What Is a Hypothesis? Definition, Types, and How It Works
Ben Orlin draws on classroom teaching experience and a deep interest in making mathematics understandable. His practical knowledge connects mathematical ideas with everyday examples, helping readers explore concepts and approach learning with curiosity.
Quick answer
A hypothesis is a testable, evidence-based proposed explanation for an observation. It can be confirmed or ruled out through experiment. It is not a wild guess, and it is not yet a theory. It sits between a first observation and a proven scientific conclusion.
What a Hypothesis Actually Is
A hypothesis is a proposed explanation for something you observe. It must be based on prior observations. The key word is testable: a statement only counts as a scientific hypothesis if some observation or experiment could show it to be wrong.
That last point trips up a lot of students. A hypothesis is not simply any guess. It is an educated, observation-based proposal that commits to a specific, checkable prediction. If no conceivable result could ever contradict it, it is not a scientific hypothesis.
In formal logic, the word carries a different, narrower meaning: in a statement like “If P, then Q,” the hypothesis is the antecedent P. That usage appears in math proofs and philosophy courses, so it helps to know which context you are in.
Hypothesis vs. Theory: The Most Common Misconception
Everyday speech treats “hypothesis” and “theory” as near-synonyms, both meaning something like “a hunch.” In science, they mean very different things, and mixing them up leads to real confusion.
A hypothesis is a starting point: a single proposed explanation waiting to be tested. A scientific theory is what a hypothesis can become after many experiments confirm it, done by many different researchers. Calling evolution or germ theory “just a hypothesis” misunderstands this progression. Theories are not upgraded guesses; they are well-tested explanatory frameworks built from many confirmed hypotheses.
Observation leads to a hypothesis. Testing either supports or refutes it. A hypothesis that survives many independent tests, across many researchers, can eventually become part of a theory.
Testability and Falsifiability
Philosopher Karl Popper argued that a hypothesis must be falsifiable: there must be a possible result that would prove it false. This is not a technicality. It is what separates science from unfalsifiable claims that can never be checked.
Consider two statements:
- “Plants grow taller when given more sunlight.” Falsifiable. You could grow plants under different light levels and measure the results.
- “An invisible force makes plants grow, but it hides whenever you measure it.” Not falsifiable. No experiment could ever contradict it.
Only the first qualifies as a scientific hypothesis. Some thinkers have added other criteria, such as verifiability and coherence with existing knowledge. But testability remains the core requirement across approaches.
A useful hypothesis lets you make predictions. It should tell you what you expect to observe before you run the experiment, not after. If you already know the outcome, the statement describes a consequence, not a hypothesis.
How Researchers Evaluate Competing Hypotheses
When more than one hypothesis could explain the same observation, researchers weigh them against several criteria. These are not arbitrary preferences; they reflect what makes an explanation genuinely useful in science:
| Criterion | What it means |
|---|---|
| Testability | Can the hypothesis be checked by observation or experiment? |
| Parsimony | Does it avoid unnecessary assumptions? (Occam’s razor) |
| Scope | Does it explain multiple known phenomena, not just one? |
| Fruitfulness | Does it open up new questions or predictions worth investigating? |
| Conservatism | How well does it fit with what is already well established? |
A hypothesis that scores well on all five is a strong candidate for further investigation. One that fails testability is not a scientific hypothesis at all, regardless of how plausible it sounds.
The Working Hypothesis: Science in Progress
A working hypothesis is a hypothesis accepted for now to move research forward. A researcher uses it even knowing it may be incomplete or wrong. Scientists use working hypotheses when they need a direction to pursue, not a final answer.
Working hypotheses are frequently discarded. That is not a failure; it is the process. A working hypothesis that turns out to be false can still point researchers toward the right questions, rule out dead ends, and generate data that supports a better explanation later.
In qualitative research, working hypotheses act as an organizing framework. They help researchers decide what to observe and record before a full theory is available.
Null and Alternative Hypotheses in Statistics
When researchers study a possible relationship between two things, such as whether a drug lowers blood pressure, they frame the question using two paired hypotheses:
- The null hypothesis (H0): There is no effect, no relationship, or no difference. It is the default assumption the test tries to challenge.
- The alternative hypothesis (H1): There is some effect or relationship. This is usually what the researcher suspects or hopes to show.
The test does not prove the alternative hypothesis true. It assesses whether the data are inconsistent enough with the null hypothesis to justify rejecting it. If the probability of seeing results this extreme by chance (the p-value) falls below a pre-set threshold, such as 0.05, the null hypothesis is rejected.
One important detail: the significance threshold must be set before collecting data, not after. Choosing the threshold once you have seen the results makes the test invalid.
The alternative hypothesis can be one-sided or two-sided. One-sided means the effect goes in a specific direction, for example, the drug lowers blood pressure. Two-sided means there is some effect, but the direction is unknown. Which form you use depends on what the research question actually asks.
Where a Hypothesis Fits in the Scientific Method
Forming and testing hypotheses is one of the central activities that defines scientific inquiry. The Learning Concepts topic hub covers related foundational ideas. A hypothesis connects two phases of the scientific method. First comes systematic observation, what you notice. Then comes experimentation, what you do to check your explanation.
The process is not strictly linear. Researchers often revise a hypothesis mid-study when unexpected results raise new questions. A hypothesis that survives testing does not become proven; it becomes better supported, and it remains open to revision if new evidence appears. This provisional quality is a feature of science, not a weakness.
Self-Check: Can You Spot a Real Hypothesis?
Use these prompts to test your understanding before an exam or assignment:
- Is it testable? Can you describe an experiment or observation that would show the hypothesis is wrong? If not, it is not a scientific hypothesis.
- Does it make a specific prediction? A vague statement like “exercise affects health” is not a hypothesis. A specific one is: “Students who sleep at least eight hours before an exam will score higher than those who sleep fewer than six hours.”
- Is it based on prior observation? A hypothesis should follow from something you noticed or read, not appear from nowhere.
- Does it distinguish from a theory? Calling a well-established scientific framework “just a hypothesis” is a common mix-up. A theory has already been tested and confirmed many times; a hypothesis has not.
- In a statistics context: can you state both H0 and H1? If you cannot write out what “no effect” looks like, you have not fully formed the hypothesis pair.
If your proposed hypothesis fails the first two checks, revise it before moving to experimental design. A poorly formed hypothesis produces results that are hard to interpret, no matter how carefully the experiment is run.
Frequently asked questions
Can a hypothesis be proven true?
Not in an absolute sense. A hypothesis that survives repeated independent testing becomes better supported and may eventually contribute to a scientific theory. But science does not offer final proof; new evidence can always prompt a revision. Saying a hypothesis is "confirmed" or "supported" is more accurate than saying it is "proven."
What is the difference between a null hypothesis and a research hypothesis?
A research hypothesis states what the researcher expects to find, such as a drug reducing blood pressure. The null hypothesis is the opposite default: no effect or no difference. Statistical tests assess the data against the null hypothesis. If the null is rejected, the result supports the research hypothesis, but the two are not the same statement.
Does a hypothesis have to be written as an if-then statement?
The if-then format is not a universal requirement. What matters scientifically is that the hypothesis is testable, based on observation, and makes a specific prediction. The format can be helpful, but it does not define whether something counts as a hypothesis.
What happens when a hypothesis is rejected by experimental results?
A rejected hypothesis is still scientifically valuable. It rules out one explanation, narrows the field of possibilities, and often points toward a better hypothesis. Working hypotheses are frequently discarded. The research process can still lead to a more accurate explanation, even if the original hypothesis fails.
How is a hypothesis different from a theory in everyday vs. scientific use?
In everyday speech, both words suggest uncertainty or guesswork. In science, a theory is a well-tested explanatory framework supported by many independent lines of evidence. A hypothesis is a single testable proposal that has not yet been broadly confirmed. Calling a scientific theory "just a hypothesis" misrepresents how much evidence supports it.
Sources
- Scientific Method (Stanford Encyclopedia of Philosophy) · plato.stanford.edu
- Hypothesis - Wikipedia · en.wikipedia.org
- Statistical hypothesis test - Wikipedia · en.wikipedia.org
Prepared from these references and checked against them. References retrieved October 11, 2026.
Spotted an error? Tell us what needs correcting.