Published in Mind, 101, October 1992, pp.651-688

 

 

Content, Embodiment and Objectivity

 —The Theory of Cognitive Trails—

 

                             Adrian Cussins

                                               

 

 

 

List of Sections

 

(1)   Experience and Thought

(2)   Mind and World: Preserving a Symmetric Metaphysics

(3)   Specifying Content by Reference to the Realm of Embodiment

(4)   S/Ojectivity of Thought: Generality and whether the realm of embodiment makes available the realm of reference

(5)   Feature-Placing:Significance without Satisfaction

(6)   Environmental Ability Range

(7)   Cognitive Trails

(8)   S/Ojectivity of Experience: Generalizing along Two Dimensions

 

(1) Experience and Thought

            Frege was a powerful advocate of the explanatory priority of a theory of representation in thought over a theory of representation in experience.[1]  It is hard to overestimate the importance of this theoretical strategy.  It has had ubiquitous consequences: for our conception of persons—and the relation of persons to other animals—for computational and other technologies, and for our modeling of  psychological, social and semantic phenomena.  Most systematic theories of representation since Frege, however divergent from each other they have been in other respects, have followed Frege in adopting the priority of thought over experience.  The principle exceptions to this have belonged to the empiricist or positivist traditions; most impressively, perhaps, Carnap's Aufbau.  But theories in those traditions have been widely viewed to be failures in their own terms, and to make untenable assumptions of the 'given' in experience, and of the duality of representational scheme and experiential content.[2]  In this paper I explore one way to do representational theory which adopts the converse of Frege's priority—i.e. which explains thought in terms of experience—but which nevertheless is not an empiricist theory.

 

             I shall call representational theory which takes experience to be prior to thought "nonconceptual and constructionist".  It is nonconceptual because it does not take the atoms of representation to be concepts (the constituents of thoughts[3]) and it is constructionist because it attempts to explain thought in terms of nonconceptual atoms grounded in experience. (Experience is taken to be prior to thought, not merely independent of thought.  I don't consider 'eclectic' theories—recently popular in the philosophy of psychology—that employ mutually independent theories of thought and experience). The version of nonconceptual constructionist theory explored here is called 'the theory of cognitive trails'.  It is, in addition[4], neither empiricist nor materialist because it adheres to a symmetric metaphysics: it neither assumes (our conception of) the mind in order to explain (our conception of) the world, nor does it assume (our conception of) the world in order to explain (our conception of) the mind.  My aim in this paper is to give enough of a sense of the theory of cognitive trails to show the possibility of the enterprise. I don't present arguments against conceptualist representational theory, nor do I provide a theory equipped in all details, applied to specific philosophical problems, and accompanied with its own metaphysics and epistemology.  But I do try to show that the semantic tradition since Frege has explored only a partial sub-space of the possible territory available to representational theory; that a range of theoretical options which have been assumed to be unavailable can be legitimately pursued; and that the payoff may be handsome.  In other words: 'One hundred years after Uber Sinn und Bedeutung how about looking over here?!'

 

            The kind of theory which I explore has generally been assumed to be impossible (if considered at all).  We can see why by asking what is entailed by the relative priorities of experience and thought.  The concept of thought is introduced as it is in Frege: thoughts are those things for which the question of truth arises.[5]   Representation in experience is representation in embodied (eg. human) mental acts of perceiving, remembering, talking, writing, imagining, playing, 'thinking', exploring and other forms of acting and animal practice.  Priority may be taken to be the relation of logical or constitutive priority familiar within analytic philosophy: roughly, A is prior to B iff the concept A can be understood without the concept B, but not vice-versa.  Or: a theoretical reconstruction of the concept A need not appeal to the concept B, but a theoretical reconstruction of the concept B should appeal to the concept A.[6]  So when theorists of representation take thought to be prior to experience, they take truth and truth-conditions (hence also their functional determinants: satisfaction-conditions and the realm of reference) to be explained independently of a theory of embodied experience and practice (or not at all).  And they suppose that representation in embodied practice is to be explained in terms of the prior, and therefore given, notions of truth and the realm of reference.  The norms of experiential practice are, from the point of view of explanation, external to the practice itself.  The concrete, embodied practices of linguistic activity are to be understood, metaphorically, as aiming at an external target: the true.  Hence, the integrity of the target is not threatened by a lack of integrity in our 'target practice'.  The metaphor of a target at which we aim is pervasive in philosophy, and in those disciplines that make use of representational theory, even in the etiology of the word "intentionality" itself.[7]

 

            To give up on the metaphor and attempt to explain truth and the realm of reference in terms of embodied experiential practice looks unlikely at best.  If you start out with a conception of experiential practice which is independent of the truth (or correctness) norms which govern it, don't you thereby lose any grip on the practice as representational, as involving action or perception?  Action is goal-directed and answerable to intentions, perception is veridical or hallucinatory.  Just as the game of chess shorn of the aim of winning is an empty syntactic computation (no longer chess), so human practice shorn of external norms is mere physiology and physics.  (A practice of 'thinking' which does not aim at the true is not thinking, merely an experiential sequence.)  If we fail to specify a representation in terms of an element of the realm of reference (that which determines truth or warranted-assertibility value), don't we thereby fail to specify something with 'aboutness' or intentionality?  The philosophical tradition has not encouraged us to believe that if we start outside the circle of intentionality we will ever be able to find our way back in.

 

            Frege adopted a more Cartesian conception of experience, not as embodied practice, but as the private 'Ideas' of an individual.  The threat, however, is the same: to attempt to explain thought in terms of Ideas is to abandon the possibility of universal science, the possibility of inter-subjective communication and, indeed, the possibility for any rational basis for agreement and disagreement.[8]  If we don't start with truth, then we will be unable to finish with truth; representation will collapse into relativistic practice, and thereby into the loss of significance altogether.

 

            I shall show that these warnings about the dangers of life outside the Fregean semantic tradition are much overblown.  The reasonable alternatives to Frege's priority are  not exhausted by explanatory autonomy and by explanatory inter-dependence between the content of experience and the content of thought, for we can make sense of the more radical thesis of the priority of experience.  If we can understand how to start our theorizing without truth and the other denizens of the realm of reference, but nevertheless with a genuine notion of significant representation, then we will be able to survive on the outside for long enough to tell whether truth can be recovered.

 

(2) Mind and World: Preserving a Symmetric Metaphysics

            Many Idealists, phenomenalists and empiricists have attempted to start their representational theory with experience rather than with thought or the conceptual elements of thought.  But they have all supposed that starting with experience involved starting with Mind. Since explaining representation in thought involves an account (eg. an ontology) of the world, the project of explaining thought in terms of experience became, for them, explaining the world in terms of mind.  That is, for those who have taken experience to be prior to thought, asymmetric representational theory has entailed an asymmetric metaphysics.  And that has been their downfall.  The world is too large to fit in human mind; their only way to preserve a realistic world has been to appeal to the more capacious mind of a Berkeleian God.

 

            Frege combines an asymmetric representational theory (the priority of thought) with a symmetric metaphysics: he presupposes both world and mind (rational judger of thoughts).  The metaphysical presupposition of both mind and world follows from Frege's starting his representational theory with the sense / reference distinction in place.  The realm of reference is the world that we talk and think about (Dummett, (1973) p.198).  Sense is the mode of presentation of the world (the realm of reference) to a mind (subject of thought).  The realm of sense is thus the realm of cognition.  In making the sense / reference distinction basic to his theory of representation, Frege takes the cognition / world distinction to be basic to (ie a starting point for) his theory.  For Frege, a condition for the possibility of representation is the separation between cognitive subject and cognized world.  Therefore, representational theory cannot explain what it is for there to be a separation between cognition and world.

 

            By contrast, the theory explored here attempts a different strategy to maintain a symmetric metaphysics in the face of an asymmetric representational theory (the priority of experience).  Empiricists presuppose the mind in order to explain the world; materialists presuppose the world in order to explain the mind and Fregean rationalists—as we have seen—presuppose both mind and world.  But the theory of cognitive trails begins its representational theory with a notion of 'experience' which does not come equipped with a ready-made distinction between sense and reference, nor does it present itself as a part of the total experience of an experient.  Rather, the theory explains what it is for a sense / reference distinction to arise in experience: it provides an account of the logical genesis of the sense / reference distinction, and thereby of the experient / world distinction.  Hence mind and world are treated symmetrically.

 

            My aim in this paper is to indicate the possibility of a certain kind of representational theory, not to argue for its adoption.  But it should be noted that if the possibility can be made out, tremendous explanatory power is gained in virtue of not presupposing the sense / reference distinction.  Not just for the purposes of a representational theory suited to explaining learning, the acquisition of new concepts, or even the first concepts, in developmental psychology.  Nor just for the purposes of a representational theory suited to explaining the evolution of cognition.  Nor just for the purposes of a representational theory suited to explaining context-sensitivity, vagueness and particularity in semantics.  But also because of the need for a theory to sustain the plausible conviction that much of our intelligence in communicating and acting consists in our ability to move between alternative conceptualizations of a problem domain, rather than consisting largely in our ability to carry out inference within a given conceptualisation.

           

(3) Specifying Content by Reference to the Realm of Embodiment

            Gareth Evans (1982) introduced the notion of nonconceptual content, although he did so in a way which invited the charge of an empiricist distinction between scheme and content.[9]  It would be unfortunate, however, if the not properly worked out explicit statements about nonconceptual content in The Varieties of Reference were to obscure some remarkable resources for a nonconceptual constructionist theory that can be teased out of Evans's posthumous work.  A book that begins with a chapter of Fregean commitments leads by Part II to a sub-text (at least) that provides for the priority of experience over thought.  In this section I set up these resources so that they can be put to work for the theory of cognitive trails.

            A representational content is a presentation (or re-presentation) of the world in experience or in thought.  Our question is this: how is a theorist to capture—by means of a canonical (theoretically privileged[10]) specification—the nature of different contents that can be carried by a representation? Almost all semantic theories do so by having the specification refer to an element of the realm of reference.  They thereby abandon our ambition of explaining, rather than presupposing, what it is for there to be a mind / world separation: an independent world given in experience to an independent mind.  Is there an alternative to specification by reference to the realm of reference which doesn't abandon content?

            Evans saw that contents can be canonically specified by the theorist’s referring to abilities of the organism, where the abilities are not  (or need not be) part of the realm of reference.  The realm of reference is that with respect to which the correctness (eg. the truth value) of the content is determined.  Evidently, a subject in thinking of a coffee mug that it is full—a thought which is sustained by the subject’s perceiving the mug—is not referring to abilities to grasp the mug, to track it as it moves, or to be selectively sensitive to changes in its appearance.  Nevertheless, Evans’s idea was that the cognitive significance of the singular mug-content could be captured, and could only be captured, by the theorist’s referring to abilities to grasp the mug or otherwise to locate it, to track the mug through space and time, and to be selectively sensitive (in judgment and action and memory) to changes in the mug’s features.  These abilities are not available to the subject as the content’s referent, but they are available to the subject as the subject’s experience-based knowledge of how to act on the object, and respond to it.  The theorist may canonically specify the content by referring to abilities, because the cognitive significance of the content consists in the experiential accessibility of these abilities to the subject in experience-based knowing-how.[11]

 

            Let’s unpack this a little:- There is a realm of reference which, in this case, consists of the coffee mug, its liquid contents, its being half-full, the table, the mug's being on the table, and so forth.  Ontology is typically concerned with how the realm of reference should be specified: in terms of things, facts, situations, states of affairs, ... , but that is not my concern.  What is important here is that the realm of reference is that which determines whether a given content is correct (true, veridical, accurate) or incorrect (false, illusory, inaccurate).  And that the normal canonical specification of contents is achieved by referring to the realm of reference.

 

            But the animal also has a range of abilities, skills, dispositions and mechanisms, in virtue of which the animal is able to grasp the content.  This range will include sensory and effector mechanisms which are sensitive to, and can store and access information from the mug; for example, information about the mug's weight, colour and position.  It will include skills to act directly on the mug, and to behave appropriately with respect to it.  We may lump these together as the content's embodiment in the organism and in its environment.   Psychology is typically concerned with how the realm of embodiment should be specified: in terms which refer only to the organism, or to the organism / environment interaction; in computational terms, physiological terms, the apparatus of information processing, ... , but that is not my immediate concern.  What is important here is that the realm of embodiment, unlike the realm of reference, is not that which determines the correctness or incorrectness of a representational content.  For, evidently, the truth of the content that there is a mug on the table does not depend on the sensitivity of the animal's sensory mechanisms; it does not depend on the animal's grasping the content at all.  The realm of embodiment is distinct from the realm of reference.[12]

 

            It has usually been supposed that to attempt a canonical specification of representational content by referring to the realm of embodiment is to make the truth or correctness of a content dependent on an animal's contingent skills, abilities and mechanisms.  On the face of it, a theorist who attempts to provide canonical specifications of content which refer to the realm of embodiment is confusing the realm of embodiment with the realm of reference.  Yet I suggested above that Evans's insight was that content could be specified by reference to the realm of embodiment!  But this is no blunder:What Evans saw was how to pull apart the specification of content from the specification of reference or truth.  If a canonical specification of a (propositional) content need not be a specification of a truth (or verification) condition, then canonical specification of a content which refers to the realm of embodiment does not entail the evident falsehood that the truth of the content depends on the character of the realm of embodiment.

 

            The philosophical literature does, of course, contain suggestions of how content specifications need not be specifications of truth conditions, but rather specifications of verification conditions or conditions of warranted assertibility.  But this is not at all what I have in mind.  All such theories are anti-realist in one stripe or another: in specifying the verification conditions they take themselves to be directly specifying the realm of reference.[13] Since the realm of reference is the world that we talk and think about, the world is impacted by the limitations of our capacities; it does not transcend them.  But Evans's insight is entirely compatible with realism.  Content specifications which refer to the realm of embodiment are not direct specifications of the realm of reference.  Not only are they not specifications of truth conditions as the realist conceives of truth conditions, but also they are not specifications of truth conditions as the anti-realist conceives of truth conditions.  They are not specifications of truth conditions, under any conception, at all.  For Evans, truth conditions (and, more generally, correctness conditions) are fixed by the realm of reference, and not by the realm of embodiment; but the cognitive significance of representation is fixed by the realm of embodiment, and not by the realm of reference.

 

             Nor is this a dual component theory of content in which explanatorily quite independent theories are employed to account for what are taken to be autonomous components of ‘content’: one component concerned with reference and truth and one component concerned with the ‘narrow psychological’ explanation of behavior or functional role.[14] The neo-Evansian is neo-Fregean at least this far: the theory of cognitive significance (sense) determines the theory of reference.  For this reason, a realm of embodiment specification of content can be (although it need not be) an indirect specification of the realm of reference (of course, we have to show how this can be).  This is quite unlike dual component theory for which functional role does not determine referential significance.  For both the neo-Fregean and the neo-Evansian what it is to be an element of the world is to be the referent of a sense-content, and what it is to be a mind is to be the subject of a sense-content: representational contents are Janus-faced in that they are symmetric with respect to mind and world.  But for the dual component theorist the explanation of the structure of the world and the explanation of the structure of sense-content are independent.  Whilst the theory of reference-content is dependent on the theory of the structure of the world, the theory of the world is explanatorily independent of representational theory.

 

            Nor is the neo-Evansian engaged in an elimination of content.  In order to see how it is possible to canonically specify representation by reference to the realm of embodiment without abandoning content, we need to proceed in stages.  In the next few sections I am concerned with relating this central idea of content specification to the notion of objectivity.  The crucial point is that content specification by reference to the realm of embodiment does not presuppose the objectivity of content (the separation of mind and world), but—rather—by understanding the different ways in which embodiment can be made available in experience, we can better understand what it is for a world to be given to a mind.  This is why content specification by reference to the realm of embodiment is not directly the specification of truth or reference (since objectivity is not presupposed), and why it is the specification of content (since objectivity is explained).

 

            Evans's idea is phenomenologically very natural and persuasive (although trained as we are in the analytic tradition, it takes a little getting used to)[15].  What does my seeing the mug as being located there consist in?  What does my seeing it as a particular consist in?  The Evansian begins like this: the subject sees the mug as graspable, as locatable, as being such as to resist manual-pressure, as being drinkable-from, as being push-and-then-fallable ...  These descriptions that the theorist uses are not likely to be elegant or simple, for they are not available to the subject as descriptive conditions on the world.  They are, rather, available to the subject in the kind of way in which I know, on the basis of my kinaesthetic experience, how to raise my left hand to a point several inches above my right ear.  I know very well how to do this.  Moreover, I know that I know how to do it (I can, for example, make myself aware in imagination of each of the stages of the action-sequence).

 

            What is for the subject experientially direct is for the theorist highly complex.  The theorist has to make available in language those abilities that were available to the subject in experience.  In shifting from one representational medium (experience) to another representational medium (language), elegance and simplicity is replaced by hyphenation and complexity.  But given the theorist’s ambitions, trading elegance for substance is good business.

 

(4) S/Ojectivity of Thought: Generality and whether the realm of embodiment makes available the realm of reference

            There is a perceptual content which we, as theorist, know to be about the coffee mug in front of the organism, which is canonically specified by reference to abilities to track the mug, to orient with respect to it and to be selectively sensitive to changes in it.  These abilities are part of the realm of embodiment and are not part of the realm of reference.  Presuppose the organism’s possession of no other contentful capacities: no general mastery of objectivity and in particular, no background knowledge about what mugs are, or liquids, quantity measurement, tables and containment.  Then ask: Does the organism enjoy a content which is specified via the realm of reference in virtue of having this experience which is specified via the realm of embodiment?  Is a portion of an independent world (that which makes contents correct or incorrect) thereby presented to a subject (that which grasps contents, and acts on their basis)?[16]  That is, does a part of the realm of embodiment being given in experience-mediated knowing-how make it the case that a part of the realm of reference (the objective world) is given to a subject?

 

            Intuitively it does not.  An objective world is given to a subject if the content presents something as being independent[17] of the subject’s particular abilities, and particular location in space and time. But given only the realm-of-embodiment-specified content, all the ‘subject’ (experiencing organism) has is an experiential awareness of how to move etc. in response to local changes in its environment.  If this is a subject’s conception of a referent, it would be a conception of something as not independent of contingent characteristics of the subject itself.  The necessary separation between subject and object would not have been achieved.  If we could presuppose the sense / reference distinction then we could say that the content is a presentation of the mug to the subject in a looks-thus-and-so-and-reach-twist-and-graspable... kind of way: the content would involve the identification of a particular.  But specification by reference to the realm of embodiment does not presuppose the sense / reference distinction: we have to say that experience presents looks-thus-and-so-and-reach-twist-and-graspable...ly, and that is not to present something (a thing) as a particular object which is, in principle, publicly available or identifiable from any perspective.  If we attempted to give the content a referent, it would be a referent that was available only to the experiencing organism, and only when it was enjoying something like the present information-link.[18]  That is to say, it would be a necessarily local (and context-dependent) object, and hence no object at all.  Hence, as far as this content goes, there would be no subject at all; only an experiencing organism, since there would be no basis for the experient to think of itself as one element amongst others in the objective world.  Objects and subjects go together, or not at all.

 

            A start on the objectivity of content is this: that the content’s referent is given as public, as something which is, in principle at least, equally available to any subjective point of view[19].  A sign of this objectivity is that the content can be incorrect: If the referent is given as a public object, then it is always possible that the subject is wrong about the object, even where the referent is the subject oneself.  What we are after is a metaphysical distance between subject and object, a distance which makes intelligible the subject’s being wrong (and therefore also being right) about the object; which provides for the possibility of truth.  Such a metaphysical distance may obtain between the subject and itself, and between the subject and some of its own states.  For this reason the term ‘objectivity’ can be a misleading label for metaphysical distancing: since what is subjective is not objective, the subject’s own states of mind are—if anything is—subjective, yet the subject’s own states of mind may be given to the subject as object.  So instead of speaking of ‘objectivity’, I shall speak of ‘subject-object-ivity' or more simply 'S/Ojectivity'[20], thus characterising contents which exhibit metaphysical distance between subject and object.

 

            To explain why realm of embodiment specified contents need not present the realm of reference, but how, nevertheless they can do so, we need a much firmer grasp on S/Ojectivity.  The Fregean tradition of taking representation in thought to be prior to representation in experience takes S/Ojectivity of content to be generality.  I will first explain this equation, then see what the consequences are for realm of embodiment specified nonconceptual contents, and finally work towards a quite different conception of S/Ojectivity which is appropriate to the priority of experience over thought.  Once we have developed an account of S/Ojectivity in experience, we can explain thought as the content of S/Ojective experience, and then compare this new conception of thought with the Fregean conception.

 

            The correctness of a thought is truth.  Familiar semantic theories explain the truth of complex (quantified, negated, hypothetical, conjunctive) sentences of a formal language (a linguistic manifestation of thought) in terms of the truth of atomic sentences,[21] where atomic sentences are formed as the concatenation of a predicative (or relational) expression and the appropriate number of singular terms.  The explanation of how the semantics of (possibly multiply) quantified sentences depends on the semantics of atomic sentences entails what has become known as ‘the generality constraint’.  Thus: “[Frege] is making the assumption that, whenever we understand the truth-conditions for any sentence containing (one or more occurrences of) a proper name, we likewise understand what it is for any arbitrary object to satisfy the predicate which results from removing (those occurrences of) that proper name from the sentence, irrespective of whether we actually have or can form, in our language a name of that object ... [Frege is assuming that] from a knowledge of the truth conditions of ‘A(c)’, we can derive a knowledge of the conditions under which the predicate ‘A( )’ will be true of all the objects in a domain, when we do not and could not have the means of referring to each of those objects” (Dummett (1981), pp.17-19)[22].  From this it follows that, where a thought is completely grasped by a subject (ie. its formal expression in a logical language is fully understood by the subject), the understanding of the predicative component of the semantically basic atomic proposition(s) is conceptually independent of the understanding of the singular component, and conversely.  That is, mastery of who or what an atomic proposition is about cannot, where perfect understanding is in question, be dependent on mastery of which property is being predicated.  And mastery of what property is being predicated in an atomic proposition cannot be dependent on understanding which individual it is being predicated of.  Conceptual mastery of the singular and predicative components is independent of each other.  What is interesting is that this commitment to generality follows directly from the explanation of the dependence of the semantics of quantified sentences on the semantics of atomic sentences.

 

            Evans (1982) elevates this commitment of semantic theory to a constitutive claim about the S/Ojectivity of thought: “we cannot avoid thinking of a thought about an individual object x, to the effect that it is F, as the exercise of two separable capacities; one being the capacity to think of x, which could be equally exercised in thoughts about x to the effect that it is G or H; and the other being a conception of what it is to be F, which could be equally exercised in thoughts about other individuals, to the effect that they are F. ... If we make the claim [that the thought that a is F and the thought that b is G are structured in this way], then we are obliged to maintain that, if a subject can entertain those thoughts, then there is no conceptual barrier, at least, to his being able to entertain the thought that a is G or the thought that b is F.  And we are committed in addition to the view that there would be a common partial explanation for a subject’s having the thought that a is F and his having the thought that a is G: There is a single state whose possession is a necessary condition for the occurrence of both thoughts. ...  It is a feature of the thought-content that John is happy that to grasp it requires distinguishable skills.  In particular, it requires possession of the concept of happiness—knowledge of what it is for a person to be happy; and that is something not tied to this or that particular person’s happiness.  There simply could not be a person who could entertain the thought that John is happy and the thought that Harry is friendly, but who could not entertain—who was conceptually debarred from entertaining—the thought that John is friendly or the thought that Harry is happy.  Someone who thinks that John is happy must, we might say, have the idea of a happy man—a situation instantiated in the case of John (he thinks), but in no way tied to John for its instantiation” (Evans (1982), p.75 and pp.102-3).

 

            In this transition from Dummett's Fregean Semantics to Evans's Generality Constraint, the semantics of quantified sentences has forced a particularly strong reading of the subject / predicate structure of truth bearers: entertaining a thought is now held to entail that one's understanding of which property is attributed to the object is entirely independent of one's knowledge of what it is for any particular thought to be true: it must be independent of one's knowledge of what it is for the property to be instantiated in any particular case.  This effectively disconnects conceptual knowledge from experiential abilities (recognitional and discriminatory abilities) and practice.  Conversely, human embodied experience and practice can count as the thinking of thoughts only derivatively; as metre rules in everyday use count as metre rules only insofar as they are the same length as the standard metre rule in Paris, or as—in one theology—animals have souls only insofar as they are made in the image of God.  Embodied practice counts as representational only to the extent that its constituents approximate the conceptual independence required by the generality of thought.  Like the theist's animals, embodied practice is assessed by standards quite external to its own nature.

 

            The neo-Fregean generality constraint provides for a clear conception of S/Ojectivity, and thereby for a workable conception of the representational as what can be treated as conforming to the autonomous standard of generality for the purposes at hand.  Since the world is what is presented in S/Ojective content, and mind is the subject of S/Ojective content, generality yields a certain metaphysical view of the mind /world relation.  And the gap between thoughts proper (which meet the standard) and what counts as representational for the purposes at hand, propagates an epistemological gap between the real world judged of by the rational thinker, and the everyday world through which animal practice stumbles.[23]  So the important question for us becomes:  Is generality what S/Ojectivity consists in?    Is there an alternative account of S/Ojectivity grounded in experience, as generality is grounded in thought taken as prior to experience?

 

            The notion of nonconceptual content fixed by canonical specifications which refer to the realm of embodiment does not by itself help with this.  Such a notion doesn't provide an alternative conception of S/Ojectivity.  It is therefore vulnerable to the charge that realm of embodiment specified contents can only count as representational if they are treated as input to a conceptual system which satisfies the generality constraint.  If the nonconceptual contents are not assessed against a background of conceptual contents then they fail the test of generality.  What we theorists describe as the experiencing organism’s perception of a mug would not, for the subject, be a perception of a mug, because it would not be a perception of an object at all.  The presentation of which object it is which is seen is not at all independent of the presentation of how the object appears (and conversely), as the presentation of John was independent of the presentation of a happy man in the thought that John is happy.  This is why Evans introduces a quite separate layer of representation—"the fundamental level"— not specified by reference to the realm of embodiment, but by reference to the realm of reference.  A subject must exercise "basic conceptual skills" [p.227] in order to transform a nonconceptual experiential content into a conceptualized judgment.  The experiential representation "serves as the input to a thinking, concept-applying and reasoning system" [p.158].  Seen this way, Evans's transformation of Fregean representational theory is only from the priority of thought over experience to a no-priority theory.[24]  If experience is to be prior, then we have to do much more than develop a mode of specification of content appropriate to experience (eg. realm of embodiment specifications).  We have to show that realm of embodiment specification can support a notion of content which is not parasitic upon thought content.

           

(5) Feature-Placing: Significance without Satisfaction

             Since thought is introduced as that for which the question of truth can arise, the priority of thought entails that the conditions for a representation being a truth bearer (the S/Ojectivity conditions) are presupposed at the start of representational theory.  Hence, under the equation S/Ojectivity = generality, something counts as representational only as it is aimed at the independent standard of generality. But if representation in experience is to be prior we cannot start our representational theory with given conditions of S/Ojectivity (for to start with them is to start with representation in thought).  We need to start with a medium of representation which (pre-theoretically) is not (need not be) S/Ojective and then consider how S/Ojectivity might be built up in such a medium.

 

            It might turn out that the S/Ojectivity that we construct is generality, or it might not.  If it does, then we shall have to consider whether the representationality of the medium with which we began can survive independently of being aimed at this standard of generality.  For if the representationality of the medium is explanatorily dependent on the concept of generality, we shall have to abandon our attempt at developing a theory of representation in which experience is prior to thought.  But if it doesn't turn out that the constructed S/Ojectivity is generality, then we shall be working within a theory in which experience is prior to thought, and for which thought content is not  identical to experiential content which satisfies the generality constraint.  We should then have to characterize the new notion of thought content (and therefore a new notion of truth) and consider what interpretation the new representational theory gives to generality, and what role it assigns to Fregean thought contents.

 

            So let us start our representational practice—for which the conditions of S/Ojectivity are not given—without the norm of truth (or other norms given in terms of the realm of reference; eg. veridicality).  But this looks hopeless.  What could it be to be a content which does not bear a truth value (and is not a constituent in a content which presents the world S/Ojectively and has a truth value)?  After all, if something is a content it is a presentation of the world.  And any presentation of the world is a presentation of it as being one way or another.  But if the world is the way in which the content presents it as being then the content is true, and if the world is not the way in which the content presents it as being then the content is false.  What sense can we give to the phrase “a non-S/Ojective presentation of the world”?

 

            In a linguistic medium it is subject / predicate structure which makes for the possibility of truth, for an atomic statement will be true iff the object referred to by the subject has the property referred to by the predicative term, or if the objects referred to by the subject terms stand in the relation referred to by the predicative term.  (Generality, as we saw, provided for a particular interpretation of subject / predicate structure).  So we should consider whether a language can function successfully even though the language does not have subject-predicate structure.

 

            Strawson introduced the idea of a “feature-placing” language.[25]  The paradigm examples of feature-placing sentences are “it is snowing” (or ‘snoweth’), “it is raining here”, “wetness!”, or “jam here”.  The terms “snow”, “rain”, “jam” and “wetness” do not serve to characterize particulars which are introduced elsewhere in the sentence.  Snow, unlike being made of snow, is a kind of stuff not a property of a particular.  Nor are these terms sortal universals like ‘man’ or ‘rabbit’.  Sortal terms do not serve just to introduce a feature, but rather a sortal property or kind, for the identity conditions for what ‘man’ applies to are determined within the meaning of ‘man’ itself.  The terms ‘man’ and ‘rabbit’ divide their reference, so that it makes sense to ask ‘how many men are there?’, or ‘how many rabbits do you see?’ but it does not make sense to ask ‘how many jams are there?’ or, ‘how many rains do you see?’.  These latter sentences can be interpreted only by adding criteria of identity and distinctness to the features; thus ‘how many pots of jam are there?’ or, ‘how many rainstorms do you see?’

 

            Strawson’s original idea was that subject / predicate sentences are semantically complex in relation to feature-placing sentences.  Whereas the semantics of subject / predicate sentences involve the identification of a particular (an object) and the ascription to it of a property, the semantics of feature-placing sentences do not involve the identification of a particular, and hence do not involve the notion of properties of particulars.  Rather, they involve the placing of features.  Typically we can form a feature-placing term by taking a term and stripping from it its conditions of identification and re-identification.  For my purposes, a feature-placing sentence is a sentence for which sensitivity in experience to the assent conditions of the sentence does not require knowledge of any identification or re-identification conditions associated with the (feature-placing) term from which the sentence is formed.  Or: for which the ability to discriminate a feature does not require the ability to identify something as a countable (referrable) item, nor the ability to re-identify it as the same again; it does not even require knowledge of what it would be to re-identify it.  Whereas the semantics of subject / predicate sentences involves instance identification, the semantics of feature-placing sentences involves incidence indication.[26].

 

            An instance of soggy or bliss or loud can only be spoken of once sortal concepts (meringues-thrown-at-face, wedding-day, bang, etc.)—concepts which divide their reference—are injected into the feature-placing repertoire.  So, if we consider sentences which are restricted to feature-placing, then the semantics will have no use for a particular being an instance of a property, hence no use for logical subjects or for predication.  The significance of such sentences is restricted to the general indication of the presence of features.

 

            One can become confused about what is at stake here by supposing that these remarks depend on weight being put on a distinction between a particular and a process, or on a distinction between a localized particular and a distributed particular.  But this is not so.  The identification of a particular is, essentially, the identification of something which can be reidentified over time and through space.[27]  The possibility of spatial and temporal reidentification is as appropriate in the case of a tropical storm as it is in the case of a chair.  A particular, then, is something—either process or object, either localized or distributed—whose identification entails the possibility of reidentification.

 

            Strawson uses the example of a 'naming-game' in which children respond to "presence of cat, or signs of the past or future presence of cat, but do not think identifyingly of particular cats."  The ability to make identifying references to cats involves the ability to recognize cat-feature; but the ability to recognize cat-feature does not require the conceptual resources for identifying reference to cats.  "We can readily enough acknowledge that the introduction of particulars is so fundamental a conceptual step as to leave the primitive pre-particular level of thought as, at most, no more than vestigial in language."

 

            We could also picture a non-linguistic mode of activity in which a theorist would use feature-placing sentences to describe the psychology of a simple marine animal. Suppose that the animal was capable of experience but had no use for the experiential detection of anything more than 'warm / cold : in front / behind' and 'bubbly / clear : in front / behind' as it floated through its liquid medium.  The terms 'in front' and 'behind', which register the animal's experience, would gain their significance entirely by means of their connection with the two forms of locomotion available to the animal: move forwards and move backwards: they wouldn't function to identify places.  (Their significance would be exhausted by a realm of embodiment specification).  These connections would exist in virtue of connections between experiential 'assent' to the eight possible (two-part) feature-placing sentences and the locomotion of the animal.  It would be quite unwarranted to suppose in such a case that experiential responsiveness to 'warm' and 'bubbly' involved a sensitivity to conditions for re-identification.  (And inappropriate to attempt to apply the generality constraint).

 

            Because there is no warrant for re-identification or for generality, the simple experiential representation system of the marine animal (or the naming-child) is not S/Ojective (see footnote 17).  Hence the feature-placing sentences of the system do not have truth-conditions.  Instead we speak of "assent conditions" (or, with even less commitment, "experiential activity threshold conditions"); that is conditions in which a feature-placing ‘subject’ will 'assent' to a sentence of this type.  And, as theorists, we may characterize these assent conditions by means of other sentences (not themselves feature-placing sentences) which have truth conditions.  But this by no means entails that feature-placing sentences themselves have truth conditions. 

 

            The basic argument for a gap between truth conditions and assent conditions is straightforward.  Given the connection between S/Ojectivity and the reidentification of particulars, the content of feature-placing sentences is not S/Ojective content.  But thought content is essentially S/Ojective content, and a thought is introduced as that for which the question of truth arises: the truth-bearer.  Hence we only talk of the truth-conditions of a feature-placing sentence in a sense which is derivative on the conceptually sophisticated language of the theorist.  But we want the semantics of a language to characterize (no more than) what is understood by masters of the language in virtue of being masters of the language.  Hence truth has no role in the semantics of a feature-placing language.

 

            It is not a rebuttal of this argument to observe that there may be a fact of the matter as to whether it is raining at a certain location, and a paradigm feature-placing sentence is standardly used to assert precisely that.  For, if a sentence like “it is raining” is being assigned a semantics according to which its significance is that at a particular location it is raining, a significance which may be either true or false, then it is being assigned a subject-predicate semantics rather than a feature-placing semantics.  Under such an interpretation, there is an implicit subject-term which serves to introduce a particular in the ordinary way.  The only difference from a paradigm subject / predicate sentence, like ‘John is bald’, is that the subject term introduces a place rather than a material object.  This, for Strawson, is to ascend to a higher semantic level than is warranted by mere feature-placing.  For Strawson, if there is an adverbial demonstrative, like ‘here’, present in the sentence, then if the sentence is to be assigned a feature-placing semantics, the adverbial demonstrative does not introduce a particular location but rather serves to indicate the general area of incidence of the feature in question.[28]  Incidence indication does not require place identification: no identifiable and re-identifiable particular is introduced.  A placing of features is not a reference to a place of features.

 

            If there are truth conditions for a sentence of the language, there must also be falsity conditions.  In a classical semantics, these will be the conditions under which the sentence is not true.  But then—given the appropriate syntax— it must always be possible to form contradictions in the language by forming the conjunction of two sentences, the conditions for the truth of one of the sentences being the falsity conditions for the other sentence.  In the classical case, this will be achieved simply by the use of a negation operator.  So if ‘wet here’—given a feature-placing interpretation—has classical truth conditions, then ‘wet here and wet not here’ should be a contradiction.  But it is not.[29]  The conditions under which ‘wet here’ are assented to are quite compatible with the conditions under which ‘wet not here’ may be assented to.  The only way to make possible the formation of contradictions in the language is to provide a semantics that allows for the identification of particular places or—correlatively—the identification of particular instances: wetness is here and wetness is not here; the same location being picked out by each use of ‘here’.  But instance identification (and place identification) is unsupportable at the pure feature-placing level.

 

            What is lacking from the feature-placing language, which results in the inability to produce contradiction in the language, is the capacity for boundary conditions for the feature.  If the assent conditions of a sentence or its negation are sensitive to the general incidence of the feature, but are not sensitive to the presence or absence of the feature throughout a bounded region, then the semantics for the language should not appeal to objects and to predication.  That is, the ability of a language to form contradictions entails the sensitivity of assent conditions to the presence or absence of a feature throughout a bounded region: a necessary condition for a subject / predicate semantics.

 

            Evans (1985), in his article “Identity and Predication” brings out these connections by considering how the radical interpreter should treat sentences formed from concatenated terms of the alien language.[30]  What pattern of assent conditions would the interpreter have to find in order to treat the concatenation of F and G as predication?  Evans writes:

    

     When we look at the assent conditions, the following picture emerges.  We find that it is not sufficient for assent to (F G), eg., ‘White Rabbit’, that F and G both be assented to nor that there be an overlap between the features associated with F and with G.  For example, ‘White Rabbit?’ can be dissented from even though some of the exposed rabbit stuff is white, even, indeed, though a decent-sized (rabbit-sized) continuous portion is white, as when, for example, several brown rabbits are so organized that their white tails are contiguous.

 

     What is required for many of these compounds is that the F feature be distributed in a characteristic way in relation to the boundaries of a SINGLE object whose presence prompts assent to the queried G terms.  And when the simple overlap principle does seem to work for the affirmative sentence (as it does, for example, for the compound ‘Bloodstained Rabbit’) we find the assent condition of the internally negated sentences (‘not-bloodstained Rabbit’) again show a sensitivity to the boundaries of an object, for assent requires the ABSENCE of the associated feature from the entire exposed surface of that object.

 

            The G term is ruled out as a feature-placing term like 'rabbiteth', in favour of linguistic items which require a subject / predicate semantics, on the basis of the observed sensitivity of the assent conditions of sentences containing a concatenation of the G term and an F term to the presence or absence of the feature associated with the F term throughout a bounded region whose characteristics are determined by the G term.  It is this requirement of sensitivity to the boundary which provides for the possibility of contradiction. In a semantics of mere feature-overlap, where there is no dependence of assent conditions on the distribution of a feature throughout a bounded region, there is no possibility of contradiction.  Thus 'whiteth rabbiteth and not whiteth rabbiteth' is not a contradiction, since it can be assented to when there is experiential sensitivity to both an overlap between whiteth and rabbiteth, and also to some rabbiteth which is not whiteth.  But if the assent conditions to sentences containing the term interpreted above as 'rabbiteth' is associated with sensitivity to a bounded region of feature-space, then the bounded region "delimits that area in relation to which one or the other, but not both, of a pair of contradictory predicates may be chosen". 

 

            If it seems odd at first to suppose that the notion of a bounded region is not available at the pure feature-placing level, it may help to recall the conceptual inter-dependence between the characteristics of location and the characteristics of what may occupy locations.[31]  In particular, if the location occupants are features, then locations will have scatter in the same sense as features are scattered.  Such scattered locations are not identifiable or reidentifiable, and so 'reference' to them does not depend on sensitivity to bounded regions of feature-space.  It would be as if in throwing the contents of a bottle of glitter, our only way of identifying the location of fragments of glitter-in-flight consisted in our ability to locate the fragments themselves.  But, as things are, we cannot identify or reidentify glitter fragments-in-flight: the best we can do is to provide a general indication (a sweep of the hand) of the incidence of the scattered glitter feature. (Imagine that no instruments are used, and that the location of  glitter-fragments cannot be identified against a background of objects).  If we were to think of the scatter-locations on the model of particular places, then, as the glitter falls, we would have to say that the locations ceased to exist.  But this is unnecessarily paradoxical, for it results from employing a subject / predicate semantics where one is not warranted.  We are not talking, in such a case, of something that ceases to exist as it loses its feature, for there was no something in the first place: no distinguishable and recognizable place or region.

 

            For a feature-placing 'subject' all experience is as glitter-in-flight experience is for us (without the use of backdrops, etc.)  But if all experience is 'feature-glitter' then the experience does not present itself to a subject.  If there are no experiences with contents of objects and places, then there are no resources for distinguishing between the contribution of the world and the contribution of the subject, as the experiencing organism moves around in its environment.  If the environment is not given in experience as a world of objects, then the subject is not given in experience as one amongst those objects.  A failure of S/Ojectivity is a failure of subject as well as a failure of object.  Hence a feature-domain is a unitary phenomenon that contains within it (but not as constituents) what—as theorists employing a higher level of analysis—we would identify as the phenomena of subject, object, location and property.  But these phenomena-of-content are explanatorily distinct only at the conceptual level of analysis of content.  At the nonconceptual level of analysis of a feature-domain the distinctness of object, subject, location and property is an explanandum.  The explanatory apparatus (the primitives) of this level describes content-phenomena for which there are no distinctions between object, subject, location and property.  There is only the feature-domain itself, and trails through it.

 

            This suggests that we treat 'feature-placing experience' not as experience within a consciousness (which would require a subject / object distinction) but as environmental experience.    Here I intend 'environmental' in the sense of 'in the environs of mind', where "environs" is used as it might be in "you may find Robin in the environs of L.A.".  Searching successfully for Robin might depend on not attempting to employ a sharp distinction between the region within the city, and its surround, or between L.A. and Santa Barbara or Irvine.  Similarly, way-finding in environmental experience need not require applying a mind / world distinction: a distinction between what is in a subject and what belongs to the object, or between one subject and another.  In section 7, I interpret the abilities of the Realm of Embodiment as abilities to find one's way (the building and use of feature-boundaries) within environmental experience.  But first I want to introduce a general metric which can be applied to way-finding abilities.

 

(6) Environmental Ability Range

            Of the four anchors of the representational theory of cognitive trails, two of the anchors are now in place.  These are first, the Evansian strategy of the nonconceptual specification of representational contents by reference to the realm of embodiment.  And secondly, the Strawsonian strategy of a representational medium based on feature-placing.  We need now to gain a much better understanding of what is involved in the construction of the conditions of S/Ojectivity within a feature-placing medium.  We saw that the introduction of predication, and the possibility of contradiction, required sensitivity to feature-boundaries.  I develop this idea in the course of the next three sections by introducing the third anchor, which I call "the PD ratio of cognitive trails through environmental feature-space" and the fourth anchor, "stabilization".

            But let me begin with "PD ratio".  Forget, for a moment, about experience and thought.  I start this section with animals and their abilities to navigate around their environments.  How might we classify the range of animals’ ability to find their way?

           

            In this section I shall presuppose as little as possible about the theory of representational content, so we are to think of the range of abilities specified operationally.  For example, if a system was placed ab initio at a particular location A in the space, would it be able to find its way to an arbitrary location in the space, G?  Of course, things cannot be as simple as this.  We must stipulate that the system finds its way to the goal location in reasonable time without exhaustive search of the territory.  We must stipulate that the system finds itself at A without any knowledge of how it moved in relation to other points in the territory in order to arrive at A.  For example, we could specify that the system must emerge at A through a manhole cover, having arrived at the manhole by traversing a system of subterranean channels (where the structure of the channels carries no information about the structure of the terrain above).   Then there is the trickier matter of how the goal location is specified for the system (it must cease searching when it arrives at the goal).  Here it is hard to abstract altogether from representational notions.  Let us say that the system can recognize the perceptible appearance of the goal location when within sensory range and unobstructed, and that the goal is specified to the system in terms of this perceptible appearance.  But these are complications.  What is important is that we can now straightforwardly establish a spectrum for the abilities to get from A to G which range from the worst case to the best possible case.

 

            The worst case (ie the most minimal navigational ability) is that in which the system can only find its way to G when it can perceive G from A.  In such a case the system’s ability does not go beyond what is required for the system to know its goal (this is why it is the worst case).  Suppose that the system can perceive the goal when and only when it is n or less metres from it.  Then we may represent the zone of competence as follows:

 

Figure 1: Worst Case

The whole territory is represented by the dotted rectangular region, the goal by the ‘G’ ellipse, and the zone of competence by the region containing all points whose distance from G is less than or equal to n metres.

 

The zone of competence is, thus, the set of points within the territory which is such that the system can navigate successfully to G if, and only if, the start location, A, is at one of the points in the set. The worst case is that in which the zone of competence is the set of points from which G is perceptible.  The system’s ability is less restricted when it is capable of recognizing one or more landmarks from its start position, and the goal from each of the landmarks.  The zone of competence might then be represented as follows:

 

Figure 2: Route-Based Ability to Navigate

The system has the ability not only to recognize the goal on the basis of perception, but also a number of landmarks.  A homing landmark is either a landmark from which the goal is perceptible, or else it is a landmark from which a homing landmark is perceptible.  Then the zone of competence is extended by a circular set of points, radius n metres, centred on each homing landmark.

 

Thus a system has a route-based ability to navigate about its territory when its total zone of competence is the non-overlapping sum (union) of more than one local zone of competence; where each local zone of competence is the set of points from which either the goal or one of the homing landmarks is perceptible.  A homing landmark is either a landmark from which the goal is perceptible, or it is a landmark from which a homing landmark is perceptible.  Figure 2 represents three homing landmarks and one non-homing landmark.

 

            There are many ways in which the system’s ability can be extended beyond the case in which it has a simple route-based (piloting) ability to navigate.  Most obviously, it could master many possible routes to the goal by mastering many landmarks.  More interestingly, it could be equipped with certain search skills through which each local zone of competence could be extended.  For example, it might be that the system could search in the space m metres from each point in the periphery of a local zone of competence. The region of extension is that region in which the system does not get lost, ie which is such that the system can find its way back within reasonable time from any point in the extension to some point in the local zone.  In some cases having an extended local zone of competence will transform the landmark on which the zone is based from being a non-homing landmark to being a homing landmark.  And thereby substantially extend the global zone of competence.

 

            Since the total zone of competence is operationally defined, and we may suppose that the whole territory is fixed pragmatically by reference to the circumstances of the system, we may classify a range of abilities in a way that depends only minimally on representational notions.  I shall say that the degree of perspective-dependence of the system’s ability to locate an arbitrary goal within the territory is given as the ratio of the total zone of competence to the whole territory.  I call this "PD ratio".  The system’s ability is maximally perspective-dependent (start-location dependent and route dependent) in the worst case which I have depicted in figure 1; in cases like this the PD ratio is close to zero.  The system’s ability is maximally perspective-independent in the best case depicted in figure 3, in which the PD ratio is equal to 1.  And the system's ability has intermediate perspective-dependence, and an intermediate value for the PD ratio, in figure 2-type cases.

 

Figure 3: Best Case

The zone of competence has spread to fill the whole territory.  There are no ‘privileged’ positions within the space such that the system must start at one of the privileged positions in order to locate the goal, or must proceed to a privileged position in following a route to the goal.  The best case is that in which there are no privileged positions wherever the goal is located.

 

 

(7) Cognitive Trails

            Apply the metric introduced in §6 not to an objective spatial territory, but to an environmental feature-domain, as introduced at the end of §5, in which there are no given distinctions between object, subject, location and property.  Hence the abilities of the realm of embodiment (RE) by reference to which nonconceptual contents are specified (§3) are not  capacities of a human body / mind (or a biologically given organism with a bodily boundary), but are way-finding abilities through an environmental feature-domain.  Since the way finding and that through which a way is found are not—at this elementary theoretical stage prior to the subject / object distinction—given separately, we can take the RE specified abilities to be identical to trails through an environmental feature-domain: a structuring of a feature-domain into a feature-space.  The way-finding abilities are the environmental structures which make way-finding possible.  Nonconceptual content is (the experiential presentation of) cognitive trails.  Thus the strategies of §3 and §5 converge.

 

            The theory does not posit an independent subject who manifests a certain PD ratio for finding a way through an independent spatial environment. Instead imagine a PD ratio for the tracking of trails and trail-blazing through a feature-domain: not a subject’s tracking trails in the S/Ojective world, but the phenomena of tracking trails and trail-blazing themselves.  Trails are both person-made and world-made, and what makes persons and worlds.  Trails are in the environment, certainly, but they are also cognitive objects.  A trail isn’t just an indentation in a physical surface, but a marking of the environment; a signposting for coordinating sensation and movement, an experiential line of force.  Hence the marking is both experiential and environmental.[32]

 

            Let us suppose that the entire feature-domain is involved in the presentation of a single object.[33]  I don’t mean that a point in the feature-domain corresponds to a part of the object, such as a corner, for parts of objects are objects too. Rather, we can think of a point in a feature-domain as being an experiential moment.  (Speaking conceptually, such a moment might be a particular object standing in a multitude of particular relations, not all of them specified, as given in the experience of a particular subject, in a particular context.  But attempts to do conceptual justice to nonconceptual representation are never very satisfying).  So let us imagine very simple objects that have no parts.  I want to recommend that we think of the relation between features and objects as distributed: many points in the feature-domain correspond to a single (simple) object, and each point in the feature-domain corresponds to many such objects.  That is, objects are distributed over features.  We need now to understand how objects can be recovered from features, without in the process presupposing an autonomous and unexplained capacity for thought.

 

            Basic to such explanation is the tracking of trails through the feature-domain.  A trail-tracking through the domain may be a coordinated interaction between subject and object (as a conceptual theorist would describe it), but one in which subject and object need not be discriminated participants.  I reach out and grasp a mug, and thereby a trail is tracked.  The trail is the coordination.  Each trail occurs over time, and is a manipulation or a trial or an avoidance or a capture or simply a movement.  It is entirely context-dependent: the pressure successfully applied to a wall does not entail the ability to apply pressure to a wall in general (in any context), and it doesn’t entail the ability to act appropriately with respect to a wall in any circumstance and given any kind of goal.  Yet a trail is not transitory (although a tracking of a trail is): the environmental marking persists and thereby the ability to navigate through the feature-domain is enhanced.

 

            As multiple trails are marked, some trails intersect. Since I want to explain the structuring of a feature-domain into a space of feature-points in terms of trails (rather than vice-versa), I don’t want to explain the intersection of trails in terms of their crossing at a point. Rather, a point in feature-space is explained initially as an intersection of two or more trails.  An intersection amongst trails is a moment of enhanced possibility for shifting from one coordination to a distinct coordination: at a certain stage in the manual application of pressure to a stone there is an enhanced possibility to grasp the stone, and at a different stage of the stone-grasping is an enhanced possibility to pick up the stone.  Coordinations intersect.

 

            A feature-domain landmark (‘f-landmark’) is an intersection of two or more trails.  A feature-domain is structured into a feature-space by means of a network of f-landmarks.  f-Landmarks are the first points in feature-space; other points are possibilities for f-landmarks.  Points not themselves on a trail come under the influence of nearby trails.[34]  There may be a set of points which are under the influence of the same set of trails; this is a region of the feature-domain.  If the influence of the nearby trails is strong enough and pervasive enough, then the organism would be able to find its way (to an arbitrary goal) within the feature-space whichever point within the region it happened to start from.  Such a region therefore has a high PD ratio and is said to be bounded by the influencing trails (equivalently: the trails dominate the space).  Thus boundaries within feature-space are capacities to navigate around the feature-space; capacities which consist in an appropriate network of intersecting trails.

 

            Given the connection between f-boundaries and predication argued for in §5, it would be natural now to explain S/Ojectivity in terms of a feature-space with PD ratio equal or close to 1; or, equivalently, as a region of feature-space which is bounded by a network of trails. We could arrange a spectrum of contents with differing degrees of S/Ojectivity varying according to their position along the dimension of PD ratio through environmental feature-space.  Contents with a higher PD ratio would be more S/Ojective than contents with a lower PD ratio.  A concept would be a maximally S/Ojective content; ie. a content with PD ratio equal to 1.  Moreover it would be natural to think of the construction of objects (in the world and in society) as the laying down of an intersecting network of trails which dominate a region of feature-space.  According to this suggestion, the feature space of an object is that (sufficiently large) region of points in a feature-domain for which there is a network of trails which dominates all the points in the region[35].  In the next section we will see that however tempting this idea may be, it is not correct.

 

            But this much is clear now: low PD ratio of a region of feature-space entails that any attempt to capture its content referentially in sentences of a language would fail the generality constraint.[36]  Reconsider our organism finding its way around a spatial territory, having a more or less perspective-dependent ability to find its way from A to G.  The organism is capable of experience, and is finding its way around on the basis of its experience of the territory.  The experience has a feature-placing semantics and we are to ask about the degree of generality of a content ent