Germination experiment with a suspicious replicate
A student investigates whether soaking seed in gibberellic acid (GA) improves germination of a slow-germinating pasture species. Four treatments: 0, 50, 100 and 200 mg/L GA. For each treatment, 50 seeds are placed on moist filter paper in a petri dish, with three replicate dishes per treatment, incubated at 22 °C. After 10 days the student counts germinated seeds. Results (germinated out of 50):
- 0 mg/L: 18, 20, 19
- 50 mg/L: 24, 26, 25
- 100 mg/L: 31, 30, 33
- 200 mg/L: 12, 14, 41
The 41 in the 200 mg/L group is far above its replicates (12, 14). The student is unsure whether to exclude it and asks for advice before writing the discussion.
1. Advise the student on whether the 41 should be excluded, and justify your advice using the principles of data handling in experimental work.
Read the worked answer
A value should not be removed simply because it is inconvenient; exclusion must be justified by evidence about how the value arose, and the decision must be transparent.
- Investigate before deciding. Check the raw record sheet, the dish label, and whether the 41 dish was handled differently (e.g. contaminated, placed at the edge of the incubator, lid loose, different seed lot, counted on a different day). If a documented procedural error is found - for example, the dish was mislabelled or a different seed batch was used - exclusion is defensible and should be stated with the reason.
- If no cause is found, keep it. A high value could be genuine biological variation (e.g. a seed lot effect, or a real promotive response in one dish). Removing it without cause biases the result and inflates the apparent consistency of the data.
- Use a formal, pre-specified criterion if exclusion is contemplated. For example, a stated outlier rule applied consistently to all treatments, not only the one that looks odd. Applying a rule only to the 41 would be selective.
- Report both ways. The safest approach is to present the analysis with and without the value and show that the conclusion is or is not sensitive to it. If the conclusion changes, that uncertainty must be reported.
Because the 41 is inconsistent with its own replicates but no cause is yet documented, the student should investigate first, and if nothing is found, retain the value and discuss it as unexplained variation rather than deleting it.
2. The student keeps all data. Explain what the 200 mg/L results suggest about the GA dose–response and how the student should present this in the discussion.
Read the worked answer
The 200 mg/L group has a mean of (12 + 14 + 41)/3 = 22.3, but with enormous spread (SD roughly 16), so the mean is a poor summary and the treatment cannot be described as clearly inhibitory or promotive from these data alone.
The pattern across treatments is the more informative finding: germination rises from about 19 at 0 mg/L to about 31 at 100 mg/L, then the 200 mg/L group is highly variable and its central tendency (median 14) is below the 100 mg/L group. This is consistent with a dose–response that peaks around 100 mg/L and becomes erratic or inhibitory at higher concentration - a common pattern for hormone-like treatments - but the 200 mg/L data are too variable to establish the shape confidently.
In the discussion the student should:
- present the 200 mg/L result with its variability (mean and SD or the individual values), not just the mean;
- avoid claiming a definite inhibition at 200 mg/L, instead describing it as 'variable and possibly reduced germination, requiring further replication';
- note the possible causes of the variability (seed lot, dish position, contamination) as limitations;
- recommend additional replicates at 150–250 mg/L to resolve the response at the high end;
- avoid extrapolating beyond the tested range.
This presents the result honestly and identifies the specific follow-up experiment needed, which is what a strong discussion should do.