Richard Menary Cognitive Integration, Extended Mind, cognitive task, cognitive tasks, environmental vehicles, manipulating, cognitive processes, Harvard Press, Richard Menary, cognitive system, Philosophical Psychology, external, notebook, manipulations, Connectionism, Cambridge MA, cognitive process, MIT Press, integration, problem solving, Cognitive integration, representations, Oxford, Artificial Intelligence, internal representations, Mind and Cognition, Cognition In The Wild, Cambridge University Press, Menary, R., John Sutton, cognitive systems, representational system, Millikan, R., Cognitive Science, external representation, Journal of Cognitive Systems Research, representation, environment, arguments, bodily processes, distributed cognition, Menary, The practice, cognitive architecture, References Adams, external representations, Oxford University Press
Cognitive Integration and the Extended Mind XI Cognitive Integration and the Extended Mind Richard Menary
1. Introduction: First wave EM A recent, radical approach to cognition has argued that some cognitive vehicles and cognitive processes
1 are external to the body. This new approach has variously been labelled as: active externalism (Clark and Chalmers 1998), vehicle externalism (Hurley 1998, Rowlands 2003), and the extended mind (Clark and Chalmers 1998). This first wave of arguments focuses on the causal interaction, sometimes known as `coupling' between bodily internal vehicles and processes and bodily external vehicles and processes I will call these first wave arguments2 extended mind style arguments. Allied to the extended mind is the emphasis on social situation and embodiment found in situated and distributed cognition
(Hutchins 1995, Gallagher 2005). Together they constitute the view that cognisers are embodied and located in a situation which has both physical and Social Aspects
and that some bodily interactions with the environment constitute cognition. Moreover, extended mind style arguments are radical because they are committed to the manipulation thesis: "Cognitive processes are not located exclusively in the skin of cognising organisms because such processes are, in part, made up of physical or bodily
1 I shall use just vehicle and process for simplicities sake. 2 Following John Sutton (this volume). 267
Cognitive Integration and the Extended Mind manipulation of structures in the environments of such organisms." (Rowlands 1999, 23) Hence, the first wave of arguments aims at establishing that some cognition is, in part, externally located. However, simply to think of this emerging view of cognition as externalist is misleading. This is because the payoff from extended mind style arguments is the integration of the bodily `internal' and `external' aspects of cognition into a whole. Therefore, I will refer to this emerging approach to cognition as cognitive integration. The second wave of arguments, have the aim of beginning the job of explaining how the bodily internal and external aspects of cognition are integrated into a whole and this integration is to be understood in terms of the manipulation of environmental vehicles. My main line of argument is to show that the upshot of extended mind style arguments leads us to understand cognition (and the mind) as hybrid involving both internal and external processes and integrated the bodily internal and external processes coordinate with one another in the completion of cognitive tasks. Secondly, we cannot make good on the manipulation thesis without understanding the normativity of the bodily manipulations of external vehicles of cognition. The motivation for cognitive integration is not that we are causally coupled to external vehicles, nor is it the view that the mind is first in the head and then gets extended out into the world (into the vehicles themselves). Adams and Aizawa's caricature of the extended mind is, therefore, an attack on a straw man (Adams and Aizawa 2001, this volume). The primary motivation for cognitive integration is that we should begin our study of cognition and the mind from the brute fact that we are embodied and embedded in the world. Our primary engagements with the world are
Cognitive Integration and the Extended Mind embodied and, unsurprisingly, our initial cognitive engagements with the world are embodied engagements, they are primarily sensorimotor engagements (see Menary 2007, chapter four). At least some of our mature cognition retains the structure of these embodied engagements in the form of manipulations of the environment, hence some cognitive processes are hybrid they are comprised of neural processes and vehicles and bodily processes on environmental vehicles. This is what I take extended cognition/cognitive integration to be about, it is the attempt to understand what the nature of the integration between these elements of a hybrid process is like. One way, suggested in the introduction to this volume, is to think of hybrid cognitive processes as enacted skills or capacities for manipulating the environment. However, we should not forget that the embodied cogniser is embedded in a physical and social environment
and that environment contains norms which determine the content of environmental vehicles and how we manipulate them. 2. What Is Cognitive Integration? A popular way of thinking of cognition is to think of cognitive systems as systems of the brain. In philosophy this usually goes hand in hand with a commitment to some form of mind-brain supervenience; if so then cognition has a natural boundary, it is contained by the brain. It follows that if you want to study cognition and the mind, then you need to study the systems responsible for cognitive and mental phenomena implemented in the brain. Furthermore, it may be that we believe that the `bounded by the brain' view is quite common-sensical, the mind is indeed `in the head'. Most philosophers and cognitive scientists take cognition to be a clump of mental acts or processes that come under broad headings such as: remembering, perceiving, learning and reasoning. Identifying what makes a process cognitive is more difficult. In their
Cognitive Integration and the Extended Mind continuing critique of the extended mind Adams and Aizawa (Adams and Aizawa 2001, this volume), stipulate that processes that exhibit the mark of the cognitive are identified as those that involve representations with non-derived [intrinsic] content. However it is not only notoriously difficult to specify just what intrinsic content is supposed to be (Hutto 1999, Dennett 1990, Mendola 2003), but also the definition looks to be unduly restrictive (Menary 2006). In general, there is no real agreement in the cognitive science community upon a definition of what a cognitive process is, nor of what the vehicles of cognition are. For example, classical computationalists take the vehicles of cognition to be symbols that have formal, or syntactic, properties in virtue of which they are processed (Fodor and Pylyshyn, 1988). Connectionists deny that the vehicles of cognition are symbols, instead they are patterns of activation distributed across nodes in a network. Connectionists understand cognitive processes to be algorithms for the spread of activation across the network (Smolensky, 1988, 1995). It is quite natural to be pluralistic about cognitive processes and vehicles; as such, there is no single genuine `cognitive kind'. In general, we might specify that cognition involves the manipulation of a cognitive vehicle in the completion of a cognitive task. The classical - connectionist debate demonstrates that there is a plurality of types of manipulations and vehicles employed by empirical theories of cognition. However, what we do have is a sense of the cognitive task as defined by Rowlands (2003, 161): "...it does seem fairly clear that the notion of a cognitive process is defined, in part, in terms of the notion of a cognitive task. A cognitive process is one that plays a fairly central role
in allowing a subject to accomplish a cognitive task."
Cognitive Integration and the Extended Mind Quite generally this amounts to perceiving the world, remembering things about the world and employing things remembered in making inferences, problem solving
and the like (Rowlands 2003)3. Perhaps there is more clarity about the notion of a cognitive system? There are two senses of cognitive system, which we need to distinguish. That of a particular cognitive system for example the memory system as it might be - and the overall system, of which these specialised sub-systems are parts. However, there is not even agreement on what a cognitive system is in either of these senses. One very general way of thinking about cognitive systems is that they are the mechanisms that underlie the processes involved in remembering, perceiving, learning and reasoning. A theory of cognitive architecture, such as classicism or connectionism, is supposed to specify the nature of these mechanisms. Classicists and connectionists have generally agreed that the mechanisms that underlie cognitive processes are all in the cranium and we shall call this cognitive internalism. Why should we think that cognitive internalism is wrong to suppose that all cognitive processes, vehicles and systems supervene exclusively on the brain? Humans spend a lot of time and effort to create linguistic and representational surrounds and then maintain and manipulate them. The exercising of the capacities to create external linguistic and representational vehicles is of course fleeting, although the long-standing disposition is not (see the introduction to this volume). We often, for example, write out mathematical problems, rather than completing them `in the head'. Another sense is the direct manipulation of the environment to complete 3 A general definition of a cognitive task can easily end up being unhelpfully vacuous. If we define the cognitive task as any task for the completion of which cognition is required, then almost every task will be a cognitive one. I think it is more helpful if we think of cognitive tasks as involving the exercise of particular cognitive capacities such as remembering a date, solving a problem, learning to do something, etc. These are tasks where the exercising of cognitive capacities is directly tied to their Successful completion
Cognitive Integration and the Extended Mind cognitive tasks. For example, expert players of the game Tetris prefer to rotate the shapes on the screen using buttons, rather than by rotating images of them `in the head.' If cognition is bounded by the brain, why do we not complete all these cognitive tasks, and many others like them, `in the head?' Cognitive integration provides an answer to this question. Its cash value is that the co-ordination of bodily processes of the organism with salient features of the environment, often created or maintained by the organism, allows it to perform cognitive functions that it otherwise would be unable to; or allows it to perform functions in a way that is distinctively different and is an improvement upon the way that the organism performs those functions via neural processes alone. Developing the integrationist position begins with the fact of our embodiment. Embodied approaches to the mind and cognition are supposed to reveal to us something profound about the embodiedness of our minds that we ought to understand the mind as shaped by the body. However, there seems to be a bifurcation of approach in the embodied mind community, there is on the one hand the phenomenologically inspired approach of Gallagher (2005) with a detailed account of how bodily activity in the environment constrains what we perceive and of what we are consciously aware. This approach takes seriously the detailed description
of embodiment with regard to cognitive and mental capacities such as perception and social cognition
. Then there is the distributed/extended approach to cognition and mind of the likes of Hutchins (1995), Clark (1997) and Rowlands (1999) who begin with the assumption that cognition is embodied but then concentrate on the ways in which we interact, bodily, with the environment. They take seriously detailed descriptions of manipulations of external representational vehicles such as diagrams, mathematical notations or written sentences with regard to mental and cognitive
Cognitive Integration and the Extended Mind capacities such as memory and belief. Extended cognition does not give a detailed account of the way in which the body shapes cognition in these cases, rather they tend to focus on how external vehicles (artefacts, representations) shape and transform cognitive capacities. The difference in approach does not constitute a profound difference. Rather, we are approaching the same phenomenon from different directions. Therefore, we need to reconceive the mind on both bodily and environmental grounds. For example, integrationists take the manipulation of external vehicles to be a pre-requisite for higher cognition and that embodied engagement is a pre-condition for these manipulative abilities. Therefore, not only would it be a mistake to disengage the body from its environment, but also to ignore the contribution of external representational systems to our cognitive capacities. A straightforward way of understanding the position of cognitive integration is in terms of bodily engagement with vehicles in the extra-bodily environment, in such a way that they are integrated into a whole. The study of bio-cultural representational systems is reliant upon a clear understanding of those systems as structured by bio-causal co-ordinations/integrations and that the functioning of the system requires the stability and availability of extrabodily vehicles and the bodily manipulation of those vehicles. This is certainly true, but these explanatory projects lack the resources to fully explain how and why we manipulate extra bodily vehicles in the way that we do. To do this satisfactorily we need to place the dynamics of the system in a wider cultural and normative setting.
Cognitive Integration and the Extended Mind
Fig. 1 Red arrows: Manipulations of external vehicles have to be understood in terms of bio-cultural norms
and external cognitive vehicles have to be understood in terms of their role in bio-causal integrations. Hence the two approaches are mutually explanatory.
Our abilities to manipulate the extra-bodily environment are normative and are largely dependent upon our learning and training histories. Hence, explanations of the dynamics of integrated cognitive systems will only be one, important, explanatory factor. There are three complementary ways in which we can understand integration: 1. Bio-causal co-ordinations/integrations: The dynamical approach analyses the reciprocal coupling between systems which are part of a larger system. They have causal influence over each other for as long as they are coupled. This is a symmetrical relation, the two systems are mutually constraining of each others behaviour. 2. Embodied Engagement: The body is integrated with the environment through its body schemas, which are unconscious sensorimotor programmes for action. These
Cognitive Integration and the Extended Mind programmes often integrate with the environment in two ways, firstly by training (or evolutionary adaptation) and secondly by norms governing practices such as driving, playing a sport such as tennis or writing, etc. 3. The Manipulation Thesis: Humans manipulate their local environment with their bodies. They might directly manipulate the physical structure
of the environment and they might use tools do this. They create artefacts, such as tools and representational vehicles. Humans very often create and manipulate external representational vehicles to complete a cognitive task. In doing so they are carrying out a cognitive practice which is governed by its own norms which I call cognitive norms. Unlike first wave arguments for the extended mind, cognitive integration does not rely upon the parity principle, the motivation is found in the brute fact of our embodiedness and embeddedness. The parity principle is apt to confuse and mislead (see the introduction to this volume) as a motivation for the extended mind. I elaborate on this in the next section. In the final two sections I outline and explain the manipulation thesis and cognitive practices. 3. Cognitive Integration and the Parity Principle As I explained in the introduction to this volume, external processes/vehicles do not get to have cognitive status conferred on them because they are relevantly similar to (supposedly) uncontroversial cases of cognitive processes and vehicles which are internal. Nor do they get to be relevantly similar because external processes/vehicles are causally coupled to internal processes/vehicles. Parity will not necessarily come from the direct similarity of the external with the internal. Internal process X may have properties a, b, c and external process Y may have properties d, e, and f.
Cognitive Integration and the Extended Mind Internalists latch on to these differences and use them to deny parity and, therefore, that there are any external cognitive processes and vehicles4. This version of the parity principle is fatally flawed because it assumes the very position it is meant to displace. The extended mind and the parity principle encourage us to think of an internal cognitive system that is extended outwards into the world. Hence it implicitly endorses a picture of a discrete cognitive agent, some of whose cognitive processes get extended out into the world. It also argues for the cognitive role of the environment by claiming that they are functionally similar to (or the same as) the functions of neural processes. The main question of the extended mind would then be: "How do processes in the world get to function like processes in the brain?" A major difference between extended mind style arguments and cognitive integration is that the latter does not depend upon the parity principle. It cannot be misinterpreted as claiming that cognition is extended from inside the head out into the world, or that external processes are cognitive because they are similar (weak version) or isomorphic (strong version) to internal processes. Cognitive integration differs from first wave extended mind style arguments because it takes the manipulation thesis to be its starting point, not the parity principle. The parity based formulation of the extended mind is a functionalist thesis, take the Otto example. C&C want to say that in the case of Otto and Inga there is a sufficient functional similarity between Otto's use of his notebook and Inga's recall from biological memory that we are inclined to say that Otto has beliefs. Otto's retrieval of information about the location of MOMA causes him to go to 53rd street and the pattern of activation in a part of Inga's brain causes her to go to 53rd street.
4 See A&A's paper this volume. 276
Cognitive Integration and the Extended Mind Otto's information retrieval
and Inga's neural activations play the same kind of causal role in producing actions. As such, the physical implementation of the causal role is irrelevant to the functional level of description Otto's use of his notebook and Inga's pattern of activation in her brain. Otto receives input from the environment, there is an exhibition on at the MOMA, he then retrieves the location of MOMA from his extended memory system, which causes him, as behavioural output, to go to 53rd street. Only at the grossest level of functional description can this be said to be true. Otto and his notebook do not really function in the same kind of way that Inga does when she has immediate recall from biological memory. There are genuine and important differences in the way that memories are stored internally and externally and these differences matter to how the memories are processed. John Sutton has pointed out that biological memories stored in Neural Networks
are open to effects such as blending and interference (see Sutton this volume). The vehicles in Otto's notebook, by contrast, are static and do no work in their dispositional form (Sutton this volume). This is of course no problem for cognitive integration which does not work from the assumption that internal and external vehicles and processes need to be functionally equivalent. They may function in very different ways, as Sutton points out. However, this is the point, it is because the external vehicles provide a different kind of functionality and that they can co-ordinate with internal processes that they are integral parts of our cognitive systems. Again, putting this complementary integration in the Wider context
of cognitive tasks and practices highlights the cognitive roles that external vehicles can play; but this is not a matter of functional similarity. It is in the details of the integration between neural processes and vehicles
Cognitive Integration and the Extended Mind and environmental vehicles via bodily manipulations that we will understand the hybrid nature of cognition. 4. The Manipulation Thesis Clark clearly is committed to the manipulation thesis: "In all these cases the individual brain performs some operations, while others are delegated to manipulations of external media. Had our brains been different, this distribution of tasks would doubtless have varied." (Clark and Chalmers this volume, 8) Clark (1997, this volume, and 2001) explains the manipulation thesis by causal interaction between organism and environment, often referred to as causal coupling. Rowlands (1999, 2003) explains the manipulation thesis in terms of the bodily manipulation of external vehicles, or information bearing structures. There is a clearer version of the parity principle at work here which also deals with Adams and Aizawa's admonition that the extended mind does not provide a clear mark of the cognitive. The mark of the cognitive is that cognitive processes involve the manipulation of information bearing vehicles in completing a cognitive task. Hybrid cognitive processes involve the integration of neural manipulations of vehicles and bodily manipulations of environmental vehicles.
"there seems to be no great theoretical divide between manipulating internal information bearing structures and manipulating external information bearing structures to make available to oneself, or to one's cognitive operations, the information that results. To claim that only the former constitutes genuine
Cognitive Integration and the Extended Mind information processing
seems little more than an internalist prejudice." (Mark Rowlands 2006, 39)
Extended mind theorists are still developing a clear sense of what external vehicles are. Sometimes they talk of external vehicles as affordances (Hurley 1998, Rowlands 1999), sometimes as external representations such as written sentences, diagrams and other notations (Clark and Chalmers 1998). In the extended Mind, Clark and Chalmers take external vehicles5 to play the role of cognitive vehicles and they take the external manipulations of those vehicles to play the role of cognitive processes. When Otto accesses his notebook to recall the address of the Museum of Modern Art
, he is manipulating an external vehicle as part of his remembering where the Museum is located. Integrating external vehicles in Otto's notebook with internal vehicles constitutes a part of the same cognitive process. The co-ordination of internal manipulations and external manipulations allows the cognitive agent to complete the cognitive task. In this case the cognitive agent, Otto, manipulates the vehicles in his notebook to retrieve the desired information concerning the location of MOMA. Therefore, the cognitive integrationist claims that for any cognitive system, some cognitive vehicles and cognitive processes are externally located. Nevertheless, the overall cognitive system is integrated because `internal' and `external' co-ordinate with one another in completing cognitive tasks. Now we need to give an account of the different kinds of manipulations of external vehicles. We can classify bodily manipulations and I identify four general classes:
5 This ambiguity needs to be eliminated, especially as some criticisms of the extended mind focus on the claim that artefacts and tools can play the role of external vehicles.
Cognitive Integration and the Extended Mind · Biological Coupling: such as extended phenotypes (Dawkins 1982), animate vision (Ballard 1991) and SMC (O'Reagan and Noe 2001). · Epistemic Actions: using the environment as its own representation, obviating the need for endogenous representations as in Tetris (Kirsch and Maglio 1994). · Self Correcting Actions: The use of language and external props to direct and structure practical actions in completing tasks. · Cognitive Practices: the manipulation of external representational and notational systems according to certain normative practices as in mathematics (Vygotsky 1978, Karmiloff-Smith 1992, Menary 2007). Examples of biological coupling run from cases such as phonotaxis in crickets (Webb 1994) and bee dances (Millikan 1993, 2004) to sensorimotor contingencies (Noл and O'Regan 2001) and animate vision (Ballard 1991). Kirsch and Maglio (1994) have dubbed the second class of manipulations epistemic actions. An epistemic action involves directly manipulating the environment to bring about a better state in a problem solving/planning task, rather than constructing an internal representation and manipulating that. An example of a self-correcting action is the role of spoken language in structuring activity, such as reminding oneself of the order in which one must conduct a sequence of actions. In these kinds of cases we use speech as a corrective tool. The classic example of a cognitive practice is Rumelhart and McClelland's (1986) example of using pen and paper to complete a mathematical algorithm. Performing long multiplication involves mastery over a notational system, which
Cognitive Integration and the Extended Mind involves cognitive norms for manipulating those notations when completing cognitive tasks. In the final section I look in more detail at the nature of cognitive practices. 5. Cognitive Practices We are able to manipulate external vehicles because we gain manipulative abilities that are governed by cognitive norms. These are norms that govern manipulations of external representations, which aim at completing cognitive tasks. This is obvious given that external vehicles, such as written language and mathematical symbols, are tokens of representational systems. Such systems have their own norms governing manipulations of token representational vehicles. Hence, they are cognitive norms, as opposed to moral or social norms. I shall call manipulations of an external representation to complete a cognitive task a cognitive practice. Otto's cognitive practice involves writing things in his notebook and then accessing them later. Otto's cognitive practice also falls under the definition of a cognitive process given above. Therefore, manipulations of internal and external vehicles are causally integrated, but we should place this within a wider cultural and normative context. This is what is missing from the first wave of extended mind style arguments, which fall prey to some of the internalist worries about parity I outlined above. Cognitive integration benefits from the central insight of the extended mind hypothesis some cognitive vehicles are bodily external and manipulations of these vehicles are part of the overall cognitive process, which includes manipulations of bodily internal vehicles. If we focus on manipulations of classical representations, such as mathematical symbols, then we can see the importance of these two points. Classical representations are best understood as notational schemes that are physically
Cognitive Integration and the Extended Mind
embodied, on paper or on a computer monitor for example. What extended mind style arguments have not explained, and what their claims of causal coupling do not show, is how we are able to manipulate a variety of notational types. There is a great variety of representational systems6 which mirrors the great variety of tasks to which we put them. Examples of such tasks include: solving problems
, making inferences, planning, working out answers to questions etc. (these are cognitive tasks). A manipulation of any of an external representation is normative, in the sense that we learn or acquire a practice that is an established method of manipulating representations to produce an end. For example, we write down the intermediate stages in problem solving, which can function as part of the working memory space making information available for further manipulation. Or we might directly manipulate the world as part of the problem solving process, rather than manipulating internal representations
. Plans are often written down and then transformed, updated and shared. Lists and diaries allow us to retrieve information that requires long term storage and is easily and conveniently accessible. The representational properties of maps allow for easy and shared navigation, allowing for the kind of detailed representations and orientations that internal representations cannot provide. In each case, there is a cognitive task that must be completed. The practice allows us to complete the task by manipulating the representation. The implementation of a cognitive practice depends upon cognitive norms that guide that practice. So, for example there are: 1. Purposive norms. The activity is engaged in for a purpose, or end. 2. Corrective norms. Norms for using representations to correct activity in pursuit of an end.
6 I take all classical notations to be representational. Hence, when I speak of a notation it should be taken to be a representation.
Cognitive Integration and the Extended Mind 3. Manipulative norms. Norms for manipulating inscriptions of a representational system. 4. Interpretative norms. Norms for interpreting inscriptions of a representational system as having some wider significance, not just within the notational system itself but also with regard to the wider world and interests of others. Manipulations of representations are embedded in a practice, which has a normative, as well as a physical/causal dimension, such as the practice of manipulating mathematical notations. The practice of manipulating a representation is normative because we learn how to manipulate the representations correctly and because of the cognitive purpose of the practice. The purpose is to achieve a particular kind of goal, such as solving a problem, planning, or making inferences which I call the cognitive task. It follows that we will need an account of how we learn cognitive practices. This will, in part, involve the acquisition of capacities to manipulate representations and thereby transform our cognitive abilities. However, acquiring these capacities should be understood in the context of the cognitive practices required to complete cognitive tasks. The practice of manipulating representations, a cognitive practice, is essentially the embodying of norms in an activity. An example of such an activity is that of writing a scholarly paper by word processing
. Which of the components play an active causal role? Presumably, thanks to the CPU, the keyboard and monitor are able to exert an effect on what I write next and the words I type which come up on the screen are an extension of short term memory. In a stronger sense my reading and re-reading what I have written gives me new ideas about what I should write next. Thus, the keyboard and monitor play an important causal role in the production of the paper. There is, however, a sense in
Cognitive Integration and the Extended Mind which this is the wrong focus of interest. Whilst it is true that tools such as keyboards and pens enable me to write, it is manipulating the written vehicles themselves that drives my cognitive processes. The sentences extend my working memory and are, of course, what can be re-written, erased, moved to another paragraph, etc. It is, moreover, precisely these kinds of manipulations that are not easily, if ever, achieved in the head. Therefore, writing as an active and creative process
is enabled by tools such as pen and paper or Word Processor
s. The written vehicles are then available for further manipulations such as restructuring, revising and re-drafting. Manipulating written vehicles is a kind of problem solving where a particular goal is aimed at: ``how do I make this piece of writing clearer?'' for example. I could, of course, compose a paper without external media. Nevertheless, not only would retaining the paper and updating it be made more difficult but, perhaps more crucially, it would take on different content. The kind of manipulations of written sentences described above require external vehicles and tools for manipulating them, without them behavioural competence will drop. However, it is not just a matter of ease that is at issue here; in an important sense, the manipulation of scripts transforms the skills needed in composing scholarly articles. The media function as enabling hardware, but the vehicles themselves enable processes that cannot be completed in the head alone. The physical act of typing necessarily involves external physical manipulations. My ability to compose a paper is severely curtailed by the absence of those external manipulations. Hence, cognitive integrationists are inclined to think that those external manipulations play an important enabling role in the processing of the task. Why could we not stick to a form of neural internalism here? There is, of course, an attenuated sense in which I can compose an article in my head. The likelihood of retaining much of the argument and structure, would, however, become
Cognitive Integration and the Extended Mind very limited. Making revisions and corrections would be almost impossible, for example: trying out ideas and then deleting them. By contrast, becoming integrated with external representations transforms my cognitive capacity to compose a philosophy paper. Importantly, there are things I can do with pen, paper, or word processor that I cannot do in my head. Stable and enduring external written sentences allow for manipulations, transformations, reorderings, comparisons and deletions of text that are not available to neural processes. A further internalist worry becomes apparent here, why should the integrationists insist that the manipulation of the written sentences be cognitive? Why cannot we just say that some of the representations and manipulations of those representations get manipulated in the environment and then function as input for further, genuine neural cognitive processes? Here we reach the nub of the issue, once the internalist accepts that the manipulation of representational vehicles is part of the process, it is very difficult for them to discount their cognitive function without invoking some form of neural chauvinism. Furthermore, the claim that manipulating written vehicles simply provides new input for neural processes does not do justice to the integration of neural processes, bodily processes and manipulations of vehicles in real time. 6. Conclusion Cognitive integration takes the first wave of extended mind arguments to establish that cognition is hybrid. However, it is not motivated by the parity principle but takes embodied engagement with the world as it is starting point. The manipulation thesis provides a further motivation, and a definition of integrated cognition. It is then the job of the integrationist to provide a taxonomy of the different kinds of manipulation and to provide empirical examples of them (see Clark, Wilson, Sutton, Hurley,
Cognitive Integration and the Extended Mind Cowley and Spurrett, this volume for examples). The manipulations must also be understood in terms of cognitive norms as well as causal explanations of the manipulation of external vehicles. This explanatory project is very different from the straw man version of the extended mind criticised by Adams and Aizawa in this volume. References Adams, F. and Aizawa, K. (2001) The Bounds of Cognition. Philosophical Psychology 14, 43-64. Ballard, D. (1991) Animate Vision. Artificial Intelligence
, 48, 57-86. Adams, F. and Aizawa, K. (This Volume). Defending the Bounds of Cognition. Clark, A. (1999) `Where brain, body, and world collide'. Journal of Cognitive Systems Research
, 1, 517. Clark, A. (2001) Reasons, robots and the extended Mind. Mind and Language. 16 (2), 121-145. Dawkins, R. (1982) The Extended Phenotype. Oxford: OUP. Fodor J. and Pylyshyn Z. (1988) Connectionism and cognitive architecture, Cognition, Vol. 28, Nos. 1-2, 3-71. Gallagher, S. (2005), How the Body Shapes the Mind (Oxford: Oxford University
Press). Hurley, S. (1998) Consciousness In Action. Cambridge MA: Harvard Press. Hutchins, E. (1995). Cognition In The Wild. Cambridge, MA: MIT Press. Karmilloff-Smith, A. (1992) Beyond Modularity: A Developmental Perspective on Cognitive Science. Cambridge MA: MIT Press.
Cognitive Integration and the Extended Mind Kirsh, D. and Maglio, P. 1994. `On distinguishing epistemic from pragmatic actions', Cognitive Science, 18, 513-549. Menary, R. (2006) Attacking The Bounds of Cognition in Philosophical Psychology. 19, 3, pp. 329-344. Menary, R. (2007), Cognitive Integration: Mind and Cognition Unbounded. Basingstoke: Palgrave Macmillan
. Mendola, J. (2003) A Dilemma for Asymmetric Dependence. Nous 37, 2, 232 257. Millikan, R. (1993) White Queen Psychology and Other Essays For Alice. Bradford Books/MIT Press. Millikan, R. (2004) The Varieties of Meaning: The Jean-Nicod lectures. MIT Press. O'Regan, J. K. and Noл, A. 2001. A sensorimotor account of vision and visual consciousness
. Behavioural and Brain Sciences 24 (2001): 9391031. Rowlands, M. (1999) The Body in Mind: Understanding Cognitive Processes. Cambridge: CUP. Rowlands, M. (2003) Externalism: Putting Mind and World Together Again. Chesham: Acumen. Rowlands, M. (2006) body language
: Representation in Action. Cambridge: Cambridge University Press. Rumelhart, D., Smolensky P. and Hinton G.E., In McClelland J. and Rumelhart D., Parallel Distributed Processing: Explorations in the Microstructure of Cognition, 1986 vol.2, pp. 7-58. M.I.T. Press. Rupert, R. (2004) "Challenges to the Hypothesis of Extended Cognition," Journal of Philosophy 101: 389-428.
Cognitive Integration and the Extended Mind Smolensky P. (1988) On the proper treatment of connectionism, reprinted in C. Macdonald and G. Macdonald (eds.) Connectionism: Debates on Psychological Explanation. Oxford: Blackwell. 28-89. Smolensky P. (1995) Constituent Structure and Explanation in an Integrated Connectionist/Symbolic Cognitive Architecture, in C. & G. Macdonald (eds.), Connectionism: Debates on Psychological Explanation. Blackwell. 223-290. Vygotsky, L. (1978). Mind In Society. Cambridge MA: Harvard Press. Webb, B. (1994). Robotic Experiments in Cricket Phonotaxis. Cliff, David, Husbands, Philip, Meyer, Jean-Arcady, and Wilson, Stewart W. From Animals to Animats 3: Proceedings of the Third International Conference
on the Simulation of Adaptive Behaviour. MA: MIT Press. 45-54