The scientific method in practice: hypotheses, experiments and replication

TUREFF• •0

School science often presents the scientific method as a tidy five-step recipe. Real research is messier, with false starts, surprises and arguments, but the core ideas hold up: make a claim that could be wrong, test it fairly, and see whether others get the same result.

Start with a question and a testable hypothesis

Science usually begins with an observation that needs explaining. Why do some plants in a field grow taller than others? A hypothesis is a proposed explanation that makes a specific prediction, for example that the taller plants are getting more nitrogen. The key feature of a good hypothesis is that it is falsifiable: there must be some possible result that would show it is wrong. "The plants grow better because of an undetectable energy" cannot be tested, so it is not a scientific hypothesis, however appealing it sounds. A good hypothesis also says what you expect to see in numbers or clear observations before you look, so that you cannot reinterpret the results to fit afterwards.

Design a fair test

An experiment changes one thing and watches what happens, while keeping everything else as similar as possible.

Independent variable: the thing you deliberately change, such as the amount of fertiliser.
Dependent variable: what you measure, such as plant height after six weeks.
Controlled variables: everything you hold constant, such as light, water, soil and seed variety.
Control group: plants given no added fertiliser, so you have something to compare against.

Protect against bias

People are very good at seeing what they expect. Good experiments build in safeguards.

Randomisation: decide by chance which plants, patients or samples get which treatment, so that hidden differences are spread evenly.
Blinding: where possible, the person measuring the results should not know which group each sample came from. In medicine, double-blind trials keep both patients and researchers unaware of who received what.
Enough repeats: one plant per group tells you almost nothing. More samples make it easier to tell a real effect from chance variation.
Pre-registration: writing down the hypothesis, methods and analysis plan before collecting data, so that the plan cannot drift towards whatever looks interesting later.

Analyse, then try to break your own result

Once the data are in, researchers use statistics to ask how likely it is that a difference this large would appear by chance alone. A result that passes this test is evidence, not proof. Good scientists then look for weaknesses: could something else explain the pattern? Did the fertiliser change soil acidity as well as nitrogen? A result that contradicts the hypothesis is still useful. Negative results stop others wasting time and help refine the next question.

Replication: the step that matters most

Replication means repeating a study to see whether the result holds. There are two broad kinds.

Direct replication: following the original methods as closely as possible, ideally in a different laboratory.
Conceptual replication: testing the same idea in a different way, for example with a different plant species or measurement.

Over the past decade or so, large organised efforts to repeat published studies in psychology, medicine and other fields have found that a substantial share of results came out weaker than first reported, or did not appear at all. This is often called the replication crisis. Common causes include small samples, flexible analysis choices and journals favouring surprising positive results. The response has been constructive: more pre-registration, larger collaborative studies, shared data and code, and journals that agree to publish a study based on its design before the results are known.

Theories, laws and changing your mind

In everyday speech a theory is a guess. In science a theory is a well-tested explanation that ties together many observations, such as evolution or plate tectonics. A law describes a regular pattern, often mathematically, without necessarily explaining why it happens. Science is provisional: any idea can be revised by strong enough evidence. That is a strength, not a weakness. Well-established theories rarely get overturned entirely; more often they are refined or shown to apply within certain limits.

Try it at home

You can practise the method with simple kitchen experiments, such as whether seeds sprout faster on damp cotton wool or in soil. Write your prediction down first, keep everything else the same, use several seeds per group and record what happens. What experiment did you first run that surprised you? Tell us in the comments.