How do you write a research hypothesis, with examples?

Direct answer: A research hypothesis is a clear, specific and testable statement that predicts the relationship between variables in your study. Praharaj and Ameen (2025) add that it is always generated before the research is carried out, so a prediction written after you have seen the results does not count as a hypothesis.

Praharaj and Ameen (2025) put the definition in one line: "A research hypothesis is a predictive statement about the expected relationship between two or more variables" (Praharaj & Ameen, 2025, p. 86). The Texas A&M University Writing Center (n.d.) draws the line between a question and a hypothesis on the basis of prior knowledge. A research question can be posed with no idea of what the data might show, whereas a hypothesis is formed from what you already know about the topic, so it predicts the outcome rather than asking about it. McCombes (2022) makes the same point from the student side: a hypothesis is a tentative answer to your research question, and it should rest on existing theory and earlier studies rather than on a hunch.

Not every project needs one. The University of Melbourne, Academic Skills (n.d.) states plainly that not all research has a hypothesis, and that whether you need one depends on your discipline and topic. Praharaj and Ameen (2025) give the usual exception, descriptive studies that only observe and record the characteristics of a population, which can work with an aim, research questions and objectives alone.

Direct answer: A good hypothesis sentence is a statement, not a question, that names the variables, the group being studied and the predicted outcome. It must be falsifiable, meaning a possible result could show it is wrong, and it should avoid conditional words such as should, might or could (Texas A&M University Writing Center, n.d.).

The University of Melbourne, Academic Skills (n.d.) reduces the standard to three tests. A hypothesis should be in the form of a statement, it should be disprovable, and it should be clear and specific in scope. Its model example reads: "Undergraduate students who participate in a stress-reduction seminar will report an increase in their well-being after six weeks." That sentence names one group, one intervention, one outcome and one time frame, so a pre and post survey could support it or refute it.

Cuttler (2017), writing for psychology students, explains why testability matters. A hypothesis is a specific prediction about what should be observed if a particular theory is accurate, which is why theories and hypotheses share an if-then relationship. The Texas A&M University Writing Center (n.d.) shows the opposite case: a moral or ethical question, such as whether students should follow an honour code, cannot become a hypothesis, because no observation could test it. The practical fix is to establish variables that can be observed or measured inside the sentence itself.

The first two stronger versions in the table come from the worked examples in McCombes (2022). The third and fourth are illustrations written for this guide.

Direct answer: Start from a focused research question, read the theory and previous studies, write an initial answer, refine it until it names the variables, the group and the predicted outcome, phrase it in a standard form, and add a null hypothesis if you will test it statistically. McCombes (2022) sets out these six steps.

McCombes (2022) uses one running example that shows each step at work. The research question asks whether students who attend more lectures get better exam results. Preliminary reading suggests attendance and performance are linked, so the initial answer becomes "attending more lectures leads to better exam results". Refining it means adding the relevant variables, the specific group being studied and the predicted outcome, which narrows the population to first-year students and the outcome to exam scores.

The fifth step is choosing a sentence form, and McCombes (2022) offers three. The if-then form states the independent variable first and the dependent variable second: "If a first-year student starts attending more lectures, then their exam scores will improve." The effect form, which is more common in academic writing, states the relationship directly: "The number of lectures attended by first-year students has a positive effect on their exam scores." The comparison form states the difference you expect between two groups. The sixth step adds the null hypothesis, the default position of no association, written H0, alongside the alternative hypothesis H1 or Ha. Our guide to writing a dissertation proposal shows where these statements sit in the proposal itself.

Direct answer: Good examples follow the same pattern across disciplines: a defined group, an independent variable, a dependent variable and a predicted direction. The table below gives one example per discipline, drawn from university guides, a journal article and Scribbr, so you can copy the structure rather than the topic.

Two patterns stand out when the rows are read together. Each example names who is studied, what changes and what is measured, which is exactly the refinement step in McCombes (2022). Each one could also fail, because a result showing no difference, or a difference in the other direction, would count against it. An environmental science student might write, as an illustration, that "urban sites with more than 30% tree canopy record lower summer surface temperatures than sites with less than 10%". That sentence follows the same pattern, and the thresholds would come from the student's own literature review.

Direct answer: Psychology hypotheses are usually derived from a named theory and tested in an experiment or survey, so the strongest ones state what the theory predicts in a specific situation. Cuttler (2017) notes that the very best hypotheses distinguish between competing theories, because one result supports one theory and undermines the other.

Cuttler (2017) gives an example from expressive writing research. If habituation theory is correct, writing about positive experiences should not improve health in the way writing about traumatic experiences does, because it would not cause people to habituate to negative thoughts and feelings. A student hypothesis built on that logic might predict fewer signs of emotional distress with each new writing session. Cuttler (2017) also describes a study by Schwarz and colleagues that asked people to recall either 6 or 12 occasions when they were assertive, which tested whether people judge themselves by the number of examples they recall or by how easily the examples come to mind. A psychology hypothesis written in that spirit names the theory, the manipulation and the measure in one sentence.

Direct answer: There is no single official list of seven. Lists online mix categories, such as simple, complex, directional, non-directional, null, alternative and associative, that overlap. For a thesis, the distinctions that matter are null versus alternative, directional versus non-directional, and primary versus secondary (Praharaj & Ameen, 2025).

Praharaj and Ameen (2025) describe the pair most methods chapters need. The null hypothesis asserts no difference or relationship between the variables, typically written as "there will be no difference between...", while the alternative hypothesis is a positive or directional statement, such as "there will be a greater reduction in mean scores". They also note that a study may state secondary hypotheses alongside the primary one, mentioned only when they are relevant. Dissertations often include both the null and the alternative form, even though many journal papers state only the alternative.

Directional and non-directional hypotheses link to one-tailed and two-tailed tests. Praharaj and Ameen (2025) report that two-tailed tests are preferred in practice, and that one-tailed tests are rarely carried out. If you need the statistical side of the null and alternative pair in more depth, read our guide on choosing the right statistical test.

Direct answer: The most common mistakes are writing the hypothesis as a question, hedging with should or might, combining two relationships in one sentence, and writing the hypothesis after seeing the results. Kerr (1998) named that last habit HARKing, short for hypothesizing after the results are known.

Kerr (1998) describes HARKing as presenting a hypothesis that was formed after looking at the data as if it had been stated in advance. It removes the point of the test, because the prediction is no longer independent of the evidence used to check it, and it contradicts the requirement in Praharaj and Ameen (2025) that a hypothesis is generated a priori. The Texas A&M University Writing Center (n.d.) adds a reporting rule that prevents a related error: when you write up results, state that H0 was rejected in favour of the alternative, rather than claiming the alternative was proved. Keeping one relationship per hypothesis also keeps each statistical test answering exactly one question, which makes your research methodology chapter easier to justify.

Many students write their hypotheses alone, and the skill is worth building yourself. A second reader can still help with the parts that are hard to see from inside a project: whether a variable is defined clearly enough to measure, whether two relationships are hiding in one sentence, and whether the hypothesis matches the design you plan to use. Through research proposal support at MAAS, you write the proposal and the hypotheses, and an expert in your field reviews your drafts and comments on how they connect to your question and method. Bring your research question, your reading list and your supervisor's feedback to the first conversation so the comments respond to your actual project.