Published in the Electronic Journal for Analytic Philosophy, January 1999
Copyright @ Adrian Cussins
This paper explores some problems with Gareth Evans’s theory of the fundamental and non-fundamental levels of thought. I suggest a way to reconceive the levels of thought that overcomes these problems. But, first, why might anyone who was not already struck by Evans’s remarkable theory care about these issues? What’s at stake here?
I shall say that experience and judgement are different modes of cognition. I mean only that they are different ways in which people have or exercise knowledge of the world. There are diverse aspects to this difference; for example, the kind of epistemic commitment involved in experiencing the world differs from the kind of epistemic commitment involved in forming judgements about the world. But that will not be my topic here. There are also differences in kinds of content: the characteristic content of experience differs from the characteristic content of judgement, which is truth-evaluable thought content. I will not here try to defend a version of Evans’s claim that the content of experience is nonconceptual as contrasted with the conceptual content of judgeable thought. Philosophers have often thought that the content of experience is subjective, at least in part, whereas the content of judgement is, or ought to be, objective. Evans, for example, emphasizes the egocentricity of how perceptual experience presents the world, and contrasts the objectivity of his fundamental level of thought. And related to this, it has often been held that judgements which are based directly on experience—for example, observational judgements—carry an essentially subjective element, whereas judgements based on (fundamental) thoughts may be wholly objective. It is this (supposed) third difference betweeen experience and judgement that I want to pursue here, in connection with the distinction between the non-fundamental and the fundamental levels of thought, as applied to thoughts about material, spatial objects.
As we will see, a central category, for Evans, of non-fundamental concepts of spatial objects are those which are based on experience, especially perceptual experience. For example, demonstrative concepts of material objects in the immediate environment of the thinker, which the thinker is now perceiving, and where the demonstration of the object relies on its being perceived. The way these concepts are of their objects is fixed by the way their grounding perceptual experiences are of their objects. Now if the intentionality of these concepts is fixed by the intentionality of the perceptual experiences on which they depend, and the perceptual experiences are subjective, then it is at least going to be a concern that thoughts based on these concepts will, in part, be subjective. Hence the felt need for an independently grounded level of thought which is not, in part, subjective. On one reading, Evans’s fundamental level is a level of concepts of objects that are wholly objective; where being “wholly” objective means being not at all subjective. It may be that insofar as human minds realize a fundamental level of thought, concepts that are “wholly” objective are rarely, if ever, grasped; nevertheless, this sort of objectivity is the ideal that governs the fundamental level of human thought. We need the fundamental level of thought to secure thought’s objectivity, as an ideal, and as best as is possible given human limitations.
So, on this reading—a popular reading of Evans—the distinction in cognition between the subjective and the objective maps on to the distinction between the non-fundamental and the fundamental levels of thought: the subjective is non-fundamental and the fundamental is objective. But, surely, what is not fundamental is not necessary, so we can explore the idea of cognition which exists entirely at the fundamental level. We can conceive of, and explore in thought, what on this view would be an idealization of human thought: a creature whose cognition was exclusively “wholly” objective because all of the creature’s thought operated at the fundamental level of thought. I will call such a creature a “Fregean Angel”. Now this is what is at stake here: Is human thought such that its idealisation would be the thought of a Fregean Angel? We could pursue some philosophical questions about the nature of human persons by engaging in a quasi-theology: if the angels are modelled on the best in humanity, then how should we conceive of the angels? Personally I am not enamoured of Fregean Angels, but not because I don’t think we should strive for what is best in our cognitive natures. I’d rather craft a different theological form, but what? And what theory of cognition could legitimate such an alternative form?
I won’t have very much to say about these last two questions here, but I want to do some preparatory work by exploring, in the context of Evans’s account, what goes wrong with the idea of the Fregean Angels. In seeing what goes wrong with Angelic thought we can better see the nature of the error in a theory of thought in which there is a “fundamental” level at which objectivity is secured, which is autonomous of subjective cognitive phenomena, and which is therefore cognitively detachable from them. Maybe then we will be better placed to offer one or two ideas about how to put this right; about an alternative understanding of objectivity’s role in cognition. If, on the other hand, we find no incoherence in Angelic thought then we may be unable to resist a conception of human thought in which the subjective and the objective elements are assigned to functionally, and philosophically, distinct parts of the cognitive system.
One way to understand what goes wrong with a metaphysical picture of the world is to fill the picture out more, find oneself getting into difficulties in filling out the picture, and then to give a principled explanation for these difficulties. Suppose we were to attempt this for a picture of objectivity secured exclusively at the fundamental level, from which all subjectivity has been detached. What could such thought and its world be like?
Think first about the basis for a distinction between qualitative identity and numerical identity under such conditions. Normally, if there are two objects which are qualitatively identical to each other, we can understand that they are not numerically identical (that they are two rather than one) because we can imagine them given in different egocentric presentations to the thinker, or with different significances for action; for example, one on their left and one on their right. But this idea is not available in our picture of a “wholly” objective world, because all subjectivity is eliminated from such a world, and the idea of being ‘to the left’ or ‘to the right’ depends on a subject’s egocentric perspective on the world. Likewise significance-for-action depends on different embodied, subjective ways of being in the world. We need some other way of making sense of these qualitatively identical objects being in different places. Again, normally, we have no difficulty with the idea of unperceived distinct places, but that is because we are already conceptually equipped to operate with the distinction between qualitative and numerical identity: we make sense of unperceived different places because we can make sense of their being distinct particular objects located at those places. But we cannot appeal to that here because we are trying to make metaphysical sense of this distinction that is presupposed in our ordinary understanding of different places. What is entirely legitimate in ordinary practice leads in this metaphysical task to vicious circularity: numerical distinctness of objects is explained by appeal to distinct places which in turn is made sense of in terms of the idea of numerically distinct objects. Both Strawson and Evans emphasize the explanatory inter-dependence of objects and places, so in conceiving our metaphysical picture we cannot rest numerical distinctness of objects on distinctness of places, nor distinctness of places on distinctness of objects. At the metaphysical level of explanation these distinctions must be provided for symmetrically. But how?
It helps to start with a very simple world—a world of shapes—and ask about the conditions under which one can make sense of the idea of two distinct objects which are identically shaped. Strawson (1959) is concerned with the possibility of this distinction given that “we cannot at any moment observe the whole of the spatial framework we use, that there is no part of it that we can observe continuously, and that we ourselves do not occupy a fixed position within it”. In contrasting a situation in which the subject could monitor continuously the spatial relations amongst objects he imagines the ‘world’ depicted in the figure below. I want to focus on a component of this ‘world’, its bound or frame of reference, which provides an absolute spatial framework for the shape-world:
Even though there is no subjectivity in this ‘world’, we can make sense of the distinction between qualitative and numerical identity for the ‘world’ by means of its frame. The questions “which triangle?” and “which circle?” make sense only because triangularity and circularity can be differentially discriminated in terms of distinct relations to the boundary frame of reference. In what follows there are two aspects to the frame of reference of this ‘world’ which will be of importance: that the objects in the shape-world are always co-presented with their frame of reference, and that the frame of reference is not itself dependent on the identity and distinctness of the objects that are located with respect to it.
If a world is presented at all to a Fregean Angel, it must be presented in such a way that the entire extent of the world is laid out before it, as the shape ‘world’ is laid out for us in the Strawson figure above. Subjects or thinkers cannot act in this world; or, if they can, their thought about the world in no way depends on their action in the world. The entire extent of the world is laid out before the thinker, quite independently of activity or experience, or embodiment: the world as a purely formal object.
In sum: if the fundamental level of thought is the level at which objectivity is secured, and if the fundamental level can be detached from any non-fundamental level, then we ought to be able to make sense of an idealised cognitive creature that operates exclusively at the fundamental level of thought and therefore whose thought is ‘wholly’ objective in that all subjectivity has been excised from the thought of the creature. If thought of any world at all entails the applicability of the distinction between qualitative and numerical identity (for if it were not applicable there could be no objects) then the world presented in the thought of such an idealised cognitive creature would have to be laid out before it so that the bounds of the whole world were presented to it along with each object. The very possibility of objects in such a world depends on their being co-presented with the frame of the whole world. And the frame of reference itself is not dependent on the objects in the world, or on action on the world: rather, it is a transcendental condition on the possibility of objects at all.
This is not how it is with us! But it doesn’t follow that our cognitive theology is not like this. The question is this: Does the explanation of why things are not like this with us appeal to our imperfections; that, for example, some of our singular thought is dependent on perceptual experience, and our experience is egocentric and partial? For, if so, the idealisation of human cognition—which abstracts from human imperfections—would still be the cognition of the Fregean Angel. And therefore the functionality of human cognition would entail the coherence of the idea of a Fregean Angel, which would stand as its model, just as soviet sculpture in the former USSR modelled workers as their perfection: heroic, unceasing and strong. Or as Evans’s Generality Constraint (see below) models the less-than-perfect compositionality of humanly-grasped-concepts? On one view, what’s going on with Evans’s fundamental level of thought is to secure a core part of human cognition which—unlike subjective experience—would, if perfected, sustain a metaphysics of the kind illustrated in the Strawson figure. A question will be, what might drive Evans to that, and is it possible to satisfy the motivation without the Angelic theology? We will know that an alternative conception of thought is needed because, as we will see, Angelic cognition loses distinctions that play an essential role in the human conception of objectivity, even as an ideal for human thought.
Here is one reason one might think that that is what is going on with Evans’s fundamental level. Notice that the frame of the world in an Angelic metaphysics need not be geometric-spatial. It might instead be arithmetical: a constitutive sequence, expressible for example with the integer numerals, with respect to which the identity of numbers is given. Identity and distinctness of numbers, like the identity and distinctness of triangle shapes, makes sense only in relation to such a canonical sequence. Evans brings out something just like this when he introduces the notion of the fundamental level via his discussion of abstract objects, and especially numbers. Since it is the discussion of abstract, formal objects that leads up to the fundamental level in Evans, we ought perhaps to pay special attention to it.
Suppose in playing a game one asked, “What number is the square of a prime, the sum of two primes, and less than twenty?” It would be a true answer, but a bizarre one, to say “the number of planets”. It would be bizarre because the connection between the concept is the number of planets and the concept is the square of a prime, the sum of two primes, and less than twenty must be mediated conceptually (one can’t “see straight off” that “the number of planets” is the answer). One could only recognize “the number of planets” as a way of giving a true answer to this question if one was in a position to judge true the mediating thought, the number of planets is nine. That is, the way of thinking of a number which is the number of planets only connects to other ways of thinking of numbers via a canonical way of thinking of numbers in terms of their location in a canonical sequence, for example the sequence of integers. Hence, the role of the concept nine in the mediating thought. To give or understand “the number of planets” as answer to the question is a two-step process: (1) the number of planets is nine, and (2) nine is the square of a prime, the sum of two primes, and less than twenty. Since a thinker has to go via the mode of presentation of the number as nine, one wonders why the answer is not expressed using the same mode of presentation; hence the appearance of bizarreness.
In order to understand what it is for numerical properties to apply to a number, one must be able to think of the number in terms of its location in a canonical sequence, so we could call this way of thinking of numbers in terms of their location in a canonical sequence, the “fundamental” way. Other ways of thinking of numbers—for example, as “the number of planets”—are non-fundamental, because it is only possible to understand how concepts of numerical properties can be predicated of them if the thinker knows what it would be to identify a number so thought of with a number thought of in a fundamental way. (This restriction doesn’t, of course, itself apply iteratively to fundamental ways of thinking of numbers).
One could go on from here to the following idea: if the thinker were attempting to think about a number in a non-fundamental way, then we should say that the thinker wouldn’t know what they were thinking about if they didn’t know which number they were thinking about, and they wouldn’t know which number they were thinking about if they weren’t in a position to understand the truth conditions of a range of arithmetical predications of the number. But since arithmetical predications to numbers make sense only via a fundamental way of thinking of numbers, a thinker wouldn’t know what they were thinking about if they didn’t know what it would be for a thought of the form <this number is d> to be true where d is a fundamental way of thinking of the number. Satisfaction of “Russell’s Principle” requires a fundamental level of ways of thinking of numbers, and it requires the explanatory dependence of non-fundamental ways of thinking upon fundamental ways of thinking. Hence the necessity of the two stage procedure noted above, which is emphasized wherever Evans talks about the fundamental level of thought.
Notice the equivalent role played by the canonical integer sequence in thought about numbers and the absolute spatial frame of reference in thought about shape-objects in the Strawson-world. Both are absolute frameworks in that the identity and distinctness of their objects is explained in terms of the frame of reference, but the frame of reference itself is not explained in terms of the objects. Both frames extend to the limits of their world (there cannot be more objects than can be discriminated with respect to their frame). And in both cases objects must be co-presented in thought with their frame of reference: even if one thinks of the objects in a non-fundamental way, one’s understanding of which object is in question (its satisfaction of Russell’s Principle) rests on understanding of an identity claim in which the object thought of non-fundamentally (the number of planets, the shape I am thinking of) is identified with an object thought of in terms of its location relative to the absolute frame of reference (the number nine, the triangle at the top left of the frame). That is, the frame must itself be represented in thought, in the same kind of way as non-fundamental concepts are explicitly cognitively represented: understanding which object is in question rests on understanding the identity proposition, and in the identity proposition the framework concept d stands symmetrically across from the non-fundamental concept. Russell’s Principle is satisfied only because the frame of reference is rendered explicit at the fundamental level of thought. The frame itself figures as a conceptual constituent in thoughts which must be understood if even non-fundamental thought about particulars is to be possible.
Impressed by this account of thought about numbers, Evans—on this reading—develops an analogous account for thought about material, spatial objects. The analog of the canonical sequence which is constitutive of numerical identity and difference is an objective frame of reference for spatial objects. It is not possible that the frame of reference for human thought extends to the limits of its world, but the frame plays the equivalent role for human thought that the absolute frame plays in the number and shape examples: it acts to secure objectivity, in the sense of securing uniqueness and distinctness for objects without relying on experiential or other non-fundamental ways of thinking. In the material, spatial-world the frame provides for the possibility of thought which is from no point of view, which—by contrast with egocentric thought—is necessary for objectivity. Moreover, Evans claims that satisfaction of Russell’s Principle for thought about material, spatial objects rests on being able to bring into ‘alignment’ one’s egocentric presentations with a conception of the spatial world which is from no point of view. And, as we will see, Evans claims that satisfaction of the Generality Constraint for thought about material, spatial objects requires a fundamental level of thought at which a thinker can make sense of a full range of predications.
So in this domain too, thought about objects involving non-fundamental concepts requires a two-stage procedure: to understand the thought that <a is F> where a is a non-fundamental concept of an object grounded in egocentric experience, the thinker must know (1) what it would be for <a = d> to be true, for a fundamental way of thinking of an object, d, which is from no point of view because it exploits an objective frame of reference, and (2) what it would be for <d is F> to be true. Thus the frame of reference must be cognitively represented, so that it is itself presented in thought (“d”) whenever one thinks of a spatial object. Experience-based thought about spatial objects is explanatorily dependent on thought about objects at the fundamental level of thought, which is “wholly” objective, because “from no point of view”, because anchored by an absolute (if still partial) frame of reference. And fundamental thought of objects is not explanatorily dependent on experience-based thought; it can detach itself from experience-based thought. The objective frame of reference, on this view, is not a special way of employing egocentric presentations, or presentations of the world as activity, because it is constituted by a fundamental kind of concept, the ds, which present objects from no point of view, and therefore not experientially and not in terms of embodied activity. In this sense, then, the fundamental level of thought about material, spatial objects, like fundamental thought about numbers and about the Strawsonian shape-world, is a formal level of thought.
We need now to explore a little how frames of reference work, so as to be able to distinguish between ‘pathological’ and ‘healthy’ uses of frames of reference in cognition. I’m going to suggest that frames of reference, if they are to successfully provide for the possibility of singular reference must (1) not be explanatorily independent of the identity of particular objects that are located with respect to the frame, and (2), must not figure explicitly in thought, by being directly cognitively represented at a special level of thought. For, if they are so represented and they do so figure, they will prevent the possibility of successful singular reference. Thus we should abandon the model of the fundamental level of thought about material, spatial objects that we arrived at via the Angelic metaphysics of the Strawson shape-world, and Evans’s discussion of thought about numbers; the model in which the frame must be independent of and co-presented with the objects. We will need some other way to secure the ‘objectivity constraints’ of Russell’s Principle and the Generality Constraint. The payoff, I hope, is to gain a richer sense of the contrast between a formalist conception of thought and of metaphysics, and a non-formalist conception in which experience provides for the possibility of objectivity. This opens up a quite different conception of experience than that which is subjective-in-the-sense-of-anti-objective.
Let’s begin this discussion by distinguishing thought which is relative to a frame of reference from thought which is dependent on a frame of reference. First a couple of examples of frame-dependence.
A subject might make a reference to a chess piece which is relative to the frame of reference established by the chess board. I am thinking here of a game of chess not as a spatiotemporally located episode, but as a formal type whose existence is independent of any particular spatio-temporal context, constituted solely as a sequence of moves which might be materially manifested in any number of different places and times. For example, the subject might say “black’s bishop moves to Queen-Bishop-4” referring to a move in a game between Kasparov and Deep Blue. The reference to the chess piece is relative to the frame of reference which is established by the chess board; not by any particular chess board situated in the space through which the thinking subject moves, but relative to a general or abstract frame which does not belong to the unified spatio-temporal framework in which actions and material objects are situated. Such a reference is not only relative to a frame of reference but is frame-dependent because the chess-piece in question cannot be identified independently of the frame of reference established by the generic chess board. The idea of that particular makes no sense at all independently of the frame of reference of the chess-board. In such cases the ability to think about the chess piece is dependent—both explanatorily and ontologically—on the ability to think about the frame of reference of the chess board, but the ability to think about this frame of reference is not dependent on the ability to think about that chess piece. This sort of asymmetric dependence is characteristic of frame-dependent reference.
Consider my ability to think about Piggy which is relative to the reference-frame established by William Golding’s book The Lord of the Flies. I can talk quite happily about Piggy, even counterfactually about what Piggy would do in various imagined circumstances, but only so long as I stay within the reference-frame established by the story. If, walking down the street, someone asks me, “Is that Piggy?”, the question makes no sense unless it is reinterpreted to mean something like "Doesn't that person look just like what you imagine Piggy to look like?", or perhaps, "Isn't that the person who was a model for Golding's depiction of Piggy?", or something of the kind. I can identify the person walking down the street without necessarily having to identify anything else, but an actual identification of Piggy would require the identification of the objects and places that constitute the story-frame with respect to which references to Piggy are made. But the story-frame exists only in Golding’s world, not in Piggy’s. The cognitive resources for epistemic access to Piggy are exhausted by the frame of reference provided by the story. Therefore, Piggy-references cannot escape the story-frame of The Lord of the Flies; references to Piggy make sense only internally to the frame. References to Piggy are not only frame-relative. They are also frame-dependent. Again we have the characteristic asymmetry of frame-dependence: Piggy-references are situated and made possible by the story-frame, but William Golding’s story-frame cannot itself be situated within a space in which Piggy acts.
One might say about thoughts of Piggy and the chess piece that they are descriptive thoughts; that they identify their objects purely descriptively. But it is important to see that not all descriptive identification is frame-dependent. What is important to the argument is not so much the distinction between descriptive and singular content as it is the distinction between frame dependent identifications and frame relative identifications of objects. Thoughts based on certain kinds of definite description or universal quantification are frame-relative. If I think about the tallest Cornell Professor, qua tallest Cornell Professor, (and I have no other way available to me of thinking of this person) then I think about some particular person only relative to a reference-frame which establishes a set or a totality: the population of Cornell Professors, with respect to which the expression "the tallest Professor" may fix a unique object. In cases like this, a frame of reference is established not by a fictional story, but in order to interpret the semantics of a definite description, or of a quantifier ("all Professors", "most Professors"). My understanding of which particular Cornell Professors there are rests, in part, on my grasp of the totality of Cornell Professors, but it is also true that my understanding of the totality which is the population of Professors can (at least in principle) rest on my understanding of what it is to be each of the members of this totality. Hence my reference to the tallest Professor is not frame-dependent: the characteristic asymmetry of frame-dependence does not obtain in these cases. In this example we use a frame-relative semantics to interpret the definite description, relying on our background understanding of particular individuals (particular Professors) to ensure that the meaning of the description is not frame-dependent.
In certain other cases escape from a frame-dependent semantics may not be so smooth. Consider thinking about a person as the tallest spy. If I am given a totality—the population of spies—then I am in a position to think about a particular as the tallest spy. But how might I be given the totality? I do not have some prior general understanding of spyhood which determines its own application to particular objects independently of knowledge of those particulars. My understanding of the predication of spyhood to Jones draws equally on my inter-dependent understandings of spyhood and of the special character, history and circumstances of Jones. My understanding of spyhood is dependent on my understanding of what it is for Jones to be a spy, and what it is for Smith to be a spy, and so on for a small number of particular spies of whom I have knowledge. I may have an understanding of these very particular cases of spy-hood but not in a way which generalizes to allow me an understanding of what it is for an arbitrary particular to be a spy. So I cannot generate an intelligible idea of the totality of spies either through a general concept of the universal spyhood which independently determines its own application to particular objects, or through an inductive understanding—for the cases of Smith and Jones and so forth—of what it is to be a particular spy. The semantics, however, still requires the totality of spies. Since the totality cannot be given in terms of what is available to be understood in the predicate ‘spy’ or in particular instances of spyhood, it must be supposed to exist independently of what can be given. If the thought of the tallest spy is not to be an illusion, the world—but not the world as experienced, or acted on or otherwise made intelligible—must somehow provide a ‘frame’ which bounds the spies from the non-spies. Such a ‘noumenal’ frame would delimit the totality of spies, it would outrun my (and anyone else’s) understanding of the predicate ‘spy’, and it would be independent of the particular known cases of spyhood that fall within it. Thus the attempted thought about an object as the tallest spy would be frame-dependent; it would exhibit the characteristic asymmetric dependence of individual object on totality or frame. A putative thought about a particular by means of the descriptive content the tallest spy would, in that case, fail in a way which is analogous to the failure of a thought about the actual Piggy: the frame required for the intelligibility of the totality of spies is a fictional projection onto the world. Or, it might be preferable to put the point a little differently (picking up, for example, on a disanalogy due to the intentional fictionality of Piggy) like this: there could be nothing in our understanding which would provide for the intelligibility of a distinction between the frame of reference being fictional and its being actual; hence the idea of its presupposing a ‘noumenal’ world. But either way, both ‘Piggy’ and ‘the tallest spy’ are frame-dependent, and either way the frame-dependence disables an understanding of these terms as referring robustly to an objective reality. That this is so does not entail a complete breakdown in meaning; that’s part of what makes these cases interesting.
This discussion suggests a hypothesis: that where non-abstract objects are in question, successful and intelligible reference to a particular requires that, if the reference is frame-relative, it not be frame-dependent. Moreover, the distinction between reference to actual material particulars on the one hand, and reference to fictional objects, or virtual objects, or representational objects (the woman in the painting) or merely ideological objects (‘most spies’, ‘most liberals’, ‘most terrorists’, ‘most freedom fighters’), or objects which are defined by the rules of a closed formal system, on the other, is to be understood in terms of the distinction between reference which is merely frame-relative and reference which is frame-dependent. The hypothesis is that it is constitutive of any intelligible notion of being actual—and of not being fictional, virtual, representational, ideological or formal—that reference to the object not be frame-dependent.
If this hypothesis is correct then the account, considered above, of the fundamental level of thought about material, spatial particulars—which was suggested by the model of thought about numbers and motivated by a certain conception of objectivity—cannot be right. In order to establish a level of thought uncontaminated by egocentricity or subjectivity, appeal was made to an absolute spatial framework with respect to which spatial objects stood in a relation of asymmetric dependence. But, as we have seen, this kind of asymmetry is characteristic of frame-dependence. Therefore, on that reading of Evans, his fundamental level of thought would be frame-dependent, and so unable to sustain distinctions between reference to actual objects and reference to fictional or merely representative objects. But the applicability of such a distinction is a necessary component in objectivity. Hence, that conception of thought would fail to satisfy its governing motivation.
It does not help to respond that Evans does not suppose that his fundamental level can, in humans, eliminate subjectivity altogether. For the model of the fundamental level would still be in intact. On that conception of thought, subjectivity is only an imperfection, and so would entail the coherence of the Angelic metaphysics. And if the argument just given is correct, the Angelic metaphysics cannot sustain the distinction between the fictional and the actual. If reference to any material thing in the world were relative to the Angelic frame of reference which limns the extent of the world, then there would be no content to the distinction between the world’s being actual and the world’s being fictional. (The problem with being an angel is that there would be no gap between dreams and reality).
I will come later to the second objection to an objectivist theory of thought, but I want now to consider (a few ideas about) how an alternative conception of thought might work; a conception which does not abandon objectivity but reconceives it, not in oppositional contrast to subjectivity, but rather in contrast to being a mere information-processor. Thought, even at its most sophisticated and most fundamental, essentially entwines elements of subjectivity with elements of objectivity. Put more strongly the conclusion would be that the objectivity of thought entails the subjectivity of thought.
We can begin this part of the discussion with what Evans would classify as non-fundamental concepts of material, spatial objects. They are non-fundamental at least because they essentially rely on experience, and embodied activity, and therefore necessarily incorporate subjective elements. With our eye towards a more positive account, we are to consider whether these kinds of concepts could satisfy Evans’s guiding motivation to secure the objectivity of thought. And therefore whether objectivity can be secured without introducing a special level of thought—the fundamental level—which is independent of experience and subjectivity.
Suppose I think of a house as number 12, 32nd Street, downtown then my identification of the house is relative to the street grid and numbering system for downtown. My ability to think of the house is not, however, wholly dependent on the frame of reference provided by the street grid, because I can gain information from the house, or information which is house-directed, which is not relative to the street grid, and I am able to coordinate my judgements and my actions with respect to these different sources of information. For example, I have available to me not only descriptions which are given relative to a street grid, but also a map of downtown which locates the house relative to other objects based on distances and directions (and not the street grid) which are given relative to the map-based frame of reference. I may also have directions for driving to the house given to me verbally by a trusted and reliable friend which consist of a sequence of instructions to turn left or to turn right or to go straight ahead. When I get near to the house I can ask passers-by if they know where number 12, 32nd street is. I may have information through memory about the appearance of the house. What is important for my understanding of which house is in question is that, whether or not I decide to visit the house or to form judgements about the house, I have an epistemic capacity to coordinate information so as to guide my actions and judgements in a way which is appropriate or sensitive to that particular house. It may be that there is some misinformation in what is available to me, and that my capacity to coordinate amongst the information sources compensates for these errors, keeping my identification of the house robustly on target. An important part of the robustness of my identification of the house may consist in my ability to recover from guidance errors due to misinformation amongst my sources. This kind of coordination and recalibration of multiple frames of reference—even though each individual frame may be subjective—is what eliminates the threat of frame-dependence from my knowledge of which house is in question. I exploit many frames in conceptualising the house, but my cognitive resources for epistemic access to the house are not exhausted by these frames of reference.
References to particulars are often not frame-dependent because they have this kind of structure: they are based on the coordinations of many different frames of reference, which are illustrated in the house example. I do not mean that they are dependent on some number of frames of reference greater than one; that would result in frame-dependence just as much as the chess example. What I mean is this: a subject’s understanding of which particular is in question does not consist primarily in knowing the position of the object relative to one or more frames of reference, but rather in the subject’s knowledge of how to coordinate different frames of reference so as to be able to judge appropriately or to act appropriately with respect to the particular object. What I want to suggest is that if there is a fundamental level of cognition, it is not a special level of thought, or a uniquely privileged frame of reference, but rather a capacity to coordinate amongst frames of reference, and, when necessary, to generate and establish new frames of reference. We can, in fact, read in this way what Evans has to say about cognitive maps which ground his fundamental level. At the level of thought and reference there would be frames of reference, all of which would include indexical, demonstrative and first-personal components. Objectivity, however, would not be secured at this level, but at a nonconceptual level of capacities and map-like representational devices whose function is to provide for holistic coordinations amongst diverse frames of reference.
Coordination of frames of reference often involves sufficient redundancy to allow for the management of error (as well as the management of subjectivity). Because the subject’s understanding of which object is in question consists in the subject’s knowledge of how to coordinate amongst multiple frames of reference, the subject’s frame-internal knowledge is not autonomous. In such examples, a subject’s ability to think about the object in question may depend on the subject’s ability to think about a frame or frames, but it is also the case that the subject’s ability to think about the frames rests on the ability to think about particulars which are structured by the frame. It is because of this that errors in a frame of reference may not be damaging to either action or judgement, that the management of error can be robust, and that we can make sense of a whole frame of reference being in error. The subject’s ability to think about the house depends initially on his ability to think about the street grid, but, as the coordinations develop, his understanding of the frame of the street grid is also based on his capacity to think of the house, and other places, as given within a map of the area, or as reachable by a sequence of turns, or as having certain characteristic appearances from privileged vantage points. Successful frame-relative thought of a particular does not exhibit the asymmetric dependence of the chess, Piggy and “the tallest spy” examples, and this affects directly the kind of normativity (error, correctness, guidance, etc.) that governs adequate singular reference.
We can see both similarities and differences between these ideas about objectivity’s resting on coordinations amongst multiple frames, none of which are privileged, and an alternative reading of Evans that moves away from the account of the fundamental / non-fundamental distinction as the distinction between subjective and objective levels of thought. The alternative reading understands the fundamental / non-fundamental distinction as the contrast between thought which is relative to an egocentric frame of reference and thought which is relative to a holistic frame of reference. On this new view of Evans, a non-fundamental concept identifies its object relative to an egocentric frame of reference. A fundamental concept identifies its object relative to a frame of reference which is both non-local and holistic: it is non-local in that the body of the subject does not have a privileged role for identifications relative to the frame, and it is holistic in that the identification of each place is supported by the potential identification of any or all of the other places. So, when Evans requires for non-fundamental thought not only knowledge of location in egocentric space but also knowledge of what it would be for something identified egocentrically to be identical to something given at the fundamental level of thought, he is requiring that the subject be able to coordinate the two frames of reference: there must be cognitive resources for identifying a position in the egocentric frame of reference with a unique position in the non-local, holistic—and, in that sense, objective—frame of reference.
What Evans has to say about cognitive maps fits much better with this alternative interpretation than with the objectivist interpretation. It is true that Evans says (on page 152) that fundamental thought which is grounded in a cognitive map “is truly objective - it is from no point of view”, but Evans also explicitly denies that ‘from no point of view’ means from a God’s-eye point of view, or even from a third-person point of view. It may well be that Evans’s notion of objectivity here—unlike the notion which is contrasted with subjectivity—is quite compatible with identifications of objects and places which depend on the subject’s current spatio-temporal location, and which are therefore indexical, or first-personal, or demonstrative. But if the cognitive map sustains indexical identifications, what becomes of the distinction between the fundamental and the non-fundamental levels of thought? What matters for Evans about the fundamental level is not non-indexicality or anti-subjectivity or avoiding the contingencies of a subject’s idiosyncratic location, but rather securing objectivity via the holistic coordinations amongst objects and places. Consider, for example, how cognitive maps are introduced on page 151:
The places which we think about are differentiated by their spatial relations to the objects which constitute our frame of reference ... Hence a fundamental identification of a place would identify it by simultaneous reference to its relations to each of the objects constituting the frame of reference. A place would be thought about in this way if was identified on a map which represented, simultaneously, the spatial relations of the objects constituting the frame of reference. This identification has a holistic character: a place is not identified by reference to just one or two objects, and so the identification can be effective even if a few objects move or are destroyed. Our identifications of places has this holistic character whenever we rely in our thinking about places upon what has come to be called a “cognitive map”: a representation in which the spatial relations of several distinct things are simultaneously represented.
Notice that there is no reference in this passage to non-indexicality or non-egocentricity. It is holism which is emphasized. It is true that several paragraphs later, Evans does say that ‘each place is represented in the same way as every other; we are not forced in expressing such thinking, to introduce any “here” or “there”’. The alternative interpretation notes that each side of the semi-colon expresses the same claim, and that therefore we should interpret ‘not being forced to introduce [indexicals or demonstratives]’ as being equivalent to ‘representing each place [and object] in the same way as every other’. What I think Evans wants is (1) to think of a cognitive map as constituted by its uses, rather than by intrinsic properties; (2) that the uses of a map will often involve indexical and demonstrative modes of presentation, but that (3) this is still compatible with the cognitive map’s securing the objectivity of the fundamental level. The ‘heres’, ‘theres’, ‘thises’ and ‘thats’ disperse across the map in different uses of it; the map itself does not fix, at a time, some place as ‘here’ or some object as ‘that one’; the “you are here” pointer is not part of the cognitive map, but has a varying location in the map depending on the changing alignment of the egocentric frame of reference with the cognitive map. The map does not force us to use ‘here’ of a particular place or ‘that’ of a particular object, for that would be a map which allowed thought from only one particular point of view. That is how an egocentric frame works, but not a frame which is ‘from no point of view’.
This kind of interpretation is most strongly supported in a passage in chapter 7 (pp. 211-212):
Why should we suppose that everything that is true can be represented [in non-indexical] terms? ... Just as our thoughts about ourselves require the intelligibility of the link with the world thought of “objectively”, so our “objective” thought about the world also requires the intelligibility of this link. For no one can be credited with an “objective” model of the world if he does not grasp that he is modelling the world he is in - that he has a location somewhere in the model, as do the things that he can see. Nothing can be a cognitive map unless it can be used as a map - unless the world as perceived, and the world as mapped, can be identified. For this reason, I think that the gulf between the “subjective” and “objective” modes of thought which Nagel tries to set up is spurious. Each is indispensably bound up with the other.
This passage denies directly the thesis of detachability required for the idea of a creature whose cognition was “wholly” objective. The passage is ignored or set aside in objectivist interpretations of the fundamental level because it runs counter to so much else in the book (given that the book was unfinished at Evans’s death, it is not surprising that there are some tensions between, for example, chapter 6 and chapter 7). Thus in section 3 of the appendix to chapter 7—though in the context of offering an alternative formulation of the fundamental level—we find, “one must conceive the states of affairs one represents in one’s “egocentric” thoughts—thoughts expressible with “this”, “here” or “I”—as states of affairs which could be described impersonally”, which suggests, given the <a = d> form of Evans’s account, that the constituent concepts of the cognitive map are impersonal, non-indexical, and non-demonstrative, or would be so treated by the version of the theory being considered in the appendix.
It is at this point in the appendix—where alternative accounts of a subject’s satisfaction of Russell’s Principle are in play—that the threat of frame-dependence arises for Evans’s fundamental level. The matter is left unresolved, finishing with the plaintive “It still seems there is something right about §6.3 ...” What I think Evans came to see was this: Holism of the cognitive map is sufficient for it to represent each object and place in the map in the same way as the other objects and places that figure in the map. Holism is thus sufficient for not being forced to introduce a ‘here’ etc. for some particular place. But, nevertheless, holism is not sufficient for objectivity because—given that the extent of the map’s representation is very partial, unlike the ‘formal’ cases we considered—the identification of the frame objects and places would still be egocentric. (There may be no determinate fact as to which objects and places are the frame objects and places; in which case the egocentricity will distribute holistically across all the objects and places. It would thus remain true that ‘each object and place is represented in the same way as every other’ but that they are all somewhat “contaminated by egocentricity”).
One must conceive the states of affairs one represents in one’s ‘egocentric’ thoughts—thoughts expressible with ‘this’, ‘here’, or ‘I’—as states of affairs which could be described impersonally, from no particular standpoint.
But there is a problem about what exactly this requirement comes to—what exactly it is to know what it is for an arbitrary element of the objective order to be this, or here, or me. ... But it is not clear what we should make of the requirement if we do not suppose that the subject can formulate, and in favourable circumstances decide the truth of, propositions of these kinds [<a = d> and <d is F>]. And it seems that we are not entitled to that supposition. Section 6.3, for instance, gives the impression that the objective or impersonal mode of thought about space can be understood as a mode of spatial thinking organized around a framework of known objects and places—the ‘frame of reference’. But such a mode of thinking will not be capable of achieving a higher degree of impersonality that that achieved by the subject’s thought about the objects and places which constitute the frame; and (especially if we think here about Twin Earth cases), it seems plausible that a subject’s right to be counted as thinking about these familiar objects and places turns partly on his conception of the role they have played in his past life—being visited by him, seen by him, etc. ... In that case, the seemingly objective mode of thinking about space is, after all, contaminated by egocentricity. [pages 264 - 265]
If the problem is just left like this, it is devastating for Evans’s account, whose point all along had been to show how to satisfy a strong reading of Russell’s Principle compatibly with realist truth conditions; to show that “the requirement of discriminating knowledge [for thought] is to be justified outside a verificationist framework” (page 106). It is because of the problem of the appendix, rather than any other difficulty in the book, that it’s being an unfinished posthumous work has prevented the project from being completed.
To escape the problem entails abandoning the form of Evans’s theory in that it entails abandoning both the fundamental level of thought at which objectivity is secured, and the two-stage procedure for understanding (<a = d> and <d is F>) which was introduced, and made most sense, in connection with thought about abstract objects. But the motivations that were important to Evans are still intact, if we shift the objectivity-securing work of holistic coordinations to a nonconceptual level—not itself a level of thought and concepts—at which frames of reference are established, aligned and re-aligned with each other. We can, then, allow that every frame of reference will include ‘subjective’ elements, without compromising the objectivity of cognition.
It is possible for a subject’s ability to think about the frame (or frames) of reference to rest on a subject’s ability to think about particulars within the frame(s)—as well as conversely—only because the subject’s grasp of which particular object is in question is not exhausted by his grasp of the object relative to the frame(s) of reference. This independent understanding of the particular object is not possible for thought about chess pieces, but is possible for thought about the house downtown. What provides for this difference of understanding? What allows for knowledge of a particular not to be exhausted by frame-relative knowledge? In discussing the example I suggested that frame-relative understanding, if it is not to be frame-dependent, must rest on a level of cognition which coordinates amongst different frames of reference. And I want to suggest that these coordinations are possible only because both object and subject are situated in an active information environment which allows the subject to orient with respect to the object. (Contrast this with the relation between subject and object in—and out of—the Strawsonian shape-world). I cognize the house conceptually by means of the frame of reference provided by the street grid: number 12, 32nd street, etc. but my cognitive resources for epistemic access to the house are not exhausted by this and other frames of reference because I also have available to me trails of information that lead through the information environment to the house, and can flexibly guide activity which is oriented towards the house. These trails of information provide me with ways of cognizing the house which are not relative to the street grid or other frames. I am suggesting, in effect, that for singular thought to succeed—for the identification of the object not to be frame-dependent—it must rest upon ways of understanding in which the world is given to subjects as trails of information through an environment of activity.
The active information environment—if it is to sustain successful singular reference to the object—must situate both the subject (in that it is a space in which the subject can act) and it must situate the object (in that it is a space with respect to which the object has its identity). This can break down in various ways; I will consider a couple of examples in a moment. But we should recognize immediately that the notion of an active information environment is as subjective as it is objective; it is certainly not a notion of an absolute spatial framework. In order to get a better idea of both subject and object being jointly situated in an active information environment it will help to contrast Evans’s notion of an information link, with the information trails, which I discuss (briefly) here.
An information link exists only when a subject is experiencing an object, typically in perception. An information link can allow a subject, without inference, to form observational judgements about the object. A subject may judge that that object on the table in front of the subject is spherical on the basis of a visual information link, where the information link sustains the observation judgement without requiring any inferential structure. S does not have to form two judgements: that the j object is spherical, and that that is the j object, on the basis of which S can infer that that object is spherical. Rather, the information link puts S in a position to straightaway judge that that object is spherical. Moreover, an information link maintains epistemic contact between subject and object over time, so that the subject has an unmediated disposition to change his observational judgements over time as the properties of the object change. Evans’s information links are individualistic: different information links for different subjects. They are largely subject-centred: the work of maintaining the link goes on mostly within the information-processing systems of each subject. And they are sub-personal: although the deliverances of a link are personal, the link itself is not part of a subject’s cognition, either in experience or in thought. For this reason we cannot talk of the content of the link itself, only of its deliverances.
For Evans, the content of the deliverances of an information link has two components: it is standardly conceptual for the predicative component, which “can be specified neutrally, by an open sentence in one or more variables” (p.124). The non-predicative component is egocentric: it presents the object as standing in a bodily relation to the subject, which may be specified within the theory of content by means of a set of bodily axes, with the subject’s body at the origin of the space.
Peacocke (1983) motivated his claim that Evans’s requirements for non-fundamental singular thought were too stringent by considering a subject in a fairground who sees an apple through a complex array of mirrors, some of which may be moving in ways unknown to the subject. In this case the subject enjoys an information link with the apple which puts the subject in a position to form, without engaging in any inference, correct observational judgements about, for example, the color and the shape of the apple. Moreover, the apple appears in the subject’s perceptual experience as having an apparent position in egocentric space, though in fact the apple is not at this apparent position in egocentric space. The subject does not know where in egocentric space the apple is, or how far it is from him; a lack of knowledge which is manifested in his inability to reliably point in the direction of the apple, or to be able to walk reliably towards the apple. On Evans’s account there can be no adequate singular demonstrative thought about the apple in this case because the subject neither knows where the object is in objective space, nor does he know where it is in egocentric space. That the subject nevertheless enjoys an information link with the apple secures at best only descriptive thoughts about the apple, such as the thought *the apple which is causally producing these images is juicy* where the subject refers singularly to a component of the information link (“these images”) but only descriptively to the apple. Peacocke’s line by contrast is that the information link is sufficient for a singular demonstrative concept so long as “some conceivable additional evidence, experiences, and devices ... would allow the subject to locate the presented object”. Now, this can’t be right as it stands. As McDowell comments, “the bare existence of any information-link at all will make it conceivable that additional evidence, and so forth, would enable the subject to locate the presented object; Peacocke’s supposed extra requirement adds nothing...”.
But nor is it correct to return to the strict position which requires actual knowledge of position in egocentric space. Given Evans’s notion of an information link it looks as if Evans’s strict position and Peacocke’s liberal position are the only options: either the information link produces an egocentric deliverance which locates the object for the subject, or else it does not. In the latter case, since the link itself is sub-personal for Evans, it would only be by “investigating the details of the perceptual link” that the information link could sustain knowledge of which object was in question. But that would be descriptive, not singular, knowledge: “If the subject’s thought can make contact with the apple in Peacocke’s fairground case, as it perhaps can if the subject knows in general outline about the peculiar nature of the perceptual link between him and the apple, the ‘know which’ requirement is satisfied only in a completely different way: here it is the idea of the information-link between subject and object, rather than the information-link itself ... that carries the subject’s thought to the right object”. But McDowell here misses the possibility that the information link can be available cognitively to the subject as nonconceptual content, and therefore not via the idea (a concept) of the link. The idea is that the information link itself (not just its deliverances, but the whole trail) is experientially available to a subject as part of the environment of activity, as salient forms of guidance within the environment of activity. Because the trail itself is experientially available, the subject has unmediated, but still rational dispositions to go this way rather than that, or to judge that there is the house. Such actions and judgements are not the products of two-stage cognitive procedures mediated by a concept j: First I judge that I should go the j way, and then I judge that that is the j way; so I go that way. Rather I am in a position to see straightoff which way to go, because I see the forms of guidance, a part of the trails, laid out before me. I do not need to see the whole trail, only enough to start me on and then, at each moment, to keep me on the trail; the information environment does the rest.
If, as theorists, we use this notion of information trails through the environment, rather than the notion of an information link, we can secure genuine epistemic constraints on singular thought without the excessive requirement of actual knowledge of egocentric location. What matters about the fairground mirror example is whether or not the subject’s information environment—as it can be made available, perhaps only in parts, in the experience of the subject—is structured sufficiently for the subject to orient in the space of the object. Do the forms of activity guidance allow the subject, knowingly and reliably, to move towards the object, correcting for missteps if they occur? Do the forms of activity guidance allow the subject, knowingly and reliably, to retrace her steps so as to bring the object into view again? Do they allow for the knowing and reliable reproduction of object-directed activity? Given features like the fairground’s rotating mirrors it is unlikely that the answers to these questions are positive in this example; but the point is that the theorist of content needs to ask these kinds of questions: it is not just the private, egocentric deliverances to occurrent experience that matters but rather how an intersubjective information environment is structured by information trails, parts of which may be available at a time in subjects’ experience. If such an information environment is well structured and allows a subject to orient their activity in an object-directed way, then the information environment provides a nonconceptual foundation for singular thought.
Similar points can be made about less gerrymandered examples. It seems correct and important that I can think a singular demonstrative thought about Hunter Rawlings even though I have never met the man, would not recognise him if he were in the room today, do not know if he is and where he is in my egocentric space, nor do I know where he is in space given objectively. And I say this even though I agree with Evans about Russell’s Principle and about the falsity of the causal theory of singular thought, according to which it is sufficient for singular thought that my mental representations be at the end of an appropriate causal chain whose initial links were caused by the object which is the referent. The reason why I can think singular demonstrative thoughts about Hunter Rawlings even though I fail to satisfy Evans’s requirements for singular thought about him is that I am plugged into a rich information environment structured by a multitude of information trails that can guide action in a Hunter Rawlings appropriate way. The example is similar to the earlier example of my singular thought about the house downtown. In this case I can satisfy Russell’s Principle with respect to Hunter Rawlings, even though I don’t know his location, because his identity is currently socially embedded within the University and I am plugged into a rich set of socially mediated information trails that guide my action in a way which is appropriate to an object with that kind of socially determined identity. The subject, in this example, is located in an active information environment in that the subject is able to find their way through the environment; the object is located in the same information environment, and its location in the environment is, in part, constitutive of the identity of the object. (It is not, of course, necessary that the subject actually does locate the object).
We are now in a position to return to the argument against a detachable fundamental level of “wholly” objective thought. There is a second argument against the ‘objectivist’ conception of the fundamental level, which rests not on the characteristic asymmetry of frame-dependence, but on the necessity in frame-dependent reference for the frame itself to be co-presented along with the object of thought. We can introduce this argument via a discussion of an inadequate active information environment; not, this time, Peacocke’s fairground case, but Evans’s television example.
When Evans argues that information links are not sufficient for singular concepts he does so in terms of an example of a soccer player seen on a TV screen, and he argues that because of the circuitousness of the information channel the subject is not in a position to think non-descriptive thoughts about the soccer player. I think that it is fair to say that many readers have not been convinced by Evans’s example, and that perhaps even more have failed to understand it properly. A proper discussion of that example would involve a discussion of the generality constraint, and I want to avoid that here. So what I will do is to alter Evans’s example in a way that avoids some of the difficulties. One source of difficulty is that people often imagine that the viewing subject may recognize the soccer player, and that this recognitional ability would ground their singular thoughts about the player. And, secondly, even if the viewing subject does not recognize the player, surely the subject is plugged into an information environment in which there are information sources who would recognize the player, and through which the television transmission can be tracked allowing the subject to discover the identity of the player by for example finding in the newspaper which match was being shown on that channel at that time. We can avoid these difficulties by supposing that the TV images are images of coke cans, not soccer players, and that there are no information trails, for whatever reason, which lead from the TV display back to the source of the transmission. You can imagine, if you like, bizarre Peacockian fairground-style shenanigans which render the information links untrackable.
Suppose that an image of coke cans appears on a television screen, (and we do not know, nor have we any way to track where the information is coming from). One of us then points (as we might say, "to a coke can") and says something (as we might say, "about the coke can"). If I have identified some particular object then I have done so only relative to the frame of reference established by the television screen: my identification is frame-relative. My capacity to think about a particular coke can, which is sustained by the information provided through the TV, must exploit the frame of the TV: for example, the can which is presented at the top left of the screen. It's not that I must entertain such a descriptive thought which locates the can relative to the screen, but just that what I do entertain can only have a semantics, an interpretation which fixes a particular coke can, relative to the informational frame of the TV. The semantic interpretation includes the informational frame whether or not I am conscious of, or explicitly think of, the frame. For without such a frame to anchor my reference, I may attempt a singular demonstrative thought about the can — *That can is made of aluminium*—, but the semantic link to the can is sustained only descriptively: *The one and only can that I am now seeing through this information channel which is at the top left hand corner is made of aluminium*. Likewise for thoughts like *The shiniest can is made of aluminium*, and *All the cans are made of aluminium*. As in the discussion of *The tallest Cornell Professor*, the content of these thoughts is fixed relative to a given totality, of cans or of Professors. But in the case of the cans, the totality is fixed not by an independent identification of individuals within the totality but by the frame of the information channel.
Notice how this case is different from my thought about the house downtown. In that case I have frame-independent knowledge of which house it is because I can exploit a complex environment of information trails that lead downtown and enable me to locate the street grid frame of reference relative to other sources of information. And someone might say that in the coke can case I can similarly exploit my knowledge of the location of the frame of reference; that is, of the location of the TV. But this won't work, and seeing why it won't work shows up some similarities to the Piggy case. There are available to me information trails which lead downtown, and in virtue of which I know where downtown is. But these information trails are not part of the semantics of my thoughts about the house: they do not enter into the determination of the truth value of these thoughts. Without the information trails there would be nothing which would constitute my semantic understanding of sentences about the house, but there is no sense in which my thoughts are about the trails. The trails make for the possibility of my singular thought, but they do so at a nonconceptual level of cognition: the trails themselves are not co-presented with the object in the subject’s thought, but only—and occasionally—in experience. Not so for the coke can thoughts: for them, the information channel is part of the semantics. It is because of this—as well as, and related to, the characteristic asymmetry—that the TV-link can sustain only frame-dependent thoughts.
The house, identified relative to the grid of streets, is in the same space as downtown. Hence our knowledge of the location of downtown can ground our knowledge of the location of the house. But the coke can is not in the same space as the TV: the epistemic resources which are sufficient for the subject to act and judge appropriately in the environment of the TV are not sufficient for the subject to act and judge appropriately in the environment of the coke cans. (There is no single active information environment which situates both subject and object, and is such that the object’s location in it is, in part, constitutive of the identity of the object). Therefore our knowledge of the location of the TV cannot ground our knowledge of the location of the can of cola: the location of the TV provides only a virtual space for the cola cans. Indeed, the information channel available to us does not tell us anything about the spatial relation (either direction or distance) of the coke can to the observer, or to any object whose location is already known to the observer. It only provides information about the location of the coke can relative to other coke cans in the virtual space defined by the frame of the TV. The information trails which provide epistemic access to the frame of reference do not provide any epistemic access to the coke cans, whereas the trails which provide access to downtown do provide epistemic access to the house. In both the coke can utterances and the Piggy-references the object is identified relative to a frame which bears the wrong kind of epistemic relation to the subject and the object. Piggy's frame of reference can itself only be identified in literary space, as an abstract production of William Golding's, etc. Knowledge of that space is of help only in identifying entities, such as a literary work of art, with an abstract ontology, rather than the material ontology of Piggy-the-schoolboy. That’s why knowledge of the book “Lord of the Flies” fails to provide us with any conception of what it would be to identify Piggy in the street. Knowledge of the literary space may help in identifying Piggy-the-literary-archetype, but then knowledge of the frame of reference provided by the TV helps in identifying the coke can-qua-visual-image. Only if we change the ontology can reference to the frame, or knowledge of the location of the frame, provide a way to avoid the frame-dependence of reference to Piggy or to the cola can: from Piggy-the-boy to Piggy-the-abstract-archetype, or from the solid-material-coke-can to the coke-can-as-set-of-pixels. In such cases thoughts about the objects (Piggy, the coke can) require the semantic presence of the frame of reference; the frame itself enters into the semantics; or, in other words, the frame itself shows up as a constituent in the cognitive level of thought. Avoiding frame-dependence requires that the frame of reference not work at this level, but at a nonconceptual level which subserves thought. Which is another reason why the Strawson shape-world and thoughts about numbers are very bad models for what is best in singular thought about material objects, of even the most fundamental kind.
 Gareth Evans “The Varieties of Reference”, Oxford University Press, 1982
 One attempt is in Cussins “The Connectionist Construction of Concepts” in Oxford Readings in Philosophy: The Philosophy of Artificial Intelligence, edited by Margaret Boden, OUP 1990. A more recent version is in “Nonconceptual Content, Frames of Reference, Trails of Information and Singular Thought” (forthcoming).
 In the following discussion of the two levels of thought, I will be assuming in most cases that it is the distinction as it applies to spatial objects that is in question. Because, for me, the point of all this has to do with the relations between experience and judgement, we should consider the fundamental level as it applies to thought about objects which can be experienced. Perhaps we can have experiences of objects which are not spatial, but these will not be central examples of experience.
 Recognition-based concepts are also non-fundamental for Evans, and are also based on experience.
 I maintain the scare quotes around ‘wholly’ to distance myself from the conception of objectivity which is required for the idea of that which is “wholly objective”.
 That the non-fundamental is non-essential to thought is borne out in the discussion of abstract objects below. Part of why I want to resist this view is that I take experience to be essential to thought, and not merely necessary for imperfect, embodied creatures such as humans.
 Strawson, Peter “Individuals: An Essay in Descriptive Metaphysics”, London: Methuen, 1959, chapter 1, p.32.
 We need some directionality to the frame, to distinguish the sides from each other. There are interesting questions about what distinctions could be grounded in symmetries and asymmetries of shapes across the ‘world’.
 Consider the discussion in Evans (1982), pp. 106-7: “An Idea of an object is part of a conception of a world of such objects, distinguished from one another in certain fundamental ways... For example, we may say that shades of colour are distinguished from one another by their phenomenal properties, that shapes are distinguished from one another by their geometrical properties, that sets are differentiated from one another by their possessing different members, that numbers are differentiated from one another by their position in an infinite ordering, and that chess positions are distinguished from one another by the positions of chess pieces upon the board.”
The examples are all abstract: color shades, geometric shapes, sets, numbers, chess positions. And when Evans introduces the idea of the fundamental ground of difference he does so in terms of the example of the number three, and the shape square: “the fundamental ground of difference of the number three is being the third number in the series of numbers; the fundamental ground of difference of the shape square is having four equal sides joined at right angles...”
And when Evans comments that “evidently, we do very often employ such fundamental Ideas of objects”, he notes in a footnote that “this is especially clear with abstract objects”. The issue in this paper has to do with the consequences of using such a model for our theories of thought about material, spatial, concrete objects.
 Evans’s theory of content is governed by his use of Russell’s principle, the principle that in order to be able to think about an object one must know which object it is. The principle establishes a direct connection between thought and knowledge, so that to ask about the conditions for singular thought about material particulars is to ask about the kinds of knowledge that are required to sustain thoughts of these kinds. Evans’s line is that the ‘know which’ requirement is satisfied by knowing where the object is, either objectively in space given from no point of view (in which case the subject has a fundamental concept of the object), or else egocentrically by means of the egocentric deliverances of an information link (in which case the subject may have a non-fundamental concept of the object). See the discussion in chapter 4 of Evans (1982).
 Strawson calls this “story-relative identification” (Strawson, 1959, p. 18)
 So, there are two crucial differences between Evans’s account and the account that I prefer. First, that, for Evans, there is a category of singular thoughts—fundamental thoughts—which do not require the cognitive work of coordination amongst multiple frames of reference. Whereas, in my favored account, singular thought always requires such cognitive work. And, secondly, that, for Evans, the objectivity of singular thought is to be secured by the use of a special—holistic—frame of reference. This holistic frame of reference provides for Evans the fundamental level of thought, and thus it is still the case on this interpretation that the fundamental level is required for securing objectivity. Whereas, in my favored account, objectivity is secured only by coordinations amongst multiple frames of reference. There is no privileged, ‘objective’ frame: all frames of reference are, in some way or other, situated with respect to the active life of environments. My argument is that Evans’s commitment to a frame which is privileged as the objective frame of reference leads disastrously to the frame-dependence of his ‘fundamental’ level of thought.
 The distinction between frame-relativity and frame-dependence is not in Evans, but this seems a natural way to describe the problem that arises in the appendix. The issue of ‘contamination by egocentricity’ also arises at footnote 19 on page 152, which directs the reader to the appendix.
 I discuss these ideas further in Cussins (1992), and in “Nonconceptual Content, Frames of Reference, Trails of Information and Singular Thought”.
 One role for the concept of trails in the theory of thought is discussed in Cussins “Content, Embodiment and Objectivity”, MIND, 101, October 1992, pp. 651-688. See also “Norms, Networks and Trails”.
 This idea is to be found in Evans (1982) especially chapters 5, 6 and 7.
 John McDowell “Peacocke and Evans on Demonstrative Content”, Mind, vol 99, number 394, April 1990, [p.259]
 McDowell, ibid, pp.257-8
 ... and Evans and Peacocke
 By coincidence both the President of Cornell, and the tallest Cornell Professor.
 Evans (1982), chapter 4.