A hypothesis is a testable prediction about the relationship between two or more variables. To write one, start from your research question, identify the independent variable (what you change or compare) and the dependent variable (what you measure), predict how they are related, and state it as a clear sentence that data could prove wrong. You then write a matching null hypothesis that predicts no effect. This guide explains each step with examples.
What a hypothesis is (and is not)
A hypothesis:
- is a statement, not a question;
- makes a prediction about variables;
- is testable: data could show it to be wrong;
- is based on theory or previous research, not a guess.
Hypotheses are used mainly in quantitative research, such as experiments and surveys analyzed with statistics. Qualitative studies usually work with research questions instead.
Key terms
| Term | Meaning | Example |
|---|---|---|
| Independent variable (IV) | What you change, compare or group by | Hours of sleep |
| Dependent variable (DV) | What you measure | Exam score |
| Alternative hypothesis (H1) | Predicts an effect or relationship | More sleep leads to higher exam scores |
| Null hypothesis (H0) | Predicts no effect or relationship | Hours of sleep have no effect on exam scores |
| Directional hypothesis | Predicts the direction of the effect | Higher, lower, more, less |
| Non-directional hypothesis | Predicts a difference without its direction | There is a difference |
How to write a hypothesis in 5 steps
1. Start with a research question
Does sleep affect exam performance among secondary-school students?
See how to write a research question.
2. Do some reading
Look at what earlier studies found. If several studies link sleep to memory consolidation, you have a reason to predict that more sleep improves exam scores.
3. Identify your variables
- Independent variable: hours of sleep the night before the exam.
- Dependent variable: exam score.
- Population: secondary-school students aged 15–18.
4. Write your prediction
State the relationship in a clear, declarative sentence in the present tense:
Secondary-school students who sleep at least eight hours the night before an exam score higher than students who sleep less than six hours.
5. Write the null hypothesis
The null hypothesis predicts no difference or relationship between the same variables:
There is no difference in exam scores between secondary-school students who sleep at least eight hours and those who sleep less than six hours the night before an exam.
Statistical tests assess whether your data give enough evidence to reject the null hypothesis.
Hypothesis examples
Directional
Employees who work from home at least three days a week report lower job-related stress than employees who work in the office every day.
Non-directional
There is a difference in reading comprehension scores between students who read on paper and students who read on screens.
Correlational
Among adults aged 65 and over, the number of minutes of daily walking is positively correlated with cognitive test scores.
Causal (experimental)
Patients who receive a daily text-message reminder take their medication on more days than patients who receive usual care.
With more than one independent variable
Both the frequency of practice tests and the amount of sleep predict final exam scores among first-year psychology students.
Weak vs strong hypotheses
| Weak | Problem | Stronger |
|---|---|---|
| Social media affects people. | Vague variables, not testable | Teenagers who use image-based social media for more than three hours a day report lower body satisfaction than those who use it for less than one hour. |
| Will exercise improve mood? | A question | Adults who take part in a six-week aerobic exercise program report greater improvements in mood than adults on a waiting list. |
| Coffee is good. | Not measurable | Drinking 200 mg of caffeine one hour before a test improves reaction time in sleep-deprived adults. |
How to test a hypothesis
Your hypothesis determines your design and analysis:
- Comparing two groups: an experiment or survey, analyzed with a t-test (or a non-parametric alternative).
- Comparing more than two groups: ANOVA.
- A relationship between two continuous variables: correlation or regression.
- Several predictors: multiple regression.
Describe your design and analysis in the methodology chapter. See how to write the methodology chapter.
Checklist
- Is it a statement, not a question?
- Does it name both variables and the population?
- Are the variables measurable?
- Could your data show it to be wrong?
- Is it based on theory or previous research?
- Is there a matching null hypothesis?
Try the free hypothesis generator
The hypothesis generator turns your topic or research question into testable hypotheses, each with its null hypothesis, its variables and a suggested way to test it.
Frequently asked questions
What is the difference between a hypothesis and a prediction?
A hypothesis is a general statement about a relationship between variables. A prediction applies the hypothesis to a specific study: what you expect to see in your data.
Do qualitative studies need a hypothesis?
Usually not. Qualitative research explores experiences and meanings with open research questions rather than testing predictions.
Can a hypothesis be wrong?
Yes, and that is still a valid result. Failing to reject the null hypothesis is a finding you report honestly.
How many hypotheses should a study have?
As many as your research questions require, usually one to four. Each should test a specific relationship.