Biochar and Lettuce: A Greenhouse Trial with Confounded Light
Year 10 students test whether biochar improves lettuce growth. They fill 24 pots with the same compost and add biochar to 12 pots at 5% by volume; the other 12 are controls. All pots are placed in one row along a greenhouse bench. The end nearest the door receives noticeably more afternoon light. Students water by hand whenever the surface looks dry, and they harvest after six weeks by cutting the tallest leaf. Mean shoot height is 18.4 cm for the biochar pots and 15.1 cm for controls. One control pot died and was excluded. No soil moisture, temperature, initial seedling mass or pot position was recorded. The students want to report that biochar increased growth by 22% and that the result is statistically significant.
1. Identify three design weaknesses in this trial and explain how each could produce the observed difference between treatments.
Read the worked answer
No randomisation or blocking along the light gradient. Pots were placed in one row, and the door end was brighter. If the biochar pots happened to occupy the brighter part of the bench, light rather than biochar could explain the height difference. Even if the treatments were alternated, position effects were not recorded or controlled.
Uncontrolled and unequal watering. Watering by hand 'when the surface looks dry' is subjective. Biochar changes water retention, so the two treatments may have received different effective water amounts even if the students tried to treat them alike. Without soil moisture data, the mechanism cannot be separated from a watering artefact.
Measurement and handling bias. Harvesting 'the tallest leaf' is a subjective, destructive measure and is not the same as total biomass. A control pot died and was excluded, which removes a poor-performing control and inflates the control mean downward or the treatment difference upward. There is no initial seedling mass, so pre-existing size differences are unknown.
A 22% difference can arise from any of these. The trial shows an association, not a demonstrated biochar effect.
2. Describe a practical redesign the students could run in the same greenhouse next term, including how to assign pots and what to measure.
Read the worked answer
Use the same number of pots but arrange them in blocks along the bench, for example eight blocks of three positions. In each block, randomly assign one biochar pot and one control pot, with the third position used for a second control or a second biochar rate. Randomise within each block using a random-number table, and record every pot's position. This makes each treatment equally exposed to the light gradient.
Before planting, weigh each seedling or record initial leaf area and assign similar-sized seedlings across treatments. Use the same compost batch and pot type. Replace hand watering with a measured schedule: weigh a subset of pots daily and add a fixed volume to bring them to a target mass, or use saucers and record water added. Record soil moisture and greenhouse temperature at fixed times.
At harvest, measure shoot dry mass and root dry mass, not 'tallest leaf'. Dry mass at 70 °C to constant mass is less subjective. Keep the dead pot in the dataset and report it. Analyse paired differences within blocks, report means with 95% confidence intervals, and avoid claiming statistical significance without an appropriate test.
3. The students calculate that 18.4 cm is 22% higher than 15.1 cm. Assess this calculation and advise how they should report the result instead.
Read the worked answer
The calculation is (18.4 − 15.1) ÷ 15.1 × 100 = 21.9%, so 22% is arithmetically correct as a relative difference in mean shoot height. It is not a valid estimate of a biochar effect because the design confounds treatment with position, watering and possibly initial size, and because one control was removed.
The students should report the absolute difference, 3.3 cm, with a measure of variability and the sample size. For example: 'Mean shoot height was 18.4 cm in biochar pots and 15.1 cm in controls, a difference of 3.3 cm (n = 12 and 11 after one control died).' They should give the standard deviation or confidence interval for each mean. If they use a blocked design next time, they can report the mean within-block difference and its confidence interval.
They should avoid the phrase 'statistically significant' unless they have run a test that accounts for blocking and have stated the significance level. They should also avoid saying biochar 'increased growth' as a causal claim; 'was associated with taller shoots in this trial' is accurate.