Phenomenology and the Cognitive Sciences

, Volume 11, Issue 2, pp 287–308

Extended cognition and fixed properties: steps to a third-wave version of extended cognition


    • Department of PhilosophyMacquarie University

DOI: 10.1007/s11097-011-9237-8

Cite this article as:
Kirchhoff, M.D. Phenom Cogn Sci (2012) 11: 287. doi:10.1007/s11097-011-9237-8


This paper explores several paths a distinctive third wave of extended cognition might take. In so doing, I address a couple of shortcomings of first- and second-wave extended cognition associated with a tendency to conceive of the properties of internal and external processes as fixed and non-interchangeable. First, in the domain of cognitive transformation, I argue that a problematic tendency of the complementarity model is that it presupposes that socio-cultural resources augment but do not significantly transform the brain’s representational capacities during diachronic development. In this paper I show that there is available a much more dynamical explanation—one taking the processes of the brain’s enculturation into patterned practices as transforming the brain’s representational capacities. Second, in the domain of cognitive assembly, I argue that another problematic tendency is an individualistic notion of cognitive agency, since it overlooks the active contribution of socio-cultural practices in the assembly process of extended cognitive systems. In contrast to an individualistic notion of cognitive agency, I explore the idea that it is possible to decentralize cognitive agency to include socio-cultural practices.


Extended cognitionFixed-propertiesDynamical propertiesCognitive transformationCognitive assembly


In philosophy of mind and cognition, the model known as extended cognition (EC) expresses the view that many of our genuinely cognitive processes are composed of physical vehicles running on machinery distributed across parts of the brain, body, and material–cultural environment (Clark 2011; Menary 2010a, b, c; Rowlands 2010; Sutton 2010; Wheeler 2010). My independent aim in this paper is to explore an issue that I do not believe has been pursued elsewhere in the literature on EC, namely the relationship between EC and, what I shall call, the “fixed-properties view of internal and external processes” (FP). Throughout this paper, I hope to show that it is possible to move EC along both faithfully and fruitfully by downplaying this tendency to treat the properties of the “internal” and “external” as fixed and instead explore an alternative route to EC.

To get an understanding of FP, we can contrast FP with what I call the “dynamical properties view” (DP). Unlike FP, DP holds that the plastic brain gets enculturated through development in socio-cultural practices (Roepstorff et al. 2010, p. 1052). This is the first part of DP. The second part is as follows. Unlike FP, DP does not assume, when having to explain the integration/assembly of cognitive systems, that the individual organism is the most active element. DP implies that the assembly of cognitive systems is the result of richly dynamical and distributed elements, where there is no collapse into individualism like in FP. To flesh out DP in more detail, I focus on two domains.

The first of these domains is sometimes referred to as cognitive transformation (Menary 2010b), where I focus on the transformation of neuronal representational capacities as a result of enculturation. To make sense of this, I discuss three examples: (1) Näätänen et al.’s (1997) study on phoneme perception suggesting a diachronically mediated re-shaping of certain cortical areas of the brain by “the brain’s” participation in structured socio-cultural practices; (2) Dehaene’s (2005) neuronal recycling hypothesis suggesting that our uniquely human capacity for exact mathematical reasoning is partly biologically inherited and partly a result of enculturation into socio-cultural practices; and (3) Wheeler’s (2004) philosophical work on language and off-line cognition, where Wheeler argues that the brain must “recapitulate” certain structural features of language in order for humans to engage in off-line cognition. The second domain is the construction of cognitive systems typically referred to as cognitive assembly (Clark 2008; Hutchins 2008, 2011a; Kirchhoff and Newsome 2011; Sterelny 2010). I want to signpost a couple of important issues here. First, unlike individualist approaches to cognitive assembly (Clark 2008), addressing cognitive assembly via DP implies a non-individualistic conception of cognitive agency. Second, unlike synchronic perspectives on cognitive assembly (Clark 2008), DP aims to explain synchronic assembly (in part) from diachronic considerations. Thus, in contrast to DP, FP is the view that the properties of biological processes and the properties of socio-cultural processes remain non-interchangeable, both diachronically and synchronically.

Scope of the critical exploration and strategy

Nothing of what I have to say in this paper will undermine the credibility of the EC research program, in the sense that what I shall critically explore are tendencies or assumptions left unquestioned. Hence, I do not claim that there is a necessary relationship between EC and FP. So, the more conservative aim is to establish that leaving FP behind opens a space for interesting research on EC topics. Let that be the first proviso. The second is that I shall not be arguing that EC in general is committed to FP. First, I do not argue that parity-based arguments are committed to FP. This is evident from my endorsement of Wheeler’s (2004) case of off-line cognition precisely because Wheeler is a parity-driven EC theorist. Nevertheless, I will argue that Clark, whose work is often directed at parity-EC, shows evidence of FP-style assumptions with respect to the domains of cognitive transformation and cognitive assembly. In cognitive transformation, the justification for this claim comes from the consideration that Clark only thinks that parity is sufficient for cognitive extension, and that the far “more interesting and plausible argument […], is the one which describes the seepage of the mind into the world by stressing that “the brain’s brief is to provide complementary facilities that will support the repeated exploitation of operations upon the world […].” (Clark 1998, p. 98; italic added). Hence, it is at the fine-grained level of Clark’s complementary arguments—and not at the coarser-grained level of functional isomorphism—that FP assumptions arise in Clark’s work. In addition to this, what warrants my gluing of Clark to FP in the domain of cognitive assembly is the individualistic conception of cognitive agency expressed in Clark’s hypothesis of organism-centered cognition: if the orchestration process of systemic parts on the synchronic level is primarily the result of organismic properties, then the dynamics of socio-cultural practices are left explanatorily static, with the result of separating unnecessarily the biological from the cultural. Therefore, it is not parity that I am questioning but rather assumptions in the complementarity model and in EC’s individualistic notion of cognitive agency. Second, I do not argue that Menary’s framework of cognitive integration is committed to FP, in that Menary is much more DP-oriented in his approach. However, on the issue of cognitive assembly, there is still a tendency to explain integration from the perspective of the individual organism. Third, because I am pursuing a critical line against complementarity EC, the arguments against Clark also apply to the model of complementarity endorsed by Sutton (2010). However, and which will become clear in the “Toward a third-wave extended cognition” section, Sutton is quite aware of the tendency within EC of fixing the realms of the “internal” and “external.”

Different waves of extended cognition

There are different versions, articulating different methodologies and research interests, in EC. Especially Sutton (2010) and Menary (2010c) should be credited for making several of these important differences come to light. In the present literature there are two articulated waves or versions of EC: Sutton and Menary call these 1st wave EC (Clark and Chalmers 1998; Clark 2008; Wheeler 2010) and 2nd wave EC (Menary 2006, 2007, 2010b; Rowlands 2010; Sutton 2010; Sutton et al. 2010).

First-wave extended cognition

First-wave EC is based on a combination of complementarity and what is known as the parity principle (PP). Opinions on these matters, however, differ somewhat. While Clark (see above) only thinks that the PP is sufficient for cognitive extension and that the better argument for EC is a complementarity-based argument (Clark 1998, p. 99), Wheeler (2010, 2011) conceives of the PP as the only viable route to EC (Walter 2010, p. 286). Nevertheless, because first- and second-wave EC share a commitment to complementarity, but do not equally share the endorsement of the PP, I focus here on the PP. The PP stresses the following: “If we confront some task, a part of the world functions as a process which, were it to go on in the head, we would have no hesitation in accepting as part of the cognitive process, then that part of the world is (for that time) part of the cognitive process.” (Clark and Chalmers 1998, p. 8; and reprinted in Clark 2008, p. 77). The PP focuses on functional isomorphism of “inner” and “outer” processes. In support of the PP, Clark and Chalmers developed the case of the neurobiologically impaired Otto and his notebook. Otto is a victim of a mild form of Alzheimer’s disease. Over time, Otto has written down useful information in his notebook in a similar way to storing information in biological memory. According to Clark and Chalmers, because the dispositional information in Otto’s notebook is functionally poised to guide action in a functionally similar way as non-occurrent beliefs in biological memory, the information in Otto’s notebook should be considered as cognitive belief-like states. Importantly, the PP by itself is not meant as an argument for EC but rather to encourage us to look at various cases of cognitive extension “behind a veil of metabolic ignorance,” (Clark 2011, p. 449). Given this picture, then, the PP acts as a kind of heuristic device (Menary 2010a, p. 5) designed to suspend what Clark calls “biochauvinistic prejudice” (2008, p. 77).

Second-wave extended cognition

Second-wave EC shares the complementarity view of Clark’s version of first-wave EC. But this second wave of EC is critical of arguments based on parity and functional isomorphism. Generally, it is EC in its integrationist, historical, and cognition-in-the-wild mode. Even so, second-wave EC is best understood not as a substantial doctrine departure from first-wave EC; but rather, as a refinement and (perhaps) attunement to a more empirically and enactive oriented approach to EC. However, even if the two waves are compatible, they are also distinct, in the sense that most proponents of second-wave EC argue that the PP is either wrong or incomplete as a motivation for EC in general (Sutton 2010, p. 200). Within second-wave EC, there are two approaches with slightly different views or inflections: the first starts from a principle of complementarity (Sutton 2010), whereas the second focuses on integration and manipulation (Menary 2007). Sutton builds his case for complementarity arguing, among other things, that the PP “does not encourage attention to the distinct features of the components in particular cognitive systems […],” and because of this “downplays—or even collapses—differences between inner and outer resources […].” (Menary 2010a, b, c, d, p. 198). As a result, the PP fails to study the deep mechanistic dissimilarities between “inner” and “outer” parts and on how they complement and “operate together in driving more-or-less intelligent thought and action.” (Sutton et al. 2010, p. 525). Against functional isomorphism, complementarity EC “both predicts and requires” (Rowlands 2010, p. 89) such disparate but complementary processes between culturally engineered props and artifacts and the brain’s biological modes of processing. For Menary, complementary is also an important aspect of cognitive integration (2006, p. 330). Sharing Sutton’s critical stance towards the PP, Menary builds his case for second-wave EC on the manipulation thesis:

“The manipulation thesis as a constituent thesis of cognitive integration is first understood to be an embodied engagement with the world, […]. Secondly it is not simply a causal relation, bodily manipulations are also normative—they are embodied practices developed through habit and training and governed by cognitive norms.” (2007, p. 84).

Important for Menary’s version of second-wave EC is that some cognitive processes are (partly) made up of—constituted, composed of—an individual’s bodily manipulation of “external” structures and that such manipulations are embedded in the wider social, semantic, and normative cognitive niche (2010d, p. 611). An important focus of Menary-style EC is the idea of cognitive transformation. In particular, Menary thinks that the PP fails to explain how bodily manipulations alter the informational and physical structure of the cognitive niche, thereby transforming human cognitive capacities. Also, nothing in the PP tells us about how such manipulations literally result in the transformation of body schemas required for manipulation of external elements and in transformation of both representational and cognitive capacities (2010b, p. 561).

Toward a third-wave extended cognition

Following Sutton (2010), this paper explores what a third-wave version of EC might look like. I take my starting point in the following specifications:

“If there is to be a distinct third wave of [EC], it might be a deterritorialized cognitive science which deals with the propagation of deformed and reformatted representations, and which dissolves individuals into peculiar loci of coordination and coalescence among multiple structured media […]. Without assuming distinct inner and outer realms of engrams and exograms, the natural and the artificial, each with its own proprietary characteristics, this third wave would analyze these boundaries as hard-won and fragile developmental and cultural achievements, always open to renegotiation.” (2010, p. 213; italics added)

The first feature mentioned is the framing “a deterritorialized cognitive science.” This is both a radical and fascinating proposal. Although I cannot deal with it sufficiently, since doing so will take us too far astray, more detail is required. Despite EC’s core claim that we should not privilege the boundaries of the biological organism in settling the question of what is cognitive and what is not, or, alternatively, what is “internal” to a cognitive system and what is “external,” certain strands of EC end up setting the bar for cognitive inclusion by appeal to what is ‘internal’ to the skin-and-skull boundary of an individual organism. This is evident if we consider first-wave EC’s parity argument for EC. Briefly, the PP works accordingly: (1) locate an “external” element taken to perform a functional role in intelligent behavior; (2) locate or imagine a scenario in which the same functional role is realized by an “internal” mechanism; (3) then ask the question, should we count the “internal” mechanism in (2) as part of the cognitive system?; (4) if the answer is yes, then (1) qualifies for cognitive inclusion. So, despite employed to overcome any sharp separation between “inner” and “outer”—between “fluid biology and stable culture” (Sutton 2006, p. 242)—the PP ends up privileging the “internal” on matters of cognitive inclusion for the “external” (Di Paolo 2009, p. 10). The relationship between what is “external” and what is “internal” to a cognitive system is most certainly complicated; nonetheless, it is not at all clear that sticking to the old “internal” and “external” distinction will be helpful in settling this debate. Moreover, it is important not to lose sight of the fact that the question of what is supposed to be “internal” to a cognitive system and what is “external” is pitched at a synchronic, individualist level, thereby encouraging an object-based view of cognitive systems. If one acknowledges, as Hutchins does (1995, 2008, 2011a, b), the socio-culturally distributed nature of cognitive systems and consequently of cognitive agency (more on this in the “Cognitive assembly: EC, FP, and DP” section), the criteria for what it means to be “internal” to a cognitive system become even more opaque. EC, of course, can still pursue the all important boundary inquiries into which processes, relations, systems, and mechanisms are cognitively relevant. However, if Sutton is correct, EC should not presuppose any clear-cut boundaries between “internal” and “external.”

This brings us to the second interesting feature, namely that third-wave EC deals with “deformed and reformatted representations.” This idea of deformed and reformatted representations takes third-wave EC’s focus on a decentralized and deterritorialized cognitive science even further. What does Sutton means by this? One way to get a grip on this allows me to comment critically on the complementarity framework endorsed by Sutton himself. Building on Donald’s (1991) theory of engrams (biological memory records) and exograms (culturally encoded public representations) as a signature mark of complementarity, Sutton points out that exograms are often discrete, modality and task independent, and much more durable than engrams. However, following Hutchins, this distinction between two disparate (but complementary) representational formats “has the unintended side effect of rendering cultural practices invisible.” (2008, p. 2017) In line with the enactivist approach (Stewart et al. 2011), what makes a material pattern into a representation is a process of enactment—of bringing forth of meaning. Within this framework, cognition is an active embodied and heavily embedded process (Hutchins 2011b, p. 429). Hence, to even engage with exograms as representations, it is necessary to bodily engage “a material pattern as a memory record” (Hutchins 2008, p. 2017) within structured and patterned socio-cultural practices. Thus, exograms are not simply “memory records outside the brain” with properties different from engrams; but rather, exograms are constantly reformatted according to the non-symbolic features of physical materials, cultural norms for their use, and embodied activities of individuals (Ingold 2000). There is a second way to understand the framing of “deformed and reformatted representations,” and it is this I shall pursue in depth in this paper. Instead of looking at the meshed composition of exograms, Sutton’s framing encourages analysis on the effects of enculturation on the plastic brain’s own representational capacities. I signpost this here, and nothing else, since a large part of the “Cognitive transformation: EC, FP, and DP” section is devoted to fleshing out the idea of the “enculturated brain.”

The third, and final, suggestion in the above passage is the claim that third-wave EC “dissolves individuals into peculiar loci of coordination and coalescence among multiple structures media.” Echoing Latour (1999) and his decentralized notion of agency, I understand this as a direct call to free EC from its present individualistic tendencies and to do so by taking socio-cultural practices, and diachronic contributions, seriously in addressing the issue of cognitive assembly and the underlying notion of cognitive agency. In fact, on a DP-oriented third-wave EC, cognitive assembly and agency are best understood as self-organizational processes distributed across brains, bodies, people, norms, and socio-cultural practices and structures—none of which have any analytical priority. However, driving assembly is not the same as effecting a significant “internal” transformation. So, apparently one can have DP-style assembly without DP-style transformation. I bring this worry here in order to avoid confusion about FP in cognitive transformation and FP in cognitive assembly. Generally, FP-style assumptions involve keeping boundaries fixed and separate. In cognitive transformation, FP arises because of a tendency to claim that the properties of “external” elements augment but do not transform the psychological innards of the individual. In cognitive assembly, in contrast, FP comes into play because of certain individualistic tendencies delegating all the causally active roles to the organism, thereby deactivating the dynamical role of the environment in co-driving cognitive assembly. Hence, FP is not only indicative of a non-transformative view of representational capacities but also of prioritizing the individual as the loci of coordination and control in cognitive assembly.

Cognitive transformation: EC, FP, and DP

I begin this section with a version of cognitive transformation endorsing DP, namely Menary’s (2010b) hypothesis of dual component process transformation. As I shall unpack this hypothesis, it contains not two but three important elements. The first implies an extension of the functional architecture by integration of new processing nodes into the cognitive system. The second implies an enhancement of individual cognitive abilities to engage in cognitive practices. According to Menary, the third, and final, element is:

“The deeply transformatory power of our learning histories in the cognitive niche is one that reformats the representational capacities of the brain in terms of public symbol systems.” (2010b, p. 576; italics added).

Everyone in EC endorse the first two elements of cognitive transformation. It is the final element—the claim that ontogenetic development within the cognitive niche reformats or transforms the representational capacities of the brain in terms of public symbol systems—that non-trivially marks the distinction between DP-inspired EC and EC with FP tendencies. Unfortunately, Menary is not very clear about what he means by this. Hence, part of my exploration (in the “Cognitive transformation, EC, and DP: the enculturated brain” section) is to examine this in more detail. In the following subsection, however, I show how FP theorizing sneaks into EC.

Cognitive transformation: EC and FP

Imagine digging a hole in your garden. When asked about where the digging power resides, how would you respond? Would you answer “in the spade,” or “in you”? According the Clark, both answers are incorrect, since digging power is the product of a larger coupled system. In the digging case, the point is that the spade makes an “ongoing and complementary contribution to that made by [the] biological body,” and “in such a case, [there is] no obvious sense in which I biologically replicate the essence of the spade’s activity.” (Clark 2006, p. 292) An important step in Clark’s reasoning about the transformatory role of language (and public symbols in general) is to view language in analogy to the spade. Hence, on Clark’s view, understanding language implies explaining the coordination dynamics between “internal” biological resources and “external” non-biological resources, enabling language to “occupy a wonderfully ambiguous position on any hybrid cognitive stage […], looking one moment like any other piece of the biological equipment, and the next like a peculiar potent piece of external cognitive scaffolding.” (2006, p. 293)

A quick objection and response

My claim is that precisely this spade-language analogy, conceiving of scaffolds as different but complementary, contains the assumption of FP in EC theorizing. Just like the properties of biological resources do not need to take on the properties of the spade’s activity, so is it possible for the biological brain to dynamically coordinate with linguistic scaffolds, and do so without alteration of the brain itself. As Clark says: “Over cultural-evolutionary time, […], language itself becomes the kind of thing (artifact) that brains like ours can learn to use, and that they can use without radically altering their basic modes of representation and computation.” (2004, p. 720; italics in original). Before going any further, I want to discuss an objection to my claim that the complementarity model of EC—crisscrossing first- and second-wave EC—displays FP tendencies. First, one could argue that a motivation for endorsing EC is that it frees the biological organism from having to carry on-board all the machinery required for higher-level problem solving. Our brains, as Clark puts it, “are generally better at Frisbee than at logic.” (1997, p. 60). Humans become good at logic not because our brains undergo a functional transformation so the brain becomes good at both Frisbee and logic, but rather by our brains continuously coupling with, decoupling from, and recoupling with public symbol systems. Second, if our brains become functionally transformed to be both good at Frisbee and logic, this would seem to screen-off the environment from its role of augmenting and empowering cognition. Although I agree with this worry in its motivational claim for EC, there are two problems with this way of conceiving of cognitive transformation from an anti-FP perspective. Giving a DP-style explanation of cognitive transformation by focusing on how some (and only some) of the brain’s representational capacities undergo a process of enculturation does not imply a reduction of the contribution of the environment to merely a form of garden-variety causal influence. This will become much clearer in the “Cognitive transformation, EC, and DP: the enculturated brain” section. But consider, briefly, Gallagher and Crisafi’s (2009) account of mental institutions. According to Gallagher and Crisafi, large-scale, socio-cultural practices—like legal institutions—carried out in interactive coordination by many people, using many kinds of cognitive tools, constrained and empowered by the activities of previous generations, could be instances of distributed cognitive systems enabling the achievement of cognitive practices. Such a case is important precisely because it “is a cognitive practice that in principle could not happen just in the head; indeed, it extends cognition through environments that are large and various.” (Gallagher and Crisafi 2009, p. 48) So, looking at how processes of enculturation might transform the brain’s representational capacities—as Menary puts it, in terms of public symbol systems—does not imply delegating the cognitive contributions of the environment a less important role than it is taken to play in the EC literature. The second problem with this kind of worry is that it itself indicates an overly synchronic way of dealing with cases of EC. Consider how the response is framed in terms of coupling with and decoupling from external resources. This imposes an overly inert picture of what is a vastly dynamical process at multiple levels. An important virtue of DP considerations is that such a perspective encourages an argument for synchronic cognitive extension based on diachronic, developmental considerations. It is from diachrony that the question of enculturated transformation processes becomes important.

Cognitive transformation: EC and FP continued

We can further substantiate the claim that there are FP tendencies in the complementarity model via recourse to a brief summary of how Clark (1997, 2006, 2008) distances his view of the cognition-transforming role of language from Dennett’s (1991, 1996) account of this issue. First, Clark agrees with Dennett on the evolutionary dimension. It is certainly true that our human cognitive profile is different from other non-linguistic primates—past and present. The point that both Clark and Dennett endorse is that it does not follow that this difference amounts to any profound and radical differences in underlying neural hardware. Rather, the “idea is that some relatively small neural (or neural/bodily) difference was the source that lit a kind of intellectual forest fire.” (Clark 2001, p. 151, 1997, p. 198; Dennett 1991, p. 219). The real disagreement between the two is on the developmental scale. On Clark’s interpretation (one he himself admits is only tentative), Dennett’s view is that developmental exposure to public language and symbol systems results in a profound reorganization of the brain’s computational and representational profile, in the sense that Dennett thinks that our largely innate and connectionist cognitive architecture is transformed—by exposure to language—to simulate “something like a serial logic engine […].” (Clark 1997, p. 197; Dennett 1991, p. 278) In contrast to Dennett, Clark sees public language (and material symbols) as essentially external cognitive tools “that complements but does not profoundly alter the brain’s own basic modes of representation and computation.” (1997, p. 198; italics added)

Cognitive transformation, EC, and DP: the enculturated brain

Case #1: off-line mathematical reasoning

According to Wheeler (2004), our human capacity to engage in off-line cognitive tasks requires transformation of the brain’s representational capacities. In fact, this is “an unavoidable consequence of the basic logic of the [EC] paradigm.” (2004, p. 709) This is an interesting argument in part because it shows DP-style considerations within EC and in part because it is pitched against what I argue is Clark’s FP assumption on cognitive transformation. As I shall reiterate this argument, it consists of the following steps (Wheeler 2004, pp. 708–9):
  1. 1.

    In cases of on-line, distributed (extended) mathematical reasoning, the “internal” mechanisms will be directly causally coupled with certain properties of mathematical symbols located in the environment.

  2. 2.

    According to Clark, it is the coordination dynamics between biological and non-biological elements that augment and empower cognition.

  3. 3.

    In cases of off-line mathematical reasoning, there are, by hypothesis, no such environmental factors onto which the “internal” mechanisms can integrate with or lock onto.

  4. 4.

    According to Clark, in cases of on-line and off-line mathematical reasoning, fundamentally the same “inner” processing mechanisms are active in both cases.

  5. 5.

    However, if certain properties of “external” mathematical symbols are essentially part of our capacity for mathematical reasoning, and if these are absent in cases of off-line mathematical reasoning, then certain “internal” surrogates must recapitulate certain structural properties of those “external” mathematical symbols.

  6. 6.

    Therefore, off-line mathematical reasoning requires “inner” representations that recapitulate certain structural properties of “external” symbol systems, namely those structural properties that are non-trivial contributing factors in on-line distributed mathematical performance.


Premises 4 and 5 carry the heaviest explanatory burden. However, we can begin by noting that the first two premises articulate an interactively coordinated system composed of domain-general connectionist-style mechanisms with public symbol systems. In accordance with EC, our human capacity for linguistic reasoning is the joint product of neural networks organized exhaustively in terms of high-dimensional patterns of activation and similarity metrics and environmentally realized symbol systems (Clark 1997; Sutton 2010).

In unpacking premise 4, I focus on responding to a potential worry one might have against DP in the domain of cognitive transformation. One could argue that the real power of Clark’s phrase “complements but does not profoundly alter the brain’s own basic modes of representation” (1997, p. 198; italics added), is that it resists the orthodox view that language-like features such as productivity and systematicity can be explained “only if a classical representational format (one which features a language-like combinatorial syntax) is understood to be a fundamental feature of the neural economy at some “higher,” non-implementational level.” (Wheeler 2004, p. 710) There is absolutely nothing in either Wheeler’s argument or in the alternative to FP I am exploring that heralds anything like a return to a “language of thought architecture” (LOT). Wheeler gives two compelling reasons for this. First, unlike a LOT, mathematical reasoning—regardless of accomplished on-line or off-line—does not require that there is any sort of domain-specific processing device organized so as to encode language-like symbols. According to Wheeler, “we can easily hold onto the Clark-inspired thought that fundamentally the same kind of processing mechanism (perhaps even the very same pattern-completing mechanism) may still be deployed […].” (2004, p. 708). This sets DP-style views in contrast with orthodox positions like a LOT precisely because a DP-oriented explanation of off-line mathematical reasoning is fully compatible with the neural architecture being at root connectionist. Second, the “recapitulation” claim inherent in Wheeler’s argument should be not understood as a claim about complete “re-programming” as Clark takes Dennett’s view to be. According to Wheeler, during on-line performance of mathematical reasoning we should not expect the psychological innards of the brain to undergo distinct processes of transformation. While on-line, it is the particular hybrid system of neural pattern-completing networks and environmental symbols that drives cognitive success. It is only during off-line cognition—in this case, mathematical reasoning processes—that a transformation of neural representational formats is required. Summarizing these two points, Wheeler claims:

“As far as I can see, while the sub-claim concerning the neutral effects of language on the brain’s own basic mode of computation can still be sustained following my critical arguments, the sub-claim concerning the neutral effects of language on the brain’s own basic mode of representation cannot.” (2004, p. 709).

This brings us to premise 5. The argumentative appeal of this premise comes from Clark’s own view that the brain’s own basic modes of representation remain unaltered regardless of being reliably and functionally integrated with public symbols in the cognitive niche or being effectively disconnected or decoupled from the exact same environmental symbol systems. Following Wheeler, if some properties of public symbol systems are distinctively part of our capacity for mathematical reasoning—that is, if our ability to manipulate mathematical symbols in the environment is part and parcel of our capacity to perform mathematical practices—and if the properties of these public symbol systems are absent in cases of mathematical off-line reasoning, then certain representations in the brain must stand in for these structural properties when conducted off-line. Thus,

“[I]n the off-line case we confront nothing less than a profound transformation in the brain’s own basic mode of representation, and that runs contrary to Clark’s avowed view.” (Wheeler 2004, p. 711).

Case #2: phoneme perception

Consider the concept of “patterned practices” (Roepstorff et al. 2010, p. 1051). Patterned practices are pervasive forms of structured practices in which most (if not all) human socio-cultural activity is both conditioned by and emerges from. So a patterned practice is a socio-cultural practice structured in a particular, collectively assembled way. As Roepstorff et al. say of these practices:

“Everyday life is continuously ordered into more or less stable patterns that are specific to particular types of situations, defining preferences, predispositions, and expectations for actors. These patterns present regularities that arise from everyday practices while at the same time shaping them. Patterns appear to be emergent phenomena that are currently not sufficiently explained by its constitutive factors, i.e., individual behavior or intersubjectivity.” (2010, p. 1051).

A patterned practice analysis is important because it sets up a perspective from which to explore important shifts in dynamical properties, viz., how patterns of practices at the socio-cultural level reshape the patterning of cortical connectivity and activity, “and in the same way the social practice forms patterns, large-scale brain signals as well as other psychophysical signals generated during particular task performance can be analyzed to expose significant patterning.” (2010, p. 1052) To underpin their claim, Roepstorff et al. turn to a study on the effects of enculturation on the brain, Näätänen et al.’s (1997) study on phoneme perception. This study indicates some important cortical transformatory effects facilitated by the brain’s participation in certain patterned, socio-cultural practices.

Based on a mismatch negativity paradigm, where subjects are exposed to unattended sounds in certain rhythmic patterns, Näätänen et al. have established that the primary auditory cortex in the left hemisphere is highly sensitive to changes in predictable sound patterns (1997, p. 432). In the first set of these experiments (1997), Näätänen et al. utilized Finnish and Estonian language speakers because of a relatively small discrepancy between the two languages in terms of vowel structure, except that Estonian vowel space includes an additional vowel, /õ/, not found in Finnish. In the experiment, the speakers were also presented as deviants a prototype of this sound, along with vowels existing in both languages (/o/ and /ö/), and a non-prototypical vowel (located between/e/ and /ö/) (1997, p. 432; Roepstorff et al. 2010, p. 1053). The upshot was that Finnish speakers showed significantly higher mismatch negativity when exposed to prototypical vowels in their native language than when exposed to the Estonian vowel /õ/. Näätänen et al. take this finding to show that this “language-specific prototypical effect suggests the existence of neural traces of language-specific phoneme representations.” (1997, p. 433)

The neuronal recycling hypothesis: arithmetic and reading

The final piece of evidence running against FP-style assumptions and favoring a DP-based explanation of cognitive transformation comes from neuroscience and the development of arithmetic and reading abilities. Especially Dehaene’s “neuronal recycling hypothesis” (NRH) is interesting here (2005). Why should one think that Dehaene’s work supports DP? One of the most compelling ideas of the NRH is that when the brain recruits and utilizes cognitive tools during ontogenetic development this may lead to a recycling or pre-empting of biological mechanisms so that these acquire a new function from their evolved function (Dehaene 2005; Nieder and Dehaene 2009). This accommodates Wheeler’s point: that during participation in socio-cultural patterns of stable mathematical practices, certain cortical areas of the brain undergo a transformation in representational format enabling off-line mathematics to occur. Second, because of these recycling processes, the organism itself may undergo certain losses in cognitive abilities. If FP had been the case, this loss of abilities would have been hard to explain. For instance, in learning to read and engaging in regular patterns of reading such practices will partially reduce perceptual abilities like object-related activations in the left inferotemporal sulcus (Dehaene 2005). Of course, one might object to my use of the NRH as evidence of DP, to the extent that the NRH invokes the idea of “neuronal niches”: the idea that cultural plasticity is not unlimited, but that certain functional and structural evolutionary mechanisms partially constrain the acquisition of novel cultural capacities. In the following passage, Dehaene makes this explicit:

“According to this […] view, the architecture of the human brain is limited and shares many traits with other non-human primates. It is laid down under tight genetic constraints, yet with a fringe of variability. I postulate that cultural acquisitions are only possible insofar as they fit within this fringe, by reconverting pre-existing cerebral predispositions for another use. Accordingly, cultural plasticity is not unlimited, and all cultural inventions should be based on the pre-emption of pre-existing evolutionary adaptations of the human brain. It thus becomes important to consider what may be the important precursors of reading and arithmetic.” (2005, p. 134)

But why think (like one of the reviewers of this paper) that this counts against DP? As far as I can tell, this worry is based on a symptom spread throughout the literature on cognitive extension, namely an overly static and synchronic conception of the relationship between brain, body and environment. If one is willing to accept that diachronic development can establish more than merely causal dependence of neural mechanisms on environmental elements, then the NRH’s emphasis on “constraints imposed on cognitive development during ontogeny” has itself emerged through dynamical interactions over large, evolutionary time-scales. The simple fact that the NRH does not express a “largely equipotential cortex” (Quartz and Sejnowski 1997) and virtually unlimited plasticity is unlikely to count against a DP explanation. Let us look at some evidence.

According to Dehaene (1997), Dehaene et al. (1999) the processes enabling mathematical reasoning are stored in a complex network of representational mechanisms: first, the capacity for exact mathematical reasoning is stored in a culturally acquired language-specific format and located strictly in left-lateralized activation in the inferior frontal lobe; second, the capacity for approximate mathematical reasoning stored in an evolutionary older and non-linguistic, visuo-spatial representational format and located primarily in the bilateral parietal lobes of the brain (1999, p. 970, 973). The model invoked by Dehaene et al. depicts mathematical reasoning as composed from three cognitive capacities (Dehaene 2001, pp. 18–20):
  • A biological precursor to discriminate small quantities, e.g., 1–2, 2–3, etc. This capacity is a cross-species capacity found in preverbal children, dolphins, chimps, rats, etc.

  • A biological precursor to discriminate magnitudes between objects and numbers, e.g., 8 is larger than 4. This capacity is also a cross-species capacity.

  • An enculturated numerical system enabling humans to reason in a discrete and symbolic format required for exact mathematical competence.

Evidence supporting the result that approximate reasoning is stored in a non-linguistic format and that exact, discrete, and symbolic reasoning is stored in a language-dependent format is given in Dehaene et al. (1999). In one experiment, Russian–English bilinguals were trained on 12 cases involving exact and approximate sums of two-digit numbers presented in one of their two languages (1999, p. 970). In the exact addition task, subjects were required to select the correct sum of two numerically close numbers, e.g., the English trained subjects were asked to add “Twelve + Four” and then select the correct sum from “Sixteen and Fourteen” (1999, p. 970). In the approximate addition task, subjects were asked to estimate the result and select the closest number, e.g., the English trained subjects were asked to add “Twelve + Four” and estimate the result by choosing the closest number “Eighteen and Six” (1999, p. 970). After being trained on the 12 cases of two-digit numbers, subjects were then tested on the same sums. What Dehaene et al. interestingly found was that on the exact addition problems, response time decreased when having to perform exact calculation in their untrained language, whereas subjects show an increase in response time when performing exact calculation in their trained language, regardless of they were trained in Russian or English (1999, p. 971). In contrast, performance on the approximate addition task was unaffected by switching between trained and untrained language. According to Dehaene et al. such studies provide evidence that:

“[D]emonstrate that exact calculation is language-dependent whereas approximation relies on non-verbal visuo-spatial cerebral networks.” (1999, p. 970)

The evidence from off-line mathematical reasoning, the view of enculturation of cortical areas in the case of phoneme perception, and the acquisition of exact mathematical reasoning lend support to DP—certain neural transformations take the form of a re-shaping or reformatting process, where diachronic development in socio-cultural practices significantly transforms a sub-set of the brain’s cortical representations from primarily non-linguistic to linguistic. A quick return to the evidence on phoneme discrimination might help shed further light on this issue. There is evidence that human infants from birth are capable of discriminating phonemic patterns in their native language and non-native languages suggesting that phoneme discrimination is based in non-verbal auditory networks (Bertoncini et al. 1987). What is interesting is that during the first year of their lives, infants show a decrease in their phoneme discrimination abilities. After their first year, the infants only show sensitivity to the phoneme structure of their native language (Werker and Lalonde 1988). Such evidence fits the enculturated brain hypothesis. As Roepstorff et al. point out: diachronic development in stable and predictive patterns of socio-cultural practices sculpts the patterns of neural activity to the ones found at the socio-cultural level:

“In the cases here discussed, the differences in patterning, exposure to different languages with different structures of the phonetic space, may co-locate with differences in culture in the sense that ‘Finns’ appear to have one type of brain while ‘Estonians’ appear to have a different type of ‘brain’.” (Roepstorff et al. 2010, p. 1053)

Determining whether such evidence of enculturation is evidence for a constructivist approach (Quartz and Sejnowski 1997), a selectionist approach (Dehaene 2005), or a nativist approach (Pinker 1994) to neuronal and cognitive development is a matter of dispute and further empirical investigation. I signpost it here, since for my purposes a solution to this debate is less important than what it illustrates—for we have here a string of evidence suggesting transformatory effects on certain of the brain’s representational capacities due to its immersion in patterned socio-cultural practices.

Cognitive assembly: EC, FP, and DP

In addition to the domain of cognitive transformation, I aim in this section to show that there are FP-style tendencies in the domain of cognitive assembly as well. Unlike cognitive transformation, where it was the assumption that non-biological elements augment but do not transform certain of the brain’s representational capacities that linked EC’s complementarity framework to FP, FP takes on a very different structure in the domain of cognitive assembly. Hence, it is essential not to conflate my dealings with FP in the domain of cognitive transformation with what is to come in this section. In the domain of cognitive assembly, FP emerges due to an individualistic conception of the cognitive agent orchestrating the assembly processes, viz., that it is primarily the biological individual that is conceived as assembling cognitive systems (Clark 2008). So, FP sneaks into the debate on cognitive assembly due to a failure of decentralizing sufficiently the notion of cognitive agency to include socio-cultural practices.

Cognitive assembly: EC and FP

Concerning the product of cognitive assembly, cognitive processing is sometimes extended across the biological boundaries of the organism to include parts of the material and social environment. This is a view endorsed by all proponents of EC. However, when it comes to explaining the process of cognitive assembly—the recruitment and coordination of information-processing resources—especially Clark steps back from this distributed nexus of disparate, heterogeneous, and self-organizing elements and gives priority to the human brain:

“Human cognitive processing (sometimes) literally extends into the environment surrounding the organism. But the organism (and within the organism, the brain/CNS) remains the core and currently the most active element. Cognition is organism centered even when it is not organism bound.” (Clark 2008, 139; italics added).

“Just as it is the spider body that spins and maintains the web that then (following Dawkins 1982) constitutes part of its own extended phenotype, so it is the biological organism that spins, selects, or maintains the webs of cognitive scaffolding that participate in the extended machinery of its own thought and reason.” (2008, p. 123; italics added).

Clark coins this view the hypothesis of organism-centered cognition (HOC). Unlike diachronic time-scales, the HOC is intended to operate synchronically. It is in the process of recruitment and assembly here-and-now that Clark thinks that the brain is primarily the responsible mechanism (Clark 2011, p. 459). In his (2008), Clark refers to a series of experiments by Gray and Fu (2004) in order to empirically substantiate the HOC. What these studies show is that the neural control system is indifferent to the source of information—whether it is in-the- head or in-the-world located information—such that the brain (based on the time–cost involved in information retrieval) determines whatever mix of resources are recruited to solve a problem. Following Gray and colleagues, Clark takes the lesson from these experiments to be that “our problem-solving performances take shape according to some cost function or functions that, in the typical course of events, accord no special status or privilege to specific types of operations (motoric, perceptual, introspective) or modes of encoding (in the head or in the world).” (Clark 2008, 121) As Clark sums up the results to support the HOC:

“Concerning the process of recruitment, it is indeed the biological brain (or perhaps some of its subsystems) that is in the driver’s seat. That is to say, it is indeed some neurally based process of recruitment that (following Gray et al.) turns out to be so pointedly unbiased regarding the use of inner versus outer circuits, storage, and operations.” (2008, 122).

Part #1 of FP in cognitive assembly

Like Dennett (1991), who sees the human mind as consisting of a semianarchic coalition of elements, interconnected and competing, Clark sacks the idea of an “all-powerful, hidden agent inside the brain whose job is to do all real thinking and which is able to intelligently organize those teams of internal and external supporting structure.” (Clark 2008, p. 136) Clark thinks that the “control is itself fragmented and distributed, allowing different inner resources to interact with, or call upon, different external resources without such activity being routed via the bottleneck of conscious deliberation or the intervention of an all-seeing, all-orchestrating inner executive.” (2008, pp. 136–7) Consequently, Clark decentralizes the cognitive agent responsible for cognitive assembly into a fragmented and distributed mix of internal self-organizing processes.

My claim is that it is at this particular stage that the first part of FP emerges in Clark’s account, since when decentralizing the inner homunculus to self-organizing and fragmented processes, he ends up bounding all of these dynamical elements within the biological boundaries of the organism: “it is the biological organism that [assembles] the […] extended machinery […].” (Clark 2008, p. 123) FP is the failure of decentralizing the notion of cognitive agency sufficiently to include patterned socio-cultured practices. Following Hutchins (2011a), the individualist assumption of cognitive agency in the HOC threatens to isolate the activity of the brain from the dynamics of socio-cultural practices on both synchronic and diachronic time-scales (2011a, p. 411). That is, if the boundaries of cognitive agency do not outweigh the individual organism, and if the dynamics of socio-cultural practices in cognitive assembly are taken to play only secondary roles, then the dynamics of cognitive assembly must derive from the biological brain and body. Hence, by endorsing an individualist notion of cognitive agency, Clark ends up privileging the organism-centered mechanisms as the driving and most active loci around which extended architectures organize and dissolve.

Part #2 of FP in cognitive assembly

According to Hutchins, “A straightforward way to deal with [cognitive assembly] is to abandon the assumption that the biological brain is the essential element. Doing so, of course, requires that one look elsewhere for the apparently impartial forces that assemble cognitive systems.” (2011a, p. 439) In shared spirits with Roepstorff et al. (2010), the proposal favored by Hutchins is that a “good start on understanding this process of recruitment would be to notice the role of cultural practices in the orchestration of soft-assembly of extended systems.” (2011a, p. 440) As Hutchins emphasizes, “The ecological assemblies of human cognition make pervasive use of cultural products.” (2011a, p. 445)

My claim is that we can find a second part of FP in cognitive assembly when considering how Clark responds to Hutchins’ challenge. First, Clark argues that Hutchins fails to appropriately separate the different time-scales of the processes concerned in cognitive assembly: to distinguish evolutionary and developmental from synchronic time-scales. As Clark says: “I think Hutchins is failing to attend to important differences concerning the shape and timescale of the processes concerned. My own targets, in the discussion in SSM of cognitive assembly, were processes operating in the here-and-now.” (2008, p. 459) Clark does not want to deny that some spinning is done by socio-cultural practices on diachronic time-scales. As Clark says, cultural practices really do provide “me with both a prestructured recipe for success, a well-honed cultural practice (schooling) to help me benefit from that recipe, and a pre-selected set of supporting materials and structures (pen, paper) all ripe for assembly into a new problem-solving whole.” (2011, 459) However, what Clark wants to deny is that the assembly process itself is being spun and maintained by cultural practices synchronically: “even here, it is still the individual biological brains […] that are, in the here-and-now, the most active orchestrating elements in this process.” (Clark 2011, p. 459) FP here is a consequence of Clark’s view of cognitive agency, since neglecting to take seriously the dynamical role of socio-cultural practices in the here-and-now, precludes taking seriously the dynamical contribution of socio-cultural practices on diachronic time-scales as well.

Cognitive assembly: EC and DP

Case #1: mental institutions

Clark’s intuition is that the dynamical processes responsible for cognitive assembly reside at only one level: the individual organism. Maybe it is possible to find individual cases where the appropriate level is indeed the individual organism. However, the problem with encapsulating cognitive agency at the individual level is that we can find cases where the individual level is insufficient in order to explain the assembly of extended cognitive systems (Protevi 2009; Theiner et al. 2010). On Hollan et al.’s account: in “distributed cognition, one expects to find a system that can dynamically configure itself to bring subsystems into coordination to accomplish various functions.” (2000, p. 175) Consequently, in such cases, cognitive agency outweighs the boundaries of skin-and-skull to include distributed processes and activities across time, space, and people. Such a scenario has recently been proposed by Gallagher and Crisafi (2009) in their discussion of mental institutions. Essentially, if the cognitively relevant and reliable interactions bring a myriad of subsystems into play, these subsystems being largely distributed across people, artifacts, patterned practices, and the activities of previous generations, and if all these subsystems are sufficiently integrated, it becomes increasingly difficult to single out the individual level as the most active element in the assembly process.

In a similar way to Hutchins’ “observation that the outcomes that mattered to [ship navigation] were not determined by the cognitive properties of any single navigator, but instead were the product of the interactions of several navigators with each other and with a complex suite of tools.” (Hollan et al. 2000, p. 183), Gallagher and Crisafi show that in some cases the individual brain only performs a sub-set of actions, while others are enacted and orchestrated by socio-cultural practices and cognitive technology. In considering legal institutions and the practice of law as mechanisms for carrying out cognitive practices, Gallagher and Crisafi argue that we should view all of the following three cases on a par with respect to cognitive effort on behalf of the individual or distributed system (2009, p. 47):
  1. 1.

    A person, let us call her Alexis, is given a set of facts and is presented a collection of evidence and is asked to judge on the basis of her own subjective sense of fairness, the legitimacy of a certain claim being made. To make her judgment Alexis must weigh the facts and consider the evidence entirely in her own head, without help or interference from others. In this process she draws up and considers three questions about the facts, tries to answer them the best she can, and that makes her decision.

  2. 2.

    Alexis is given a set of facts and is presented a collection of evidence and is asked to judge the legitimacy of a certain claim that is being made. This time, however, she is given the three questions by a group of experts who provide a set of possible answers from which she may choose. She may also decide to formulate her own set of answers.

  3. 3.

    Alexis is given a set of facts and is presented a collection of evidence and is asked to judge the legitimacy of a certain claim that is being made. As in (2), she is asked to consider the same three questions by a group of experts who inform her of a set of pre-established possible answers from which she can choose, and a set of pre-established rules she must follow in answering the questions. The rules specify that she must answer each question in one of only two ways, choosing from the set of possible answers. Alexis is not allowed to formulate her own alternative set of answers.


According to Gallagher and Crisafi, cases (2) and (3) are cognitive, in the sense that they are cognition producing—they involve information-processing resources, problem solving, reasoning, etc.—and they are cognition produced—they are assembled from the activity of many people at both diachronic and synchronic time-scales and their cognitive artifacts. On the issue of cognitive assembly—what or who it is that spins, selects, and coordinates the assembly process in the case of legal practices—this could not take place primarily in the individual biological organism. In fact, patterned practices would seem to do much of the spinning and maintaining in the case of legal practices. Essentially, DP-driven explanations of cognitive assembly enable at least two things: first, decentralizing cognitive agency to include socio-cultural practices; and second, extended cognitive systems that exist at a larger scale than merely individualist approaches focusing on integration or incorporation of tools into the profile of a single individual.

Case #2: memory in Elizabethan theater companies

Consider now Tribble’s (2005) impressive analysis of collective memory or joint action in Elizabethan theater companies (Sutton 2010). Applying the framework of distributed cognition (Hutchins 1995) to understand the processes of collective action and memory in Elizabethan theater companies, Tribble provides a novel and highly impressive analysis of what not only have impressed but (more importantly) puzzled historians of English drama.

These companies would perform a staggering number of plays; up to six different plays per week; and without frequent time for rehearsals and the full texts of these plays. As Tribble informs us, “between 1594 and 1597 a leading player such as Edward Alleyn has to secure and retain command of about 71 different roles, of which number 52 or 53 were newly learned.” (2005, p. 136) Given the rather staggering demands on human memory, this raises the puzzle: how did these groups actually manage to learn or memorize so many plays with so many changing roles? Tribble offers an analysis of how socio-cultural practices, including the regular use of artefacts “form elements of a cognitive structure that, in constraining and limiting, also enables an extra-ordinary level of achievement.” (2005, p. 142) Especially the role of stage doors and the playing platform; plots or scripts; and a culturally situated, intergenerationally mediated apprenticeship system is analyzed. In the study, the stage doors play a powerful double-role. First, over diachronic time, the interactions within the cognitive niche alter the informational structure of the situation. By continuously exploiting the size and locational distribution of doors, the local environment comes to drive and constrain, orient and re-orient, the behavioral opportunities of the actors. Second, in line with Hutchins (2011a) and Sterelny (2010), Tribble shows how the ability of the actors to establish and maintain their grasp of the situation is acquired through participation in socio-cultural practices in inter-personal space. A third component of Tribble’s analysis pertains to how a specific culturally and cross-generationally mediated apprenticeship system enculturates the process of inclusion into Elizabethan theater practices (Tribble 2005, pp. 153–55). The apprenticeship system is hierarchically structured. Following Hutchins’ (1995) analysis of a similar system aboard naval vessels, Tribble shows how novices come to acquire the necessary skills by being thrown into situations—guided by more experienced actors—were it becomes a real possibility for the novice actors to perform successfully both computationally and theatrically.

It is not clear why socio-cultural practices (including socio-cultural technologies) could not play a dynamically orchestrating role in the assembly of extended cognitive systems—not even if these are partly constructed over diachronic time-scales. Right here, however, we are moving beyond the HOC—it is very hard (if not impossible) to shoehorn collaborative memory in the case of Elizabethan theater groups with the HOC. As Hutchins pinpoints, “few of the dynamic loops that link people to their environments are invented by the people who exploit them.” (2011a, p. 441) In Elizabethan theater companies, a substantial part of the assembly process seems to be performed by dynamics of an apprenticeship system functioning on both intergenerational time-scales and synchronic time-scales. Moreover, when Sterelny remarks that it is true when Clark claims “that an organic brain chooses and assembles the resources that make much problem solving possible […], but that in many critical cases, those brains belong to members of the previous generation, not to the agent faced with the problem […], (2010, p. 479),” he means quite literally that the cultural practices of prior generations shape our epistemic access to the world by constraining “what to attend to and to see when so attending.” (Hutchins 2011a, p. 441) Consequently, the most active orchestrating elements in the process of joint action and memory in Elizabethan theater companies outweigh the organism.


In this paper, I have explored some of the paths a distinctive third wave of extended cognition might take. The critical-constructive view offered is that there is room enough in extended cognition to do so without discrediting the validity of extended cognition in general. In cognitive transformation cases, I argued that a potentially problematic assumption concerning the complementarity framework is that it presupposes that non-biological resources augment but do not transform the brain’s representational capacities. In contrast to this tendency, I have attempted to show that a much more dynamical model is available for the extended cognition theorist, namely to view the enculturation processes on the brain as transforming the brain’s representational capacities. In cognitive assembly cases, I argued that another problematic tendency is an overly individualistic notion of cognitive agency—one that places all the active and dynamical elements inside the organism, thereby treating socio-cultural practices as only indirectly involved in the assembly of extended cognitive systems. In contrast to this individualistic starting point, I have argued that a much more distributed model is available to the fan of extended cognition, namely to conceive of cognitive agency as socio-culturally distributed across social groups, cognitive tools, and patterned practices.


An Australian Research Council (ARC) Discovery Project Grant (DP1095109) and an International Macquarie University Research Excellence Scholarship (no. 2011180) has funded this project. Thanks to Richard Menary, John Sutton, and Will Newsome for helpful discussions. I am especially grateful for the highly constructive comments made by two anonymous referees. Any mistakes are mine and mine alone.

Copyright information

© Springer Science+Business Media B.V. 2011