Change one thing, measure one thing, and keep everything else exactly the same. Learn those three roles and you can run a fair test of almost anything — a cookie, a paper plane, a plant, even a new medicine.
Start bakingYou change one thing, you measure one thing, and you lock everything else — so when the answer moves, you know exactly what moved it.
The thing you change on purpose is the independent variable. The thing you measure to see what happened is the dependent variable (it depends on your change). Everything you deliberately keep the same is a controlled variable. A "variable" just means anything in your experiment that could vary — anything that could be different from one go to the next. Your whole job is to decide which one variable gets to move, which one you'll watch, and which ones you'll pin down. Get those mixed up, and no matter how careful you are with everything else, your experiment can't actually answer your question.
That sounds strict, and it is — but it's also a kind of superpower. Once you can spot the three roles, a messy real-world question turns into something clean and testable. Let's meet the three roles, then go bake a cookie and feel them for yourself.
Every experiment is a tiny story with the same three characters. Once you've been introduced, you'll recognise them everywhere — in your science class, in a cooking video, in a sports try-out. Here they are.
The independent variable is the single thing you change on purpose. It's called "independent" because it doesn't depend on anything else in the experiment — you decide it, before you even start. You usually pick a few different settings to try (scientists call these the "levels"). In the cookie test coming up, the independent variable is the oven temperature, and the levels are the temperatures you choose to bake at.
Quick memory trick: Independent — I change it.
The dependent variable is the thing you measure to find out what happened. It's called "dependent" because its value depends on what you did to the independent variable. You never set this number by hand — you read it off something: a ruler, a stopwatch, a scale, a thermometer, or just your own eyes. In the cookie test, the dependent variable is how brown the cookie gets.
Quick memory trick: it depends on your change, so you have to wait and see.
The controlled variables are everything else you deliberately keep identical from one go to the next. They're sometimes called "constants." There can be a whole dozen of them — same flour, same sugar, same tray, same shelf, same baking time — and they don't get any glory, but they're the reason your one comparison actually means something.
Quick memory trick: controlled = you keep them under control, frozen in place.
A controlled variable (a single thing you keep the same) is not the same as a control group. A control group is a whole extra set-up that gets the "normal" or "no change" treatment, so you have something fair to compare against — like the plant you give the usual amount of water, or the patients who get a fake pill instead of the real medicine. Both ideas help you stay fair, but one is a thing you hold steady and the other is a comparison you set up. We'll meet the control group again near the end.
Drag the orange temperature dial and watch the cookie brown. The teal browning readout is what you measure — it moves on its own, you never type it in. The green items are locked: tap one and see why it won't budge.
You don't set this number — you read it off the result.
Tap a locked item to see why it stays the same.
At 175°C the cookie reads 46% browned. Only the temperature is free to move — the recipe is locked, so the colour change can only be the oven's doing.
Feel what's happening as you drag: you are the only thing setting the temperature, but you have no direct control over the browning. You set the cause; the experiment hands you back the effect. That gap between "the knob I turn" and "the number I read" is the whole heart of an experiment.
Three parts of this cookie experiment — give each one the right role.
Nice — all three sorted. That's a fair test. 🎉
A single batch on its own doesn't tell you much. The magic happens when you compare two settings of the independent variable, with everything else identical. So let's do exactly that with the oven. Bake one batch at a cool 160°C and one at a hot 210°C — same dough, same time, same tray — and read the browning off each.
(Don't take my word for it — drag the dial back up to the oven and check those numbers yourself.) Now read the result out loud, the scientist's way: "When I raised the temperature by 50 degrees and kept everything else the same, the browning jumped from about 18% to about 82%." Notice the shape of that sentence. It names the change you made, it names the measurement you got, and it says "everything else the same." That little phrase is what lets you finish with a conclusion: a hotter oven makes a browner cookie.
Here's the part that's easy to miss. That conclusion is only trustworthy because the recipe was locked. If Batch B had also used more sugar, or a darker tray, or two extra minutes in the oven, then the deep-brown colour might have come from any of those — and your tidy sentence would fall apart. The comparison is only clean because exactly one thing was allowed to differ.
Picture your question travelling through the experiment. You turn the change knob, the experiment runs, and out comes a measurement. If only one thing was free to change, the result has only one possible explanation — your change caused it. But if two things changed at once, you're stuck: you can't tell which one mattered. The controlled variables are the locks that keep that path clean and single-file.
Watch the signal travel: you change the temperature, the experiment runs with everything else locked, and you read the measurement. One open path, one clear answer.
This is why scientists change only one thing at a time. It's slower, and honestly it can feel like a lot of fuss — but it's the only way to be sure your one change is really the cause, and not some hidden difference you forgot about. A locked variable is a suspect you've ruled out before the investigation even begins.
It really feels like it would. If you're testing whether a new paper-plane design flies farther, why not also grab nicer paper and throw it a bit harder while you're at it? More changes, surely a bigger, quicker answer — right?
"Changing several things at once is faster." It looks efficient, but it quietly destroys your experiment. If you change three things and the plane flies farther, you've learned that something helped — but you have no idea which one, or whether two of them helped and one actually hurt. You'd have to start over to untangle it. "Fast" turns into "useless," and useless is the slowest result of all.
Let's make it real. You and a friend want to know which paper plane flies farther — the pointy Dart or the wide Glider. The design is the one thing you want to test, so that's your independent variable — it's allowed to differ. Everything else needs to be the same, or you won't know what caused the winner to win. Right now the test is a mess. Tap each item to lock it to Same until the test becomes fair.
See what happened? The fair test isn't the one with the most changes — it's the one with the fewest. When only the design differs, a win for the Dart can only mean one thing. That's the whole trade: change one variable at a time, and every result you get is one you can actually trust.
Even with everything locked, the real world is a little wobbly. Bake the exact same cookie three times and you still won't get three identical numbers — your oven has hot spots, the dough blob is never perfectly even, the tray sits a centimetre to the left. So a single measurement can fool you. The fix is simple and powerful: repeat each setting a few times, then take the average — the typical, middle value.
Three bakes at the same temperature gave 80%, 84% and 79%. None is "the answer" on its own — but their average, about 81%, is something you can trust. Repeats do two jobs at once: they pull you toward the true value, and they show you how spread out your results are. If your three numbers were 80, 81 and 79, your test is steady. If they were 30, 80 and 95, something else is wandering around loose — a clue that one of your "controlled" variables isn't really under control yet.
Two quick habits that keep a test fair, on top of repeating it. First, spread your independent variable across a sensible range — test five different temperatures, not two, so you can see the whole pattern instead of guessing between two dots. Second, measure the dependent variable the same way every time: same ruler held the same way, same "browning" judged against the same chart. A wobbly measurer is just another variable you forgot to lock.
Once you've got results, you draw them — and a graph has its own version of the three roles. The independent variable (the thing you changed) goes along the bottom, on the x-axis. The dependent variable (the thing you measured) goes up the side, on the y-axis. There's a little rhyme for it: the cause goes across the bottom, the effect climbs up the side.
Below is your cookie experiment as a graph. Pick a temperature, then hit Bake this batch to drop a dot: its left-right spot is the temperature you chose, and its height is the browning you got. Bake a handful across the range and watch the dots draw the story.
No dots yet. Choose a temperature and bake a batch — each batch becomes one point on the graph.
As the dots line up, the shape they make is your finding. Here it climbs: low temperatures sit near the bottom, high ones near the top, and the line rises from left to right. That rising line is a sentence in picture form — "as temperature goes up, browning goes up." A graph lets anyone see in two seconds what would take a paragraph to explain, and it's only possible because you were disciplined about which variable went where.
Cookies were just practice. The three roles don't care what you're testing — swap the topic and the cast stays exactly the same. Read each of these and notice how the Independent, Dependent and Controlled always slot into place.
Spot the pattern? The independent is always the thing you'd set before you start. The dependent is always the thing you'd wait and measure at the end. And the controlled list is always the longest — it's everything else.
A friend wants to know whether plants grow taller with more water. Give each part its role.
You've got it — same trio, totally different experiment. 🌱
This isn't just a classroom exercise. The exact same three roles are how grown-up scientists, doctors and engineers decide what's actually true — and they spend enormous effort locking variables, because real answers depend on it.
Change: who gets the real pill versus a fake one (the control group). Measure: who actually gets better. Controlled: similar patients, same dose schedule, same check-ups — so the cure can be credited to the medicine, not to luck.
Companies show one group of users a blue button and another group a green one, keeping the rest of the screen identical, then measure which gets more taps. It even has a name — an "A/B test" — and it's just one independent variable with everything else locked.
Try one new warm-up routine (the change), time your sprint (the measure), and keep the track, the shoes and the time of day the same (controlled). Now you actually know if the warm-up helped.
Scientists can't put the whole Earth in a lab, so they work hard to isolate one factor at a time — changing one input in a careful model or sample while holding the others steady — to tell apart what's really driving a change.
Can I have more than one independent variable?
When you're learning, keep it to one at a time — it keeps every result clean. Advanced scientists sometimes change several on purpose, but they need clever maths to untangle which change did what. Start with one; you can level up later.
What if I just can't keep something the same?
Sometimes you can't — maybe you can't control the weather outside. When that happens, you do your best to match things up, and you write it down as a limitation. Honest scientists always tell you what they couldn't control, so others can judge the result fairly.
Does the dependent variable have to be a number?
Usually it's something you can count or measure, because numbers are easy to graph. But it can also be a category — yes or no, which colour, alive or wilted. Numbers just make the pattern easier to picture on an x-and-y graph.
How is "controlled variable" different from "control group" again?
A controlled variable is a single thing you keep the same across every trial. A control group is a whole extra set-up that gets no change, on purpose, so you've got a fair baseline to compare your changed group against. You often use both in the same experiment.
The independent variable — the single thing you turn on purpose.
The dependent variable — the result you read off to see what happened.
The controlled variables — everything else, kept identical so the test stays fair.