Experiment 20: Design Your Own Experiment — Science Fair Capstone

Everything in this course converges here: the capstone asks you to be the scientist, end to end — a question of your own, stated as a testable hypothesis, a fair test designed around it, data collected in honest trials, and a conclusion that reports what the data actually said. The simulation is your planning studio: it contains an experiment-plan builder, a data table and a charting tool, and this lesson walks a complete example through all three.

The physics behind the practical

There is no new law in this lesson — the physics is the *method* every previous lesson used. A scientific question becomes testable when it is phrased as a hypothesis with a mechanism: "If [I change X], then [Y will change in this way], because [physical reason]." The "because" is what makes it physics instead of fortune-telling: it names the law (momentum, Hooke, v = fλ, I = V/R…) whose behaviour you are predicting.

A fair test changes exactly one variable — the independent variable — measures one dependent variable, and holds everything else (controlled variables) fixed, because any second change is an alternative explanation for whatever you observe. Single measurements prove little: random error scatters individual trials, so scientists run repeated trials (3 is the school minimum, 5 is better, 10 if the measurement is quick) and summarise with a mean, quoting the spread as a first uncertainty. A result is only believed when the difference between conditions is large compared with that spread — the discipline you met in the paper-airplane ranges and the pendulum's repeated timings. And a hypothesis the data reject is not a failed experiment: it is a successful measurement that ruled something out. Nobel prizes have been built on exactly that.

Variables and controls

In the virtual lab

The capstone sim is a planning studio, and this lesson uses it as one. Work the Experiment-plan builder: choose an Independent variable from the menu (balloon air %, ramp angle, string length, coil turns, light distance, fertiliser…), then a Dependent variable it could plausibly affect (travel distance, slide time, pendulum period, paperclips held, plant height, bulb brightness). Set Trials per condition — notice the choices are 3, 5, 10: plan for 5 if time allows. Tick the controlled-variable checkboxes deliberately, one by one — each unticked box is a confound you are choosing to live with. Type your hypothesis into the builder in the If… then… because… frame and press Generate my experiment plan. Then run your real experiment (bench or any earlier lesson's sim), enter results in the Data table with Add trial row, Plot my data to chart means, and keep running notes in the Observation Log. The worked example below shows what a finished pass through the studio looks like.

Hands-on: run your plan

Execute the plan you generated — here, the worked example runs on the pendulum bench from Lesson 1 (string, stand, stopwatch, three bobs). Whatever your own question, the rules are the same: measure before you theorise, write numbers down as they happen (never "fix them up" later), flag any trial that went visibly wrong (and say why you are excluding it, if you do), and keep the setup photograph or sketch with your data. Household items only; if your question involves anything hot, fast, sharp or electrical beyond battery packs, it needs an adult's approval before it needs data.

A worked investigation

Question: does bob mass affect a pendulum's period? Hypothesis: If the bob mass increases, then the period will not change, because T = 2π√(L/g) contains no mass term. Data — string 0.50 m, three 10-swing timings per mass:

Step 1 — Means. (14.2 + 14.1 + 14.3) ÷ 3 = 14.2 s; the other masses likewise 14.2 s. Period = 14.2 ÷ 10 = 1.42 s at every mass.

Step 2 — Spread as uncertainty. Trials within a mass vary by at most 0.2 s in 14.2 — about ±0.1 s, or ±0.7%. Any real mass effect smaller than ~±0.7% would hide inside this noise; none larger appeared.

Step 3 — Theory cross-check. T = 2π√(0.50/9.8) = 2π × 0.226 = 1.42 s. Measurement and the mass-free formula agree to the last digit the stopwatch can support.

Step 4 — Conclusion. The hypothesis stands: across a fourfold mass range, period is constant within ±0.7% — and the honest phrasing is "no effect detected at the ±0.7% level", not "proved exactly zero", because no measurement proves exactness. A photogate re-run could tighten the bound tenfold; that is the evaluation's suggested improvement, together with checking one more mass and a different length to confirm the L-dependence the formula does predict.

Conclusion and evaluation

The capstone's product is not the (unsurprising) result — it is the chain: mechanism-bearing hypothesis → one-variable plan → repeated trials → mean-and-spread analysis → a conclusion sized to the uncertainty. Weak links in typical first attempts: a hypothesis without a mechanism ("I think it will go further"), controls listed but not actually held (the release angle drifting), and conclusions that overclaim ("proves mass never matters" from three timings). Every one of those is fixable in the plan builder *before* a single measurement — which is precisely what the studio is for.

Safety

Design your risk assessment with your question: household materials only, battery packs (never mains) for electrical questions, adult supervision for anything hot, sharp, or launched, eye care around springs and projectiles, and water kept away from electronics. If a test cannot be made safe, it is the wrong question — redesign it.

Common mistakes

• No mechanism — "If… then…" without "because…" is a guess, not a hypothesis.
• Two variables changing together (heavier bob *and* longer string) — the result becomes uninterpretable.
• One trial per condition, then a grand conclusion — single trials are anecdotes.
• Deleting outliers silently — a rogue value gets flagged, investigated and reported, or repeated; never quietly erased.
• Confusing a rejected hypothesis with a failed experiment — the experiment succeeded; the idea failed. Science advances exactly that way.
• Precision theatre — quoting means to 4 decimal places from a ±0.1 s stopwatch.

Exam Experience

Question 1 — analysis [5 marks]. A student investigates whether ramp angle affects a toy car's speed at the bottom. Their data table lists angle, distance travelled and time; they conclude "steeper is faster". Identify two weaknesses in drawing that conclusion and the extra analysis needed.

Scoring guide:

• Speed was never computed: distance ÷ time must be calculated for each run before comparing "faster" [1]
• If distance varied between runs as well as angle, two variables changed — the runs are not comparable [1]
• Single trials (no repeats) mean no mean and no spread to judge the difference against [1]
• Extra analysis: compute mean speed per angle and compare differences with within-angle spread [1]
• A conclusion is justified only if between-angle differences clearly exceed trial-to-trial spread [1]

Question 2 — evaluation and design [5 marks]. Turn the vague question "do bigger parachutes fall more slowly?" into a complete fair-test plan.

Scoring guide:

• Hypothesis in If… then… because… form: if canopy area increases, descent time increases, because a larger canopy traps more air and drag balances weight at a lower speed [1]
• Independent variable: canopy area (defined sizes, e.g. 20 × 20, 30 × 30, 40 × 40 cm) [1]
• Dependent variable: fall time from a fixed drop height (and/or computed descent speed) [1]
• Controls: same canopy material and shape, same payload mass, same drop height, same release method (no throw), still air [1]
• Trials: at least 3 drops per size, mean and spread reported, times compared against spread [1]

Multiple choice 1. The most important feature of a scientific hypothesis is that it:

• A. Is always correct
• B. Makes a testable prediction with a mechanism ("because…") that an experiment can support or refute
• C. Is written in complicated language
• D. Agrees with what the teacher expects

Answer: B. A hypothesis earns its keep by risking falsification for a stated physical reason; correctness is the experiment's business.

Multiple choice 2. An experiment's data clearly reject the hypothesis. The experiment is:

• A. A failure that must be repeated until the hypothesis wins
• B. A successful measurement — ruling an idea out is a result
• C. Proof the equipment was broken
• D. Wasted time

Answer: B. Negative results are knowledge: the fair test did its job, and the next hypothesis starts smarter.

Going deeper: presenting like a scientist

A finished investigation deserves a finish. Science-fair judges and exam markers alike look for the same skeleton: a question sharp enough to test; a hypothesis carrying a mechanism; a method another student could repeat from your words alone — variables named, controls listed, trials counted; results in a table first and a graph second (the table is the evidence, the graph is the argument); and an evaluation that names the weakest measurement and the specific fix. Photographs of the setup, dated log entries and the raw data (crossings-out and all) count for more than a polished poster with suspiciously perfect numbers.

Present your capstone that way and you will notice something: the evaluation section writes itself, because every lesson in this course ended with one. That habit — asking "what limits this result, and what would I do next?" before anyone else asks — is the actual subject of Physics Lab. The pendulums, springs, circuits and waves were the exercises. This is the muscle.