Experimental Design, Variables and Controls
This domain asks you to design and critique experiments. Most questions give a scenario and require you to sort the variables, identify the comparison group, and point to the flaw that would block a causal conclusion.
The three variable types
- Independent variable (IV) - what the experimenter deliberately changes; it needs at least two levels.
- Dependent variable (DV) - what is measured in response; it must be measurable and defined operationally ("plant height in mm from soil to growing tip", not "growth").
- Controlled variables - everything held constant across groups: temperature, light, volume, species, instrument.
Control group versus controlled variables
A control group is a separate set of subjects that receives no treatment, or a placebo. It exists so that change can be attributed to the IV. Controlled variables are the constants. Marks are routinely lost by swapping these two terms.
Design quality
- Random assignment of subjects to groups prevents systematic differences between them.
- Blinding participant and tester reduces expectancy effects; a placebo is required in human trials of subjective outcomes.
- Repeat trials and a sample large enough to average out natural variation.
- One IV at a time; if two factors change together, attribution is impossible.
- Measure the DV the same way, with the same instrument and observer, in every group.
Worked example
A student predicts that fertilizer increases tomato yield. She takes 30 identical seedlings in identical soil and pots, all on one bench, waters each with 200 mL daily, and gives 15 plants fertilizer A and 15 plain water. All the plants on the sunny side of the bench grow better. Identify: IV = fertilizer A versus none; DV = number of ripe tomatoes per plant after 8 weeks; constants = species, soil, pot size, water volume, duration. The control group is the water-only plants. The flaw is that light exposure is confounded with treatment. The fix is to randomise pot positions or rotate them regularly, so light is spread across both groups.
Traps
- Drawing a conclusion about the IV when a second factor varied.
- No baseline measurement, so pre-existing differences stay invisible.
- Generalising from a small convenience sample to all plants, people or ecosystems.
- Measuring the DV only once at the end when a growth curve would be more informative.