If your stats textbook defines the null hypothesis as "the assumption of no effect" and you're still unsure what to write for your actual study, that gap between definition and sentence is exactly where students get stuck. The null hypothesis (H0) is the specific statement your statistical test is designed to reject. It says: there is no relationship, no difference, no effect. Your job is to collect enough evidence to reject it — or to fail to reject it and be honest about that too.
PaperDraft is a writing assistant, not a paper generator — the draft is your starting point, not your submission. You are responsible for editing, verifying sources, and following your school's academic integrity policy.
This guide gives you worked null hypothesis examples across psychology, biology, nursing, and education, plus the structural rules that hold across every field. For the broader decision of whether you even need a hypothesis, see research question vs hypothesis. For the full paper arc, see how to write a research paper.
The Structure Every Null Hypothesis Shares
Before the examples, the mechanics. A null hypothesis follows a predictable template:
"There is no [difference / relationship / effect] between [variable A] and [variable B] in [population]."
Paired with an alternative hypothesis (H1):
"There is a [difference / relationship / effect] between [variable A] and [variable B] in [population]."
A few rules hold across all fields:
- H0 and H1 are mutually exclusive. If one is true, the other is false.
- You never "prove" the null. You either reject it (evidence of an effect) or fail to reject it (insufficient evidence).
- The null is about the population, not your sample. You sample to infer about the population.
- Statistical significance is a threshold, not a truth. A p < .05 lets you reject H0 at the 5% significance level — it doesn't mean the effect is real, large, or important.
Psychology Null Hypothesis Examples
Psychology covers experiments, surveys, correlational studies, and interventions. Each has a typical H0 shape.
Experiment (between-groups design)
Study: Does a 4-week mindfulness intervention reduce state anxiety in undergraduates compared to a waitlist control?
- H0: There is no difference in mean state anxiety scores between the mindfulness condition and the waitlist control condition at post-test.
- H1: State anxiety scores will be lower in the mindfulness condition than in the waitlist control condition at post-test.
Test: Independent-samples t-test. Reject H0 if p < .05.
Correlational study
Study: Is social media use associated with depressive symptoms in adolescents aged 13-17?
- H0: There is no correlation between daily hours of social media use and PHQ-9 scores in adolescents aged 13-17.
- H1: There is a positive correlation between daily hours of social media use and PHQ-9 scores in adolescents aged 13-17.
Test: Pearson correlation. Reject H0 if the correlation is significantly different from zero.
Within-subjects design
Study: Does sleep deprivation affect working memory performance?
- H0: There is no difference in n-back accuracy between participants' rested and sleep-deprived sessions.
- H1: N-back accuracy will be lower in the sleep-deprived session than in the rested session.
Test: Paired-samples t-test.
Biology Null Hypothesis Examples
Biology often tests differences in means, counts, or concentrations across conditions or species.
Comparing two populations
Study: Do freshwater fish from agricultural watersheds show higher tissue microplastic concentrations than fish from protected watersheds?
- H0: There is no difference in mean microplastic concentration (particles/g tissue) between fish from agricultural and protected watersheds.
- H1: Microplastic concentration will be higher in fish from agricultural watersheds.
Test: Independent-samples t-test or Mann-Whitney U if non-normal.
Factorial design
Study: Does nitrogen fertilizer affect maize yield, and does the effect depend on rainfall?
- H0 (main effect of fertilizer): There is no difference in mean yield across fertilizer levels.
- H0 (main effect of rainfall): There is no difference in mean yield across rainfall categories.
- H0 (interaction): The effect of fertilizer on yield does not depend on rainfall.
Test: Two-way ANOVA. Each null is tested separately.
Nursing Null Hypothesis Examples
Nursing research typically tests interventions, risk factors, or outcomes across populations.
Intervention study
Study: Does a family-presence protocol reduce postoperative delirium in elderly cardiac surgery patients?
- H0: There is no difference in the rate of postoperative delirium between patients receiving the family-presence protocol and those receiving standard care.
- H1: Patients receiving the family-presence protocol will have a lower rate of postoperative delirium than those receiving standard care.
Test: Chi-square test for proportions, or logistic regression if adjusting for covariates.
Observational study
Study: Is shift length associated with medication error rates among hospital nurses?
- H0: There is no association between shift length (8 vs 12 hours) and medication error rate per 100 patient-days.
- H1: Medication error rate per 100 patient-days will be higher on 12-hour shifts than on 8-hour shifts.
Test: Poisson regression or comparison of rates with appropriate adjustment.
Education Null Hypothesis Examples
Education research often tests teaching methods, interventions, or predictors of academic outcomes.
Comparing instructional methods
Study: Does flipped-classroom instruction improve final exam performance compared to traditional lecture in introductory statistics?
- H0: There is no difference in mean final exam scores between flipped-classroom sections and traditional-lecture sections.
- H1: Final exam scores will be higher in flipped-classroom sections than in traditional-lecture sections.
Test: Independent-samples t-test, or ANCOVA if controlling for baseline differences.
Predictor study
Study: Does autonomy-supportive teaching predict intrinsic motivation in asynchronous online courses, controlling for prior GPA?
- H0: Autonomy-supportive teaching does not predict intrinsic motivation scores after controlling for prior GPA.
- H1: Autonomy-supportive teaching positively predicts intrinsic motivation scores after controlling for prior GPA.
Test: Multiple regression. Examine the coefficient and p-value for the autonomy-supportive teaching predictor.
Null hypotheses written but the blank Methods section is still staring at you? PaperDraft gives you a structured first draft — thesis stub, IMRaD skeleton, opening sections in academic register — so you can spend your time sharpening your analysis plan instead of formatting. It's a drafting assistant, not a submission. Try PaperDraft — free
Common Mistakes Students Make With Null Hypotheses
The mechanics are simple. The mistakes are consistent.
Stating H0 as "the opposite of what I expect." H0 is not your intuition. It's the "no effect" statement specific to your design. "There is no difference" or "there is no correlation" — not "mindfulness won't work."
Mixing up H0 and H1. The null always asserts no effect. The alternative always asserts the effect you're testing. Swap the names and your whole analysis logic inverts.
Writing a directional H0. The null is non-directional by convention. "There is no difference" — not "anxiety will be higher." Direction belongs in H1.
Claiming to "prove" H0. A non-significant test does not mean the null is true. It means you didn't have enough evidence to reject it. Sample size, measurement error, or effect size may be the culprits.
Mismatching H0 and your statistical test. If your H0 is about means, your test should compare means. If it's about proportions, use a proportion test. Alignment matters — see sample size in research for power considerations.
Too many nulls, no correction. If you test 20 null hypotheses at p < .05, you'll expect one false positive by chance. Use a correction (Bonferroni, FDR) when testing multiple hypotheses.
How a Drafting Assistant Fits
A drafting tool can scaffold your hypothesis section with the H0/H1 template, suggest the language that matches your design (between-groups, within-subjects, correlational), and maintain consistent formatting across the introduction and Methods. What it cannot do is pick the right statistical test for your data, verify your assumptions, or catch an underpowered design before you run the study. PaperDraft handles the structure and academic register. You handle the statistical judgment and the honest interpretation of whatever the data show.
FAQ
Can I "accept" the null hypothesis?
No. You can fail to reject it, which means the evidence wasn't strong enough to conclude an effect exists. That's not the same as proving no effect.
What does p less than .05 actually mean?
It means the probability of observing your data, or something more extreme, if the null hypothesis were true is less than 5%. It does not mean there's a 95% chance your hypothesis is correct.
Do I need to state H0 explicitly in my paper?
Depends on the field. Clinical and health sciences usually ask for both H0 and H1 in the Methods. Psychology and education often state only H1 (the research hypothesis) and let the null be implicit. Check your style guide.
What if my test statistic is non-significant?
Report it honestly: "The difference was not statistically significant, t(48) = 1.23, p = .22." Then discuss whether the study was underpowered, the effect was smaller than expected, or the theoretical prediction may be wrong.
How do I decide between a one-tailed and two-tailed test?
One-tailed requires a strong theoretical justification for direction in advance. When in doubt, use two-tailed — it's more conservative and more widely accepted.
Once your null and alternative hypotheses are clean, the Methods section has a clear target. For the next piece — justifying your sample size so your test has power — see sample size in research.