Reply to Commentaries on "Empiricism vs. Pragmatism"

Back to CQ Homepage

Back to Archives' Table of Contents Reply to Commentaries on "Empiricism vs. Pragmatism"

By Teed Rockwell

      Dear CQ participants,

     Thanks for your many thoughtful comments, and special thanks to those who re-sent your messages after my data disaster. (note: the data disaster acounts for the fact that this reply contains references to many commentaries that are not posted.) I've noticed one thread that runs through many of your reactions to my first post, which I am going to deal with first because it reflects on the validity of much of what I'll be doing.

      Bob Kane remarks:

      Few if any philosophers, psychologists, or cognitive scientists think any longer that perception is purely passive reception of data. It is active integration and interpretation by way of neural networks. So Dewey and James are beating a dead horse when they attack sensationalistic empiricism *on grounds of* its passivity.

      James Garson also used the same necro-equine metaphor in an earlier correspondence about my plans to critique sense datum theory from a neurological perspective, as did Bernard Baars.

      Ullin Place remarks:

      I gave up that idea { of sense data as a foundation for knowledge} fifty years ago when I heard John Austin give his `Sense and sensibilia' Lectures in Magdalen College, Oxford, in 1947.

      So if everybody assumes that Sensationalistic Empiricism is a dead horse, why am I bothering to bring this topic up?

      1) BECAUSE DEAD HORSES DON'T ALWAYS LIE DOWN.

      There is a great difference between consciously acknowledging that a theory is wrong, and actually rooting it out of your presuppositions. In fact, there are still domains of discouse that presuppose something like what Dewey calls sensationalistic empiricism. (Which, as far as I can see, is indistinguishable from sense datum theory. Because this later term is better know than Dewey's, and somewhat less clumsy, I shall use it for convenience's sake.) Two subjects in particular come to mind:

      A)Information theory: Sense Datum theory is still lurking in any theory of knowledge which assumes that knowledge is achieved by a computer-like device that collects independent bits of information, and then manipulates them with the mind (or CPU)in accordance with of the laws of logic. There are some significant disanalogies, or course. The "bits" that are stored at the addresses in a computer are qualitatively simpler than the raw feels of sense data. But this simplification is, I think, usually glossed over by assuming that the computer is only a stylization, which will become more sophisticated as technology develops. Consequently, it would be difficult to abandon sense datum theory and still believe, (as for example Fodor appears to) that the digital computer captures all that is essential to cognitive processing. In fact. I did receive a post from a Computer person, (lost during the great data disaster), which defended sense datum theory because it was essential to computer science.

      Computer science will remain an important scientific discipline even if it is significantly misleading about the principles that govern neurological systems, because computers have become so important to us in their own right. But any epistemology that rejects The Myth Of The Given will have to precisely articulate the ways in which computers are not like conscious biological systems before it can get the dead horse of sense datum theory to lie down in peace. (Another piece of useful machinery that appears to give support to sense datum theory is the movie camera. The fact that it can create something remarkably like our own manifest image of shaped and colored experience by stringing together a series of discrete frames makes it very tempting to assume that our brains are doing the same thing.)

      B) Qualia: there are a lot of respected philosophers who believe in Qualia (David Chalmers, Frank Jackson, John Searle, Ned Block, to name a few) and as far as I can tell the common view of Qualia essentially defines them as sense data when seen from the first person perspective. The qualia critiqued in Dennett's classic article "Quining Qualia" share most of the essential characteristics of sense data, and Dennett claims that once these characteristics are dismissed, there is nothing left in the concept of qualia that is worth keeping. I think that is an overstatement; chapter 4 of Owen Flanagan's "Consciousness Reconsidered" points the way to a concept of qualia with none of these characteristics, and it may be that other philosophers have similar theories. (Any comments from Chalmers on this?). Also of interest are these selections from Jim Garson's last CQ post.

      Are there sense data? This reminds me of the question: are there qualia? And of course the right answer is yes and no.

      What one wants to object to is the dogmas of empiricism that assign to the so-called "given" certain fanciful properties: privacy, direct access, incorrigibility, being pre-theoretical etc. These remind me of qualities of qualia that Dennett objects to. It is not that qualia aren't THERE. It is just that our presuppositions about their qualities are all wrong. (OK, if the wrongly assigned properties are supposed to be constituitive of qualia then they aren't there, and something else is the thing we need described and explained.) The same moral can be applied to the "given". Once this is clear, there is room for an epistemological theory that takes full account of the Quine Sellar critique, but still insists that truth is not a matter entirely of internal coherence, or perhaps to put it another way it is a theory that priveledges certain "experientially" derived beliefs, without making them incorrigible, private, non-theoretical etc.

      The question is: What is it that priveledges certain "experientially" derived beliefs? Despite all its many problems, Sense Datum theory has an answer to that question, and until we can come up with a better one, we cannot stop ourselves from thinking in terms of the old theory. Ullin Place, after stating that he himself had abandoned sense datum theory years ago, makes this comment.

      As soon as I read the late Donald Broadbent's 1971 book DECISION AND STRESS, it seemed to me obvious that his concept of a "state of evidence" on the basis of which the brain categorizes sensory inputs corresponds rather precisely to the notion of raw uninterpreted sensory experience. This notion I take to be implicit both in James' description of the consciousness of the child as a "big blooming buzzing confusion" and in Wilhelm Wundt's distinction between Immediate and Mediate Experience, where Mediate Experience is experience interpreted as a sensory encounter with external reality and Immediate Experience is the same experience interpreted as what it really is, a process taking place within the observer's own consciousness. In ordinary language it is implicit in the distinction we draw between `physical' pleasure and pain which does not depend on how the stimulus is interpreted and `mental' pleasure and pain which DOES so depend.

      I think that there is an important difference here between James' concept and the other two, because James believed only that uninterpreted experience *as a whole* was a blooming buzzing confusion. As Elizabeth Minnich pointed out in her CQ post, There were no uninterpreted bits of experience for James; for him the act of interpretation was what broke the whole into bits.(Marcel Kinsbourne proposed a similar theory at his presidential address last year at the Society for Philosophy and Psychology meeting. There was no reference to James in it, however, apparently it just seemed to him to be the best theory to explain the data. Did anyone get a copy of that paper?)

      From Place's description, however, it appears that both Broadbent and Wundt did accept something like Sense Datum theory, and that this theory shaped and conditioned how they interpreted their experiments. If philosophers decide a theory is dead, that won't necessarily stop scientists from accepting its assumptions. And this can happen even when the data is crying out to be interpreted with an alternative theory, unless the scientists are themselves philosophically sophisticated, or remain in dialogue with philosophers.

      2)IT IS NO GOOD POINTING OUT THAT THE HORSE IS DEAD WHEN YOU DON'T HAVE ANYTHING ELSE TO HITCH YOUR WAGON TO.

      After making the above quoted point about Neural Networks, Bob Kane then asks:

      If we add the active element into immediate sensory experience does this come closer to {Dewey and James'} idea of experience, or is it some third option they do not consider?

      I don't think you can add an active element to immediate sensory experience, because once the experience is mediated by being acted upon, it is by definition no longer immediate. I do think, however, that the distinction I quoted from Dewey in the last CQ post was an attempt to offer an alternative option, which we could describe at least partially as grounding "knowing-that" knowledge in "knowing-how" knowledge. Sue Pocket apparently found this idea too counterintuitive to take seriously; she thought that the fact that the word "experience" applies to both sense data and skilful abilities was simply a pun. But I think that because Cognitive Science has made so much progress since Dewey's time, we might now actually be able to explain in biological detail how the two forms of knowing are related. Jed Harris points out that Andy Clark's book "Being there:etc." has already suggested some answers, so perhaps Andy will favor us with a few comments the next time around. I also received some other offerings of Cogsci facts that could be parts of the puzzle.

      From Jim Garson:

      First we need to get clear about what the {Knowing-how vs. knowing-that} distinction amounts to and why it matters. I am reminded of the fruitless procedural-declarative controversy in AI. The moral was that finding the dividing line is not easy, and whether you find it or not is not a prerequisite for doing good AI. One interesting feature of connectionist models is that the distinction becomes less tenable. To obtain the distinction you need to separate the data from the doings of the program. But it is exactly this distinction that tends to be blurred in connectionist models. So perhaps the answer is that the difference between knowing how and that is a matter of degree, and not something you could specify in neural terms. If so the theory of the relation is something like the theory of the relation between waves and particles at the sub-atomic level. Waves and particles are aspects of an underlying reality that is of neither kind.

      Maurice Willey cited Stephen Pinker for evidence that sense data cannot stand on their own because "Inherited structures serve as the starting point for acquiring experience." The relationship between the a priori and the acquired is, I think, closely related to the relationship between Knowing-how and knowing-that. In fact, they are too closely related for me to be sure how to separate them, so I will deal with that at a later time. (Gary Schouberg also takes me to task for confusing these two distinctions, and Rick Norwood makes some very interesting observations that straddle both questions when he discusses the relationship between mathematical and empirical knowledge .)

      Willey then eloquently demonstrates that something like sense datum theory is still alive and well in the minds of some practicing scientists.

      All data used in the building of experience flows through the senses, there is no other creditable source, despite the many claims to the contrary. Sensory data is built into percepts which are what we are consciously aware of. Thereafter, we build our epistemic knowledge from the perceived 'connectedness' of percepts. Regularities with respect to other data points get noted and start to cause concept building.

      One of the problems with this description (which could have come right out of Locke) is that if we have to "build" experiences from discrete unconnected bits, there won't be any intrinsic connections between those bits for us to perceive. Kant thought the only way to solve this problem was to assume that the mind superimposed order onto those bits. James solved this problem by claiming that experience was fundamentally unified, and that we divided it up by thinking of it in terms of how it meets our goals and purposes.

      From Sergio Chaigneau:

      From the point of view of psychology, it is difficult to support the notion of pure sensations. Interactionists, such as Gibson (1966), showed that perception (mostly visual) needs motor input in order to be constructed. The classical experiment involves a subject that tries to adjust a luminous rod to a vertical position. This is done in a dark environment (no other visual clues to judge verticality) and with a remote control. When subjects are given a small electric discharge on the neck area, they tend to fail the vertical adjustment by some degrees. These results can be interpreted as if subjects integrated information about activity on the retina and head position, in order to act. A small electric discharge is interpreted as muscle contraction, which usually means having your neck bent. Now, if your head is slightly bent sideways (like this ` / `), anything aligned with your head`s main axis cannot be on the vertical plane. Therefore, it has to be corrected. Since subjects in this experiment were fooled into "thinking" that their heads were slightly bent, they erroneously corrected the rod`s position.

      These results are consistent with the Lateral Geniculate Nucleus` role (LGN) in the visual system. The LGN is not just a relay station for visual input, but it also integrates information form motor and other areas. About 80% of incoming information to the LGN is not from the retina, but from cortical areas; mainly motor areas (Varela, Thompson and Rosch, 1991).

      Gibson thought the Nervous System extracts invariant characteristics from a continuously changing environment. These invariants are the objects of our perception (maybe, we can even consider ideas as invariants). The French philosopher Ulmo (1969, specially chapter 9) has extended the notion of invariant (apparently independently from Gibson) to cover not only individual knowledge, but also scientific knowledge. According to Largeult any area of knowledge can be understood as a set of operations or actions you perform. Since he frames his discussion within logical rules, he argues there are logical systems (a coherent set of operations) which allow the user to restore a previous state (a reversible operation would be the most simple example, as in Piaget). In any scientific field, what you learn is to perform a set of operations that lead to certain observations. These observations are the invariants (i.e., theoretical or concrete objects). Once these objects are established, the field can be defined by its system of operations or by its objects. { This sounds a lot like Pierce, especially his essay "What is Pragmatism", which was recommended to me by Markate Daly -WTR}

      I think that from this point of view, knowledge is not denied in principle. It can be a bit shaky if invariants seem to depend completely on the set of operations you perform (e.g., experimental artifacts). But if different invariants, which are the result of different sets of operations, can be interpreted as the same object (converging evidence), then the object gains credibility as a real object. {Millikan makes a similar point in her paper mentioned below-WTR} (snip)

      I really think "knowing that" is always based on operations (e.g., sensory-motor coordinations). A good example is depth perception. When you perceive depth, especially in close range (1 or 2 yards maybe), you just know that something is within your arm`s range (you can also make precise judgements about relative distances of objects within that range). What you don`t know is how you obtain that knowledge; it seems direct, pure sensation, but we know it rests on complex sensory-motor coordinations created through our developmental history. This is a case where you actually can`t tell how, you just "know that". {I think you actually have knowing-how and knowing-that reversed here, but the point is a good one. For some fascinating answers to this mystery, see Melvyn Goodale's "the Visual Brain in Action"-WTR}

      References

1. Gibson, J.J. (1966) "The senses considered as perceptual systems". Boston: Houghton Mifflin Co.

2. Ulmo, J. (1969) "La pensee scientifique moderne". Paris: Flammarion.

3. Varela, F.J., Thompson, E. & Rosch, E. (1991) "The embodied mind: Cognitive science and the human experience". Cambridge, MA: MIT Press.

      I was delighted to see Sergio bring up J.J. Gibson (who was on my reading list for another topic to be discussed later on in CQ), and even more delighted to see that Ruth Millikan's contribution also reinterpreted Gibson with her usual balance of care and audacity. Ruth sent me a long (and I believe as yet unpublished) paper on the knowing-how vs. knowing-that distinction which was truly awe-inspiring.(Some Different ways to Think) In it she proposes a theory which makes knowing-how knowledge the foundation of knowing-that knowledge with a detailed biological plausibility that would probably even convince Sue Pockett. I'm not going to say anymore about it now, because I will discuss it at length in the next post.