Statistical significance describes a result that would be unlikely to appear if there were truly no effect or no difference in the population being studied. It is the standard decision rule researchers use to judge whether an observed pattern likely reflects more than chance.
The concept sits at the heart of hypothesis testing. A researcher starts with a null hypothesis, usually stating that there is no effect, and then asks how compatible the collected data are with that assumption. When the data would be very surprising under the null hypothesis, the result is called significant.
Importantly, significance is a statement about the data relative to a model. It is not a measure of how large, important, or true an effect is.
The p-value is the probability of obtaining a result at least as extreme as the one observed, assuming the null hypothesis is true. Smaller values indicate that the data are less compatible with the null hypothesis.
The significance level, written as alpha, is the threshold chosen before the analysis. In many fields it is set at 0.05. If the computed probability falls below alpha, the null hypothesis is rejected. Some disciplines, such as particle physics and genomics, use far stricter thresholds to limit false positives across many tests.
Setting alpha in advance matters. Choosing it after seeing the data undermines the logic of the test.
A statistically significant finding is not automatically meaningful in practice. Clinical significance asks whether an effect is large enough to matter to patients, practitioners, or decision-makers.
For example, a very large trial might show that a drug lowers blood pressure by a tiny amount, with a very small probability value. The result is real in a statistical sense, yet too small to change treatment. Reporting effect sizes alongside test results helps readers judge practical importance.
Significance testing is widely used and widely misunderstood. Frequent errors include:
Each of these misreadings can lead to overstated conclusions, and reviewers are increasingly alert to them.
Clear reporting lets readers evaluate results on their own terms. A strong results section states the test used, the exact probability value rather than only "p < 0.05", the effect size, and a confidence interval.
It should also explain how the sample size was determined and whether any corrections for multiple comparisons were applied. The same details are what readers look for when they critique a research paper.