Nitrate runoff and duckweed: a confounded bench experiment
Ana is investigating whether nitrate runoff from farmland increases the growth of duckweed (Lemna minor). She collects pond water from a single site, fills 30 identical 500 mL glass jars with 400 mL, and adds potassium nitrate to make three treatments: 0, 0.5 and 5.0 mg NO3− per litre, with 10 jars per treatment. All jars are placed on one bench beside a south-facing window. After 14 days she counts the fronds in each jar and calculates the mean per jar. Her results are: 0 mg/L, 121 fronds; 0.5 mg/L, 148 fronds; 5.0 mg/L, 209 fronds. She concludes that nitrate almost doubles duckweed growth. Her teacher notices that the jars were arranged in treatment blocks from left to right, and that a pillar casts shade over the left third of the bench in the afternoon. In the 5.0 mg/L treatment, shaded jars averaged 152 fronds and unshaded jars averaged 258 fronds. Ana has no data on light intensity or water temperature in individual jars, and she counted fronds without knowing which treatment each jar belonged to.
1. Identify the two most serious weaknesses in Ana's design and explain, using her numbers, how they could produce the pattern she observed.
Read the worked answer
Two weaknesses
Light was confounded with treatment. The jars were laid out in treatment blocks from left to right, and the pillar shaded the left third of the bench. The 5.0 mg/L jars therefore received less afternoon light on average. In that treatment, shaded jars averaged 152 fronds while unshaded jars averaged 258. The shaded mean is almost identical to the 0.5 mg/L mean (148) and only 31 fronds above the 0 mg/L mean (121). The overall difference, 209 versus 121 fronds (a 73% increase), could therefore be produced largely by light rather than by nitrate.
There was no randomisation and no independent replication of the water source. All 30 jars contained water from one pond, so the 10 jars per treatment are pseudoreplicates for any conclusion about ponds. Frond number is also not the same as biomass, and evaporation or temperature differences between bench positions were not measured.
A fair test needs random placement and rotation of jars, measured light at each position, and replication across several ponds or water sources.
2. Describe two changes that would make the experiment a fair test of nitrate, and state what result would support a causal effect.
Read the worked answer
Changes
- Randomise jar positions and rotate them every two days, or use a growth cabinet with uniform light; record light intensity at every jar.
- Add a KCl control at the same potassium concentration as the highest nitrate treatment, so that the effect of K+ and ionic strength is separated from the effect of nitrate.
- Use at least five nitrate concentrations (for example 0, 0.1, 0.5, 2.0, 5.0 and 10 mg/L) and measure dry mass as well as frond number.
Result that would support causation
A monotonic dose-response in which increasing nitrate increases growth, the KCl control shows no increase, the measured light explains little of the variation, and the same pattern appears in water from at least two other ponds. If the 5.0 mg/L effect disappears once light is equalised, nitrate is not the cause.
3. State the population to which Ana's result can be generalised, and explain the limitation.
Read the worked answer
Ana can generalise only to duckweed growing in water from that particular pond under those laboratory conditions. She cannot generalise to farmland ponds in general, because pond water differs in nutrients, alkalinity, microbial community and duckweed genotype; her jars also lack sediment, herbivores and competition. The unit of inference for a pond-level claim is the pond, not the jar. To generalise, she would need replicate ponds or at least replicate water sources, and ideally a field survey showing the same association between nitrate and duckweed cover across many ponds.