Abstract
This paper develops the claim that epistemic feelings are affective experiences. To establish some diagnostic criteria, characteristic features of affective experiences are outlined: valence and arousal. Then, in order to pave the way for showing that epistemic feelings have said features, an initial challenge coming from introspection is addressed. Next, the paper turns to empirical findings showing that we can observe physiological and behavioural proxies for valence and arousal in epistemic tasks that typically rely on epistemic feelings. Finally, it is argued that the affective properties do not only correlate with epistemic feelings but that we, in fact, capitalise on these affective properties to perform the epistemic tasks. In other words: the affective properties in question constitute epistemic feelings.
Introduction
Increasingly, epistemic feelings are shown to underpin our capacity for metacognition and our pursuit of epistemic and intellectual goods: they are responsible for our immediate sense of knowing, familiarity, understanding, coherence and rightness (e.g. Ackerman & Thompson, 2017; de Sousa, 2008; Michaelian & Arango-Muñoz, 2014; Proust, 2013). Various descriptions of epistemic feelings have been proposed such as “feelings concerning the subject’s own mental capacities and mental processes” (Michaelian & Arango-Muñoz, 2014, p. 97) or “feelings that enter into the epistemic processes of inquiry, knowledge and metacognition” (de Sousa, 2008, p. 189).
1
I will understand epistemic feelings as feelings that signal epistemic properties broadly construed.
2
Now, how do epistemic feelings feature into our ontology of mind, i.e. what
I am not the first to assimilate epistemic feelings with affective experiences. The grounds for this association have not been bulletproof, however. Some just assume that epistemic feelings are affective (Arango-Muñoz, 2014; Dokic, 2012; Dub, 2015). Others employ an “affective by association” strategy by grouping epistemic feelings together with more established affective experiences such as surprise (Carruthers, 2017a; de Sousa, 2008; Prinz, 2007, 2011). Yet others rely on a handful of empirical findings and considerations that taken by themselves appear inconclusive (Proust, 2015). So while the idea behind Affectivism is not new, it lacks solid footing. Here, I aim to provide such a footing. For that I bring the accumulating but scattered evidence together and reinforce the case for Affectivism.
Having a strong case for Affectivism matters. The idea that epistemic feelings are affective experiences is not unanimously accepted. In fact, some assume it to be false (Clore, 1992; Clore & Huntsinger, 2007; Stepper & Strack, 1993) while others refer to epistemic feelings as introspective evidence for the existence of distinctive
Although this issue might appear theoretical at first, it harbours practical implications for the ways we approach many psychopathologies. Conditions such as bipolar disorder, schizophrenia, obsessive-compulsive disorder or Capgras syndrome are marked by unusual patterns in what subjects consider right, known or familiar. It seems plausible that (alterations in) epistemic feelings have a role to play in properly conceptualising these states (e.g. Dub, 2015; McLaughlin, 2010). Against this background, understanding epistemic feelings as affective experiences might shed new light on the nature of such psychopathologies and allow us to make targeted adjustments to the ways we approach them.
Here is how I will build my case for Affectivism about epistemic feelings: In section 2 I will provide the reader with a better grasp of epistemic feelings. In section 3 I will establish some diagnostic criteria by outlining what is characteristic about affective experiences: valence and arousal. Based on that I will argue that epistemic feelings display these marks of affective experiences. For that, I will address a challenge coming from introspection in section 4: introspectively, it does not seem obvious that epistemic feelings are affective. I will try to undercut the force of this observation by appealing to the mild nature of epistemic feelings and by providing some phenomenal exhibits that are introspective evidence
Grasping Epistemic Feelings
A good way to get a grasp of epistemic feelings is to consider some typical situations in which one would experience these familiar phenomena. Remember for instance the last time you encountered a person seemingly for the first time but had the impression that you had seen her before. Such FEELINGS OF FAMILIARITY (FOF) (Whittlesea & Williams, 1998) happen to everyone from time to time and can be directed at all kinds of things (e.g. people, songs, places, odours). Consequently, we take FOFs to mean that we have encountered a certain content before. A closely related feeling is the puzzling DÉJÀ-VU EXPERIENCE (Brown, 2003) where, against your better knowledge, it seems to you as if you have already been in the situation you find yourself in now.
For another epistemic feeling, think back to your time in school. In situations when a teacher was probing the knowledge of one of your classmates, asking her questions such as “When did the French Revolution start?” or “What is the capital of Australia?” it might have occurred to you that you knew the answer. Importantly, this feeling struck you
Here are some other examples of epistemic feelings:
FEELING OF (NOT) UNDERSTANDING: The feeling that you have (not) understood a certain content. FEELING OF COHERENCE: The feeling that some content is coherent or stands in relation to another content. FEELING OF RIGHTNESS/WRONGNESS (FOR/W): The feeling that some content is right/wrong.
Now, what
Characterising Affective Experiences
In this section, I will briefly outline features that are characteristic to affective experiences. In the sections that follow, I will then use these features as diagnostic criteria and argue that epistemic feelings have these features and are thus affective experiences.
First off, affective experiences are
Arguably, the central feature of affective experiences is
It is important to emphasise that when I talk of valence I mean valence as a
Another characteristic phenomenal aspect of affective experiences is
Note that both properties of affective experiences, valence and arousal, are
I say that valence and arousal are
The Challenge from Introspection
Why think that epistemic feelings are affective experiences? As proponents of cognitive phenomenology are right to point out: when one introspects, it does not seem obvious that they are. 5 In comparison to affective experiences such as migraines, fears or orgasms, epistemic feelings are not obviously experienced as positive, negative or arousing. This datum threatens to undermine the case that epistemic feelings are affective experiences at the outset and thus needs to be addressed first.
The force of this observation is weakened by acknowledging that, usually, epistemic feelings have only a subtle positivity or negativity and degree of arousal. In other words, epistemic feelings typically come in the form of
There is a lesson here. Part of the problem for acknowledging mild affective experiences lies in the approach traditionally taken towards affective experiences. When we think of affective experiences, the focus tends to lie on a few paradigm cases of affective experiences such as pain and fear. But in what sense are pains and fears paradigms of affective experiences? Without doubt, they exhibit the features characteristic of affective experiences—valence and arousal—to an
Now, we might be able to triangulate this mild part of our affective life that is often lost to introspection by considering this: Phenomenally obvious paradigmatic affective experiences are relatively rare occurrences in comparison to, say, perceptual experiences and thoughts which are with us literally all the time. However, we know something about affective experiences that appears somewhat at odds with this apparent scarcity. Importantly, we can see the feature in question instantiated in paradigmatic affective cases: Affective experiences are typically caused by perceptual experiences and thoughts and they interact with these states in significant ways. 6 Now, we have perceptual experiences and thoughts all the time. If these are involved with affective experiences, does it mean that they get only involved with them under exceptional circumstances? Does it mean that outside of these exceptional circumstances we go about our business as some kind of “Kantian Angels” driven purely by thought and perceptual experiences—only to be sometimes thrown off our enlightened path by affective seizures?
A more natural construal is that our ever-present perceptual experiences and thoughts lead to affective experiences that are just as ever-present. Most of them, however, are not present as phenomenal ruptures but as gentle guides of thought and action. The reason why we tend to think about affective experiences in the former “violent” way might be because we tend to study the tip of the affective iceberg that happens to be more phenomenally salient.
From an evolutionary standpoint it appears plausible that we have affective experiences that are concerned with epistemic properties. Epistemic properties are of relatively high survival value to our species, a species that strongly relies on social coordination and the exchange of information. Furthermore, the importance of epistemic properties—in contrast to e.g. specific colours—is relatively invariant across contexts. It seems thus plausible that we have evolved a suite of affective states that swiftly detect these properties in our external and internal milieus (Sperber et al., 2010).
This perspective also brings to the fore that the function of affective experiences is not to be violent but to make things salient and prepare us to adaptively respond to them (Brady, 2009; Kozuch, 2020). Consequently, they typically direct our attention towards something else than themselves, towards something that matters. It is thus not surprising that we are only able to get a good look at them in exceptional circumstances—such as when they
Now, add to this our documented unreliability to introspect the nature of our experiences, especially affective experiences (Haybron, 2008; Schwitzgebel, 2008), and you get a sense for why becoming aware of mild affect—while beneficial for theoretical and personal reasons—is not at all an easy task.
So if epistemic feelings are affective experiences and, furthermore,
For the remainder of this section let me note that so far, I have been fighting a defensive battle concerning the power of introspection to shed light on the affective nature of epistemic feelings. Yes, epistemic feelings are usually mild affective experiences—typically the positive or negative valence integral to them does not come “in a very large quantity (or a high intensity), explosively” (Bramble, 2013, p. 212). This is, however, not to say that epistemic feelings cannot be reasonably intense, giving us some introspective evidence for their affective nature. To demonstrate this, I ask you to read the following passage and try to understand what it is about: A newspaper is better than a magazine. A seashore is a better place than the street. At first it is better to run than to walk. You may have to try several times. It takes some skill but it is easy to learn. Even young children can enjoy it. Once successful, complications are minimal. Birds seldom get too close. Rain, however, soaks in very fast. Too many people doing the same thing can also cause problems. One needs lots of room. If here are no complications it can be very peaceful. A rock will serve as an anchor. If things break loose from it, however, you will not get a second chance. (Bransford & Johnson, 1972, p. 722)
How do you feel? Probably confused, unable to understand—this FEELING OF NOT UNDERSTANDING is a negative epistemic feeling (e.g. Silvia, 2010). Now try to attend to what phenomenally happens when I give you the following hint: kite. You likely feel much better now; suddenly everything seems to fall into place. What you just experienced is a reasonably intense FEELING OF UNDERSTANDING (e.g. Dodd, 2014).
My favourite illustration of a FEELING OF WRONGNESS (FOW) is, alas, not well compatible with the present format. It consists in making you look at upward flowing water. 7 Looking at it, you supposedly experience a clearly unpleasant FOW about what you see. Presumably, many of us experience similar (but less intense and continuous) FOWs on seeing (or imagining) things such as a crooked picture or cars driving on the left/right side of the street.
Consider now FEELINGS OF RIGHTNESS (FORs). Think about, for instance, the last time you were arranging furniture until it “looked” or felt right. The internet has recently spawned a genre of video clips that capitalises on the FORs of the audience. These clips show events and actions that typically involve the meticulous manipulation of physical objects such as peeling wood. In fact, “Oddly Satisfying” videos have become prominent enough to be featured in
In fact, descriptions of highly intense instances of FORs occurring during ecstatic seizures (Picard, 2013) or intoxication (James, 1882, pp. 206–208) allow for an instructive peak into the affective nature of epistemic feelings. The extremely magnified feelings in these reports are the result of exceptional mental conditions. It stands to reason, however, that they are just extraordinary representatives of feelings that, in much lower dosages, are ordinary and regular ingredients of our phenomenal life. What makes these testimonies so remarkable is that we encounter FORs with their phenomenal volume turned up very high and a positive valence that comes “in a very large quantity (or a high intensity), explosively”.
These examples demonstrate that epistemic feelings can in fact be reasonably intense and clearly positive, negative and arousing.
In this section I addressed the worry that the affective nature of epistemic feelings is not introspectively obvious. In its course I strived to illustrate and explain that epistemic feelings are usually (but not always) only mildly valenced and arousing. These reflections serve to give at least initial plausibility to the idea that epistemic feelings are affective experiences. Luckily, however, there is harder evidence available. I now turn to empirical findings.
The Correlation of Affective Properties and Epistemic Feelings
In the following two sections I will make a case for epistemic feelings being affective experiences by reviewing empirical findings. In this section, I will establish the case for a covariation between affective measures and epistemic feelings. Then I will present studies that suggest that the relationship is not merely correlational but that the affect constitutes epistemic feelings.
As discussed in section 3, it is distinctive of affective experiences to possess a phenomenal valence. Thus, if it can be shown that epistemic feelings have phenomenal valence, then this can be taken as evidence that they are affective experiences.
As in general with phenomenal qualities, valence cannot be measured directly. One thus has to rely on indirect evidence by measuring observable variables assumed to be associated with valence. Luckily, several valence-associated variables have been identified in the form of psychophysiological and behavioural responses (Mauss & Robinson, 2009).
It is generally assumed that valence is bodily realised (e.g. Craig, 2009; Damasio & Carvalho, 2013). Now, epistemic feelings co-vary with bodily changes in the form of specific interoceptive alterations and facial muscle activity (Fiacconi et al., 2016, 2017; Forster et al., 2016; Topolinski et al., 2009; for a review Winkielman et al., 2003).
9
Topolinski et al. (2009), for instance, presented subjects with word triads that either did or did not share a common remote associate (e.g. coherent triad SALT, DEEP, FOAM implying SEA vs. incoherent triad DREAM, BALL, BOOK). They found that relatively to presenting incoherent triads, presenting coherent triads activated the smiling muscle,
Epistemic feelings have also been shown to lead to increased liking, a behavioural measure of positive valence (e.g. Forster et al., 2013; Trippas et al., 2016, experiment 1; Topolinski & Strack, 2009b, experiment 1, 2009c, experiment 3; Winkielman et al., 2003). Trippas and colleagues, for instance, presented subjects with simple arguments that were either logically valid or invalid. 10 They did not ask subjects to reason about the arguments but simply to rate how much they spontaneously liked them. They found that relatively to invalid arguments, valid arguments were liked more.
Another indicator of valence in epistemic feelings is the finding that they lead to affective priming effects (Topolinski & Strack, 2009c, experiment 2). In affective priming, subjects evaluate a target stimulus with an affective connotation, say, the word “poison” as positive or negative after being presented with an affectively-laden prime stimulus, say, the word “cake”. If target and prime have the same/opposite affective connotation then the evaluation of the former is facilitated/hampered. Such priming effects can e.g. be read out from a subject's response time in making target evaluations. Assuming that the previously mentioned coherent word triads trigger positive epistemic feelings, Topolinski and Strack used word triads as affective primes and negative and positive words as targets. They found that while incoherent word triads did not lead to changes in response time, coherent triads slowed subjects down when they had to subsequently evaluate a negative word. 11
Another established physiological proxy of affective experiences is the skin conductance response (SCR) which is associated with the second characteristic of affective states: felt arousal. Ordinarily, the occurrence of FOFs co-varies with a discriminatory SCR for familiar and unfamiliar stimuli (e.g. faces or words) (Ellis et al., 1999; Morris et al., 2008). Capgras patients display a similar SCR to familiar and unfamiliar faces indicating, among other things, the absence of a FOF on whose basis they could discriminate between familiar and unfamiliar individuals (Ellis et al., 1997). The patients recognise the familiar person (or sometimes pet or object) visually but the usual affective response ordinarily elicited by the sight of the individual in question (
I think these findings show that epistemic feelings co-vary with affective properties such as valence and arousal. This, in turn, strengthens the case for the thesis that epistemic feelings
Affective Properties Constitute Epistemic Feelings
What we have seen so far is that affect arises during epistemic tasks. However, this does not establish that the affective properties play any genuine epistemic role. In fact, it might be mere correlation. The observed changes in affective markers might not be part of epistemic feelings but rather consequences of other things that happen during the epistemic task. Perhaps the subjects in the experiments are simply happy or frustrated as a
I think these are legitimate considerations—it is plausible that there might be episodes of happiness and frustration as well as excitement and anxiety during the experiments. That is, there might be affective experiences that occur during the experimental tasks that are not epistemic feelings. However, I think that this is well compatible with the idea that the epistemic feelings on which the epistemic tasks themselves capitalise are affective as well. This is, (some of) the observed affective properties indeed
Particularly instructive evidence comes from two kinds of misattribution studies: The first kind of studies observes false positives of epistemic properties based on incidentally induced affect. That is, inducing nondiagnostic affect leads subjects to incorrectly judge that an epistemic property is present. The first part of this section will be concerned with these studies. The second part of this section will be dedicated to the second kind of misattribution studies. These go the other way around: the researchers make the subject believe that the affect they experience during an epistemic task is not diagnostic for the presence of an epistemic property. This turns out to strip the subject of her ability to accurately detect the epistemic property, indicating that epistemic properties are detected based on affect, and, since the affect can be misattributed, that the affect in question is conscious.
The first kind of studies generates a misattribution of seemingly non-affective properties such as familiarity, coherence and grammaticality based on induced positive or negative affect.
13
In the familiarity studies, novel stimuli are rated as more familiar (or unfamiliar) as a result of the affect manipulation. This holds true for various affect manipulations: i) making participants contract the smiling muscle,
Crucially, Duke and colleagues and Topolinski and Strack explicitly demonstrate that the effect of induced affect closely mirrors the effects of processing fluency (as well as actual familiarity, coherence and grammaticality) on familiarity, coherence and grammaticality judgments (Duke et al., 2014; Topolinski & Strack, 2009a). This needs a little unpacking. To understand the importance of this finding, we need to familiarise ourselves with the construct of processing (dis)fluency (Alter & Oppenheimer, 2009). Processing fluency is a process property that refers to the “ease”, understood as relative speed, with which a given cognitive process is executed. 14
There are a couple of things that we know about processing fluency. For instance, it is a prominent proximal cause of epistemic feelings, leading to judgments of epistemic properties such as familiarity or coherence (e.g. Unkelbach & Greifeneder, 2013). Now, there is something else that we know about processing fluency: it has been found to trigger positive affect (Winkielman et al., 2003). We can now connect the dots between these two observations.
When the researchers induce fluency-independent affect, they find that it mirrors the effects of fluency on judgments of epistemic properties. This parallel effect suggests two things: First, the induced affect seems to be
Against this background, it appears likely that epistemic feelings in general (i.e. also those not caused by fluency) are constituted by transient, context-specific positive or negative affect.
This point is reinforced by the second kind of misattribution studies, to which I now turn. While in the first kind of studies the subjects misattribute seemingly non-affective epistemic properties based on affect, in this kind of studies the misattribution goes the other way around: Informative affective reactions are discounted by being misattributed to an irrelevant source (Topolinski & Strack, 2009b, 2009c). In these studies, the experimenters ask subjects to make semantic coherence judgments by discriminating between word triads that either share a common remote associate (e.g. SALT, DEEP, FOAM implying SEA; coherent triad) or not (e.g. DREAM, BALL, BOOK; incoherent triad).
In the fluency reattribution condition, the subjects are told that the “easiness of reading and the fluency with which the meaning of words is recognized” (Topolinski & Strack, 2009b, p. 614) is due to a task-irrelevant source: background music. In the affect-reattribution condition, the subjects are told that the positive affect that might arise in the course of the task is due to the background music. The authors show that while misattributing fluency has no effect on performance, misattributing affect essentially strips subjects of the ability to detect the property of semantic coherence (above chance level).
Importantly, the aim of the researchers was to find out what is
First, this strengthens the initial case made on the basis of the findings by Duke and colleagues and Topolinski and Strack by suggesting that epistemic feelings consist in context-specific, transient positive or negative affect. Fluency is not a cue available in experience to use for judgment. What is available is the result of fluency: positive affect. The researchers additionally back this conclusion with the finding that coherent triads are liked more than incoherent triads but are not rated as more fluent in processing (Topolinski & Strack, 2009b, experiment 1).
Reinforcing and extending this point, Balas and colleagues find that altering the semantic coherence task to include word triads that themselves are neutral but have an affect-laden common remote associate has a characteristic impact on judgments of semantic coherence: 15 there is an increase in accuracy and speed for triads with positive associates relative to those with neutral and negative ones. On this basis, the authors argue that “fluency-based positive affect can be strengthened or weakened by affective responses induced through partial activation of an affectively valenced memory content (i.e., solutions to triads).” (Balas et al., 2012, p. 318) This, in turn, brings the point home (in line with Duke et al., 2014 and Topolinski & Strack, 2009a) that “fluency of processing is not the only source of affective response that can influence intuitive judgements” (Balas et al., 2012, p. 312).
Together these findings imply that seemingly non-affective epistemic properties such as coherence are (sometimes) detected based on
Second and perhaps even more important: The valence in epistemic feelings needs to be conscious in order to make them affective
However, we cannot explain the present experimental findings simply by relying on them. On the contrary, the mentioned studies demonstrate that the affect integral to epistemic feelings is conscious. This is because the subjects can misattribute the conscious affective signals that they would usually use to make conscious judgments. This contrasts with e.g. their inability to misattribute and use the unconscious processing fluency directly. Subjects cannot misattribute something that is unconscious since there is nothing to (correctly or incorrectly) attribute in the first place. The present finding, thus, rules out something that might seem like a possible explanation when one considers unconscious valence. Instead, what we observe in the experiments appears to be the result of affective experiences—epistemic feelings.
Conclusion
Here, I have provided a case for the idea that epistemic feelings are affective experiences. I first outlined the characteristic features of affective experiences: phenomenal valence and felt arousal. Using these as diagnostic criteria I proceeded to make the case that epistemic feelings possess said features. To give this idea initial plausibility, I explained why the affective nature of epistemic feelings might not appear introspectively salient: epistemic feelings are usually only mildly valenced and arousing. I also provided some phenomenal examples where the affective nature of epistemic feelings is introspectively salient.
I then turned to empirical findings to show that epistemic feelings covary with affective markers. Specifically, epistemic feelings covary with interoceptive changes, variations in SCR and facial muscle activity, proxies for the affective properties of valence and arousal. Furthermore, positive epistemic feelings lead to increased liking and can serve as positive affective primes—behavioural proxies for the presence of valence.
I went on to make the case for the covariation between epistemic feelings and affective properties being not just a correlation but a constitution relationship. For that, I presented studies that observe false positives of epistemic properties based on incidentally induced affect. That is, inducing nondiagnostic affect leads subjects to incorrectly judge that an epistemic property is present. This speaks in favour of an affective constitution of epistemic feelings.
Secondly, I made the case that the constitutive affect in question is conscious. It thus not only causally biases epistemic behaviour but phenomenally constitutes epistemic feelings that provide conscious guidance for the subject's epistemic behaviour. To build the case for this idea, I recounted studies where the following happens: the experimenters make the subject believe that the affect they experience in an epistemic task is
This conclusion is not without consequence. I mentioned at the outset that epistemic feelings are plausibly involved in psychopathologies such as bipolar disorder, schizophrenia, obsessive-compulsive disorder or Capgras syndrome. Specific aspects of these conditions can be cast in a new light by applying what we know about affective experiences to epistemic feelings. If e.g. the delusions characteristic to manic episodes are not put in place by faulty reasoning but by abnormal affective experiences, say aberrant feelings of rightness or wrongness, then quite different considerations apply when making an assessment. Arguably, we have significantly less intentional control over the ways we feel than over the ways we reason. Additionally, affective experiences are typically imbued with motivational force and are thus particularly hard to override (Brady, 2009; McLaughlin, 2010). This perspective has thus implications for the agency and responsibility we ascribe to somebody in a manic episode. Similar considerations apply to the ways we go about treatment. As e.g. demonstrated by exposure therapy, maladaptive affective experiences can be changed but they are sensitive to very different kinds of evidence than is reasoning.
All this shows is that identifying epistemic feelings as affective experiences is good news because it allows us to apply the wealth of theoretical and empirical resources that we have for the latter to understand the former. At the same time, we realise how the affective realm expands into domains traditionally considered the province of “cold” cognition. It turns out that affect is an integral part of our intellectual and epistemic lives.
Footnotes
Acknowledgements
I would like to thank Mark Wulff Carstensen, Ophelia Deroy, Jérôme Dokic, Elisabeth Pacherie, Pablo Fernandez Velasco, Jurgis Karpus and the participants of the 2019 Annual Meeting of the Association for the Advancement of Philosophy and Psychiatry (AAPP) for helpful comments on previous drafts of this article.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the H2020 Marie Skłodowska-Curie Actions under grant number 675415 and the Agence Nationale de la Recherche under grant agreement ANR-10-IDEX-0001-02 and ANR-17-EURE-0017.
