To write a hypothesis, turn your research question into a clear, specific, and testable prediction about the relationship between two or more variables, usually phrased as an if…then statement that your study can either support or refute. In short: a hypothesis is your educated guess, written precisely enough that data can prove it wrong.
This guide covers exactly how to write a hypothesis from scratch — the six steps from question to statement, the difference between the null and alternative hypothesis, the three ways to phrase one, real worked examples (from school science fairs to PhD research), a copy-ready template, and the mistakes that get a hypothesis rejected by markers. It is written for UK students at every level, using British English and academic-integrity-safe guidance.
What Is a Hypothesis (and What It Is Not)
A research hypothesis is a specific, testable prediction that answers your research question, grounded in existing theory, prior studies, or preliminary observation. It is not a wild guess, an opinion, or a question — it is a precise statement that proposes a relationship between variables and that your data can confirm or contradict. A good hypothesis must be based on reasoning, facts, and established theory, and it must be testable through data analysis, observation, experiment, or other recognised scientific methodologies that can support or refute it.
The simplest way to recognise a hypothesis is by what it does: it predicts. A research question asks “Does sunlight affect plant growth?”; the hypothesis answers, “Plants that receive more sunlight grow taller than plants that receive less.” That shift — from open question to falsifiable claim — is the heart of learning how to write a hypothesis.
The Anatomy of a Hypothesis: Variables
Every hypothesis is built from variables — characteristics or facts that can change or vary and that you propose are related. Most studies involve at least two: the cause you control and the effect you measure. Understanding these is the single most useful thing you can do before writing, because the variables literally become the two halves of your sentence.
- Independent variable (IV): the factor you change or manipulate — for example, hours of sunlight exposure.
- Dependent variable (DV): the factor you measure — for example, plant height in centimetres.
- Confounding (extraneous) variables: anything else that could affect the DV — soil quality, watering frequency, plant species — which you must control or hold constant.
This relationship between an IV and a DV is most obvious in experimental designs, where the researcher actively manipulates the independent variable. In correlational studies you do not manipulate anything — you measure whether two variables move together — but you still name both clearly. Choosing the right approach is part of your wider research designs decision, and it shapes how directional your hypothesis can be.
How to Write a Hypothesis: 6 Steps
Here is the core process for how to write a hypothesis for a dissertation, an undergraduate lab report, or a school science project. The same logic scales from a Year 5 science fair to PhD research — only the complexity of the variables changes.
Step 1: Start with a focused research question
Begin with a specific, researchable question about a topic that genuinely interests you. It should be clear, narrow, and answerable with evidence — for example, “Does the amount of sunlight a plant receives affect its growth?” A vague question (“Is sunlight good for plants?”) produces a vague hypothesis, so sharpen the question first.
Step 2: Do preliminary background reading
Before you predict anything, find out what is already known. Read prior studies, theories, and observations connected to your question. This stops you from proposing something already disproved and gives your prediction a defensible rationale — markers want to see that your hypothesis is informed, not invented.
Step 3: Define your variables
Identify your independent and dependent variables explicitly, with measurable values where possible. For the plant example: IV = amount of sunlight (2, 4, or 8 hours per day); DV = plant growth (height in centimetres). Operationalising variables — saying exactly how you will measure them — is what makes the later hypothesis testable.
Step 4: Formulate the hypothesis
Now write the prediction. A hypothesis states the expected relationship between your variables, most often as an if…then statement: “If plants receive more sunlight, then they will grow taller.” Put the independent variable in the “if” clause and the dependent variable in the “then” clause.
Step 5: Make sure it is testable and falsifiable
A good hypothesis can be empirically tested and, crucially, disproved. You should be able to design a study whose results could contradict it. If no possible result could prove your statement wrong, it is not a scientific hypothesis. For the plants, you can expose groups to different sunlight levels and measure growth — a result that could clearly go either way.
Step 6: Account for confounding variables
Finally, list the factors other than your IV that could influence the DV, and decide how to control them. Soil quality, water frequency, and plant type all affect growth, so you would hold these constant across groups. Naming confounders shows examiners that any effect you find can be attributed to your independent variable, not to noise.
Filled in: “If students sleep at least eight hours before an exam, then their test scores will be higher than those of students who sleep fewer than six hours, because adequate sleep improves memory consolidation and concentration.”
Three Ways to Phrase a Hypothesis
The same prediction can be worded in several valid ways. Use whichever fits your design, but keep the independent variable first and the dependent variable second. Below, the same idea — that attending Zumba fitness classes improves health — is phrased three ways.
| Phrasing style | When to use it | Example |
|---|---|---|
| If…then | Experiments where you manipulate the IV | If a person attends Zumba fitness classes regularly, then their health will improve. |
| Direct relationship | Correlational studies that link two measured terms | The number of Zumba classes attended has a positive effect on a person’s health. |
| Group comparison | Studies comparing two or more groups | People who attend most Zumba classes will have better health than those who attend few. |
Directional vs. Non-Directional Hypotheses
A directional (one-tailed) hypothesis predicts the direction of the effect — more, less, higher, lower. A non-directional (two-tailed) hypothesis predicts that a difference or relationship exists, but not which way it goes. Use a directional hypothesis when prior research gives you good reason to expect a specific direction; use a non-directional one when the evidence is mixed or the area is under-researched.
Research question: Does listening to music while studying affect how well students retain information?
Directional: Students who listen to music while studying will retain less information than those who study in silence.
Non-directional: There will be a difference in information retention between students who study with music and those who study in silence.
Null vs. Alternative Hypothesis
If your research uses statistical analysis, you must state both a null and an alternative hypothesis. The two are mirror images: the alternative is the effect you expect to find, and the null is the default position that no effect or relationship exists. You never “prove” the alternative directly — instead you gather evidence to reject the null. The figure below shows how the two divide every possible outcome between them.
By convention, H₀ denotes the null hypothesis and H₁ (sometimes written Hℹ or Hₒ) denotes the alternative. Writing both keeps your study honest: it forces you to specify what “no effect” would look like before you see the data.
H₀ (null): The number of Zumba classes a person attends has no effect on their health.
H₁ (alternative): The number of Zumba classes a person attends positively affects their health.
More Hypothesis Examples Across Levels
The clearest way to learn how to write a hypothesis is to read good ones. These span school projects through to higher-level research, so you can see the same structure at every stage.
| Level | Research question | Hypothesis |
|---|---|---|
| School / science fair | Does the amount of sunlight a plant receives affect its growth? | Plants that receive more sunlight will grow taller than plants that receive less sunlight. |
| School / science fair | Does water temperature affect how fast sugar dissolves? | Sugar will dissolve faster in hot water than in cold water. |
| Undergraduate | Do students who eat breakfast perform better in exams? | Students who eat breakfast will score higher in exams than students who skip it. |
| Undergraduate | Does screen time before bed affect sleep quality? | Greater screen time in the hour before bed is associated with poorer self-reported sleep quality. |
| Postgraduate | Does remote working affect employee productivity? | Employees who work remotely three or more days a week report higher task productivity than fully office-based employees. |
What Makes a Strong Hypothesis: A Checklist
Before you commit a hypothesis to your proposal, test it against these five criteria. A statement that fails any one of them usually gets flagged by markers.
- Clear: it states a definite relationship between named variables.
- Testable: you can design a study to investigate and measure it.
- Specific: it avoids vague terms and predicts a precise outcome.
- Falsifiable: some possible result could prove it wrong.
- Relevant: it answers your research question and aligns with existing knowledge.
Common Mistakes to Avoid
Most weak hypotheses fail for a handful of predictable reasons. Watch for these:
- Writing a question instead of a statement (“Does X affect Y?” is a question, not a hypothesis).
- Being too vague — “Diet affects health” names no measurable variables.
- Making it untestable or unfalsifiable, so no result could ever contradict it.
- Predicting more than one relationship in a single sentence, which becomes impossible to test cleanly.
- Forgetting the null hypothesis when your study uses statistics.
- Confusing the independent and dependent variable, which reverses the logic of the prediction.
“A hypothesis is a tentative statement about the relationship between two or more variables. It is a specific, testable prediction about what you expect to happen in a study.” — American Psychological Association (APA)
Where the Hypothesis Sits in Your Research
A hypothesis is not a standalone sentence — it is the hinge between your question and your method. It grows directly out of your research question and your literature review, and it then dictates everything downstream: the variables you measure, the data you collect, and the statistical test you eventually run. Get the hypothesis right and every later chapter has a clear target to aim at; get it wrong and your method ends up answering a different question from the one you posed.
For quantitative dissertations, the hypothesis usually appears at the end of the introduction or the literature review, immediately after you have justified it with evidence. It must be consistent with the variables you go on to define in your methodology and with the analysis you plan in your statistical analysis. Note that not every study needs a formal hypothesis: purely exploratory or qualitative research often uses research questions or aims instead, because the goal is to discover patterns rather than test a fixed prediction. Use a hypothesis when you have enough prior knowledge to predict a specific outcome.
| Stage | What you produce | Example |
|---|---|---|
| 1. Research question | An open, focused question | Does sleep duration affect exam performance? |
| 2. Variables | Named IV and DV | IV: hours slept; DV: exam score |
| 3. Hypothesis (H₁) | A testable prediction | More sleep before an exam leads to higher scores. |
| 4. Null hypothesis (H₀) | The “no effect” default | Sleep duration has no effect on exam scores. |
| 5. Test & conclusion | Evidence to reject or retain H₀ | Data show a significant effect → reject H₀. |
Remember that rejecting your own hypothesis is not a failure. If your data do not support your prediction, you have still produced a valid, publishable result — one that can refine the theory and shape future research questions. Examiners reward a well-reasoned, well-tested hypothesis far more than a “correct” one that was never genuinely at risk of being disproved.
For deeper background on how predictions are evaluated once your study runs, see our companion guide to hypothesis testing, which walks through significance levels, p-values, and the decision rules that determine whether you reject the null. This article stays focused on writing the hypothesis itself; the testing guide takes over from there.
Struggling to write or test your hypothesis?
Our expert UK academics help students frame, refine, and test hypotheses across every discipline — with 100% satisfaction guaranteed.
