Andrew Clark

Where does the mind stop and the rest of the world begin? The question invites two standard replies. Some accept the demarcations of skin and skull, and say that what is outside the body is outside the mind. Others are impressed by arguments suggesting that the meaning of our words "just ain't in the head", and hold that this externalism about meaning carries over into an externalism about mind. We propose to pursue a third position. We advocate a very different sort of externalism: an _active externalism_, based on the active role of the environment in driving cognitive processes.

Recent studies such as Thelen and Smith, Kelso, Van Gelder, Beer, and others have presented a forceful case for a dynamical systems approach to understanding cognition and adaptive behavior. These studies call into question some foundational assumptions concerning the nature of cognitive scientific explanation and the role of notions such as internal representation and computation. These are exciting and important challenges. But they must be handled with care. It is all to easy in this debate to lose sight of the explanatorily important issues and to talk at cross purposes, courtesy of the various ways in which different theorists often concieve key terms.

Mind, it has recently been argued1, is a thoroughly temporal phenomenon: so temporal, indeed, as to defy description and analysis using the traditional computational tools of cognitive scientific understanding. The proper explanatory tools, so the suggestion goes, are instead the geometric constructs and differential equations of Dynamical Systems Theory. I consider various aspects of the putative temporal challenge to computational understanding, and show that the root problem turns on the presence of a certain kind of causal web: a web that involves multiple components (both inner and outer) linked by chains of continuous and reciprocal causal influence. There is, however, no compelling route from such facts about causal and temporal complexity to the radical anti- computationalist conclusion. This is because, interactive complexities notwithstanding, the computational approach provides a kind of explanatory understanding that cannot (I suggest) be recreated using the alternative resources of pure Dynamical Systems Theory. In particular, it provides a means of mapping information flow onto causal structure -- a mapping that is crucial to understanding the distinctive kinds of flexibility and control characteristic of truly mindful engagements with the world. Where we confront especially complex interactive causal webs, however, it does indeed become harder to isolate the syntactic vehicles required by the computational approach. Dynamical Systems Theory, I conclude, may play a vital role in recovering such vehicles from the burgeoning mass of real-time interactive complexity.

Is consciousness all in the head, or might the minimal physical substrate for some forms of conscious experience include the goings on in the (rest of the) body and the world? Such a view might be dubbed (by analogy with Clark and Chalmers’s ( 1998 ) claims concerning ‘the extended mind’) ‘the extended conscious mind’. In this article, I review a variety of arguments for the extended conscious mind, and find them flawed. Arguments for extended cognition, I conclude, do not generalize to arguments for an extended conscious mind.

Marinov''s critique I argue, is vitiated by its failure to recognize the distinctive role of superposition within the distributed connectionist paradigm. The use of so-called subsymbolic distributed encodings alone is not, I agree, enough to justify treating distributed connectionism as a distinctive approach. It has always been clear that microfeatural decomposition is both possible and actual within the confines of recognizably classical approaches. When such approaches also involve statistically-driven learning algorithms — as in the case of ID3 — the fundamental differences become even harder to spot. To see them, it is necessary to consider not just the nature of an acquired input-output function but the nature of the representational scheme underlying it. Differences between such schemes make themselves best felt outside the domain of immediate problem solving. It is in the more extended contexts of performance DURING learning and cognitive change as a result of SUBSEQUENT training on new tasks (or simultaneous training on several tasks) that the effects of superpositional storage techniques come to the fore. I conclude that subsymbols, distribution and statistically driven learning alone are indeed not of the essence. But connectionism is not just about subsymbols and distribution. It is about the generation of whole subsymbol SYSTEMS in which multiple distributed representations are created and superposed.

O'Regan and Noe present a wonderfully detailed and comprehensive defense of a position whose broad outline we absolutely and unreservedly endorse. They are right, it seems to us, to stress the intimacy of conscious content and embodied action, and to counter the idea of a Grand Illusion with the image of an agent genuinely in touch, via active exploration, with the rich and varied visual scene. This is an enormously impressive achievement, and we hope that the comments that follow will be taken in a spirit of constructive questioning. Overall, we have two main reservations.

The paper considers the problems involved in getting neural networks to learn about highly structured task domains. A central problem concerns the tendency of networks to learn only a set of shallow (non-generalizable) representations for the task, i.e., to miss the deep organizing features of the domain. Various solutions are examined, including task specific network configuration and incremental learning. The latter strategy is the more attractive, since it holds out the promise of a task-independent solution to the problem. Once we see exactly how the solution works, however, it becomes clear that it is limited to a special class of cases in which (1) statistically driven undersampling is (luckily) equivalent to task decomposition, and (2) the dangers of unlearning are somehow being minimized. The technique is suggestive nonetheless, for a variety of developmental factors may yield the functional equivalent of both statistical AND informed undersampling in early learning.

Recent advances in cognitive science and cognitive neuroscience open up new vistas for human enhancement. Central to much of this work is the idea of new human-machine interfaces (in general) and new brain-machine interfaces (in particular). But despite the increasing prominence of such ideas, the very idea of such an interface remains surprisingly under-explored. In particular, the notion of human enhancement suggests an image of the embodied and reasoning agent as literally extended or augmented, rather than the more conservative image of a standard (non-enhanced) agent using a tool via some new interface. In this essay, I explore this difference, and attempt to lay out some of the conditions under which the more radical reading (positing brand new integrated agents or systemic wholes) becomes justified. I adduce some empirical evidence suggesting that the radical result is well within our scientific reach. The main reason why this is so has less to do with the advancement of our science (though that certainly helps) than with our native biological plasticity. We humans, I shall try to show, are biologically disposed towards literal (and repeated) episodes of sensory re-calibration, of bodily re-configuration and of mental extension. Such potential for literal and repeated re-configuration is the mark of what I shall call "profoundly embodied agency," contrasting it with a variety of weaker (less philosophically and scientifically interesting) understandings of the nature and importance of embodiment for minds and persons. The article ends by relating the image of profound embodiment to some questions (and fears) concerning converging technologies for improving human performance.

Mind, it is increasingly fashionable to assert, is an intrinsically embodied and environmentally embedded phenomenon. But there is a potential tension between two strands of thought prominent in this recent literature. One of those strands depicts the body as special, and the fine details of a creature’s embodiment as a major constraint on the nature of its mind: a kind of new-wave body-centrism. The other depicts the body as just one element in a kind of equal-partners dance between brain, body and world, with the nature of the mind fixed by the overall balance thus achieved: a kind of extended functionalism (now with an even broader canvas for multiple realizability than ever before). The present paper displays the tension, scouts the space of possible responses, and ends by attempting to specify what the body actually needs to be, given its complex role in these recent debates.

Fodor’s theory makes thinking prior to doing. It allows for an inactive agent or pure reflector, and for agents whose actions in various ways seem to float free of their own conceptual repertoires. We show that naturally evolved creatures are not like that. In the real world, thinking is always and everywhere about doing. The point of having a brain is to guide the actions of embodied beings in a complex material world. Some of those actions are, to be sure, more recondite than others. But in every case the contents of thoughts still look to depend, in some non-unique but vitally important way, on the kinds of doings they support.

What is the relation between the material, conventional symbol structures that we encounter in the spoken and written word, and human thought? A common assumption, that structures a wide variety of otherwise competing views, is that the way in which these material, conventional symbol-structures do their work is by being translated into some kind of content-matching inner code. One alternative to this view is the tempting but thoroughly elusive idea that we somehow think in some natural language (such as English). In the present treatment I explore a third option, which I shall call the "complementarity" view of language. According to this third view the actual symbol structures of a given language add cognitive value by complementing (without being replicated by) the more basic modes of operation and representation endemic to the biological brain. The "cognitive bonus" that language brings is, on this model, not to be cashed out either via the ultimately mysterious notion of "thinking in a given natural language" or via some process of exhaustive translation into another inner code. Instead, we should try to think in terms of a kind of coordination dynamics in which the forms and structures of a language qua material symbol system play a key and irreducible role. Understanding language as a complementary cognitive resource is, I argue, an important part of the much larger project (sometimes glossed in terms of the "extended mind") of understanding human cognition as essentially and multiply hybrid: as involving a complex interplay between internal biological resources and external non-biological resources.

Like those famous nations divided by a single tongue, my paper and Professor P.M. Churchland's deep and engaging reply offer different spins on a common heritage. The common heritage is, of course, a connectionist vision of the inner neural economy- a vision which depicts that economy in terms of supra-sentential state spaces, vector-to-vector transformations, and the kinds of skillful pattern-recognition routine we share with the bulk of terrestrial intelligent life-forms. That which divides us is, as ever, much harder to isolate and name. Clearly, it has something to do with the role of moral talk and exchange, and something to do with the conception of morality itself. Most of this Reply will be devoted to clarifying the nature of the disputed territory. First, though I shall rehearse some points of agreement concerning moral talk and progress.

Embodied agents use bodily actions and environmental interventions to make the world a better place to think in. Where does language fit into this emerging picture of the embodied, ecologically efficient agent? One useful way to approach this question is to consider language itself as a cognition-enhancing animal-built structure. To take this perspective is to view language as a kind of self-constructed cognitive niche: a persisting though never stationary material scaffolding whose critical role in promoting thought and reason remains surprisingly ill-understood. It is the very materiality of this linguistic scaffolding, I suggest, that is responsible for some key benefits. By materializing thought in words, we create structures that are themselves proper objects of perception, manipulation, and thought.

Language, according to Jackendoff, is more than just an instrument of communication and cultural transmission. It is also a tool which helps us to think. It does so, he suggests, by expanding the range of our conscious contents and hence allowing processes of attention and reflection to focus on items which would not otherwise be available for scrutiny. I applaud Jackendoff s basic vision, but raise some doubts concerning the argument. In particular, I wonder what it is about public language that uniquely fits it to play the functional role which Jackendoff isolates — why couldn't expression in a private inner code induce the same computational benefits? I suggest a weaker position in which the communicative role of public language moulds it into a suitably expressive resource, such that natural language emerges as the logically and technologically contingent filler of a functional role which could, in principle, be filled by other means. I also compare and contrast Jackendoff's position with some related ideas due to Daniel Dennett and others, concluding with a sketch of my own view of language as an external artifact whose computational properties complement those of the basic biological brain.