Mind-body problem-Chomsky/Nagel

apeiron

This is not a study of how mind - conscious experience - arises. A description of the apparent contents is not the same thing as a description of the production of the contents. Which again, to remind, is what the OP is about.

So the cognitive paradigm is based on the belief that an "internal processing architecture" can account for the mind as a material phenomenon.

It would seem that a behaviourist description of the mind would apply even if minds were immaterial souls. If I say I see red or have a toothache because my verbal behaviour is being reinforced by a social community in the presence of an inner qualia, then that qualia could equally well be the product of a complexity of neural activity, or some immaterial soul-field.

That is why Behaviourism seems detached from the questions that cognitive science returned to.

Of course, behaviourism with a small "b", would correctly point out that architecture-based approaches to mind need to be ecologically valid, embodied, evolutionary-rooted, etc. And with computationalism and functionalism, cogsci strayed away into abstract, disembodied and otherwise unrealistic thinking.

Thinking about it this way, I can see a big part of my objections to Skinnerian Behaviourism is that its idea of behavioural context was so shallow. This was why in the 1970s I went off looking for the way that the brain adapts to its contexts over multiple temporal scales, from the evolutionary through the developmental and habitual right up to the anticipatory.

Yes I agree that Fodor was a wrong turn and modularity did appear to dominate the conversation. But it is interesting how many on Bohm2's millenium list were hierarchical and distributed architecture thinkers. This list makes cogsci look much more balanced and reasonable than I remember.

Again, this seems to be social dynamics at play. Fields become remembered for their extremes. They become generalised in the academic memory so that what was asserted as "right" can be then definitely rejected as wrong, so allowing paradigms to shift, "progress" to be made.

And I don't think cogsci ever really recovered and took the right turn. A belief in strong modularity has persisted into evolutionary psychology and cognitive neuroscience.

In mind science, there are still the unresolved tensions caused by some deep polarities. Is the brain's architecture distributed or modular, computational or dynamic, material or informational, nature or nurture, hardwired or learnt?

The reasonable view is that it is always somehow both. But reductionist logic does not allow that answer. The law of the excluded middle must apply. One proposition must be true, the other false.

Which is why I focus on more complex models of causality - hierarchical or systems causality - where dichotomies are not a bug but a feature. They are the process by which hierarchies naturally arise.

Again, this was explicit in Luria's classic, The Working Brain. His first law of brain organisation was that it is a functional hierarchy. His second and third laws were then about the fundamental dichotomies in this organisation - plasticity~stability and fringe~focus.

Great, then you should have no problem offering specific examples of Behaviourism in practice.

Is it a reasonable question to ask how operant constructs like mands and tacts have fruitfully led to new neurolinguistic insights? How have they guided us in investigating functional brain architecture, in making sense of the brain's complexity.

Mr.Samsa

I said that his review is given little-to-no weight in academia, not his work in general. And yes, it was an influential piece of work but that wasn't quite what I was getting at - I was more highlighting the fact that the ideas contained within it have mostly been rejected and discredited (obviously with Skinner's ideas coming out trumps in a lot of areas).

So the review was definitely influential, in that it's viewed as being one of the works that sparked the "cognitive revolution", and it also resulted in a massive amount of research in areas of language - like the language acquisition device, etc. Most of his ideas in the review were found to be poorly thought out and ruled out pretty quickly, or they were wrong and unproductive, and ruled out after research came back negative.


Except the behaviorist account also discusses the internal processing architecture, just using different terminology. That's what is confusing me here. You seem to be presenting cognitive science and behavioral psychology as conflicting paradigms, and promoting the cognitive perspective, but ignoring the fact that cognitive science and behavioral psychology are interchangeable. They both study the same phenomena.

And this is ignoring the fact that cognitive psychologists use behaviorism as their philosophy of science.

I'm not sure behaviorism could apply to immaterial minds because it studies the inner workings of the mind, and if the entity is immaterial, then it wouldn't be able to study it. Remember, behaviorism isn't a black box approach - it studies what goes on inside the mind in the exact same way cognitive scientists do.

If it helps, keep in mind that there is no practical reason for having separate labels for cognitive science and behavioral psychology. They are the same field (hence why behavioral psychologists and cognitive scientists hop between the fields with ease, no particularly new training, or changing their methods or subject matter). The only reason there are separate labels is for political and historical reasons.

Shallow in what sense? Since behaviorism includes neuroscience, how does investigating the brain constitute escaping behaviorism?

But didn't you accuse behaviorism of being reductionistic? How can it be both reasonable (in that it balances nature and nurture, hardwired and learnt, etc), but also reductionistic (which you claim doesn't allow that)?

What particular area are you interested in?

Well Verbal Behavior led directly to Relational Frame Theory. This article discusses some of its findings in relation to neuroscientific findings. And obviously, the fact that Skinner's approach resulted in us understanding how language is learnt and made it possible for us to develop language therapies - surely a fairly impressive neurolinguistic insight.

bohm2

Just to add some more papers on this off-topic detour (which is fine since it's forcing me to read and learn something), I came across this paper. The author discusses this whole issue including MacCorquodale’s paper and Chomsky's response (or really dismissal) to it.

On Chomsky’s Appraisal of Skinner’s Verbal Behavior:A Half Century of Misunderstanding
http://www.ncbi.nlm.nih.gov/pmc/arti...-29-02-253.pdf

Mr.Samsa

Cheers, yeah that's quite a good paper too but I didn't link to it because I think it focuses a little bit more on the political side of the issues, and more on Chomsky's tone and attitude, whereas MacCorquodale just deals with the arguments. But it's still a great paper.

apeiron

The point was how they don't study them in the same way.

Thanks for proving my case. Here we have behaviourism not discovering anything about the brain, but relying on an identification already made by neurolinguistics to justify their redescriptions.

The difference is neurolinguistics has a whole cognitive theory around N400s and other ERPs. They were critical in showing that the brain is acting in predictive Bayesian fashion to make best guesses of sentence semantics. There are neural net simulations that attempt to model the processing architecture, such as the Unification Space Model.

So on the one hand, we have behaviourists trying to justify bits of jargon by refering to what is already known in other fields. And on the other, we have science that has novel and surprising observations which are leading to architectural models that are then validated by simulations. And the people doing that science all call themselves cognitive researchers.

bohm2

In the most recent issue of "Language Learning and Development-Special Issue: Cognition and Language" there are some papers discussing this modularity debate. You guys might find the Chomsky piece in the same issue interesting (but not freely available for some reason unlike the other 2 papers?). The first link/piece below trying to accomodate both views is also worth reading:
The Modularity Issue in Language Acquisition: A Rapprochement? Comments on Gallistel and Chomsky
http://www.bcs.rochester.edu/people/...t_%20LLD11.pdf

Prelinguistic Thought
http://ruccs.rutgers.edu/~galliste/G...ic_Thought.pdf

I'm guessing here that in this quote Chomsky is taking a shot at Bayesianism?
Language and Other Cognitive Systems. What Is Special About Language?
http://www.tandfonline.com.myaccess....41.2011.584041
Video Version of that paper-at ~1:19 there's an interesting section about Turing and the meaninglessness of the question whether computers/animals can think:

apeiron

Thanks for pointing this out. From Newport's gloss, it seems that Chomsky has in fact shifted his position so that it fits the evolutionary view exactly as I argued in previous posts. The brain is generally arranged hierarchically and it was the vocal tract - the imposition of a serial constraint - which was the crucial difference that underpins syntactic communication.

So now we can all be Chomskyites, even if it took him 42 years to get there.

I thought the Newport summary was excellent. But it shows the need for a concept of hierarchical organisation that is rather more sophisticated, more organic.

As I said, dichotomies are treated as a bug rather than a feature so long as science tries to force the question into an either/or form - is it modular, or is it non-modular (distributed)?

You can call it semi-modular, semi-distributed. But that is not very satisfactory. Or instead, as with nature~nuture and all the other dichotomies that crop up in scientific description, you can say it is 100 per cent of both. Each is equally strongly true of the system in question.

It is quite clear that to be conscious of the world, the brain has to work equally hard at two things - integration and differentiation. It must see the parts and see the whole. So it should be no surprise that it is organised along these lines.

It has suprisingly located responses - like a "Jennifer Aniston cell" (http://www.newscientist.com/article/dn7567). But also every neuron, every brain area, is immensely connected.

Modularity is reductionist jargon as it claims the brain to be an assembly of functional components. The language of hierarchies - where global constraints shape up local degrees of freedom - captures the reality that brains are both differentiating and integrating with equal vigour. A fact imprinted on their organisation as you might expect.

MarcoD

Maybe I am getting old, or he is, but I found it a very hard speech to follow. (Pretty sure he was jet-lagged.) The basic stand-off between behaviorists and cognitive scientists seems to boil down to the question whether all behavior is conditioning. From Chomsky's linguistical view, that also relates to the question whether language exists, it probably also bears relation to whether the question has meaning in either of two settings, where I think that behaviorists are probably inclined to deny the existence of language.

I am looking at it again, to try to decipher what he really said.

He rejects cognitive science studies, which take a holistic approach, as being unscientific. I am inclined to agree.

He states that there has been no linguistical evolution in humans for the last tens of millenia. I am inclined to disagree.

The combinatorial argument to speech I disagree with.

The existence of a Universal Grammar is probably right, though I am not sure what is meant with it.

I am not sure whether a UG is necessary, or reducible to discrete entities. Conversely, I am inclined to think that there must be an underlying principle to thought which can be abstracted from wetware neurological reasoning.

I have the feeling the brains is an organized mess, so explanations which are either pure neurological or modular are bound to fail.

A reduction of whether language exists to the existence of UG, in purely linguistical terms, boils down to semantics. The existence of language probably implies the existence of UG with a sufficiently broad definition of UG.

The minimalist program is 'scientific,' is the same as an endorsement, though I agree with it.

He tries to assign meaning to a meaningless sentence somewhere.

Perfectly good thoughts probably don't exist, so expressing them should be impossible too. He managed to express an inexpressible thought anyway.

Language is 'sound with meaning' vs 'meaning with sound.' Somewhat of an existential debate where one isn't sure what essence, or existence, is. I have little doubt that evolution doesn't care whether communication, or linearized internalized meaning, is more relevant since both add to your advantage.

The structural distance principle I find very interesting, and I agree a lot with him in that it says a lot about the computational process of thought. I am not intelligent enough to understand the merge combinators without projected examples.

The interaction between the emminent Chomsky and the next generation of scientists is hilarious.

apeiron

That's a good way to put it. Conditioning implies that every action has some specific cause - and that's all you need to talk about. Whereas cognitive approaches are interested in the general causes.

Conditioning tends to encourage blank slate thinking because it seems that any kind of behaviour could in principle be learnt. As a style of analysis, it does not consider global constraints, and so the possible variety of local behavours seems quite unconstrained.

This is like the now outdated "modern synthesis" Darwinian evolution where selection pressure could chip away at the genome to produce any kind of organism in principle. The organism responded to its environment, its context, in an atomistic fashion, one trait at a time.

But cognitive science, and the current evo-devo approach to evolution in general, accept the existence of global constraints on what is actually possible. There is a systemic relationship that limits (as well as guides) the kinds of bodies that can evolve, and the kinds of behaviours that can be learnt.

Well, of course, the problem with cognitive science is there are some like Chomsky who turn constraints into rules. They just take the global aspects of a system as something which exists in a Platonic fashion, not something that emerges due to an evolutionary/developmental process of local~global interaction.

Adding to the confusion, they then try to stick these rules inside some black-box component - a functional module. In the systems view, the constraints are constituted in a holistic fashion. They are the general architecture. But once you start trying to treat constraints as something separate, something that exists rather than emerges, then you have to find a location for them within the system. You have to stick them away in a private box which you claim is a rule-implementing device.

So no surprise that Chomsky is so murky as he tries to navigate this contorted view of what is going on (while the Behaviourists seem by contrast, childishly simplistic).

The modelling needs to focus on how particulars become generals, and generals in turn shape the particulars. As for example with generative neural network approaches to modelling the mind. Or with hierarchical accounts of the brain that are in fact now the norm in cognitive neuroscience.

So there are three camps. Those who believe that the particulars of a system explain everything. Those who believe the generals explain everything. And those who believe that systems are about the interaction of the particular with the general. We learn from experience, and we experience by applying the lessons that have been learnt.

Supporters of Chomsky and Skinner will of course point out how their heroes are always much misunderstood because really they were arguing for this third way.

MarcoD

Ah, I was still writing while you posted. I don't know a lot about Skinner vs cognitive science.

When it comes to conditioning, I am in the cognitive science camp. There are many examples in human life where thought processes overrule what is otherwise 'conditioned' behavior. Going on a diet after Christmas, going cold turkey after a drug addiction, and so forth. I guess you could reduce that to conditioning too, but I would like to see those arguments first.

Of course both sides have an argument: We know people calm down when you play Bach, hence we experiment with that in criminal environments -or play music in stores,- but we also know that people can be smarter than their animal behavior, so we appeal to that too.

I am not sure you don't read too much into Chomsky's 'Platonic' approach. Chomsky seems to postulate that a Universal Grammar should exist, but I am not sure how much he thinks of it as a thing which is really universal (as in that even alien intelligences should develop it), Platonic, or an emergent property of human mind/body(tongue) interaction. He sure did point out that linearization is a necessary byproduct of our communication organ, whereas structural distance is a necessary product of our neurological organ; that is not Platonic. (Though I think he tried to state that language was 'born' perfect, which I would disagree with.)

Chomsky is a linguist, and a hard-core one too. (Which I find a bit fifties, though I tend to agree with him.) He's probably only seen as murky by psychologists. Conversely, I am not sure in present-day psychology he can, or should, have an opinion. Though his old opinions are certainly worth study.

Then again, I sometimes find psychology is on the same level of academic worthiness as free-time studies. But that's a personal thing.

Retrospectively, also, I don't find Chomsky's ideas very murky. He just postulates, and overemphasizes, the role of language.

Nice, guess everybody is in the third camp then. From the little I now know about it, I know I am. Thanks for the explanation.

bohm2

The major difference is still the debate between empiricism versus rationalism/nativism. The rest is "window-dressing", I think. Everyone agrees that all behavior is caused by the interaction of a genetically-derived structure with its environment but nativists like Chomsky believe that everybody else gives way too much importance to "external stimulation," to environmental cues, and too little to the genetically-derived "internal structure of the organism, the ways in which it processes input information and organizes its own behavior." So for him, the internal structure is not some sort of amorphous blob ready to be molded by its environment, but an organism adapted to exploit that environment in its own unique way, and this fundamental principle applies to the mind-brain (CNS) as well as to all other organismal structures. And both should be treated and studied similarly.

apeiron

Yes, as Newport points out, his tune is changing. He is now in fact endorsing Lieberman's story of 40 years ago.

Before, he was saying it was all about a module/organ that could handle hierarchical syntax. Now he has moved closer to the idea that the whole brain is hierarchically organised, and the suddenly new thing in evolutionary history was the serial constraint placed upon that hierarchical organisation by a vocal tract.

So in saying this, Chomsky has finally come around to agreeing with existing thought in paleolinguistics.

But of course, to be consistent with his long history of scorn for this hypothesis, he has to say that the novel constraints created by the vocal tract are "peripheral". So really, we only have to pay attention to the hierarchical organisation of the brain.

Yet the whole point is that the constraints are indeed "on the periphery". They would have to be to be able to constraint the functioning of the brain in a radical new way. You have to stand outside what you seek to control.

So Chomsky is still seeking to downplay the significance of "computational linearisation" when it is the whole point really. It is what actually arose as the difference in evolutionary terms. The structural efficiency of hierarchical organisation is an important fact too - but it is an important fact about the brain in general, not the language function in particular. It was not the evolutionary novelty whose social and genetic history we must trace.

apeiron

Yes, but even when you get into genome-level learning of the species, as opposed to the developmental-level learning of the individual, you still have the same dichotomy to sort out.

Are you/Chomsky saying that this genomic learning is empirical or rational? Is it the bit by bit, trait by trait, construction of something due to particular experiences, or is there instead some kind of global organisational constraint that acts as a Piagetian structural attractor?

As Chomsky says, there is a computational efficiency argument when it comes to hierarchical organisation. But that is not a lesson that can be learnt empirically. Or can it in fact? Well, it certainly is a form that must emerge because all less efficient organisations get weeded out.

So strong rationalism (of the true Platonic kind) says the existence of these kinds of fundamental truths - the efficiency of hierarchical organisation - exist "somewhere" that is external to the systems they constrain. Which is what makes them so mysterious. It is the old debate about the nature of maths - is it Platonically existent or merely socially constructed?

But a systems view of constraints is that they emerge - reliably. When things self-organise, they fall into predictable and "logical" arrangements. And some kind of "least mean path" principle is at the heart of all our physical laws. When symmetry breaks, it follows the most efficient available course.

So a balanced approach treats empiricism vs rationalism as a false dichotomy. That is, we don't have to make a choice that sees one as wrong, the other is right. At every level of the story, both exist. And at every level of the story, both are in interaction.

By nativist, you simply mean genome-level learning/adaptation. And evo-devo spells out how that is an interaction of the "empirical" and the "rational". Or rather, that selection can only tune the parameters of self-organising limit cycles.

I guess you could call the selection "empirical" - the outward experiencing - and the self-organisation "rational" - the inward knowing. But these particular terms do start to seem rather strained.

MarcoD

Though I know little of the subject, I am inclined to disagree with either of you. For communication, it is likely necessary to somehow linearize thought. At the same time, for 'higher thinking' processes, rationality, it is also likely necessary that linearization of thought is necessary, as modelling the world linguistically probably also gives an evolutionary advantage.

To be honest, as stated elsewhere, evolution doesn't stop. Never. So it is highly likely that both processes occurred in tandem, along with a lot of other processes. Maybe the Neanderthal became extinct since he, or she, was less capable of communication as of thought due to organic limitations of the tongue and brain. And probably, even today, people who lack mental or speech capabilities die earlier with respect to better equipped individuals.

My point: Evolution dictates that both the thinking and the communication organ developed in tandem, and keep on evolving, so the question of what came first seems rather moot. Or rather, it's the combination of both which does the 'trick.'

apeiron

But are you talking now about biological or cultural evolution? Genetics or memetics?

The socio-cultural evolution of human thinking could only really begin with the invention of language. And indeed, the story is one of exponential change. That's plain enough from the archaeological record.

Mr.Samsa

But the point is that they do. How else would there be such a massive amount of crossover between cognitive and behavioral researchers? Surely if they don't study them in the same way, it would be impossible or difficult for them to collaborate as often as they do.

Yes, that's called "science". When you are developing a new theory, you need to demonstrate that it can account for the range of data we currently have.

I don't understand how you are so confused over this. Why would learning theorists come up with architectural models of the brain? This is like suggesting that the sociologists are "blank slatists" because they haven't contributed anything new to genetics. Science doesn't just reduce down to lower levels of explanation like that, and you can't expect a particular field to make predictions about an area it's not related to.

Learning theorists are focused on what learning mechanisms produce behaviors, and language in this case. The behaviorist research here resulted in breakthroughs like long-term potentiation and in-vitro reinforcement, which are vital for any process that requires an element of learning - this is the kind of thing that it would add to neurolinguistics. It won't add knowledge of modular concepts within the brain because that's not what they're studying.

Who would hold that position? Behaviorism is predicated on the understanding that not all behavior is conditioned, and in fact the entire field would collapse if this were true. This is because learning begins with the biological systems that underpin (to put it simply) our concept of pleasure and pain. We also have innate senses of things that are inherently 'pleasurable', like eating food, having sex etc, and these form primary reinforcers or unconditioned stimuli.

Such a position would require us to ignore numerous breakthroughs in innate behaviors which have come about through behaviorist research, like the discovery of the Garcia effect (the finding that we are naturally predisposed to learning a taste-sickness association to avoid poisoning ourselves), "preparedness" (the finding that we are naturally predisposed to learning some associations like fear of snakes more readily than others), tool-use and culture in New Caledonian crows, and even the matching law, which underpins our major theories of choice, is argued to be an innate feature of organisms.

As Skinner argued, behavior is necessarily a combination of environment and genetics/biology (he also included "culture", but I think that essentially falls into environment). No behaviorist would argue that all behavior is conditioning as such a position would be ridiculous and unworkable.

I'm not sure what this means. If behaviorists deny the existence of language, then why do they spend so much time studying it?..

Conditioning doesn't imply that, unless you interpret conditioning as simple stimulus-response psychology (which, as discussed above, nobody holds to anymore since the behaviorists disproved their position). Our behavior, especially in the messy real world, is largely driven by the general context, reinforcement histories, genetic predispositions, etc.

It's not like behaviorists argue that, say, Action X is caused by Stimulus Y.

What? Who the hell would suggest such a thing?

The approach of learning theorists is to test what behaviors could be learnt. I know it's a subtle distinction, but it's an important one. For example, if someone suggests that grammar is an innate aspect of humans, then this suggestion needs to be tested. A good way to test it is to find an exception to it, i.e. finding our black swan. So we take an organism which has no evolved 'grammar module' according to the nativist researcher, and we see if we can manipulate the environmental variables so that the organism can pick up grammar - as Herbranson did with pigeons.

This isn't to say that anything can be learnt, or even that we're trying to demonstrate that, but we're simply testing hypotheses put forward by other researchers who claim that they can't be learnt. Look at tool-use in the New Caledonian crow, where evolutionary psychologists argued that it must be innate. We separated chicks at birth, and put them in a situation where they could build tools to catch their food. What happened was that they still attempted to build the tools (by carving out notches in a pandanus leaf), but the creation was sloppy and the physics behind the tool was often wrong (i.e. the "hooks" of the leaf were on the wrong side so it couldn't hook on to a grub). From this we could conclude that there appears to be an innate preference for modifying leafs and resources in their environment, but that this behavior requires an element of learning to perfect and successfully use. And this finding is then strengthened by the finding that there were essentially different isolated "cultures" of crows across the island that had generated different designs over generations.

So no. No behaviorist ever has, currently does, or ever will believe that any behavior (even in theory) could be learnt. And as mentioned above, a blank slate approach to behaviorism would cause all behaviorist theory to collapse in on itself. It just could not function under the assumption that blank slatism was true.

Come on.. you're just taking the piss now, right?

A fundamental approach to behavioral research is the understanding of biological constraints on the organism - in a highly simplistic way, this is why rats aren't taught to fly using tree bark as a reinforcer. To argue that it ignores such constraints is like attacking optimal foraging theory for "assuming" that behavior is always aimed at being optimal, when the point of the analysis is to theorise what a 'perfect' behavior would look like and see how the actual behavior deviates. This discrepancy leads to clues as to what is causing a particular behavior. Look at the work of the Brelands that discovered "instinctive drift", where the finding was that 'natural' behaviors will sometimes come to the surface - this was a constraint on a form of learning, and it's something that needs to be taken into account when studying behavior.

Nice poisoning of the well there, but tell me: if someone explicitly states that their entire philosophy is dependent on a combination of both genetics/biology and environment, then how are we to interpret this? That they're lying and really they're blank slatists, despite the fact that they released several books specifically to refute the claim that they are blank slatists?

As for the implication that Skinner is my "hero", this is of course ridiculous. He was a scientist with some good ideas, and some horribly flawed ones (e.g. his views on punishment, cognitive science, etc). His view is mostly historical now of course, and he's only relevant to this discussion because Chomsky was attempting to address his arguments. Defending Skinner against the charge of blank slatism is no different from defending Darwin from a charge of being a creationist due to his apparent claims of intelligent design when we quotemine his discussion of the complexity of the eye. That is, I defend him because the charges are so ridiculously wrong, that they don't deserve to be in a forum that is supposedly filled with intelligent and scientific-minded people. As an individual, he was an interesting person who kickstarted a hugely important and influential field, but his ideas are largely outdated and have been replaced now. The field has moved on to quantification and prediction, understanding context and constraints, and looking at what behaviors are learnt and which are innate.

MarcoD

I was pointing out that human rationality developed as a byproduct of language, sentence manipulation, and that manipulation is the result of the interaction between the communication device and the thinking device. Since it is unlikely that nature professionally designed one of them before the other, both devices interact, and the combination of both is what really gives an evolutionary advantage, we can assume that early humans had lousy working proto-tongues and proto-brains, subsequently evolved to what we have now, and will evolve further into better talking and better thinking individuals, hopefully.

So I would say the genetics and memetics work hand in hand. Better brain/tongue (genes), better -more elaborate- thoughts (memetics), better survival.