What a 'model' actually is
Think of a model train set. It isn't a real train — it has no roaring engine, no hundred tonnes of steel, no passengers. And yet, by keeping only what matters (wheels on rails, a loop of track, a throttle you can turn), it lets a child *understand* how trains move without ever building a railway. A scientific model is exactly this: a deliberately simplified stand-in for something too big, too fast, or too hidden to grasp all at once. You strip away everything inessential, keep the bones, and end up with a thing you can hold in your hand — or in this case, run on a computer — and poke at safely.
The whole field of [[computational-neuroscience|computational neuroscience]] is the art of building such stand-ins for the brain. Not to replace the real thing, but to make it *thinkable*. A brain has tens of billions of cells firing many times a second; you cannot hold all of that in your head, and you cannot easily run experiments on a living one. A model is a small, obedient version you *can* run — speed it up, slow it down, switch a part off, ask "what if?" a thousand times before breakfast. The map is never the territory, but a good map is the only way to plan a journey across a territory too vast to see all at once.
Theory and experiment: two hands, one job
It's tempting to think experiments are the 'real' science and theory is just decoration on top. But picture trying to clap with one hand. Experiments *gather* the facts — they record a cell crackling, they watch a region light up — yet a pile of facts on its own doesn't explain anything, the way a heap of bricks isn't yet a house. Theory is the blueprint that says how the bricks fit, *and* which brick to fetch next. The two are a single rhythm: measure, explain, predict, measure again.
Here is the most famous handshake in all of neuroscience. In the 1950s, two scientists measured electrical currents flowing across the membrane of a giant squid nerve, then wrote down equations that *reproduced* those currents and, crucially, predicted the shape of a nerve's spike before anyone could see it directly. That work — the [[hodgkin-huxley-model|Hodgkin–Huxley model]] — is the moment theory and experiment fused into one. The math didn't replace the squid; it explained the squid, and then forecast what fresh squids would do. That is what a theory buys you that a notebook of measurements cannot: a *prediction* you can go and test.
Three questions to ask about any brain part
When something in the brain baffles you, it helps to know *what kind* of question you're even asking. A classic lens — usually called the three [[levels-of-analysis|levels of analysis]] — splits every puzzle into three clean layers. Take a familiar machine first, a hand calculator, to feel how they differ. You can ask *what* it does (turns two numbers into their sum), *how* it does it (electricity flowing through tiny switches in a particular pattern), and *why* it's built that way (it's faster and surer than counting on fingers). Same device, three completely different answers — and you need all three to truly understand it.
THE THREE QUESTIONS (the same part, three layers)
WHAT ──► What problem is this part solving?
(e.g. "recognise a face")
HOW ──► What recipe / wiring carries it out?
(e.g. "these cells, connected like so")
WHY ──► Why is the brain built this particular way?
(e.g. "this is fast, cheap, and robust")Now point that lens at vision. *What* is the eye-and-brain system doing? Turning a flat smear of light into a guess about the solid world out there. *How* does it do it? Through layers of cells passing and reshaping signals, each one nudging the next. *Why* is it arranged this way rather than another? Because this arrangement is fast, sips little energy, and keeps working even when cells die. The beauty of these three questions is that they keep you from mixing answers up — and they hand you a shelf for every later topic on this rung, so each new idea has an obvious place to sit.
A brain is a thing that changes over time
Here is the single most freeing idea in this whole rung. A brain is never frozen — every cell's voltage is rising and falling, signals are racing along, the whole thing is in restless motion, *always*. Anything that keeps changing by its own rules, moment to moment, is what scientists call a dynamical system. A swinging pendulum is one. So is the weather, a sloshing cup of coffee, a population of rabbits and foxes rising and falling. The branch of brain science that treats the brain this way — as a pattern that *evolves over time* — is [[dynamical-systems-neuroscience|dynamical systems neuroscience]], and it gives us a way to follow the dance instead of just freezing a single snapshot.
Why does this matter so much? Because it tells you what the equations are *for*. They aren't there to make the brain look fancy; they are simply the most honest way to write down a rule of change — "given where things are right now, here is where they'll be a heartbeat later." That single sentence is the seed of everything ahead: a cell's voltage chasing its own spike, a network settling into a memory, a thought taking shape. Once you see the brain as something that flows according to rules, equations stop being a wall and start being the natural language for describing the flow.
The road ahead, and why it's worth walking
Everything on this rung grows from the four ideas you now hold. Some guides zoom all the way in to a single cell and ask how its [[action-potential|action potential]] can be captured in math, from the detailed Hodgkin–Huxley recipe down to a stripped-bare 'fill the bucket until it tips' sketch. Others zoom out to whole networks that learn, the ancestors of today's artificial neural nets. Others ask whether the brain is secretly a prediction machine — a guesser that constantly bets on what's coming next, an idea known as the [[bayesian-brain|Bayesian brain]]. And some look outward, to maps of every connection and to chips wired to mimic neurons.
And the payoff reaches far past the lab. The same lessons we draw from modelling brains flow straight into [[brain-inspired-ai|brain-inspired AI]] — the machines that recognise faces and finish your sentences borrow their core trick, layers of simple units learning together, from neuroscience. The traffic even runs both ways: insights about thinking machines feed back into how we read living brains, including the devices that let a paralysed person move a cursor with thought alone. Wherever you're headed — medicine, AI, or pure curiosity — the habit of turning a brain into a model you can run is one of the most powerful moves you'll ever learn. That's why it's worth the climb.
- A model is a deliberately simplified copy of the brain you can run, poke, and ask 'what if?' — true to one thing even while wrong about many.
- Theory and experiment are two hands of one job: experiments gather facts, theory explains them and predicts the next fact to test.
- Sort every puzzle into what (the problem), how (the wiring or recipe), and why (the reason it's built that way) — three layers, never mix them up.
- A brain is a dynamical system — a pattern that evolves by its own rules over time — so equations are simply the honest way to write down 'how it changes next'.