1 Introduction

This paper examines two rather neglected phenomena that in many respects resemble each other, metaphysically, phenomenologically, and neurobiologically: episodic emotional (‘affective’) memories and imagined emotions. Both states seem to play significant and ubiquitous roles in our everyday lives, such as allowing us to ‘re-live’ important aspects of our past experiences, and to make important decisions by imagining the emotional responses of our future selves. Like emotions in general, they seem to disclose important self-relevant values, but ones with specific past and future-oriented temporal dimensions. Both states, however, also clearly differ in various respects from everyday emotional responses towards occurrent events, situations, and objects. They thus raise important questions concerning the difference, if there is one, between so-called ‘real’ and ‘as if’ emotions, and any account of these phenomena should therefore help to shed light on the nature of emotion as such.

Yet, curiously, the existence of ‘as if’ emotions has been explicitly denied, with sceptics claiming that the purported states are in fact cases of genuine ‘real’ emotional responses to remembered or imagined contents. (Currie & Ravenscroft, 2002; Debus, 2007) Equally surprising is the fact that there are few attempts to defend ‘as if’ emotions against such attacks, and correspondingly few positive accounts of the phenomena themselves, either in the philosophical or empirical literature. (Goldie, 2012; Arcangeli & Dokic, 2018; Loev et al., 2022) As we shall see, these accounts are all explanatorily inadequate in different ways. A proper defence and account of ‘as if’ emotions is therefore overdue.

Comparing affective memories and affective imaginings directly, I contend, allows us to grasp important phenomenological similarities between them, and in doing so to counter scepticism about their existence, in part by undermining certain misguided assumptions about them and about the general nature of emotion on which such views depend. It also helps point the way to a more plausible explanation of their nature and function, which I will argue depends on the fact that both affective memory and imagined emotion essentially involve what I will call ‘affective bodily imagery’.

There is currently almost no recognition, let alone discussion, of the existence of affective imagery in the philosophical or empirical literature, but I will contend that positing it is required to explain what makes these states distinctive and phenomenologically different from ‘real’ emotions. Specifically, I will claim that when we experience emotional memories or imagine feeling an emotion, the various bodily feelings comprising ‘real’ emotions will in these cases be the images of such feelings, and hence the phenomenological character and content of our state will be imagistic, rather than perceptual. It is important to make clear that I’m aiming to give an account of ‘as if’ emotions rather than ‘real’ emotions and I will side, without argument, with what I take to be the majority view amongst emotion researchers: namely that emotions, and conscious emotional experiences, in some way essentially involve bodily feelings. However, I will not commit to any particular view of this involvement since I will suggest that the account I offer is broadly compatible with all such ‘feeling theories’. (For references see p. 3 below; Deonna and Teroni 2012 for an overview)

My primary aim is to defend the claim that it’s the presence of imagistic bodily feelings that makes ‘as if’ emotions distinctive and provides the key phenomenological contrast to ‘real’ emotions. Moreover, our capacity to remember and imagine such imagery plays a significant role in allowing us to perform certain important functions, such as engaging in empathetic understanding or making decisions about our future selves. Outlining the nature of affective bodily imagery will thus enable us to explain the nature and function of emotional memories and emotional imaginings. In doing so we will also be able to illuminate some important connections between memory, imagination, imagery, and emotion.

The paper is structured as follows. In Section 2, I introduce the notion of ‘as if’ emotion via examples of affective memory and imagined emotion, and contrast these with cases of ‘real’ emotion. Section 3 addresses sceptical views denying the existence of such states and I show that these views all fail, partly for the same reason: they mischaracterise the very phenomena they seek to deny. Section 4 looks in detail at one of the main positive accounts of affective memory and argues that it too fails to properly characterise and explain the phenomenon. I then draw some important implications concerning the relationship between emotion, imagery, and perspective. Section 5 examines a recent account of imagined emotion, and points to significant ways in which, while promising, it remains explanatorily inadequate. Section 6 offers an original, empirically informed account of the nature of affective bodily imagery, showing how it explains the nature, phenomenology, and function of ‘as if’ emotion. I conclude by considering some objections and some important potential connections between memory, imagination, and emotion.

2 Characterising ‘as if’ emotion

What is an affective memory? The question seems deceptively simple, especially if we look at specific cases where one remembers being in some clear emotional state in the past. For example, I remember feeling happy when swimming in the Ionian Sea last summer. In this case, the happiness was not just a reaction to the beautiful sensory surroundings but was also partly a result of the feeling of relief I felt from having completed a significant piece of work before going on holiday. The overall positive affective experience I am now remembering as ‘happiness’ thus comprises a number of different valenced (and non-valenced) aspects. Some of these are best characterised as primarily sensory: the pleasing visual sensation of the deep blue sky and the sunlight glinting on the sea, the felt bodily freedom and relaxation of floating, the warm sun on my skin. Others are not obviously directly connected to bodily sensations and might be described roughly as psychological or cognitive: relief from the stress of the deadline, satisfaction in producing something difficult, intellectual pride, and no doubt a range of other related perceptions, thoughts and attitudes. Emotional states are often a complex mixture of many different elements. (Deonna and Teroni 2012)

What is important to note for now is that the memory is not just a semantic memory that I was in such and such an emotional state. The memory itself gives me pleasure now as I seem, in some sense, to relive those bodily sensations and the overall feeling of calm and contentment that characterised the experience. It thus requires us, in some sense, to re-experience the original state ‘from the inside’, from our own first person subjective point of view. As such, it shares a similar phenomenology – comprising bodily sensations, psychological feelings, emotional reactions, and perhaps also an awareness of the remembered (intentional) objects and causes of those states (the sun on the sea, the finished paper, etc.) – with the original emotional experience, plausibly in virtue of inheriting these features from that experience.

Yet, the memory also clearly differs in marked phenomenological ways from the original experience, even if it can be a challenge to articulate these differences precisely. Remembering the bodily feeling(s) involved in floating or swimming is obviously different from experiencing the actual bodily feeling(s) of floating or swimming, and the remembered vision of the glinting sunlight is clearly different from the actual perceptual experience of the sunlight. One natural way of putting it is to say that the remembered experiences are somehow fainter or less vivid, less intense or detailed than the original experiences. They are fainter or less vivid in something like the way that visual images have often been said to be in relation to their perceptual counterparts. (McGinn, 2004; Nanay, 2015, 2016; Kind, 2017; Pearson, 2019)

Arguably the memories seem like this precisely because they essentially involve images. However, although that is obviously the case for visual memory that deploys visual imagery, it might seem less obvious that the memory of the feeling of floating, let alone the memory of feeling relieved or proud or calm involves images. Of course, we may have a visual imagery of ourselves floating, and may simply remember now that we felt relaxed and relieved and happy at the time, but this does not seem to properly capture the remembered experience. We seem to re-experience phenomenologically similar feelings of relief or calmness or bodily movement or sensation in remembering them. I want to suggest, for the moment without argument, that many of these feelings are also imagistic in nature, and that it is primarily because they are structured differently from visual imagery that they have been overlooked in discussions of affective memory and imagined emotion.

Swimming in the Ionian Sea made me happy, and this is an overall characterisation of an emotional state with many components, as I noted above. Most researchers agree that emotions are essentially complex states constituted by many different elements – perceptions, physiological responses of arousal and action readiness, various attitudes and cognitive states and actions, such as appraisal and evaluation. Yet there is a lot of debate about the nature of this constitution and which components, if any, are necessary or sufficient for emotions in general and/or particular emotion types. Arguably, however, most contemporary emotion theories hold that bodily changes, and the feelings of such changes, are in some way central to – and on some views partly constitutive of – an emotional response. For example, somatic feeling theories (Damasio, 1994; Prinz, 2004; Laird, 2007; Friedman, 2010), many appraisal theories (Ellsworth & Scherer, 2003; Scherer, 2004; Lewis, 2005; Moors et al., 2013. See Lowe & Tiemke, 2011 for discussion), constructivist theories (Barrett, 2005, 2017), the attitudinal theory (Deonna and Teroni 2012), and Kriegel’s ‘New Feeling Theory’ (Kriegel, 2015b) all give prominent (albeit different) roles to bodily feelings.

It seems to me that all of these theories are right to reject traditional cognitivist theories (e.g. Nussbaum, 2001) that exclude bodily feelings from playing any essential role in emotions. I will not argue for that here, and hence I will thus just assume that there aren’t any conscious emotional experiences in which bodily feelings don’t figure in some way. Nor will I commit to any one of these theories, since I think that the account I develop below is broadly compatible, or can potentially be made so, with all of these ‘feeling’ theories. For now, it is important to note only that insofar as emotions arguably comprise certain bodily states that are in some way felt, it seems plausible that remembered emotions will comprise memories of these feelings. Although the other components – e.g. perceptual and cognitive – also play essential roles in the original and remembered emotional experiences, and perhaps also contribute to their phenomenology (cf. Kriegel, 2015b), I will eventually claim that the phenomenological distinctiveness of ‘as if’ emotions is due primarily to the nature of the remembered bodily feelings.

I will have more to say about the phenomenology of imagined emotion in Section 6, but in most salient respects it does not seem to differ in any clear way from that of remembered emotion. It seems that we can give exactly the same phenomenological description of my Ionian Sea memory, but as a case of imagining rather than remembering. One way, of course, in which it does differ is simply that in the one case we are normally aware that we are remembering, while in the other case we are normally aware that we are imagining. It is an open question, however, whether this awareness itself in either case has any distinctive phenomenological profile, such as a particular feeling of agency, or feeling of pastness, for example. (McGinn, 2004; Bayne & Montague, 2011; Farkas, 2013; Dokic & Martin, 2015; Kriegel, 2015a; Nanay, 2015, 2016; Teroni, 2017; De Brigard, 2017; Arango Munoz & Bermudez 2018; Hopkins, 2018; MacPherson & Dorsch, 2018; Pearson, 2019; Kind, 2020; Dijkstra, 2022)

For now, all we need to bear in mind is that insofar as sensory (visual, tactile etc.) imagery is involved in both imagining and remembering of the kind we are discussing, and if there is a role for cognitive (and metacognitive) phenomenology in either state, it might be that the overall phenomenology of affective memory is in general different from that of imagined emotion only in virtue of differences that are, so to speak, extraneous to the differences in the specifically emotional aspects of these states; for example, the phenomenology of effort or agency or pastness. This, in fact, is what I will suggest later by delineating what is specific to the phenomenology of the affective imagery that makes ‘as if’ emotion states phenomenologically distinctive and different from ‘real’ emotion states.

3 The case against ‘as if’ emotion

Leaving aside questions about whether memory ‘preserves’ past experiences and in what sense we can plausibly be said to ‘relive’ them (Matthen, 2010; Soteriou, 2018), why might one deny that there are examples of affective memory of the kind I just presented?

One type of sceptical view targets a purported analogy between emotional memories and everyday perceptual memories. My memory of seeing the Colosseum closely resembles, in many ways, the original visual experience. Yet, of course, in many ways the visual memory differs significantly from the original perceptual experience, much in the way that images differ from percepts. As we have just seen, it has often been claimed that images are fainter, less vivid, and less detailed or saturated. When bringing the image of the Colosseum to mind in memory, it is only ‘as if’ we are now seeing the Colosseum. The question at issue is then whether ‘as if’ emotional memories stand in the same kind of relation to occurrent ‘real’ emotional experiences as visual memories do to occurrent visual experiences.

Debus (2007) has provided the most significant articulation of this sceptical position, which claims that unlike in cases of standard perceptual memory there is no real distinction between ‘as if’ and ‘real’ emotional experiences. Starting with the position that remembered emotions are occurrent emotions, her view is that examples of the kind I have just given are best thought of as perceptual memories of the thing or event that triggered the original emotional reactions, and these memories now give rise to occurrent emotional reactions to the remembered content. But since occurrent emotional reactions (qua emotions) are not themselves memories, they clearly cannot be instances of the distinctive state we assume they are, namely that of being a specifically emotional memory: that is, a memory that is also an emotion, or an emotion that is also a memory. Thus, if all supposed instances of distinctively emotional memories turn out to be merely occurrent emotional reactions to the remembered content, there are no distinctively emotional memories.

If we accept what appears to be the simple fact that memory and emotion are different things, and hence that occurrent emotions are not memories, it seems quite straightforward to arrive, from the starting point that affective memories are occurrent emotions, at the claim that they therefore cannot be memories. Yet as Debus herself points out, her opponent might accuse her of beginning on the wrong foot, since such emotions as she characterises them do not demarcate the phenomenon we are interested in, that of distinctively emotional memories. As she notes, our everyday experience seems to suggest the existence of such states, and we readily make a difference between (A) ‘full-blown’ past directed emotions and (B) ‘as-if’ emotions, which are something like the ‘feelings again of an old emotion’, as in the commonplace example I gave above.

Debus explores and rejects one possible explanation that is highly pertinent for our discussion, the idea that emotional memories are characterised by some faint proprioceptive memory experiences:

I have an affective memory (an ‘as if’ emotional response) whenever I “perceptually (proprioceptively) remember some event (i.e. a certain bodily change) which was a constitutive part of my experiencing a certain emotion at the relevant time in the past…[for example] a subject who remembers a situation in which she was very much afraid might, when remembering the event, have a proprioceptive memory of her heart beat’s speeding up. The subject’s heart beat at present remains constant, but she perceptually (namely, proprioceptively) remembers how her heart beat accelerated at the relevant time in the past.’ (p.776).

This view seems both plausible and suggestive, and I will go on to defend an account along these lines, focussing not just on proprioception but also on feelings of ‘action tendencies’ and ‘interoception’ – the awareness via feeling of certain visceral bodily changes. Debus, however, argues that it is misleading to call these ‘as if’ emotions, since they are in fact proprioceptive perceptual memories of relevant bodily changes which were constitutive parts of a subject’s experiencing a certain emotion in the past. As such, she says, they cannot count as APD-emotions since “…we will all agree (or so I hope) that a perceptual memory of a past event (e.g., a proprioceptive memory of a past bodily change) is not an occurrent emotion.’ (p.777).

This response is clearly question-begging, for even if it is true that a mere perceptual memory is not itself an occurrent emotion, ‘as if’ emotional memories are precisely not supposed to be occurrent emotions, if ‘occurrent’ is being used, as it appears to be here, as simply equivalent to ‘real’ or whatever is not ‘as if’. But of course ‘as if’ emotional memories, assuming they exist, are occurrent experiences. There is, however, a more general puzzle here. Why is Debus happy to allow ‘as if’ proprioceptive memories, but not allow ‘as if’ emotional memories, particularly if emotions are held to be partly constituted by perceptual (including proprioceptive) states? She doesn’t really say, and there are few arguments to be found in the literature.

One possible answer is given by Teroni (2021), who notes that emotions are essentially valenced; they are intrinsically pleasant or unpleasant. This, he claims, raises a potential problem for ‘as if’ emotions, since it is hard to make sense of the idea that one is in a state that bears a striking phenomenological similarity to a (un)pleasant state, but where it is only “as if” one were (dis)pleased. Roughly, the thought is that valence is an experiential property and there is no ‘appearance vs. reality’ distinction for these properties. Thus, to be phenomenologically similar to a (un)pleasant state would imply that it actually hurts or feels good, but the ‘as if’ locution seems to deny that this is the case.

Another response, though explicitly directed at imagined rather than remembered emotions, is given by Currie and Ravenscroft (2002) who claim that emotion has no imaginative counterpart, because imagination is in their words ‘transparent’ to emotion: “emotions are peculiar states in that they are, so to speak, their own counterparts… when I put myself imaginatively in the position of someone being threatened, it is genuine fear I come to experience, not an imagination-based substitute for fear.”(p.159) They contrast this purported ‘transparency’ with the example of pain, contending that when we imagine feeling pain we do not as a result actually feel pain.

Obviously, this is not an argument so much as an appeal to our phenomenological intuitions and it seems perfectly legitimate to ask why it seems, if it does, that you can imagine feeling pain but not have imagined emotions. Interestingly, Currie and Ravenscroft don’t say much more about this, and they spend their time giving arguments in favour of the reality of our emotional responses to imagining – as in the case of fiction-directed emotions – rather than arguing against imagined emotions as such. But the state we are concerned with is that of imagining or remembering an emotion, not having an everyday occurrent emotional response to remembered or imagined content.

Indeed, it seems as though various authors may simply have conflated these two different phenomena, which might well explain the focus on fiction-directed emotions in the case of imagining, or in Debus’ account the focus on autobiographical past-directed real emotional reactions to remembered content rather than the emotionally-laden remembered experiences themselves. So these positions, curiously, both sidestep the phenomena we are interested in, rather than targeting them directly, and hence both manage to leave untouched the reality or otherwise of ‘as if’ emotions.

Nonetheless, this ‘transparency’ view of emotion suggests one possible assumption underpinning these sceptical views. Roughly, the thought might be that if a state feels like an emotion, then it is an emotion, since there are no other ‘fainter’ states that differ sufficiently in quality or quantity from a genuine occurrent emotion such that we could count them to be a different kind of ‘as if’ emotional state. In other words, if we can’t tell the difference between, say, real fear and ‘as if’ fear, then there is no good reason to make room for ‘as if’ states in our ontology.

However, this looks like a very dubious conclusion, because even if it is difficult to demarcate clearly the phenomenology of remembered or imagined emotional states, it is difficult to deny that there seem to be such experiences and that they do differ in important ways from ‘real’ occurrent emotional states. Moreover, if we take Teroni’s point about the supposed transparency of valence, then the example of pain is problematic, for pain too is essentially valenced, but then what sense can we make of ‘as if’ pain, if it is not genuinely painful? If we can make sense of it, as indeed it seems we can, and in a fairly clear phenomenological way, then there seems to be no obvious reason for thinking we cannot make sense of ‘as if’ emotions.

Again, therefore, it is not obvious what the purported asymmetry between perceptual memories (or imaginings) and emotional memories (or imaginings) is supposed to consist in. Indeed, even if Teroni’s point about valence is correct, emotions have other aspects that could vary in their ‘as if’ condition. Looking closely at these other aspects, as we will do below, should then help us differentiate, phenomenologically, ‘real’ from ‘as if’ emotions.

Before doing that, however, it is worth noting that according to these sceptical views we should also be dubious about experiential memory and experiential imagining more generally, since experiences too are often valenced or occurrently felt when we ‘relive’ them or imagine them. When it comes to experiential memory, it looks difficult to deny that the phenomenology resembles that of the original experience to some degree and in some ways. Yet one is certainly both remembering this original experience experientially and clearly not now having the actual original experience.

A helpful illustration of the irreducibly experiential nature of emotional memory is given by nostalgia. There is no strong consensus about what nostalgia is, but a rough-and-ready definition would refer to it as a mixed ‘bittersweet’ emotion that is characterised by something like a sad reflection on or wistful yearning for an irretrievably past positive experience. (e.g. Batcho, 2020; Wildshut & Sedikides, 2020). The sadness here is both caused by and directed at this irretrievability, and it is not just (or not necessarily) the irretrievable object or event or person or situation that was positively valued that the sadness is about, but that of the positive experience itself. For, after all, those objects and persons might not themselves be irretrievable; what is irretrievable is the past experiential state directed at them. Arguably, therefore, the pleasant experiential aspect of a nostalgic memory is inseparable from the sadness about the irretrievability of the pleasant experiences constituting this content.

In this way, the remembered happiness is transformed or coloured or infected by one’s current wistful state, which now involves a recognition of irretrievable pastness that the original happy state did not possess. This remembered happiness is thus now ‘layered’ by an awareness of its pastness and irretrievability, and this layering helps constitute the present nostalgic memory. As Goldie (2012) puts it, the feeling that one now has about what one remembers can infuse the episodic memory itself, so that the ‘feeling can be invested both into the content of what is remembered, and into the way of remembering what is remembered.’ (52) Thus, since nostalgia targets remembering an experience, it requires experiential remembering and hence presupposes affective memory.

The very possibility of the emotional state of nostalgia, at least as characterised, thus undermines the kind of scepticism about remembered emotion defended by Debus. On her view, for example, our emotional memories must themselves be affectively neutral – if they are not just in her terms ‘occurrent’ emotional responses – but this just does not seem to do justice to the affective presentation of the remembered emotional content, the emotional way in which they are remembered.

On the one hand, nostalgia is not merely an occurrent experience of sadness in response to the semantic memory that I was happy, nor to the perceptual memory of the situations or objects or persons that are remembered. If this were the case, the prevailing occurrent emotional response to this content would, on her view, be sadness. But nostalgia is, unarguably, a genuinely mixed but homogenous bittersweet emotion. On the other hand, nor does it seem to be a simple combination of occurrent happiness caused by this memory together with an occurrent sadness that the experience remembered is past. If this were the case, it would result in a non-homogenous, unstable mixed emotion, with each emotional component focussed on different intentional objects and our attention oscillating between them. Instead, as I have just described it, the sadness necessarily involves a perspective on that happiness as lost and has as a constitutive part that remembered experience of happiness.

By way of summary, then, not only have we not yet seen any good reason to deny the existence of ‘as if’ emotions, either remembered or imagined, the sceptical views examined above fail to directly target the phenomena we are interested in at all.

4 Affective memory and emotional perspectives

In one of the few explicit defences of affective memory, Arcangeli and Dokic (2018) argue that it involves a particular way of remembering that depends on certain rational relations between three different perspectives we can adopt when remembering:

P1. The perspective of the character (the subject represented in the memory e.g. my earlier self’s pleasure in swimming).

P2. The perspective of the narrator (the way of representing what is remembered e.g. viewing with irony that joy which ultimately would not last very long).

P3. The perspective of the author (the remembering subject e.g. myself, remembering now my joy in swimming last year).

In brief, they claim that we have a case of affective memory whenever P2 rationally ‘constrains’ P3. The account is best illustrated by one of key examples they provide.

4.1 Example – maria

“Last night Maria was driving on the highway heading home. As she was very tired, she fell asleep on the wheel for only a few seconds. As a consequence, her car was about to hit the crash barrier, but fortunately she was soon back in control and her trajectory adjusted at the last minute. When Maria [author] now remembers the scene, she represents it [from the perspective of a virtual narrator] as an extremely dangerous episode. However fear was not part of her [the character’s] original experience. On the contrary, she [author] remembers [the character] having been relaxed and sleepy, not afraid of hitting the crash barrier. Fear is not part of her [author] present experience either. Maria [author] is not now afraid of hitting the crash barrier. On the contrary, she [author] is relieved by the fact that a serious accident has been avoided. Here, all three emotional perspectives seem to be distinct. The author’s perspective (P3) involves relief, the virtual narrator’s perspective (P2) involves fear, and the character’s perspective (P1) involves calmness and peace of mind.” (pp.147 − 48).

Maria’s relief, they claim, is a “rational response to the virtual narrator’s fear. What makes Maria’s memory affective is the fact that it involves emotionally constrained imagination.” Interestingly, as they acknowledge, this example also shows that “a given memory can be affective even if the remembered situation did not involve the character’s having an emotion.” (p.149).

“When the narrator’s perspective (P2) is emotional and constrains the subject’s emotional perspective (P3), there is a special emotional way of remembering an event. On this view, memories are affective when the narrator’s perspective is emotionally loaded and directly affects the subject’s emotional perspective, independently of whether a past emotion is part of what is remembered.” (p.149).

This account is rather perplexing. First, it does not obviously succeed in avoiding the sceptical claim that these are cases of occurrent real emotions directed at remembered contents. Second, this emotional way of remembering does not even seem to meet the minimal criterion for explaining the phenomenon we are interested in, which is that of a distinctively affective state. For, as we have just seen in this example, P2 can rationally constrain P3 even if the remembered situation did not involve an emotional response (since Maria at the time was asleep).

So it looks like this theory too is in danger of simply missing the phenomenon we are interested in, and to that extent we could safely put it aside. Yet this would be too quick, for a crucial part of the account is that the affective work, so to speak, is being done by the “emotionally loaded” narrator’s perspective. It is thus helpful to see why the focus on perspective is significant, and how the way it is used here is problematic.

The empirical literature on memory and emotion distinguishes ‘field’ memories from ‘observer’ memories. (Nigro & Neisser, 1983; Holmes & Matthews, 2005; LaBar & Cabeza, 2006; Libby & Eibach, 2011; Sutton, 2014). The former are memories from the first-person perspective. For example, when I remember swimming in the Ionian Sea last summer, some of my memories will be of how things looked and felt from my visual and bodily perspective at the time. ‘I’, however, will not appear as an object in such memories, but I will be an object in observer memories, which capture things from the third person perspective. So, for example, in such a memory I may ‘see’ myself from above, swimming in the sun. This distinction is also applicable to imagined scenarios, where I can choose to imagine from a field or observer perspective.

In the account given by Arcangeli and Dokic the key examples involve what seem to be observer memories, because the emotionally laden narrator’s perspective is what constrains the subject’s emotional response. But two obvious problems lurk here. First, how does the account explain cases where there is no obvious observer memory? For it seems that I can remember now the situation or object or event that I experienced earlier by remembering it solely from the field perspective of my earlier self. If I remember the feeling of pride when being presented with the award that day, for example, I might see myself from the observer perspective accepting it, but I might instead simply remember the presenter, the setting, and the crowd from the first-person point of view (of my earlier self). So even the subject’s perspective need not itself be an observer memory when it simply involves recalling the character’s first-person experience, the field perspective, which after all is/was one’s own. Second, how is the account supposed to deal with cases of affective field memory, such as the one just given, that involve no narrator perspective, since they require only the remembering subject’s perspective and that of the character’s i.e. the subject’s earlier self?

In the example of Maria, the character has no experience at all since she is asleep, so the fear of the narrator is not itself a memory and hence a fortiori not an affective memory that the subject has. More importantly, we might further wonder whether there is a distinct narrator perspective here at all, rather than just the subject’s post-hoc perspective on the event, which involves something like imagined fear and maybe also occurrently felt ‘real’ relief.

In short, not all emotional memories seem to require a narrator perspective, including even those where an observer memory is involved, for that of the remembering subject seems to suffice. The theory of affective memory being offered here is simply too contrived, proposing unnecessary extra perspectives were none are needed, and thereby failing to account for what seem to be the most straightforward and typical cases in which the remembering subject simply recalls the original field perspective. Indeed the theory treats even such basic cases in a way that seems to mischaracterise the phenomenon for, counterintuitively, the affective nature of the memory itself is not what makes the memory affective; it’s rather the role that this memory (whether affective or not) plays in constraining the remembering subject’s perspective.

Rather than being a rational constraint, what seems to be required is something more like an affective colouring. As in the case of nostalgia discussed above, the remembering subject’s emotional experience is coloured by the remembered emotional experience. What explains the remembering subject’s emotional state is, in part, the irreducibly affective nature of the memory. And yet it is also important to note that the emotional ‘colouring’ can go both ways. As Goldie (2012) argues – in the context of exploring the nature of dramatic irony in autobiographical memory – what you now know, think, or feel may also infuse the memory of a past event such that, in effect, you remember it as you now feel it. As such, where one remembers from an observer perspective, this can incorporate and affect the field perspective.

This points to an important way in which philosophers have been operating with a mistaken picture of the role of emotional perspective in memory. In these discussions the paradigm examples of remembering – and indeed imagining – are almost always visual, and the most significant feature here is the distal and ‘external’ nature of visual perception: the fact that in our experience vision seems to present an ‘external’ world to us more or less transparently. And visual imagery, for all of its differences, seems to mirror this, presenting us with an ‘as if’ experience of the external world. That is, the content of visual imagery is experienced as if it is of something external. (See Nanay, 2015)

The field/observer distinction we examined earlier concerns the spatial nature of the visual imagery involved in all of the central examples of episodic memory, and yet we are trying to understand not visual memory as such, but rather emotional memory. Since all emotional memory involves first person affective experience, however, this essentially spatial distinction between field and observer memory does not obviously apply to it, or at least not in the same way. When I remember now feeling happy in the past, I might take an observer perspective in remembering myself swimming in the Ionian Sea, but the happiness I remember experiencing is experienced now, in memory, from the inside, from a first-person perspective.

Sutton (2010;, 2012;, 2014) makes a similar point when talking of adopting perspectives in imagination, showing that, in respect of the internal embodied, proprioceptive or kinaesthetic perspective, many athletes and dancers deliberately remember or imagine their own performances ‘from the outside’ (i.e. from an observer perspective) visually, even while they actively cultivate a rich subjective sense of their movement activity. That is, they imagine the experience of movement from the inside whether or not they adopt an observer perspective on themselves as well. The internal emotional perspective can coincide with either a field or observer visual perspective. So when we remember emotionally, whether we adopt a field or observer perspective, the original emotional experience will be remembered, so to speak, first-personally by the remembering subject.

This is important because it demonstrates that the focus on visual imagery in these discussions leads to a misleading way of thinking about the nature of affective memory or affective imagining. For visual imagery is not itself affective, and the affective nature of ‘as if’ emotion states is experienced in an essentially ‘internal’, first-person, embodied way. It is partly this focus on visual memory and visual imagining, I suggest, that has blinded philosophers to the existence of a quite different type of imagery; namely bodily imagery, the importance of which will be outlined below.

5 Imagining emotions

Goldie (2012) does not himself explicitly defend the existence of distinctively affective memories, which he seems to take for granted. Yet he does put his account of how different perspectival and emotional layers can affect each other to work in defending the existence of imagined emotions, primarily by highlighting structural similarities between the operation of emotional perspectives in imagination and in memory. To borrow one of his examples: I might remember now, with shame, making a fool of myself while singing drunkenly yesterday, where the feeling of shame now colours, but doesn’t completely conceal, the original carefree joy I felt. Likewise, I can imagine making a fool of myself, and feel something like occurrent embarrassment at the image, although in my imagination, from the first-person point of view of my imagined self, I find what I am doing amusing. Analogously to affective memories, imagined emotions require us to imagine ‘from the inside’ undergoing some emotional response.

We have already seen Currie and Ravenscroft’s notion of ‘transparency’ being directed, supposedly, against there being imaginative equivalents of emotional states, and Goldie explicitly targets this by claiming, against them, that “… it does seem possible for me to imagine something threatening, and to imagine feeling afraid of the threatening thing that I imagine, where the experientially imagined fear is part of the content of what I imagine, and not an external response to what I imagine…. imagining being afraid just is imagining having a fearful experience” (p.82) [italics mine].

Making the emotion part of the content of the imagining is certainly one way of highlighting the fact, already discussed, that Currie and Ravenscroft’s view does not really address the issue of imagined emotions, but rather sidesteps it by focussing instead on the reality of our emotional responses to imagined content. It is, however, not exactly clear what Goldie means by having the emotion be part of the content of what is imagined. It is certainly not sufficient to merely focus on the concept of the emotion or merely to imagine that one is afraid, for what seems to be required is some experienced phenomenological resemblance to the non-imagined state.

The question now, therefore, is what more we can say about this imaginative emotional content that explains how we distinguish, phenomenologically, ‘as if’ emotions from their ‘real’ counterparts.

The only comprehensive attempt to provide an account of imagined emotions, of which I’m aware, is that of Loev et al. (2022) who have recently argued that imagined emotions must possess at least some of the phenomenal properties essential to ordinary emotional experiences. They identify three primary ones: phenomenal valence, bodily sensations, and motivation. What then makes the difference between non-imagined and imagined emotional experiences, on their view, is that, first, only the latter is constitutively defined in terms of imagined content that is itself specified in terms of these emotional phenomenal properties. Non-imagined emotions, in contrast, although specified in terms of the same phenomenal properties, are not themselves constitutively defined in terms of imagined content. That is because they are not essentially causally or constitutively linked to such content.

Second, they contend, imagined emotions are themselves voluntary actions, to the extent that imagining itself is a voluntary action, while non-imagined emotions are not, even if they can be responses to what is imagined. Only in the case of the former does the imaginer exert control over the emotional state itself, such that they are as an agent directly responsible for what is felt. And third, following directly from this, imagined emotions have a specific emotion-like state as their aim and intentional object, whereas non-imagined emotions are directed at any emotion-inducing intentional object that gives rise to independently specified emotional consequences.

The role of intentionality here is complex and will be explored more fully in the final section. As the authors note, whereas non-imagined emotions are understood to be reactions to a temporally prior intentional object, it seems possible that imagined emotions co-exist with or even exist without such an object. This is presumably because one can recreate the phenomenal properties of an emotion in imagination, to some extent, without requiring an intentional object. That is, it seems possible to imagine, for example, being in a state of anger by re-recreating how this state feels, in terms of its valence and bodily profile and perhaps also action tendencies, without (visually) imagining any particular object or person or situation at which it is directed. This is an important point, to which I will soon return.

Nonetheless, they suggest that such objectless imaginings will be less vivid, long-lasting, temporally coherent and specific, partly because emotional valence occurs primarily in relation to relevant properties of the stimulus. As such, objectless imagined ‘as if’ states will be more difficult to secure for complex emotions like nostalgia, and for richer and more detailed emotionally relevant scenarios. To fully imagine the feeling of being the angry Napoleon upon his defeat at Waterloo, they note, requires further ‘scaffolding’, such as attending to bodily sensations of exhaustion and frustration and imagining the ‘rain and mud of Waterloo’.(p.7).

There is much that is right about this account of imagining emotions, but it remains rather incomplete. In particular, although they talk about the phenomenal similarities between imagined and non-imagined emotions, the phenomenological differences between them are glossed over rather briefly in favour of the definitional or metaphysical differences we’ve just seen.

The authors do note that we distinguish imagined from real emotions in just the way we distinguish states of imagining from non-imaginative states; namely, by simply being aware that we are indeed performing the mental action of voluntarily imagining. That is, as noted earlier, there are various cognitive and metacognitive mechanisms – an awareness and feeling of agency, a feeling of pastness, intention and so on – in virtue of which we (normally) know which state we are in. Even if we accept that these mechanisms result in feelings with their own phenomenological profiles, however, I suggest that this does not provide a complete account of the phenomenological differences between ‘real’ and ‘as if’ emotions, for what is particular to the affective emotional phenomenology of these states is still missing. Nor does it offer an explanation of how we imagine emotions, of what precisely is involved in enabling us to ‘conjure’ or ‘recreate’ these affective states.

The authors do promise to give an account of how the features of affective imagining are ‘concretely realized by our minds’ (p. 6), arguing that our ability to engage in emotional imagining plays a crucial function in emotional regulation and affective forecasting. Although this is no doubt a highly plausible functional explanation of the existence of these capacities, and something I will return to in the next section, we still lack the promised account of how the mind does in fact imagine emotion. I suggest that the missing ingredient is what I call ‘affective bodily imagery’.

6 Affective bodily imagery

In affective memory and affective imagining, I contend, the content of our ‘as if’ emotions will be, and appear to us to be, imagistic rather than perceptual. This, I suggest, is the best way of articulating Goldie’s idea of emotional content, and the most important structural and phenomenological similarity between emotional imagining and emotional remembering is that both employ the same ‘affective bodily imagery’. What makes ‘as if’ emotions phenomenologically distinctive and different from ‘real’ emotions is the presence and nature of this imagery. The phenomenological and metaphysical differences between remembered and imagined emotions, then, will be whatever it is that in general makes the difference between imagining and remembering.

As I noted earlier, there are many different theories of what emotions are, but many contemporary emotion theories allow some kind of central or important role for bodily feelings, while disagreeing on the exact role. Whether emotions are taken to be partly constituted by bodily feelings, somehow causally dependent on them, normally composed in part by them, or even identical to them, and whether feelings and emotions can be unconscious, are all subjects of immense debate. (e.g. Contrast Prinz 2005 and Damasio, 2010. See also Deonna and Teroni 2012; Lowe & Tiemke, 2011 for good overviews) Exploring this further would take us well beyond the scope of this paper but, as I said earlier, I do not think I need commit to any particular view here, since the account I develop below is broadly compatible, or can potentially be made so, with all of these ‘feeling’ theories.

Depending on how exactly bodily feelings fit into one’s favoured theory, the exact role of bodily imagery in any remembered or imagined emotional state will differ accordingly. So, for example, on an appraisal theory, an experience would be of an ‘as if’ emotion if the typical suite of bodily feelings characteristic of that emotion figures imagistically and perhaps also connects up in the right way to other imagined or remembered contents, such as the intentional object of the emotion. On a constructivist theory of emotion, an as’ if’ emotion will involve the relevant bodily imagery being interpreted as being that (‘as if’) emotion. On the attitudinal theory, the relevant bodily imagery will be that of some holistic world-directed bodily attitude.

It is not always clear which bodily feelings are being referred to by these emotion theories, but at least three types (mentioned earlier) seem to play central roles in most emotional states: interoceptive, proprioceptive, and feelings of action tendencies. Proprioception requires little explanation and incorporates many types of bodily movement and position, including that involved in posture, gesture, muscle contractions, facial expressions, and actions. In terms of interoception, the empirical and philosophical literature clearly associates emotions with particular changes in the activity of the visceral motor (autonomic) system, such as increases or decreases in heart rate, cutaneous blood flow (blushing or turning pale), piloerection, sweating, and gastrointestinal motility can all accompany various emotions. (Damasio, 1994, 2004; Prinz, 2004; Friedman, 2010)

Perhaps one of the most important types of bodily feeling is that involved in an awareness of action tendencies, what one is physiologically primed to do in a situation. These may include for example appetitive or approach behaviours, and defensive or avoidance behaviours. Indeed many theorists hold action tendencies to be somehow essential to emotion (Frijda, 2004 and, 2007; Ekman, 2003; Ellsworth & Scherer, 2003; Lewis, 2005; Lowe et al., 2007; Laird, 2007). Although the relationship between emotion, action tendencies, proprioception, and interoception is extremely complicated (Lowe & Tiemke, 2011), for our purposes we can simply refer to the combination of all these types of bodily feeling as ‘somatic-visceral’.

Recall that Currie and Ravenscroft think that, unlike emotions, pain experiences are easily imagined and readily distinguishable from real pain perceptions. What enables us to distinguish ‘as if’ from ‘real’ pain experiences? Apart from the fact that we are normally aware of when we are imagining or remembering an experience, as opposed to actually having the experience, via the mechanisms alluded to above, there seem to be clear phenomenological differences between the two. ‘As if’ pain is fainter, less intense, and essentially sustained by our attention and intention; but, I suggest, ‘as if’ pain is not just like an image of real pain, it is an image of real pain. We can imagine and remember having pain experiences in virtue of having the relevant imagistic experiences of bodily feelings: an interoceptive-like awareness of painfulness, a proprioceptive-like awareness of the bodily location, and a feeling of an action tendency, such as of wanting to withdraw from the source of the pain or wince from the pain itself.

If you can have imagistic experiences of painful bodily feelings, there’s no reason to think that you cannot have them for the various bodily feelings present in other states such as hunger or thirst, sensations of heat and cold, feelings of being tickled, or being immersed in warm water. You should have no problem in now summoning up, in memory or imagination, a ‘feeling’ of hunger or thirst or pain or nervous butterflies. But just such bodily feelings also arguably figure in cases of anger or fear or joy or any number of other emotions in which such feelings are clearly enough demarked.

There is a great deal of empirical evidence for the existence of motor and proprioceptive imagery. (Shenton, J. et el. 2004; Mulder, 2007; Ganea & Longo, 2017). Imagining moving your hand, for example, seems to involve the same neural mechanisms as those involved in really moving your hand, and yet clearly the imagined experience feels distinct from the real experience, in just the ways, I have indicated, imagistic experiences feel different from ‘real’ experiences. Interestingly, there is almost no discussion of visceral bodily imagery in the philosophical or empirical literature. (See Vianna et al., 2008) We can, however, draw on the extremely influential account of bodily feelings developed over a number of years by Damasio (1994, 2004, 2010; Bechara and Damasio 2005) that underpins his view of emotions, for his theory has the resources – albeit not explicitly so deployed by him – to explain the kind of affective bodily imagery I wish to defend.Footnote 1

Roughly, for Damasio, conscious emotion experiences involve the somatic-visceral feelings that result from and help represent certain homeostatic changes to internal bodily states. He argues that the brain represents these internal states by what he calls the “body loop”. In addition to the primary inducers of emotional states, such as external perceptual stimuli, there are secondary inducers which comprise affective responses caused by the recall of past events or anticipation of future states. According to Damasio, therefore, the brain also utilises what he calls an “as-if loop” which involves something like the representation of an expected or anticipated, but not actual, body action: ‘It is ‘as if’ the body had really been changed but it was not.’ (Damasio, 1999: 79–80).

In short, on this view, in addition to representing the actual bodily states constituting emotions that result from exteroceptive (i.e. perceptual) experience, the brain also makes use of an ‘as if’ loop to represent bodily states – including action tendencies – that are rather the result of inner representations, such as we find in memory and imagination. (See also Lowe & Tiemke, 2011) Indeed, although he is referring rather loosely to feelings as neural representations, Damasio sometimes uses the term ‘image’ when talking about the role of feelings in emotions. He says, for example, that ‘proprioceptive processing entails images of specific body components such as joints, striated musculature, some viscera’ (2010: 76) and as far as the body is concerned ‘feelings are images of actions rather than actions themselves; the world of feelings is one of perceptions executed in brain maps’ (2010: 110).

Nonetheless Damasio, and others following him, do not really say anything about the phenomenology involved in such ‘as if’ feeling states, and this is potentially problematic. In one of the few studies that seem to touch on the idea of bodily imagery, for example, Vianna et al. (2008) hypothesized that vivid examples of emotional imagery would be accompanied by strong increases in gastrointestinal (GI) and sympathetic nervous system (SNS) activity. They found that SNS and GI changes, measured by skin conductance and electrogastrogram, respectively, correlated positively with subjective ratings of arousal during the imagery. The problem is that although imagistic emotional experience is involved in this study, it is not clear whether the resulting feelings are ‘real’ or ‘as if’, since the bodily changes measured here are real ones.

It thus seems possible that, on Damasio’s view, the representations of the as-if body loop could result in the very same bodily feelings that would occur as a result of the perceptual representations of the non-as-if loop. If this were the case, then we could distinguish ‘real’ from ‘as if’ affective states not by intrinsic phenomenological differences but only by other perceptual, cognitive and metacognitive markers in virtue of which we could tell, for example, whether we were really perceiving or merely imagining some object or bodily state. One could perhaps argue that the overall phenomenology would differ, particularly if one were convinced of the existence of cognitive phenomenology. (Bayne & Montague, 2011) But even this would allow the sceptic to get a foothold on the idea that there were no genuine emotionally relevant phenomenological differences between ‘as if’ and ‘real’ emotions.

The first thing to note is that this study talks of emotional imagery rather than affective bodily imagery as such, and was explicitly looking at real bodily changes and hence ‘real’ bodily feelings in response to ‘emotional imagery’. More importantly, however, even apart from what seem to be evident phenomenological differences of the kind we have discussed, I think there are plausible functional considerations that help explain why we should expect marked phenomenological differences between ‘real’ and ‘as if’ affective states, and why it makes sense to think of the latter as imagistic in nature.

The mechanisms involved in monitoring and representing the bodily states constituting emotional feelings should be expected to manifest differently, depending on whether those states are caused by ‘external’ perceptual input or by the internal states registered by the ‘as if’ loop. It is after all clearly important for us to be able to distinguish them. The most efficient way of manifesting this difference, arguably, would be in terms of a difference in the bodily feelings arising from those different ‘bodily’ states. Amongst other things, it should thus be entirely unsurprising that the ‘as if’ states, in being disconnected from actual actions and perceptions, will normally feel like attenuated imagistic versions of the ‘real’ states they resemble. After all, in such cases we aren’t being directly confronted with something in the external environment bearing immediately upon us and forcing us to act.

This brings us, however, to a further problem. In lieu of active interoceptive attention, visceral feelings often seem peripheral at best to such experiences, and as a matter of empirical fact they can be difficult to detect, let alone delineate. Moreover, some people are naturally much worse at interoception than others. (Kriegel, 2015b: 136; Larsson et al. 2021; Garfinkel et al., 2022; Sukasilp & Garfinkel 2022) One might thus object that insofar as we often seem to be only dimly aware, if at all, of visceral feelings in ‘real’ emotional experiences, it is highly unlikely that we would be aware of them at all in attenuated ‘as if’ versions of such experiences.

Indeed, even if we are inclined to think that visceral feelings are at least sometimes more prominent than this suggests, it seems possible that in certain cases we might confuse ‘real’ visceral feelings for ‘as if’ ones, particularly insofar as it is possible to have and experience both simultaneously. (Cf. The ‘Perky effect’ See Dijkstra et al. 2022) For example, in a case of imagined fear, one’s affective bodily imagery might help give rise to actual feelings of fear. In this light, it is our purportedly poor ability to distinguish supposedly different feelings phenomenologically in such cases that has no doubt played a significant role in philosophical views denying the existence of distinctive ‘as if’ emotions, as discussed above.

More generally, it has been noted that one problem with ‘feeling’ theories of emotion is that there do not seem to be enough bodily states or bodily feelings – or the feelings themselves do not seem to be sufficiently specific or fine-grained – to differentiate all of the emotions that we distinguish conceptually. On the one hand, different emotions seem to have closely overlapping somatic characteristics: for example, fear and anger are both marked by arousal, accelerated heart rate, sweating, and so on. On the other hand, although it might seem that, for example, jealousy and anger are as different from each other as sadness from nostalgia, it is difficult to articulate extra bodily feelings that might explain the difference. Further, some so-called ‘intellectual’ or ‘contemplative’ emotions do not seem to be accompanied by identifiable bodily feelings at all. (Charland, 2005; Deonna & Teroni, 2017; Colombetti & Harrison, 2018; Mitchell, 2021)

In addition to bodily feelings, therefore, we normally also require recourse to the material and intentional objects of our responses in order to differentiate and demarcate ordinary emotions. We differentiate fear from anger, for example, partly in terms of the perceptually experienced and cognitively appraised causes and objects of these states. As Kriegel (2015b) points out, bodily feelings are obviously insufficient by themselves to account fully either for the specificity of many contextually dependent emotional responses, or for the identity and phenomenology of complex emotions such as indignation, pride, or nostalgia. On this basis, he has argued that although proprioceptive feelings may play a role in emotional experience, emotional phenomenology can’t consist solely in bodily phenomenology, and he claims that cognitive and conative forms of phenomenology are more important. (Kriegel, 2015b)

For what it is worth, I happen to broadly agree with Kriegel’s view that emotional phenomenology is a complex mix of different types, but I suggest that he thinks bodily feelings are relatively peripheral to emotional experiences because he overlooks the important distinctions between the various types that we have been discussing.

First, even if we allow that visceral feelings are often relatively insignificant or unclear in emotion experience, arguably proprioceptive feelings and feelings of action tendencies are much more prominent. The tendency to, say, strike out physically in extreme anger will clearly be attenuated in even a very vivid memory of such anger that involved actually striking out, but it is indubitably there nonetheless. In cases of ‘as if’ emotion, these feelings of action tendencies will often be present in the sense that, for example, we now seem to feel the physical urge again, but in the form of proprioceptive imagery or in the imagistic feeling of wanting to or being about to strike. Indeed, what Kriegel means by conative phenomenology can arguably be captured to some extent in this way. We feel the desire-like wanting to strike out in anger, or to withdraw in shame.

Second, most cases of affective memory and imagined emotion will indeed involve many phenomenological components, such as sensory-perceptual imagery, and the cognitive phenomenology involved in the beliefs, attitudes, desires, evaluations and other states relating to the situation that the ‘as if’ emotion is directed at. It is important to emphasise, however, that in the case of ‘as if’ emotions, as discussed earlier, it seems possible to remember or imagine the relevant bodily feelings independently of remembering or imagining any particular object or cause of such feelings.

Crucially, the fact that we can do this provides a further reason to believe in the existence of affective bodily imagery because these experiences are not tied – causally or intentionally – to objects in the same way that ‘real’ emotions normally are and hence not reducible to ‘real’ emotions in the way that sceptical views would have us believe. Indeed, this feature of affective imagery, its capacity to be open to some extent to interpretation, is itself a well-discussed feature of imagery in general, namely that the meaning or reference of imagery is in some way dependent upon – or perhaps even entirely determined by – one’s intention and interpretation. (Arcangeli, 2019, 2021; Kind, 2017).

This is important because it points to a plausible evolutionary-functional explanation for our ability to invoke such imagery, namely its role in empathetic understanding and ‘affective forecasting’. The former needs little discussion here, but the latter refers to our ability to make future-oriented decisions by being able to anticipate the affective impact that certain events will have on us by allowing us to ‘experience’ these events in imagination. (Goldman, 2006; Wilson & Gilbert, 2005; Loev et al., 2022).

Although both empathy and affective forecasting will normally involve imagining specific situations and objects that cause the relevant ‘as if’ emotional reactions to them, as Loev et al. note, I think they underappreciate the significance of the fact that the bodily feelings they refer to can be separated from such objects. For one clear advantage of being able to experience objectless affective imaginings, is that they enable us to feel a familiar emotional response to an unfamiliar object. In order to understand, for example, the political emotions of someone from the opposite side of the political spectrum, I might take the feeling of a familiar positive emotion such as love and try to associate this feeling with an object I would not normally have this emotion towards, such as the Monarchy or Nation State. Such an ability, I think, is clearly centrally involved in both empathetic understanding and affective forecasting.

There is even neurobiological evidence for our ability to (re)construct such affective responses, as Gerrans (2018) writes:

‘when we imagine or reflect on an experience we do not re-create a complete representation of the experience. Rather we (re)construct the affective response to that experience. In other words we represent not what happens to us but how it matters to us. We empathize with our past or future self by activating circuitry that represents not body state per se, but the significance of body state. To do this we need to be able to control activity in the AIC independently of perceptual/sensory elicitation. In order to do this, we exploit an adaptation that allows us to regulate AIC activity independently of its normal interoceptive (bodily) afferents.’ (169).

Given the central role of affective bodily imagery in such ubiquitous and important states as affective memory and emotional imagining, therefore, it is unsurprising that these states have the distinctive phenomenology they do primarily in virtue of this imagistic bodily phenomenology.

Moreover, it is easy to see why this phenomenology is more central to ‘as if’ emotions than cognitive or sensory-perceptual phenomenology. For, on the one hand, states of non-emotional memory and non-emotional imagining will share their non-emotional imagistic features with their comparable ‘as if’ affective counterparts, and hence, for example, will have similar cognitive or sensory phenomenological profiles. The visual image of the Colosseum in my emotionless memory will – all things being equal – be identical to the visual image of the Colosseum that I happily remember. On the other hand, ‘real’ emotions differ phenomenologically from ‘as if’ counterparts in ways that point to the imagistic nature of the latter.

Returning to the question of what makes affective memory different from affective imagining, I see no reason to think that they differ in terms of their imagistic affective bodily phenomenology. As intimated earlier, therefore, the overall phenomenology of affective memory would seem to differ from that of emotional imagining only in virtue of differences that not emotionally affective.

One might wonder, nonetheless, whether remembered emotions are more specific or fine-grained, being bound up with very particular (cognitive-perceptual) contexts in the past. (Cf Kriegel, 2015b) Insofar as both states rely on imagery, however, there seems to be nothing essential to imagining that precludes there being an equal level of specific cognitive-perceptual detail, even if it may be a psychological fact that imagining normally requires a degree or kind of effort in constructing imagery that remembering does not. (e.g. Dokic & Martin, 2015) Such effort, and the feeling of effort, however, would seem to be a feature of the metacognitive phenomenology involved in imagining as such, rather than an essential part of the affective bodily phenomenology the two states have in common. Moreover, it seems plausible, phenomenologically and from the point of view of cognitive efficiency, that this affective bodily imagery is the same in each state, and hence that emotional imagining utilises memories of bodily feeling. Such a discussion would take us beyond well beyond the current one, but if this could somehow be independently demonstrated, it might provide a reason to side with certain ‘continuists’ in the debate about the relationship between memory and imagination. (See De Brigard, 2017 for an overview)

Finally, can there be ‘as if’ emotions without affective bodily imagery? I have aligned my view with feeling theories of emotion and have thus assumed that there aren’t conscious emotional experiences in which feeling states don’t figure in some way. This forces me to say, of course, that purported cases of affective memory or emotional imagining which do not possess some awareness of bodily feelings are not really emotional. And indeed, if we accept the account I have given, this should seem quite persuasive. If I claim to remember or imagine feeling, say, nostalgic or proud but cannot detect any affective (imagistic) bodily feeling, these are cases of remembering that or imagining that and hence aren’t the states we are interested in. Indeed, it might even be possible to use this insight about the central role that bodily feeling plays in cases of ‘as if’ emotion to develop an indirect argument for feeling theories of emotion as such. But this would be the task for another paper.