· By Thesisroom
Research Hypothesis: How to Write a Testable One, With Examples
How to write a research hypothesis for a thesis or proposal: null and alternative pairs, when qualitative studies skip one, eight steps and a worked example.
This guide has a printable version: open the printable hypothesis checklist
A research hypothesis is a testable statement of the result you expect, written before you collect data. A methods guide by Barroga and Matanguihan in the Journal of Korean Medical Science calls it “an educated statement of an expected outcome.” In a thesis or proposal, it names the variables, predicts how they relate, and gives a tentative answer to your research question that your data can then test.
This guide is for Master’s and doctoral candidates writing a proposal or the early chapters of a thesis. It covers how a hypothesis differs from a research question, null and alternative pairs, when a thesis needs no hypothesis at all, eight steps to write one, and a worked example.
What is a research hypothesis in a thesis?
It is your prediction about the answer to one research question, phrased so the data can support it or not. National University’s dissertation guide quotes Creswell’s definition: “A hypothesis is a formal statement that presents the expected relationship between an independent and dependent variable.”
Barroga and Matanguihan list what makes a hypothesis good. They write that “The hypotheses must be testable and specific,” and that good ones are “empirically testable,” “backed by preliminary evidence” and “testable by ethical research.” They also say research questions and hypotheses “should be developed before the start of the study.”
A thesis hypothesis does not come from nowhere. The George Washington University’s Graduate School of Education and Human Development puts it this way in its dissertation guide: “Hypotheses are deduced from theory or induced from accumulated knowledge.”
How is a research hypothesis different from a research question?
The question asks; the hypothesis answers in advance. Barroga and Matanguihan define a research question as “what a study aims to answer after data analysis and interpretation.” A hypothesis, in their words, “provides a tentative answer to the research question to be tested or explored.”
Here is the difference on one topic:
- Research question: Is weekly quizzing related to test anxiety among first-year students in an introductory statistics course?
- Hypothesis: First-year students in a section with weekly low-stakes quizzes will report lower test anxiety at the end of term than students in a section without them.
The question can be answered with “yes”, “no” or “it depends”. The hypothesis commits to one answer and a direction, so the data can contradict it. If you cannot imagine a result that would count against your statement, it is not yet a hypothesis.
What are null and alternative hypotheses?
They are the two halves of one test. The null hypothesis (H0) says there is no relationship or difference; the alternative (H1 or Ha) says there is one.
Barroga and Matanguihan define the null as “A negative statement indicating no relationship or difference between 2 variables” and the alternative as one that “predicts a relationship between 2 study variables.” National University’s guide asks for both: “for every research question, there is a set of null and alternative hypotheses statements.”
Written as a pair for the quiz example:
- H0: There is no difference in end-of-term test anxiety between first-year students in the weekly-quiz section and students in the section without weekly quizzes.
- H1: First-year students in the weekly-quiz section will report lower end-of-term test anxiety than students in the section without weekly quizzes.
What is the difference between directional and non-directional hypotheses?
A directional hypothesis says which way the effect goes; a non-directional one says only that there is an effect. Barroga and Matanguihan describe a directional hypothesis as one that “Identifies study direction based on theory towards particular outcome.” For a non-directional one, the “Nature of relationship between two variables or exact study direction is not identified.”
The H1 above is directional: it predicts lower anxiety. The non-directional version would say test anxiety will differ between the two sections. Choose a direction only when your literature review gives you a reason to expect it, and say what that reason is in the sentence before the hypothesis.
Does every thesis need a research hypothesis?
Not always, and the guides below do not agree on every design. National University’s guide expects hypotheses for every quantitative research question, while the George Washington University guide says they are used only sometimes outside experiments. Rules differ between universities and programs, so read your own handbook before you decide.
- Quantitative studies: National University’s guide says “It is not generally permissible to have quantitative research questions for which there are no hypotheses,” nor hypotheses “that don’t connect to a research question.”
- Design matters: The George Washington University guide says “Hypotheses are used in experimental research and sometimes in quasi-experimental research and nonexperimental research.”
- Qualitative studies: National University’s guide says that because qualitative studies “do not involve any testing of relationships between variables,” “qualitative RQs are not written by hypotheses.”
- Rare exceptions: The University of Northern Iowa’s research methods text says qualitative research relies on broad “how” or “what” questions, and that “while hypotheses in qualitative research may exist, they are rare.”
- Hypotheses at the end: The George Washington University guide notes that “qualitative studies may result in emergent hypotheses articulated at the conclusion of the study during the interpretation phase (Chapter 5).”
Barroga and Matanguihan add that qualitative work can be hypothesis-generating: an exploratory study may allow “formulation of a formal hypothesis potentially testable in a future quantitative approach.” For mixed-methods work, the George Washington University guide says that research “may use both approaches.”
How do you write a research hypothesis? Eight steps
Work from the question to the test, and fix the wording before any data arrive.
- Start from one research question your design can answer. A hypothesis can only be as clear as the question above it. If the question is still vague, go back to the gap it comes from; our guide on how to find and test a research gap shows how to narrow it.
- Find out what theory and earlier studies predict. National University’s guide says the predictions “should be firmly rooted in the researcher’s understanding of the literature and theory.” Write down the two or three studies or the theory your prediction rests on. Our worked theoretical framework examples show how a theory turns into constructs you can measure.
- Name the variables and who they are measured in. Barroga and Matanguihan’s “simple hypothesis” “Predicts relationship between single dependent variable and single independent variable.” Say what each variable is, how you will measure it, and in which population.
- Decide on a direction only if the literature gives you one. If earlier work points one way, write a directional hypothesis. If it is mixed, a non-directional hypothesis is the honest choice.
- Write the alternative and the null as a pair. Keep the same variables, population and measure in both, so the only change is “difference” against “no difference”.
- Match each hypothesis to the analysis that will test it. Every hypothesis needs a named test or analysis in your method section. Our parent guide on writing a data analysis plan before you have data shows how to pair each hypothesis with its test.
- Read the hypothesis against the rest of your proposal. The variables must be ones your instruments measure, in the people you plan to sample, over the time you have. Thesisroom’s Scope Check reads a proposal and quotes pairs of passages where what it asks and what it plans sit apart, each with one question about how they join.
- Fix the wording before you collect data. Presenting a hypothesis formed after the results as if you had it all along has a name. Kerr’s paper on HARKing defines it as “presenting a post hoc hypothesis (i.e., one based on or informed by one’s results) in one’s research report as if it were, in fact, an a priori hypotheses.”
What does a weak hypothesis look like next to a testable one?
A weak hypothesis leaves out the variables, the population or the direction, or uses a word no instrument can measure. The examples below are invented to show the repair.
| Weak version | What is wrong | Testable version | What was fixed |
|---|---|---|---|
| Social media affects students. | No population, no measure, no direction | Among first-year undergraduates at one university, more daily hours on social media will be related to lower self-reported sleep quality. | Named the population, both variables and the direction |
| Mindfulness training is beneficial for nurses. | “Beneficial” cannot be measured | Ward nurses who complete an eight-week mindfulness course will report lower scores on a named burnout scale than nurses on the waiting list. | Replaced a value word with a measured outcome and a comparison group |
| Does remote work affect productivity? | It is still a question | In the two firms studied, remote support staff will close more tickets per week than office-based support staff. | Turned the question into a prediction with a countable outcome |
| Small firms that adopt digital payments grow faster and hire more staff. | Two predictions in one sentence, so partial support is unclear | H1a: Small firms that adopt digital payments will show higher sales growth over two years. H1b: They will add more staff over the same period. | Split into two hypotheses, each with its own test |
| The new reading app will work because students enjoy it. | “Work” is vague, and “because” adds a second claim that is not tested | Fourth-grade students who use the reading app for one term will score higher on the school’s end-of-term reading test than students who do not. | Named the outcome and dropped the untested reason |
The pattern is the same in every row: a population, a cause or grouping, a measured outcome and, where you have grounds, a direction.
A worked example: from question to hypotheses
This example is invented. A Master’s candidate in education teaches two sections of an introductory statistics course at her university. She wants to know whether weekly low-stakes quizzes are linked to test anxiety.
- Her first draft: “Quizzes help students.” It names no outcome, no population and no comparison, so no result could count against it.
- Her research question: Is weekly quizzing related to end-of-term test anxiety among first-year students in an introductory statistics course?
- Her variables: The grouping is the section (weekly quizzes or no weekly quizzes). The outcome is each student’s score on one published test-anxiety questionnaire, given in the last week of term. She names the questionnaire and its source in her method section.
- Her direction: In this invented case, the studies in her literature review point toward lower anxiety with frequent low-stakes testing, so she writes a directional H1 and cites them in the paragraph before it.
- Her pair: H1 says the weekly-quiz section will report lower end-of-term test anxiety than the other section. H0 says there will be no difference between the two sections.
- Her limit, stated up front: Students were not randomly assigned to the two sections. So she predicts a difference between groups, not that quizzes cause lower anxiety, and she lists this in her limitations.
Her supervisor’s one change was in step 3: measure anxiety at the start of term too, so a difference at the end is not just a difference the sections began with. That adds a second, linked hypothesis about the change in scores, which she pairs with its own test.
If she had instead asked how students experience weekly quizzes, she would write open “how” and “what” questions and no hypothesis, in line with the qualitative guidance above.
How many hypotheses should a thesis have?
None of the guides cited here sets a number of hypotheses. Each hypothesis pair belongs to one research question, so the number follows the number of questions you can test well.
National University’s guide says “it is recommended that a single, central research question be developed.” It also says “you should limit the number of research questions to avoid scope creep and taking on too much with your research.” It also asks for a null and alternative pair for every quantitative question. Your program may set its own expectation, so ask your supervisor or committee early.
A useful test: if you cannot name the analysis for a hypothesis in one line, you probably have one hypothesis too many, or two hypotheses hiding in one sentence.
Where does the hypothesis go in a thesis or proposal?
Valdosta State University’s format introduces the questions and hypotheses in Chapter 1, and the George Washington University guide also states them in full in Chapter 3, the method chapter. Rules differ between universities, and these are two named examples, not a general rule.
- Valdosta State University: Its suggested Chapter 1 format has a “Purpose of the Study” section: “There should be one or two paragraphs to introduce the research questions and hypotheses.” A numbered “Research Questions” list follows. It adds that headings “may vary somewhat depending on the nature of the study.”
- The George Washington University (GSEHD): Chapter 1 has a “Purpose and Research Questions” section, and Chapter 3 states the “specific research questions and/or hypotheses to be investigated” under quantitative research. It adds: “Your research questions must be presented exactly as you state them in Chapter 3.”
The same guide says its dissertation proposals “include the elements normally found in Chapters 1, 2, 3, and the References.” So at GSEHD, a proposal’s hypotheses sit where its question and method sit. Our guide to the format for a PhD research proposal shows where the research questions section falls and how much space it usually gets.
What are the common weaknesses in a thesis hypothesis?
Most weak hypotheses fail one of the tests in the eight steps. These are the ones to look for in your own draft:
- Written after the results: Presenting a reworded hypothesis as if it were your original prediction is HARKing, as Kerr defines it. Keep your original wording and report the result as it came out.
- No measurable variable: Words like “beneficial”, “effective” or “better” need an instrument and a score behind them.
- No population: A hypothesis about “students” or “firms” in general cannot be tested on the sample you actually have.
- Two predictions in one sentence: If one half is supported and the other is not, you cannot say what happened. Split them.
- A direction with no reason: A directional hypothesis needs literature or theory behind it, cited before you state it.
- No link to a research question: National University’s guide treats a hypothesis with no question above it as not okay.
- A claim of cause the design cannot carry: Barroga and Matanguihan describe a causal hypothesis as one where “An effect on dependent variable is predicted from manipulation of independent variable.” Without that manipulation, predict a relationship or a difference instead.
A result that goes against your hypothesis is not a failed thesis. Our guide to defending quantitative research in a viva gives wording for explaining a null result to examiners.
Printable hypothesis checklist
Print this page and tick each box for every hypothesis in your proposal or method chapter.
- It answers one named research question.
- It is a statement, not a question.
- It names the population or sample it applies to.
- It names the independent variable or grouping.
- It names the dependent variable and how it will be measured.
- Any direction (higher, lower, more, less) is backed by cited literature or theory.
- It has a matching null hypothesis with the same variables and measure.
- It makes one prediction only, or is split into H1a and H1b.
- It claims cause only if the design manipulates the cause.
- A named test or analysis in the method section will test it.
- If your handbook requires it, the wording is the same in Chapter 1 and the method chapter.
- It was written before data collection, and the date of the final wording is recorded.
- My program’s handbook allows or requires hypotheses for this design.
| Hypothesis label | Research question it answers | Variables and measure | Direction and reason | Test or analysis |
|---|---|---|---|---|
| H1 | ||||
| H2 | ||||
| H3 |
Frequently asked questions
What is a research hypothesis in simple words?
It is your best prediction of the answer to a research question, written before you collect data. It names what you will measure, in whom, and what you expect to find, so the results can support it or count against it.Is a hypothesis the same as a research question?
No. A research question asks what you want to find out. A hypothesis gives a tentative answer to that question, sometimes with a direction, and your analysis tests it. National University’s guide asks for a null and alternative pair for each quantitative research question.Can a qualitative thesis have a hypothesis?
It can, but it is rare. Qualitative studies usually use open “how” and “what” research questions instead. Some end with hypotheses that emerge from the findings, for a later study to test. Check what your program’s handbook expects for your design.What is the difference between H0 and H1?
H0, the null hypothesis, states that there is no relationship or difference between the variables. H1, the alternative hypothesis, states that there is one, and may give its direction. Both use the same variables, population and measure.How many hypotheses should a thesis have?
None of the guides cited here sets a number of hypotheses. Each hypothesis pair belongs to one research question, and National University’s guide advises limiting the number of questions to avoid scope creep. Your supervisor or committee may have a view for your field, so ask early.Can I change my hypothesis after I see my results?
You should not present a changed hypothesis as if it were your original prediction. That practice is called HARKing. Report the hypothesis you set, the result you got, and any new idea the results suggest as a question for future work.Sources
- Barroga E, Matanguihan GJ. A Practical Guide to Writing Quantitative and Qualitative Research Questions and Hypotheses in Scholarly Articles. Journal of Korean Medical Science
- National University Library: Research Questions/Hypotheses, Dissertation and Applied Doctoral Project Components
- University of Northern Iowa: Research Methods, Chapter Five: Qualitative Methods (Part 1)
- The George Washington University, Graduate School of Education and Human Development: Dissertation Content and Style Guide
- Valdosta State University: A Suggested Format for Chapter 1 of the Dissertation
- Kerr NL. HARKing: Hypothesizing After the Results are Known. Personality and Social Psychology Review
How this guide was made: Thesisroom drafted this guide with the help of AI writing tools. Every fact, figure and source in this post was checked against its original source before publishing.