A null hypothesis is a default statement that there is no effect, no difference, or no relationship between the things you’re studying. It’s the starting assumption in statistical testing, and the entire point of a study is to determine whether the evidence is strong enough to reject it. In notation, it’s written as H₀.
The Core Idea Behind H₀
Think of a null hypothesis as a “nothing is happening” claim. If you’re testing whether a new drug lowers blood pressure, the null hypothesis says the drug has no effect and any difference you observe between the treatment group and the control group happened by chance. If you’re studying whether height and shoe size are related, the null hypothesis says there’s no relationship between them.
The word “null” literally means “nothing.” The null hypothesis assumes the status quo: no change, no difference, no connection. It’s not what the researcher believes or hopes to find. It’s the position that has to be disproven with data before any other conclusion can be drawn.
Null vs. Alternative Hypothesis
Every null hypothesis has a partner called the alternative hypothesis, written as Hₐ or H₁. These two work as a complementary pair, each stating that the other is wrong. The alternative hypothesis is what the researcher actually expects to find, based on prior knowledge or a theoretical prediction. It’s the “claim” being tested.
Here’s a concrete example. Say you want to know whether factory workers’ salaries differ by gender:
- Null hypothesis (H₀): There is no difference in salary between male and female factory workers.
- Alternative hypothesis (Hₐ): Male factory workers have a higher salary than female factory workers.
Or if you’re studying a biological relationship:
- Null hypothesis (H₀): There is no relationship between height and shoe size.
- Alternative hypothesis (Hₐ): There is a positive relationship between height and shoe size.
The null hypothesis is always the default. The alternative hypothesis is what you conclude only if the data provide strong enough evidence against H₀.
How Researchers Test It
Testing a null hypothesis comes down to a single question: if H₀ were true and nothing were really going on, how likely would it be to see results as extreme as the ones you got? That likelihood is the p-value.
Before running a study, researchers set a threshold called the significance level (written as α). This is the cutoff for how unlikely the results need to be before you reject the null hypothesis. The most common threshold is 0.05, meaning there’s a 5% or lower probability that the observed results happened by chance alone. Some fields use stricter thresholds like 0.01.
If the p-value falls below the significance level, you reject the null hypothesis and conclude in favor of the alternative. For example, if you set your significance level at 0.05 and your calculated p-value is 0.002, you reject H₀. But if that same p-value were 0.03 and you had set a stricter threshold of 0.01, you would not reject it. The threshold matters, and it’s chosen before the data are collected, not after.
Why You “Fail to Reject” Instead of “Accept”
You’ll notice statisticians never say they “accepted the null hypothesis.” The correct phrasing is “failed to reject” it, and the distinction is important. You can’t prove a negative. A lack of evidence that something exists is not the same as proof that it doesn’t exist. Your study may simply have missed a real effect because the sample was too small, the measurements weren’t precise enough, or the effect itself was subtle.
A useful analogy is a courtroom verdict. A jury doesn’t declare a defendant “innocent.” They declare “not guilty,” meaning the evidence wasn’t strong enough to convict. In the same way, failing to reject the null hypothesis means your sample didn’t provide enough evidence to conclude that an effect exists. It does not prove the effect is absent.
Type I and Type II Errors
Because statistical testing deals in probabilities rather than certainties, two kinds of mistakes are possible.
A Type I error (false positive) happens when you reject the null hypothesis even though it’s actually true. You conclude there’s a difference or effect when there isn’t one. The probability of making this mistake equals your significance level, α. So with a threshold of 0.05, you accept a 5% chance of a false positive.
A Type II error (false negative) happens when you fail to reject the null hypothesis even though it’s actually false. A real effect exists, but your study missed it. The probability of this error is called β. Reducing it requires either a larger sample size or studying an effect that’s large enough to detect reliably.
Statistical Power and Sample Size
The ability of a study to correctly reject a false null hypothesis is called statistical power, calculated as 1 minus β. Researchers generally aim for a power of 0.80, meaning an 80% chance of detecting a real effect if one exists.
Two things heavily influence power: sample size and effect size. Effect size is how large the real difference or relationship is. When the effect is large, even small samples can detect it. In one analysis, just 8 samples were sufficient to reach adequate power when the effect size was 2.5. But when the effect size dropped to 0.2, even 30 samples weren’t enough. This is why studies that find “no significant difference” don’t automatically mean there is no difference. They may simply have been too small to detect a subtle one.
How to Write a Null Hypothesis
Formulating a good null hypothesis starts with a clear research question. If your question is “Does exercise improve sleep quality?”, your null hypothesis states that exercise has no effect on sleep quality. The alternative states that it does. A few principles make it stronger:
- Be specific about the outcome. “No difference in total hours of sleep” is more testable than “no effect on sleep.”
- State it in terms of equality. The null hypothesis always posits that two groups are equal, or that a correlation is zero. In notation, that looks like μ₁ = μ₂ (population means are equal) or ρ = 0 (no correlation).
- Make it falsifiable. You need to be able to collect data that could, in principle, contradict it.
In published research, the null and alternative hypotheses typically appear at the end of the introduction, after the study background has made clear why the question matters.
Limitations of Null Hypothesis Testing
The null hypothesis framework has been the backbone of scientific research since Ronald Fisher developed the theory behind p-values in the 1920s, with Jerzy Neyman and Egon Pearson later formalizing hypothesis testing as a method. But it has well-documented limitations that have drawn increasing scrutiny.
In 2016, the American Statistical Association took the unusual step of releasing a formal statement on p-values, the first of its kind from the organization. It outlined six principles, several of which directly address common misunderstandings. A p-value does not measure the probability that the null hypothesis is true. It does not measure the size or importance of an effect. And scientific conclusions should not rest on whether a p-value crosses the 0.05 threshold alone. As the ASA’s executive director put it, the p-value “was never intended to be a substitute for scientific reasoning.”
The statement encouraged researchers to supplement or replace p-values with other approaches when appropriate: confidence intervals, Bayesian methods, likelihood ratios, and measures that estimate the size of an effect rather than simply flagging whether it clears an arbitrary bar. The null hypothesis remains a foundational concept in statistics, but understanding its limits is as important as understanding how it works.