Robert Stalnaker


I start with a bit of philosophical jargon, first introduced by Gareth Evans, but used since by many others who cite Evans, including Christopher Peacocke, John McDowell, and Michael Tye. My initial question was, what do these philosophers mean by "nonconceptual content," and its contrast, "conceptual content"? What kinds of objects are these different types of content, and how are they used to characterize perception and thought? It is controversial among those who talk of nonconceptual content whether there is such a thing, and whether perceptual states have a kind of content that is different from the kind that characterizes belief states and speech acts. But Evans gives us no direct and explicit characterization of the notion of nonconceptual content that he introduces - at least none that I can find. And it is not clear to me that the different philosophers using this term mean the same thing by it. Without some account of what nonconceptual and conceptual contents might be, it is difficult to have more than a general impression of what this controversy is about.

Some things Evans says suggest that it is mental states, rather than their contents, that are conceptual or non-conceptual, and sometimes he substitutes "non-conceptualized" for non-conceptual, but it is clear that he thinks there are two kinds of content, and not just two kinds of states that content is used to characterize, or two ways in which content might be expressed. "The process of conceptualization or judgement," he says, takes the subject from one kind of state (with a content of a certain kind, namely non-conceptual content) to his being in another kind of cognitive state (with a content of a different kind, namely, conceptual content)." [Gareth Evans, Varieties of Reference, p. 227. All page references to Evans are to this book] John McDowell, on the other hand, argues that the process of judgment does not introduce a new kind of content, but "simply endorses the conceptual content, or some of it, that is already possessed by the experience on which it is grounded." [John McDowell, Mind and World, p. 49] It is issue behind this dispute that I want to try to get a little clearer about.

Let me confess at the beginning that I will not propose answers to my questions about how these philosophers should be understood. I am puzzled by much of what they say - I have the impression that their arguments are being guided, on both sides, by conceptions of content and its role in the explanation of perception and thought that have underlying presuppositions that I don't share, and don't fully understand. So my strategy will be indirect: rather than trying to ferret out those presuppositions by detailed examination of the texts, I will spell out my own assumptions about representational content, and ask how, given the way I understand this notion, a distinction between conceptual and nonconceptual might be drawn, and what role it might play in the explanation of the relation between perception and thought. I will begin with what I take to be some platitudes about content, assumptions that I would expect to be disputed only by a philosopher who rejected the whole idea of representational or intentional content. After a while, more controversial assumptions may emerge, but I hope we will be able to identify the point at which disagreement begins.

The notion of propositional content begins with the idea that what is said in a speech act - the proposition expressed - can be abstracted from two different aspects of the way it is said: first from the means used to express it, second from the force with which it is expressed. The same proposition can be expressed by different sentences of the same or different languages, and the same proposition can be the content of an assertion in one context, and of a supposition, a component of a disjunctive assertion, or a request in other contexts. Furthermore, the contents expressed in speech acts with different force are the same kinds of things as the contents of mental states of different kinds, such as belief, desire, intention, hope and fear. Just as what is said can be separated from how it is said, so what is thought can be separated both from the means of mental representation and from the kind of mental state (belief, wish, tacit presupposition, hope or fear) that the proposition is used to specify. Just as you and I might say the same thing, even though you say it in French and I say it in English, so you and I might believe the same thing even though the systems of mental representation in which the information is encoded in our respective minds is different. And just as I may assert what you merely suppose, so I might believe what you doubt, but hope for. And it seems at least prima facie reasonable to say that when something merely looks to me to be a certain way, even though I don't really believe that it is that way, then there is perceptual state with a certain content that might have been, but is not the content of any of my beliefs.

So what might these things be - things that are the contents of speech acts and mental states of various kinds? There are many different theories about what propositional content is, but two things seem common to all theories that take content seriously at all: first, a content is an abstract object of some kind (as contrasted, for example, with a sentence token, or a mental representation); second, it is essential to propositional contents that they have truth conditions. Perhaps they are truth conditions, perhaps something more fine-grained that allows for the possibility that different propositional contents may have the same truth conditions. Either way, what is assumed is that for any state, act or object with propositional content, one can ask whether or not things are as the state represents things to be, and this is to ask whether the truth conditions of the propositional content are satisfied.

What are truth conditions? Different things might be meant by this expression; here is one: think of the meaning of a sentence as a recipe for determining a truth value as a function of the facts. The recursive semantic structure of the sentence encodes such a procedure. One might identify the recipe with the truth conditions, since it spells out the procedure, or conditions, for determining whether the sentence is true. Here is a contrasting explanation: one might instead identify the truth conditions of a statement simply with the circumstances (the way things must be) for the proposition expressed to be true - the conditions under which the statement is true. Different recipes determined by statements with different constituent structure might end in the same place, no matter what the facts (as, for example, with statements of the forms ~(PÚ Q) and (~PÙ ~Q)). Such statements will have different truth conditions in one sense, but the same truth conditions in the other. I will make only the weaker assumption that propositional contents have truth conditions in the second sense.

Now there are many kinds of abstract objects that have, or determine, truth conditions in this sense - different candidates for a kind of representational content. Some of them may be appropriately called "conceptual" in some sense; others might appropriately be called "nonconceptual." For example, one might define complex objects, nested ordered sequences that reflect the recursive semantic structure of the sentences with which the structure is associated. The ultimate constituents of such structures might consist wholly of senses or concepts. Maybe conceptual content is an object of this kind, though it remains to be said what senses or concepts are. Alternatively, one might take the ultimate constituents of such structures to be individual objects and properties and relations (the referents of names and the properties and relations expressed by the predicates in the relevant sentences). Perhaps this is a kind of nonconceptual content. And there might be mixed cases - structures that contain concepts or senses (associated with predicates) and individuals (associated with names).

These different candidates for a kind of content are not independent; there will be correspondences between contents of the different kinds. In some cases there will be straightforward ways of determining a content of one kind as a function of a content of one of the other kinds. In particular, whatever senses are, they determine referents; whatever concepts are, it seems reasonable to say that concepts of the appropriate kind determine properties. (A concept of cat determines, I assume, the property of being a cat.) If this is right, then a structure that is made up of senses or concepts will determine a unique structure of the kind made up of individuals, properties and relations, though the reverse will not be true. So there is a clear sense in which structures made of senses or concepts are more fine-grained than those made of individuals and properties and relations, with the mixed cases falling between the two.

All of the candidates considered so far build into the content a recipe for determining truth conditions. One might instead take the recipe determined by some sentence as part of the means by which content is determined, rather than as essential to the content itself. One might, that is, identify the content with the truth conditions themselves - the possible circumstances that must be realized in order for some expression or thought with that content to be true. This is the most coarse-grained conception of content - the outer limit on a conception of content that meets the minimal conditions that we are requiring that any conception of representational content meet. Any conception of representational content meeting these conditions will determine a unique content of this most coarse-grained kind, so this is a kind of content that everyone should agree can be used to characterize mental and linguistic states, acts and events that can be said to have representational content of any kind. I will use the label "informational content" for content as truth conditions, propositions as functions from possible circumstances to truth values, or equivalently, as sets of possible situations. I suppose that if this is a kind of content, it is a kind of nonconceptual content, although since it can be used to characterize any kind of representational act or state, its use says nothing one way or another about whether any kind of act or state essentially involves the exercise of conceptual capacities (whatever this might mean).

Thus far I have been talking about the kinds of abstract objects that might be thought to be the representational contents of acts and states that have representational content, but I have said nothing about the states themselves, or about what it is about a cognitive, perceptual, or motivational state, or a speech act or act of judgment in virtue of which it has some particular representational content. Recall that part of the initial motivation for developing a conception of content at all was the idea that content could be abstracted from the force with which it is expressed and from the attitudes that are characterized in terms of content - a conception that might be used to describe states and acts of different kinds, and that was intelligible independently of its use to describe any representational states or acts. But of course the interest of these abstract objects will derive from their use for describing in a revealing way the phenomena they are used to describe, and for bringing out the relationships between different acts and states that are involved in representation. So I turn now to questions about the role of content in characterizing representational events and properties, beginning, again, with some platitudes.

Statements involving sentential complements (for example, statements of the form x believes that P, it appears to x that P, x asserted that P), state that a certain relation holds or held between x and something denoted by the term "that P." Sometimes the problem of intentionality is posed as the problem of how it is possible for a person to be related, as a matter of contingent fact, to the kind of abstract object denoted by a that-clause. The puzzlement is exacerbated by the causal metaphors philosophers often use to describe the ways we are related to propositions: they seem to be things we can get our hands around: we grasp propositions, we gather information, process it, and send it. Content travels in vehicles. Information saturates our thoughts, [Evans, p. 122] seeping into them like some kind of spiritual fluid. But the sober reality behind the metaphors need not be so mysterious. To be related to an abstract object is just to have a property that can be determined as a function of such an object, in the way that (to use a now familiar analogy) the property of weighing 75 kilograms can be determined as a function of the number 75. One way to get at the question, what is content? is to ask how whatever it is that is denoted by such sentential complements as "that the cat is on the mat" determines the properties that are ascribed in predicates like "believes that the cat is on the mat."

Consider a simple and straightforward example borrowed from Evans. [p. 122] I am thinking about something I can see: a black and white cat sleeping on a mat. I see that the cat is sleeping. Perhaps I suspect or speculate that the cat is a favorite of Queen Elizabeth. I entertain the possibility that the cat is ginger, rather than black and white, and judge that it is not (or in Evans's terms, I "grasp [the thought that the cat is ginger] as false."). Various properties are ascribed to me with the help of reference to some kind of abstract object that has truth conditions. Whatever is going on in me when I entertain the possibility that this cat is a favorite of Queen Elizabeth, we can describe it by attributing to me a relation to the proposition that the cat is a favorite of Queen Elizabeth (a proposition that is true in possible worlds in which that cat is one of her favorites and false in possible worlds in which it is not). The problem is to say what the world must be like for me to be related in the right kind of way to such an abstract object.

The thought episodes and belief states in such cases are, Evans notes, based on some information that the subject receives. "People are," he says, "in short and among other things, gatherers, transmitters and storers of information. These platitudes locate perception, communication, and memory in a system - the informational system - which constitutes the substratum of our cognitive lives." [p. 122] Evans says little about what information is, or what informational states are, but here is a simple minded and crude version of a familiar story: one thing contains information about another if there are causal and counterfactual dependencies between the states of one and the states of the other. An object contains information about its environment if the object is in some state that it wouldn't be in if the environment weren't a certain way. x carries the information that P if the object would not be in the state it is in if it were not the case that P. Some objects are sensitive to a range of alternative states of the environment in a way that makes them apt for transmitting or storing detailed information about some aspect of the environment. The pattern of light and dark on the ground on a sunny day, for example, carries information about the shape of the tree since there are systematic counterfactual dependencies between a range of alternative possible shapes of the tree and a corresponding range of alternative patterns on the ground. Obviously, artifacts such as thermometers and cameras are designed to be sensitive to the environment in just this way, and they are naturally described as devices designed to carry information. Equally obviously, animals have evolved a diverse range of systems - perceptual systems as well as internal monitoring systems of various kinds) that carry and use information in this sense. For a philosopher looking for a naturalistic account of intentionality, this conception of information and informational states provides a natural starting point.

Of things that carry information, we can say what information they carry. Information itself is something described with that-clauses - the information that a black and white cat is sleeping on a mat, or that this black and white cat is sleeping on a mat, or the information that the tree trunk is shaped roughly like a Y, that the temperature is seventeen degrees centigrade. Informational states have content - presumably (at least in some cases) a kind of content that is in some sense nonconceptual, since it would not be reasonable to attribute conceptual capacities to the patterns of light on the ground in virtue of the counterfactual dependencies that make it the case that those patterns have informational content.

Of course things that lack conceptual capacities (such as books) still might carry conceptualized information, and so might be correctly describable in terms of some notion of conceptual content. But whatever conceptual content turns out to be, it seem reasonable to think that for anything that has conceptual content, the fact that it does must have its origin in something with conceptual capacities, as the information contained in books has its origin in the thoughts and intentions of the members of the community in which they are written. (Pace George Washington, who is alleged to have said that all knowledge has its origin in the knowledge in books). If the notion of an informational system is, as Evans suggests, to "constitute the substratum of our cognitive lives," and if the notion of information is to contribute to an explanation of the source and nature of the content of full blooded intentional states - acts of judgment, states of belief - then it is important that our account of what constitutes the carrying of information not presuppose, or be derivative from, states of mind.

Artifacts that are designed to record, display or store information (fuel gauges, thermometers, cameras, compact disks) are among the best examples of information carrying systems, and are often used to illustrate the strategy for explaining intentionality in terms of systems that function to carry information. Since the design of such things is explained in terms of the intentions of the designers, their information carrying capacities are in a sense dependent on the intentional states of persons. But it would be a mistake to think that the information such artifacts carry is derivative in the same way as the capacity of a book to carry the information expressed by the sentences written in it. Thermometers and cameras are designed with an information carrying purpose in mind, but facts about the way such devices happened to come by their informational capacities are inessential to the explanation of what it is that constitutes those capacities. A natural thermometer or camera brought into being by some fortuitous process (Swamp-thermometer, or Swamp-camera) would carry the same information in the same way as actual thermometers and cameras. The word-like marks in Swamp-book, on the other hand, don't mean anything, and they don't carry information that has anything to do with what such marks usually mean. I don't know whether Evans is making this mistake when he says "what gives a photograph its content is, of course, something quite different from what gives states of our brain their content. (The former is parasitic on the latter)" [p. 125,n. 8], but I think this claim is at least misleading in that it ignores the distinction between the inessential way in which the capacities of cameras are parasitic on those of our brains, and the important way that the capacities of books are parasitic on ours, but the capacities of cameras are not.

Information is by definition veridical. According to the simple story I have sketched, x cannot carry the information that P unless it is true that P. If this concept is to provide a basis for an account of representational content, we need to complicate the story, but the strategy for doing so is straightforward. As Fred Dretske has emphasized, even in the simple story, any characterization of the information carried by some object will presuppose a distinction between the facts that form the background conditions (or channel conditions) for the causal structure in virtue of which the object carries information and the facts that constitute the information carried. The correctness of the characterization of the content of the information carried will be relative to such presuppositions. (So we say that the thermometer (which is in fact functioning normally, and which registers 17) carries the information that the temperature is 17 degrees centigrade, even though if, contrary to fact, the temperature were 27 degrees, and certain particular anomalous conditions also obtained, the thermometer would still be in the state it is in. The presupposed background conditions must obtain for information to be carried in the strict sense, but one can use the same content ascriptions, and the same distinction, without making the assumption that relevant conditions in fact obtain. One can say, that is, that x indicates that P if it is in a state that would carry the information that P if the appropriate background conditions obtained. If the conditions do not obtain, then what is indicated may be misinformation rather than information. But the essentials of the story remain the same.

Perceptual systems are paradigm cases of systems apt for receiving information, and statements about perceptual achievements are cases of content attribution that most straightforwardly fit the information theoretic picture. To say that O'Leary sees that the zebra is striped is to say, at least, that O'Leary receives via the visual system, the information that the zebra is striped. In this kind of straightforward perceptual statement there is no question of misinformation or misperception: the zebra must be striped for the statement to be true. Furthermore, if it is true that O'Leary sees that the zebra is striped, he must also come to believe - in fact, to come to know - that the zebra is striped. Suppose O'Leary thinks this: "That animal sure looks striped, but who ever heard of a striped horse. It is probably just the way the sunlight is filtering through the trees that makes it look that way." O'Leary is wrong to doubt his senses, let us assume. It really is a striped zebra that he sees; lighting is normal, and things are just as they appear to be. I assume that in such a case it would be wrong to say that O'Leary sees that the zebra is striped, though it might still be right to say in such a case that O'Leary received the information, through the visual system, that the zebra is striped. Despite the fact that O'Leary withholds judgment, it is still true that had the zebra not been striped, it would not have looked the way it does.

In the normal case - the one that is correctly describable by the statement, "O'Leary sees that the zebra is striped" - the information that the zebra is striped is received through the visual system and results in the knowledge that the zebra is striped. A case might be abnormal in at least two independent ways. First, the "information" might be misinformation: to use the terminology suggested above, it might be that the state of the visual system merely indicates that the zebra is striped, meaning that it is in a state that, under normal conditions, would constitute receiving the information that the zebra is striped. Second (as in our example), it might be that the information (or misinformation) did not result in belief, and so not in knowledge. O'Leary's case deviates from the norm in the second way, but not in the first. It is also true that in this case, O'Leary falsely believes that it deviates from the normal one in the first way.

To describe what happens in a way that does not exclude either kind of deviation from the normal case, one might say that it looks to O'Leary as if the zebra were striped, or that it appears visually to O'Leary that the zebra is striped. (though this is not how O'Leary would put it, since he doesn't realize that what appears to him to be striped is a zebra.) But how exactly are such statements about how things appear to be analyzed?

One idea, considered and rejected by Evans, is this: statements about the way things seem should be understood as dispositions to believe, or as something like defeasible or prima facie beliefs. For it to seem to you that P is for you to be in a state that would, except for some intervening factor, result in the belief that P, or perhaps for you to be in a state that produces an inclination to believe that P. More specific seeming or appearing states such as its looking to the you that P, would, on this strategy, identify a particular source for the inclination: something seems to be so as a result of the way things look. But on this proposal, one would still derive the content of the state from the content of the belief states that it would produce under normal conditions. Evans argues that this gets things backwards. Belief states are sophisticated cognitive states, while "the operations of the informational system are more primitive." [p. 124] But I think the main problem Evans has with this attempt to explain the content of seeming states is not that the perceptual systems are intrinsically more primitive, but that they are earlier links in an informational chain. The informational content of a state should be explained in terms of the causal source of the state, and not its normal result. Consider the analogy between testimony and the senses, an analogy Evans alludes to and takes seriously. [p. 123,n. 5] The senses are like witnesses who tell us things that we may accept or reject. Just as in the normal case, when it looks to be that P, one comes to know that P, so in the normal case, when one is told that P, one comes to know that P. This implies that to be in the state of having been told that P is to be in a state that would under normal conditions result in the knowledge that P. But it would obviously be absurd to try to explain the content of the witness's testimony in terms of the content of the knowledge state in which it would normally result. Similarly, it gets things backwards to try to explain the way things look in terms of what one would come to believe if one judged that things are as they look.

This seems persuasive, but the way the contents of the components of some informational system are determined, and the way such contents are related to the way things seem, may be more tangled than this analogy suggests. Consider another kind of situation in which information received by the visual system fails to result in knowledge, this time a case where the explanation for the failure is that the subject lacks the conceptual resources to form the relevant belief. Eucalyptus trees, let us suppose, have a quite distinctive look. If the tree in the garden were of any other kind, then things would look differently to O'Leary than they do, so we might say that O'Leary's visual system receives the information that there is a Eucalyptus tree in the garden. But O'Leary doesn't come to know, or believe, that there is a Eucalyptus tree in the garden, since he doesn't have a concept of a Eucalyptus tree (by that name, or any other). While O'Leary's visual system succeeds in discriminating Eucalyptus trees from other kinds, O'Leary himself does not. Now it does not seem to me that this situation is correctly described as a case in which it appears to O'Leary that there is a Eucalyptus tree in the garden. Why not? One might think that it cannot be right that things seem to someone to be a certain way unless the person has the capacity to endorse the appearance - to judge that things really are that way. If this is right, then the knowledge and beliefs that normally result from perceptual states may constrain the content properly attributed to the kinds of states that are ascribed when one says how things look. appear or seem to be.

But whether this is right or not, it is clear that one cannot, in all cases, identify the way things seem, or look, with the information received or delivered by some component of the informational system, some link in a chain of information transmission that in normal cases results in knowledge. It is surely essential to seeming that the way things seem be accessible to consciousness, but this need not be true of information bearing states that are part of a process that normally results in knowledge. Consider a very early stage of the visual processing system such as the retina. Suppose the image of a tree is on our subject's retina: it is in a state that it would normally be in when, and only when, the subject is confronted by a tree (as we may suppose he is). So the subject's retina indicates, and carries the information, that there is a tree before him. But suppose a distortion is introduced very early in the process, so that the subject is under the impression that he is looking at something that bears no resemblance to a tree. There is no sense in which it would be correct to say, in this case, that it seems to him (retinally speaking?) that there is a tree before him. But the retina surely does carry information, and is in a state that has representational content. The retina is part of the informational system that "constitutes the substratum of our cognitive lives." In the normal case when the information on the retina is successfully transmitted up the line, it contributes to determining the content of the subject's states of knowledge and belief, and to the way things appear to be to him. And since I assume that it will be agreed that the retina does not have conceptual capacities, I assume that it is safe to conclude that states of the retina have only nonconceptual content.

Evans regards it as important to identify information-bearing states of perceptual system with states of seeming since he is anxious to avoid the traditional epistemologist's picture according to which the subject receives, through the perceptual systems, sensory data that is "intrinsically without objective content," but which forms the basis for inferences about the world that causes them. "The only events that can conceivably be regarded as data for a conscious, reasoning subject are seemings - events, that is, already imbued with (apparent) objective significance." [p. 123] But information-bearing states of all kinds, even those of things that are too primitive for anything to seem to be some way to them, are imbued with objective significance. This will be true whether one is talking about the rational judgments of an articulate and conceptually sophisticated person, or about the way things look to such a person, or to an animal, or about the image of a tree on a retina, or about that of the moon on the surface of a pond. To attribute informational content to the state of someone or something is to make a claim about a relation between that person or thing and its environment, and so is to make a claim that is in part about the environment - about the kinds of things that are found in the environment, and about the way the states of the person or thing are disposed to reflect the properties of those things. So, for example, if it is true that it appears to O'Leary that the zebra is striped, then there must be zebra that O'Leary is looking at. The claim that O'Leary's retina indicates that the zebra is striped has the same consequence. Of course O'Leary's visual system might be in an intrinsically indistinguishable state even if there were no zebra present, even if he were seeing nothing, but was hallucinating. In that case it couldn't be true, or even intelligible, to say that it appears to O'Leary that the zebra is striped. We would have to say that it looks to O'Leary like there is a stripped zebra in front of him, and this would be to attribute a different property to him - to make a different claim, one that relates his visual system to a different piece of information - a different informational content.

One might be tempted to think that the real content of O'Leary's state of seeming is one that abstracts away from the environmental dependence. The zebra can't be part of the real content of the way things look to O'Leary, since things could look just as they do even if it weren't the zebra that looked that way, or even if nothing looked that way. But it would be a mistake to yield to this temptation, since one cannot eliminate the environmental dependence without getting rid of informational content altogether. (Suppose O'Leary were in an intrinsically similar state in a world in which pigs looked exactly like zebras in fact look, and vice versa. Then it would look to O'Leary like there was a striped pig before him. This counterfactual possibility should not incline us to say that it is compatible with the way things look to O'Leary that there is a striped pig before him.) If one succeeded in purging the content of perceptual states of their environmental dependence, what would be left is sensory data, "intrinsically without objective content." One would be left with the dreaded myth of the given.

Ascriptions of informational content are external in that the concepts used to ascribe them are not thereby attributed to the subject. When I make a claim about the informational states of some subject, I use my concepts to describe the way her environment is disposed to affect her, but the concepts I use may or may not be ones that she shares with me. That is why it does not matter very much whether the parts of the informational system to which we attribute content and that we take to be the proximate sources of the information that is the content of our beliefs and knowledge are themselves accessible to introspection - are properly described as states of seeming. We use conceptual resources to refer to the contents of states of seeming, and also of more primitive informational states, but we do not thereby refer to some kind of content that has the concepts we express as constituents. We can give a reasonably clear account of a kind of abstract object that satisfies our conditions for being a kind of representational content, and that is apt for describing both primitive states of informational systems and the states of belief and acts of judgment of sophisticated reasoners.

My concluding question is this: why shouldn't one take the contents of belief and judgment to be the same kind of content as the kind used to characterize the more primitive information carrying states? Let us grant (without looking too hard at what this means) that states of belief and judgment are essentially conceptual - states and acts that require the capacity to deploy concepts, and that manifest the exercise of this capacity. That does not by itself imply that the concepts that subjects deploy and are disposed to deploy when they are in such states or perform such acts are thereby constitutive of the contents that are used to describe the states and acts. Even if both I and the subject to whom I attribute beliefs must be assumed to have conceptual capacities, it might be that my concepts and hers are different - that we cut the world up in somewhat different ways. If your concepts are different from mine, then I would be unable to use my concepts to attribute beliefs to you if in doing so I were saying that you had beliefs with those concepts as constituents. But we might think of concepts as part of the means used to refer to informational contents, in the way we do when we ascribe content to more primitive informational states that do not involve the deployment of concepts at all. In general, we might say, to attribute content is to characterize various kinds of internal states of others by describing how they tend to vary with certain alternative states of the environment (or more generally, the world). We use our own conceptual resources to distinguish the alternative states of the world, but do not thereby imply that the subject uses the same means to distinguish the alternatives. We might say this even about information-bearing states that belong to what I refuse to call the space of reasons, or the realm of spontaneity.

Gareth Evans proposed that we distinguish different kinds of informational states, that are characterized by different kinds of content. John McDowell argued, in his criticism of Evans, that both kinds of states had the same kind of content: content is conceptual all the way down. I am inclined to agree with McDowell that the different kinds of states have the same kind of content, but I am suggesting that it is nonconceptual all the way up.



Department of Linguistics and Philosophy

Massachusetts Institute of Technology