Road Salt and Duckweed: A Single-Replicate Dose-Response Experiment
Biology students test whether road salt runoff harms duckweed, a floating plant used as an indicator species. They prepare four tanks with 500 mL of pond water and equal starting duckweed masses. Tank 1 gets no salt; Tanks 2–4 get NaCl to reach 0.5, 1.0, and 2.0 g/L. Each tank sits on the same windowsill for 10 days. At the end, students count fronds: 48, 51, 33, and 12 respectively, in order from 0 to 2.0 g/L. During the experiment, the lid on the 1.0 g/L tank was left off for three days and the water level dropped noticeably; no water was added. Each concentration was run in a single tank. The class must decide what the data support about salt and duckweed growth, how confident they can be in the pattern, and what to change before repeating the investigation.
1. Describe the pattern in the frond counts and state the strongest claim the data support. Is the claim 'salt always reduces duckweed growth' supported?
Read the worked answer
From 0 to 2.0 g/L the counts are 48, 51, 33, and 12.
- 0.5 g/L: 51 vs 48 is +3 fronds, about +6%.
- 1.0 g/L: 33 vs 48 is −15, about −31%.
- 2.0 g/L: 12 vs 48 is −36, about −75%.
The decline is steep at the high end but not monotonic: the lowest salt level is slightly above the control. With one tank per level, a 3-frond difference is well within tank-to-tank variability, so the 0.5 result is best read as 'no detectable harm,' not as evidence that salt helps.
The strongest defensible claim is that high salt concentrations (1.0 and 2.0 g/L) are associated with substantially fewer fronds, with the effect growing at 2.0 g/L. 'Salt always reduces growth' is not supported: it ignores the flat low-end result, assumes a threshold that one tank per level cannot locate, and uses 'always' where the data show only an association in one 10-day trial.
2. Explain how the uncovered 1.0 g/L tank affects the interpretation. In which direction does it bias the result, and what does that do to the dose-response pattern?
Read the worked answer
Evaporation removes water but not dissolved NaCl, so as the level dropped, the salt concentration in that tank rose above the labeled 1.0 g/L. By day 10 the tank is no longer a 1.0 g/L treatment, and nobody measured the actual concentration.
The bias runs toward exaggerating harm at that level. If the true concentration drifted, say, toward 1.3–1.5 g/L, then some of the 33-frond decline should be credited to a higher dose, and the real effect at 1.0 g/L may be smaller. That makes the dose-response curve look steeper and more regular than it is. There is a second, smaller confound: lowered water volume and any resulting crowding or root exposure could stress the plants independently of salinity.
The 2.0 g/L result - a 75% drop - is large enough that a similar drift is unlikely to erase it, so the high-dose conclusion survives. The middle point does not, and the class should not report a precise threshold.
3. Propose a redesigned experiment that would let the class make a stronger causal claim. Specify the changes and explain what each one fixes.
Read the worked answer
Redesign:
- Three or more tanks per concentration (12+ tanks total), with treatments assigned to positions at random and rotated every two days. Replicates let the class estimate tank-to-tank variability instead of reading noise as an effect, and rotation prevents a sunny or drafty windowsill spot from being confounded with a treatment.
- Control evaporation identically: cover all tanks with vented lids, or top up every tank to a marked line with distilled water on a fixed schedule. That removes the concentration drift that spoiled the 1.0 g/L tank.
- Measure the actual variable: check conductivity or salinity at the start and end of the trial, so the reported dose is the dose the plants experienced.
- Measure growth, not just a final count: record frond number and wet mass at day 0 and day 10 and report change, since equal starting mass was only asserted.
- Hold light, temperature, and water chemistry constant and use the same pond-water batch.
With replicates and measured doses, a consistent decline at 1.0 g/L would support a causal claim; without them, the class can only report an association at high concentrations.