Hypotheses are tested by developing scenarios in which certain outcomes are consistent with the hypothesis being true or false, then by experimenting or researching to see whether the hypothesis matches reality. In statistics, hypothesis testing has a specialist meaning that entails making a probability assessment of assumptions before experimenting.
The classic example of a statistical hypothesis test is that of a coin toss. Coins have two sides, so the null hypothesis is that the coin is balanced and each side has a 50 percent chance of coming up in a toss. If the experiment is then run, and the results diverge significantly from the results predicted by the null hypothesis, then a second hypothesis must be developed to account for the unexpected outcome.
In this scenario, a quarter may be flipped 100 times to determine whether it truly is balanced. The null hypothesis is that the population of outcomes should fall close to 50 heads and 50 tails if the coin is balanced. Any significant divergence from the predicted 50-50 split is likely due to some factor such as an unbalanced coin. If the test results in 80 heads and 20 tails, the null hypothesis is said to have been tested and rejected.