A hypothesis is your testable, educated guess about what will happen in an experiment and why. It’s not a random prediction. It’s a specific statement that connects a change you plan to make (your independent variable) to an outcome you expect to observe (your dependent variable), grounded in what you already know about the topic. If you’re staring at a lab report or science project wondering how to fill in this section, here’s how to build one that works.
What Makes a Hypothesis Scientific
A hypothesis is scientifically valid only if it can be proven wrong. This property, called falsifiability, is what separates a real hypothesis from a hunch or an opinion. If no possible result from your experiment could contradict your statement, it isn’t a hypothesis. “Plants need positive energy to grow” can’t be tested because there’s no way to measure “positive energy.” But “plants given 8 hours of sunlight daily will grow taller than plants given 4 hours” can absolutely be proven wrong by running the experiment and measuring the results.
Three criteria make a hypothesis usable in science:
- Testable: You can design an experiment or observation to check whether it’s true.
- Falsifiable: There is some possible evidence that would show it’s wrong.
- Specific: It identifies exact variables and a clear, measurable outcome.
The If-Then-Because Structure
The simplest way to write a strong hypothesis is to follow three parts: what you’ll change, what you expect to happen, and why you expect it. This is commonly written as an “if-then-because” statement.
If we change a specific independent variable, then we will observe a change in the dependent variable, because of what we already know from prior research or observation.
The independent variable is whatever you deliberately change or manipulate in your experiment. The dependent variable is the outcome you measure. For example, if you want to know whether vehicle exhaust affects childhood asthma rates, the concentration of exhaust is the independent variable and the incidence of asthma is the dependent variable.
Here’s what this looks like in practice: “If we increase the concentration of salt in water, then the boiling point will rise, because dissolved particles require more energy to transition to a gas phase.” The “because” clause is what elevates your hypothesis above a simple prediction. A prediction just estimates an end result (“I predict the water will boil slower”). A hypothesis explains the relationship between cause and effect.
Examples Across Different Fields
A good hypothesis looks slightly different depending on the discipline, but the structure stays the same. In a health study, you might write: “Increasing apple consumption in adults over 60 will result in fewer doctor’s visits.” In a workplace psychology study: “Employees who have flexible working hours will report greater job satisfaction than employees who work fixed hours.” In a business context: “Low-cost airlines are more likely to have delays than premium airlines.”
Notice that each of these identifies a specific group, a specific variable being changed or compared, and a specific measurable outcome. None of them are vague. None of them are impossible to test. And critically, each one could be wrong, which is exactly what makes them useful.
Null and Alternative Hypotheses
If your experiment involves collecting data and running statistics, you’ll likely need two formal hypotheses. The null hypothesis (written as H₀) states that there is no relationship between your variables, that nothing interesting is happening. The alternative hypothesis (H₁ or Hₐ) is what you actually think is true and what you’re trying to demonstrate.
For example, if you’re testing whether a new fertilizer increases tomato yield, your null hypothesis would be: “The fertilizer has no effect on tomato yield.” Your alternative hypothesis would be: “The fertilizer increases tomato yield.” The null always contains some form of equality (no difference, no effect), while the alternative uses directional language (greater than, less than, not equal to).
Your experiment doesn’t “prove” the alternative hypothesis. Instead, you collect data and determine whether there’s enough evidence to reject the null. If the data strongly contradicts the idea that nothing is happening, you reject the null and conclude that your alternative hypothesis is supported. If the data is inconclusive, you decline to reject the null. This framework exists because, as philosopher Karl Popper argued in 1934, evidence can disprove a scientific claim but can never fully prove one beyond all doubt.
How a Hypothesis Fits Into Bigger Ideas
People sometimes confuse hypotheses with theories or laws, but these are three distinct levels of scientific knowledge. A hypothesis is an untested prediction about a single experiment. A scientific law is a single, proven statement about how the universe behaves, verified across a wide variety of situations. Newton’s law of universal gravitation, for instance, is one equation that holds true everywhere we’ve tested it.
A scientific theory is much larger. It’s an entire framework of laws, principles, and facts united into a self-consistent system that accurately describes a whole field of study. The theory of evolution, for example, isn’t a guess. It’s a massive collection of verified evidence. Your hypothesis, by contrast, is the smallest unit of this process: one specific, testable claim about one specific experiment. If it survives testing and joins a growing body of evidence, it may eventually contribute to something much bigger.
Building Your Hypothesis Step by Step
Start with your research question. What are you actually trying to find out? Then do some background reading so your prediction is informed rather than random. This is where the two types of reasoning come in. Inductive reasoning means you’ve observed specific patterns (every time I water this plant more, it seems to grow faster) and you’re forming a general principle from those observations. Deductive reasoning works the other way: you start with a known principle (plants need water for cell growth) and apply it to predict a specific outcome in your experiment.
Both approaches are valid, and most good hypotheses use a combination. Your background research gives you the “because” part of your statement, and your specific experimental setup gives you the “if-then” part.
Once you have a draft, check it against three questions. Can I design an experiment to test this? Could the results prove me wrong? Have I clearly identified what I’m changing and what I’m measuring? If the answer to all three is yes, your hypothesis is ready. Not every inquiry needs a traditional hypothesis. Some research explores whether a relationship exists at all, or investigates historical or population-level questions where controlled experiments aren’t possible. In those cases, a research question alone is the appropriate starting point.