Bodily cognition

Traditional cognitive science has been largely organized around the idea of the brain as a computing device and cognitive systems as functionally organized systems of data-processing. There is an emerging alternative to this paradigm that is described as “4E Cognition,” where the four “E’s” refer to cognition that is embodied, embedded, enactive, and extended. For example, there is the idea that perception of a fly ball is constituted by bodily awareness of arms and legs as well as neurophysiological information processing of visual information; that a paper scratch-pad used to assist a calculation is part of the cognitive process of calculation; or that a person’s reliance on her smartphone for remembering names incorporates the smartphone into the extended process of recognizing an acquaintance on the street.

The 4E-cognition approach is well represented in The Oxford Handbook of 4E Cognition, edited by Albert Newen, Leon de Brun, and Shaun Gallagher, which provides an exposure to a great deal of very interesting current research. The fundamental idea is the questioning of the “brain-centered” approach to cognition that has characterized much of the history of cognitive science and neuroscience — what participants refer to as “representational and computational model of cognition”; RCC. But the 4E approach rejects this paradigm for cognition. 

According to proponents of 4E cognition, however, the cognitive phenomena that are studied by modern cognitive science, such as spatial navigation, action, perception, and understanding others emotions, are in some sense all dependent on the morphological, biological, and physiological details of an agent’s body, an appropriately structured natural, technological, or social environment, and the agent’s active and embodied interaction with this environment. (kl 257)

Here is a summary statement of the chief philosophical problems raised by the theory of “4E cognition”, according to the introduction to the volume provided by Newen, de Brun, and Gallagher:

Thus, by maintaining that cognition involves extracranial bodily processes, 4E approaches depart markedly from the RCC view that the brain is the sole basis of cognitive processes. But what precisely does it mean to say that cognition involves extracranial processes? First of all, the involvement of extracranial processes can be understood in a strong and a weak way. According to the strong reading, cognitive processes are partially constituted by extracranial processes, i.e., they are essentially based on them. By contrast, according to the weak reading, they are non-constitutionally related, i.e., only causally dependent upon extracranial processes. Furthermore, cognitive processes can count as extracranial in two ways. Extracranial processes can be bodily (involving a brain–body unit) or they can be extrabodily (involving a brain–body–environment unit).


Following this line of reasoning, we can distinguish between four different claims about embodied cognition:


a. A cognitive process is strongly embodied by bodily processes if it is partially constituted by (essentially based on) processes in the body that are not in the brain;


b. A cognitive process is strongly embodied by extrabodily processes if it is partially constituted by extrabodily processes;


c. A cognitive process is weakly embodied by bodily processes if it is not partially constituted by but only partially dependent upon extracranial processes (bodily processes outside of the brain);


d. A cognitive process is weakly embodied by extrabodily processes if it is not partially constituted by but only partially dependent upon extrabodily processes. 

The last version of the claim (d) is identical with the property of being embedded, i.e., being causally dependent on extrabodily processes in the environment of the bodily system. Furthermore, being extended is a property of a cognitive process if it is at least partially constituted by extrabodily processes (b), i.e., if it extends into essentially involved extrabodily components or tools (Stephan et al. 2014; Walter 2014). (kl 259)

These are metaphysical problems on the whole: what is the status of cognition as a thing in the world, and where does it reside — in the brain, in the body, or in a complex embedded relationship with the environment? The distinction between “constituted by” and “causally affected by” is a metaphysically important one — though it isn’t entirely clear that it has empirical consequences.

Julian Kiverstein’s contribution to the volume, “Extended cognition,” appears to agree with this point about the metaphysical nature of the topic of “embedded cognition”. He distinguishes between the “embedded theory” (EMT) and “extended theories” (EXT), and proposes that the disagreement between the two families of theories hangs on “what it is for a state or process to count as cognitive” (kl 549). This is on its face a conceptual or metaphysical question, not an empirical question.

I show how there is substantial agreement in both camps about how cognitive science is to proceed. Both sides agree that the best explanation of human problem-solving will often make reference to bodily actions carried out on externally located information-bearing structures. The debates is not about how to do cognitive science. It is instead, to repeat, a debate about the mark of the cognitive: the properties that make a state or process count as being of a particular cognitive kind. (kl 590)

Embedded and extended theorists therefore agree that internal cognitive processes will often not be sufficient for explaining cognitive behaviors. (kl 654)

It might be thought to be analogous to the question, “what is the global trading network?” (GTN), and the subsequent question of whether systems of knowledge production are part of the global trading network (constitutive) or merely causally relevant to the GTN (extended causal relevance). But it is difficult to see how one could argue that there is a fact of the matter about the “reality of the global trading system” or the “mark of the cognitive”. These look like typical issues of conceptual demarcation, guided by pragmatic scientific concerns rather than empirical facts about the world.

Kiverstein addresses this issue throughout his chapter, but he arrives at what is for me an unsatisfactory reliance on a fundamental distinction between conceptual frameworks and metaphysical reality:

I agree with Sprevak, however, that the debate between EXT and EMT isn’t about the best conceptual framework for interpreting findings in cognitive science. It is a debate in metaphysics about “what makes a state or process count as mental or non-mental” (Sprevak 2010, p. 261) (kl 654)

The central claim of this chapter has been that to resolve the debate about extended cognition we will need to come up with a mark of the cognitive. We will need to say what makes a state or process count as a state or process of a particular cognitive kind. (kl 951)

But debates in metaphysics are ultimately debates about conceptual frameworks; so the distinction is not a convincing one. And, contrary to the thrust of the second quote, it is implausible to hold that there might be a definitive answer to the question of “what makes a state count as a state of a particular cognitive kind.” (Here is an earlier post on conceptual schemes and ontology; link.)

What this suggests to me is not that 4E theory is misguided in its notion that cognition is embedded, embodied, extended, and enactive; rather, my suggestion here is that the metaphysical questions about “constitution of cognition” and “the real nature of cognition” might be put aside and the empirical and systematic ways in which human cognitive processes are interwoven with extra-bodily artifacts and processes be investigated in detail.

Also interesting in the volume is Tadeusz Wiesław Zawidzki’s treatment of “mindshaping”. This topic has to do with another aspect of extended cognition, in this case the ability humans have to perceive the emotional and intentional states of other humans. Zawidski takes on the more traditional idea of “mindreading” (not the spooky kind, just the idea that human beings are hard-wired to perceive behavioral expressions of various mental states when performed by other people). He argues instead that our ability to read other people’s emotions and intentions is the result of a socially/culturally constructed set of tools that we learn. And, significantly, he argues that the ability to influence the minds of others is the crucial social-cognitive ability that underlies much that is distinctive in human history.

The mindshaping hypothesis rejects this assumption [of hardwired interpersonal cognition], and proposes an alternative. According to this alternative, our social accomplishments are not due to an individual, neurally implemented capacity to correctly represent each other’s mental states. Rather, they rely on less intellectualized and more embodied capacities to shape each other’s minds, e.g., imitation, pedagogy, and norm enforcement. We are much better mindshapers, and we spend much more of our time and energy engaged in mindshaping than any other species. Our skill at mindshaping enables us to insure that we come to have the complementary mental states required for successful, complex coordination, without requiring us to solve the intractable problem of correctly inferring the independently constituted mental states of our fellows. (chapter 39)

Here is how Zawidzki relates the mindshaping hypothesis to the 4E paradigm:

The mindshaping hypothesis is a natural ally of “4E” approaches to human social- cognition. Rather than conceptualize distinctively human social cognition as the accomplishment of computational processes implemented in the brains of individuals, involving the correct representation of mental states, the mindshaping hypothesis conceptualizes it as emerging from embodied and embedded practices of tracking and molding behavioral dispositions in situated, socio-historically and culturally specific human populations. Our socio-cognitive success depends essentially on social and hence extended facts, e.g., social models we shape each other to emulate, both concrete ones, e.g., high status individuals, and “virtual” ones, e.g., mythical ideals encoded in external symbol systems. And social cognition, according to the mindshaping hypothesis, is in a very literal sense enactive: we succeed in our socio-cognitive endeavors by cooperatively enacting roles in social structures. (chapter 39)

 This is an interesting approach to the important phenomenon of interpersonal perception. And it has immediate empirical implications: are there cross-cultural differences in “mindshaping” practices? Are there differences within a given culture according to socially relevant characteristics (gender, race, class)? Is it possible to track historical changes in the skills associated with human “mindshaping” practices? Were Victorian aristocrats different in their mindshaping capacities from their counterparts a century earlier or later?

There are many instructive implications of research within the umbrella of 4E cognitive science. But perhaps the most important is the license it gives researchers to think more broadly about knowledge, perception, intention, belief, and emotion than the narrowly neurophysiological versions of cognitive science would permit. This perspective allows researchers to pay attention to the interdependencies that exist between consciousness, thought, bodily action, joint activity, social context, and artifact that are difficult to incorporate into older cognitive theories. The model of the mind as the expression of a brain-computer-information machine is perhaps one whose time has passed. (Sorry, Alan Turing!)

Ian Hacking on natural kinds

Ian Hacking has written quite a bit on the topic of “kinds” (link), beginning with “A Tradition of Natural Kinds” in Philosophical Studies in 1991 (link) and most recently with his lecture to the Royal Institute of Philosophy in 2006 (link). He is also one of the most interesting theorists of “constructivism” — a sort of mirror opposite to the position that the world consists of things arranged in natural kinds (The Social Construction of What?). So it is worthwhile examining his view of the status of the idea of “natural kinds”.

Before we get to natural kinds, Hacking thinks it is a good idea to consider an idea that emanates from Nelson Goodman in Ways of Worldmaking, the idea of “relevant kinds”. Hacking discusses this concept at length in Social Construction (128 ff.). Fundamentally the idea of a relevant kind is an ontologically non-committal interpretation of concepts; it is a contingent and interest-driven way of classifying things in one way rather than another.

So what does the idea of a natural kind add to the notion of a relevant kind? A preliminary definition might go along these lines: a natural kind is a group of things sharing a set of properties or capacities. A natural kind is a set of things sharing a common structure or a common set of causal properties. Metal is a natural kind; green things is not. In the 1991 article Hacking lists a number of characteristics that are often thought to attach to natural kinds: independence, definability, utility, and uniqueness (110-111). The final principle is the most comprehensive, and also the least plausible:

Uniqueness. There is a unique best taxonomy in terms of natural kinds, that represents nature as it is, and reflects the network of causal laws. We do not have nor could we have a final taxonomy of anything, but any objective classification is right or wrong according as it captures part of the structure of the one true taxonomy of the universe. (111)

(Hacking explicitly rejects the uniqueness thesis.)

Hacking traces the language of kinds and natural kinds to J. S. Mill and John Venn in the middle of the nineteenth century. He quotes Peirce’s effort to improve upon Mill’s definition of natural kinds, based on the idea that the objects encompassed within a kind have important properties that are naturally related to each other:

The following definition might be proposed [for ‘real kind’]: Any class which, in addition to its defining character has another that is of permanent interest, and is common and peculiar to its members, is destined to be conserved in that ultimate conception of the universe at which we aim, and is accordingly to be called ‘real’. (119)

Here is how Hacking distinguishes between Mill and Peirce:

A Mill-Kind is a class of objects with a large or even apparently inexhaustible number of properties in common, and such that these properties are not implied by any known systematized body of law about things of this Kind. A Peirce-kind is such a class, but such that there is a systematized body of law about things of this kind, and is such that we may reasonably think that it provides explanation sketches of why things of this kind have many of their properties.

In the 2006 article Hacking offers a clear definition based on William Whewell’s reasoning:

A kind is a class denoted by a common name about which there is the possibility of general, intelligible and consistent, and probably true assertions. (13)

And here is his reading in 2006 of John Venn’s view of natural kinds:

‘There are classes of objects, each class containing a multitude of individuals more or less resembling one another […]. The uniformity that we may trace in the [statistical] results is owing, much more than is often suspected, to this arrangement of things into natural kinds, each kind containing a large number of individuals.’ (17)

Now let’s turn to Hacking’s views fifteen years later in “Natural Kinds: Rosy Dawn, Scholastic Twilight” (link). This piece extends his historical analysis of the evolution of the concept, but here Hacking also lets us know more clearly what his own view is on natural kinds. He argues for two fundamental theses:
  1. Some classifications are more natural than others, but there is no such thing as a natural kind.
  2. Many philosophical research programmes have evolved around an idea about natural kinds, but the seeds of their failure (or degeneration) were built in from the start.

The first is a declaration about the world: the world does not divide into distinct categories of things, as postulated in the uniqueness principle above. The second is a declaration about a philosophical tradition: the line of thought he scrutinizes leading from Mill through Peirce and Russell to Kripke and Quine has led to irresolvable inconsistencies. The topic has become a degenerating research programme.

One of the most interesting recent views on kinds that Hacking discusses is that of Brian Ellis in Scientific Essentialism. Hacking summarizes Ellis’s essentialism in these terms:

It emphasizes three types of natural kinds. Substantival natural kinds include elements, fundamental particles, inert gases, sodium salts, sodium chloride molecules, and electrons. Dynamic natural kinds include causal interactions, energy transfer processes, ionizations, diffractions, H2 +Cl2 ⇒ 2HCl, and photon emission at λ = 5461Å from an atom of mercury. Natural property kinds include dispositional properties, categorical properties, and spatial and temporal relations; mass, charge; unit mass, charge of 2e, unit field strength, and spherical shape. (27)

Also interesting is Richard Boyd’s “homoeostatic property cluster kinds”, a concept that seems to apply best in evolutionary biology. Boyd’s view appears in “Realism, Anti-Foundationalism and the Enthusiasm for Natural Kinds” (link), a response to Hacking’s 1991 article.  Hacking summarizes Boyd’s view in these terms: “In his analysis, kinds, and in particular species, are groups that persist in a fairly long haul. The properties that characterize a species form a cluster. No distinctive property may be common to all members of the species, but the cluster is good for survival” (30).

So what is Hacking’s view, all things considered? He is fairly consistent from 1991 to 2006. Hacking’s view in 1991 seems to have a pragmatist and anti-realist orientation: things are organized into kinds so as to permit human beings to use and manipulate them. Kinds, uses, and crafts are intimately related.

It is important that some kinds are essential to some crafts. Those are the kinds that we can do things with. It is important that some kinds are important for knowing what to expect from the fauna and flora of the region in which we live.

And in 2006 he ends the discussion with this conclusion:

Although one may judge that some classifications are more natural than others, there is neither a precise nor a vague class of classifications that may usefully be called the class of natural kinds. A stipulative definition, that picks out some precise or fuzzy class and defines it as the class of natural kinds, serves no purpose, given that there are so many competing visions of what the natural kinds are. In short, despite the honourable tradition of kinds and natural kinds that reaches back to 1840, there is no such thing as a natural kind. (35)

So Hacking’s view is a kind of conceptual constructivism. We construct schemes of classification for various pragmatic purposes — artisanship, agriculture, forest and wildlife management. Schemes have advantages and disadvantages. And there is no definable sense in which one scheme is uniquely best, given everything that nature, biology, and society presents us with.

I’ve argued for a long time that there are no “social kinds” (link). My fundamental reason for this conclusion is somewhat different from Hacking’s line of thought: I emphasize the fundamental heterogeneity and plasticity of social objects, leading to the result that there is substantial variation across the members or instances of a social concept (state, revolution, riot, financial crisis). Social things do not have essential natures, and they do not maintain their properties rigidly over time. So we are best advised to regard sociological concepts in a contingent and pragmatic way — as nominal schemes for identifying social events and structures of interest, without presuming that they have fundamental and essential properties in common.