lunes, 20 de agosto de 2012

The Destruction and Reconstitution of Experience: On Sellars' Account of Ur-Conceptuality and Sensibility

On Sellars, Sensibility and Correlationism


    I have been re-considering Sellars' account of "ur-conceptuality" from the Lever of Archimedes in the Carus Lectures, after evaluating my contention, presented during the Bonn summer school about a month ago, that these are not to be understood in terms of 'pre-linguistic' cognitive capacities, but are rather more like 'rudimentary concepts', already linguistically mediated, if not fully functional in the logical space of reasons.  In what follows I intend to present this issue which, I believe, ties up some essential knots on the questions about the relationship between Sellars' nominalism and his realist metaphysics.

1. Ur-Concepts and Primitive Representation
           In a discussion with Roderick Firth, Sellars sets out to separate the primitive conceptual capacity of pre-adult humans to discern secondary qualities as component-parts of physical objects. He thereby seeks to pin down precisely how it is that appearances already enjoy a primitive ontological status, the better to eventually offer successor concepts in order to supersede not only the base conceptual attributions that children make, but also our available 'adult' concepts for secondary qualities.  The ur-concepts that Sellars attributes to the infant Jones Junior are as follows[1]:

1.       Junior has an ur-concept of volumes and expanses of red stuff.
2.       Junior has an ur-concept of seeing a volume of red stuff.
3.       Junior has an' ur-concept of a physical object as an individuated volume of color stuff which is endowed with certain causal properties.
4.       Junior has an ur-concept of seeing a volume or expanse of red stuff not only as a volume or expanse of red, but as a constituent of a physical object.
5.       Junior has an ur-concept of what it is to see of a physical object a volume or expanse of red which is one of its constituents. If the constituent is the surface of an opaque object, e.g., an apple, it is the very redness of the apple.
6.       Junior has an ur-concept of what it is to see the very redness of an object.

      The problem at hand is what precise status these ur-concepts have; are they linguistic capacities or representations? Or are they another species, perhaps psychological, of intentional representation? What is behind Junior's capacity to do all of these things?

   At Bonn, Ray Brassier flirted with the idea that these might be understood best as pre-linguistic capacities, while I contended the opposite. Essentially, I think my original contention was fundamentally correct, but I think I can see why Sellars' own formulation of the problem might nevertheless tempt us to identify ur-concepts with pre-linguistic capacities, especially judging from what he says in other writings. Moreover, in his influential study, James O'Shea identifies ur-concepts with pre-linguistic capacities, which would seem to settle the case once and for all. I think the issue is more complicated, and ultimately important.

   The two crucial texts that I think help us figure out what's going on, however tentatively, are Some Reflections on Language Games and the Mental Events paper. These will also shed a lot of light into why many consider Sellars' venture into process metaphysics to constitute a relapse into a form of the Myth of the Given, or finally a form of adherence into naturalist prejudices that strictly speaking are 'pre-Sellarsian'. The separation between left and right wing forms of Sellarsianism can be mapped to some of the neighboring issues as well. But anyhow, I think these texts allow us to see why the thesis that ur-concepts are pre-linguistic is infelicitous.

   In Some Reflections on Language Games, Sellars considers the original empiricist or naive realist appeal to pre-linguistic ur-concepts as facilitating the capacity to associate undefined descriptive predicates ('red) with items in the world. The basic idea that these thinkers shared is that the meaning of our observation statements derives from our primitive cognitive capacity to apprehend the categorical structure of the world, and specifically the content concerning perceptible qualities. The subsequent idea was that the predicates involved in observation statements (as paradigmatic ways in which causal stimuli trigger linguistic responses or "language-entry transitions") acquire meaning  as we learn to obey "semantical rules" which in involve recognizing colored things. Say, our concept of 'red things' acquires meaning as we learn to explicitly obey the rule red objects are to be called 'red', and so on. This is not to say, of course, that we need to have a concept of the rule in order to obey it; but that our behavior exhibits conformity to the rule in a sense in which we count as having internalized it, as exhibited by regularities of behavior, and specifically in learning to recognize objects of the right sort. But what grounds the surreptitious pre-linguistic capacity for recognition that subjects make explicit as they learn to deploy observation talk? Sellars critically considers two such candidate accounts for these ur-concepts that allow pre-linguistic categorical apprehension:

1) A symbolic structure or language -  Under this hypothesis, there is a fundamental language or symbolic system that is not itself acquired by obeying rules, but that rather conditions that possibility of identifying different terms in different languages as being about the same thing, beyond their functional role. For if learning to use a concept requires identifying by way of its use an appropriate kind of object/properties or class of objects/properties, then we need to explain how we come to identify such objects or properties of being of that kind.

         If in order to learn the meaning of 'rot' we obey the rule 'red objects are to be called rot' then, in pains of regress, we need to say that the symbolic language by virtue of which we identify 'red objects' is unique, and not itself learned by virtue of obeying a further rule, for this would obviously unchain an infinite regress. This must hold necessarily since obviously the ur-concept of 'red object' would have to be different than the concept we learn by virtue of obeying the semantic rule in which the ur-concept is formulated, in order not to fall to the inconsistent hypothesis that we acquire concepts on condition that we already have them. Thus, some basic language is primitive in the sense that it is not acquired by obeying rules and in that it conditions all further learning by relating the functional role of a term in another language to the objects/properties that we identify primitively in the base language.

      It is precisely on the basis of such an ur-conceptual economy we can then learn the functional role that specific linguistic tokens bare to empirical reality, i.e. the contents of descriptive predicates (redness, red items...etc). I think this preemptively anticipates the nativist hypothesis championed by Fodor and Lepore, among others, according to which innate conceptual structures precede and condition that capacity for learning, and thus full-blown linguistic rationality.

2) The capacity to apprehend items as belonging to kinds or resembling classes of particulars  - Sellars castigates this position as a variant of the 'mental eye' view which ascribes to the mind the capacity to immediately apprehend the categorical structure of the physical world, sense qualities, or whatever else, by fixating itself inwardly and outwardly at once. But it seems clear that were Sellars to endorse this, he would be delivered right back into both the matrimonial account of meaning and the epistemological Myth of Given. For it would require postulating both that we have immediate awareness of abstract entities which furnish the categorical structure of reality ('redness'), and that it is by virtue of such awareness that words acquire meaning. But this is to reactivate knowledge by acquaintance.

       At this juncture it becomes clear that the the first, proto-Fodorian hypothesis must also postulate a pre-rational cognitive symbolic economy that simply staples mind into world. But this is another version of the Myth of the Given since, it seems that whatever this system is, it works to grant a sort of luminosity into at least the undefined color qualities of objects.

    It is important to note that in SRLG, Sellars rejects both accounts of pre-linguistic ur-concepts as facilitating the rule obeying usage of observational language. The solution seems to be clear: we have to account for language acquisition not as the explicitation of a semantic protocol for recognizing 'red things' via some dubious 'semantic intuition'. We don't learn to say red because we learn to obey the rule that we should utter 'red' when we see a red thing. For then the ur-conceptual protocol as a foundational discourse is introduced simply to halt the inevitable regress that follows from the thesis that we need a meta-language to learn a language. The obvious alternative thesis is to say that an observation-language is, if not learned as rule-obeying,  a conditioned responsive behavior. We can relax our account to claim that although one does not obey rules to apply observational discourse, one must nevertheless be conditioned to respond to the right sorts of entities in the world. Thus, to say that "...the fact that the word 'red' means the quality red may be identified with the fact that 'red' is a conditioned response to red things" (Sc. 38). But while Sellars accepts that indeed the proper application of the word involves being capable of reliably responding to the occurrent stimuli, he warns against the dubious claim that this entails that words acquire meaning by being associated with things. Reiterating the distinction between acting in accordance to rules and acting in accordance with conceptions of rules, Sellars' suggestion is that in acting in accordance to rules we are simply conditioned to respond to red things in the right circumstances, rather than assigning a term to an object we have already recognized in advance by some prior intentional mechanism. This is obviously a corollary of Sellars idea that there is no form of intentionality that precedes the linguistic; neither psychological nor practical.

       What is going on, then, when Sellars speaks of ur-concepts in Junior's talk?

     Now, it seems tempting to claim that when formulating his account of ur-conceptuality to open the space for the analysis of sensibilia or sensa, Sellars reverts into something like the second variant of the empiricist/ naive-realist account outlined above, in virtue of which we have some kind of immediate awareness of the categorical contents of reality. This might be reinforced by considering that Sellars also wants to account for something like pre-linguistic, "animal representations" or proto-cognitions roughly around the same time as the Carus Lectures, in Mental Events (1981). And yet it seems just implausible that Sellars would have suddenly relapsed into a mind's-eye view, of all people! How are we, then, to account for the ur-conceptual status of Junior's experiencing of sensible qualities as constituents of physical objects?

   I think Sellars' remarks in S. 38 of SRLG point towards the right solution. Although it is true that our we cannot have a concept of red without being capable of reliably responding to instances of red, it does not follow that we must equate the meaning of the word with a relation to an item in the world. We can simultaneously accept that our capacity to use undefined predicate-quality words as conditioned by capacity to differentially respond to environmental stimuli in a way that precedes rational rule-obeying, while maintaining that the meaning of the word is nothing but its role in a conceptual economy. Sensibility itself plays no epistemic role, even if it is necessary to acquire knowledge. Sellars must preserve his nominalist account: meaning is functionally specified within a conceptual economy.

     But this means that Junior's ur-concept of red is not 'pre-linguistic' in the sense that it involves the operations of a nativist ur-language (hypothesis 1), nor the fixation of the mind's eye upon the categorical structure of the given via intellectual intuition or some other pre-linguistic intentional mechanism (hypothesis 2). Junior clearly is conditioned to respond with the verbal output 'red' when he sees red things, and can do so somewhat reliably. He uses language already in a self-conscious manner, even if lacking full fledged capacities to enter into deliberative reason-giving behavior.

          Furthermore, his ur-concept is not one of something looking red, but of something being red, since he lacks the contrastive concepts of 'looks/is' introduced upon further conditioning. Junior attributes the redness he sees to the objects he interacts with rather than to private experiences, fields of sense, or whatnot. This is not to say that Junior has a fully-developed concept of 'physical object, extended in space' at that stage, or something so sophisticated. It's clear Junior is not yet capable to playing the game of giving and asking for reasons, or of using the contrastive concepts of 'looks/is' to enact withdrawal of endorsement, or to characterize episodes as merely ontensible seeings. He reports redness as pertaining to objects, and is flummoxed when some evil grown-up tampers with lighting conditions so that what he takes to be a red object suddenly appears blue.  He must then simply retort to thinking that a blue object is before him, and needs eventual introduction into the contrastive concepts to distinguish between mere seemings/lookings and actual features of objects, by modifying his entitlement towards the content of the claim. Thus, Sellars' 'ur-concepts' are utterly linguistic capacities; they cannot be understood as primitive capacities at the level of mere sentient registration and causal responsiveness. Junior's repertoire is surely already more sophisticated than that.

     So what does then, this ur-concept of redness, encompass? Sellars list is amply clarifying.
  It clearly already involves the capacity to see a red thing as red, or that it is red, since Junior attributes redness to that which he sees. Although Sellars characterizes this as-content as being 'of a physical object' this should be taken in tandem with his qualification in section 32, where he makes it clear that Junior has the notion of an object as a 'determinate thing-stuff'; but not the full-grown concept of physical we attribute to adults and which evidently Junior could not have. This is of a piece with his reductive semantics of sensa that he provisionally already proposed in Some Reflections on Perceptual Consciousness (1975).

   Junior also has the capacity to see of the thing the redness which is a extended surface of the object, i.e. Junior is not shocked if when splitting an apple in half the inside of the apple is not red too. Junior sees of the object the facing side, which is non-conceptually given to him through his senses. But having an ur-concept of what it is to see of an object, Junior thus both is capable of attributing redness to the object, and at the same time realizing that this redness pertains to something like the extended surfaces of opaque objects, rather than being a thorough characteristic of objects through and through. This is crucial, since it is the first step towards untethering expanses of sense-qualities from objects; a move that Sellars subsequently exploits for metaphysical speculation. This is what is fundamentally added by conditions (5) and (6).

      Junior's ur-concept of red is thus conditionally acquired and not rule-obeying, linguistic insofar as it involves predicative attributions and the capacity to see things as such-and-such, yet too rudimentary to count as an 'adult concept' since it lacks the contrastive concepts that allows one to formulate one's endorsement or lack thereof, on the basis of warrant tethered to a notion of standard conditions, and so on.  And precisely because Junior's ur-concept of red is simply a 'rudimentary' rather than 'foundational' concept that Sellars can resist the myth of the categorial given: it's not that Junior, before learning language, already accesses the categorical structure of the physical world. Junior is conditioned to respond in ways through which he becomes capable of characterizing the contents of his experience, but this already involves linguistic mediation/acculturation.  Junior has no determinate category of colored physical objects extended in space, he has an determinate concept of something-being-colored as a colored-expanse more generally. This is the only way to reconcile the psychological nominalism with the holistic account of linguistic rationality, I think.

       On this account, it becomes a lot clearer just where the Carus Lectures depart from the account of the Myth of Jones, and the looks-talk analysis. Sellars' eventual supplement to the Myth of Jones is to say that before the contrastive concept of looks, which can be applied to color-qualities as easily as dispositional ones, we have a proto-concept of color-qualities or 'the proper sensibles' characterizing them as constituent parts of physical objects, if not identical with them. There is a positivity of appearance that precedes the epistemic regulation of looks talk, which does require fully grown linguistic rationality or integration in the game of giving and asking for reasons. Understanding this also allows us to understand why looks-talk, when applied to color-concepts or sensible-predicates operates withdrawal of endorsement in the form of ostensible seeings, i.e. to merely see red is not just to refrain from claiming that something is red, but to report on the autonomy of the colored expanse from the properties of the object.  The occurrent properties of sense can thereby be examined positively in a metaphysical account.

2. Against Correlationism 
         On a different but related note, this allows us to see how Sellars is precisely not a correlationist. Correlationism requires three conditions:

1) the sense dependence of objects on concepts,
2) the reference dependence of concepts on sensibility, and
3) the ontological identification of sensibility with subjective appearances. Sellars' trick is to accept the first claim, qualify the second, and reject the third.

       Yes, knowledge of anything requires concepts. This is true, but somewhat trivial. Even people like Tyler Burge think knowledge requires justification of the sort only sapient creatures enjoy. Yes, the empiricist is right in insisting that our concepts and so our knowledge of the world begins in sensible experience. But, against the sense-datum theorist or phenomenalist, this is not to say that there is a reference dependence of concepts on sensibility, but only another form of epistemic dependence. That our knowledge is anchored in the world through sensibility is not to say that all knowledge is of the sensible. This is not to refuse that there is such a thing as sensibility, or to say it cannot be metaphysically investigated. Rather, it is simply to say that sensibility can condition our knowledge of that which is not sensible. We can  investigate the non-apparent structure of appearances as belonging to the domain of physical nature, just like we can investigate the micro-physical constitution of the manifestly described world of middle-sized objects and properties apart from their phenomenological conceptual envelopment. Thus, against (3), it is simply false to equate the content of sensibility with 'subjective appearances', where the latter are understood as self-presenting episodes.

         The properties of sense-qualities, although relative to sentient organisms, are irreducible to the phenomenological categories that furnish our commonsense description of the world. Granted, phenomenology reveals that there is a dimension of appearing that must be accounted for. But this is not to say we are forced into construing appearances as subjective correlates. They can be accounted  at the genetic neurocomputational level in terms of how environmental inputs relate to our cognitive faculties by triggering appropriate neuronal onsets.

          Now, Sellars thinks on top of this you can actually characterize the apparent particularity of sensa as ultimately non-object bound physical phenomena, which have irreducible qualitative properties. Sensa turn out to be features that are objectively constituted only in relation to sentient organisms, but they are no less objective for that. In other words, that sensa are dependent on sentience does not render them subjective in any interesting sense. Again, no basic categorical stratum of the sort phenomenologists fetishize over can hope to undergird the objectivity of the physical within which even appearances are constituted.

     Whether that much is tenable remains an open metaphysical question. But the point is that once we have  defused the ontological equation between sensibility and subjective appearances, both forms of epistemic dependence won't do to motivate correlationism. The link between knowledge and conceptuality, and that between conceptuality and sensibility requires that we investigate precisely how observation statements, qua language entry transitions, are articulated with complex causal mechanisms that relate environment to organism. Even if it turns out that the presentational content of experience is fundamentally tethered to sense-data, this is not to assert that sensibility has an autonomous categorical status apart from the physical, nor that the content of our judgments is reducible to sense data just because we require sense data to make them: we can examine the link between appearance and reality by examining the reality of appearing, refusing the phenomenological hypostasis of the categorical status of subjectivity as foundational, by reintegrating sensibility to physical theory, acknowledging it as our anchor to the world, rather than our solipsistic prison.