Experiments that distinguish explanations
Turn a build failure into a testable comparison with controls, recorded conditions, and evidence that can separate competing causes.
- Design a repeatable one-factor experiment with a prediction, control, observations, and a conclusion limited to the evidence.
Understand
The working idea
An experiment is useful when possible results distinguish explanations. If a small fan stops intermittently, “try things until it works” can hide the cause. Competing hypotheses might be a loose connector, low source voltage, or a blocked rotor. Each predicts a different pattern. A connector fault may respond to gentle cable movement; a weak source may show voltage sag; a blockage should correlate with mechanical resistance.
Start with a specific question and response variable. Define the factor you will change, hold other relevant conditions as constant as practical, and state predictions before collecting data. A control or baseline shows what happens without the change. Repetition reveals variability. Record actual observations separately from expected ones, including inconvenient outcomes. An experimental conclusion should say which hypothesis the result supports under tested conditions, not declare universal proof.
Quantitative comparisons need normalization. If comparing two toy-car wheel materials, use the same car mass, surface, start method, and distance. Measure travel time or stopping distance, not “felt faster.” Randomize trial order when warming batteries or changing surfaces could create a trend. For a sensor, a build decision such as adding a shield should follow a measurable improvement across representative lighting, not one successful demonstration.
A common failure is changing several things together—battery, code, wiring, and sensor angle—then crediting the last change. Another is confirmation bias: repeating failed trials but stopping after the first desired result. Predefine the number of trials and decision rule. Null results are information; they may reject a hypothesis or reveal that the measurement lacks sensitivity. The method does not remove judgment, but it makes the chain from observation to design choice inspectable and repeatable.
Follow the reasoning
Worked example
Two wheel materials are tested for 1.0 m travel. Foam trials take 4.2, 4.0, 4.1 s; plastic trials take 5.0, 4.8, 4.9 s. Compare mean speed under the tested conditions.
- 1
Foam mean time = (4.2 + 4.0 + 4.1)/3 = 4.1 s, so pooled distance/time speed ≈ 1.0 m / 4.1 s = 0.244 m/s.
- 2
Plastic mean time = (5.0 + 4.8 + 4.9)/3 = 4.9 s, so mean speed ≈ 0.204 m/s.
- 3
Relative difference is (0.244 − 0.204)/0.204 ≈ 0.196, or about 20%.
Foam produced about 20% higher pooled distance/time speed in these six controlled trials. This ratio is not the arithmetic mean of the individual trial speeds.
Sanity checkThe shorter mean time must correspond to higher speed over the same distance. The data do not establish performance on other surfaces or loads.
Try it yourself
Independent check
Move to the bench
Determine whether a paper hood changes a light sensor's repeatability without claiming more than the data show.
- low-voltage light sensor or phone light-sensor app
- paper hood
- fixed lamp
- ruler
- timer
- data table
Experiment
- Choose one fixed sensor-to-lamp distance and define the reading spread as the response variable.
- Predict how the hood will affect the mean and range, then precommit to five alternating trials per condition.
- Collect actual hood/no-hood readings without moving lamp or sensor marks.
- Deliberately shift the lamp for one marked trial; identify it as a protocol fault and analyze both with and without it rather than silently deleting it.
- Compare means and ranges, then write a conclusion limited to this geometry and lamp.
Observe
- Expected: the hood may reduce stray-light variation but can also lower the mean; actual direction and magnitude must come from the recorded trials.
- Expected fault clue: the shifted-lamp trial changes both conditions' comparability and should be traceable in notes.
Plan a controlled test of whether a vacuum-blower model's inlet shape changes airflow, naming one response measure, three controls, and a stopping rule.
- Use a cool low-voltage lamp or ambient light; keep paper away from hot bulbs.
- Do not infer eye-safety, appliance safety, or outdoor reliability from this bench experiment.
Keep evidence
Bench note
Record what you expected, what you actually measured or observed, and one difference worth investigating. This is self-reported evidence, not a certification.