It is often noticed that a phenomenological contrast seems to exist between feelings—that is, they are not equally emotional. Are not M-feelings in general as “cold” as the proprioceptive feeling that my right arm is being extended? Or can they also be “hot”—that is, involve valence, i.e., be pleasant or unpleasant? Our proposal of a common expressive evaluative format suggests that all the feelings vary in affect in roughly the same way, because they all include valence in their informational structure. Stepper & Strack (1993), however, have emphasized that epistemic feelings are “cold”. Feelings like effort, familiarity, surprise, or feeling of knowing “have no fixed valence”, in the sense that they don’t feel particularly good or bad. Linguistic research on the emotional lexicon is invoked as congruent evidence: for words referring to readiness, success, and a desire to deal with new information (like “alert” “confused”), i.e., terms expressing metacognition, affects are not “focal”, which implies that they are not centrally emotional (Ortony et al. 1987).
There is abundant evidence, however, that feelings of fluency increase perceivers’ liking of the objects perceived. Familiar items (other things being equal) are found to be more pleasant than new ones. An initially neutral stimulus is felt to be pleasant after repeated exposure. This “exposure effect”, first demonstrated by Zajonc, has been attributed to increased perceptual fluency (Zajonc 1968). This affective effect of fluency has since been found to apply to any dimension of a perceptual input. The sense of beauty in a symmetrical face or in a landscape, or the pleasure felt in contemplating a picture seem to be inherent to the feeling of fluency generated in the perception. As noted above, psychophysiological measures in the facial muscles provide additional evidence for the affective character of the feeling of fluency (Reber et al. 2004; Winkielman & Cacioppo 2001; for a review see Oppenheimer 2008).
An interesting, untested, speculation intended to explain the presence of cold and hot versions of feelings is that valence, although never fully absent from monitoring, is modulated by dynamic aspects of the task under evaluation (Carver & Scheier 1990; Carver & Scheier 2001). On this view, affective feelings can appear in physical and cognitive action, and probably also in somatosensory experience, when certain dynamic conditions for affective reactions are present. But what are these conditions?
Let us first examine an area where these dynamic conditions seem to have a minimal role. This is the area of first-order motor control (including the initiation of an action, the monitoring of its development, and of goal completion). As with any other form of control, motor control involves specialized feelings, in the above sense of subjective experiences with a distinctive embodied phenomenal quality (see Pacherie 2008). At first glance, these feelings do not typically seem to be affective.[35] Why is this so? According to Carver and Scheier, this can be explained by the dynamics of a monitored activity that generates feelings. Affective feelings are part of a second-order type of feedback, having, in their terms, “the meta-monitoring function” of “checking on how well the action loop is doing at reducing the behavioral discrepancy that the action loop is monitoring”. This meta-loop, then, monitors a particular aspect of one’s progress in relation to one’s distal goal: it represents “the rate of discrepancy reduction in the behavioral (monitoring) system over time”. This dynamic representation is what a feeling is equipped to offer: the intensity and quality of a positive, or a negative, feeling express how far above, or how far below, the observed rate of discrepancy reduction is, with respect to some reference value. One consequence of this view, if it turns out to be experimentally validated, is fascinating and deep: affect in action does not depend merely on the amount of discrepancy being reduced. An agent may be an inexperienced performer in a task; if the velocity of her progress to the goal is higher than expected, she will feel more confident, and have retrospectively more positive feelings when reaching her goal than a competent performer whose progress to the goal is as steady as predicted.
There is a second type of affect, according to Carver and Scheier, that the dynamics of prediction can generate. Acceleration is the rate of change of velocity. Feelings express such acceleration when the rate of discrepancy reduction increases beyond expectancy—a sense of exhilaration then occurs. Lucky athletes, who break several records within days, experience this. Symmetrical feelings of sinking, or despair, arise when the rate of discrepancy reduction decelerates unexpectedly and falls below the expected threshold more quickly than anticipated. In summary, cold motor feelings are generated when one is routinely acting on the world, when things develop as expected, except for small motor adjustments. Hot action feelings are generated when action monitoring involves unexpected dynamics of reduction or increment of likely success or failure.
How does this theory apply to M-feelings? A similar contrast may exist in M-feelings. Carver and Scheier’s model allows us to predict that M-feelings can have colder and warmer varieties, depending on the dynamics of the discrepancy reduction that they express. As seen above, there are two varieties of M-feelings, distinguished by their function. Some, like FOKs, have a predictive function. Others, like FORs, perform retrospective evaluation. Neuroscientists explain these feelings through the rate of the accumulation of evidence, measured through the comparative activity of the neural assemblies involved in cognitive decision. (This rate of accumulation has to be compared with a stored standard in order to produce a reliable feeling of confidence.) From this widely accepted model, it follows that the rate of reduction of discrepancy toward a confidence threshold is automatically computed, and plausibly expressed through somatic markers that themselves have a varying intensity.
If this reasoning is correct, then although all M-feelings do not often have a definite “hot” quality comparable to fear and love, they always have a valence, according to whether they predict an agent’s progress towards or away from her cognitive goal. To find more intense M-feelings, however, one needs to look at the dynamics of meta-monitoring, which is when an agent expects a given rate of reduction of the discrepancies toward her cognitive goal, and either observes a rate that is well above the expected rate or well below it. In these cases, the sense of confidence that the positively surprised agent experiences is modulated by an intense, highly motivating affect of joy and renewed passion for the associated cognitive activity; while the uncertainty of the negatively surprised agent is associated with an intense, highly demotivating affect of discouragement, or loss of interest. Note how crucial an intense feeling of this kind can be, especially with regard to future motivation. It can precipitate in children a passion for learning; or it can lead them to reject an activity, or even a whole group of similar activities, because of the threatening affect associated with the activity, often combined with a still more threatening social affect (the sense of being an inferior, incompetent performer, or of being stupid). This kind of meta-monitoring cognitive affect, important as it is in predicting and fuelling epistemic motivation, is not easily observable in experimental settings, because it is elicited in middle or long-term forms of cognitive tasks, such as studying at school in a given grade, learning algebra, etc. This may in part explain why Stepper and Strack have failed to encounter it.
To summarize: noetic feelings, like all feelings, have an evaluative function. They are the output of a monitoring process, which expresses how likely it is that an agent’s cognitive preferences or goals will be (or have been) fulfilled in a given task and context. They all have a valence, but their affective tonality is more intensely felt in special cases that arise when meta-monitoring makes “intensively new” affordances salient. The rate or the acceleration with which an observed initial discrepancy differs from a predicted standard value may either exceed the expected value, thereby producing positive feelings of confidence or feelings of knowing, or be insufficient to reach this value, producing negative feelings of uncertainty. The intensity of positive or negative affect in M-feelings thus depends on particularly unexpected properties of the underlying cognitive activity.