Glossary

A hypothesis is a prediction the data could contradict

A hypothesis predicts a specific relationship between variables in terms precise enough that data could contradict it. Falsifiability separates it from a hunch; being written before the analysis separates it from a conclusion. Exploratory and interpretive studies often have none, and inventing one afterward misrepresents how the work was done.

Why it matters when you are writing

Statistical testing works with a pair. The null hypothesis states there is no effect or no difference; the alternative states there is. A test estimates how surprising your data would be if the null were true, which is why a non-significant result means you failed to reject the null, not that you proved it. Writing “the results proved there is no difference” is a claim the machinery cannot support, and reviewers strike it.

A directional hypothesis predicts which way the effect goes and buys statistical power in exchange for committing you in advance. A non-directional one predicts only that something differs. Choose before you look at the data, and say which you chose — switching after seeing results is the practice pre-registration exists to prevent.

The bigger honesty issue is HARKing: hypothesizing after the results are known, and presenting a pattern you discovered as something you predicted. It makes an exploratory finding look confirmatory and is one reason findings fail to replicate. The alternative costs nothing: report the analysis as exploratory and say the hypothesis it generates deserves a confirmatory test.

IN PRACTICE

Vague prediction, testable hypothesis

Before

Video use is bad for students’ sleep.

After

H1: Among first-year undergraduates, each additional hour of daily short-form video use is associated with longer self-reported sleep latency. H0: Daily video use is unrelated to sleep latency.

OFTEN CONFUSED WITH

What Hypothesis is not

Research question: A research question asks what is going on; a hypothesis commits to a specific answer that the data can refute.
Theory: A theory is a broad explanatory framework supported by accumulated evidence; a hypothesis is one narrow, testable prediction, often derived from a theory.
Assumption: An assumption is a condition you take as given for the analysis to be valid; a hypothesis is the thing you are actually testing.

Questions about hypothesis

Does every study need a hypothesis?

No. Exploratory, descriptive, and most qualitative studies proceed from research questions. Adding a hypothesis they never had misrepresents the design.

What is the null hypothesis?

The statement of no effect or no difference, which significance testing evaluates. Failing to reject it is not the same as demonstrating it is true.

Can I change my hypothesis after seeing the data?

You can generate a new one, but presenting it as your original prediction is HARKing. Report post-hoc findings as exploratory and say they need confirmatory testing.

Related terms

Research question

A research question states, in one interrogative sentence, what a study exists to find out — narrow enough that you can say what evidence would answer it and what would not.

IMRaD

IMRaD names the four-part structure of most empirical papers: Introduction, Methods, Results, and Discussion.

Meta-analysis

A meta-analysis statistically combines the results of separate studies measuring the same effect, weighting each by its precision, to produce a pooled estimate with a confidence interval.

Thesis statement

One or two sentences near the end of your introduction carry the arguable claim the whole paper defends: that is the thesis statement.

Write it once.
Prove you wrote it.

Copyeditor is an academic writing workspace — structured drafting, citations that build their own bibliography, and a signed record of how your document came together.

Apply for beta

*Beta opens this season — starting with students and researchers.