Imagine a student says, ‘Our experiment was random, so the results must apply to every plant.’ That conclusion sounds scientific, but ‘random’ can describe two entirely different decisions made at two different points in a study.
Random sampling is choosing which members of a larger group will enter a study. Random assignment is deciding, by chance, which study condition each selected member receives. Sampling mainly affects what population the study can reasonably describe; assignment helps make a comparison of conditions fairer. One does not automatically supply the other.
A diagnostic question with two doors
Ask, ‘Who got included?’ That is the sampling question. Then ask, ‘What happened to those included?’ That is the assignment question. A learner who can answer both separately already has a stronger foundation for evaluating an experiment than someone who only repeats the word ‘fair’.
An investigator might take plants that happen to be available on a classroom bench. That is a convenience selection, not a random sample of every plant of the species. The investigator could still use a random procedure to allocate those classroom plants to two light conditions. The assignment is random even though the selection was not.
Worked example: 24 classroom seedlings
Here is an illustrative, invented investigation, not a report of real experimental results. A class has 24 similar bean seedlings grown in a school tray. The class wants to compare two safe light conditions over 14 days, using the same soil type, watering schedule and measurement method where feasible.
- Selection: all 24 available seedlings are included. Because they are the available classroom plants rather than plants randomly chosen from a defined wider population, do not call this random sampling of all bean plants.
- Assignment: label the seedlings 1 through 24. Randomly shuffle 12 ‘Condition A’ labels and 12 ‘Condition B’ labels, then assign one label per seedling. Each group has 12 plants.
- Measurement: record an initial height in centimetres and a final height after the stated 14 days, using the same measurement convention.
- Outcome: calculate each seedling’s change as final height minus starting height, then compare the groups’ changes rather than comparing final heights alone.
For a calculation check, suppose one illustrative plant grows from 8 cm to 11 cm. Its change is 11 − 8 = 3 cm. Another grows from 10 cm to 12 cm: its change is 2 cm. The second plant ends taller but grows less over the interval. This is why the question and the measured quantity must agree.
If Group A’s 12 changes total 36 cm and Group B’s 12 changes total 24 cm in a hypothetical dataset, the mean change is 36 ÷ 12 = 3 cm per plant for A and 24 ÷ 12 = 2 cm per plant for B. The observed difference in means would be 1 cm per plant. These invented totals illustrate arithmetic only. They do not establish that either light condition is generally superior.
Why assignment helps but does not perform magic
Random assignment reduces the risk that a person deliberately puts the strongest seedlings into a preferred condition. It does not guarantee two small groups will be identical. Random chance can still produce imbalance; unequal positions, watering, pests or measurement decisions can also influence results. Describe these possibilities rather than claiming randomisation eliminated every source of bias.
If every plant in Condition A sits on one shelf and every plant in B sits on a different shelf, differences in shelf temperature or airflow may travel with the light condition. Independent replication or appropriately mixed positions may be needed before making a stronger causal claim. The design must remain practical and safe for the school.
Progressive guided-to-independent practice
- Guided: for a survey that invites only students from one club, identify the sampling process. If those volunteers are then allocated by a shuffled list to two learning activities, identify the assignment process.
- Supported: draw two boxes labelled ‘choose participants’ and ‘allocate conditions’. Place each study action in the correct box.
- Independent: design a fair comparison of two paper-towel materials, first stating which sheets are included and then how test order or conditions will be allocated.
- Transfer: read a news claim about an experiment, explain what the sample represents, how the conditions were assigned and which conclusion the evidence can actually support.
For parents and educators: signs of learning
Ask the student to explain the difference without using the phrase ‘making it fair’. Look for three observable behaviours: they identify the population, describe the inclusion method, and describe a separate allocation method. If any one is unclear, return to a simple drawing of the two decisions before adding statistical vocabulary.
Our related Core Aim article baseline reading versus control group untangles another common experimental mix-up. Setting a stopping rule before collecting results helps pupils make the design clearer before seeing the outcome.
Limitations and FAQs
FAQ: Does a random sample always represent the whole population perfectly? No. Chance variation, incomplete lists and missing responses can still matter. FAQ: Can an experiment be random without sampling randomly? Yes: random allocation of a convenient group is possible. FAQ: Does random allocation prove a result caused by the treatment? No; a careful design, reliable measurements and appropriate analysis are still essential.
The US Office of Research Integrity’s educational explanation of random assignment describes chance-based allocation in experiments. For wider reading, use the eduKate Education Hub and the Clementi reference on checking the thinking process, not merely recording an answer. The most important question is simple: was chance used to choose the participants, to allocate the conditions, or both?