Date: September 16, 1998 01:06 PM
Author: John Schmidt (schmidt@wsuhub.uc.twsu.edu)
Subject: A Neurobiologist's View of Eliminativism

(Another view of eliminativism.)

I came to Phil. 590I with some huge holes in my understanding of the philosophy of mind. I had assumed a rather stark division of philosophical stances concerning mind/brain: either people were reductionistic materialists or they were dualists. The idea of eliminativism (1) had slipped past me.

I had read books such as Consciousness Explained (2) and The Engine of Reason (3) and made the assumption that Dennett and Paul Churchland were reductionists, playing out the game of trying to understand how the physical processes inside brains can account for our subjective experiences and the complexities of human behavior. To start reading Rey (1) and find him calling Churchland and Dennett eliminativists was a real shock.

Can I be forgiven for missing eliminativism? Dennett mentions it once in Appendix A of Consciousness Explained (2) where he refuses to wear the label. In his 1995 book (3), Churchland states his goal to be an attempt to deal with the problem of understanding how a brain sustains a thinking self. I read that (and the whole book) as a reductionistic program. I was stunned to read Rey's claim (1) that Churchland is the Poster Boy for total eliminativism ("denying the existence of mental phenomena entirely", page 69). As of now (having read the first 4 chapters of Rey's book) I suspect that what we have here is two groups of people using the same words in two different ways, resulting in confusion and poor communication. Frankly, such confusion is often an indication that people are constructing arguments based on wishful thinking and emotions rather than facts and solid arguments.

I like "intuition pumps" and analogies for helping us to get through confused situations. The arguments between mind/brain monists and dualists are in many ways similar to the clash between Democrats and Republicans. Republicans try to pin derisive labels on Democrats such as "commies", "pinkos", "socialists", and "liberals". After a while, some Democrats may accept the spin provided by Republican propaganda, and they may begin to wonder if they are part of some extreme political party that does not really fit with their personal beliefs. Of course, the same goes for Democrats painting Republicans as Right Wing Extremists. Thus we can have Republicans and Democrats abandoning their political parties in order to escape the labels that their political enemies have used. And there are certainly some Republicans and Democrats who are Right Wing or Left Wing extremists, so there can also be motivation to leave an old political party and found a new one that does not make room for one or another extreme wing of the old party.

All the same dynamics seem to apply to mind/brain monists and dualists. We have had several centuries of mud-slinging and an accumulation of every imaginable extreme view under these two labels: monist and dualist. Rey's book (1) seems to center itself on a "new" debate between eliminativists and functionalists. The skeptic (me) must wonder if functionalism is just the current brand of dualism while eliminativism is just the latest name that dualists are trying to pin on reductionistic materialists.

Behaviorism: boon or boondoggle?

An aspect of Rey's presentation of eliminativism that has captured my attention is the attempt to identify behaviorism as a key pillar of eliminativism. As a boy, I found some of John Watson's and B. F. Skinner's books on my father's book shelf. I found the austerity of their approach to behavior elegant and powerful and a refreshing escape from the baroque excesses of Freudian and Pop Psychology. It was in this forge of audacious over-simplification inherent in the behaviorist research program that my personal approach to the mind was formed.

In this context of looking for accessible toe-holds for our assault on The Great Mountain of the Mind, I came to accept a key principle of experimental science, that when we are confronted by a complex system we wish to understand, it is sensible to start by retreating from our loftiest goals of understanding and illumination. It is practical and efficacious to first study any available simpler but related systems rather than the complete complex system. It pays-off to make simplifying assumptions and restrict attention and efforts, at least initially, to the simplest aspects of the complete complex system. Of course, all of this gross over-simplification is done with the full intention of eventually, at a later date, and after we think we have learned something from the simpler systems, returning our attention to the complete complex system.

There is a serious problem in executing a research program that makes use of such a strategic retreat. Many people, at the start of the retreat, will fail to see that you are actually getting anywhere. And what if the complete complex system is so complex that the retreat lasts for decades? Many nay-sayers will declare the whole campaign an obvious failure before it even begins to make serious advances. This is exactly where the reductionistic and materialistic approach to mind/brain monism finds itself. Behaviorism was an important part of the strategic retreat: ignore complex mental brain states and see what sense you can make of behavior in terms of simple stimulus/response relations. This research strategy has paid off brilliantly. The study of a simple form of learning and memory in the peripheral nervous system of the marine sea slug, Aplysia (6), allowed our first glimpse of how cells and molecules can produce a memory. What was learned in that simple system has now been shown to be applicable to human learning and memory, and so the more difficult task of understanding learning and memory in the mammalian central nervous system is now well underway, after a boost from Aplysia. And even now, as we open up the brain for close scrutiny, the behaviorist methodology of stimulus/response analysis is used every day by working neurobiologists to probe the interactions of the many sub-systems with which a functioning brain constructs a mind. It is painful for me to watch Rey (1) try to ignore the huge historical and on-going importance of behaviorism while he executes his condemnation of Radical Behaviorism.

I cannot escape the suspicion that Rey is starting with Radical Behaviorism and constructing a straw man in the form of Total Eliminativism which he will then morph into Radical Connectionism which he will then deftly topple, leaving the field open to what ever brand of functionalism he cares command forward from his imagination. If this is all his book has to offer, I will be disappointed.

I do not accept Rey's characterization of behaviorism as an attempt to avoid mental phenomena because "they did not seem to serve as the proper objects of scientific inquiry". I have always thought of the origins of behaviorism in another way. At the start of this century nobody had the tools required for study of the physical processes that occur inside the brains of behaving animals. Therefore, the sensible strategic decision was to put on-hold our desire to understand brain function in detail and, instead, treat brains as black boxes. Provide inputs (sensory stimuli) to brains of experimental animals and record the resulting outputs (motor behavior). Based on these stimulus/response data, try to make educated guesses about what must be going on inside the black box (the brain). This logical research strategy in no way equates with eliminativism. It never denies that the brain contains complex physical structures and processes that science must someday (when the tools become available) attack by direct investigations. Behaviorists did not start with a denial of mental processes, just a pragmatic avoidance of a (then) experimentally inaccessible bundle of technical difficulties better left for future generations.

Yes, there were behaviorists who tried to push the "brain as black box" strategy to the furthest extremes. How much of animal behavior can we explain by S/R coupling without having to actually know anything about the mechanisms of brains? But very few (if any) behaviorists actually fit into the Total Eliminativistic Garb that Rey tries to dress them in.

Beyond Behaviorism: a time to tear-down or a time to build-up?

It is amusing that Rey champions Chomsky and linguistics as the great vanquishers of RBism. This comes across like a Tide Detergent commercial proclaiming Tide as the savior of long-suffering grass-stain-on-the-pants-leg-inflicted house wives. The commercial never mentions that there are less expensive detergents on the market that work better and that Tide is a horrible polluter of lakes and rivers. Rey seems not to notice that Chomsky's research strategy ("we need to study each of the factors in relative isolation and only then consider how they might interact") is exactly the same kind of reductionistic research strategy used by behaviorists. The only difference is that the linguists brought to the table a few new tools (such as the study of competence rather than performance) for probing the black box of the brain that were not available to the classical behaviorists. This is how science works. One generation studies a complex system as an impenetrable black box. The next generation applies new tools to the task of opening up the black box. Eventually the contents of an old black box is made accessible to detailed analysis. Is it wise to gloat and condemn the generations that came first?

It is also a serious miss-reading of the history of biology for Rey to suggest that cognitive neuroscience is just some extension of Chomsky's program in linguistics. Rey, as a philosopher, is particularly intrigued by human language use and inordinately impressed by Chomsky's black box construction, the "LAD". The study of animal behavior first by natural philosophers and later by biologists (from Aristotle to Darwin and Piaget and E. O. Wilson and right on past Chomsky with hardly a side glance) is a rich story within which Chomsky stands as a minor bit actor. I doubt if there was ever a biologist who needed Chomsky to point out the importance of brains and brain processes in animal behavior.

The issue of nature vs. nurture pervades all of the study of animal behavior. It is a gross misdirection for Rey to suggest that RBism had so warped the scientific community's perceptions that everyone with an interest in animal behavior was hopelessly wedded to empiricism while it took some sort of revolution in cognitive science to recognize the importance of the structure of our evolutionarily designed brains. Some of the most famous and dramatic observations made by behaviorists centered on innate responses of new "born" animals to stimuli. Example: hatchling bird chicks who are normally fed by parents often peck (response) at anything that looks like the beak of an adult of their species (stimulus).

Some Common Ground: the Problem of Categorization

Where does all of Rey's posturing lead us? Finally, after 124 pages (1), Rey picks up the scent. Rey starts with the same jumping-off point as does Gerald Edelman (Page 34 of Neural Darwinism (4)), "categorization". How do animals recognize individual examples of sensory stimuli as being classifiable into a finite collection of categories? How does the sea gull chick recognize a beak when it sees one? Both Rey and Edelman specifically cite Gibson's 1979 book "The ecological approach to visual perception" as a classic, recent, and incorrect but logically extreme view of the all too common idea that the sensory organs and sensory input processing systems of brains are able to simply recognize key features embedded in sensory inputs and successfully assign them to their correct naturally occurring (think of Platonic Forms) categories. The problem is that there are no "naturally occurring categories" in the Platonic sense that appeals to people like Gibson. The chick recognition system for beaks has been designed by evolutionary selective pressures and the beak detector system of chicks has to be constructed from cells and molecules inside each developing chick brain. Thus, the category "beak" is a Rube Goldberg construct of practical benefit to birds, not some Platonic ethereal Form placed into bird brains by the Great Programmer in the Sky. While animals have a few such innate feature detection systems (for example, humans have a face detector), most of the categorization of sensory inputs that animals achieve has to be carried out by brain systems which learn to recognize patterns in sensory inputs and construct from scratch feature detectors that are well suited to the environment the animal finds itself in. It is only the fact that our brains are so good at this and that any two human brains are likely to contain many of the same feature detectors (because we share the same environment) that it seems obvious to us that there are natural categories of objects in the world.

No Memory, No Mind

From the common starting point of categorization, Rey and Edelman take one more important step together. Both agree that memory is central to mind. Both Rey and Edelman stress the fundamental importance of understanding how a brain can interact with a complex environment and successfully store memories in response to those interactions with the environment so as to allow an animal to successfully respond to future challenges from the environment by making use of remembered past experiences. From this point of common ground, Rey resolutely walks West while Edelman heads East. (See chapter 10 of Dennett's The Intentional Stance (12) for a description of this coordinate system. What is to the North? People from Brentano to David Chalmers who have suggested that there are aspects of the mind that conventional reductionistic materialism cannot deal with. To the South is the land of eliminativism with respect to "a science of intention" and propositional attitudes, a place where various "radical behaviorists" and philosophers such as Quine and Paul Churchland have ventured.)

Rey suggests that what we want to know is how to construct a learning machine that could learn human language by listening to other people use language. This is where I came to Artificial Intelligence research and neurobiology in the late 1970's. I also decided that language learning is what we really want to understand, but a tough nut to crack. Rey seems willing to stick with the problem of language learning and play the very type of black box game that he condemns the behaviorists for having played. Rey cannot open up the brain and see the mechanisms by which human children learn language, so he is content to take the results of linguistic black box analysis and construct a functionalistic theory of what must be going on in human brains when they use language. Rey immediately endows certain key terms (such as "representations") with the full powers of Platonic Forms, including his belief that they are "causally efficacious".

In contrast, Edelman and his fellow biologists (including me) and philosophers like Dennett and Churchland, say: "Not so fast!" It is a mistake to jump right into this attempt to construct a functionalistic theory of mind based on your black box analysis and introspection-derived intuitions about how humans use language and reason. We need to take the time to really figure out how memories are formed inside brains. Only then will we be able to determine how a mind is constructed from brain systems that have the power to create and utilize memories.

Neural Networks, Artificial vs. Biological

From this point of divergence between Rey and the non-functionalists, Edelman and Churchland are both headed back to the lab. Churchland, as an outside commentator on the connectionistic, parallel-distributed-processing school of artificial intelligence research, is enthusiastic about the power of such "neural network" models (3). Edelman, who has actually made a study of biological brains, is certain that the types of networks studied within artificial intelligence are too simple to accomplish what a biological brain can do (5). Edelman and his co-workers have developed a series of neural network models (4) that differ fundamentally from those popularized by Churchland (3). Interestingly, Edelman faults the AI neural nets for reasons that Rey would appreciate. The typical AI neural network must be designed by the programmer for a certain task and then trained by a programmer on a "learning set" of carefully prepared stimulus/response pairs. A non-local mechanism such as back-propagation of errors is then used to adjust the "synaptic connections" within the network so that the network is trained to "learn" to correctly categorize future inputs and match the correct responses to each category.

What does Edelman see as the basis for categorization and memory? Edelman's claim is that the memory storage mechanisms of biological neural nets are based on locally acting synaptic modification rules. Such local memory mechanisms can only construct a memory system when embedded in a neural network with the special architecture that we observe inside biological brains. This architecture has two key aspects: 1) rich pathways for feedback between interconnected brain regions, and 2) and specific patterns of local interconnections within each of those more widely distributed brain regions. The local synaptic modification rules that function within local regions of brain tissue are now being revealed by neurobiology (for example, long-term potentiation of synapses (6), LTP). The anatomical connections within and between brain regions are also being mapped by neuroanatomists. The combination of identifying locally acting synaptic modification rules and mapping of the network architecture is exactly what solved the problem of explaining simple learning in Aplysia (Essentials of Neural Science and Behavior, chapter 36).

In addition to the biology research that is needed to flesh-out Edelman's theory of mind, Edelman and his co-workers have constructed proof-of-concept neural network models that show that it is possible to make computer-based robotic systems that use biologically realistic neural networks to interact with an environment and successfully construct categorizations of sensory inputs and use them to adaptively and successfully control behavior (See the so-called "Darwin" series of models described in Neural Darwinism (4)). Others are now following along, working methodically into the central nervous system along Eric Kandel's (6) and Edelman's (4) proclaimed path to success (7).

Edelman's approach is entirely bottom-up. According to this view (which I share), self-wiring (using adhesion proteins and molecular coordinating signals during embryonic development of the nervous system) modular components and local (neurotransmitters, receptors, etc.) synapse modification rules are what brains need to construct learning machines. People like Rey, in contrast, who want to work top-down are tempted to turn to advisors like Wittgenstein for guidance in finding clever top-down rules for categorization. I imagine that Rey would dearly love to be able to publish a list of such imagined rules as an appendix to the next edition of Contemporary Philosophy of Mind. Biologists, in contrast, are content to wait for experimental neurobiologists to publish all of the brain's synaptic mechanisms and map all of the brain's intricate wiring. I am confident that it will then be possible to replicate the circuitry of biological brains in robotic constructs which will be capable of confronting a complex environment and able to learn how to adapt in the same way humans do. Who will be shown to be correct, Rey or Edelman? I suspect we will know for sure sometime in the next century or two when a robot with human-like intelligence is constructed and it learns to speak a human language and is granted human legal rights and protections.

Consciousness, can we get there from here?

However, there is still a lot of work to be done to get from Aplysia to Chomsky's LAD and issues like finding a mechanistic explanation for the type of consciousness we experience subjectively. The most audacious attempt to bridge this gap was made by Edelman (8). However, Edelman's attempt to describe how the type of analysis that solved the question of how simple learning works in the peripheral nervous system of Aplysia can also solve the problem of human language, consciousness and other higher brain functions is only a very rough sketch of a possible theory (8). Possible mechanisms for higher brain functions and consciousness such as the one suggested by Crick (9) are needed and remain to be discovered. While neurobiologists are sure that we are on the right path, that reductionistic materialism will solve the mystery of how a brain makes what we experience as mind, there is still much hard investigative work to be done.

But where are we today, at the end of Chapter 4 in Rey's book (1)? Rey seems unaware of Edelman's existence. Both Rey and Edelman, from two different directions, condemn the brand of connectionism that Paul Churchland likes. Edelman is devastating in his attack against functionalists (see the postscript of Bright Air, Brilliant Fire (5)). Rather than try to reproduce Edelman's rather rambling assault on functionalism, I will attempt a simplification. Edelman points out the fact that functionalists are playing a black box game, assuming that the mechanisms used by biological brains to construct a mind are irrelevant to some higher level of functionality that could exist in either a biological brain or a robot. The key then, is what is meant by "irrelevant"? Here is where causation rears its ugly head. If we assume that it is cells and molecules or transistors and diodes that cause things to happen in minds that are generated by biological or computerized brains, then a description of those mechanistic components will explain for us how mind is embodied, completing the reductionistic materialist program. If Rey is correct, we can ignore the physical mechanisms of mind and we will find some set of causal rules by which higher order mental states (like beliefs and thoughts) cause human behavior. For me, such a contest fits nicely into the age-old debate between dualism and a monism rooted in reductionistic materialism. It seems that Rey is using the inadequate types of neural networks that have been studied by artificial intelligence researchers as a bogeyman against which he can array the forces of functionalism. By arguing that there is something scientific in his view of functionalism, Rey attempts to evade being labeled a dualist, even though after 134 pages (1), Rey still has not given me a hint about how "mental states" can be causally efficacious if not by physio-chemical (reductionistic materialism) processes in the brain. I wonder if the situation will look any different to me after I have completed Rey's book.

John's View of Eliminativism

Eliminativism is not a word that shows up much in the world of Science. I am very interested in the importance of constructive interactions between philosophers and scientists. If philosophers (like Rey) find it useful to emphasize issues like eliminativism, I want to know why. I like the way philosophers study every possible angle of a problem (unencumbered by the biases of working scientists) and put a name on every part of complex issues. As far as eliminativism goes, I feel like a person who has just discovered that he knows how to speak prose. Clearly, we all deal with eliminativism on a continual basis in the memosphere. Human societies can contain certain belief patterns for centuries and then discover that the time is suddenly ripe to eliminate a mistaken idea. Since science is a great creator of change, we can wonder why scientists make so little use of an apparently useful term like eliminativism. I would like to know how long the term "eliminativism" has been in use by philosophers. Dennett indicates that it is less than 40 years old, at least as used within the philosophy of mind (12). Maybe it will eventually catch on with scientists.

However, one of my concerns is that philosophers and scientists often use the same words to mean different things. Certainly, elimination has a common usage in our language already. On this basis alone it might face poor prospects for use as a technical term by scientists. However, scientists do often use words that have a common meaning as technical terms with a specific biological meaning. Also there is already a way in which the word "elimination" is used to refer to removal of waste from an animal's body. It may be asking to much for a biologist to ignore this meaning and take to referring to themselves as eliminativists. Also, there is at least one other meaning of the term "elimination" within the history of biology (10). Ernst Mayr (page 488 of The Growth of Biological Thought) has used the term "elimination" to refer to the now defunct (but once influential) notion that there is a natural force which removes from biological systems all deviations from normal. I think it is fair to say that "eliminativism" is philosophical jargon, really just short-hand for saying something like, "the natural tendency for people to strategically decide to stop trying to defend and explain a previously held but now defunct idea or belief". If the term "eliminativism" is to be used by scientists, it must become attached to a replicating meme within the culture of science. Since there is now a new science of memetics (11), it may be possible to find a rational way to spread the term "eliminativism" within science. But I also think that in coining the technical term "eliminativism", philosophers have identified and named a key component of the memosphere, a component that is so well embedded in our human way of life, that most people have a hard time even being aware of it. The main problem is that the eliminative process as identified by philosophers is usually a slow social process that escapes the attention of human nervous systems which are tuned to more rapid environmental changes. Nobody spontaneously notices the slow changes that occur in human language over the course of decades. Also, memologists are often more concerned with the sudden explosion of new memes within the memosphere rather than the slow slipping away of old and neglected memes. Maybe a memetic engineer can do something (rename it?) to make eliminativism attain its rightful prominence within all domains of human discourse.

If, by the end of Rey's book, I find myself unable to accept the way he wants to include causal powers within the meaning of functional components of his theory of mind such as "beliefs", without allowing that such causal powers must ultimately reside in the physical brain components, then I will be forced to admit that by Rey's definition, I am an eliminativist with regard to mentalistic terms (as he defines them) like belief. However, I will not willingly accept his definition of terms such as belief and I will continue to use the word belief according to my own reductionistic and materialistic view of how brains construct the beliefs that we find in our minds. This is memetic warfare and a strain in the relations between philosophy and science. It is a problem we must work through together. If philosophers and scientists cannot speak with each other successfully, I fear that the public at large will never be able to understand what I believe are important issues that human society will have to face as we come to better understand our brains and minds.

 post script
Go here for my comments after having completed Rey's book.

Works cited.

1. Georges Rey (1997) Contemporary Philosophy of Mind ISBN 0-631-19069-4.

2. Daniel C. Dennet (1991) Consciousness Explained ISBN 0-316-18066-1.

3. Paul M. Chruchland (1995) The Engine of Reason, the Seat of the Soul ISBN 0-262-03224-4.

4. Gerald M. Edelman (1987) Neural Darwinism ISBN 0-465-04934-6

5. Gerald M. Edelman (1992) Bright Air, Brilliant Fire ISBN 0-464-00764-3

6. Eric R. Kandel, James H. Schwartz, and Thomas M. Jessell (1995) Essentials of Neural Science and Behavior ISBN 0-8385-2245-9

7. Patricia S. Churchland and Terrence J. Sejnowski (1992) The Computational Brain ISBN 0-262-03188-4

8. Gerald M. Edelman (1989) The Remembered Present ISBN 0-465-06910-X

9. Francis H. C. Crick (1994) The Astonishing Hypothesis ISBN 0-684-19431-7

10. Ernst Mayr (1982) The Growth of Biological Thought ISBN 0-674-36446-5

11. Aaron Lynch (1996) Thought Contagion ISBN 0-465-08466-4

12. Daniel C. Dennett (1987) The Intentional Stance ISBN 0-262-54053-3

Go to John's Book Page.

Go to John's Home Page.