciences, 23 (11), 571-579. of consciousness (pp. 307-324). Cambridge, MA: MIT
10.1016/S0166-2236(00)01657-X Press.
LeDoux, J. E. (1996). The emotional brain. New York, (2009). The representational basis of brute metacogni-
NY: Simon & Schuster. tion: A proposal. In R. Lurz (Ed.) Philosophy of animal
Logan, G. D. & Crump, M. J. (2010). Cognitive illusions minds: New essays on animal thought and consciousness
of authorship reveal hierarchical error detection in (pp. 165-183). Cambridge, UK: Cambridge University Press.
skilled typists. Science, 330 (6004), 683-686. (2013). The philosophy of metacognition. Mental
10.1126/science.1190483 agency and self-awareness. Oxford, UK: Oxford Uni-
Mangan, B. (1993). Taking phenomenology seriously: The versity Press.
“fringe” and its implications for cognitive research. Reber, R., Fazendeiro, T. A. & Winkielman, P. (2002). Pro-
Consciousness and Cognition, 2 (2), 89-108. cessing fluency as the source of experiences at the fringe
10.1006/ccog.1993.1008 of consciousness. Psyche, 8 (10), 1-21. 10.1155/6152
(2000). What feeling is the “feeling of knowing”? Reber, R., Schwarz, N. & Winkielman, P. (2004). Pro-
Consciousness and Cognition, 9 (4), 516-537. cessing fluency and aesthetic pleasure: Is beauty in the
10.1006/ccog.2000.0488 perceiver’s processing experience? Personality and So-
Murata, A., Moser, J. S. & Kitayama, S. (2013). Culture cial Psychology Review, 8 (4), 364-382.
shapes electrocortical responses during emotion sup- 10.1207/s15327957pspr0804_3
pression. Social Cognitive and Affective Neuroscience, Recanati, F. (2002). Does linguistic communication rest
8 (5), 595-601. 10.1093/scan/nss036 on inference? Mind & Language, 17 (1-2), 105-126.
Nieuwenhuis, S., Ridderinkhof, K. R., Blom, J., Band, G. 10.1111/1468-0017.00191
P. & Kok, A. (2001). Error-related brain potentials are Schwartz, B. L., Travis, D. M., Castro, A. M. & Smith, S.
differentially related to awareness of response errors: M. (2000). The phenomenology of real and illusory tip-
Evidence from an antisaccade task. Psychophysiology, of-the-tongue states. Memory & Cognition, 28 (1), 18-
38 (5), 752-760. 10.1111/1469-8986.3850752 27. 10.3758/BF03211571

www.open-mind.net
Schwarz, N. (2004). Metacognitive experiences in con- Winkielman, P. & Cacioppo, J. T. (2001). Mind at ease
sumer judgment and decision making. Journal of Con- puts a smile on the face: Psychophysiological evidence
sumer Psychology, 14 (4), 332-348. that processing facilitation elicits positive affect.
10.1207/s15327663jcp1404_2 Journal of Personality and Social Psychology, 81 (6),
Schwarz, N. & Clore, G. L. (2007). Feelings and phenom- 989-1000. 10.1037/0022-3514.81.6.989
enal experiences. In A. W. Kruglanski & E. T. Higgins Wolpert, D. M., Ghahramani, Z. & Flanagan, J. R.
(Eds.) Social psychology: Handbook of basic principles (2001). Perspectives and problems in motor learning.
(pp. 385-407). New York, NY: Guilford. Trends in Cognitive Sciences, 5 (11), 487-494.
Smith, E. R. & De Coster, J. (1999). Associative and rule 10.1016/S1364-6613(00)01773-3
based processing. In S. Chaiken & Y. Trope (Eds.) Zajonc, R. B. (1968). Attitudinal effects of mere expos-
Dual-process theories in social psychology (pp. 323- ure. Journal of Personality and Social Psychology, 9 (2
336). New York, NY: Guilford. pt 2), 1-27. 10.1037/h0025848
Solomon, R. C. (2007). True to our feelings: What our (1980). Feeling and thinking: Preferences need no
emotions are really telling us. New York, NY: Oxford inferences. American Psychologist, 35 (2), 151.
University Press. 10.1037/0003-066X.35.2.151
Stepper, S. & Strack, F. (1993). Proprioceptive determ-
inants of emotional and nonemotional feelings.
Journal of Personality and Social Psychology, 64 (2),
211-220.
10.1037/0022-3514.64.2.211
Strawson, P. F. (1959). Individuals. London, UK: Meth-
uen.
Thompson, V. A., Prowse Turner, J. A. & Pennycook, G.
(2011). Intuition, reason, and metacognition. Cognitive
Psychology, 63 (3), 107-140.
10.1016/j.cogpsych.2011.06.001
Tye, M. (2008). The experience of emotion: An intention-
alist theory. Revue Internationale de Philosophie, 243
(1), 25-50.
Unkelbach, C. (2007). Reversing the truth effect:
Learning the interpretation of processing fluency in
judgments of truth. Journal of Experimental Psycho-
logy: Learning, Memory, and Cognition, 33 (1), 219-
230.
Unkelbach, C. & Greifeneder, R. (2013). A general model
of fluency effects in judgment and decision making. In
C. Unkelbach & R. Greifeneder (Eds.) The experience
of thinking: How the fluency of mental processes influ-
ences cognition and behavior (pp. 11-32). London, UK:
Psychology Press.
Whittlesea, B. W. & Williams, L. D. (2000). The source
of feelings of familiarity: The discrepancy-attribution
hypothesis. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 26 (3), 547-547.
10.1037/0278-7393.26.3.547
(2001). The discrepancy-attribution hypothesis: I.
The heuristic basis of feelings and familiarity. Journal
of Experimental Psychology: Learning, Memory, and
Cognition, 27 (1), 3-13. 10.1037/0278-7393.27.1.3


Proust 2014 The Representational Structure of Feelings

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Proust 2014 The Representational Structure of Feelings

Încărcat de

Drepturi de autor:

Formate disponibile

The Representational Structure of

Keywords Iuliia Pliushch

Feeling “Feeling” denotes a reactive, subjective experience with a distinctive embodied

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

tion increases beyond expectancy—a sense of experiences is modulated by an intense, highly

ingredients of a nonlinguistic expressive mode Acknowledgement

References Ekman, P. (1992). An argument for basic emotions. Cog-

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

Proust, J. (2015). The Representational Structure of Feelings.

S-ar putea să vă placă și