Mental Representation

Mental representation is the idea that the mind carries something—a sign, a picture, a code, a map—that stands for things in the world.

12 frameworks, from 1250 to today · Philosophy Of Mind · Philosophy

still a working view ended · each leads to its entry below

Overview

Mental representation is the idea that the mind carries something—a sign, a picture, a code, a map—that stands for things in the world1. When you believe it will rain, your belief is about rain; when you see a dog, your perceptual state is of a dog. This aboutness is what philosophers call intentionality, and the central question of this subfield is how a physical system, such as a brain, can manage it. The stakes are large. If thoughts are representations, we can explain how minds are useful—they track the world and guide action. If they are not, it is unclear what distinguishes a thought from a stomach rumble. And if representations can be carried by silicon as well as by neurons, then minds might not be exclusively biological. The history of the subfield is a long argument about what representations are, where they get their content, and whether we need them at all.

Images, Ideas, and the Classical Picture

The oldest approach treats mental representation as a kind of inner picture or likeness. Aristotle held that perception takes on the form of the perceived object without its matter, in something like the way wax takes the shape of a seal. Medieval scholastics developed this into a theory of species—likenesses that the senses and intellect receive. The early modern period kept the picture but made it private: Descartes's ideas are mental contents directly present to the mind, and Locke's "ideas" are whatever the mind is thinking about, derived ultimately from sensation and reflection.

This representational theory of mind came with a problem. If what we directly perceive is an inner image, how do we know it resembles anything outside? The seventeenth-century occasionalists and later Berkeley pressed this worry hard. It also generated a puzzle about what does the representing. A picture in the head is just more physical stuff; something must interpret it. The classical picture struggled to say what that interpreter is, short of a little person—the notorious homunculus—which merely postpones the question.

Ordinary Language and the Attack on Inner Pictures

By the mid-twentieth century, a powerful reaction set in. Wittgenstein's later work, especially the Philosophical Investigations (published 1953), attacked the idea of a private inner object that gives a word its meaning. On his account, meaning is anchored in public use, in the practices of a community; an inner picture that no one else could check does no explanatory work. Gilbert Ryle's The Concept of Mind (1949) made a related point in a different idiom, arguing that talk of inner representations often rests on a category mistake—treating a disposition or a skill as if it were a hidden thing.

This ordinary-language critique did not kill the representational theory, but it changed the terms of debate. Any account of mental representation would now have to say how inner states connect to public behavior and public criteria, not just how they resemble objects.

The Computational Turn and the Language of Thought

The decisive break came with the cognitive revolution of the 1950s and 1960s. The computer offered a new model: a physical system could manipulate symbols according to rules and thereby compute. If the mind works like this, representation need not be a picture; it can be a symbol in a formal system.

Hilary Putnam's early work on functionalism and his 1960 paper "Minds and Machines" helped launch this view, and Jerry Fodor gave it its most influential statement. In The Language of Thought (1975), Fodor argued that thought has a compositional structure: complex thoughts are built from simpler constituents, and the content of the whole depends on the content of the parts and their arrangement. To explain this, he posited a language of thought—an internal symbol system with a syntax, "mentalese," in which thinking literally consists in manipulating symbols12. The theory's strength was that it explained productivity (we can think infinitely many novel thoughts) and systematicity (if you can think "John loves Mary," you can think "Mary loves John").

Fodor combined this with a representational theory of mind: to have a propositional attitude such as believing that p is to stand in a computational relation to a mental representation with the content p. The theory was powerful, but it left the hardest question open: what makes a mental symbol mean something?

Naturalizing Content: Information and Teleology

The 1980s were dominated by attempts to answer that question without appeal to an interpreter—to "naturalize" intentionality.

Informational semantics, associated above all with Fred Dretske's Knowledge and the Flow of Information (1981), proposed that a state represents whatever it carries information about, in roughly the sense of information theory: a state covaries reliably with a condition in the world. A mental state is about dogs if it is reliably caused by dogs. The difficulty is the problem of misrepresentation. If representation is mere covariation, a state can never be wrong—but beliefs can be false. Dretske and others wrestled with this, and one influential response appealed to the function of the state.

Teleosemantics, developed by Ruth Millikan in Language, Thought, and Other Biological Categories (1984), and later by David Papineau and others, grounded content in biological function: a state represents what it was selected to indicate. Just as a heart is for pumping blood, a perceptual state is for detecting certain conditions; misrepresentation occurs when the mechanism fails to perform its function. This gave a naturalistic account of error, but raised questions about how much of human thought is shaped by evolutionary selection.

Wide Content and the Externalist Turn

A parallel development challenged the assumption that mental content is fixed entirely by what is inside the head. In The Meaning of "Meaning" (1975) and later work, Hilary Putnam introduced the famous Twin Earth thought experiment: a person on Earth and a molecule-for-molecule duplicate on Twin Earth, where the watery stuff is not H₂O but a different compound, would both say "water," but arguably their words and thoughts mean different things. Tyler Burge extended this content externalism to many ordinary concepts, arguing that what we believe depends partly on the linguistic and social environment.

If externalism is right, the boundaries of the mind do not coincide with the boundaries of the skull. This set the stage for a more radical proposal.

The Extended Mind and Enactivism

In 1998, Andy Clark and David Chalmers published "The Extended Mind," arguing that cognitive processes can extend into the environment3. Their central example was Otto, a patient with Alzheimer's who relies on a notebook to store and retrieve information; Clark and Chalmers argued that, under the right conditions, the notebook plays the same functional role as biological memory and should count as part of his cognitive system3. The extended mind thesis was controversial—critics asked whether coupling to an external resource is enough for genuine mental representation—but it forced the field to reconsider where representation lives.

A different challenge came from enactivism, developed by Francisco Varela, Evan Thompson, and Eleanor Rosch in The Embodied Mind (1991), and later by Daniel Hutto and Erik Myin. Enactivists argue that cognition is fundamentally sensorimotor engagement with the world, not the manipulation of inner representations. Perceiving is something we do, not something that happens inside us. Some enactivists allow minimal representations; others (the "radical" wing) reject them outright. The debate is whether representation-talk is a useful explanatory tool or a legacy of the inner-picture tradition the ordinary-language critics attacked.

Connectionism and the Return of the Network

Meanwhile, the symbol-manipulation picture was being challenged from within cognitive science. Connectionism—the use of artificial neural networks with distributed representations—emerged in the 1940s and 1950s with McCulloch and Pitts's logical neurons and Rosenblatt's perceptron, and returned powerfully in the 1980s with work by Rumelhart, McClelland, and the Parallel Distributed Processing group4.

Connectionist networks represent information not as discrete symbols but as patterns of activation across many units. This had explanatory virtues: it handles pattern recognition and graceful degradation (networks degrade gradually rather than crashing) better than classical symbol systems. Fodor and Zenon Pylyshyn's 1988 book Connectionism and Cognitive Architecture: A Critical Analysis pressed the objection that connectionist networks cannot account for the systematicity and compositionality of thought without implicitly containing a symbol system. The debate reshaped both sides: later connectionist and hybrid models attempted to show how distributed representations could support structured thought.

Predictive Processing and the Representational Present

The most recent major framework is predictive processing, developed in the 2000s and 2010s by Karl Friston, Andy Clark, Jakob Hohwy, and others. On this view, the brain is constantly generating predictions about the causes of its sensory input and updating them as prediction errors arrive5. Perception is "controlled hallucination" corrected by the world; action is the fulfillment of predictions. The framework is representational in a distinctive sense—the brain encodes probabilistic models of the world—but it is also deeply embodied and action-oriented, drawing together threads from connectionism, enactivism, and the Bayesian turn in cognitive science.

The field today is pluralistic. Classical computationalism and the language of thought remain live research programs, defended by Fodor's successors and by philosophers of cognitive science such as Zenon Pylyshyn. Teleosemantics and informational semantics continue to be refined. Externalism about content is widely, though not universally, accepted, and the extended-mind debate has matured into careful work on the conditions under which environmental resources count as part of cognition. Enactivism and predictive processing represent the most active frontier, with a productive tension between those who see representation as the core explanatory posited and those who see it as a largely dispensable posit. The open questions are the old ones in new form: what exactly is a mental representation, what gives it content, and whether explaining the mind requires positing representations at all.

References

  1. Mental representation (introduction, Representational theories of mind), Wikipedia
  2. Language of thought hypothesis, Wikipedia
  3. Extended mind thesis ("The Extended Mind"), Wikipedia
  4. Connectionism, Wikipedia
  5. Predictive coding, Wikipedia

The frameworks of Mental Representation

Each framework is a way of working the field has used: when it held, and the question it answers.

  1. 1250-Present

    Intentionality

    How can a mental state be about something?

    Fly to its tower in Nousfarer

  2. 1600-1900

    Representational Theory of Mind

    What are the contents the mind directly grasps?

    Fly to its tower in Nousfarer

  3. 1949-1975

    Ordinary language philosophy

    How does public use fix the meaning of words?

    Fly to its tower in Nousfarer

  4. 1960-Present

    Computational Theory of Mind

    How does a physical system compute with symbols?

    Fly to its tower in Nousfarer

  5. 1975-Present

    Anti-individualism

    Does what we think depend on our environment?

    Fly to its tower in Nousfarer

  6. 1975-Present

    Language of thought hypothesis

    In what inner code are thoughts written?

    Fly to its tower in Nousfarer

  7. 1981-Present

    Informational Semantics

    Does a state mean whatever it reliably indicates?

    Fly to its tower in Nousfarer

  8. 1984-Present

    Teleosemantics

    Does biological function fix what a state represents?

    Fly to its tower in Nousfarer

  9. 1986-Present

    Connectionism

    Can networks of units represent without symbols?

    Fly to its tower in Nousfarer

  10. 1991-Present

    Enactivism

    Is cognition sensorimotor action rather than inner representation?

    Fly to its tower in Nousfarer

  11. 1998-Present

    Extended Mind

    Do cognitive processes extend beyond the skull?

    Fly to its tower in Nousfarer

  12. 2005-Present

    Predictive coding

    Does the brain minimize prediction error?

    Fly to its tower in Nousfarer

How the ideas connect

Each arrow runs from an earlier idea to the later one it shaped. The list below says what passed between them.

IntentionalityRepresentational Theory of MindOrdinary language philosophyComputational Theory of MindAnti-individualismLanguage of thought hypothesisInformational SemanticsTeleosemanticsConnectionismEnactivismExtended MindPredictive coding

Beyond the field

Where Mental Representation's ideas came from in other fields, and what they led to there: the strongest 12 of 37 links.

A city in Nousfarer: a red-carpeted stair climbing terraces to a crowned tower on an acropolis, more towers round it, forest and snowy mountains beyond.
A city in Nousfarer, the game played in the browser.

Mental Representation is a planet in Nousfarer

Its city has a tower for each of these 12 frameworks. Land, walk in, and climb one concept at a time.

Land on Mental Representation

Related fields

Fields its ideas are linked to

More in Philosophy Of Mind

See where Mental Representation sits on the Map of Human Knowledge