Biology · Life, Evidence & Biological Scale
Comparisons, controls, variation and causal reasoning
A plant on a windowsill and a plant in a cupboard differ in more than light. Temperature, water loss and starting condition could differ too. If their extensions differ, which change explains the result? The comparison has not separated the alternatives.
No earlier biology course is needed. The data lesson explains its own arithmetic. Chemistry and mathematics offer useful next steps, but you can read this chapter before completing their maps.
Back to the Biology mapBy the end, you can
Recognise a useful comparison and explain why additional measurements do not always mean additional independent evidence.
Specify what changes and what is measured
The experimental variable is the condition we assign; here it is a specified light treatment. The response is what we measure; here it is the change in stem length during seven days. A comparison group provides a reference condition. It need not be a group receiving no light, and it does not mean that nothing else happens to it.
In the imagined design, individually potted seedlings come from the same seed lot. Light conditions are applied separately to each pot, and both conditions occur across positions. Age, substrate, water and observation times are matched. Temperature is monitored because a light arrangement that also changes temperature would test a combined change.
Random allocation reduces systematic selection differences. It does not guarantee identical groups, especially with small samples. Random sampling concerns which individuals enter a study; random allocation concerns which treatment the selected individuals receive. One does not substitute for the other.
A lamp can change more than the light
Read the diagram in words
A candidate causal diagram has arrows from a lamp arrangement to light and temperature, and arrows from both conditions to stem extension. If the arrangement changes both, the contrast does not isolate a light effect. Arrows represent possible influences, not effects established by the supplied data.
- Lamp arrangement → Light condition
- Lamp arrangement → Temperature
- Light condition → Stem extension
- Temperature → Stem extension
Count the units that receive the treatment
An experimental unit can receive a treatment independently of the other units. If a light condition is applied separately to each pot, the pot can be the unit for that treatment. If a whole chamber receives one light condition, the chamber is the unit for that treatment. Ten seedlings within one chamber do not become ten independent assignments of the chamber's light.
Measuring the same plant repeatedly can show its change over time or help assess measurement consistency. It does not create new plants. Calling repeated measurements independent replicates would overstate how much independent evidence we have.
A useful design therefore asks two questions: what received the treatment, and what was measured? The answers may be different. More readings can improve a record without repairing a comparison based on one chamber per condition.
Variation is part of the result
Individual seedlings need not give the same measurement. Show their values, rather than retaining only the averages. Differences between organisms, local conditions and measurement processes can all contribute to the spread. We need a design and further investigation to distinguish these contributions.
Our small constructed sample is for learning how to reason. It supplies no formal statistical test or calculated interval for a population effect. Overlapping individual values alone neither demonstrate nor rule out statistical significance. Describe the sample accurately and keep any wider conclusion conditional.
Worked example
Twenty seedlings, two chambers
Suppose ten seedlings share a lower-light chamber and ten share a higher-light chamber. The chamber identities and treatments are inseparable. A difference might reflect another chamber condition. Taking three readings from each seedling gives sixty readings, but still only one chamber per treatment. The extra readings cannot supply the missing independent chamber comparison.
Check your understanding
You measure one seedling ten times. Have you created ten independent seedlings?
Show reasoning
No. You have repeated observations of one seedling. The experimental unit and the dependency between readings remain.
Does random treatment allocation guarantee that a small sample has perfectly balanced starting conditions?
Show reasoning
No. It limits selection bias but chance imbalances remain possible. Record starting conditions and assess the design.
What to carry forward
Causal reasoning needs a meaningful comparison, relevant controls and correctly identified independent units.
References
Resources supporting this lesson’s explanations and examples.