Extended Minds Without Vehicle Externalism?

[Post by Neil Levy]

Brains blogger Ken Aizawa is well-known for (among other things) his opposition to the extended mind thesis. To refresh your memories, the extended mind thesis is the view, associated especially with Andy Clark (but also subscribed to by fellow-blogger Rob Wilson, as well as a growing number of others), that the mind is not confined within the skull, but spills out into the world. Clark and Dave Chalmers originally defended this view in a paper in Analysis, in which they introduced a parity principle: if something plays a role in cognition such that, were it internal to the skull we would classify it as part of the mind, it would be merely internalist prejudice if we fail to classify it as part of the mind merely because it is not internal to the skull.

Now, the extended mind thesis is often taken to be equivalent to vehicle externalism (the term is, I think, Hurley’s). At least one of Ken’s central arguments (in work with Fred Adams), the argument from the mark of the mental, is I think best seen as an argument aimed at vehicle externalism. The argument, roughly, is that anything that is genuinely mental must have intrinsic content, and external vehicles cannot have such content. But need an interesting externalism be a vehicle externalism? Mightn’t there be a way to get an interesting extended mind that is consistent with the claim that mental states are intrinsic content involving, in the following (strong) sense: that all mental states have their representational content in mentalese?

These thoughts are spurred by reading Peter Carruthers’ new book, The Architecture of the Mind (actually just chapter four, available on his website; the book itself hasn’t made it down under yet). Carruthers argues that sentences in natural language play an essential role in integrating the content of different conceptual modules, and this integrating function explains the flexibility, creativity and power of human thought. But he doesn’t claim that we actually think in natural languages. Representations in mentalese do all the real cognitive work. Human beings, he thinks, rehearse natural language utterances to themselves, thereby making them globally available to conceptual modules (via a process of retranslation into mentalese). The important point, for him, is that a mentalese utterance with the content p could only come about via the content-integrating mechanism of natural language; without natural language, the atomic components of p could not have been combined into a single thought. The thought has intrinsic content. But this is a content that it could only have via its route through natural language, which doesn’t have intrinsic content.

Two questions: First, might this serve as a model for how external resources can contribute to the power of human thought? Might external models and representations in formats different to natural language play a content integrating role, thus allowing us to entertain thoughts – in mentalese – with a content otherwise unavailable to us?

And second: Does this count as an extended mind thesis? On this view, my computer (say) with its graphic representations of data, might play an analogous role in enabling my thought processes to natural language. It would be a component of my thinking. Shouldn’t this count as an interesting externalism?

19 Comments

  1. Rob Wilson

    Answers: yes and yes. In fact, I think that this is the kind of take on the extended mind that is what makes the extended mind thesis an interesting, empirically fruitful hypothesis about where cognitive systems begin and end.

    But although I also haven’t read Carruthers’ book (but I guess Edouard and I will both be doing so some time in the next 6 months, since we’ll be serving as critics in an author-meets-critics session on it at the APA Pacific in April), I don’t think this is how HE would view his own work. If that’s all it takes for a mind to be extended, then the EM thesis marks no real departure from business as usual in cognitive science.

    On vehicle externalism: I’m pretty sure this does derive from Hurley’s 1998 book and/or several articles she had out around the same time. (But surely “vehicle externalism” itself hearkens back to Dawkins’ view of organisms as the “vehicles” of selection (where it is replicators that are in the driver’s seat). Although the metaphor is evocative in both contexts, I’ve always thought that it was limiting and in many ways misleading. Vis-a-vis natural selection, organisms aren’t simply receptacles for genes, and vis-a-vis the EM thesis the (external) vehicles of cognition aren’t simply bits of the world that are storage devices. The vehicle talk is what makes folks like Ken and Fred look for ways in which “things” outside the head can’t serve as such vehicles, or are asymmetrically parasitic on “internal vehicles” of cognition. I think that Clark and Chalmers bought into this way of thinking in their initial arguments, and that a healthier way of thinking here lies in a focus on the *activity* of cognition, and what it takes to realize particular cognitive activities. This is why I would say that the interpretation of Carruthers’ own work given above sounds like it advances thinking about the extended mind.

  2. Ken

    Neil,

    Thanks for this nice post. It raises many of the issues that separate Fred Adams and me (A&A) from the EC folks.

    Just let me note first some things about our view.

    First, A&A think that cognitive processes involve intrinsic (or non-derived) content, as you say. We don’t think that external vehicles cannot have non-derived content. It is merely that as a matter of contingent empirical fact in the cases EC folks propose, they do not.

    Second, we think, for example, computing sums in the head is a cogntively different process from computing sums using pencil and paper. It is because using pencil and paper involves fewer demands on memory (among other things) that leads people to use pencil and paper. Cognitive processes do things with representations that non-cognitive processes do not.

    I mention this since this second conditions, it seems to me, is somewhat underplayed in the (emerging) literature.

    So, that is all quick and dirty.

  3. Ken

    Now to the two questions.

    1. First, might this serve as a model for how external resources can contribute to the power of human thought?

    Yes.
    It seems to me that extended cognition folks want to say these things and they are fine. Concede that using pencil and paper to compute sums enables you to compute sums you could not compute otherwise. Concede that natural language use enables mental states otherwise impossible. Fine. A&A can live with that.

    2. Does this count as an extended mind thesis?
    Yes. But, the other issue EC folks care about is where cognitive processes are. EC folks often also want to say that cognitive processes are not all in the head. A&A think this rare, if ever, happens in the real world, as opposed to thought experiments.

    One of our most basic observations is that just because the pencil and paper help you compute sums does not mean that cognitive processes extend from the brain into the pencil and paper. Manipulating pencil and paper can be a “powerful” process without being a cognitive process. That’s the short of it.

    So, using natural language can be a “powerful” process without thereby being a cognitive process.

    So, EC folks can have this weaker claim, that external objects and processes are useful for cognitive processes. Who would doubt that? So, in anwer to “Shouldn’t this count as an interesting externalism?” A&A say no.

    Gotta run ….

  4. mike bruno


    Mightn’t there be a way to get an interesting extended mind that is consistent with the claim that mental states are intrinsic content involving, in the following (strong) sense: that all mental states have their representational content in mentalese?

    That certainly seems like a live possibility. Andy Clark doesn’t like that view because he also rejects LOT, I think Kim Sterelny holds something in the vicinity, and Chris Maloney defended a similar view a number of years ago.

    Another possibility which grants that the mental should be demarcated in terms of a derived/underived content distinction would be to connect underived content to phenomenal consciousness. To do this, one would first need to characterize some important aspects of phenomenal consciousness in terms of interactions between organisms and environments. The interactions must be constitutive elements of those aspects of phenomenal consciousness. Perhaps this could be done in a broadly representationalist framework. Then, one would need to show how those same aspects of phenomenal consciousness are necessary for underived content.

    Re: Question 1

    Perhaps, but active externalists will probably claim that positing mentalese in this case is explanatorily superfluous.

    Re: Question 2

    An interesting externalism, yes. But I don’t think it should count as active externalism. Clark and Chalmers distinguish their view from passive forms of externalism, which includes most well-known versions of content externalism. Active externalism requires that some types of mental states a subject may be in constitutively depend on interactions between the subject’s brain and anything outside the head (an alternative version would be in terms of bodies, not brains). It seems to me that the Carruthers-inspired externalism fails to be sufficiently interactive to count as active externalism.

  5. kenneth aizawa

    Incidently, Fred and I have speculated between ourselves that having a combinatorial syntax and semantics is part of the mark of the cognitive.  But, we figured that this would just catch us too much flack and not really strengthen our case defense of orthodox cognition-in-the-head or sharpen our critique of the case for extended cognition.

    Note that Andy at least at times thinks that phenomenally conscious processes are in the head.  It’s in the C&C paper, too, I believe.

  6. kenneth aizawa

    Rob,

    Do you see any difference between questions about where cognitive systems begin and end and where cognitive processes begin and end?

    For example, in many cases, X processes occur only in a limited region of an X system.  Air conditioning processes occur only within a given region of an air conditioning system.  Computing processes occur only within a given region of a computing system.

  7. Random note: This reminds me a little of the notion of an “exocortex” in Charles Stross’s sci-fi novel Accelerando, downloadable here:

    https://www.accelerando.org/book/

    Basically, the main character (and most everyone else in the novel) is constantly interfaced to a swarm of intelligent agents, which are used to keep track of information and take care of tasks. It eventually gets to the point where most of what makes the main character -himself- is embodied within his exocortex. At one point in the novel his exocortex is stolen, and he’s left a wandering zombie without any idea of who he is or what he’s doing.

  8. Neil

    Thanks for all the replies (and thanks to Eddy for posting the entry for me). A few responses.

    Mike Bruno argues that the version of externalism suggested isn’t active enough. He writes: Active externalism requires that some types of mental states a subject may be in constitutively depend on interactions between the subject’s brain and anything outside the head.

    I don’t see why the process isn’t active in this sense. On Carruthers’ view, “a particular tokening of an inner sentence is (sometimes) an inseparable part of the mental episode that carries the content of the thought-token in question. So there is often no neural or mental event at the time that can exist distinct from that sentence […] which carries the content in question”

    It seems to me that on this view the subject’s mental states do constitutively depend on interactions between the subject’s brain and natural language. And if natural language, why not other formal (and indeed informal) systems? Carruthers predicts that the thinking of native users of Sign will involve “the manipulation of visual (or kinaesthetic) images”, and adds that some ordinary thinkers may sometimes employ visual images, including written languages, in their language-based thinking.

  9. Neil

    Now esponses to Ken. First, it occurs to me that if we can get active externalism without vehicle externalism, we also solve the second major objection (from A&A and also Rupert) to the extended mind: the argument from causal individuation. Since internalists think that neural processes are causally individuated, if each external process causes a neural state or process, we can expect the lining up of mental states with causally individuated states.

    Second, Ken writes “we think, for example, computing sums in the head is a cogntively different process from computing sums using pencil and paper. It is because using pencil and paper involves fewer demands on memory (among other things) that leads people to use pencil and paper. Cognitive processes do things with representations that non-cognitive processes do not”

    Here’s what I say about this argument in my forthcoming book:

    to the extent to which the parity thesis is true, the external mind thesis loses its interest. It is precisely because the external scaffolding has different properties from the internal resources that it is worth extending – or embedding – our minds. Consider, for instance, Rupert’s (2004) claim that we ought to distinguish between the mind, on the one hand, and the set of external resources for thinking, on the other, on the basis that there is a set of empirical generalizations that characterize internal cognitive processes only, including characteristic ways in which they go wrong (e,g., interference effects typical of forgetting across all kinds of learning situations). If we obscure the difference between the internal and the external, he argues, we shall not be able to answer important questions, such as why these ‘learning curves, interference patterns, and so forth, are unavoidable when we rely on internal storage, while entirely optional […] in the case of external storage’ (2004: 418). Maybe so; maybe that is a reason to retain a science of the naked brain. But pointing to this difference between the internal and the external cannot be a refutation of an interesting externality thesis. Indeed, it is more like a confirmation of it. Why extend your mind if there’s nothing to gain by doing so? It is because extending our mind extends our cognitive powers that it is worth doing, and extending our cognitive powers requires that the extended resources have different properties to the unadorned brain. You can’t refute the extended mind hypothesis by pointing out that it’s interesting”

  10. Neil

    Finally,Ken’s claim that Clark style aliens with bit-mapped texts in their head are mere thought experiment, and A&A are instead interested in defending a contingent claim about human beings, here and now. Again, allow me to quote myself:

    “I wonder, though, whether they might not feel some pull toward Clark’s view of the case if they compared it to actual cases of mental recall here on Earth. Some people – mainly children, but some adults as well – have what is known as eidetic recall. In certain circumstances, and with varying degrees of reliability, they are able to ‘picture’, in their ‘mind’s eye’, a scene that is no longer before them. Of course, we are all capable of mental imagery, but eidetikers seem capable of something beyond most of us, in most circumstances: they are able to extract information from their mental image of which they were not previously aware. They can, for instance, do exactly what Clark’s imagined aliens do: see the individual letters of texts they have briefly scanned and later read them. One woman was even able to see the 3d image hidden in a stereogram, by committing one random dot pattern to memory, and then merging it with a second, visually presented, dot pattern to form the image (Bourtchouladze 2002: 110-1). Now, Adams and Aizawa do not claim that mental states are necessarily internal to the skull. Instead, they argue, as a matter of fact all actual mental states of human beings are internal (Adams and Aizawa, forthcoming). They therefore deny the relevance of Clark’s thought experiments: no matter how things stand with the alien residents of possible worlds, here on earth mental states have the features they claim. But eidetikers, here and now, seem to be capable of representations with the features that Clark imagines. Moreover, it seems to me that some of my representations of texts are closely analogous to the representations of Clark’s aliens. I know some songs and other texts ‘off by heart’, a legacy of school assemblies. If you ask me what these songs are about, I would have to rehearse them, aloud or mentally, in order to discover the answer. If that’s right, I have representations inside my head that do not (right now) have intrinsic content. If we should count these representations as mental, then even if I am only capable of thought at all because some of my representations have intrinsic content, and even if all my mental states must involve intrinsic content, such content is not ‘the mark of the mental’ at all – and nothing stands in the way of our treating states which include external representations and artefacts as mental states”

  11. kenneth aizawa

    Neil,

    Let me reply piecemeal.

    Now esponses to Ken. First, it occurs to me that if we can get active
    externalism without vehicle externalism, we also solve the second major
    objection (from A&A and also Rupert) to the extended mind: the
    argument from causal individuation. Since internalists think that
    neural processes are causally individuated, if each external process
    causes a neural state or process, we can expect the lining up of mental
    states with causally individuated states.

    Just to be clear, our idea is that, say, cognitive processes (and neural processes) are one kind of causal process and muscular processes are another kind of causal process.  The extended cognition hypothesis A&A care about is where cognitive processes begin and end, not where causal processes begin and end.  So, to establish some case of extended cognitive processing, one has to show that a cognitive process, not merely a causal process, extends into the body and, perhaps, environment.

    This bears on what comes next.

  12. kenneth aizawa

    Next,

    to the extent to which the parity thesis is true, the external mind
    thesis loses its interest. It is precisely because the external
    scaffolding has different properties from the internal resources that
    it is worth extending – or embedding – our minds.

    This is a line that Andy has taken (I forget in just which paper off the top of my head) and so has John Sutton.  Perhaps there are others.  I think Sutton calls this a complementarity argument.  In the A&A forthcoming book =), we critique the complementarity argument as well.

    We say that there are cognitive processes that occur in the brain and non-cognitive physical processes in, for example, the production of marks on the paper with the pencil.  A things stand, this complementarity does nothing to support the idea that cognitive processes extend from the brain into the hand, pencil, and paper.

    Take an example.  In an air conditioner, there is an evaporation coil and a compressor (among other things).  It is because of their complementary roles in evaporating and liquifying a refrigerant that they can work together (with other components) to have the power to cool the air in a house.  Still, this complementarity gives us no reason to think that the process of evaporation extends from the evaporation coil into the compressor or that the process of liquifaction extends from the compressor into the evaporation coil.

    It is because extending our mind extends our cognitive powers that it
    is worth doing, and extending our cognitive powers requires that the
    extended resources have different properties to the unadorned brain.
    You can’t refute the extended mind hypothesis by pointing out that it’s
    interesting”

    While A&A don’t propose to be “language police,” we think that this is a rather loose way of talking.  Fine for most purposes, but a potentially question begging description.  We think that humans have a relatively fixed set of cognitive powers (confined to the nervous system in typical cases), but that we can get more out of those relatively fixed cognitive powers by combining them in clever ways with non-cognitive processes.  We wouldn’t try to refute the hypothesis that cognitive processes extend into the body and environment by saying it is interesting.

  13. kenneth aizawa

    Neil,

    Our point about rejecting thought experiments was basically to note that it is not enough to imagine a situation in which there is a cognitive agent that does not meet the hypotheses of our theory of cognition.  That should be doable, since we mean to be offering a falsifiable theory.  To falsify our theory, you could find a cognitive agent that does not have cognitive processes that involved non-derived content.

    The eidetiker cases are very interesting and a nice step beyond what Clark does, but I am not sure what you take these cases to show.  Could you spell this out a bit?  Do you propose that these mental images have only derived content?  That’s what we think Clark wanted in his thought experiment.  Is the problem that, in these cases, we have cognitive processes that involve only derived representations?

    Thanks,
    Ken

  14. Thanks to Neil Levy and Rob Wilson for drawing attention to my book *The Architecture of the Mind*, and for raising the issue of its implications for the EC thesis. Let me say first that the book isn’t at all about the latter. (What it is about is developing and defending the most plausible form of massive mental modularity thesis, and showing how it can accommodate the creativity and flexibility of the human mind.) I myself feel that my position doesn’t lend support to an EC thesis, however. What I stress is the role of activated action schemata in generating images (either of heard sentences, or of transformations of visual or other images), which are then globally broadcast to a suite of central systems, which in turn activate yet further action schemata. These cycles of mental rehearsal are claimed to be constitutive of so-called “System 2” thought processes. And in the simplest cases (such as removing, in a “think aloud” protocol, the inhibitory signals that normally prevent action schemata from generating actual actions) widening the cycle to include external events (real utterances) adds nothing significant to it. From a cognitive perspective it is just as if the utterance had not been made aloud (with the exception of some peripheral differences, in motor cortex and early vision).

    Of course it is true in some cases that the external events enable cognitive sequences that couldn’t take place otherwise (as with doing a long division sum with pencil and paper). But I doubt that studying the entire process involving the pencil and paper is of any interest to cognitive science. All the hard work goes on in explaining how patterns of stimulation on the retina lead to percepts, which in turn lead to appropriate thoughts, which lead to activations of action schemata which then control a movement. What takes place in between (graphite marks appearing on the paper, and reflecting light in distinctive ways) is of no interest to cognitive science, even if it is true that without it, the entire sequence couldn’t keep going.

    Peter

  15. Neil

    Ken,

    I don’t want to say that there are cognitive processes that involve only derived content. I think we agree that cognitive processes must involve intrinsic content. However, given that there can be internal representations that do not have intrinsic content until they are manipulated in some way, the fact that some external representations also do not have intrinsic content until they are manipulated ought not to be a reason for treating these external representations different to internal. But I want to say that these internal representations are mental.

    Of course, you might deny that. And here I want to stress my agreement with one point you’ve (that is, A&A) have made repeatedly. We really need some kind of criteria for the mental (failing which we may simply be talking past each other). Of course, I don’t agree that intrinsic content or causal individuation are necessary conditions of the mental (though they may be sufficient).

  16. I intepret more the question you posed to Rob as a rethorical one rather than an unanswerable question to those defending an “extended view of mind” just because even for the most “neuro-fan”, today is not clear where boundaries are. Vision, is a cognitive process, begining from a single anatomical pathway, but lately affect the motor system (Bridgeman 2002),the so called: sensorimotor transformations or loops.

    This discovered fact is translated in contemporary theories of psychology in unexpected ways, where the boldest theories postulate a share circuitry or mechanism between perception and action suported by multiple empirical eveidence (the premotor theory of attention[Rizzolatti] among them, saying that there is close relationship between attention and oculomotor system).
    Vision and action are coupled.
    In relation to systems or whole organisms some concepts in evolutionary biology, such as the the extended phenotype of Dawkins (1982), are pointing in the direction of thinking about the organism (system) and the enviroment (world)in a more mutualistic relation.

    This means that within the “head” boundaries are blurred. Are boundaries between the organism and the world blurred?
    Is it possible to expand this notion and incorporate the world as platform for ignite or help more with cognitive procesess than ussualy thought? One thing is clear, without the brain the world has not sense, but without the world the brain is out of context.

  17. Neil

    Peter,

    Congratulations on a great book. As I said, I’ve only read ch. 4, which I got from your website, but I rushed to download it because I’ve been reading your M&L papers on modularity. I think the book is likely to prove the single most important contribution on the topic since Fodor.

    Doesn’t your work support, first, a *genetic* externalism, in the following sense: agents are only able to entertain thoughts with some contents (combining the outputs of various conceptual modules) because of internalising a resource that is initially external to the agent? And, second, mightn’t it support a more radical externalism by showing how other representational resources might be used to combine concepts that are not combinable in natural language – one element might be higher dimensional mathematics, or indeed primitive visual processing when represented in (say) Marr’s vocabulary?

    I noted above that you say that ordinary thinkers might already utilise images of natural language in their thinking. Isn’t there at least the possibility that they might use other representational systems?

  18. Hi guys. Typos are very important to all written form. It gives the reader something to look for so they aren’t distracted by the total lack of content in your writing. Help me! I find sites on the topic: Westminster wall clock. I found only this – large wall clock. Clocks the perfect addition to any shop, garage,, or man cave. Designated trademarks and brands are the property of their. Thank you very much :cool:. Radman from Kuwait.

  19. Nice, this place is full shit, make me oral sex.
    I am from Sudan and learning to write in English, give please true I wrote the following sentence: “A distinctive and functional accent, this unusual clock will enhance the look of your kitchen, hallway or living room.”

    With respect :-D, Melba.

Comments are closed.

Back to Top
%d bloggers like this: