Comments on David Chalmers’ “Perception and the Fall from Eden”

Susanna Siegel * 15 February 2004


There would be no Edenic content, hence no two-stage view, if there was no such thing as the perfect veridicality conditions of experiences, where these diverged from ordinary veridicality conditions. (From now on, I’ll just take as definitional that perfect veridicality conditions, if there are such, diverge from ordinary veridicality conditions).

A central substantive claim in Chalmers’s paper is that perceptual experiences have perfect veridicality conditions. And not just ones associated with color experience, but ones associated with many other aspects of visual experience, as well as with perceptual experiences in other modalities as well.

In defending this view, I think Chalmers brings into focus quite a strong claim about the relation between phenomenology on the one hand and accuracy conditions on the other.

Say that an experience has accuracy conditions if there are ways the world would have to be, for the experience to be accurate, and accuracy is truth. And say it’s definitional of events that are perceptual experiences that in undergoing them, the subject has some sort of phenomenology or other. I’m going to take for granted that it makes sense to talk about experiences having exactly the same phenomenology, or being phenomenally the same (some philosophers think it doesn’t make sense to talk this way).


The strong claim can be seen by contrasting it with five other claims about the relation between these two things. (The strong one I want to talk about is number (5) below).


(1)   Typically, perceptual experiences have accuracy conditions.

(2)   All perceptual experiences have accuracy conditions.

(3a) For any two phenomenally identical experiences, there are some accuracy conditions that they share.

(3b) Any two phenomenally identical experiences have the same accuracy conditions.

(4)   All perceptual experiences have accuracy conditions that directly mirror their phenomenology.

(5)   For any two phenomenally identical experiences, there are some accuracy conditions that they share that directly mirror their phenomenology.


(p. 12: “What view of the contents of perceptual experience best mirrors its presentational phenomenology? Here I think the answer is clear. The view of content that most directly mirrors the phenomenology of color experience is primitivism…Taking the phenomenology completely at face value, visual experience presents a world where perfect redness and blueness are instantiated on the surfaces of objects, as they were in Eden. These are simple intrinsic qualities whose nature we seem to grasp fully in perceptual experience. For the world to be exactly the way my phenomenology presents it as being, the world would have to be an Edenic world in which such properties are instantiated.”)


All of these claims are controversial. Charles Travis in a recent paper denies claim (1), and claims J. L. Austin agrees with him. Claim (2) is denied by Tyler Burge in ‘Vision and Intentional Content’, where he argues (as I understand it) that visual experiences that aren’t appropriately hooked up to the world are not assessable for accuracy in the sense above. Claims (3a) and (3b) are denied by people who think both that a duplicate of someone whose experience has a certain phenomenology will be a phenomenal duplicate, but that to share accuracy conditions that experiences (or their subjects) must stand in the same external relations to thing or properties in the world.


Chalmers defends claim (5) (and so (4), which follows from it, since by definition all experiences have some phenomenology or other).  This is the strong claim about the relation between phenomenology and accuracy conditions. Not only does phenomenology suffice for accuracy conditions – as someone who believed (3a) or (3b) could agree. It suffices for accuracy conditions that stand in the mirroring relation to phenomenology.


I don’t know exactly what the mirroring relation is, but as I understand it it’s supposed to some non-contingent, non-arbitrary relation between the quality of the phenomenology, and the accuracy condition (or part thereof). The phenomenology of, say, a visual experience, is supposed to have isolable aspects, and then there are supposed to be properties (‘perfect properties’) that correspond to each of those aspects. So for example, if an aspect of the phenomenology of my (say) seeing my shirt is phenomenal pinkness, then there is supposed to be a property – perfect pinkness – that corresponds to that aspect. Chalmers isn’t completely explicit about what other aspects phenomenology has, but there are supposed to be quite a few other ones—not just ones broadly associated with color.


It seems to me that a doubt could be raised about Chalmers’s reason for thinking that (5) (and so likewise (4)) is correct, and the first question for Chalmers is how he responds to this doubt.


I take it that the reason he has for believing (5) is, he thinks we have intuitions about ‘how the world would have to be, taking phenomenology completely at face value.’(p. 12). Another way he has of putting it is that we have intuitions about how the world would have to be, for it to be ‘exactly the way that my phenomenology seems to present the world as being.’ (p. 12).


As evidence for our having these intuitions he points to the fact that people who consider the matter tend to find it at least a little bit compelling to suppose that when experiences represent color properties, the properties they represent are ‘intrinsic properties whose nature we grasp fully in experience’  (p. 12). The evidence isn’t supposed to be that lots of philosophers believe this view about which color properties are represented in color experience. Rather (I take it…) the support for the view (that we have intuitions about how the world would have to be if we took phenomenology at face value) is supposed to be that whatever we think of the primitivist view about the contents of color experience, we should see that it’s strength is that its proposal ‘vindicates color phenomenology’.


It seems to me that this appeal to the strength of primitivism isn’t very strong support for our having intuitions of the sort Chalmers says we have. (Obviously he doesn’t rest his case on our having all the intuitions he needs to say we have, if experiences really do have perfect veridicality conditions – just ones about color). Suppose you ask of a color experience, how would the world have to be, for the world to be exactly the way phenomenology ‘seems to present the world as being’. Here’s one answer: ‘there’s no single way the world would have to be for the experience to be accurate. There are lots of different ways the world could turn out to be, with the experience turning out to be accurate. It just depends on the properties that are involved in causing this sort of experience in me.’ That answer doesn’t involve any substantial intuitions about perfect veridicality conditions, and doesn’t purport to identify perfect properties. It says that phenomenology doesn’t take a stand on exactly which properties surfaces of objects have. It doesn’t leave this totally open, but it also doesn’t purport to identify them. 


So there’s a doubt that we have intuitions that the strong veridicality conditions of color experiences involve positing perfect color properties. If we had those intuitions, then I think it wouldn’t seem as reasonable to answer the question by saying ‘there’s no single way the world would have to be for the color experience to be accurate’ – even when we’re really paying close attention to our phenomenology. But this does seem reasonable.


At one point in the paper Chalmers says something that directly addresses this. When he is characterizing the phenomenology of color experience, he says,


“when I have a phenomenally red experience of an object, the object seems to be simply, primitively red. The apparent redness does not seem to be a microphysical property, or a mental property, or a disposition, or an unspecified property that plays an appropriate causal role. Rather it seems to be a simple qualitative property, which a distinctive sensuous nature. We might call this property ‘perfect redness’.” (p. 12).


Suppose that this description of the phenomenology is right. I think Chalmers would say, if this is the right description of the phenomenology, and if we’re supposed to take phenomenology at face value when we consider how the world would have to be for the experience to be accurate, we should say that the object of the phenomenally red experience has to have perfect redness in order for the experience to be accurate.


It seems to me that there are two criticisms one could make of this bit of reasoning. First, one might deny that the description of the phenomenology really provides independent support for the claim that experiences have perfect veridicality conditions. The description of the phenomenology isn’t the unproblematic gloss it purports to be if it just builds in that experience has veridicality conditions, since as we saw this is a claim that some people (presumably reasonably) deny. If there’s a genuine debate here then it’s the same phenomenology whose character is in dispute (ie, it’s not that Charles Travis has dfrnt phenomenology than Chalmers has). 


A different criticism one might make of this response grants that the description is perfectly fine, but interprets it in a way that removes its support for the conclusion that perfect redness is represented in experience. If we ask how the world would have to be for a phenomenally red experience to be accurate, it seems we could easily lack intuition that the world would have to contain perfect redness in order for the experience to be veridical.  It seems plain that if you asked someone what property a ripe tomato looks to have, we wouldn’t say that it looks to have a microphysical property, or a mental property, or a disposition, or an unspecified property that plays a causal role. What’d we say, probably, it just that it looks red. This is a way of interpreting Chalmers’s description of the phenomenology –as a description of what we’d say if asked to characterize how the tomato looks. [This is how Strawson seems to think of the contents of experience in “Perception and its Objects”]. Here it seems that someone could agree with Chalmers’s claims about appararent redness, but deny that the world would have to contain an instance of the perfect redness in order to be veridical. One might answer the question “how would the world have to be for the experience to be exactly the way phenomenology presents it” by saying, ‘the tomato would have to be red’ – and that leaves open what property redness is.


Let me sum up what I’ve said so far. I’m not denying that phenomenology could illuminate how it is that experiences have veridicality conditions, in addition to sufficing for them. That’s an interesting view. I’m just questioning Chalmers’s support for it. His support for it seems to be that we have a certain kind of intuition about veridicality, one that specifies perfect properties. I don’t think we have those intuitions, or at least, I don’t that think they’re as firm as Chalmers needs them to be to support his claim that there are perfect veridicality conditions associated with experiences.


I think the heart of the issue between Chalmers and someone who would be inclined to answer the way I said was reasonable has to do with background commitments about what can make it the case that an experience represents the properties it does. I think Chalmers thinks that there are some properties represented in experience which are such that you have to be in causal contact with their instances to represent them, there are other properties for which this isn’t true. I’m inclined to agree with this—there are probably various relations between properties and experiences that secure representation. One thing that’s strong about the claim (5) is that it proposes that phenomenology can do this all by itself—whether or not having the phenomenology itself requires standing in external relations to things. (As I formulated (5) it doesn’t make this part explicit, but that’s Chalmers’s view). . So someone might who got queasy over that sufficiency claim would get queasy over (5) – as well as over (3a) and (3b) (if we likewise add on to them the condition that phenomenology can be shared by duplicates).


But it seems to me that there’s a related thing that someone might reasonably get queasy about, and this is another thing that’s strong about claim (5). It sounds to me that the bit about mirroring is supposed to illuminate how experiences could have accuracy conditions at all, by pinpointing a (putative) set of accuracy conditions that ‘mirror’ phenomenology—ie, that not only stand in a non-contingent relation to them (as would any contents mentioned in (3a) or (3b), such as Chalmers’s ‘Fregean’ contents), but also a non-arbitrary connection to them. The mirroring part of (5) sounds to me like a theory of intentionality for experiences with respect to one sort of content, namely the edenic kind. It’d a pretty simple theory, and one that it’s hard to see how to amplify. The theory says, the fact that these contents stand in a manifestly non-arbitrary relation to phenomenology illuminates how experiences could have at least that sort of accuracy condition in the first place. Of course the price of getting this illumination is that the contents are only perfectly veridical! But for all that they’re still veridicality conditions.


So second question for Chalmers is, do you think of the mirroring relation in either of these ways: 1-as illuminating how experiences could have at least any sort of accuracy condition at all. or 2-as entailing that the phenomenal property mirrored by perfect property is identical with experientially representing that perfect property (so that if phenomenal redness is mirrored by perfect redness, the phenomenal redness is identical with the property of experientially representing perfect redness). This second option would avoid there being any priority of phenomenology over representation, which is something Chalmers seems to want (going by what he says at the very end of another recent paper on the experience and representation, "The Representational Character of Experience").


The third question for Chalmers concerns the gloss on perfect colors. These are glossed as ‘intrinsic properties whose nature we seem to grasp fully in experience’ (p. 12). It’s definitional of perfect properties that they exactly mirror phenomenology. But is it definitional that they seem to be grasped fully in experience? Neither of the further characterizations of mirroring (both of which seem to be suggested by things he says in the paper) seems to require that we seem to grasp a property’s nature fully in experience. I’m a little bit in the dark about what it is to grasp a property’s nature fully in experience (hence what it is to seem to do so), so it would be nice to hear more about what this is too.


Maybe Chalmers thinks that for a property to perfectly mirror phenomenology, one would have to grasp its nature fully in experience, because he thinks there is a sense in which we seem to grasp fully in experience the nature of phenomenology that gets mirrored. I can see how a premise about our fully grasping the nature of phenomenology would support a conclusion about our fully grasping the nature of perfect redness, if phenomenal redness were identified with perfect redness. But presumably Chalmers doesn’t want to make this identification, since phenomenal redness is a property that’s actually instantiated but perfect redness isn’t. So fourth question is, what’s the status of the epistemic claim about full grasp of perfect colors – fallout from the definition of perfect properties? Or empirical fact about color phenomenology only, with no bearing on other perfect properties? Or something else.


I want to end by asking about how the account is supposed go for aspects of experience besides color, in particular how it’s supposed to go for what Chalmers calls the ‘objectual aspects of our experience.’ The objectual aspects of experience are those aspects associated with the representation of objects.


Chalmers expresses sympathy with the view, which he attributes to John Campbell and Mike Martin, that “a merely existential characterization of phenomenal content does not fully respect the directness of an experience of an object…experience does not merely present that there is an object at a certain location with a certain color: it presents that that object is at a certain location with a certain color.” If experiences have the perfect veridicality conditions that Chalmers says we intuit they do, then in experience we seem to be ‘directly acquainted with objects, and no mediation is involved.’ (47)


One way to question this would be to say that Chalmers and Campbell and Martin are all wrong about the phenomenology – that it really is respected by a merely existential characterization of phenomenal content. Another way to question is to ask what exactly the phenomenal difference is supposed to be between the way Chalmers, Campbell and Martin take the phenomenology to be, and the way they think their opponents take it to be. I find this pretty obscure. Both demonstratives and certain existentially quantified formula (eg, the Russellian expansion of ‘the table over there’) are devices for representing particular objects. Of course there are modal differences, but those seem irrelevant when considering phenomenology. Chalmers seems to think that if the part of the perfect veridicality conditions associated with representation of objects really was existentially quantified, then we’d expect our phenomenology to be different than it is. (If it weren’t different, then the existentially quantified conditions could mirror phenomenology). Maybe you can say more what the latter contents would leave out.


Chalmers also says, in discussing the objectual aspects of experience, that while in Eden we are directly acquainted with objects, out of eden “our epistemic grip on objects is not as direct as it is in Eden” (p. 47). This suggests that out of eden, when we (say) see objects, there is something missing from our epistemic situation with respect to them that we would have if things were as phenomenology presents them. What is missing?