Representation: How One Thing Stands for Another
Published:
A map represents a city.
A photograph represents a scene.
A word represents a concept or object.
A mathematical model represents a system.
Representation is one of the deepest tools of intelligence because it allows one thing to stand in for another.
But what makes representation possible?
The answer is not one single relation.
Representation can depend on:
- resemblance,
- convention,
- causal connection,
- structural correspondence,
- use.
Representation Is Not Identity
A map is not the territory.
A word is not its referent.
A model is not the system it describes.
Representation works precisely because one thing can substitute for another in thought or action.
Confusing representation with identity leads to category mistakes.
Resemblance
Some representations work through resemblance.
A portrait resembles a person.
A scale model resembles a building.
A map preserves spatial relations.
But resemblance is neither necessary nor sufficient.
A cloud can resemble a face without representing one.
Convention
Words often represent through convention.
There is nothing tree-like about the sound:
“tree.”
Speakers learn the mapping.
The relation is social and historical.
Convention can create representation without physical resemblance.
Causal Connection
Some representations arise through causal traces.
A footprint represents the passage of an animal.
A photograph represents a scene partly because light from the scene caused the image.
A thermometer reading represents temperature through a causal measurement process.
Structural Correspondence
A representation can preserve relations rather than appearance.
A subway map may distort geographic distances while preserving station order and connectivity.
Its value lies in structure.
Representation can therefore be isomorphic or partially structure-preserving without being visually realistic.
Models
Scientific models represent systems selectively.
A model may preserve:
- causal structure,
- geometry,
- statistical relationships,
- dynamics.
It intentionally ignores other details.
Good representation is often selective.
Maps
A map that included every detail of a territory would be useless.
Its power comes from abstraction.
Roads matter.
Individual stones usually do not.
Representation compresses reality by choosing relevant features.
Scale
Representation depends on scale.
A world map represents countries.
A city map represents streets.
A floor plan represents rooms.
No single representation is best for every purpose.
The question determines the abstraction.
Perspective
A photograph represents from one viewpoint.
A map may use a projection.
A graph emphasizes some variables.
Representation is always shaped by perspective and method.
This does not make it arbitrary.
Different representations can be more or less accurate for specific purposes.
Fidelity
A representation can be evaluated by fidelity.
Does it preserve the relevant properties of the target?
A medical image may sacrifice color realism but preserve tissue contrast.
A circuit diagram ignores physical layout while preserving electrical relations.
Fidelity is purpose-relative.
Abstraction
Abstraction removes detail.
The letter:
A
does not depend on exact handwriting.
Many shapes count as the same letter.
Representation requires treating different physical tokens as instances of one abstract type.
Categories
Representations often depend on categorization.
A traffic sign depicts a pedestrian using a simplified figure.
It does not represent one particular person.
It represents a category.
Abstraction allows one symbol to apply to many cases.
Internal Representations
Cognitive science often speaks of mental representations.
A brain may represent:
- location,
- object identity,
- goals,
- expected outcomes.
But the exact form of mental representation is debated.
It may involve:
- symbolic structures,
- distributed neural patterns,
- sensorimotor states.
Neural Representation
A neuron or population may respond systematically to certain features.
For example:
- orientation,
- location,
- sound frequency.
Scientists may say the activity represents that feature.
But representation here is inferred from reliable relations and functional role.
The neuron does not contain a tiny picture.
Representation and Action
Representation is not only about passive description.
A navigation system represents roads so that an agent can choose routes.
A motor plan represents possible movement.
Representations often matter because they guide action.
Misrepresentation
A representation can be wrong.
A map can contain an incorrect road.
A belief can misdescribe the world.
A sensor can report the wrong temperature.
The possibility of error is a hallmark of genuine representation.
Natural Signs vs Symbols
Smoke can indicate fire without convention.
A word represents through learned convention.
Both carry information.
But their representational mechanisms differ.
This is why semiotics distinguishes different sign relations.
Representation Without a Mind?
Can a system represent something without a conscious interpreter?
Biology suggests yes.
DNA sequences and cellular signals have functional roles.
A receptor responds differently depending on molecular input.
The representational relation may be implemented by a causal system.
Teleosemantics
One philosophical approach, teleosemantics, links representation to biological function shaped by evolution.
A signal represents something because systems evolved to respond to it in ways connected to survival or reproduction.
This attempts to naturalize meaning.
It remains debated.
Representation in Machines
Machines use representations constantly.
A memory address represents a location.
A bit pattern represents a number.
A data structure represents relationships.
The machine need not be conscious.
Representation can be operational.
Data Structures
A graph data structure can represent:
- roads,
- friendships,
- dependencies,
- computer networks.
The same formal structure can stand for very different domains.
Interpretation determines the referent.
This shows how representation separates structure from content.
Mathematical Representation
Mathematics represents quantities and relations abstractly.
The equation:
[ F=ma ]
represents a relation among force, mass, and acceleration.
The symbols are not physical forces.
They encode a structure that can apply to many systems.
Representation and Truth
A representation can be:
- accurate,
- inaccurate,
- incomplete,
- idealized.
Truth is not identical to detail.
An idealized model can be highly informative even when literally false in some respects.
A frictionless plane does not exist.
The model can still explain motion.
Multiple Representations
The same system can have many valid representations.
A molecule can be represented by:
- chemical formula,
- structural diagram,
- 3D model,
- quantum state.
Each reveals different information.
Representational pluralism is often scientifically necessary.
Translation Between Representations
One representation can be transformed into another.
Text becomes bytes.
Coordinates become maps.
Equations become graphs.
This requires rules connecting representational systems.
Encoding is one such transformation.
Representation and Compression
A good representation often compresses.
Instead of storing every pixel of a circle, store:
- center,
- radius.
A short representation captures a large structure.
This links representation to information and computation.
Representation and Interpretation
A representation is not complete without a system that uses the relation.
The same marks can mean different things in different contexts.
Meaning often lies in:
representation + interpreter + convention + context.
The Next Question
If representations can be transformed from one form into another, we need rules for those transformations.
Text becomes binary.
Sound becomes digital samples.
DNA becomes protein sequence through cellular machinery.
This process is:
encoding and decoding.