Thursday, January 12, 2012

An old essay of mine on eliminative materialsim

I realize not all of you have access to this online library.

Interestingly enough, in Bill Ramsey's treatment of EM in the Stanford Encyclopedia, he references this article, but not one that I wrote later (but was published earlier), in which I replied to Bill's own critique of the self-refutation objection.

I would be interested in BDK's take on this exchange.


Blue Devil Knight said...

It would take me a while to properly go through all that.

I should ask for clarification: does your belief in belief/assertion commit you to the view that we literally think in a propositionally structured representational format (a language of thought)? It could be neural or dualistic or whatever, it doesn't really matter.

Or do you admit of the possibility that thought could be fundamentally nonpropositional (analagous to our visual experiences, say), but we still make assertions and have beliefs?

For instance, you could think propositions are abstract objects, and to have a belief in such a thing is to be in a nonpropositionally structured mental (or neuronal) state with the right sort of relation to that propositional content.

This is a pretty key point, and if I knew where you stood it would help me interpret your view.

It will likely take me a good week to work through those papers with the right level of detail to say anything worth saying, but knowing your overall approach from the above questions is often helpful in talkiing to people about eliminativism.

Blue Devil Knight said...

Note, just in case it got missed. I am not an eliminative materialist about beliefs or propositional attutude psychology more generally. I am officially agnostic on the question, though I lean against flat-out eliminativism (in humans anyway).

And I was doing so well staying away from this damned place.

Blue Devil Knight said...

Also, where is the second paper the link is to philpapers but I can't access the pdf???

Could you email it to me possibly?

Victor Reppert said...

I don't have it to e-mail.

Blue Devil Knight said...

Well, I'll take a look at just the one with the pdf provided, the 1992. So far I'm impressed with it after the first part.

mattghg said...

if beliefs are called into question then it seems hard to assume that assertion, or meaning, or anything in the intentional area is safe

What is the current dialectical status of the above? I ask because a lot of the reaction to Rosenberg indicates that he denies that there is linguistic meaning at all.

I think this is somehow related to BDK's question(s).

Blue Devil Knight said...

Matt: I'm not going to worry about Rosenberg, who doesn't have a particularly good grip on the issues, frankly, or the neuroscience.

To answer your question, With Churchland, there is linguistic meaning that is parasitic upon the content of mental states. This is the standard view of linguistic meaning, held by nearly everyone, and Churchland is no exception.

The core disagreement concerns the structure of those internal meaningful states. On one hand are the language-of-thought advocates (e.g.,Fodor), who think that the content of mental states is propositionally structured, such that the brain literally employs sentence-like strings of discrete symbols to do its thinking.

Churchland disagrees (for many many reasons he outlined in his book 'Scientific Realism and the Plasticity of Mind' he published in 1979). He thinks that neuroscience will show that the fundamental computational units are not internal sentences, but something different, that the propositional kinematics of propositional attitude psychology will be replaced by the neurocomputational kinematics of the high-capacity, analog, parallel information processor in your skull.

That was the core debate between Churchland and the rest of the philosophy of mind community. It was originally posed against Fodor and the like, other naturalists, without concern for questions of materialism versus dualism about mental contents.

Getting back to your question, he would say we express meanings, but that the meaning of linguistic strings depend upon the meaning of the neuronal states, that linguistic meaning is what you get when you try to squeeze information from the brain's high-capacity information processor through the filter that translates it into the relatively impoverished low-capacity communication medium of ordered string of symbols that is public language.

Just to prime the pump a bit: think of your visual experience. Does it seem propositionally structured? Intuitively, it seems not. It seems to be a high capacity representation of what is going on in the world, one that linguoform sequences of symbols would be relatively impoverished to express. What if mental states are all like that?

That's the positive story in caricature form.

Churchland does go pretty far out there sometimes in the implications of his story. E.g., suggesting that people's talk of 'beliefs' (which, he claims, assumes propositinally structured mental states) will be replaced. But such things are actually inessential, but because more provocative get the most ink.

mattghg said...

BDK, thanks for the outline of Churchland's thinking. In the light of it, I'm struggling to see what's eliminative about 'eliminative materialism'.

the propositional kinematics of propositional attitude psychology will be replaced by the neurocomputational kinematics of the high-capacity, analog, parallel information processor in your skull

Why would that be an elimination of propositional attitude psychology and not a reduction of it?

Perhaps what I'm asking is, why does Churchland think that talk about beliefs assumes that mental states are propositionally structured?

I think I have a fairly good idea as to what Fodor's views are. I'm a grad student in a Linguistics department where he's revered.

mattghg said...

For instance, you could think propositions are abstract objects, and to have a belief in such a thing is to be in a nonpropositionally structured mental (or neuronal) state with the right sort of relation to that propositional content.

FWIW, this doesn't seem like an outrageous view to me.

B. Prokop said...

"[T]he brain literally employs sentence-like strings of discrete symbols to do its thinking."

It's interesting that when I think in Russian (the only language other than English that I am fluent in), it's different than when I think in English. And I know many, many other bilingual persons who testify to the same phenomenon.

Also, since I've started studying Latin last year, I've noticed that the mind works quite differently in that language as well.

Blue Devil Knight said...

Matt: as usual good points.

Why would that be an elimination of propositional attitude psychology and not a reduction of it?

He thinks the sentence-based view will turn out to look radically inadequate once we have the literally true neurocomputational description, so the former will go the way of phlogiston.

Perhaps what I'm asking is, why does Churchland think that talk about beliefs assumes that mental states are propositionally structured?

This is a great question, I think it gets at a serious weakness in his overall presentation.

Largely b/c the traditional view of thinking as mental sentence-tokens was how belief/intentional states were cached out in the technical literature (especially starting with his advisor, Wilfrid Sellars, whose influence would be hard to overstate).

Where I think he is weak is in assuming that even "folk" think beliefs are internal sentences, when this is actually a fairly high-falutin' notion. Fodor and Churchland both seem wrong that the brain-sentences view is some intuitive truth of folk psychology.

That's why I tend to avoid the term 'folk psychology' and just use the technical terms of language of thought and such.

Bob: it is an interesting question whether language-use requires brain-sentences. This is one reason I lean against EM wrt humans--because of our language use.

But again, only LEAN against it, technically agnostic. You could think about language in a non-linguoformal medium, after all.

Many people, incidentally, think the EM and language-of-thought views are consistent: that, as Matt suggested, the brain-sentence view is a kind of bird's-eye-view of the brain's computational architecture, but if you zoom in you will see it is actually nonsentential.

Smolensky has done the most to push this in a concrete direction.

Blue Devil Knight said...

Note you don't have to be an eliminative materialist about internal sentences.

You can be an eliminative dualist too, think that the mind is structured in nonpropositional format, but is not a neuronal phenomenon but some sort of dualism.

Blue Devil Knight said...

Wow, deja vu.

Related threads here ("Note you can also be an eliminative dualist wrt propositional attitudes, so eliminativism is orthogonal to the ontological questions of the reducing base").

Here ("brain inferences are not in a propositional format but a different analog representational format (the way you might imagine your visual experiences as not propositionally structured). But this is able to interact with (produce and understand) public language even though the brain doesn't trade in linguoformal representations (just as we can produce and understand "walking" behavior even thought the brain doesn't walk).")

William said...

Personally, I like the dual coding theory (,
with the mind using both language-structured and non-language-structured concepts of various sorts. This fits the different types of problems that occur with various regional problems.

So both beliefs and the vaguey stipulated representations favored by the Churchlands would be real things, like both the air vibrations and the music are real things, unless you feel eliminativist about music, I suppose. To me, music feels more real than air vibrations do, odd, huh?

Blue Devil Knight said...

William I agree that some kind of dual representational system is promising, in some form.

But you said:
vaguey stipulated representations favored by the Churchlands

That isn't quite right. The fully articulated view can be found in the following:
"State-Space Semantics and Meaning Holism" (1993) Philosophy and Phenomenological Research.

"Conceptual Similarity Across Sensory and Neural Diversity: The Fodor-Lepore Challenge Answered," (1998) Journal of Philosophy

Fodor-Lepore reply to above, and response from Churchlands found in 'Churchlands and their Critics'

'Neural Worlds and Real Worlds' (Nature Neuroscience 2002).

'Neurosemantics: On the mapping of minds and the portrayal of worlds' (2007) (Chapter 8 in Neurophilosophy at Work).

Blue Devil Knight said...

Bibliography of neurosemantics here.

Blue Devil Knight said...

Note that Churchland pointed out the possibility of being an eliminative dualist in his 1979 book, so that isn't my brilliant idea.

He wrote, of language of thought theories, that they might:
be replacable by some more general theory of "ectoplasmic essences"...We might call this possibility 'eliminative dualism'! It is perhaps not surprising that this possibility has gone unremarked, since the preservation of the common-sense ontology of the mind has always been part of the dualist's sales pitch (p 108)

But of course Churchland doesn't take dualism seriously, but was making the conceptual point that there is nothing about his rejection of sentential mechanics that was essentially materialistic.

Victor Reppert said...

Strangely enough, I don't know that you are doing justice to your old teacher's position, because the way you make it sound as if, when you boil Churchland down, it amounts to saying that Fodor is wrong. But central to his position is the idea, not just that Fodor is wrong, but that "folk psychology" is a false theory. Now unless he really thinks "folk" are good Fodorians (well, maybe Fodor thinks that!) it's hard to see that that is all it amounts to.

When I think, I am aware of a logic framework in which my thought occurs. The true-false distinction seems to essential to my thinking processes.

You also get, in Churchland, statements to the effect that elimiinativism is going to lead also to the elimination of truth-talk. Again, this could be Churchlandese for something more benign, but what would that be?

William said...

My criticism of Churchland is that he vaguely gestures (I mean, just suggests a theory is possible) at neuroscience as giving a substitution for semantic meaning, by replacing verbal semantics with some other (algorithmic?) abstraction. This to me just replaces one abstraction with another.

I see no reason to do so except for following the implications of a materialistic dogmatism.

Blue Devil Knight said...

William: he really doesn't just vaguely gesture. He developes, in some detail, a positive theory, in concert with ideas from neural network models. It's in the papers I . Where it is a bit vague and undetailed is how he sees the relation between this neurocomputational properties and the properties of public language. More on that below.

Victor: Thanks for the comment, though I am still curious what your answers are to my questions above, they'll determine what you think is crazy/reasonable.

[T]he way you make it sound as if, when you boil Churchland down, it amounts to saying that Fodor is wrong.

The core idea was the rejection of sentential theories of cognition, and then he would go on to explore the implications of that. So it is right that the core idea of his eliminativism applies to "language of thought" theories, with Fodor as prime target.

He did explicitly claimed that folk psychology was a species of language of thought (following Sellars). Which as I said above, seems wrong.

On your concerns about truth, that is an interesting one.

Because he is an eliminativist about brain-sentences, he questioned whether predicates that apply to public sentences will ultimately apply to thoughts if they are nonsentential.

Is a picture true or false? Is your visual experience true/false? If such things are nonpropositional, and you think it therefore can't be true/false, then how do you describe how "good" its fit with the world is? These are the kinds of things he has to worry about, and anyone that thinks visual experiences are not propositionally structured represenations has the same problems.

He just has them worse.

A strength, and weakness, of his position, is his focus on developing a theory that would not privilege one species. Rather, he wanted to look more generally at how brains exploit and use information, something that could apply to rattlesnakes, tree frogs, etc.. Within this non linguicentric framework we need to tell the story about how language use and understanding would emerge as an idiosyncratic feature of one particular species.

The problem is, it isn't clear his eliminativism extends very well to language-using species with complex grammars. Truth, at the very least, should apply to our public sentences, which display the serial symbol-string type format he so much decries. He has trouble dealing with such things, including scientific theories like relativity, which are typically expressed and evaluated in public sentences.

Same for logic. Does thinking of sentences/propositions imply you think in sentences/propositions? Clearly not (that's the point of quote above about biking).

Blue Devil Knight said...

Just to be clear again, I'm not defending the antisentential view of Churchland, and all the implications, or putative implications, of such a view. Just saying it isn't as crazy as often portrayed, even if there are problems with it.

Blue Devil Knight said...

One proposition/belief/sentence Churchland would have to say is true: The Patriots just demolished the Broncos.

That is sufficient to refute eliminative materialism.

Blue Devil Knight said...

Note while I am ready to defend eliminativism as a reasonable position, I cannot defend everthing that Churchland thought this implied. E.g., that belief/desire psychology (or more generally, folk psychology) is a subset of sentential psychology.

That said, I am not committed to folk psychology as having any kind of sacred purchase on what I should believe about minds. I am not convinced there is even a well-defined thing as folk psychology.

And if it is well-defined, it would be surprising if folk psychology were fundamentally right, given that (aristotelian) folk physics and (vitalistic) folk biology get things so fundamentally wrong.

Victor Reppert said...

I had drafted a post at halftime entitled "Can God save the Broncos?"
(Tebow et al.) Of course, this begs the question against your Patrianity, according to which God the Father coaches the Patriots, and God the Son quarterbacks them. Don't know who the Holy Spirit is supposed to be.

Blue Devil Knight said...

Holy Spirit is Robert Kraft! :)

Blue Devil Knight said...

I will read and have comments on your essay, with the above as backdrop, sometime tomorrow.

Note while in this thread we haven't been taking the
folk-->language of thought
view very seriously, it used to be much more popular. It has become a cottage industry to argue about this, whether our theories that we use to explain other people's behavior are LOT-like theories or not. Or even whether they are really theories. The arguments about our commonsense theory of mind are legion at this point.

But on the face of it, we do seem to ascribe thoughts to people in propositional format, we say they are "inside" the person in some sense, etc.. Why did Dave punch that guy? He thought that the guy insulted his daughter. We ascribe truth values to these internal states. "It turns out Dave was wrong, that guy didn't actually insult his daughter."

So taken naively, we do seem to posit internal (or private) quasi-sentential states when we ascribe beliefs.

And this was the dominant view when Churchland started writing, and it is especially important to know he started writing under his graduate advisor, Wilfrid Sellars, the original source of this 'theory theory' view.

Blue Devil Knight said...
This comment has been removed by the author.
Blue Devil Knight said...
This comment has been removed by the author.
Blue Devil Knight said...

I just read the paper, not super-closely as I don’t have the time. With that caveat, here is my impression, happy to have it crushed.

For those that haven't read the paper, one popular argument against eliminativism about beliefs is that it is self refuting. E.g., if EM is true, then beliefs do not exist. In which case your expressions have no meaning, including your support for EM. So it is self-refuting to say EM is true.

One of Churchland's responses is an analogy with vitalism: a vitalist could say your claim that vital fluids don’t exist is self-refuting. After all, that claim would have no meaning if you didn't have vital spirit, because to utter anything you must be alive.

According to Churchland, these arguments beg the question: Churchland thinks that there are meaningful utterances, but they do not depend on having internal propositionally structured mental representations such as beliefs. To assume that meaningful utterances require beliefs is to beg the question against his theory. Victor's paper performs an analysis of 'begging the question' and suggests that Churchland's analogy assumes too many controversial things to work. Frankly, I am a bit underwhelmed.

Note that Churchland’s analogy is not meant to provide a deductive proof, but simply demonstrates how unconvincing the self-refutation argument looks from within Churchland's own theory. And if Churchland shows that, from within his own theory, the self-refutation argument doesn’t work, that is all he needs to do. Consider the argument against Churchland (T is EM, Y is ‘Expressions have no meaning’):
T implies Y.
Y is false
Therefore, ~T.
But the problem is that Churchland’s theory X does not imply Y. That is his point. Sure, you could tweak Churchland’s theory so it did imply Y, but then it wouldn’t be a refutation of Churchland, but someone else.

So Victor’s point, that there are questionable auxiliary assumptions that Churchland makes, and if you reject those the self-refutation argument might work, is a bit beside the point. Using that strategy, you can refute theory T’, not T. You will have effectively assumed ~T to get the Y you want so you can conclude ~T. This is clearly to beg the question.

Victor's suggestions have an interesting implication that seems obviously right: we might preserve talk of beliefs if we reject some of the auxiliary assumptions. Hell, we could define beliefs as (nonpropositionally structured) internal representations in a neural network’s state space (i.e., use Churchland’s own nonpropositional state space semantics to define beliefs). After all, talk about 'belief' per se isn't all that big a deal for Churchland. It is the underlying theory of mental sentences that he has so consistently attacked. And which Victor didn’t seem to mention once in his paper. And has not really discussed in this thread, or at all at his blog if memory serves. That is to miss the core of EM even if it hits some peripheral targets such as whether we preserve talk of beliefs.

Blue Devil Knight said...

Looking back u have endosed some kind of propositional view, at somebryan level if notur psychological, so my last claim is an overstatement.

Blue Devil Knight said...

Srry on phone autocorrected last post

Matt DeStefano said...


Because he is an eliminativist about brain-sentences, he questioned whether predicates that apply to public sentences will ultimately apply to thoughts if they are nonsentential.

From what I've read of Churchland (and I'm a big fan - so quite a bit!), he wants to think of conceptualization as a 'map', and it doesn't make sense to apply the binary judgments of true/false to a map. "Is that map of Alaska true?" Rather, it would be degrees of approximation that would separate more successful maps from less successful maps.

Patricia Churchland has written that all evolution cares about are the four F's (fighting, feeding, fleeing, and reproducing), and that whatever "truth" is, is merely hindsight. This makes sense if we look at representation non-propositionally, but rather as a 'map' (neural networks approximating practical environments).

Incidentally, this view of our representation of external reality as a 'map' saves us from all sorts of problems regarding reliability. (Plantinga's EAAN, Bealer's self-consciousness problem, etc.).

Blue Devil Knight said...
This comment has been removed by the author.
Blue Devil Knight said...

Matt, yes, you are right about what he wants to replace things with.

I think people tend to make a bit too much of the four F's stuff. It's not like that is inconsistent with evolving more accurate representations of the world. If the frog snaps its tongue to the left b/c its sensory system represents things that way, but the fly is on the right, it won't feed, for instance.

By analogy, pumping blood is not one of the four Fs, but can be indirectly selected by evolution because of the contributions to said desiderata (well, those, and just plain old not dying as would happen if we removed your heart).

Blue Devil Knight said...

Incidentally, I just learned from Matt that Churchland has a book coming out that will pull his semantic theory together in one place, undiluted: Plato's Camera: How the Physical Brain Captures a Landscape of Abstract Universals.

Based on the Table of Contents, it will address many of hte concerns people have brought up over the years, and seems to address more thoroughly than his previous work the role of culture and language in our cognitive lives.