Direct comparison
Null vs. Alternative Hypothesis
How H0 and H1 differ in hypothesis testing, and how to operationalize both precisely in a research proposal or paper’s hypothesis section.
Side-by-side comparison
| Dimension | Null Hypothesis (H0) | Alternative Hypothesis (H1/Ha) |
|---|---|---|
| What it states | No effect, no difference, or no relationship in the population | An effect, difference, or relationship exists in the population |
| Role in the test | The hypothesis the statistical test directly evaluates | The claim supported only if H0 is rejected |
| Possible test outcomes | Reject H0, or fail to reject H0 — never "accept" or "prove" H0 true | Supported (not proven) when H0 is rejected at the chosen significance level |
| Directionality | Always states equality / no difference, regardless of test type | Can be directional (one-tailed: greater than / less than) or non-directional (two-tailed: not equal) |
| Error if the decision is wrong | Type I error (false positive) — rejecting a true H0 | Type II error (false negative) — failing to reject a false H0, i.e. missing a real effect |
| Origin | Central to Fisher’s 1920s significance testing (no formal alternative used) | Added by Neyman & Pearson (1928–1933) specifically to define Type II error and statistical power |
| Common student mistake | Stated as generic "no difference" without specifying the variable, population, or comparison — not actually testable | Stated as a vague directional claim not tied to a specific measurable outcome |
Common questions
FAQ
Can a statistical test prove the null hypothesis is true?+
No. A test can only reject H0 or fail to reject it. Failing to reject H0 means the data didn’t provide strong enough evidence against it — it is not proof that H0 is true, and could reflect low statistical power rather than a genuine absence of effect.
Do I state the null or the alternative hypothesis first in a paper?+
Convention varies by field and journal, but many methods sections state the substantive research hypothesis (H1) first as the motivating claim, then formally state H0 as the hypothesis the test statistically evaluates. Some journals expect both stated explicitly and symmetrically. Check your target journal’s author guidelines.
What’s the difference between a one-tailed and two-tailed alternative hypothesis?+
A one-tailed (directional) alternative specifies the direction of the expected effect (e.g., "higher than"); a two-tailed (non-directional) alternative only states that a difference exists, without specifying direction. The null hypothesis in both cases remains the same equality statement — the choice affects the alternative and how the p-value is calculated.
Why does my null hypothesis need to be operationalized?+
Because "no difference" is not testable until it specifies which variable, how it’s measured, in which population, and between which groups. An unoperationalized null hypothesis usually signals the outcome measure hasn’t been precisely defined in the methods section either.
