The Germination Data That Cannot Say What It Claims
Class VIII (40 students, eight groups) set up germination trays for a 'what do seeds need?' investigation. The compiled class data are: A - 20 seeds on moist cotton in a dark cupboard at room temperature, 18 germinated by day 6; B - 20 seeds on moist cotton on the window sill, 5 germinated; C - 20 seeds on dry cotton in the dark, 4 germinated; D - 20 seeds on moist cotton in the dark, but from a packet opened the previous year and stored in a damp cupboard, 2 germinated; E - 20 seeds on cotton moistened with water a group had used to rinse a soapy bottle, 0 germinated. The window sill was 3–4 °C warmer in the afternoon, and one group used 30 seeds, not 20. The science exhibition is in five days, twenty fresh seeds remain, and there is no thermometer in the school. A group wants to display trays A and B side by side and label B 'light stops germination'.
1. Which comparisons in this data set are valid, and what can A versus B legitimately support?
Read the worked answer
Almost nothing here is a controlled comparison, and the honest answer to the class is that the design, not the seeds, is the problem.
- A vs C (moisture): the intended contrast, and the cleanest one - but 4/20 germinated on dry cotton, so either the cotton was not fully dry or the seeds had residual moisture. That anomaly is worth investigating rather than hiding.
- A vs B (light): confounded. The window sill was 3–4 °C warmer in the afternoon; temperature affects germination rate, and there is no thermometer to quantify or control it. B also may have had more evaporation from the cotton.
- A vs D (seed age and storage): confounded by two variables at once, so D says only 'this batch was poor'.
- E (soap residue): uncontrolled on two axes - concentration unknown, and one tray only.
- The 30-seed group: percentages, not raw counts, must be used, since 2/30 and 2/20 are not comparable as counts.
So A vs B supports exactly one statement: in this classroom, on this sill, during this week, fewer of these seeds germinated. It cannot attribute the difference to light. The 4/20 in tray C and the unexplained 30-seed tray are the two details that most convincingly demonstrate to students why a single comparison is not an experiment.
2. With twenty fresh seeds, five days and no thermometer, redesign the investigation. Which single question would you test and why, and what would you still not be able to conclude?
Read the worked answer
Choose one question that the remaining resources can actually answer, and say so explicitly. The best choice is 'do seeds of this batch germinate under our conditions, and does moisture matter?' - because seed viability is the confound that wrecked tray D, and establishing it first is what a real investigator would do.
Design: four trays of five seeds each, all on the same shelf, same day, same cotton, two moist and two dry; record daily counts and note visible mould. Five seeds per tray is small, so treat it as a pilot and state that in the display. In parallel, do a float/soak check on the remaining seeds to get a crude viability signal without a thermometer.
Two deliberate sacrifices: (a) light is dropped as a variable - there is no way to separate it from temperature here, so it becomes a question for a later, better-equipped investigation; (b) replication is too small for a strong claim. What the class still cannot conclude is anything about temperature, about light, or about seeds in general - only about this batch, this shelf, this week.
Teaching point to make out loud: choosing a question you can answer beats producing a big table of numbers whose cause cannot be identified.
3. The group insists on displaying B as 'light stops germination'. How do you respond in the five days available, and what scientific norm are you actually teaching?
Read the worked answer
Do not simply forbid the label - that teaches compliance, not science, and loses the best teaching moment in the unit.
Response: ask the group to state what the two trays differed in, and keep a list on the board until temperature appears. Then change the exhibit rather than the claim. Relabel to 'fewer seeds germinated on the window sill - we think light, but the sill was also warmer' and add a clearly drawn 'what we would need to test this properly' panel with the missing thermometer, matched trays and the extra seeds. Present the flawed design as the exhibit: 'What went wrong in our experiment and how we would fix it.' Most exhibitions show polished results; a class that can diagnose its own confounded design demonstrates more scientific control than one showing a tidy table.
If a student insists on the original label, the response is procedural rather than authoritarian: 'Which observation would prove you wrong, and did we collect it?' That question is the norm being taught - that a claim must be tied to a comparison differing in one thing, that confounds must be named, and that a stated limitation is strength rather than failure.
Also protect the other groups: several will have clean A vs C data and should not feel their work is diminished by the class's collective redesign.