Formulating and Testing a Hypothesis
The whole logic of hypothesis testing sits in six short lines of the supplied teaching source, and one verb in them carries more weight than everything else. Choose the wrong verb at the end of your study and every claim you make is overstated.
The supposition to be tested
Supposition is a careful word. It is not a belief, a finding or a conclusion. It is a statement you are proposing on the understanding that the study will decide against it if the evidence goes the other way — and a statement written so that the evidence is capable of going the other way.
That last clause is the whole discipline. A supposition that no realistic result could contradict is not a hypothesis, whatever it looks like on the page. It is a position, and a study built around it will produce confirmation regardless of what is in the data.
The four terms the source's cycle depends on
- Hypothesis
- "The supposition to be tested." A statement proposed in advance of the data, used to make a prediction that observing an outcome can check.
- Empirical research
- The source's label for the analysis stage — collecting data and interpreting it using statistical methods. Named as the activity, not defined further at this point in the material.
- Reject
- The verdict where the outcome is inconsistent with the hypothesis. The source states this as a consequence, without qualification.
- Support
- The verdict where the outcome is consistent with the hypothesis. Note the grammar the source uses: the experiment is said to support the hypothesis, not the other way round.
The five moves the source sets out
From supposition to the next study
State the supposition
Write the hypothesis before the data exists. It has to be capable of producing a prediction, because prediction is what the outcome will be checked against.
Collect data to test it
The purpose of collection is fixed in advance by the hypothesis. This is what makes the design deductive rather than exploratory — see Inductive and Deductive Reasoning in Research.
Analyse and interpret
The source specifies statistical methods and calls the activity empirical research. It names no method and no test, which is the largest silence on this page.
Report the confirmation or rejection, and evaluate
Two obligations, not one. Report what the analysis says about the hypothesis, then evaluate it — say what the result means, including when it is not the result you expected.
Discuss avenues for further research
The source's wording is permissive — the researcher may discuss them. In practice this is where a rejected hypothesis earns its place, by naming what should be asked instead.
Notice that the cycle closes by opening another one. A hypothesis that is rejected has not wasted the study; it has removed a candidate explanation and pointed at the next question. That framing matters more for practitioners than for anyone else, because delivery culture treats a disproved assumption as a mistake rather than as a finding.
Reject, support, and the verb the source never uses
The two outcomes are not mirror images, and the source's grammar shows why. Rejection is a conclusion about the hypothesis: it did not survive contact with the outcome. Support is a statement about one experiment: this experiment was consistent with it. Another experiment, or a better instrument, can still take it away.
WHAT EACH OUTCOME ENTITLES YOU TO WRITE
| Outcome | The source's verdict | What you may write | What you may not write |
|---|---|---|---|
| The data is inconsistent with the hypothesis | The hypothesis is rejected | "H2 was rejected. The observed difference ran in the opposite direction to the one predicted." | "The hypothesis was disproved" as a claim about the world, or a quiet decision to drop H2 from the write-up |
| The data is consistent with the hypothesis | The experiment supports the hypothesis | "H1 was supported by these data", or "partially supported", with the conditions of the study attached | "H1 was proved", "this confirms that…", "the study establishes that…" |
| The data is neither clearly consistent nor clearly inconsistent | Not addressed by the supplied source | A statement of what the analysis showed and why it does not settle the hypothesis either way | A verdict of any kind. An indeterminate result reported as support is the most common overclaim in a student thesis |
Column two is the source's wording. Columns three and four are this page's application of it; the source states the rule and does not work through how to write it up.
Why a hypothesis is only ever provisional
Two things are being defined here at once. The first is usefulness: a hypothesis earns its keep by allowing prediction, not by being interesting or by sounding plausible. If nothing follows from it that could be observed, it is not useful in the source's sense.
The second is the clause "within the accuracy of observation of the time". Support is always relative to the measurement available when the test was run. Better instruments, larger datasets, longer periods and cleaner records can all move a hypothesis from supported to superseded without anyone having made an error.
- Hypothesis predicts accurately
- Observation improves
- Prediction stops holding
- New hypothesis challenges it
- New one supersedes if it predicts better
The supersession test is comparative, and specifically so: the new hypothesis wins to the extent that it makes more accurate predictions than the old. Not because it is newer, not because it is more sophisticated, and not because the people proposing it are more senior. That is a usable standard for a practitioner assessing a contested claim in their own field.
Writing a hypothesis that can be rejected
The source defines the hypothesis and its outcomes without setting out how to compose one. What follows is this page's synthesis of what the source's own rules require — a hypothesis must be a supposition, it must allow a prediction, and an outcome must be capable of being inconsistent with it.
FIVE PARTS OF A TESTABLE STATEMENT
| Part | What it fixes | Failure if it is missing |
|---|---|---|
| Population | Who or what the statement is about, bounded by sector, organisation type, project class or period | The claim is about everything, so no dataset can contradict it |
| The thing that varies | The condition, practice or characteristic you expect to make a difference | You cannot say what would have to change for the outcome to change |
| The thing measured | The outcome, defined in a form that already exists in records or can be collected consistently | The measure gets chosen after the data arrives, which is the integrity problem the material raises and never resolves |
| Direction | Higher, lower, more frequent, earlier — a stated expectation rather than "there is a relationship" | Almost any result can be read as consistent, so the test decides nothing |
| Comparison | The group, period or condition the prediction is measured against | A number with nothing to compare it to cannot be inconsistent with anything |
This page's synthesis. The supplied source states the definition and the outcome rule; it does not prescribe a structure for writing a hypothesis.
What the source leaves to the statistician
One consequence is worth stating plainly. Because the criterion for inconsistency sits outside this material, it also sits outside your proposal unless you put it there deliberately. The research design section is where it belongs, alongside the analysis method it depends on — see The Research Design Section.
Stating and adjudicating hypotheses in your document
A hypothesis is a structural device as well as a logical one. Where you put it, and whether you come back to it, determines whether a reader can check your study against its own intentions.
The hypothesis set, before the proposal goes in
- Each hypothesis is one statement, with one population, one measure and one direction
- Each one is traceable to the literature that motivated it, not to a hunch about your own organisation
- You can name the result that would reject each one, and you would accept that result if it arrived
- The measure exists now, in records you have confirmed access to, or in an instrument you have designed
- The criterion for deciding inconsistency is written down and dated before any data is collected
- The hypotheses are numbered, and the numbering is the same in the design section and the conclusion
- Nothing in the set can only be answered by asking people whether they agree with it
When a hypothesis is the wrong instrument
Hypothesis, or question
What to carry forward
- A hypothesis is the supposition to be tested — proposed before the data, written so the data could contradict it.
- Five moves: state it, collect to test it, analyse and interpret statistically, report the confirmation or rejection and evaluate it, then discuss further research.
- Inconsistent outcome, hypothesis rejected. Consistent outcome, the experiment supports it. Never proves.
- Support is relative to the accuracy of observation at the time. A better instrument can supersede a supported hypothesis without anyone having erred.
- A useful hypothesis allows prediction; a new one supersedes the old only to the extent that it predicts more accurately.
- Build the statement from five parts — population, the thing that varies, the thing measured, direction, comparison. Missing any one makes rejection impossible.
- The source names statistical methods and defines none. Fix your criterion for inconsistency, in writing, before collection begins.
Frequently asked questions
Why can I not say my research proved the hypothesis?
Because the supplied source never claims that outcome is available. It states that an inconsistent outcome rejects a hypothesis and a consistent outcome means the experiment supports it. Support is a statement about one study run with the measurement available at the time, and a later study with better observation can supersede it. Writing "proved" claims something your design cannot deliver.
Is a rejected hypothesis a failed research project?
No. Rejection is a result — it removes a candidate explanation and gives the discussion of further research something concrete to work with. The material's own cycle ends with discussing avenues for further research, which is exactly what a rejection opens up. What damages a study is quietly dropping a rejected hypothesis from the write-up rather than reporting it.
Do I need hypotheses in my research proposal?
Only if your design is deductive and your analysis can actually reject them. Hypotheses suit studies that derive a directional expectation from existing theory and test it statistically. Where the literature has not characterised the phenomenon, or where the evidence is accounts rather than counts, research questions do the job better and make no promise the design cannot keep.
How do I decide when an outcome is inconsistent with my hypothesis?
Not from this material. It specifies analysis using a variety of statistical methods and names none of them, so it supplies no threshold, no test and no significance criterion. That decision rule has to come from a statistics text, your supervisor or your field's methods literature — and it must be written down before you collect anything, because a threshold chosen after seeing the data is not a threshold.
How many hypotheses should a master's project have?
The supplied material sets no number. What it does say about research questions is that they should be few so that the focus is manageable, and the same constraint applies with more force to hypotheses, because each one needs its own measure, its own analysis and its own verdict in the conclusion. Count the analyses you can actually run and let that set the limit.
Where should hypotheses appear in the document?
The most useful observed pattern places each hypothesis immediately after the literature that motivates it, so a reader can see it was derived rather than assumed, and then restates all of them verbatim in the conclusion with a verdict attached. That makes the study checkable against its own intentions in a single page and makes negative results presentable.
References and source attribution
- Bryman, A. 2016, Social Research Methods, 5th ed., Oxford University Press, Oxford — cited in the supplied source; the standing reference for deductive design and hypothesis formulation.
- Trochim, W. M. K. 2006, Research Methods Knowledge Base — cited in the supplied source, including its material on problem formulation, of which hypothesis statement forms a part.
- Quinlan, C. 2011, Business Research Methods, 1st ed., Cengage Publishing — cited in the supplied source as the text behind its methodological framework.
- O'Leary, Z. 2017, The Essential Guide to Doing Your Research Project, 3rd ed., Sage Publications, London — one of the general research methods texts cited across the supplied source.
- The supplied teaching source: consolidated weekly teaching notes and slide material on developing a research proposal, which supplies the hypothesis definition, the five-step cycle, the reject-or-support rule and the supersession passage quoted on this page.
Suggested questions for Ask KEVOS
- Rewrite my hypothesis so that a result in the opposite direction would clearly reject it.
- Check my hypothesis for the five parts — population, variable, measure, direction and comparison.
- My conclusion says the study proved something. Rewrite it using the source's reject and support vocabulary.
- Should this project use hypotheses or research questions, given my data and sample size?
- Turn my three compound hypotheses into separate statements, each with its own measure and verdict.
- Draft the paragraph I would have to publish if my main hypothesis were rejected.