Designing the Mental Model

On architecture, internal representations, and the spaces we learn to carry in our heads

Architecture is largely a practice of making representations. Sketches, plans, sections, perspectives, building information models and digital twins all attempt, in different ways, to make a space intelligible before it has been fully experienced. Representation, in this sense, is not merely the language through which architecture is communicated. It is one of the ways architecture thinks.

But architects are not the only ones producing models.

The mind does something remarkably similar. We do not move through a building, a street or a city by retaining an exact record of its physical reality. Instead, we gradually construct an internal representation of it: incomplete, selective and often geometrically inaccurate, yet sufficiently coherent to let us find our way, anticipate what lies beyond what we can see, and understand how one place relates to another.

This creates an intriguing parallel between digital twins and mental ones.

Geographic information systems, building information models and digital twins translate physical environments into representations that can be examined, tested and acted upon. The mind, through experience, produces its own working model of space. Neither needs to preserve reality in its entirety. What matters is the construction of a model through which reality becomes usable.

A space does not enter the mind all at once

A mental model of a place accumulates.

On a first encounter, we may retain little more than an entrance, an exit, a primary route, a staircase or an unusually memorable point. The building initially exists as fragments: a corridor remembered here, a turn there, a glimpse of another floor whose relationship to the first is not yet entirely clear.

Movement begins to connect them.

With repetition, the building changes from something merely seen into something that can be anticipated. We begin to know where a corridor is likely to lead before reaching its end, which stair belongs to which part of the building, or what lies beyond a threshold that is no longer visible from where we stand.

Some buildings allow this internal model to form almost effortlessly. Others remain strangely foreign even after repeated encounters.

That difference is not accidental.

Clear circulation, distinguishable landmarks, legible transitions and spatial hierarchy give the mind relationships it can retain. Repetitive routes, ambiguous junctions and environments without meaningful distinctions demand something else: that the user repeatedly solve the building again.

Architecture, then, does more than organise physical space. It also establishes the conditions under which a representation of that space can be formed in someone else’s mind.

The useful inaccuracy of memory

Yet the building we carry in our heads is not the building itself.

Knowing a place does not mean retaining its correct scale, proportions or geometry. The mind appears to privilege usefulness over geometric fidelity. It may matter far more that a corridor connects to the stair than whether that corridor is twelve or twenty metres long; more that two rooms can be reached from one another than whether we remember the exact angle between them.

This becomes obvious when we try to externalise a place we know well.

We may navigate it without hesitation and still draw it badly. Distances stretch. Corners shift. Entire spaces disappear. A route that is perfectly clear while walking becomes strangely distorted once transferred to paper.

The model was never designed to be a survey.

It was designed to work.

The difficulty is not simply seeing the space. It is constructing a model of it that can reliably predict what comes next.

When the model meets the map

This gap between functional understanding and geometric accuracy helps explain why architectural representations remain useful even after a space has been physically experienced.

The mind fragments, simplifies and reorganises. It completes surfaces outside the field of view, imagines the continuation of a turn, and infers connections it has not yet directly seen. A plan or model does something fundamentally different: it offers scale, geometry and relationships simultaneously.

Looking at a plan of a building after walking through it can therefore produce a peculiar moment of recognition.

So that is where that corridor was.

A space that had existed as a sequence of experiences suddenly becomes visible as a whole.

Maps, models and digital twins do not simply provide additional information about an environment. They can confront the internal model with a geometrically controlled representation. Distances can be corrected, disconnected fragments joined, and relationships that were previously inferred can become explicit.

The process can also run in the opposite direction. A drawing may be geometrically precise while saying very little about how the building will actually be understood in use.

A digital twin may describe geometry, systems and measurable behaviour with extraordinary precision. Yet the model produced by a person moving through the same building is organised according to another criterion: what needs to be remembered in order to act.

Placing these representations beside one another opens up a useful design question. Not simply whether the building is correctly modelled, but whether its organisation can become intelligible to someone encountering it for the first time.

Designing for a model we cannot draw

This does not mean architects can — or should — design the exact image that will form in another person’s mind.

Spatial understanding varies with experience, attention, mobility, familiarity, purpose and countless other conditions. The internal representation will always belong to the person constructing it.

But architecture controls much of the material from which that representation must be built.

An entrance can announce itself or disappear into a façade. Two routes can establish a hierarchy or appear interchangeable. A stair can reveal the relationship between floors or conceal it. A threshold can register as a transition or pass almost unnoticed.

These decisions do not determine thought. They make certain spatial relationships easier — or harder — to learn.

And spatial understanding itself can be practised.

Walking through a building, reading its plan, imagining its volume, returning to the model and then comparing that representation with physical experience gradually trains the ability to move between what is seen, what is drawn and what is understood.

Three-dimensional thinking is therefore not simply the ability to operate modelling software. It is the ability to establish relationships between these different forms of spatial knowledge.

To read a plan and imagine its volume; to walk a route and later sketch it; to look at a façade and speculate about the organisation behind it; to read a section alongside a plan — each requires the mind to move between representation and experience.

Perhaps this is why architectural representation remains valuable even in a world increasingly capable of reproducing physical environments with extraordinary precision.

A perfect model of a building and a useful model of a building are not necessarily the same thing.

One describes the space.

The other allows us to know where we are within it.

References & Further Reading

Gibson, James J. The Ecological Approach to Visual Perception. Houghton Mifflin, 1979.

Lynch, Kevin. The Image of the City. MIT Press, 1960.

O’Keefe, John, and Lynn Nadel. The Hippocampus as a Cognitive Map. Oxford University Press, 1978.

Tolman, Edward C. “Cognitive Maps in Rats and Men.” Psychological Review 55, no. 4 (1948): 189–208.

Designing the Mental Model was originally published in Bootcamp on Medium, where people are continuing the conversation by highlighting and responding to this story.