Interpreting an Experiment With a Confound
A student designs an experiment to test whether fertilizer increases bean plant growth. She plants 10 beans in pots by a sunny window and 10 beans in pots in a shaded corner. She adds fertilizer to the sunny group only and plain water to the shaded group. After three weeks, the sunny, fertilized plants average 22 cm, while the shaded plants average 14 cm. She concludes the fertilizer caused the extra growth. Her teacher points out that two variables differ between the groups. The student must decide what the experiment can and cannot show and how to redesign it.
1. Identify the variables that differ between the groups and explain why the conclusion is not justified.
Read the worked answer
Two variables differ: fertilizer treatment (fertilized versus plain water) and light exposure (sunny window versus shaded corner). Because both change together, the results cannot be attributed to either one alone. The 8 cm difference (22 minus 14) could come from extra light, from fertilizer, or from some combination.
This is a confounding problem. A valid conclusion about fertilizer requires that the only systematic difference between groups be the fertilizer. Here the design also lacks a control for light, so the experiment is confounded. The student's conclusion overreaches: the data are consistent with fertilizer helping, but they do not isolate its effect. A careful answer should say the experiment is inconclusive about fertilizer specifically, not that fertilizer has no effect.
2. Describe a redesign that would let her test the fertilizer claim, including what she should keep the same and how she would interpret results.
Read the worked answer
She should hold light constant by placing all 20 pots in the same light conditions, for example all by the sunny window or all under identical grow lights. Then she should randomly assign 10 plants to fertilizer and 10 to plain water, keeping pot size, soil, water volume, and temperature the same. Random assignment reduces the chance that pre-existing differences between plants drive the result.
After three weeks she compares average heights. If the fertilized group is consistently taller and the difference is larger than normal plant-to-plant variation, that supports the fertilizer claim. She should also consider repeating the experiment or using more plants, since 10 per group is small and a single run could reflect chance. The key principle: change one variable, control the rest, and assign randomly.