The What And Where Of Mental States: On what is distinctive about the extended mind thesis

Karina Vold (McGill University)

[Jump to comments]

 

1. Introduction

Andy Clark and David Chalmers’ (1998) extended mind thesis provides an answer to the question ‘where is the mind?’ The thesis maintains that while minds may be centrally located in one’s brain (and body) they can sometimes “extend” to be located in objects beyond their core biological shells. Functionalism and the multiple realizability thesis are often used to support the extended mind, as is the case in Clark & Chalmers’s much discussed parity argument, but although familiar these are not uncontroversial views in the philosophy of mind.  In this paper I present an argument for the extended mind thesis that does not rely on either of these. The argument instead requires what I call the ‘multiple localizability thesis’, which says that particular kinds of mental states need not be ‘strictly’ or ‘uniquely’ located in any particular place, e.g. the brain or one of its regions. I argue that evidence of neuroplasticity shows that mental states are ‘multiply localizable’ and that this claim should be less contentious than multiple realizability, even if it is rarely stated explicitly. In advancing this new argument for the extended mind thesis I hope to clarify what I believe is distinctive about the view: that it gave us new insight about the location of mental types. Thus distinguishing between multiple realization and multiple localization helps illuminate that the extended mind is essentially a thesis about the location of mental states, while functionalism and multiple realizability are about their nature and composition, respectively. And because the matters of location and composition can come apart, an argument for the extended mind need not take any position on the composition of mental states and so the view can remain uncommitted as to whether mental states are multiply realizable or whether they can be characterized functionally. I end by considering four objections and offering replies.

2. Clark & Chalmers’ parity argument

I will briefly rehearse Clark & Chalmers’ parity argument for the extended mind thesis, which says that an object[1] located in an agent’s environment can count as partially constitutive of an agent’s mental states. To defend this view Clark & Chalmers reason: “If, as we confront some task, a part of the world functions as a process which, were it done in the head, we would have no hesitation in recognizing as part of the cognitive process, then that part of the world is (so we claim) part of the cognitive process.” (p. 8) Most now refer to this as the ‘parity principle’. Others sometimes call it the ‘fair treatment principle’ (e.g. Drayson 2010, p.  371), as it maintains that we should treat equivalent processes with “the parity they deserve”, irrespective of whether they are internal or external to the skull (Clark & Chalmers 1998, p. 6). The ‘parity argument’ for Clark & Chalmers’s thesis can be summarized as follows:

(P1)  What makes some information count as a belief is the role it plays, and there is no reason why the relevant role can only be played from inside the brain/body. (p. 14)

(P2)  Information stored in objects beyond the body can play the same role as information constituting an ordinary mental state (from inside the brain/body).

(C)  Therefore, objects located in an agent’s environment should sometimes count as partially constitutive of an agent’s mental state.

The first premise seems to express a commitment to a variety of functionalism—this will be discussed in the next section. To support the second premise, Clark & Chalmers describe several cases in which, they argue, an object in the environment does play the same role for an agent that neurons in the brain usually would (something we would surely count as a part of the mind). Their most well known example is that of Otto and Inga. Inga decides to go to an exhibition at the museum and to do so “She thinks for a moment and recalls that the museum is on 53rd Street, so she walks to 53rd Street and goes into the museum.” Meanwhile, we imagine that Otto suffers from Alzheimer’s disease and has to rely on information he stores in a notebook to help structure his life. When he decides to go to the same exhibition he consults his notebook, where he wrote the directions for how to get there, he then walks to the museum and heads inside. (Clark & Chalmers 1998, p. 11-2) Clark & Chalmers argue that in the relevant respects information in Otto’s notebook “functions just like” the information in Inga’s brain that constitutes an ordinary belief,[2] and thus both should count equally as part of the constitutive machinery of his mind.

To be clear, the extended mind thesis could be supported by other arguments besides this ‘parity argument’. I will give a new argument for the thesis myself in the final section of this paper. Others have already done so including, but not limited to: Chemero (2009), Clark (1996), Hurley (1998), Hutchins (1995), Menary (2007), Noe (2004), Palermos (2014), Rowlands (2010), Sutton (2007), Wheeler (2005), and Wilson (2004).  For obvious reasons I will not be able to survey all of these here, but in my assessment Clark & Chalmers’s parity argument has received the most attention in philosophical discussions (and perhaps more widely as well), furthermore my new argument is inspired by the parity argument. Thus, for these reasons I focus on this argument and version of the thesis.

3. The relationship between the extended mind thesis, functionalism, and multiple realizability

There is a tendency to think of the arguments for the Extended Mind thesis in general, and the parity argument in particular, as requiring functionalism. Wheeler (2010a), for example, agrees that, as he says, “The claim (extended cognition) is in some way a form of, dependent on, entailed by, or at least commonly played out in terms of, functionalism is now pretty much part of the received view of things” (p. 245; Drayson 2010 makes similar remarks). In this section I will address the perceived relationship between the extended mind and the parity argument, on the one hand, and functionalism and multiple realizability on the other.

First of all, in the philosophy of mind, functionalism is a view about the nature of mental states. It maintains that what makes something a mental state of a particular type is the functional role that it plays in the system that it is a part of, rather than, say, any of its physical properties (or the properties of its constitutive base). The multiple realizability thesis, following Putnam (1967) and Fodor (1974), states that higher level properties, in this case mental properties, possess multiple lower-level (in this case, physical) realizers or instantiations. If the physical realizers are “drastically heterogeneous”, then no bi-conditional between higher-level mental types and lower-level physical types can be established, and in these cases we say the higher-level type is ‘multiply realizable’. (Lyre 2009)

Now, what is the relationship between functionalism and multiple realizability? Because the functional role abstracts away from reference to physical properties (at least for so-called ‘role functionalists’) particular types of mental states could turn out to have multiple types of lower-level physical instantiations (or realizers). In other words, functionalism can nicely accommodate the possibility of multiple realizations of mental types. But, it could also turn out that each type of mental state has only one type of lower-level physical instantiation, just as a matter of contingent fact. In this sense functionalism could be true even when multiple realization (of mental types) is contingently false (or, more strongly, multiple realizability might be false given the laws of physics). Furthermore, it seems clear that multiple realizations of mental types might be possible even without functionalism (Wheeler 2010b agrees, as does Drayson 2010, and Churchland 2005). This is because multiple realizability is a thesis about the composition of mental states: in a sense, it says that the composition of higher-level mental types is not fixed, so that there is not a one-to-one mapping between higher-level mental types and the lower-level physical types that compose them; whereas functionalism is a view about the nature of mental states (more discussion in Drayson 2010, p. 376).

We are now in a position to return to the main question: what is the relationship between the extended mind and the parity argument, on the one hand, and functionalism and multiple realizability on the other? Clark & Chalmers (1998) do not explicitly present their argument as relying on a functionalist view about the nature of the mind. But Chalmers (2008) says that their argument for the extended mind thesis requires nothing more than the “very weak functionalism captured by the parity principle.” (p. xv) And Clark (2008b) goes on to describe his view as ‘extended functionalism’, and seems to favor ‘common-sense functionalism’. (Chalmers 2008) Most commenters have agreed. Rupert (2004) argues the parity argument for the extended mind “contains a clear functionalist strain”. Gertler (2007) argues that the parity principle and (something like) premise one of the parity argument (as presented in the previous section) commits one to role functionalism. Shapiro (2008) says that support for the extended mind “often rests on functionalist theory of minds.” (p. 7) Weiskopf (2008) argues “The argument for (the extended mind) thesis rests on a simple, orthodox functionalist principle. (p. 266) (More discussion of some of these views in Drayson 2010, p. 369) Rowlands (2010) says, “The most important assumption underlying the thesis of the extended mind seems to be functionalism about mental states and processes” (p. 99) And, according to Rowlands, “This commitment to functionalism… is not peculiar to Clark and Chalmers’s development of the extended mind… most arguments for the extended mind are functionalist ones”. (p. 100)

Some commenters further argue that functionalism entails the extended mind thesis. Sprevak (2009) argues that the extended mind thesis is entailed by functionalism, such that either the extended mind thesis is true or functionalism is false. Weiskopf (2008) seems to agree (p. 267). Sprevak furthermore says that “the most plausible justification” of the extended mind thesis is “the functionalist argument” (p. 527). In contrast to Sprevak, Wadham (2016) argues that the (common-sense) functionalism that the extended mind relies on actually implies that the thesis is false.  If Wadham is right then the extended mind thesis will need to be supported without functionalism, or at least would have to appeal to a different version of functionalism (some possibilities discussed in Drayson 2010 and Farkas 2012).

While Wheeler (2010b) seems to agree that the common opinion is that the parity argument requires functionalism, he suggests that one could advance an argument for the extended mind thesis with just multiple realizability. (p. 8) Drayson (2010) agrees with Sprevak (2009) that versions of functionalism might entail the extended mind, but also agrees with Wheeler that the parity argument need not rely on functionalism. She, thus, disagrees with Chalmers (2008) analysis and instead argues that the parity argument does not need even a ‘very weak’ functionalism. This is because one could reject functionalism and still argue for extended cognition[3] using the parity principle. Drayson’s diagnosis for why there is a general tendency to think that the extended mind requires functionalism is as follows: “The principle requires that the same cognitive processes can be realized in more than one way, and functionalism is an obvious way to accommodate multiple realizability, so it is natural to interpret the “parity principle” as a functionalist assumption.” (p. 376). Thus, Wheeler and Drayson seem to agree that the parity principle (and parity argument) only requires multiple realizability, not functionalism.

In what follows, I develop the extended mind thesis using a parity-style argument that does not require functionalism or even multiple realizability. I think this is important because, although familiar, neither functionalism nor multiple realizability are uncontroversial views in the philosophy of mind. Functionalism, for example, faces a number of long-standing objections, e.g. Searle’s (1980) Chinese room argument and Block’s absent qualia (1978). And I go a step further than what Wheeler and Drayson suggest by also developing the argument without multiple realizability because multiple realizability, like functionalism, is also contentious. Some, for example, Shagrir (1998), Bechtel & Mundale (1999), Shapiro (2004), and Polger (2009), express skepticism about the evidence of brain plasticity that is often cited in support of the multiple realization of mental types, e.g. by Block & Fodor (1972), and, as a more recent example, in Weiskopf & Adams (2015). Furthermore, I think there is a general tendency to think that in order for the mind to ‘exit’ the skull one would have to make use of a different type of realizer, and therefore that the extended mind theorist has to rely on some variety of multiple realization of mental state types. But I think this rests on a misunderstanding of the central insight of the extended mind thesis. It is worth noting that there are already some non-functionalist arguments for the extended mind thesis. For example, so-called “coupling arguments”, e.g. Clark 2008, do not seem to require functionalism. Sutton’s (2007) argument also may not require functionalism as it depends on the “complementarity” of inner and outer processes rather than functional equivalence. Rowlands (2010) also develops a non-functionalist version of the extended mind.[4], [5]

4. Against ‘strict’ (or ‘unique’) localizationism: the multiple localizability thesis

In neuroscience, cerebral functional localizationism is the view that each mental function depends on a particular region of the brain. Nowadays this tradition is familiar and even commonplace but it in fact had a tumultuous history and was long resisted.[6] Strict localizationism ascribes to the idea that there is a one-to-one mapping of cognitive function (or mental types) to brain location and vice versa. So each region of the brain, however finely-grained, is responsible for a particular type of higher-level mental or cognitive function and each higher-level mental or cognitive function is brought about by lower-level activity in a particular location in the brain. Strict localizationism seems to presuppose that there is cerebral anatomical specializationism–that certain regions of the brain are typically responsible (or better suited) for bringing about certain mental activities–but the opposite is not true. This is because anatomical specialization does preclude the possibility of (what contemporary cognitive neuroscientists call) ‘degeneracy’, a one-to-many mapping between higher-level cognitive functions and lower-level neural processes, or ‘pluripotency’, a many-to-one mapping.[7]

It is unclear how widespread a belief in strict localizationism is today. But I argue that various kinds of empirical evidence suggest the view is false and that certain mental functions are instead ‘multiply localizable’ rather than strictly localizable.[8] That is, a given mental type need not always be realized by physical structures located in the same place, for example within a particular region of the brain. Instead, a given mental type can be realized by physical structures located in different places. Some evidence in support of this is the observation that the lateralization of linguistic functions differs between right and left-handed subjects (Polger 2009, p. 462, Shapiro 2004 p. 59). So the same functions are located in different hemispheres in right and left-handers.

But more compelling support for the ‘multiple localization’ of mental functions comes from evidence of brain plasticity that indicates the ‘movement’ of function. For example, normally linguistic function is located in the left hemisphere of right-handers, but it has long been known that this can change: suffering a traumatic injury early in life can cause linguistic functions to relocate and be established in a different brain region (see Chugani et al. 1996, also discussed in Block & Fodor 1972, pp. 160-161). This kind of neuroplasticity, known as ‘homologous area adaptation’ (Grafman & Litvan 1999, Grafman 2000), shows that mental functions (or ‘mental types’ if one wants to avoid talk of ‘functions’) can shift locations and are, in this sense, not strictly localizable, but rather are ‘multiply localizable’. Here is an argument from homologous area adaptation for the possibility of the multiple localization of mental function

(P1) A subject, s, has a linguistic function, f, which is typically be realized by a neural structure located in a particular brain region, r1.

(P2) But s suffers an injury and r1 is left damaged.

(P3) Because of this injury, f shifts and is eventually realized by a neural structure located in a homologous region in the opposite hemisphere, call it r2.

(C) :. f is a linguistic function that could be realized by neural structures located in either region r1  or r2.

So as long as linguistic functions are mental functions (presumably they are[9]), then this kind of brain plasticity shows that at least some mental functions are multiply localizable, and thus not strictly localizable.

More evidence of the multiple localizability of mental states comes from another kind of plasticity known as ‘cross-modal reassignment’ (Grafman & Litvan 1999, Grafman 2000). Sharma et al. (2000) and von Melchner et al. (2000) rewired the visual systems of ferrets to project into the area that would be the location of the auditory cortex in non-manipulated ferrets. The “rewired ferrets” are able to process some visual stimuli by means of structures located in what is typically the auditory cortex. So their visual functions are able to shift locations and in this sense are ‘multiply localizable’. Philosophers sometimes cite rewired ferrets as evidence of multiple realizability, e.g. Weiskopf & Adams (2015).  Shapiro (2004), however, argues that this is not a case of multiple realizability because the rewiring in the auditory cortex ends up not being a distinct kind from the wiring that was lost in the visual cortex. It is the same kind of wiring—the same neural structure—only in a new place. (pp. 61-64) This is evidenced by the fact that the auditory cortex of the rewired ferrets “develops the columns of orientation-sensitive cells” (p. 64) that are typical of the visual cortex (further discussion in Polger 2009, p. 468). So Shapiro (2004) argues that rewired ferrets do not provide conclusive evidence of the multiple realizability of visual function. Whether or not this is right, it is clear that the visual functions in question are being instantiated in a spatially distinct region of the brain, and thus that they are multiply localizable.[10] This shows how multiple localizability could be true even if multiple realizability is not. Here the crucial distinction between the two is that in the first case the same function is ‘moving’ from one set of neurons to another set of the same kind of neurons (or neural activity) as opposed to a function moving from one set of neurons to another set of a different kind of neurons (or a different kind of neural activity).[11]

5. How the matters of localization and realization (or composition) come apart

Notice that even a type-type identity theorist, who opposes multiple realizability, can accept the multiple localizability thesis: indeed, as long as one is willing to give up strict localizationism, one can accept multiple localizability. If all pain states are numerically identical to the physical states of c-fibres firing, for example, then pain must be located wherever c-fibres are firing. Notice that there is nothing that prevents c-fibres from being the sort of thing that could fire in the right way (under the right conditions) in numerous places, thereby being compatible with multiple localizability. Suppose we discover that c-fibres are only located in the left hemisphere of left-handed people and only located in the right hemisphere of right-handed people. Or suppose we discover that c-fibres are located in our feet, not just in our brain. For an identity theorist these differences in location, though perhaps interesting, would be completely irrelevant to the claim that pain is identical to c-fibres firing. Thus, pain could be located in different regions of the brain (or elsewhere), without being multiply realized.[12] A further point here is that there need not be any commitment to internalism, the view that the mind is located within the skull, built in to identity theory, or any commitment about the location of lower-level types at all.[13]

This shows how the matter of location is orthogonal to the issue of composition, or realization: localization concerns where things are placed in space, while composition concerns what things are made of (or realized by). We cannot infer from a claim about the composition of a thing (or specifically, a part within a system) where that thing (part) is located. Likewise, we cannot infer from the precise location of a thing what that thing is composed of. Thus, taking a position on one of these issues does not entail a position on the other and, in this sense, the two issues can come apart and be handled separately. Multiple localization and multiple realization ‘come apart’ in that they are not extensionally equivalent: one can be true in a world where the other is false.

In a paper that challenges arguments for multiple realizability, Shagrir (1998) argues “psychological states and processes could be realized in different areas of the brain without being multiply realized.” (p. 448) He thus recognizes how the multiple localization of mental function is a distinct issue from multiple realization, and that multiple localization does not support multiple realization. Meanwhile, Aizawa (2009) aims to challenge the claim that studies of brain mapping somehow undermine the hypothesis that mental states are multiply realized (his specific target is Bechtel & Mundale 1999). Aizawa argues that strict localizationism (what he calls ‘unique localizationism’), if true, would not logically entail that multiple realization is false. He further agrees that the matters of localization and realization ‘come apart’ in the relevant sense,

But what does unique or multiple localization of a psychological function have to      do with unique or multiple realization? Very roughly speaking, unique localization means something like always occurring in the same place, unique realization means something like always constructed in the same manner. They are entirely separate distinctions. (p. 501)

Thus Aizawa argues that whether strict localization or multiple localization is true, the multiple realizability thesis could still be true (or false) (p. 502) Aizawa brings out this distinction with the example of automobile engines. In the actual world, automobile engines are multiple realized and multiply localized: automobile engines are multiply realized in that some “use fuel injection where others use carburetors, some are water cooled where others are air cooled, some have variable valve timing where others do not”, and they are multiple localized as “some cars have engines in the front and some have them in the back.”  But we can imagine a world where multiply realized engines might all be found only in the front of the car, “giving rise to the unique localization of multiply realized engines.” (quotes from p. 501) Or we could equally imagine a world where all engines are of the same kind, that is, built the same way and of the same materials, and yet these engines are sometimes located in the front of the car and sometimes in the back. And finally, we could imagine a world in which automobile engines are both uniquely localized and uniquely realized. It is because the location of the engine within the automobile is orthogonal to the issue of its composition that we can imagine these possible worlds. Aizawa and I agree that the same can be said of mental functions.[14]

6. On the location of mental types

I maintain that the extended mind thesis is primarily a view about the location of mental states. It is intended to provide an answer to the question ‘where is the mind’? And the response it offers is that the location of mental types is not fixed to any one place, for example, within the skull or to a particular region of the brain. There is not a definite or fixed boundary of the mind, either at the skull, skin, or elsewhere. This is why it is possible for objects located in an agent’s environment to subserve, constitutively, an agent’s mental state. For example, Inga realizes the very same belief (that the museum is located on 53rd street) internally that Otto and his notebook realize externally. Thus the example requires that the same belief can be located in two different places and so the belief cannot be strictly localizable. The same is true of the other examples that Clark & Chalmers (1998) discuss, e.g. their Tetris example involves the same cognitive process to take place in distinct locations and in one case the process just happens to take place outside the head. And it seems most cases of cognition extension (or at least those that would be supported by Clark & Chalmers’ parity argument[15]), will similarly require that mental states are multiply localizable.

Multiple realization is about the composition of mental states, and so, because the matters of location and composition come apart, the extended mind does not need to commit to any particular position about the composition of mental states and thus does not need to be committed to the possibility of multiple realization. As functionalism is a view about the nature of mental states, the extended mind also does not require a commitment to this: the view could be largely if not entirely neutral, and need not be committed to any particular position on the nature of mental states (seems Drayson 2010 and Farkas 2012 both agree).

Some might argue, alternatively, that the extended mind thesis should be understood as primarily a claim about the composition of mental states, and not a claim about location. On this interpretation, the important claim is that cognition can be realized not just in the brain, but by various instruments, like notebooks, which just happen to be outside the brain. Rowlands (2010), for example, argues that the thesis is primarily about composition, not location:[16] “(E)xtension is a spatial concept and, so, is closely tied to that of location. And the issue of the location of cognitive processes can easily sidetrack us into concerns that we should not have.” (p. 83) Rowlands states at several points that “what is of primary importance to this thesis is the issue of composition, not location.” (p. 84)[17]

One problem we encounter if we adopt Rowlands’ ‘composition interpretation’ is that the extended mind thesis does not appear to have contributed any major new insight to contemporary discussions: it would just be a consequence of multiple realizability and functionalism, ideas that have been on the table for decades in philosophy of mind. The important insight that multiple realizability gave us was that not just the brain can realize mental types, but other kinds of lower-level stuff, for example, silicon chips might as well.  It is not clear that the extended mind puts forth any novel insight into either the composition or the nature of mental states. This is precisely why, thinking that extended mind is just an implication of functionalism, Shapiro & Spaulding (2009) question the novelty of the extended mind thesis,

If all that matters to minds are computational processes, or if minds just are particular sorts of computational processes, one must begin to wonder about the novelty of (Clark’s view). Functionalists, as will be known by those who remember Ned Block’s example of a nation that realizes a mind, would have no difficulty accepting the possibility that minds extend beyond heads. Is Clark simply recounting in greater detail how this might be true?[18]

Thus the composition interpretation cannot account for what is distinctive about the extended mind thesis. On this interpretation it seems the thesis has offered no new insight. I think Clark & Chalmers’s thesis did offer novel insight, only it was insight about the location of mental types. On my interpretation, the important claim is that cognition can be realized not just in the brain, but also in one’s use of various instruments, like notebooks and devices, which are located outside the brain.  Thus, the ‘composition interpretation’ might struggle to account for the amount of attention and discussion that Clark and Chalmers’ view generated. Meanwhile, the ‘location interpretation’ does a better job of explaining what is new about the extended mind thesis. Now it is true that in an important sense the insight the extended mind gives us is that location doesn’t matter. But this is still important insight. It is on par with what multiple realizability tells us about composition: that, in an important sense, composition doesn’t matter.

Wheeler (2010a), for example, seems to agree with my ‘location interpretation’ when he says that extended cognition “is a view about the whereabouts of thinking and thoughts” (p. 2). On Wheeler’s (2010a) model, “parity is conceived not as parity with the inner simpliciter, but rather as parity with the inner with respect to a locationally uncommitted account of the cognitive.” (p. 8)[19] Parity demands that we be ‘locationally uncommitted’, and not specifically ‘compositionally uncommitted’, because to do otherwise would beg the question at hand—where is the mind? Presumably, this is because the key issue for the extended mind theorist is just this: an issue about the location of the mind.[20]

I also think the ‘location interpretation’ is Clark’s own view. Interestingly, Farkas (2012) agrees that Clark endorses the ‘location interpretation’ (what she might call the ‘realizer’ interpretation) but she nonetheless argues that his interpretation is wrong and that “Clark was misrepresenting the essence of his position” (p. 439) Farkas rejects the ‘location interpretation’ on the grounds that it would not be philosophically significant because it does not conflict with any current philosophical view (p. 438) (although she does not consider internalist orthodoxy or so-called ‘brain/skull chauvinism’, or she does not consider them ‘philosophical views’). Farkas reasons there must be another interpretation of the extended mind thesis that can account for why some “hail it as a significant insight” (p. 439). She does not adopt the ‘composition interpretation’ but rather a third option: the extended mind thesis “has to do with how we conceive the nature of (mental) states”. (p. 435) Farkas argues the “literal spatial extension” of the realizer of the functional role is not the central insight of the extended mind thesis (the ‘location interpretation’), instead the thesis is really about metaphorically ‘extending’ the relevant functional role: “we extend what counts as a functional role or a dispositional profile that qualifies a state to be a certain kind of standing state…this is the real lesson of the Otto-Inga case.” (p. 441) I think this is wrong for several reasons, for example, I agree that the extended mind does not conflict with any of the views that Farkas canvasses but I think this is precisely because it is about location and a view on the location of the mental states, as we have seen, is a separate issue from that of their composition or nature. But for now it is enough to note that Farkas and I agree on how Clark interprets his own thesis, only I think that he interprets it correctly.[21]

7. A new argument for the extended mind thesis

Consider the ‘silicon brain chip’ thought experiment (Pylyshyn 1980). Imagine that neurons in your brain are deteriorating and affecting your mental capacities (e.g. your memory), so doctors begin to replace some of your natural neurons with silicon chips. These silicon-chips are programmed to perform all the same functions of the neurons they are replacing. They work as “artificial neurons”. We can imagine that in an extreme case this replacement process continues until no part of your biological brain remains. Philosophers have different views about what will happen to your mental life while this goes on (more discussion in Chalmers 1996). Role functionalists would say that so long as the chips really do perform all of the same functions as the neurons they replace then your mental life would not be impacted. If role functionalism is right then simply changing the physical realizers should not matter, as long as they are responsible for performing the relevant kind of functional role. The example of the silicon brain chips requires that mental states are multiply realizable. But it does not require that they are multiply localizable, since (we can imagine) the silicon chips are replacing neurons where they are within the skull.

Now imagine that instead of entering your skull to replace your deteriorating neurons where they are, doctors consider it less intrusive to the biological shell and, therefore, preferable to do as much as they can externally. So doctors decide it will be safer to attach your remaining well-functioning neurons through tiny electrical nodes to an implant that threads a wire out your ear. They then attach this wire to an external device that contains your programed silicon chips and which attaches to your person. We can imagine that after the operation, you come to with a small wire now reaching out of one of your ears. Attached to this wire is a small device that hugs your ear, much like a hearing aid. Doctors call it an ‘iCog’.[22] Of course the iCog is not currently available and may never be, but they are conceivable and both logically and metaphysically possibly. Furthermore, iCogs would be objects in an agent’s environment—beyond her biological brain-and-body—that function just like the objects constituting her ordinary mental states (in this case, the neurons in her brain). Thus, iCogs are possible cases of an extended mind.

The silicon-based iCog requires a commitment to both the multiple localizability thesis and the multiple realizability thesis. But one could imagine that the iCog is itself made of neurons. Either artificially made or naturally occurring biological neurons could be connected through one’s ear by a long axon, also naturally occurring. In this case the iCog is just a (protected) clump of neurons attached to one’s ear, either by an axon or by a string of more neurons, and so multiple realizability is not required, since all mental states could be made of the same kind of neural states. All that is required is multiple localizability. But notice that the neuron-based iCog is still an example of the extended mind thesis. It is an example of a mental state that is partially constituted by an object located beyond the skull. And it doesn’t require functionalism or multiple realizability.[23] We can use this thought experiment to articulate a new argument for the extended mind thesis, which can be summarized as follows:

(P1) (At least some) mental types are multiply localizable.

(P2) The skull does not represent a theoretically significant boundary.

(P3) An object that surely counts as (at least partially) constitutive of a mental state (e.g. a neural structure in the brain) could just as well be located in the agent’s environment (e.g. in the iCog).

(C)  Therefore, objects in an agent’s environment can count as partially constitutive of an agent’s mental state.

The first premise was supported in section four with neuroscientific evidence of brain plasticity. The second premise is a claim that Clark & Chalmers (1998) articulate; to deny this outright would beg the question at hand—‘where is the mind?’—unless one can give an independent reason for thinking the skull is theoretically significant (further discussion in Wheeler 2010a). The third premise is supported by the example of the iCog.  This premise also relies on the intuition that neurons in the brain surely count as (at least partially) constitutive of (some) mental states (I take this to be widely accepted by philosophers of mind these days). None of these premises commit one to functionalism or multiple realizability. Indeed even an identity theorist could support the argument and conclusion.

8. Objections and replies

I will consider several possible objections to the argument presented in the previous section and to my understanding of the extended mind thesis as primarily a claim about location.[24]

i. First objection:

The ‘multiple localization’ argument still relies on functionalism or, at the very least, on functionalist intuitions.  The story of the iCog began with the silicon brain chip thought experiment that involves the replacement of functionally equivalent parts. Then using the intuition that this would preserve mental state types the scenario was ‘weakened’ so that that the realizers are (arguably[25]) of the same type. The argument for the preservation of mental state type in the iCog case, however, still relies on accepting functionalism as a premise. So even if we accept that the final neuron-based version of the case does not involve multiple realization, the reason for thinking that there is anything mental there at all must still be that there is no functional difference in realizer.

Reply: The multiple localization argument and the iCog example are perfectly compatible with functionalism, but neither require functionalism. It is true that the story I tell to motivate the iCog case begins with Pylyshyn’s thought-experiment intended to support functionalism, but an identity theorist could tell their own version of the same story; so too could a connectionist; and so on. For the identity theorist, the reason for thinking that there is some mental function being partially realized by the iCog is not that there is no functional difference in realizer, but rather because the realizers are of the same physical types, and mental states are identical to these physical types. So the identity theorist would see no difference between the c-fibres in one’s brain and the c-fibres in one’s iCog, as long as both sets of fibres can fire in the appropriate ways and circumstances.

ii. Second objection:

iCogs might expand the brain, but they do not extend the mind and, therefore, the iCog thought-experiment does not support the third premise. One reason they only expand and do not extend is that there continues to be a physical continuity between the internal brain and the iCog, whereas there is no such physical continuity between Otto and his notebook. For this reason, one might think that an internalist on holiday could be quite happy if one could make the iCog work–since it is just the brain doing what it does, moved out the skull. In contrast, for the externalist, this scenario just falls short of the brain in the vat.[26]

Reply: iCogs are objects that are not located within the boundaries of the skin. They could be made in a petri dish and never get inserted into the organism, so there would not always have been a physical continuity. Thus, iCogs are objects in the environment external to the organism that become appropriately coupled with the agent’s previously existing internal resources. In a sense they do expand the brain, just as a Universal Serial Bus (USB) key ‘expands’ the hardware, or the physical realizer, of a computer. But we do not say that USB keys are located ‘inside’ the computer, even when they are plugged into it. Thus, iCogs do support the third premise because they are examples of objects beyond the natural boundary of the organism that, were they to be located in the head of the agent, would surely count as constitutive of the agent’s mental states.

In fact, Clark & Chalmers (1998) discuss some similar example themselves,

In the distant future we may be able to plug various modules into our brain to help us out: a module for short-term memory when we need it, for example. When a module is plugged in, the processes involving it are just as cognitive as if they had been there all along. (p. 11)

This suggests that they would consider a plug-in module, something like an iCog, a genuine case of cognitive extension. Consider as well, Clark (2009):

An external silicon circuit is added that restores the previous functionality. (The agent) can now divide just as before, some small part of the work is distributed across the brain and the (external) silicon-circuit: a genuinely mental process (division) is supported by a hybrid bio-technological system. That alone, if you accept it, establishes the key principle of Supersizing the Mind.

The iCog is just this, only the neuron-based iCog is entirely biotechnology, it contains no silicon. This too should establish the key principle of the extended mind thesis. Another way to bring out the fact that the iCog is an extension not just a way of expanding the brain is to simply move it further away from the head. In his Foreword to Clark’s Supersizing the Mind, Chalmers (2008) has an example just like this: “one could imagine that some of the neural correlates of consciousness are replaced by a module on one’s belt.” (p. xiv) So we might wear the relevant external module on our belts rather than around our ears. (Maybe we even clip our iCogs on to our belts like we do with audio players and the axons are strung through protected wires out of our ears and plugged into the module in a way that ends up looking much like ear phones do.) The point is just that this would count, at least according to Chalmers’ (2008), as a case of cognitive extension.

iii. Third objection:

The localization argument ends in a modal conclusion: that objects in an agent’s environment could count as partially constitutive of an agents’ mental state.  iCogs are not actual possibilities in the world today and may never be. So the multiple localization argument does not show that cognitive extension is actually happening or that it ever will happen. At best it shows that it could happen. But, even critics of the extended mind would accept that, such as Adams & Aizawa (2008) who label themselves ‘contingent intracranialists’.

Reply: The multiple localization argument shows how it is principle possible for the extended mind thesis to be true even if functionalism and multiple realizability turn out to be false. The argument shows how the view itself does not rest on or require ether functionalism or multiple realizability. The iCog can be understood as a heuristic device intended to clarify what is distinctive about the extended mind given that it is conceptually distinct from functionalism and multiple realizability. It is meant to support the third premise of the argument that says that an object that surely counts as constitutive of a mental state type could just as well be located in the agent’s environment. All that is needed to this end is the modal claim that iCogs could one day instantiate an agent’s mental states. Whether or not iCogs will ever be realities is not important. The argument shows why and how that the mind could extend without multiple realizability or functionalism.

iv. Fourth Objection

The extended mind thesis is interesting because it claims that external objects in the most robust sense (e.g. notebooks, cell phones, or other people) can be partially constitutive of mental states. The multiple localization argument only supports a weak and uninteresting variety of the extended mind. The argument cannot support any of the interesting examples of cognitive extension that Clark & Chalmers give. Thus, even if it is granted that the neuron-based iCog is an example of the extended mind thesis, this is only in the strictest sense of the term—there are mental states that can be partially constituted by an ‘object’ that is located ‘beyond the skull’—but this example is not interesting, and so it cannot explain what is so interesting about the extended mind thesis.

ReplyIt is true that the iCog example is compositionally conservative, in that it is a case of the extended mind that preserves ‘unique realization’, when compared to more compositional extravagant cases such as Otto and the notebook. The result is that there can be a spectrum of cases: the neuron-based iCog only requires multiple localizability, but without multiple realizability the thesis is conservative when compared with the more extravagant range of cases that Clark & Chalmers’ functionalist parity argument supports. So one must accept both multiple localizability and multiple realizability for Otto’s notebook to count as a case of extension.

The goal of stripping the extended mind thesis of functionalism and multiple realization has been to show what is really distinctive about the thesis. Some might assume that in order for the mind to ‘exit’ the skull one would have to make use of a different type of realizer, and therefore that one has to have a commitment to some variety of multiple realization. I would venture to say that most of the arguments for the extended mind make use of the possibility of multiple realization.[28] But iCogs show how we could preserve ‘unique realization’ (of physical, lower-level kind) and still move outside of the skull. Thus the example pulls apart two issues that were thought to go hand in hand. And in doing so it shows what is distinctive about the extended mind thesis: it is, in essence, a thesis that has given us original insight about the location of mental states. For this reason the extended mind thesis may not be so controversial after all, once we understand that many of the interesting cases (and much of the controversy) fall out of the commitment to multiple realizability (and often the claim about functional equivalence, as in the second premise of the parity argument, as discussed earlier. But the extended mind thesis itself is still interesting, as it shifts focus off of the issues of the composition or nature of mental types and on to the issue of their location.[29]

 

 

Works cited

Acharya, S., S. Shukla, S.N. Mahjan, & S.J. Diwan. Localizationism to Neuroplasticity—The Evolution of Metaphysical Neuroscience. The Journal of the Associations of Physicians of India, 60 (2012): 38-46.

Adams, F. & K. Aizawa. The Bounds of Cognition. (Blackwell Pub., 2008).

Adams, F. & D. Weiskopf. An Introduction to the Philosophy of Psychology. (Cambridge University Press: 2015).

Aizawa, K. Neuroscience and Multiple Realization: A Reply to Bechtel and Mundale, Synthese 167 (2009): 493–510.

Bechtel, W., & J. Mundale. Multiple Realizability Revisited: Linking Cognitive and Neural States. Philosophy of Science 66.2 (Jun., 1999): 175-207.

Bechtel, W., & R. C. Richardson. Discovering Complexity. Decomposition and Localisation as Strategies in Scientific Research. (MIT Press, 1993).

Block, N. Troubles with Functionalism. Minnesota Studies in the Philosophy of Science 9 (1978): 261-325.

Block, N. & J.A. Fodor. What Psychological States Are Not. Philosophical Review 81 (1972): 159-81.

Broca, P. Sur le principe des localisations cérébrales. Bulletin de la Société d’Anthropologie 2 (1861): 190–204.

Chalmers, D. The Conscious Mind. (Oxford University Press, 1996).

Chalmers, D. Foreword to Andy Clark’s Supersizing the Mind. In Clark, A. Supersizing the Mind: Embodiment, Action, and Cognitive Extension. (2008): pp. ix-xvi.

Chemero, A. Radical embodied cognitive science. (MIT Press, 2009).

Chugani, H.T., R. A. Muller, D. C. Chugani. Functional brain reorganization in children. Brain Development 18.5 (1996): 347-56.

Churchland, P.M. “Functionalism at Forty.” Journal of Philosophy, 102 (2005): 33-50.

Clark, A.  Being There. (MIT Press, 1996).

Clark, A. Pressing the flesh: exploring a tension in the study of the embodied, embedded mind. Philosophy and Phenomenological Research. 76.1 (2008a): 37-59.

Clark, A. Supersizing the Mind: Embodiment, Action, and Cognitive Extension. (New York: Oxford University Press, 2008b).

Clark, A. Letters. London Book of Reviews 31.6 (2009) Available online.

Clark, A. & D. Chalmers. The extended mind. Analysis 58 (1998): 7-19.

Clarke, E. & L. S. Jacyna. Nineteenth-Century Origins of Neuroscientific Concepts. (University of California Press, 1987).

Craver, C. Explaining the Brain. (Oxford: Oxford University Press, 2007).

Descartes, R. Meditations on First Philosophy. translated by J. Cottingham (Cambridge: Cambridge University Press, 1996).

Drayson, Z. Extended Cognition and the Metaphysics of Mind. Cognitive Systems Research 11.4 (2010): 367-77.

Farkas, K. Two Versions of the Extended Mind Thesis. Philosophia 40.3 (2012): 435-47.

Fodor, J.A. Special Sciences, or the Disunity of Science as a Working Hypothesis. Synthese 28 (1974): 97-115.

Friston, K. J., & C. J. Price. Degeneracy and redundancy in cognitive anatomy. Trends in Cognitive Science 7 (2003): 151–2.

Grafman, J. Conceptualizing functional neuroplasticity. Journal of Communication Disorders 33.4 (2000): 345-56.

Grafman, J. & I. Litvan. “Evidence for Four Forms of Neuroplasticity.” In Neuronal Plasticity: Building a Bridge from the Laboratory to the Clinic. J. Grafman and Y. Christen (Eds.). (Springer-Verlag Lublishers, 1999).

Hutchins, E. Cognition in the Wild. (MIT Press, 1995).

Hurley, S.  Vehicles, Contents, Conceptual Structure and Externalism. Analysis 58 (1998): 1-6.

Jarvis, B. Evaluating the Extended Mind. Philosophical Issues 24.1 (2014): 209-229.

Kaplan D., & W. Bechtel. Dynamical Models: An Alternative or Complement to Mechanistic Explanations? Topics in Cognitive Science 3 (2011): 438-44.

Kaplan, D. & C. Craver. The explanatory force of dynamical and mathematical models in neuroscience: a mechanistic perspective. Philosophy of Science 78.4 (2011): 601-28.

Kim, K.H., N. R. Relkin, K. M. Lee, & J. Hirsh. Distinct cortical areas associated with native and second languages. Nature 388 (1997): 171-74.

Kiverstein, J., & M. Miller. The Embodied Brain: Towards a Radical Embodied Cognitive Neuroscience. Frontiers in Human Neuroscience, 9.237 (2015): 1–11.

Levin, H.S., J. Scheller, T. Rickard, J. Grafman, K. Martinkowski, M. Winslow, & S. Mirvis. Dyscalculia and dyslexia after right hemisphere injury in infancy. Archives of Neurology, 53 (1996): 88–96.

Lyre, H. “The “multirealization” of multiple realizability” A. Hieke and H. Leitgeb (eds.): Reduction and Elimination in Philosophy and the Sciences (Proceedings of the 31st Wittgenstein-Symposium). (March 2009). Ontos verlag, Frankfurt. [Preprint] URL: http://philsci-archive.pitt.edu/id/eprint/4494 (accessed 2015-07-20).

von Melchner, L., S. L. Pallas, & M. Sur. Visual behavior induced by retinal projections directed to the auditory pathway. Nature 404 (2000): 871-5.

Menary, R. (2007). Cognitive Integration: Mind and Cognition Unbounded. Palgrave Macmillan.

Menary, R. (ed.) (2010). The Extended Mind. MIT Press.

Mundale, J. Concepts of Localization: Balkanization in the Brain. Brain and Mind 3 (2002): 313-30.

Noë, A. Action in Perception. Cambridge, MA. (MIT Press, 2004).

Overgaard, M. & J. Mogensen. A Framework for the Study of Multiple Realizations: the Importance of Levels of Analysis. Frontiers in Psychology 2.39 (2011): 1-10.

Palermos, S.O. (2014). Loops, Constitution and Cognitive Extension. Cognitive Systems Research 27: 25-41.

Polger, T.W. Evaluating the Evidence for Multiple Realization. Synthese 167 (2009): 457-72.

Price, C. J., & K. J. Friston.  Degeneracy and cognitive anatomy. Trends in Cognitive Sciences 6.10 (2002): 416-21.

Price, C. J., & K. J. Friston,  Functional ontologies for cognition: The systematic definition of structure and function. Cognitive Neuropsychology 22.3 (2005): 262–75.

Putnam, H. Minds and Machines. In Dimensions of Mind. Ed. S. Hook, (New York University Press, 1960).

Putnam, H. The Nature of Mental States. (Originally published as “Psychological Predicates”). In Art, Mind, and Religion Eds. W. H. Capitan and D. D. Merrill. (University of Pittsburgh Press, 1967).

Pylyshyn, Z.W. Computation and cognition: issues in the foundations of cognitive science. The Behavioral and Brain Sciences 3 (1980): 111-69.

Rowlands, M. The New Science of the Mind. (MIT Press, 2010).

Rupert, R. Challenges to the hypothesis of extended cognition. Journal of Philosophy 101.8 (2004): 389-428.

Rupert, R. D. Systems, functions, and intrinsic natures: On Adams and Aizawa’s The Bounds of Cognition.  Philosophical Psychology 23.1 (2010): 113–123.

Shagrir, O. Multiple realization, computation and the taxonomy of psychological states. Synthese 114 (1998): 445-461.

Shapiro, L. The Mind Incarnate. (MIT Press, 2004).

Shapiro, L. Functionalism and the boundaries of the mind. Cognitive Systems Research 9.5 (2008): 5-14.

Shapiro, L. Dynamics and Cognition. Minds & Machines (2013) 23: 353–375.

Shapiro, L. & S. Spaulding. Review of Andy Clark: Supersizing the Mind: Embodiment, Action, and Cognitive Extension. Notre Dame Philosophical Reviews. (2009). Available online at: https://ndpr.nd.edu/

Silberstein, M., & A. Chemero. Constraints on Localization and Decomposition as Explanatory Strategies in the Biological Sciences, Philosophy of Science, 80 (2013): 958-70.

Sprevak, M. Extended Cognition and Functionalism. Journal of Philosophy 106 (2009): 503-27.

Sutton, J. Batting, habit, and memory: The embodied mind and the nature of skill. In J. McKenna (Ed.), At the boundaries of cricket 10.5 (2007): pp. 763–786). Special issue of Sport in Society.

Theiner, G. Res cogitans extensa : a philosophical defense of the extended mind thesis. (Frankfurt, New York: P. Lang, Internationaler Verlag der Wissenschaften, 2011).

Vold, K. (2015). The Parity Argument for Extended Consciousness. Journal of Consciousness Studies 22.3-4, 16-33.

Wadham, J.  Common-Sense Functionalism and the Extended Mind. Philosophical Quarterly 66.262 (2016): 136-151.

Walker, A.E. The Genesis of Neuroscience. (The American Association of Neurological Surgeons and A.M. Walker, 1998).

Weiskopf, D. Patrolling the mind’s boundaries. Erkenntnis 68.2 (2008): 265-76.

Wernicke, C. Der aphasische Symptomencomplex. Eine psychologische Studie auf anatomischer Basis. (Breslau, M. Crohn und Weigert, 1874. (Title in English: The aphasic symptom complex: a psychological study from an anatomical basis.

Wheeler, M. Reconstructing the cognitive world. (Cambridge, MA: MIT Press, 2005).

Wheeler, M. In Defense of Extended Functionalism. In Menary, R. (Ed.). The extended mind. Cambridge: MIT Press. (2010a): 245-70.

Wheeler, M. Minds, Things, and Materiality. In L. Malafouris and C. Renfrew. (Eds.), The Cognitive Life of Things: Recasting the Boundaries of the Mind. Cambridge: McDonald Institute Monographs. (2010b) (Reprinted in J. Schulkin (ed.), Action, Perception and the Brain: Adaptation and Cephalic Expression. Basingstoke: Palgrave Macmillan.)

Wilson, R. A. Boundaries of the mind: The individual in the fragile sciences–cognition. (Cambridge, UK: Cambridge University Press, 2004).

 

 

Notes

[1] I use the term ‘object’ loosely here (and throughout). First of all, it is the information in the object that partially constitutes the mental state, not necessarily the entire object itself. Second, I remain open that ‘objects’ could include properties, e.g. a particular ordering of objects.

[2] To support this claim they outline several ways in which the information in the notebook plays a similar role for Otto as the internally stored information does for Inga: (1) information store in the notebook is a constant in Otto’s life, just as Inga’s brain is for her; (2) information in the notebook is highly accessible to Otto, as information store in Inga’s brain is (for the most part) easily accessible to her; (3) Otto relies on the information in the notebook and doesn’t hesitate to immediately endorse and make use of the information, as Inga relies on her own beliefs; and (4) the information was previously endorsed by Otto and is there a result of having been previously endorsed and then stored, just as is the case with the information Inga makes use of. (p. 17)

[3] Drayson (2010) is careful to distinguish between ‘extended cognition’ and ‘extended mind’. There may be good reason to want to make this distinction but I have not done so as I don’t believe anything I say here hinges on this distinction. So Drayson’s further analysis is that this tendency (to connect parity with functionalism) stems from the common reading of the parity principle as a functionalist claim about the nature of mental states rather than as a computationalist claim about the nature of cognitive algorithms.

[4] Rowlands (2010) cites two reasons for wanting to do so (p. 105): He believes that the “reliance on a liberal form of functionalism” leaves the extended mind thesis (1) “vulnerable to charges of question-begging, and (2) “incompatible with the thesis of the embodied mind, a thesis that can accept functionalism only in more chauvinistic forms.” This is a tension Clark (2008a) points out.

[5] Furthermore, there are versions of the extended mind that may not rely on the possibility of multiple realization, including the ‘continuous reciprocal causation’ argument (see, for example, Palermos 2014) as well as Menary’s (2007) cognitive integration approach. As we will see, this interpretation depends in part on how one understands ‘multiple realizability’, a phenomenon that is itself, in a sense, ‘multiply realizable’ (Lyre 2009).

[6] Some of the key figures and movements in establishing cerebral functional localizationism include Franz Joseph Gall and his theory of organology; Fritsch & Hitzig (1870) studies on canines; and Paul Broca’s (1861) and Carl Wernicke’s (1874) discoveries. (See discussions in Clarke & Jacyna 1987; Walker 1998; and Acharya et al. 2012).

[7] ‘Degeneracy’ must be distinguished from multiple localizability as no distinction is made between the issues of realization and location by many discussing degeneracy (e.g. Kiverstein & Miller 2015, Friston & Price 2003, Price & Friston 2002). This merits a lengthier discussion, which I will take up elsewhere.

[8] There are, of course, other options too. ‘Wholistic’ or  ‘global’ views that maintain that the whole brain contributes to mental functions, or that a collection of regions scattered around the brain might contribute to bringing about a mental function, e.g. distributed neural networks and dynamic systems theory may fall somewhere within these family of views. I believe at least some varieties of wholism are compatible with ‘multiple localizability’ and this relationship merits a lengthier discussion then I can give here.

[9] I use the example of a linguistic function here, I think most would count this as a ‘mental’ or ‘cognitive’ function. Grafman (2000) also discusses another case of homologous area adaptation involving a shift in ‘visuospatial skill’, studied by Levin et al. (1996).

[10] Overgaard & Mogensen (2011) have a related discussion of these issues and seem to agree with Shapiro’s analysis. They argue that this kind of ‘rewiring’ shows that ‘relocalization’ of mental function (either as a result of modified sensory input or from regional brain injury) is possible but that relocalization is not conclusive evidence of multiple realization precisely because it might be accompanied by a ‘rewiring’ of the local circuitry within the novel substrate that results in a neural circuitry resembling (or of the same kind as) the original, only in a new place. (p. 3)

[11] There is a ‘big’ question here about what exactly ‘the same kind’ of neurons (or neural structure, or neural activity), or a ‘difference in kind’, amounts to and what is relevant for deciding this—whether it is the kind of neurons present, the neural microstructure or class, the anatomy of the circuitry, a specific activity, or so on.  Whether or not these various forms of neuroplasticity are evidence of multiple realizability will depend on what counts as ‘sameness in kind’ or a ‘difference kind’ at the neurobiological level (for discussions see Shagrir 1998, Shapiro 2004, Polger 2009). My initial review of the literature suggests that there is not a consensus on how to understand ‘multiple realizability’ (Lyre 2009). This issue merits more discussion, so I will not settle this here and will instead stick to using what might fairly be described as toy examples, e.g. talking about ‘c-fibres firing’ versus ‘o-fibres firing’, or activity in a silicon-chip, as all different lower-level kinds without explaining what makes them ‘different in kind’.

[12] In this example, multiple realization would require that pain be identical to a different kind of neural activity, such that a creature without any c-fibres might still experience pain. Imagine a creature instead has o-fibres and their experience of pain is identical to o-fibres firing, then pain could be identical to either c-fibres firing or o-fibres firing, which we should understand as two distinct lower-level types, so this would show that multiple realization is possible (and type-type identity theory would be false). But, admittedly, this example avoids the ‘big’ question of what exactly the difference is between c-fibres and o-fibres that makes them count as different neurobiological types.

[13] For a mind-brain identity theorist, of course, the claim is that all mental states are identical to brain states and thus the view is internalist for this reason. But one would need to give an independent reason for embracing this internalist commitment, otherwise it cannot be assumed without begging the question against the extended mind theorist. So in this section I am discussing ‘psychoneural identity theory’ which says that the mental states types are identical to neural states types, and that mental states are thus located wherever the neurons that they are identical to are located (whether that is within the skull or not).

[14] Aizawa is a critic of the extended mind thesis (he calls himself a ‘contingent intracranialist’), so the claim that the issues of location and composition ‘come apart’ is, in this case, a point of agreement between an internalist and an externalist.

[15] Given the structure of the parity argument, this would make sense. But one might argue that not all versions of extended cognition, e.g. those defended by Rowlands (2010), Chemero (2009), or Menary (2007), require that mental types are multiply localizable. This might be true, but I believe all these theories offer their own insight into the location of mental states, e.g. that the boundary of mental states is not fixed.

[16] Rowlands seems to agree that the matters of location and composition ‘come apart’ (see pp. 83-4).

[17] Rowlands is actually discussing what he calls the ‘almagamated mind’ thesis, but this is meant subsume both the embodied and the extended mind theses (p. 84).

[18] Discussed in Shapiro 2011, p. 226; for a similar analysis see Weiskopf 2008, p. 267.

[19] I think there is further evidence that Wheeler (2010a) might support the ‘location interpretation’ on p. 8 and pp. 19-20.

[20] Similarly, while not explicitly endorsing it, there is some evidence that Menary (2010), Rupert (2010), Theiner (2011), and Jarvis (2014, p. 209), might endorse the ‘location interpretation’ as well.

[21] I have further discussion on Clark’s view and supporting quotes in section VIII.ii

[22] I first introduced ‘iCogs’ to make a different point regarding the possibility of a parity argument for extended consciousness, see Vold (2015).

[23] It should be clear that the iCog does not presuppose that any mental types (or functions) can be precisely located in the brain or that the brain is organized into separable modules. I think the iCog is just as compatible with distributed neural networks, for example.

[24] Thanks to audiences at the 2015 Fall Colloquium Series at the Institute for Philosophy II at Ruhr University, Bochum; at the 2016 Eastern Division meeting of the American Philosophical Association; and at the 2016 Science of Consciousness Conference in Tucson, AZ, as well as commenters and some anonymous reviewers for raising some of these concerns.

[25] I suppose there may be questions over the actually possibility of neuron-based iCogs without multiple realizability, e.g. as to whether natural neurons could live, or function appropriately, outside the skull.

[26] This exact phrasing comes from an objection raised by an anonymous reviewer, to whom I owe thanks.

[27] This objection was raised by Steven James (West Chester University of Pennsylvania), who I thank for very helpful comments at the 2016 Eastern Division Meeting of the American Philosophical Association.

[28] As mentioned, some notable exceptions might include the ‘continuous reciprocal causation’ argument (see, for example, Palermos 2014) as well as Menary’s (2007) cognitive integration approach.

[29] A predominant methodology in contemporary cognitive neuroscience is to give a mechanistic explanation of psychological functions where a mechanistic account involves two essential steps: localization and decomposition (Bechtel & Richardson 1993; Craver 2007; Silberstein & Chemero 2013). One might think that focus in the philosophy of mind has long been on the composition of psychological functions, while the extended mind thesis, and perhaps other situated views of cognition as well, have shifted focus off of the issue of the composition and on to the issue of their location.

4 thoughts on “The What And Where Of Mental States: On what is distinctive about the extended mind thesis”

  1. Vold offers us a new argument for the extended mind, one based on (what she calls) the multiple localizability of mental states, where a type of mental state is multiply localizable if and only if it can be instantiated by physical structures located in different places. The argument that Vold develops counts as a new argument for the extended mind because many of the most prominent arguments for that view either depend on, or in some way buy into, some sort of functionalism about, or at least the multiple realizability of, mental state types. According to Vold, multiple localizability is not sufficient for functionalism or multiple realizability. So, if these other theses are controversial (which of course they are, in some quarters), in a way that multiple localizabiity isn’t (Vold argues that it’s supported by the empirical evidence), then her new argument has the advantage. Moreover, Vold claims, given that functionalism and a commitment to multiple realizability are widely held (if sometimes controversial) positions, and given that the possibility of the extended mind is usually recognized to be a consequence of these positions, the claim that the extended mind turns on multiple localizability explains what’s distinctive about the hypothesis in a way that the foregrounding of these other properties doesn’t. In short, the extended mind thesis is primarily a view about the location of mental states, not a view about their nature (functionalism) or composition (multiple realizability).

    Much of Vold’s reasoning is persuasive. In what follows, however, I shall concentrate on what I take to be a residual problem. In effect, what I have to say is a version of the fourth objection that Vold herself considers, so my main point will be to explain why I think her response to that objection falls short.

    In part, Vold motivates the idea that multiple localizability does not entail multiple realizability by creatively redeploying a familiar example from the recent history of philosophy of mind. In her words: “If all pain states are numerically identical to the physical states of c-fibres firing, for example, then pain must be located wherever c-fibres are firing. Notice that there is nothing that prevents c-fibres from being the sort of thing that could fire in the right way (under the right conditions) in numerous places, thereby being compatible with multiple localizability.” This shows that multiple localizability is compatible with type-identity, and therefore can obtain without multiple realizability. Couple this to Vold’s iCog example, in which neural matter performing a particular psychologically relevant function is located outside the body and then hooked into the brain via wires (or even wireless) technology, and, assuming everything works, one has a case of extended cognition on the basis of multiple localizability, but without multiple realizability and thus, as things are standardly set up, without functionalism. Game over?

    Well, yes and no. The ‘no’ here is not a denial of the point that one can have extended cognition without multiple realizability. That strikes me as correct. But there is an obvious problem (highlighted by the fourth objection that Vold considers), namely that cases of neural tissue functioning cognitively while located outside of the body don’t look much like the sorts of cases that have been central both to the appeal of, and to the challenge posed by, the extended mind hypothesis, cases such as Otto’s notebook, wearable computing and so on, all of which involve artefacts. It certainly looks as if an argument for extended cognition in these cases – an argument for the claim that external artefacts may count as genuine parts of our cognitive architectures – will have to appeal to multiple realization, or at least they will do if anything like the parity principle is in play, since in such instances we are encouraged to describe two different material arrangements – one wholly neural, one spread out over brain, body and world – as two different physical instantiations of the very same psychological phenomenon. So, even if one can have extended cognition without multiple realization, purely on the basis of multiple localization, one cannot have cases of artefact-involving extended cognition on the basis of multiple localization alone. For that (the kind of extended cognition we really care about), one needs multiple localization plus multiple realization.

    Confronted with this worry, Vold replies as follows:

    [The iCog] example pulls apart two issues [localizability and realizability] that were thought to go hand in hand. And in doing so it shows what is distinctive about the extended mind thesis: it is, in essence, a thesis that has given us original insight about the location of mental states. For this reason the extended mind thesis may not be so controversial after all, once we understand that many of the interesting cases (and much of the controversy) fall out of the commitment to multiple realizability (and often the claim about functional equivalence, as in the second premise of the parity argument, as discussed earlier. But the extended mind thesis itself is still interesting, as it shifts focus off of the issues of the composition or nature of mental types and on to the issue of their location.

    So Vold agrees that the cases that have propelled the extended mind to fame (or notoriety) depend on multiple realizability, not just multiple localizability. However, if I understand her response correctly, her thought is that even where multiple realizability is necessary to secure extension, it’s the subsequent conclusion about the location of the mental states concerned that makes the view interesting and distinctive. This is true, I think – the extended mind is interesting and distinctive because it tells us that mental states are not always where we thought they were – but once one puts things like this, one wonders just what we are to make of Vold’s ‘new argument’. For even though multiple realizability is not necessary for multiple localizability, multiple realizability remains necessary for some – and, as Vold seems to concede, the interesting – cases of the extended mind. In the end Vold is, I think, right about the following: (i) the extended mind hypothesis is an interesting and distinctive view because of what it claims about the location of mental states; (ii) one can have extended cognition without multiple realizability; (iii) this can happen on the basis of multiple localizability. All of these things are well-worth saying, and Vold says them clearly and persuasively. This is already a genuine contribution to the literature. But, as an argument for the extended mind, Vold’s new argument, based as it is wholly on multiple localizability, doesn’t take us as far as we would like. Getting us the rest of the way depends on the very notion – multiple realizability – that Vold wishes to banish to the conceptual periphery. Indeed, it’s that very notion that delivers the interesting and distinctive result about the spatial location of mental states, in the cases we care most about.

  2. Even after 18 years, Clark and Chalmers’ argument for the Extended Mind thesis continues to generate debate. As with many other influential arguments, the focus of discussion has shifted from disputes about whether the argument succeeds to debates about how to interpret the argument and, relatedly, about its implications. Vold’s insightful paper addresses some of these latter questions.

    My plan in these brief comments is to discuss the philosophical significance of Vold’s central claims, and to register a worry about how her proposed interpretation of the Extended Mind thesis fits with her characterization of the iCog argument. Reading Vold’s paper prompted me to think about another aspect of the relation between realization and location; I end these comments with a few speculative remarks on this issue.

    Vold undertakes two different kinds of projects here. The first kind of project is interpretive. She advances a specific interpretation of the Extended Mind thesis, and argues that this interpretation is preferable to its primary competitor. The second kind of project Vold pursues here is more purely philosophical. Vold argues that multiple realizability and multiple localizability are orthogonal claims, in that neither entails (or requires) the other. Partly in support of this claim, she develops an argument for the Extended Mind thesis that explicitly avoids commitment to multiple realizability (or functionalism).

    I begin with the interpretive issue. Vold argues that the Extended Mind thesis should be construed as the claim that mentality extends beyond the brain, in a spatial sense. This is the “location interpretation” of the thesis. She contends that this is better than the “composition interpretation”, which reads the thesis as claiming that mentality can be (partly) constituted by, or realized in, stuff other than ordinary grey matter, such as notebooks or silicon chips.

    At the risk of entering the domain of hermeneutics, I confess that it’s not entirely clear to me what it would mean for an interpretation of the Extended Mind thesis to be correct, or for one interpretation to be better than another. What is at issue here is not Clark and Chalmers’ original intent. At least, Vold does not make reference to original intent in making her case. She cites Clark’s interpretation of his own view, but doesn’t appear to regard this as deciding the issue. And the primary examples used in the original 1998 article, the Tetris computer and the notebook, involve both extra-cranial location and alternative composition. So these examples are compatible with both interpretations.

    Vold supports her own interpretation mainly by arguing that it better accounts for “what is distinctive about the extended mind thesis” (p. 16). On the composition interpretation, the Extended Mind thesis is a straightforward consequence of multiple realizability (though perhaps not of functionalism). Vold’s rationale thus seems to be essentially sociological: we should adopt the interpretation that best explains why the Extended Mind thesis was regarded, by the relevant philosophical community, as noteworthy or even radical. Yet as Vold observes, Farkas uses a similar argument to challenge the location interpretation.

    It seems to me that there is no single best interpretation of the thesis, as commentators may well construe it differently. These differences will be reflected in the implications they attribute to it, and the kinds of factors they take to support or challenge it. Vold’s discussion is nonetheless valuable, in that it highlights the threat of ambiguity and, hence, the importance of explicitly specifying how one is reading the Extended Mind thesis. For example, one may seek to challenge the Extended Mind thesis by rejecting multiple realizability; but this strategy will not work (in any direct way) if that thesis is given a location interpretation.

    This brings me to Vold’s more philosophical project. The disparity between the composition and location interpretations corresponds to a disparity between multiple realizability and multiple localizability. These are orthogonal, in that neither requires (or entails) the other. Vold’s example of the iCog illustrates that multiple localizability does not require multiple realizability. In this example, a neural structure located outside the skull partly constitutes a mental state. Since the iCog is composed of the same kind of stuff (with the same kind of structure) as ordinary mental states, the possibility of an iCog does not require multiple realizability. I think Vold is right about this: the iCog example does show that multiple localizability is compatible with unique realizability (the denial of multiple realizability).

    However, the disparity between the iCog and the standard cases of extended mentality raises questions about the location interpretation of the Extended Mind thesis. As Vold herself observes,

    the neuron-based iCog only requires multiple localizability, but without multiple realizability the thesis is conservative when compared with the more extravagant range of cases that Clark & Chalmers’ functionalist parity argument supports. (p. 15)

    The iCog argument supports a thesis of multiple localizability only: it shows that mentality can be partly constituted by factors outside the skull. The notion that this thesis is more “conservative” than the thesis supported by Clark and Chalmers’ arguments (which I presume is the Extended Mind thesis) seems to undermine Vold’s argument for the location interpretation of the Extended Mind thesis. After all, the comparative conservatism of the multiple localizability thesis implies that the location interpretation does not capture the sense that the Extended Mind thesis is, as Vold says, a “novel” and “important insight” (p. 10).

    Perhaps I’m misunderstanding the dialectic here. And even if my worry is sustained Vold’s discussion of the interpretation issue is valuable. In my view, it is valuable because it draws attention to an ambiguity, one that could easily lead to the conflation of two orthogonal questions: about how mental states can be realized and about where they can be located. Vold nicely disentangles these strands of Clark and Chalmers’ argument, and demonstrates their independence.

    Coda: multiple realizability without multiple localizability?

    Vold’s iCog argument focuses on the possibility of multiple localizability without multiple realizability. In claiming that these are orthogonal, she also maintains that mentality could be multiply realizable without being multiply localizable. Her discussion of this centers on an analogy given by Aizawa: it could be that there were differences among how car engines operated (some use carburetors, others use fuel injectors, etc.), yet every car engine was located in the same place within the car.

    Aizawa’s example concerns actuality rather than possibility: it is a case where engines are multiply realized, but not uniquely located. This example isn’t intended to show that multiple realizability does not entail multiple localizability. But it’s clear that multiple realizability does not entail multiple localizability. After all, issues about composition are fundamentally different from issues about location.

    Still, I think it would be surprising if mental states were multiply realizable yet not multiply localizable. Suppose that pain is multiply realizable in that both c-fiber stimulation and (a particular kind of) silicon chip activation realize pain. What could possibly prevent extra-cranial silicon chip activation—or, for that matter, extra-cranial c-fiber stimulation—from realizing pain? That is, how could the potential realizer’s spatial location prevent it from realizing pain? The obvious possibility is that realizing pain requires standing in certain relations to other parts of the brain or body, such as nerve cells, and a potential realizer must be located within the skull in order to stand in these relations. But spatial location seems crucial to such relations only insofar as those relations are not multiply realizable. Suppose that silicon chips could be appropriately activated (or c-fibers appropriately stimulated) by a kind of transmission from nerve cells that was less direct than ordinary neurotransmission, and could pass through the skull—perhaps involving radio waves. In that case, silicon chip activation (or c-fiber stimulation) that realized pain could occur outside the skull.

    My point is just this. While multiple realizability does not entail multiple localizability, it may be that any principled restrictions on the latter will stem from restrictions on the former. For it is difficult to reconcile unrestricted multiple realizability, multiple realizability of all related states and processes, with unique localizability of mental states or processes.

    The iCog possibility is easy to accept because it is, as Vold says, relatively conservative. To my mind, the question of where mentality could be located is less freighted with philosophical implications than the question of how mentality could be constituted. I don’t see any reason to think that there are limits as to the location of mentality, even relative to a body that affects, and is controlled by, the mental states in question. My speculative suggestion is that, if there are any in principle limits to the location of mental states or processes, relative to the thinker’s body, these restrictions will likely stem from some restrictions on realizability as regards processes associated with those states.

    This last suggestion is not at odds with anything Vold says; it lies outside her paper’s main focus. But these further reflections were prompted by Vold’s rich and engaging paper.

  3. There seem to me to be two principal issues in Vold’s paper. The first concerns the potential implications of multiple realizability for extended cognition and the second is a new argument for extended cognition.

    To properly appreciate Vold’s take on these issues, we should recall the rough and ready distinction between the claim that it is possible that cognition extends and the claim that cognition actually extends. The distinction is rough and ready, since, for one thing, we would need to consider whether the possibility is supposed to be logical possibility, nomological possibility, conceptual possibility, metaphysical possibility, or what have you. But, let me set aside the particulars of the distinction. Both claims have been advanced by advocates of extended cognition, but the claim that it is possible for cognition to extend is less controversial than is the claim that cognition does extend. Adams and Aizawa, for example, have granted that it is possible for cognition to extend, but that as a matter of contingent empirical fact, there appear to be no instances of this.

    Consider, then, how this bears on Vold’s new argument for extended cognition based on the iCogs thought experiment. This seems to make the case that extended cognition is, in some sense, possible. This seems right, but to my knowledge no one has made the case that extended cognition is impossible. But, suppose that there are those who take extended cognition to be, in some sense, impossible. Why is the iCogs thought experiment any more illuminating that the infamous Inga-Otto thought experiment? And why is this any more illuminating than simply supposing that cognition would take place outside the skull, if someone were simply to have one’s skullcap surgically removed? The paper, and the interest in the possibility of extended cognition, would be enhanced if a) there were some plausible reason to think that extended cognition is impossible, but b) this plausible reason was shown to be untenable by the iCogs example.

    Consider, next, the implications of multiple realizability for extended cognition. (Here, for the sake of brevity, I set aside the potential implication of functionalism for extended cognition.) Many philosophers seem to have the view that if cognitive states/processes are multiple realizable, then this supports extended cognition. Yet, Vold draws attention to a distinction, drawn elsewhere, between multiple realizability, on the one hand, and multiple localizability, on the other. Very roughly speaking, multiple realizability has to do with having a multiplicity of possible ways of “constructing” something. By contrast, multiple localizability has to do with having a multiplicity of possible locations of a thing (relative to some landmark). A simple example from Aizawa, 2007, used by Vold, should provide the gist. One might take four cylinder and six cylinder engines to illustrate the multiple realizability of internal combustion engines; whereas the front or rear placements of the engines illustrate the multiple localizability of internal combustion engines.

    Vold rightly points out that multiple realizability does not entail multiple localizability. Just because one can build four cylinder and six cylinder engines (multiple realizability), for example, does not entail that they can be placed in both the front and the rear of an automobile (multiple localizability). Furthermore, the hypothesis that cognition can extend is a thesis about the localizability. Thus, where many suppose that the multiple realizability supports the possibility of extended cognition, it does not.

    Vold might easily get even stronger conclusions. Multiple realization is stronger than multiple realizability. The former entails the latter, but not vice versa. Moreover, multiple localizability is weaker than multiple localization. The latter entails the former; but not vice versa. But, if one follows Vold’s earlier argumentation, one can also see that even the stronger assumption of the multiple realization does not entail multiple localization or even multiple localizability.

    To summarize, it seems to me that Vold has two distinct proposals in this paper. One is for a new argument for extended cognition; the other concerns the implications of multiple realizability. The new argument does not seem to bolster an interesting hypothesis. It makes the case that it is (in some sense) possible for cognition to extend, but this does not seem to be a controversial claim. The implications of multiple realizability, however, are more interesting. It shows that multiple realizability, and indeed multiple realization, by itself, does not support extended cognition. This conclusion does challenge what seems to be a popular view in the extended cognition literature.

  4. I’d like to thank Prof. Ken Aizawa, Prof. Brie Gertler, and Prof. Mike Wheeler for their insightful and generous commentaries on my paper. I offer a short response to each below. Together they have given me much food for thought and left me excited to return to my paper to address the concerns that they have raised. I also want to thank Cameron Buckner, Nick Byrd, and John Schwenkler, as well as the chair of my session, Andrew Christman, for organizing this fine conference and for including my paper in such a fantastic line-up.

    I would also like to make a correction to my citations: I have not cited the correct paper by Prof. John Sutton, I should have cited one of his papers on his alternative complementarity account of the extended mind, such as any of these papers:

    J. Sutton (2006). Distributed cognition: domains and dimensions. Pragmatics and Cognition, 14, 235-247.

    J. Sutton (2010). Exograms and interdisciplinarity: history, the extended mind, and the civilizing process. In Richard Menary (Ed.), The Extended Mind (pp.189-225). Cambridge, MA: MIT Press.

    J. Sutton, Celia B. Harris, Paul G. Keil, & Amanda J. Barnier (2010). The psychology of memory, extended cognition, and socially distributed remembering. Phenomenology and the Cognitive Sciences, 9.4, 521-560.

    Response to Wheeler:

    Many thanks to Prof. Wheeler for his thoughtful comments. He has made me rethink, most of all, how I should characterize the main objective of this paper. He seems willing to accept what I take to be the central claims of my paper: “In the end Vold is, I think, right about the following: (i) the extended mind hypothesis is an interesting and distinctive view because of what it claims about the location of mental states; (ii) one can have extended cognition without multiple realizability; (iii) this can happen on the basis of multiple localizability.”

    I agree that the new argument I offer has the limitations that Wheeler articulates, namely, that “purely on the basis of multiple localization, one cannot have cases of artefact-involving extended cognition”, and that, “the cases that have propelled the extended mind to fame (or notoriety) depend on multiple realizability, not just multiple localizability.” His concerns are understandable. I will take time to consider how I might meet these concerns and to be clearer about what my objectives are in the next draft of this paper.

    Response to Gertler:

    I am very thankful for these excellent comments by Prof. Gertler. They have given me much to think about both as I continue to revise this paper and in terms of the future directions of my research. Gertler’s comments consist of, first, a response to my paper where she raises at least two concerns and, second, some additional reflections on the relation between multiple realizability and multiple localizability and the issue of the location of mentality more generally.

    Gertler reads me as undertaking two different kinds of projects in my paper: an interpretive project and a more philosophical project. With respect to the first project, I appreciate her remark that it is not clear what makes one interpretation (of the extended mind thesis) better than another. I would say that Clark’s interpretation of his own view should be decisive, only others (I cite Farkas here) disagree on this. In the end, Gertler says, “there is no single best interpretation of the thesis, as commentators may well construe it differently.” This is an important issue for my project that I will need to spend more time considering in future revisions.

    With respect to the second “more philosophical project” I undertake in my paper, Gertler raises a concern about my proposed “location interpretation” of the Extended Mind thesis and how this fits with my iCog example. If I understand Gertler correctly, her concern is that the argument I offer will not be able to support the range of cases that Clark and Chalmers offer in their original paper. I agree with this point. My takeaway is that I need to reconsider what I take my main aim to be in this paper and then to do a better job at articulating that aim.

    In the final part of her comments Gertler remarks on the relation between realization and location. She thinks, first of all, that it would be surprising if mental states turned out to be multiply realizable and yet not multiply localizable. Furthermore, she suggests that, “if there are any in principle limits to the location of mental states or processes, relative to the thinker’s body, these restrictions will likely stem from some restrictions on realizability as regards processes associated with those states.” I agree with her suggestion here. But her final discussion also raises some questions that have been weighing on my mind about the potential philosophical implications of where mentality could be located. I will undoubtedly spend more time thinking about these issues moving forward.

    Response to Aizawa:

    I owe a special thanks to Prof. Aizawa not just for his insightful comments but also for his ‘Reply to Bechtel and Mundale’ paper, which was a helpful resource for me in writing this paper.

    First, Aizawa has challenged the usefulness of an argument, like mine, for the claim that it is possible that minds could extend, saying, “to my knowledge no one has made the case that extended cognition is impossible.” So there is a need for me to identify a real target. His constructive suggestion is that, “The paper, and the interest in the possibility of extended cognition, would be enhanced if a) there were some plausible reason to think that extended cognition is impossible, but b) this plausible reason was shown to be untenable by the iCogs example.” This is something I can work on.

    Finally, Aizawa has offered some further useful comments on the distinction between realization and location. What I find most interesting is his remark that “if one follows Vold’s earlier argumentation, one can also see that even the stronger assumption of the multiple realization does not entail multiple localization or even multiple localizability.” I think this is right. In the end, Aizawa comments, together with Gertler’s remarks on this topic, have convinced me that the relationship between realization and location merits further discussion.

    A final thanks to Aizawa. He has given me some constructive suggestions and identified several important areas where I could have been clearer.

Comments are closed.