The scientific method, step by step

Updated November 2026

The scientific method is the way scientists test ideas against evidence: make an observation, propose an explanation that could turn out wrong, predict what you should see if it is right, then check. Real research loops back on itself many times, so think of it as a cycle of habits rather than a fixed recipe.

The short version

• Typical steps: observe → question → hypothesis → prediction → test → analyze → conclude → share → replicate.
• A good hypothesis is testable and falsifiable: some possible result would show it is wrong.
• A fair test changes one thing (the independent variable), measures the outcome (the dependent variable) and keeps everything else the same, with a control group for comparison.
• Science does not "prove" ideas once and for all. Ideas that survive many tests and are replicated by others become well supported.

The steps

1. Observe. Notice something: a pattern in data, a result that does not fit, a question from everyday life.
2. Ask a question. Make it specific. "Why are plants sick?" is too broad; "Does this fertilizer change how tall bean seedlings grow in three weeks?" can be answered.
3. Form a hypothesis. A proposed explanation or answer that can be tested. Read what is already known first; your question may already have been studied.
4. Make a prediction. State what you would observe if the hypothesis is right, usually as "if … then …". This is where the hypothesis becomes testable.
5. Test it. Run an experiment, or, where experiments are impossible (astronomy, much of epidemiology), make new observations that could contradict the prediction.
6. Analyze the results. Summarize the data, look at graphs, and use statistics to judge whether a difference is larger than you would expect from chance variation (see the null hypothesis and p-values).
7. Draw a conclusion. Do the results support the prediction or contradict it? What are the limits of the study?
8. Share and get peer review. Present the work (see giving a clear science talk) and submit it to a journal, where other experts check the methods and reasoning before publication.
9. Replicate. Other groups repeat the work, ideally with new samples. A result that holds up again and again earns trust.

A cycle, not a recipe

Textbook lists make science look like a straight line. In practice, an unexpected result sends you back to the question; a failed test leads to a new hypothesis; analysis reveals a new pattern to observe. Researchers also move between two kinds of reasoning: inductive (from many observations to a general pattern) and deductive (from a general idea to a specific prediction you can check). Not every science is experimental. Field biology, geology and astronomy often test ideas by making careful new observations rather than by manipulating anything.

Falsifiability

A hypothesis is useful only if some result could count against it. "Plants grow better with this fertilizer" is falsifiable: if fertilized plants grow no taller than unfertilized ones in a fair test, the claim fails. "Plants grow better when they feel loved" is not, unless you define how to measure it. A result that agrees with a prediction supports a hypothesis but does not prove it, because another explanation might predict the same result. Rejecting one hypothesis also does not confirm a rival one; each needs its own test.

Controls and variables

An example experiment

A teacher notices that students seem to remember a sung list better than a spoken one. Here is how that observation could become a fair test.

• Question: Do students recall the eight Krebs cycle intermediates better a week later if they learn them with a song?
• Hypothesis: Learning the list to a simple, repeated melody improves later recall compared with learning the same words spoken.
• Prediction: If the hypothesis is right, the song group will recall more intermediates in the correct order after one week.
• Design: 60 volunteers are randomly assigned to two groups of 30. Both get the same list, the same 10 minutes of study, the same rhythm and the same quiz a week later. The only difference (independent variable) is melody versus speech. The outcome (dependent variable) is the number of intermediates recalled in order, scored by someone who does not know which group each student was in.
• Analysis: Compare group averages, report the difference with a confidence interval, and run an appropriate test.
• Conclusion: Whatever the result, report it, including limits: one class, one list, one week.

Real studies along these lines have had mixed results. One study of about 1,000 mostly school-age students (Crowther and colleagues, 2016) found that music and non-music videos produced similar quiz gains, although students preferred the music, which is exactly why careful comparisons and replication matter. See what the research says about songs and learning.

Memory aid

"Only Quiet Hippos Predict The Answer Correctly, So Remember": Observe, Question, Hypothesis, Prediction, Test, Analyze, Conclude, Share, Replicate. Then picture the last word bending back to the first to make a loop.

A verse to remember it

Check yourself

Sources

• 1.1 The Science of Biology, Biology 2e, OpenStax
• The real process of science, Understanding Science, University of California Museum of Paleontology