On this page
Your school asks for a science project, or you want to enter a fair. The hardest part is often not the experiment but the question. "Something about plants" or "solar energy" is a topic, not a question you can test. This post shows how to turn an idea into a testable question, design a fair experiment, and check your results, using a worked pendulum experiment with real numbers.
What makes a good project question?
A testable question names one thing you will change and one thing you will measure. Compare:
Topic: "Pendulums."
Vague question: "What affects a pendulum?"
Testable question: "How does the length of a pendulum affect the time it takes for one swing?"
The thing you change is the independent variable (length). The thing you measure is the dependent variable (time for one swing). Everything else, such as the mass of the bob and the starting angle, becomes a controlled variable you keep the same.
A good question is also one you can answer with what you have: a ruler, a stopwatch, string and a few weights.
What is a fair test?
A fair test changes only one variable at a time. If you change both the length and the mass, and the swing time changes, you cannot tell which change caused it. Researchers call this the control of variables strategy.
It does not come naturally. Chen and Klahr (1999) studied 87 children aged seven to ten as they designed experiments. Children who got explicit instruction in the strategy, together with probe questions, learned to design unconfounded experiments and transferred the skill to new situations. Probe questions alone did not help. In a later study, Klahr and Nigam (2004) found that many more children learned the strategy from direct instruction than by discovering it on their own, and those who learned it by instruction were as good at judging other students' science fair posters as the few who discovered it.
So it is worth learning the rule directly: change one thing, keep the rest the same, and say so in your report.
A worked example: pendulum length and period
Question: How does the length of a simple pendulum affect its period, the time for one full swing?
Prediction from theory: For small swings, T = 2π√(L/g), where g ≈ 9.8 m/s². So the period should depend on length but not on the mass of the bob.
L = 0.25 m: T = 2π√(0.25/9.8) = 1.00 s
L = 0.50 m: T = 2π√(0.50/9.8) = 1.42 s
L = 1.00 m: T = 2π√(1.00/9.8) = 2.01 s
Method.
Tie a bob to a string and measure L from the pivot to the centre of the bob.
Keep the bob mass and a small starting angle (about 10°) the same for every trial.
Time 10 full swings, not one. If your reaction time adds an error of about 0.2 s to each timing, spreading it over 10 swings cuts the error per swing to about 0.02 s.
Repeat each length three times.
Sample data for L = 0.50 m: 14.3 s, 14.1 s and 14.2 s for 10 swings. Mean = (14.3 + 14.1 + 14.2) ÷ 3 = 14.2 s. Period T = 14.2 ÷ 10 = 1.42 s, matching the prediction.
Checking the pattern. Doubling the length from 0.50 m to 1.00 m does not double the period. It multiplies it by √2 ≈ 1.41, since 2.01 ÷ 1.42 ≈ 1.42. A graph of T against L curves, but a graph of T² against L is a straight line, because T² = (4π²/g) × L. With T = 2.01 s at L = 1.00 m, T² = 4.04 s², so g = 4π² ÷ 4.04 ≈ 39.48 ÷ 4.04 ≈ 9.77 m/s², close to 9.8 m/s² (the small gap comes from rounding T). Getting g back from your own data is a strong check that the experiment worked.
A second question for free. Keep L fixed and change only the mass. The theory predicts no change in period, and testing that prediction is a good fair-test extension.
What are the common mistakes?
Changing two things at once. Changing the string and the bob between trials.
Measuring once. A single timing hides random error. Repeat and take the mean.
Large starting angles. The formula above is for small swings. At large angles the period gets longer.
Measuring length to the wrong point. Measure to the centre of the bob, not its top.
Hiding odd results. If one trial is far off, report it and suggest why, rather than deleting it.
Claiming too much. Three lengths show a pattern; they do not prove a law. Say what your data supports.
How do you write it up?
Use the same order a scientist would: question, prediction, method (with controlled variables listed), results table and graph, analysis, and limitations. A short paragraph on what you would do with more time shows you understand the weak points.
If you want to explore a question before building anything, a simulation lets you try many values quickly. The Fermi Simulation Lab has interactive simulations for ideas that are easier to see than to read, which can help you choose which variable is worth testing with real equipment. You can also write your own: our post on building a physics simulation in Python shows how.
Where can school students take a project further?
The Department of Science and Technology's INSPIRE-MANAK scheme aims to motivate students in classes 6 to 12 to submit original science ideas. Schools nominate up to five ideas through the official portal, with at most two from Class 11 or 12 science students. Check the portal for the current year's nomination dates.
This week, write one testable question for your project and list its independent, dependent and controlled variables.
