Take out a map of Singapore.
Now place it on the table.
Where is the rain?
Where is the heat rising from the pavement at two in the afternoon? Where are the smells from the hawker centre, the arguments inside one apartment, the strain in a bridge joint, the price of a bowl of noodles, the memory attached to a school gate, the traffic noise under an expressway, the direction of a morning commute, the fear somebody felt in a hospital room twenty years ago?
They are not there.
Yet the map can still be excellent.
This is the strange power of representation. A useful representation succeeds partly because it does not contain everything.
It preserves what matters for a purpose and leaves enormous amounts of reality outside the frame.
That is not a flaw to be eliminated.
It is the beginning of workable thought.
Quick Read
A representation is something that stands for, carries information about, depicts, models or otherwise makes some other thing cognitively available.
A word can represent a concept or referent. A map represents spatial relations. A graph represents selected quantities and their relationships. An equation can represent a physical pattern. A photograph represents a visual scene from a viewpoint. A memory represents aspects of a past experience. A diagram can represent a mechanism. A story can represent events, motives and possibilities. A scientific model represents selected structure in a system.
But the representation is not the represented thing.
One-sentence answer: A representation is a selective form in which information about something becomes available for thought, communication, memory, comparison or action.
The most important word in that sentence is selective.
The World Is Larger Than Any Representation of It
Imagine trying to build a representation of a tree that contains absolutely everything about the tree.
Every leaf.
Every cell.
Every molecule.
Every insect currently touching it.
Its exact geometry at every moment.
The history of every branch.
The genetic variation inside its tissues.
The water moving through it.
The sunlight striking each surface.
Its chemical exchanges with soil, fungi and air.
The representation would become as complicated as the tree—and perhaps more complicated, because now we would also need a way to store and interpret the representation.
Representation is useful because it reduces.
A botanical illustration may preserve form. A diagram of xylem and phloem preserves transport relationships. A map records location. A DNA sequence records selected molecular information. A poem may preserve what the tree felt like to a human observer. A timber inventory may preserve height, diameter and commercial category.
Same tree.
Different representations.
Different purposes.
Representation Is Not Copying
One of the easiest misconceptions is to think a good representation is simply a faithful miniature copy.
Many powerful representations do not resemble what they represent.
The numeral 7 does not look like seven objects.
The word rain is not wet.
A subway map may distort geographic distance to make routes legible.
A chemical formula does not resemble the substance it describes.
A line graph showing temperature over time does not look like temperature.
An algebraic function can capture a relationship without visually resembling the physical system.
Representation therefore depends on conventions, mappings, structural correspondences, learned meanings and interpretive practices—not only resemblance.
The Map Is Useful Because It Is Not the Territory
A literal one-to-one map of Singapore at full scale would be useless for navigation because it would be Singapore.
A map becomes useful by shrinking, selecting, symbolising and organising.
Roads may be widened visually. Buildings may disappear. Elevation may be omitted. Administrative borders may be emphasised. Names may be placed where they are readable rather than physically located. A transit map may preserve sequence and connectivity while sacrificing geographic shape.
Every map therefore answers an implicit question:
What does the user need to do?
Drive?
Walk?
Find a train connection?
Understand elevation?
Study rainfall?
Plan land use?
Analyse historical change?
A representation is good relative to a job.
A Perfect Representation for One Job Can Be Terrible for Another
This gives us one of the most important ideas in Cognitive Art: representational fitness.
Suppose you want to travel from Punggol to Orchard Road.
A geological map of Singapore may be highly accurate and scientifically valuable. It is still the wrong representation for your immediate task.
Suppose you are diagnosing why a bridge is deteriorating.
A tourist photograph might be beautiful and correctly depict the bridge. It may not reveal stress, material fatigue, loads, drainage or internal defects.
Suppose a student is learning fractions.
The symbolic representation 3/4 may be compact, but a bar model may reveal the part–whole structure more clearly. Later, the symbolic form becomes more efficient.
The same learner may need different representations at different stages.
Good teaching is therefore partly the art of choosing the representation that makes the required relationship visible.
Internal and External Representations
We should distinguish two broad families.
Internal representations
Cognitive science often uses the language of mental representation to explain information-bearing states and structures involved in perception, memory, reasoning and action. The exact nature and explanatory role of mental representations remain philosophically and scientifically debated. There is no need to pretend otherwise.
For everyday purposes, it is enough to notice that a person can think about something that is not currently present. You can remember your classroom while sitting at home, imagine tomorrow’s appointment, rotate a shape mentally, rehearse a sentence before speaking or compare two routes without physically travelling both.
External representations
Humans also move cognition into the environment.
We draw.
Write.
Count.
Make tables.
Build maps.
Record measurements.
Construct diagrams.
Invent notation.
External representations let information persist outside biological memory. They can be shared, revised, compared and inspected by multiple people. Civilisation becomes vastly more capable once representations can travel independently of the individual mind that first produced them.
Writing Is a Representation Technology
Think about the sentence you are reading now.
The marks on the screen are not the thought itself.
They are an external symbolic representation designed to trigger reconstruction in another mind.
The writer selects words.
The words carry conventional meanings and relationships.
The reader brings vocabulary, context, prior knowledge and inference.
Understanding emerges from the interaction.
This explains why communication can fail even when every word is spelled correctly.
The representation may be grammatically valid yet badly matched to the reader.
Too much assumed knowledge.
Wrong level of detail.
Ambiguous reference.
Unclear structure.
Missing evidence.
Representation quality is partly receiver-dependent.
The Same Thought Can Be Represented in Different Forms
Suppose a car travels at constant speed.
We might represent that situation as:
- a sentence,
- a table of time and distance,
- a distance–time graph,
- an algebraic equation,
- an animation,
- a physical demonstration,
- a verbal analogy.
Each form makes some relationships easier to see and others harder.
The equation is compact.
The graph makes rate visible geometrically.
The table makes individual values explicit.
The animation supports intuitive temporal understanding.
Expertise often includes the ability to translate between representations while preserving the underlying relationship.
Translation Is a Test of Understanding
A student who memorises an equation may appear knowledgeable.
Ask the student to explain the equation in words.
Then draw the relationship.
Then identify it in a real situation.
Then predict what changes when one variable doubles.
If understanding survives the translation, the knowledge is probably more robust.
This works in English too.
Can a student turn a paragraph into a one-sentence summary without destroying the central idea?
Can a narrative be converted into a timeline?
Can an argument be converted into claim → evidence → reasoning?
Can a diagram be explained verbally?
Translation exposes whether the learner owns the relationship or merely recognises the surface form.
Representation Has Resolution
A representation can be coarse or fine.
“The economy grew” is low resolution.
A detailed statistical account distinguishing real output, sectors, population, prices, productivity, income distribution and time period is higher resolution.
“The student is weak” is low resolution.
“The student reads fluently but loses causal relationships across long expository passages and gives answers unsupported by textual evidence” is higher resolution.
Higher resolution is not automatically better.
If somebody asks whether to bring an umbrella, a meteorological model containing every atmospheric variable is unnecessary. A well-calibrated local rain forecast may be the better representation for action.
Resolution must match purpose.
Every Representation Has a Viewpoint
A photograph seems objective because a camera records light mechanically.
But somebody selected the camera position.
The lens.
The frame.
The moment.
The exposure.
What sits outside the photograph is invisible.
The same applies to data.
A chart selects variables, dates, categories, scales and baselines. The numbers may be correct and the representation can still foreground one story while hiding another.
Viewpoint does not mean falsehood.
It means selection.
Responsible interpretation asks:
- Who made this representation?
- For what purpose?
- From what position?
- Which variables were included?
- Which were excluded?
- What scale was chosen?
- What alternatives would make a different structure visible?
Representation Is Compression With Consequences
Most useful representations compress reality.
A weather icon compresses atmospheric complexity into a symbol. A school grade compresses many performances into a category or number. A medical chart compresses a person’s health history into recorded observations. A national statistic compresses millions of lives into an aggregate. A résumé compresses years of work into a few pages.
Compression is necessary.
But what gets discarded matters.
A grade can hide improvement. An average can hide inequality. A diagnosis can hide individual variation. A headline can hide uncertainty. A map can hide terrain. A category can hide borderline cases.
Whenever a representation becomes powerful in a system, ask what information the compression removed.
Memory Is Not a Video Archive
People often speak as though a memory were a stored recording that can simply be replayed.
Human memory is more complicated.
Contemporary research treats memory representations as structured, variable in precision and influenced by prior knowledge and later processing. Visual memory, for example, is increasingly described using models involving noise, hierarchy and priors rather than a simple set of perfectly preserved internal pictures.
This matters because remembering feels like returning to the event.
But what returns is a representation of the event.
That representation may preserve the gist and lose detail. It may combine information. It may be influenced by later knowledge. Confidence can exceed accuracy.
The lesson is not “memory is useless.”
The lesson is to stop treating remembered experience as an untouched original.
The Brain Does Not Need a Tiny Cinema
When people hear “mental representation,” they may imagine a little internal screen on which pictures appear.
That metaphor is attractive and incomplete.
Cognitive science has proposed many forms of representation: symbolic, imagistic, distributed, model-like, sensorimotor and others. Philosophers continue to debate what mental representation is, what content consists of and how representation contributes to explanation.
A world-class reader should know that the field is not settled into one cartoon.
We can still use the higher-level insight safely:
The mind can preserve and transform information about things that are not identical to the immediate sensory world.
Representation Makes Absence Thinkable
This is one of humanity’s astonishing capabilities.
You can discuss a person who is not in the room.
A city you have never visited.
A species that is extinct.
A civilisation that no longer exists.
A galaxy you cannot see unaided.
A future building.
A fictional universe.
A number larger than anything you could physically count one-by-one.
Representation releases cognition from the immediate here and now.
Numbers Are Extraordinary Representations
Consider the numeral:
1,000,000
Seven symbols allow the reader to engage with a quantity that would be exhausting to represent using one million drawn dots.
Place-value notation is a cognitive technology.
It compresses quantity into a form that supports calculation.
Algebra goes further. A letter can represent an unknown or varying quantity. Equations can preserve relationships across infinitely many possible cases.
Mathematics becomes powerful partly because its representations are deliberately constructed to make structure manipulable.
A Diagram Can Think With You
Draw four boxes connected by arrows.
Something changes.
Relationships that had to be maintained verbally now become spatially visible. You can point. Compare. Reorder. Notice a missing link. See a loop.
The external representation is doing cognitive work.
This is why diagrams are not decorative additions to “real thinking.” They can alter the problem-solving process itself.
A good diagram reduces working-memory demands by stabilising relationships in the environment.
That is why drawing can rescue a student who is stuck.
The student has not suddenly become more intelligent.
The intelligence has acquired a better surface to work on.
Graphs Reveal Structure and Manufacture Illusions
Graphs are among civilisation’s most powerful representations.
They turn quantities into spatial relationships.
A trend that is difficult to see in a table may become immediately visible in a line. Distribution becomes visible in a histogram. Association becomes visible in a scatter plot.
But a graph can also mislead.
Change the axis.
Choose a different starting date.
Aggregate categories.
Use totals rather than rates.
Plot nominal values instead of inflation-adjusted ones.
The data may remain technically correct while the visible story changes dramatically.
Representation literacy therefore belongs inside data literacy.
Models Are Representations With Explanatory Ambition
A model usually does more than display.
It tries to capture structure sufficiently well that we can explain, calculate, predict, simulate or intervene.
A globe models Earth.
A circuit diagram models electrical relationships.
A food web models ecological interactions.
An economic model isolates selected relationships.
A climate model represents physical processes mathematically and computationally.
A model should not be criticised merely for leaving things out.
All workable models leave things out.
The serious questions are:
- Did it leave out something that matters for this use?
- Does it preserve the relevant relationships?
- Where does it stop working?
- How large are its errors?
- What observations would force revision?
A Representation Has an Operating Range
A school map may work on campus and nowhere else.
A simple physical model may work at ordinary speeds but fail near the speed of light.
A rule of thumb may work for common cases but fail at the extremes.
A child’s concept of “bird” may work until the first penguin appears.
Representations become dangerous when their operating range is forgotten.
The representation continues to look familiar, so the user assumes it remains valid.
This is a common pattern in expertise failure: a successful model from one context is transferred into another without checking whether the relevant conditions survived.
Categories Are Representations That Reshape Attention
Learning a category does not merely attach a label to an unchanged perception.
Research on category learning suggests that experience and selective attention can change the psychological and neural representation of objects, enhancing dimensions that matter for learned distinctions and reducing the influence of irrelevant dimensions.
This is profound.
Learning can change what becomes easy to see.
A birdwatcher sees species distinctions where a novice sees “small brown bird.” A radiologist sees patterns in an image that a layperson cannot use. A language teacher hears grammatical structure where a beginner hears a vague sentence. A mechanic hears engine behaviour inside what another person calls “a strange noise.”
Expertise is partly representational reorganisation.
Representation Can Create Blindness
The same learning that reveals structure can hide alternatives.
If a doctor sees every complaint through one familiar diagnosis, representation has become tunnel vision. If an economist sees every human decision only as price response, important motives disappear. If a teacher sees every mistake as carelessness, different learning failures collapse into one explanation.
The better the representation once worked, the harder it can be to notice that it is now wrong.
This is why mature expertise needs representational humility.
Not “I have no model.”
But:
I have a model, therefore I know what I might be missing.
Language Represents and Also Frames
Words do not merely point neutrally.
Choice of vocabulary can foreground different aspects of the same event.
“Crowd,” “audience,” “mob,” “protesters,” “supporters” and “residents” can refer to overlapping groups while inviting different interpretations.
“Investment,” “cost,” “subsidy,” “support,” “bailout” and “spending” similarly carry different frames.
This does not mean language makes objective reality impossible.
It means careful readers separate the event from the vocabulary used to represent it.
Ask what happened before accepting what it has been called.
Representation and Power
Representations do not merely describe social systems.
They can influence them.
A credit score represents aspects of financial history and may affect access to loans. An examination score represents performance under specified conditions and influences educational pathways. A risk category can influence inspection, insurance or treatment. A map can define administrative responsibility. A legal classification can change rights and obligations.
Once a representation becomes operational, people begin living with its consequences.
That raises a moral question:
How much of a human being or real situation has the representation failed to carry?
A score can be useful without becoming the person.
A category can guide attention without exhausting identity.
A dataset can inform policy without containing every lived reality.
The Dashboard Problem
Modern institutions love dashboards.
They are excellent cognitive tools.
Complex operations become visible as a handful of metrics. Managers can detect changes quickly. Schools can track attendance and results. Hospitals can monitor waiting times. Cities can track traffic or energy use.
But a dashboard creates a new temptation:
what is measured becomes what is real.
Anything not represented on the dashboard becomes cognitively quiet.
The dashboard is not lying.
It is omitting.
Leadership therefore requires looking both at the representation and occasionally outside it.
The Photograph Problem
A photograph has extraordinary persuasive power because it feels like direct access.
But even an unedited photograph is a slice.
One place.
One direction.
One moment.
One field of view.
Ten seconds earlier may tell another story.
Two metres to the left may tell another story.
The photograph can be genuine and still be insufficient.
This distinction matters in news, history, law and everyday conflict.
The Headline Problem
A headline represents an article in a few words.
That is extreme compression.
A good headline helps a reader decide whether the article is relevant. A bad headline substitutes for the article and leaves readers with a simplified or distorted model.
The mistake is not only writing a bad headline.
It is treating the compressed representation as though it carried the full evidence.
The Examination Problem
An examination is also a representation.
It samples performance under constrained conditions and uses that sample to infer something about knowledge or capability.
This can be enormously useful.
It is still a sample.
A mark does not contain every aspect of the learner. It may not show curiosity, persistence, oral reasoning, creativity, collaboration, prior opportunity, sleep or what the learner could do with more time.
None of this means examinations are meaningless.
It means we should understand what the representation was designed to represent and resist turning it into a total biography.
The Model–Reality Loop
A mature representation is not static.
It enters a loop:
World → observation → representation → interpretation → action → new observation → correction.
The loop matters because representations inevitably omit.
Reality gets a vote.
The bridge behaves differently than predicted.
The student makes an unexpected error.
The customer does not buy.
The experiment fails to replicate.
The weather changes faster than the model expected.
Now the representation must bend.
A representation that cannot be corrected by the world stops being a model and starts becoming a belief shield.
Representation Errors: Seven Ways the Map Fails
1. Omission Error
A variable that matters is absent.
2. Distortion Error
A feature is present but its magnitude or relationship is misrepresented.
3. Scale Error
The representation is built at the wrong spatial, temporal or organisational scale.
4. Staleness Error
The representation was once accurate but the world changed.
5. Reification Error
The representation is mistaken for the thing. The score becomes the student. The label becomes the person. The model becomes reality.
6. Receiver Error
The representation is good but the user lacks the conventions or background knowledge needed to interpret it.
7. Purpose Error
A representation designed for one job is used for another.
The Representation Audit
When something important is being represented, ask:
- What is the real-world target?
- What form is carrying information about it?
- Which features are preserved?
- Which features disappear?
- What is the intended use?
- Who is the intended receiver?
- What resolution is required?
- What viewpoint is embedded?
- What assumptions make the representation work?
- Where does it fail?
- How will reality correct it?
This audit works on maps, essays, diagrams, data dashboards, memories, school marks, forecasts, news reports, models and plans.
A Classroom Example: The Water Cycle
A textbook diagram of the water cycle may show ocean, evaporation, condensation, clouds, precipitation and runoff.
Excellent.
Now ask what the diagram omitted.
Groundwater complexity. Local weather. Different timescales. Soil properties. Human extraction. Vegetation. Atmospheric dynamics. Ice. Pollution. Geographic variation. Seasonal change.
Does that make the school diagram wrong?
No.
It makes it a teaching representation at a particular resolution.
The learner’s next developmental step is not to reject simplified diagrams.
It is to know they are simplified.
A Mathematics Example: The Number Line
The number line turns numerical order and distance into space.
That is extraordinary representational engineering.
Negative numbers become locations. Addition can be movement. Inequalities become regions. Absolute value becomes distance from zero.
The representation reveals relationships that are harder to hold in verbal form.
But the number line is not the numbers themselves.
It is a spatial model of selected numerical structure.
An English Example: Plot Versus Story
A student reads a short story and writes a five-line plot summary.
The summary can accurately represent the sequence of events and still miss almost everything that makes the story literature: tone, ambiguity, imagery, pacing, point of view, symbolism, emotional tension and language.
The summary is not bad.
It is narrow.
If the examination asks for plot, it may be sufficient.
If the question asks how the writer creates unease, the representation must change.
A History Example: The Timeline
A timeline represents temporal sequence beautifully.
It is weaker at representing simultaneous processes, geography, uncertainty, causation and lived experience.
A good historian therefore uses multiple representations:
- timeline for sequence,
- map for space,
- table for comparison,
- primary source for perspective,
- argument for causation,
- statistics for scale.
No single representation needs to do everything.
A Science Example: The Atom
Students encounter multiple models of atoms as their education progresses.
Early diagrams may show a central nucleus and electrons around it. Later learning complicates the picture dramatically.
The progression is educationally important because it teaches a deep scientific habit:
A model can be useful before it is complete.
Science does not wait for a total representation of reality before thinking.
It builds models, tests them, discovers limits and improves them.
The Developmental Progression: Learning to Distrust the Picture Without Losing It
Primary: learn that symbols stand for things
Children become fluent with letters, numerals, diagrams, maps, pictures and simple models. The educational goal is stable mapping between representation and meaning.
Lower secondary: compare representations
Students should increasingly translate between text, graph, diagram, equation, table and model. They learn that one representation reveals some structure while another reveals something else.
Upper secondary: question assumptions and operating range
The learner becomes responsible for model limits, measurement choices, textual perspective, uncertainty and inference.
Pre-university and adulthood: choose and build representations
The problem no longer arrives with the correct diagram attached. The learner must decide what should be represented, what may be safely omitted and which form makes the relevant structure visible.
The Best Representation Is Often a Set of Representations
Complex reality resists single views.
To understand a city, we may need geographic maps, transport networks, demographic data, historical records, economic flows, environmental measurements and human stories.
To understand a student, we may need work samples, examination results, classroom observations, conversation, error patterns, prior learning history and evidence of transfer.
To understand a disease, medicine may use symptoms, imaging, laboratory data, history, physical examination and population evidence.
Each representation acts like a window.
Wisdom is not choosing the one true window.
It is knowing what each window can and cannot show.
Representation and Reality Must Remain Connected
A representation can become so elegant that people fall in love with it.
The theory is beautiful.
The dashboard is clean.
The categories are tidy.
The organisational chart makes sense.
Then reality becomes inconvenient.
People do not behave according to the chart.
The data moves the wrong way.
The student fails the supposedly mastered transfer task.
The bridge cracks.
The forecast misses.
At that moment, world-class thinking protects the world from the representation, not the representation from the world.
A Practical Exercise: Represent One Thing Five Ways
Choose something familiar.
Your journey to school.
A Mathematics problem.
A family routine.
A news event.
Now represent it five ways:
- one sentence,
- a diagram,
- a table,
- a timeline or sequence,
- a set of numbers or categories.
Then ask what each form makes visible.
Ask what each form destroys.
You will begin to feel representation rather than merely define it.
From Representation to Extraction
Now the World of Cognitive Art has two movements:
World → Token → Representation.
We have selected a workable unit.
We have created or received a form that carries information about it.
The next question is:
What can we pull out that matters?
That is extraction.
And once repeated extractions reveal structure that survives across many cases, we approach abstraction.
Frequently Asked Questions
Is a representation always visual?
No. Representations can be linguistic, numerical, spatial, auditory, symbolic, diagrammatic, physical or distributed across several forms. A spoken description and a mathematical equation can both represent without being pictures.
Is a representation the same as reality?
No. A representation preserves selected information about a target. Confusing the representation with the represented thing is a major reasoning error because every workable representation omits, transforms or organises information.
Can an inaccurate representation still be useful?
Yes, if the inaccuracies are controlled and irrelevant to the purpose. A schematic map may distort distance while accurately preserving route connectivity. A simplified scientific model may omit mechanisms that are unnecessary at an introductory level. The key is knowing what has been simplified and when the simplification becomes unsafe.
Why do experts use multiple representations?
Because complex problems contain more structure than one form can display conveniently. Different representations expose different relations. Combining them reduces the risk that one viewpoint silently becomes the entire world.
Is memory a representation?
Memory is commonly discussed in representational terms, but it should not be imagined as a perfect recording. Contemporary research emphasises variable precision, reconstruction, hierarchy and the influence of prior knowledge.
Why do diagrams help learning?
They can externalise relationships that otherwise have to be maintained mentally. This can reduce cognitive load and make structure easier to inspect. A poor diagram can also mislead, so visual clarity is not proof of correctness.
What does representation have to do with writing?
Writing is an external representation technology. Writers turn observations, memories, claims and imagined situations into language that readers must reconstruct. Good writing therefore depends on choosing what to include, what to omit, how to organise it and what the reader already knows.
Can representations change perception?
Learning categories and task-relevant dimensions can change what distinctions become salient. Research on category learning shows that representations are not simply static containers; learning and attention can reshape them.
What is representational humility?
It is the habit of using a representation confidently enough to act while remembering that it is selective, conditional and correctable. It rejects both extremes: pretending models are reality and pretending models are useless because they are incomplete.
Research Notes and Further Reading
The scientific and philosophical literature on representation is large and contested. The Stanford Encyclopedia of Philosophy provides a useful overview in Mental Representation, including the diversity of representational proposals in cognitive science and ongoing debates about content and explanatory role.
For current work on memory representation, Brady, Robinson and Williams review noisy and hierarchical visual memory across timescales. For how category learning can reshape representational geometry, see Folstein, Gauthier and Palmeri on category learning and neural representations in visual cortex. A broader review of concept learning and interacting brain systems is available in Brain Mechanisms of Concept Learning.
These sources do not imply that one single theory of representation has won. That is precisely the point. The reader-facing framework here is a synthesis for learning: representations make selected structure usable, and their usefulness depends on purpose, mapping, receiver, resolution and correction by the world.
Final Thought: Intelligence Needs Something Smaller Than Reality
A human mind cannot contain the whole world at full resolution.
Neither can a book.
Nor a map.
Nor a theory.
Nor a photograph.
Nor a dataset.
Nor a school report.
So thought survives by making something smaller.
We draw the road.
Name the feeling.
Plot the data.
Write the equation.
Tell the story.
Remember the event.
Build the model.
Then we use that smaller thing to navigate something much larger.
The danger begins when we forget the reduction happened.
The opportunity begins when we remember.
Because if one representation hides the structure we need, we can build another.
Turn the sentence into a diagram.
Turn the table into a graph.
Turn the label into a set of observations.
Turn the memory into a question.
Turn the model back toward the world.
That freedom—to change how reality is represented without pretending reality itself has changed—is one of the great tools of thought.