Embodiment and the Perceptual Hypothesis

I. THE PERCEPTUAL HYPOTHESIS

The claim I consider in this paper is that we can sometimes literally see some of each others' mental states. Call this the perceptual hypothesis. Green, for example, argues that 'it is possible literally to perceive the emotions of others'. 1 Stout claims that we 'can literally perceive someone's anger in his face'. 2 Similar claims have been made by several other authors. 3 But say the perceptual hypothesis just concerned how things seemed to us when we looked at others, or what introspection told us about the nature of our awareness of others' mental states. In this light, it may seem relatively uncontroversial that we sometimes see aspects of each others' mental lives. At the same time, it wouldn't seem particularly enlightening.

We often come to be aware of others' mental states without doing any conscious work. We rarely have to put any effort into becoming aware that someone is angry, say. At the personal level, others are given to us as minded. And in most cases, some specific aspects of their present state of mind are also given to us.

As thinkers, we often take facts about others' mental features as our starting point. When we consciously reason about others' mental lives, we do this sometimes by calling on generalisations, sometimes by putting ourselves in their shoes. But even when we consciously reason about others' mental lives, the initial evidence we make use of almost always mentions some of their mental states. Why is Jack angry? Why does Jill believe her glasses are still in the car?

We are also often happy to say such things as that we could see that James was angry, that Jill looked elated, that we could sense Jack's disgust. In this way, we might think that 'the alleged imperceptibility of all psychological states is not in fact a bit of common sense; everyday discourse goes the other way '. 4 All this seems relatively uncontroversial, at least as a description of how things seem to normal subjects, or of the kinds of things normal subjects are prepared to say. But it seems unenlightening, because it is compatible with radically different models both of the mechanisms which cause our awareness of others' mental states, and the epistemic explanation of our knowledge of those states.

Debates about the kinds of mechanism that best explain our awareness of others' mental states tend to revolve around whether our awareness should be modelled as a consequence of inference or simulation. It is commonly accepted that much of the processing which results in that awareness will be sub-personal or tacit; that it will not be the kind of thing that is open to introspection, or that has any distinctive phenomenological character. The aim is to posit a model of these unobserved processes which best explains the phenomenon. The argument concerns the nature of this model. To engage with this debate, the claim that you sometimes see others' mental states will have to engage with these different ways of understanding our sub-personal or tacit cognitive mechanisms. This is no easy task. From a mechanistic perspective, there is little agreement about how to delineate perceptual from non-perceptual mechanisms. On some accounts, this will depend on whether a process meets some or all of Fodor's criteria on modularity. In this sense, both inferential and simulative models could be compatible with calling your awareness of some of another's mental state 'perceptual'. 5 On the other hand inferences, for example, work over states with structured representational contents. If the kinds of processes which result in your awareness of Jack's anger are not best modelled as working over states with structured representational contents, then in this sense, your awareness is non-inferential. In this sense, it may be distinctively perceptual. In any event, there will be no simple move from claims about the natures of subjects' personallevel experiences of others to the claim that their awareness of others' mental states is distinctively perceptual, hence not inferential or simulative. 6 My own focus is on the perceptual hypothesis insofar as this relates to epistemic debates concerning how we can know about each others' mental states. In these debates, it has tended to be assumed that our knowledge is inferential. Call those who accept this Inferentialists, and the view they accept Inferentialism. One question that concerns Inferentialists is whether those inferences are permitted to us by analogy with our own mental lives and behaviours, or because they form part of a theory that would best explain others' behaviour. It is commonly accepted that your states of knowledge can be secured inferentially without your having to do any inferential reasoning. In this way, the relevant inferences may be tacit, sub-personal, or merely 'structural'. 7 As Pryor puts it, you 'can … form beliefs which are supported by "inferential relations" without engaging in any inference '. 8 For the epistemologist, an inferential relation is something that holds between states with propositional content. The Inferentialist presupposes that in order to show how you count as knowing that Jack is angry, we 5 For Fodor's original account, see The Modularity of Mind (MIT Press, 1983). For a recent discussion of the relationship between simulation and modularity, see A. Goldman's Simulating Minds (Oxford UP, 2006), at pp. 101-11. 6 Cf. M. Herschbach, 'Folk psychological and phenomenological accounts of social perception ', Philosophical Explorations, 11 (2008), pp. 223-35, but see also Gallagher and Zahavi's 'The (In)Visibility of Others: a Reply to Herschbach', Philosophical Explorations, 11 (2008) need to point to further propositional states, which don't regard Jack's mental states, and a generalisation that connects these with the knowledge to be explained. In other words, his presupposition is that our basic propositional perceptual states don't concern others' mental lives but just their behaviours, bodily features and so on. Such propositional perceptual states are epistemically basic or primary. States concerning others' mental lives are epistemically dependent on them; secondary.

The warrant of a state with the content that Jack is angry cannot simply be inherited from the warrant of states with contents concerning Jack's behaviours, bodily features and so on. There is an epistemic 'gap' here, epistemic slack. In explaining your warrant, the epistemologist's job is to take that slack in, and this requires him to posit a legitimate inferencehowever probabilitisticwhich would bridge the gap. In order to explain how knowledge of Jack's mental states emerges from states concerning only his behaviour, bodily features and so on, a legitimate inference needs to be posited which connects the latter to the former. 9 Prima facie, at least, there seems no reason to claim that sub-personal or tacit states cannot relate to propositions. So someone engaged in this debate can happily accept that the only thing we are consciously aware of, when we look at Jack, is his anger. As Price put it: I am constantly taking for granted the existence of all sorts of foreign emotions and foreign thinking-processes. I take their existence for granted without the least hesitation or doubt. But this is a very different thing indeed from knowing them by acquaintance. 10 On the grounds that it was subjectively spontaneous, we could happily call your awareness of Jack's anger perceptual. Nonetheless, it might still turn out to be epistemically inferential. 11 In order to engage with the epistemological debate, the perceptual hypothesis needs to be understood as denying that all your knowledge of others' mental states is inferential. It needs to be understood as the claim that sometimes, you can non-inferentially know, on the basis of seeing someone, that they are in some particular mental state. Call those who espouse this view Perceptualists. If your awareness of Jack's anger were secondary, and depended epistemically on a more basic awareness of his behaviour or facial expression, this dependence would make your knowledge of Jack's anger inferential. The Perceptualist is someone who denies that your awareness of Jack's anger must depend epistemically on a more basic awareness of Jack's non-mental features. To engage with the Inferentialist, the Perceptualist must deny the Inferentialist's presupposition.

For these reasons, and restricting my attention to visual perception, I take the Perceptual Hypothesis to be the claim that: [PH] We sometimes see aspects of each others' mental lives, and thereby come to have non-inferential knowledge of them.

Denial of the Inferentialist's presupposition is compatible with the claim that your awareness of others' mental states is dependent on some very complex sub-personal processes. Not all sub-personal processes work over states that could count as having propositional content, for example. So not all sub-personal states could count as the inputs to a legitimate inference on which your awareness of others' mental states epistemically depends.

I will sketch two reasons why we might be tempted to accept the Inferentialist's presupposition. Broadly, these are phenomenological or introspective on the one hand, and physicalist on the other.

From a personal level perspective, it may seem that we can be 'directly acquainted with another person's joy in his laughter, with his sorrow and pain in his tears, with his shame in his blushing', 12 and so on. Personal level reflection on our own mental states, however, may suggest just the reverse. If we take this introspective stance, at least some of our mental states can come to seem constituted by that stance; a stance we can each have only to our own mental states. 13 As Kripke put the intuition, 'pain is not picked out by one of its accidental properties; rather it is picked out by the property of being pain itself, by its immediate phenomenological quality'. 14 When we introspect, the distinction between the act of being aware of our pain, say, and the pain itself, can seem to dissolve. Without a sense of that distinction, the possibility of another having anything like direct access to our pain appears also to dissolve. There seems no distinction between directly accessing that state and simply being in itboth seem constituted by having an experience with a particular phenomenological character. And if you're in a mental state, it's yours. So any state you directly access is a state of you. In this way, it doesn't seem as if you can ever directly access another's mental state. 12 A second kind of motivation for accepting the Inferentialist's presupposition is physicalist. The traditional physicalist claim was that mental states are either type or token identical with neural states. But states of the nervous system cannot ordinarily be seen. They are hidden beneath the skin, and in any event cannot be visibly discerned, at least with the naked eye. But if state P and state M are identical and you cannot see P, you cannot see M.

I find neither of these kinds of motivation for the Inferentialist's presupposition persuasive. I think that in fact, PH is plausible. For now, however, I consider a way of trying to defend PH that I don't think will work. This is the view that we can defend PH if, but only if, we accept that some mental states are partly constituted by structures and processes which are outward, and in this way visible. Call the claim that some mental states are partly constituted by structures and processes which are outward the Embodiment Hypothesis. Call the view that we can defend PH just in case the Embodiment Hypothesis is correct the Embodied Perception Theory. In what follows, I argue that even if we accept the Embodiment Hypothesis, we should not accept Embodied Perception Theory.

In the process of showing how Embodied Perception Theory fails, however, at least some of the motivation for denying the plausibility of PH comes to look misguided. In spite of their differences, both the Inferentialist and the Embodied Perception Theorist share a common assumption about what would need to be the case if you could see others' mental states. The assumption they share should be rejected.

The strategy for the rest of the paper is as follows. In the next Section, I sketch what I mean by embodiment, and the way embodiment is supposed to make plausible the Perceptual Hypothesis. In §III, I make some distinctions concerning the different ways in which you might see things, and how seeing things in these different ways may affect the epistemic explanation of your knowledge concerning them. Over the course of § §IV, V and VI, I argue that depending on how the Embodied Perception Theory is cashed out, the Embodiment Hypothesis is either insufficient and unnecessary to support the Perceptual Hypothesis, or sufficient yet unnecessary. Either way, then, it is not necessary to accept the Embodiment Hypothesis in order to accept PH. I conclude with some remarks about where this leaves Inferentialism.

II. EMBODIMENT

'Embodiment' is a term put to several uses. I mean to focus on just one of them, but I distinguish it from one other. It is sometimes claimed that perceptual experiences may be embodied, in the sense that they may be enactive. To say that a perceptual experience is embodied in this sense is to say that it is constituted in part by our expectations or dispositions to interact in particular ways with the objects we perceive. Call this the Enactive Hypothesis.

It is sometimes argued that PH is plausible on condition that our perceptual experiences are enactive. Gallagher, for example, claims that in social contexts, we have resonant affective and sensorimotor responses to others, and that this fact makes PH plausible. 15 The idea is that such resonant responses both track aspects of others' mental lives and serve partly to constitute our perceptual experiences of others. Hence our resonant responses enable our perceptual experiences to relate us directly to those aspects of each others' mental lives which they track.

In contrast to the Enactive Hypothesis, what I want to focus on is the claim that some mental statesparticularly emotional onesare extended; are partly constituted by extra-cranial and extra-neural states that may 'reach out to the surface' of someone's body. I will call this the Embodiment Hypothesis. As well as involving phenomenological feels, Mitchell Green argues that emotions also involve 'behavioural tendencies, facial expressions, physiological responses, etc.' 16 For Krueger and Overgaard, [emotional] processes consist partly of internal operations-including not only neural operations but a phenomenological profile given to the subject of the processbut also partially consist of publically-perceivable bodily operations that are ultimately also part of its hybrid structure (p. 257).

The Enactive and Embodiment Hypotheses are sometimes treated as two sides of a coin. But they are separable. The truth of the Embodiment Hypothesis would commit us to the claim that some mental states are not fully inside the head, while the truth of the Enactive Hypothesis does not. It is possible to hold that our perceptual experiences are enactive while denying that any mental states reach outside the head. And it is possible to hold that some mental statesincluding perceptual experiencesreach outside the head while denying that perception is enactive.

The line of thought which links the Embodiment and Perceptual Hypotheses is a simple one. Seeing a part of something can allow you to 15 See also Stout, p. 40; McGann and De Jaegher, 'Self-Other Contingencies: Enacting Social Perception', Phenomenology and the Cognitive Sciences, 8 (2009), pp. 417-43. For a slightly different account of the role of enactive perception, see Smith. 16 Green, 2007, at pp. 91-92. see the thing of which it is a part. 17 The Embodiment Hypothesis is that some mental states are partly constituted by structures or processes that are 'outward'. In other words, it is the view that some mental states have parts that in some sense come right out to people's surfaces. But what is outward may be visible. If some mental states have visible parts then by seeing those parts, we might be able to see the mental states of which they are parts. The term 'part' is used very broadly here, to include not just particular physical objects and statesnoses, mouths, grimaces and so onbut also particular events or processes which involve our surfaces arm waving, grimacing, blushing, and so on. The claim that embodied mental states are only partly constituted by what is visible is supposed to block the potential charge of behaviourism. 18 If 'in some cases we perceive emotions by means of part-whole perception', argues Green (p. 89) then 'the claim that we can perceive emotions is compatible with the possibility of their containing components that are not perceptible' (ibid). Embodied mental states can still have insides; they can still have phenomenological qualities. 19 The Enactive Hypothesis was supposed to show how our perceptual states were rich or 'smart' enough to reach out to others' mental states. By contrast, embodiment as it is here understood is supposed to bring others' mental states out to meet us. What I will call the Embodied Perception Theory is that the truth of the Embodiment Hypothesis is both necessary and sufficient to make the Perceptual Hypothesis plausible.

In what follows, I argue against Embodied Perception Theory. I argue that depending on how it is cashed out, the Embodiment Hypothesis is either insufficient and unnecessary for the plausibility of PH ( §IV), or sufficient but still unnecessary for the plausibility of that hypothesis ( §V). So the claim that PH can be defended just if the Embodiment Hypothesis is true turns out to be false.

The claim that you see something by seeing a part of it is most often made in connection with seeing objects or events. We talk about seeing the table by seeing its top; seeing the mouse by seeing its tail. We talk about seeing the race by seeing the cyclists going past; seeing the soup cooking by glimpsing its bubbling away. The plausibility of Embodied Perception Theory is usually set up by noting these kinds of relations.

In Self-Expression, for example, Green claims that just as we can see an apple by seeing its facing surface, so we can see someone's disgust by see- 17 See Green, 2007, at p. 89. 18 This charge has recently been put by e.g. P. Jacob, 'The Direct-Perception Model of Empathy: a critique', Review of Philosophy and Psychology, 2 (2011), pp. 519-40. 19 Cf. Krueger and Overgaard, at p. 245.

ing its expression in their face (pp. 86-87). Krueger and Overgaard claim that just as we can see an iceberg by seeing its tip, so we can see someone's anger by seeing the outward bodily parts of that angerthe flushing cheeks, tensing of hands and so on:

We see icebergs by seeing proper parts of themthe parts above the surface of the water. And this, on the view we are defending, is precisely what goes on in cases of … social perception: we see others' emotions by seeing proper parts of their emotions. We see tips, but we don't see the whole iceberg (p. 255).

Call this the Fossil View. You cannot see fossils when they're completely encased in rock. When they're completely encased in rock, they're hidden. But fossils can be seen, because they are not always completely encased in rock. Sometimes, some parts of a fossil can be exposed. They can come right out to the surfaces of the rocks that contain them. Even when they are encased in rock, fossils can be seen so long as they are not completely encased in rock.

The Fossil View is that our ability to see aspects of each others' mental lives is like our ability to see fossils. When mental states are completely contained within the head, they cannot be seen. But just as a fossil can be seen when some part of it becomes exposed at a rock's surface, aspects of others' mental lives can be seen when some parts of their mental states come out to their surfaces.

I think there are several problems with the Fossil View, both as it relates to how we might see aspects of each others' mental lives, and what would need to be the case if those aspects were to be seen. To understand this we need to address certain complexities to do with what kinds of things we see, how we might see them, and how this affects the epistemic role that seeing such things can play.

III. SEEING AND KNOWING

The Embodied Perception Theorist claims that you can sometimes see another's mental state by seeing its surface parts. In this way, he claims that you can sometimes have non-inferential knowledge about another's mental life. There are a number of ways of interpreting the Embodied Perception Theorist's claim. This is because there is more than one thing we can mean by the claim that you see something. We need to know in what way you are supposed sometimes to see others' mental states, in what way you are supposed to see those states' outward parts, and what exactly the connection is between seeing these two kinds of thing.

When we say that you have seen the apple, the iceberg, or Jack's anger, we might mean this to come in a transparent or an opaque context. 20 In a transparent context, we can substitute terms for what you've seen and preserve truth. What we mean to identify are just the objects to which you aresomehowvisually related. In a transparent context, if you see the apple and the apple is a pippin, you see the pippin. If you see the iceberg and it is currently the largest object afloat in the North Atlantic, you see the largest object afloat in the North Atlantic.

Dretske called this kind of seeing non-epistemic. 21 The fact that you transparently see the apple entails nothing about what you might believe, what you would have warrant for believing, or what you might know. All your transparently seeing the apple or the iceberg implies about you is that you have visually differentiated those objects. 22 Call seeing, in such transparent contexts, seeing n . If you see n the house and that was the house that Jack built, you see n the house that Jack built.

Alternatively, the context in which we claim that you see the apple, the iceberg or Jack's anger might be opaque. In such a context, we cannot substitute terms for what you've seen and preserve truth. The term we use for what you've seen, in this context, is supposed to pick out not just the object that you see, but the way in which you have seen it; what you have seen it to be. When we claim that you've seen the apple, we might mean to select the way in which you have identified the apple that you see n . We might mean to point to the fact that you've seen the apple as such.

The examples of transparent seeing I gave relate you to objects. But there is something essentially featural about opaque seeing. The claim is that not only have you seen n the object, but that you have seen something of that object; its being an apple. So in an opaque context, the claim that you've seen the apple involves the claim that you've identified one of that object's features.

Call seeing, in such opaque contexts, seeing i . I will talk of your seeing i the apple as such. I will also talk of your somehow seeing i the apple. By this I will mean that you have identified the apple in some way, have seen one of its features, while leaving it open which of those features you have seen it to have. Furthermore, because seeing i is featural, I will sometimes talk of your seeing i a particular feature of something, or your seeing i F, where F is the feature of the object you see i it to have. Unlike seeing n , seeing i is epistemic. If you've seen i the apple as such, this is sufficient to warrant a belief, regarding that object, that it is an apple. If you've seen i the apple as such, then the thing you see n is an apple. So if you do come to hold that belief, then in the absence of any defeaters, you will count as knowing that the object you see n is an apple. Likewise, if you see i the red thing as such and you form the belief that it is red then in the absence of defeaters you know, of that thing, that it is red.

With all this in mind, let us return to the Perceptual Hypothesis. Say you come to know of Jack, on a particular occasion, that he is angry. If the Perceptual Hypothesis applies to this situation, then you will have seen i Jack's anger. You will have correctly identified Jack as being angry.

We might think that this is all the Perceptualist's claim need amount to. But there is more than one way to count as seeing i an object's feature. And one of these ways will secure you knowledge of the obtaining of that feature that is epistemically inferential. The Perceptualist needs to ensure that it is not in this way that you see i Jack's anger.

Dretske argued for a distinction between primary and secondary forms of epistemic seeing (pp. 153ff). Say you see that the fuel tank is half full. If you see this fact by seeing n your fuel tank, you see it primarily. If you see this by seeing n the gauge on your dashboard, you see what you do secondarily. As Dretske forged it, the distinction concerns the different ways in which the objects you see n relate to the facts that you see. You see what you do of an object O1 secondarily if your seeing this depends on your seeing n some non-identical object O2. If there is no such dependency if you see what you do of O1 just by seeing n O1you see what you do of it primarily.

But an analogous distinction between primary and secondary seeing can be applied also to how you see i the features of the objects that you see n . Consider the following two cases, in both of which you come to know, of the fuel tank, that it is half full:

[C1] You see n the fuel tank. The fuel tank is transparent. You see how full the tank is by seeing the level of the liquid in the tank [C2] You see n the fuel tank. The fuel tank is opaque. However, the colour of the opaque fuel tank tracks the level of the liquid in that tank. The tank is red when, and only when, the tank is half full. You see how full the tank is by seeing its redness.

In both cases, you see n the object about which you know. In both cases, it is by seeing that object that you see the fact that it is half full. In Dretske's original objective sense, both would count as cases of primary epistemic seeing. There is nonetheless an important distinction between C1 and C2. This revolves around which feature of the fuel tank you see, such that you see i its being half full. In the first case, you just see how full the tank is; just see i that feature of it. Seeing i that feature doesn't depend on seeing i any other feature of the fuel tank.

In the second case, however, you see i how full the tank is by seeing i a distinct feature of it; its colour. Seeing i how full the fuel tank is depends on seeing i this distinct feature of it. Even though C2 would count as primary in an objective sense, there is nonetheless a good case for saying that in fact, it is secondary.

In drawing Dretske's original objective distinction, we made use of facts about which objects you saw n . You see n just those objects that you visually differentiate. But you can see facts about objects that you fail see n . In drawing a distinction between the different ways in which you might count as seeing i an object's feature, I will say that you sometimes but not always see n the feature that you see i the object to have. I will say that you primarily see i F only if you see n F. If your seeing i F depends on your seeing i some distinct feature, you secondarily see i F.

By virtue of visually differentiating an object, you become visually related to that object; you see n it. By virtue of visually discriminating one of an object's features, you become visually related to that feature; you see n it. If you see some feature, and it is Jack's favourite feature, then you see Jack's favourite feature.

It is more difficult to visually discriminate a feature than it is to visually differentiate the object of which it is a feature, and the relation secured is more fragile. You can see n a car from close up or when it's a speck in the distance, in broad daylight or in the pitch dark, with sunglasses on, through infra-red night vision goggles, or through your wife's prescription glasses. You can see n it by seeing its front, back, top or underside, its wing-mirror, its headlights or its tail-lights. You can see n it through a violently distorting lens, or after its involvement in a violent crash.

If you see n the car, and the car is red, then you see n the red car. But you don't necessarily see n its redness. If the car is just a speck in the distance, it may well be too far away for you to see n its colour. Likewise you can't see n the car's colour if you see n it by seeing n its headlights in the pitch dark, or just see n it from underneath.

You can be visually related to some of an object's features but not others. To see this, compare the way in which a normally sighted person and an achromatopsic might get related to an object's colour. Just as blind people cannot visually differentiate objects, achromatopsics cannot visually discriminate colours. This needn't stop them seeing i some objects to be of particular colours. They are perfectly capable of stopping at traffic lights. On visual grounds, they can correctly identify the traffic light to be red. However, their ability to see i the state of the traffic light is dependent on their ability to see i the position of the traffic light's illuminated light, not its colour. For this reason achromatopsics could at most only secondarily see i the reds of traffic lights.

The achromatopsic's situation is somewhat like yours in C2. In C2, you visually discriminate the redness of the fuel tank, and only thereby see i how full it is. In C1, on the other hand, you visually discriminate the feature you see i . You see i it by seeing n it. For this reason, we can say that in C1 but not in C2, you primarily see i the fuel tank's being half full.

The Perceptualist claims that you see i Jack's anger. But it will make a difference to his claim whether you primarily or only secondarily see i his anger; whether or not you actually see n it. This is because secondary seeing i grants you only inferential knowledge. 23 If you only secondarily saw i Jack's anger, your knowledge of his anger would be inferential, and PH would be false.

Take the case where you see how full the tank is by seeing the state of the gauge on your dashboard. From the fact that you see the state of the gauge on your dashboard, we cannot explain how it is that you know about the state of your fuel tank. There is epistemic slack here. To take it in, the epistemologist will have to posit something with the structure of an inference. It is only if there is a legitimate inferencehowever probabilisticwhich would connect what you primarily see i of the state of your gauge with the state of your tank, in this situation, that you could count as knowing the latter by seeing the former. Given that in order to show how you count as knowing how full the tank is we have to posit a legitimate inference running to this from what you see n , your knowledge will turn out to be inferential.

C2 is just the same. From the fact that you see the colour of the tank alone, the epistemologist cannot explain how it is that you count as knowing how full the tank is. It is only if there is a legitimate inference which would connect the colour of the tank to its fullness that you could count as knowing how full it is. So it is only by mentioning this inference that the epistemologist could explain how you know what you do. In C2, your knowledge is inferential.

If achromatopsics can sometimes see i colours, the mere claim that you might sometimes see i another's anger might be thought relatively uncon-23 Cf. McNeill, troversial. It is compatible with the truth of Inferentialism, and the denial of PH. If PH is true, then you sometimes not only see i others' mental features, but see n them. In other words, you must visually discriminate them.

In this Section, I have distinguished seeing n objects from seeing i them to have some particular feature. And I have distinguished between two ways in which you might see i an object to have some particular feature. I marked this distinction by saying that in some but not all cases, you visually discriminate the very feature you see i the object to have; that in some but not all cases, you not only see i that feature, but also see n it.

With all this in mind, I consider two possible ways of cashing out the Embodied Perception Theorist's claim. In the next Section, I discuss the possibility that you see n Jack's anger by seeing n some outward part of his mental state. In Section V, I discuss the possibility that you see n Jack's anger by somehow seeing i some outward part of his mental state. I conclude that neither gloss will work.

IV. TRANSPARENTLY SEEING A MENTAL STATE'S EMBODIED PARTS

As currently understood, the Embodied Perception Theorist claims that you may primarily see i Jack's anger by seeing some of the surface structures or processes which partially constitute his mental state. What is left unclear is in what sense it is supposed that you see those surface parts of Jack's mental state, such that you count as seeing n his anger. The analogy with seeing objects which the Embodied Perception Theorist employs might suggest that it is by seeing n the surface parts of Jack's mental state that you see n his anger. This, at least, is a natural reading of the claim that by seeing the iceberg's tip you see the iceberg. It seems as if, on some particular occasion, you count as being visually related to the iceberg merely by dint of being visually related to that top part of the iceberg. In virtue of the fact that you have visually differentiated one part of the iceberg, you count as visually differentiating the iceberg. So imagine that merely by dint of being visually related to those surface parts of Jack's mental state, you counted as being visually related to his anger. The Embodied Perception Theorist could claim that you not only saw i Jack's anger, but also saw n it. In this way, your knowledge of Jack's anger would be non-inferential.

In this Section, I argue that it is not a sufficient condition on your seeing n Jack's anger that you see n a part of Jack's mental state. Firstly, it is possible to see n a part of Jack's mental state without seeing n the mental state of which it is a part. Secondly, even if by seeing n some part of it you counted as seeing n Jack's mental state, this would not be sufficient for seeing n his anger; for seeing that feature of him.

Consider the following two cases:

[C3] You see n some trees, and the trees you see n are a part of the wood, but you don't see n the wood

[C4] You see n the red book. But you don't see n the redness of the book In C3, you see n some of the parts of the wood, but you don't count as seeing n the wood. If this is possible, then you don't count as seeing n the wood merely in virtue of seeing n some of its parts. In other words if C3 is right, your seeing the trees on this occasion is not sufficient to count you as seeing the wood. It is possible. Transparently seeing some part of an object does not logically entail that you have seen n the object of which it is a part. Transparency guarantees that substituting co-referential terms preserves truth. But terms for an object do not co-refer with terms for any proper sub-set of an object's parts. So there is nothing about transparency that rules out cases like C3. And indeed it seems compelling that you might find yourself in a situation like C3; that you might fail to see n the wood for the trees. Seeing n some part of an object is not in all cases sufficient for seeing n the object of which it is a part. On some occasions and in certain situations it is sufficient, in others it is not. More would need to be said. 24 In C3 you see n some part of an object while failing to see n the object itself. This shows that from the fact that you see n some surface part of Jack's mental state alone, it doesn't follow that you see n the mental state of which it is a part. It would have to be shown why this case is like the iceberg case and not like C3.

Compare C4. In C4, you see the book, and the book is red. But it doesn't follow from this that you've seen the redness of the book. Terms for an object don't co-refer with terms for an object's features. And it is compelling that you could in fact find yourself in a situation like C4. 24 Dretske (p. 27) thinks it will depend partly on which part of the object you see, partly on you, and partly on the circumstances in which you perceive the object. Jackson (p. 19) thinks you need to see 'a reasonably substantial part' of the object. B. McLaughlin, in 'Perception, Causation, and Supervenience', Midwest Studies in Philosophy, 9 (1984) pp. 569-92, argues that it is a matter of convention.

Lighting conditions might be such that you don't get visually related to that feature of the book. Or you might be achromatopsic. It is in general never a sufficient condition for seeing any particular feature that you see the object of which it is a feature.

The example in C3 concerns the conditions on seeing n objects. The example in C4 concerns the relations between seeing objects and seeing their features. When you come to know that Jack is angry, what you are aware of is the fact that Jack has a particular mental feature. You are aware of his being angry. Suppose for now that in seeing n an outward part of a particular complex which also has neural, psychological and phenomenological components, 25 you see n that complex itself. Could this amount to your seeing Jack's anger? The short answer is 'no'. What you see n , if you see n this complex of structures and processes, is a three-or four-dimensional lump of Jack; a special kind of object.

If the Embodiment Hypothesis is correct then Jack is angry, on this occasion, in virtue of having this complex of structures or of undergoing these varied physiological and psychological processes. But from the fact that you see n this complex, nothing follows about what features you've seen n . So from the fact that you've seen n this complex, it doesn't follow that you've seen n Jack's being angry.

Think back to the iceberg case. In seeing n the iceberg's tip, you see n the iceberg. It doesn't follow that you visually discriminate its being an iceberg. Partly in virtue of its tip being above water, the iceberg has the feature of being semi-submerged. But it doesn't follow from your seeing n the iceberg and its exposed tip that you visually discriminate the iceberg's being semi-submerged; it doesn't follow that you see n that feature of the iceberg.

C3 manifests the fact that seeing n an object's part is not in all cases sufficient for seeing n the object of which it is a part. C4 manifests the fact that seeing n an object, or a part of an object, is never sufficient for seeing n any particular feature of it. Even if you count as seeing n Jack's mental state, considered as a complex of structures or processes, this is not sufficient for seeing n the feature Jack has in virtue of having these structures or undergoing these processes.

This brings out an important ambiguity in our talk of mental states. We might mean to talk about the state Jack is inhis being angry, desiring ice cream, believing that the shop is shut. Alternatively, we might mean to talk about the kinds of states that can be 'inside' Jackparticu-lar bundles of neurons or muscles; particular events and processes which take place in parts of his body. While the latter might realise the former, visual discrimination of the latter does not amount to visual discrimination of the former. The Perceptualist's aim was to explain how we had knowledge of other people's mental features by claiming that we sometimes see n those features. The claim that we sometimes see n the complex structures inside Jack, or the complex processes they undergo, does not secure the Perceptualist's claim.

This also brings to light an ambiguity in what we mean when we say that you've seen n Jack's facial expression, or his behaviour. When we say that you've seen n Jack's facial expression, we might mean to say that you've seen n some part of Jack that can be visually differentiated. It would be like saying that you'd seen n his nose or his toes; like saying that you'd seen n the mole on his cheek. Alternatively we might mean to say that you've visually discriminated some outward feature of Jackhis smiling, or wincing, or frowning, for example. Likewise, when we say that you've seen n Jack's behaviour, we might mean to say that you've seen n some event of which Jack is a part; for instance, the event of Jack's poisoning Jill. Or we might mean to say that you've seen n some feature of that eventits involving Jack's arm moving downwards, say, or its murderousness.

You can see n the frowning parts of Jack without seeing i his frowning. You might be looking at his face through a specially distorting lens. By the same token, you can see n the event of Jack's frowning without seeing n him frown. When looking at him through that distorting lens, you might see n the configuration of his face change without being able to see n exactly how it is changing. You can see n the change which is his frowning without seeing n it as such.

The discussion in this Section shows one way in which the Fossil View is misguided. When we claim that by looking at the rock you see the fossil, we may mean that you have seen n the fossil, or that you have seen i the fossil as such. C3 shows that seeing n the exposed part of the fossil on a particular occasion may or may not be sufficient to count you as seeing n the fossil. C4 shows that even if you do count as seeing n the fossil, this could never be sufficient to count you as seeing i it in any particular way; as seeing its being a fossil or anything else. Likewise, seeing n some outward part of the structures and processes in virtue of which Jack is angry on a particular occasion may or may not be sufficient to count you as seeing n the structures or processes in virtue of which Jack is angry. But seeing n the structures or processes in virtue of which Jack is angry could never be sufficient to count you as seeing n Jack's anger. Yet it is seeing n this feature of Jack that the Perceptual Hypothesis requires of you.

V. OPAQUELY SEEING A MENTAL STATE'S EMBODIED PARTS

In the previous Section, I argued that seeing n the structures or processes in virtue of which Jack counts as being angry was not sufficient for seeing n Jack's anger. But in claiming that you see Jack's anger by seeing some of his outward structures or processes, the Embodied Perception Theorist might mean to claim that you visually identify those structures or processes in some particular way; that you not only see n them but somehow see i them.

In this Section, I argue first that seeing i a feature of an object F1, partly in virtue of which the object has some further feature F2, is not in all cases sufficient for seeing i F2. Second, I argue that F1 must be a very specific kind of feature, if seeing i F1 is sufficient for seeing i F2. Third, I argue that while the outward structures and processes which are parts of Jack's mental state could have such a feature, so too could outward structures and processes of Jack which are not parts of his mental states. It follows that to the extent that seeing i some feature of the embodied parts of Jack's mental state might be sufficient for seeing i his anger, seeing i such parts cannot be a necessary condition on seeing i people's anger. So the Embodied Perception Theorist's claim that embodiment makes the difference between the perceptibility and imperceptibility of mental features would be undermined.

Consider the following two cases:

[C5] You primarily see i the four right-angled corners of the square as such. But you don't see i its squareness An object is square partly in virtue of having four right-angled corners. Nevertheless, seeing i an object's having four right-angled corners is not sufficient for seeing i its squareness. Oblongs also have four right-angled corners, and you can be in a situation where the square appears oblong. In such a situation, you are not visually related to the object's squareness you don't see n that feature of it; don't visually discriminate it. Nonetheless, you can be visually related to its having four right-angled corners; can still see n that feature of it.

[C6] Via an electron-scanning microscope, you see n the detailed microstructural features of some part of the red book's surface. But you don't see n the book's redness.

It is (I will assume) the microstructural features of an object's surface that make it the colour it is. An object is red in virtue of having particular microstructural features. But you don't count as seeing n the colour of the object solely in virtue of seeing n those microstructural features of it. You can see n the microstructural features without seeing n the redness. Think of Mary in her black and white room. 26 She doesn't see n anything's colour. So she doesn't primarily see i red things as such. But Mary might learn about the connections between objects' microstructural features and their colours. Such knowledge might become so in-grained that on inspecting some book's surface microstructures with the help of her electron microscope, she spontaneously comes to be aware of that book's being red. In this sense, we might be happy to say that she has seen i the book's redness. But she will only have seen i this secondarily. The epistemologist will have to posit a legitimate inference which connects the microstructural features she primarily sees i with the colour she comes to be aware of the object having. Consequently, her knowledge that the book is red will be epistemically inferential.

The danger for the Embodied Perception Theorist is that things will be epistemically similar when it comes to seeing i Jack's anger by seeing i features of his outward parts. What is needed is some way of seeing n Jack's anger solely in virtue of seeing n some feature of Jack's outward structures or processes. If, by seeing n some feature of Jack's outward structures or processes, this just made it true that you saw n Jack's anger, then we might be able to retain the idea that your seeing i Jack's anger was dependent on seeing i something of his outward structures or processes, while denying the need for an inference connecting the two things you'd have seen.

The outward parts of Jack's embodied mental state do have such a feature. So there is a way of seeing i the outward parts of Jack's anger that would allow you non-inferentially to know of Jack's anger. If the Embodiment Hypothesis is correct, then some of Jack's outward structures and processes are aspects of him in virtue of which he is angry. In other words, it is a feature of those outward structures and processes that they partly constitute Jack's being angry. Imagine that you saw n that feature of them, and thus that you primarily saw i those structures and processes as aspects of Jack's anger. There would be no epistemic slack between seeing i this and knowing that Jack was angry. The fact that x exists and x is an aspect of y logically entails the existence of y. So seeing i an aspect of Jack's anger as such is sufficient to ground your knowledge that he is angry. There is a difference between the content of what you are claimed to see, here, and your knowledge of his anger. But the slack is not epistemic. Hence while a (deductive) inference may need to be cited in order to explain how you were caused to be in this state of knowledge, no legitimate inference needs to be cited in order to explain how your knowledge counts as such.

If you saw i an aspect of Jack's anger as such, this would be sufficient grounds for claiming that your knowledge of Jack's anger was noninferentially perceptual. Earlier I claimed that a necessary condition on having this kind of knowledge of Jack's anger was that you saw n his anger. If this is right, then we have sufficient grounds here for claiming that you saw n Jack's anger. Your perception of Jack's anger would be 'indirect', yet your knowledge would be both genuinely perceptual and epistemically non-inferential. The conditions on PH would have been met.

But now imagine that the Embodiment Hypothesis is false. Then Jack's outward structures and processes on a given occasion will be mere effects of his anger. As it happens, however, those structures and processes nonetheless have the feature of being effects of Jack's anger. So imagine that you could primarily see i that feature of them. If y is the effect of x, then x exists and is the cause of y. Seeing i an effect of x as such is sufficient to ground knowledge of the presence of x. There is no epistemic slack between the two states. Hence while an inference would need to be cited in order to explain how you came to be in this state of knowledge, no legitimate inference needs to be cited in order to explain how your knowledge counts as such. If you saw i an effect of Jack's anger as such, this would be sufficient grounds for claiming that your knowledge of Jack's anger was non-inferentially perceptual. So we would have sufficient grounds here to claim that you saw n Jack's anger. Even if your seeing n Jack's anger was somehow 'indirect', it would nonetheless secure noninferential, perceptual knowledge of Jack's anger. Again, the conditions on PH would be met.

The claim that you see i either aspects of Jack's anger as such or bodily effects of Jack's anger as such would be equally good ways of cashing out the thought that you can sometimes see another's anger by seeing that anger expressed, where the expression is conceived along McDowellian lines to entail that which it expresses. 27 Neither feature amounts to Jack's being angrythey are distinct; but neither feature 'falls short' of Jack's being angry eitherboth entail that Jack is angry.

Return now to the Fossil View. In the previous Section, I argued that while you might see n the fossil by seeing n its exposed part, the analogy here wouldn't help the Embodied Perception Theorist. You can see n the fossil without seeing n any particular feature of it. But what we wanted was for you to be able to see n a feature of Jack by seeing n some part of his surface. Now imagine that you see n some features of the exposed part of the fossil. You can primarily see i the contours and colours of that part. Thereby you might come to see i its being a fossil. But there is epistemic slack between the two things you are claimed to see i here. If you see i its being a fossil by seeing the colours and contours of its exposed part, your knowledge of its being a fossil will be epistemically inferential. Now imagine that you split open a rock, and inside is a fossil, covered by a very thin, even layer of the rock you split open. You can see n people fully clothed; see knights in shining armour. 28 So suppose you count as genuinely seeing n the fossil, thinly veiled. The contours you now see i are effects of the fossil. However, if you see i them to be effects of a fossil, you might thereby count as seeing i the fossil as such. In this way you might get non-inferential knowledge that the thing you see n is a fossil.