The Duckweed Fertilizer Trial: One Jar, Two Explanations
Priya wants to test whether fertilizer runoff makes duckweed grow faster. She fills two 500 mL glass jars with pond water. Jar A gets 2 drops of a liquid fertilizer labeled 5-10-5 (nitrogen-phosphorus-potassium) and is placed on the left side of the windowsill. Jar B gets no fertilizer and sits 1.2 m to the right, partly shaded by a computer monitor. She adds 10 duckweed fronds to each jar and counts fronds every second day for ten days, photographing each jar with her phone and counting by eye. On day 6 the water in Jar A has dropped about 15 mL, so she tops it up with tap water. Day 10 counts: Jar A 64 fronds, Jar B 22 fronds. She writes, 'Fertilizer tripled duckweed growth, so runoff will cause algal blooms.' Her teacher notes the fertilizer bottle also contains a blue dye and that Priya's phone photos of Jar A were taken at an angle.
1. Identify two flaws in this design that could plausibly produce the 64-versus-22 difference, and explain the mechanism by which each would inflate Jar A's count.
Read the worked answer
Flaw 1 - light is confounded with treatment. Jar A sits at the sunny left end of the sill; Jar B is 1.2 m right and shaded by a monitor. Duckweed growth rate is strongly limited by light, so Jar A may simply be photosynthesising more. Light is not a controlled variable here, it is a second difference that coincides exactly with the fertilizer difference, which means the two effects cannot be separated.
Flaw 2 - the day 6 top-up with tap water. Jar A lost about 15 mL to evaporation, so Priya replaced it with tap water. That adds dissolved minerals and changes concentration; it also means the two jars were never held at equal volume, so nutrient concentration in A is unknown at the end.
A third candidate - counting bias. She knows which jar is treated, counts by eye from angled phone photos, and Jar A has a blue dye that changes contrast. Even a small systematic over-count compounds over ten days of doubling.
Note that none of these flaws means the fertilizer did nothing. The direction is plausible. The problem is that the experiment cannot attribute the size of the difference to the fertilizer.
2. Rewrite the design as a controlled experiment. State how many jars you would use, how treatments are assigned, how light and water are equalised, and how fronds are counted without bias.
Read the worked answer
Replication. Use six jars per treatment (twelve total), each an independent unit. One jar per treatment gives no estimate of natural variability, so a difference of any size could be chance.
Assignment. Label jars A1-A6 and B1-B6, then randomly assign which physical jars get fertilizer, and randomly re-randomise their positions on one bench every day. Randomisation breaks the link between treatment and any position effect, and daily rotation averages out the light gradient instead of letting it load onto one jar.
Water and nutrients. Use the same pond water for all jars, mark the fill line on each jar, and top up every jar to the line with distilled water at every counting session so all jars experience the same dilution. Measure added fertilizer by volume with a pipette, not drops.
Measurement without bias. Cover the labels, code the photos, and have a classmate who does not know the codes count each jar twice. Better still, at day 10 harvest and dry all duckweed to constant mass and weigh it, since frond number is subjective (one frond can split into two) while dry mass is not.
3. Priya's conclusion says the fertilizer caused 'tripled growth.' Which parts of that claim are supported, and what single extra measurement would most strengthen a causal claim?
Read the worked answer
Supported: almost nothing quantitative. The treated jar ended with more duckweed, and the direction is consistent with nutrient limitation, but light, water volume and counting method all differ, so the ratio 64:22 cannot be read as a treatment effect size. A ratio from single unreplicated jars has no uncertainty attached, so 'tripled' overstates precision.
Not supported: the leap from a jar to 'algal blooms in runoff.' That is a different organism, a different scale, and a flowing system rather than a closed jar.
The most useful single addition is a dose-response with measured nutrient concentration: four or five fertilizer levels (0, 1, 2, 4, 8 drops per litre), replicated, plus a measured nitrate or phosphate reading at the start and end. This does two things at once. It confirms the treatment actually raised dissolved nutrients (rather than only adding dye), and it shows whether growth rises smoothly with dose - the pattern expected if nutrients are the cause. A single yes/no pair of jars can never distinguish cause from coincidence; a monotonic dose-response curve is much harder to explain away.