Embodied, embedded, enacted, extended cognition

Intuition Beginner

A person reaches for a coffee cup. The cup has a handle; the hand is shaped to grasp handles; the table is at waist height; the room is lit so the cup stands out. Cognition, in one orthodox picture, is what happens in the brain after the eyes report the cup and before the hand moves: an internal model computes a plan and issues motor commands. The embodied reply is that this "after and before" framing leaves out almost everything. The hand's shape, the table's height, the handle's affordance, and the lighting are not inputs to cognition but parts of it. The thinking happens across the brain, the body, and the surroundings.

The thesis has four strands that share a first letter. Embodied: cognition depends on the kind of body an organism has; a bat's echolocating mind is not a rearranged human mind. Embedded: cognition unfolds in an environment that constrains and scaffolds it; an office worker thinks differently with paper and Post-its than without. Enacted: cognition is something an organism does — it acts to find out about the world rather than constructing an inner picture from sensory inputs. Extended: when the environment reliably does part of the cognitive work, that part counts as cognitive too — a notebook used to remember counts as part of the memory system.

Why this matters: it changes what counts as a cognitive explanation. If cognition extends into the body and the tools, then a theory of mind that looks only at neural firing patterns is looking in the wrong place. The four strands push philosophy of mind outward, from a private inner theatre to a regulated coupling of organism and world.

Visual Beginner

The picture shows two panels. The left panel depicts the orthodox "sandwich" picture: perception comes in (an arrow from eye to brain), cognition happens in a central box labelled "reasoning and planning", and action goes out (an arrow from brain to hand). The middle box does the real work. The right panel depicts the 4E picture: a person using a notebook, the notebook connected by a bidirectional arrow to the person, the person connected to a cup on a table by reaching hands and directed gaze. There is no central reasoning box; cognition is the whole loop.

The contrast is structural. On the left, mind is a box; on the right, mind is a loop. The 4E programme argues that only the loop picture can account for what cognition actually does.

Worked example Beginner

Work through Clark and Chalmers's Otto example, the cleanest case for the extended mind.

Step 1. Consider two people. Inga has intact biological memory. She hears of an exhibition at the Museum of Modern Art, recalls the address from her memory, and walks to 11 West 53rd Street. Otto has Alzheimer's disease and carries a notebook in which he records every new fact. He hears of the exhibition, looks up the address in his notebook, and walks to 11 West 53rd Street.

Step 2. Note the functional parallel. Both Otto and Inga have a store of information. Both consult it when they need an address. Both then act on the retrieved content. Both believe the museum is at 11 West 53rd Street before they consult their store, and both rely on the store to make the belief usable.

Step 3. Apply the parity principle. The principle states that if a part of the world functions as a cognitive process would, and we would have counted that process as cognitive if it had been in the head, then we should count the external part as cognitive too. Otto's notebook functions as Inga's biological memory. By parity, the notebook is part of Otto's memory system; his beliefs are constituted by the notebook's entries, not merely caused by them.

What this tells us: the location of a cognitive process (in neurons versus in a notebook) does not by itself settle whether the process is cognitive. Function and reliable access do. The Otto argument is the flagship of the extended mind and the load-bearing thought experiment of the 4E programme.

Check your understanding Beginner

Formal definition Intermediate+

The 4E programme groups four overlapping theses about the locus and nature of cognition. Each is defined against the computational-representational theory of mind (CRTM), on which cognition is the manipulation of internal symbols according to syntactic rules, with perception supplying input and action consuming output.

Definition (Embodied cognition). A cognitive system is embodied when the form of its cognitive processes depends constitutively on the kind of body it has. On the strong reading, the body does not merely implement a brain-designed program; the body's morphology, sensorimotor repertoire, and metabolic needs are part of what makes the processes cognitive. The thesis opposes brain-body independence, on which the same cognitive process could in principle be realised by any sufficiently equivalent substrate.

Definition (Embedded cognition). A cognitive system is embedded when its processing is scaffolded and constrained by its environment in a way that changes the computational load on the organism. A navigator using landmarks solves a simpler problem than one working blind. The environment is a resource the organism uses, but the locus of cognition remains internal.

Definition (Enactive cognition). A cognitive system is enactive when perception and action are not separable input and output phases of an internal computation but two aspects of a single sensorimotor loop. Meaning is enacted through the organism's active regulation of its coupling with the environment. Enactivism rejects the claim that perception begins with sensory inputs that must be interpreted to yield a representation; perception is already meaningful because the organism has learned the sensorimotor contingencies of its engagement.

Definition (Extended cognition). A cognitive process is extended when an external structure contributes to the process in a way that satisfies the parity principle: were the contribution done internally, it would be counted as cognitive. The external structure is then part of the cognitive system, not merely a tool the system uses.

Counterexamples to common slips Intermediate+

• Conflating embedded and extended. Embedded cognition is the conservative reading: the environment scaffolds cognition but the cognitive boundary is at the skin. Extended cognition is the radical reading: the boundary includes reliably coupled external resources. The two are distinct theses, and most critics accept embedded while rejecting extended.
• Conflating embodied and extended. Embodied cognition emphasises the role of the body's morphology; extended cognition emphasises the role of external tools. A theory can accept one without the other.
• Reading the parity principle as a causal claim. The parity principle is constitutional, not causal. It says the external resource is part of the cognitive system, not merely a cause of internal cognitive events. The disagreement between extended and internalist views is precisely about whether the notebook is constitutive of Otto's memory or merely a reliable input to it.

Key argument Intermediate+

Argument (Clark-Chalmers 1998, the parity argument for the extended mind).

1. Premise (functional equivalence). Otto's notebook plays the same functional role in his address-retrieval behaviour that Inga's biological memory plays in hers: it is a portable, reliably available, automatically consulted store of information that Otto endorses as his own.

2. Premise (parity). If a process were counted as cognitive when carried out in the head, then an external process meeting the same functional description should also count as cognitive. The skin is not a cognitive magic boundary.

3. Intermediate conclusion. Otto's notebook satisfies the functional description of a memory store. By parity, the notebook is part of Otto's memory system.

4. Conclusion (extended cognition). Therefore some cognitive processes (Otto's recall of the museum address) extend into the external environment. Cognition is not skull-bound.

Reconstruction. The argument turns on a single normative principle: the question of what counts as cognitive should be answered by functional role, not by anatomical location. If function fixes cognitive status, and Otto's notebook has the right function, then the notebook is cognitive. The internalist reply (Adams and Aizawa) attacks the parity principle by arguing that the mark of the cognitive involves non-derived, intrinsic content of a kind external tools lack; the notebook's entries are derived from Otto's reading, and derived content cannot constitute cognition. The extended reply is that this mark of the cognitive begs the question against functionalism, since it is precisely what the dispute is about.

Bridge. The parity argument builds toward 29.14.02 the cognitive revolution, where the computational theory of mind it challenges was first articulated, and appears again in 20.13.01 philosophy of mind foundations as the strongest externalist pressure on physicalist internalism. The foundational reason the argument cuts deep is that functionalism, the dominant philosophy of mind since the 1970s, already commits to function-over-substrate, and the parity principle is exactly that commitment applied to the skin boundary. This is exactly the move that identifies mind with functional organisation rather than neural tissue, and the bridge is from individual cognition to the wider realisation relation that a functionalist cannot easily bound — the so-called "cognitive bloat" worry that if a notebook is cognitive, so is a smartphone, a library, or the internet.

Exercises Intermediate+

Advanced results Master

Result 1 (Varela-Thompson-Rosch enactivism, 1991). Cognition is the enactment of a world and a mind through the embodied, situated activity of a living organism. The thesis couples three commitments: (a) perception is action-guided, not representation-constructing; (b) cognitive structures emerge from recurrent sensorimotor patterns; (c) the organism and environment form a coupled dynamical system whose behaviour cannot be decomposed into input, processing, and output stages without remainder. The position is grounded in Merleau-Ponty's lived-body phenomenology and in Maturana and Varela's autopoiesis — the self-producing organisation of living systems.

Result 2 (Clark-Chalmers parity principle, 1998). If, in solving a cognitive task, a part of the world functions as a process would be expected to function if it were internal, and we would have counted that internal process as cognitive, then we should count the external part as cognitive too. Applied under the four coupling conditions (portability, automatic endorsement, ease of access, trust), the principle yields the extended-mind conclusion: cognitive processes can extend into reliably coupled external resources. The principle inherits functionalism's commitment to multiple realisability and pushes it past the skin boundary.

Result 3 (Gibson's affordances, 1979). The environment offers the animal opportunities for action — surfaces afford support, objects of graspable size afford grasping, gaps afford passage. Affordances are relational: they are defined by the fit between environmental layout and the animal's body. Perception, on Gibson's view, is the direct pickup of affordances, not the construction of an inner representation from sensations; the information in the optic array already specifies the affordances, and no inferential enrichment is needed.

Result 4 (Friston's free-energy principle, 2010). A self-organising system at non-equilibrium steady state minimises the variational free energy of its sensory observations, equivalently maximising the evidence for its generative model of the world. Perception minimises free energy by updating beliefs; action minimises free energy by selecting observations consistent with the model (active inference). The principle subsumes perception, action, learning, and homeostasis under a single imperative, providing the most influential mathematical formulation of the enactive intuition that organisms regulate their coupling with the environment.

Result 5 (Lakoff-Johnson conceptual metaphor, 1980, 1999). Human abstract concepts are systematically structured by metaphors grounded in bodily experience. The source domains of these metaphors are typically sensorimotor (orientation, motion, containment, manipulation), and the target domains are abstract (time, causation, mathematics, mind). Conceptual metaphor theory provides empirical evidence that even abstract cognition bears the stamp of the body, supporting the embodied thesis against the view that higher cognition is amodal and body-independent.

Result 6 (O'Regan-Noë sensorimotor theory, 2001). Perceptual experience is constituted by the perceiver's implicit mastery of sensorimotor contingencies — the lawful ways sensory stimulation changes as the perceiver acts. The theory dissolves the "hard problem of perceptual qualia" by reframing qualities as modes of active engagement: the redness of red is what one does with red-reflecting surfaces under changes of lighting and viewpoint. The position extends enactivism from action to perceptual phenomenology.

Synthesis. The 4E programme builds toward 29.03.02 pending visual perception by recasting the perceiver as an active agent whose body and skills constitute, rather than merely inform, what is seen, and appears again in 29.14.02 the cognitive revolution as the major post-computational movement that absorbed and redirected the symbolic paradigm. The foundational reason the four strands travel together is that they share a single negative thesis — cognition is not the manipulation of inner symbols by a disembodied central processor — and a single positive thesis — cognition is the adaptive regulation of a brain-body-environment coupling. This is exactly the structure that identifies cognition with dynamical engagement rather than with computational content, and putting these together with Friston's free-energy principle, the bridge is that even the most mathematical 4E formulation reduces to a regulation imperative: the organism acts to keep itself within the states its generative model expects. The pattern generalises across the four strands — embodied cognition grounds cognition in morphology, embedded grounds it in environmental scaffolding, enacted grounds it in sensorimotor coupling, extended grounds it in reliable tool use — and the central insight that unifies them is that a cognitive system's boundary is a research question, not a metaphysical given.

Full proof set Master

Proposition (Free-energy minimisation implies active inference). Under the active-inference framework, an organism whose internal states minimise variational free energy with respect to both its beliefs (perception) and its observations (action) acts so as to bring future observations into the high-probability manifold of its generative model.

Proof. Let the organism be partitioned into internal states $I$, sensory states $S$, active states $A$, and external states $E$, with the Markov blanket $\{S,A\}$ rendering $I$ and $E$ conditionally independent: $p(I\mid E,S,A)=p(I\mid S,A)$. The variational free energy is $F(q,o)={\mathbb{E}}_q[\ln q-\ln p(s,o)]\ge -\ln p(o)$, with equality iff $q(s)=p(s\mid o)$. Minimising $F$ with respect to $q$ performs Bayesian inference over $E$ given the blanket. Minimising $F$ with respect to action $A$ changes future observations $o$; because $F\ge -\ln p(o)$, lowering $F$ through action raises the model evidence $p(o)$, which means selecting observations the generative model rated as probable. Therefore action drives the organism toward its expected states, and the perception-action loop enacts homeostatic regulation. $\square$

Proposition (Parity principle and the cognitive-bloat reductio). The parity principle, applied without restriction, entails that any reliably coupled external resource is cognitive; the cognitive-bloat objection is that this includes too much (smartphones, calculators, the internet).

Proof sketch. Let $P$ denote the parity principle and $C$ the claim that a coupled resource is cognitive. $P$ licenses the inference from functional role to cognitive status. Applied to Otto's notebook, $P\to C_{\text{notebook}}$. Applied to a smartphone running a GPS app that the user consults automatically, $P\to C_{\text{smartphone}}$. Applied to a library, $P\to C_{\text{library}}$. The bloat reductio is that there is no principled $P$-based stopping rule short of these consequences. Clark and Chalmers respond with the four coupling conditions (portability, automatic endorsement, ease of access, trust), which are intended as a principled restriction on $P$'s application; the question is whether these conditions are principled or merely ad hoc. The disagreement is ongoing and turns on whether function or constitution is the mark of the cognitive. $\square$

Connections Master

• Philosophy of mind — consciousness, physicalism, and the mental 20.13.01. The 4E programme is the strongest externalist pressure on the internalist physicalism that anchors the foundations unit. The constitutional question the foundations unit raises — what makes a mental state the state it is — is answered very differently by internalist functionalism and by extended cognition, and the disagreement is the active frontier of the field.

• Visual perception: depth cues, object recognition, top-down and bottom-up processing 29.03.02 pending. The Gibsonian and sensorimotor accounts of 4E cognition emerged from the empirical study of perception and reshaped it: affordances, optic flow, and active vision all draw the perceiver into the environment rather than isolating the retinal image. The empirical and theoretical programmes reinforce one another.

• The cognitive revolution and the computational theory of mind 29.14.02. The 4E programme is the major post-computational movement in cognitive science, absorbing the symbolic paradigm of the 1950s-70s and redirecting it toward embodied and dynamical systems accounts. The transition from symbolic AI to situated robotics (Brooks) and from classical computationalism to predictive processing is the empirical wing of the 4E critique.

Historical & philosophical context Master

The 4E programme has two principal lineages. The phenomenological lineage runs from Edmund Husserl through Maurice Merleau-Ponty, whose 1945 Phenomenologie de la Perception ^{[MerleauPonty1945]} argued that perception is structured by the perceiver's bodily engagement with the world, not by the synthesis of sensory givens under intellectual categories. Merleau-Ponty's "lived body" (corps propre) is the substrate of all experience and cannot be derived from a sensation-then-interpretation model. The ecological lineage runs through James J. Gibson, whose 1979 The Ecological Approach to Visual Perception ^[Gibson1979] argued that the optic array already specifies affordances for an animal of a given kind, so perception is direct and does not require inferential construction.

The two lineages converged in Francisco Varela, Evan Thompson, and Eleanor Rosch's 1991 The Embodied Mind ^{[VarelaThompsonRosch1991]}, which combined Merleau-Pontian phenomenology, Gibsonian ecological perception, the autopoiesis of Maturana and Varela, and Buddhist contemplative practice into the enactivist research programme. The extended-mind thesis, the most philosophically provocative strand, is due to Andy Clark and David Chalmers in their 1998 Analysis paper ^{[ClarkChalmers1998]}. The mathematical consolidation of enactivism under active inference is due to Karl Friston from 2005 onward ^{[Friston2010]}, who showed that a single free-energy minimisation imperative subsumes perception, action, learning, and homeostasis under a self-organising framework.

The conceptual-metaphor programme of George Lakoff and Mark Johnson ^{[LakoffJohnson1999]} supplies the empirical wing of the embodied thesis: experimental linguistics shows that even abstract concepts are structured by body-grounded metaphors. The deeper lineage of the 4E critique runs through Heidegger's Being and Time (1927), whose account of Dasein as a being-in-the-world already dissolved the subject-object split that the computational theory of mind would later reinstate.

Bibliography Master

@book{MerleauPonty1945,
  author = {Merleau-Ponty, M.},
  title = {Ph\'enomenologie de la Perception},
  publisher = {Gallimard},
  address = {Paris},
  year = {1945},
}

@book{Gibson1979,
  author = {Gibson, J. J.},
  title = {The Ecological Approach to Visual Perception},
  publisher = {Houghton Mifflin},
  address = {Boston},
  year = {1979},
}

@book{VarelaThompsonRosch1991,
  author = {Varela, F. J. and Thompson, E. and Rosch, E.},
  title = {The Embodied Mind: Cognitive Science and Human Experience},
  publisher = {MIT Press},
  address = {Cambridge, MA},
  year = {1991},
}

@article{ClarkChalmers1998,
  author = {Clark, A. and Chalmers, D.},
  title = {The Extended Mind},
  journal = {Analysis},
  volume = {58},
  number = {1},
  pages = {7--19},
  year = {1998},
}

@article{Friston2010,
  author = {Friston, K.},
  title = {The free-energy principle: a unified brain theory?},
  journal = {Nature Reviews Neuroscience},
  volume = {11},
  pages = {127--138},
  year = {2010},
}

@book{LakoffJohnson1999,
  author = {Lakoff, G. and Johnson, M.},
  title = {Philosophy in the Flesh: The Embodied Mind and Its Challenge to Western Thought},
  publisher = {Basic Books},
  address = {New York},
  year = {1999},
}

@article{OReganNoe2001,
  author = {O'Regan, J. K. and No\"e, A.},
  title = {A sensorimotor account of vision and visual consciousness},
  journal = {Behavioral and Brain Sciences},
  volume = {24},
  number = {5},
  pages = {939--973},
  year = {2001},
}

@article{AdamsAizawa2001,
  author = {Adams, F. and Aizawa, K.},
  title = {The bounds of cognition},
  journal = {Philosophical Psychology},
  volume = {14},
  number = {1},
  pages = {43--64},
  year = {2001},
}