Scientists don't have one model — they have a toolbox. The trick is knowing which tool the job wants. There are really four kinds, and a good model from any of them lets you do two things: explain what's going on, and predict what happens next.
The Four Kinds of Model
Physical models are things you can hold — a globe, a plastic skeleton, a model bridge, a DNA kit. Great when you need to see the shape and how parts fit together in three dimensions. Diagram models are labelled pictures — the particle model, a food web, a circuit diagram, the water cycle. Great when you need to show how parts connect or what flows where. Mathematical models are an equation or a graph — speed = distance ÷ time, a line on a graph, a formula for how a population grows. Great when the thing is really about numbers and how they change together. And computer models (simulations) are programs that run the rules over and over — weather forecasts, traffic flow, disease spread, climate. Great when there are too many moving parts for your head, so you let the machine grind through them.
The Method: Match the Model to the Purpose
When you have to choose or build a model, run the same three moves every time.
One — name the purpose. What exactly do you need the model for? To show a shape? To explain a flow? To predict a number? To run a messy "what if" with thousands of parts? The job decides everything else.
Two — pick the kind that fits the purpose. Shape you can turn over → physical. Connections and flows → diagram. Numbers that change together → mathematical. Too-many-parts "what ifs" → computer simulation. There's often more than one workable answer, but usually one is the obvious best fit.
Three — decide what to keep and what to drop. Now do the rung-1 move on purpose. Keep only what the job needs; drop the rest so the model stays simple enough to actually use. A globe for teaching the continents keeps their layout and drops the cities — adding every street would wreck it.
A Worked One, Slowly
Question: you want to predict how high a ball will be one second, two seconds, three seconds after you throw it straight up. What kind of model, and why?
Name the purpose: you want to predict a number — a height — at different times. That's numbers changing together, which points straight at a mathematical model: an equation or a graph of height against time. A physical model (a foam ball) shows you the throw but won't predict the height at 2.4 seconds; a diagram shows the arc but gives no numbers. Then keep and drop: the equation keeps the height, the time and gravity, and happily drops the colour of the ball, the wind and how sweaty your hands were. Name the purpose, match the kind, trim it down — that's the whole method, and it works for any model you'll ever be asked to build.