You start with a number of predetermined options (hypotheses) and then use statistical techniques to determine which of them has more or less evidence for being true. There are forms of hypothesis testing used in Bayesian statistics, but the term is more usually associated with traditional statistics, and in particular with what is sometimes called null hypothesis significance testing or NHST. Here you start with a null hypothesis (usually what you would like to be false, such as your system not being any better) and have an alternative hypothesis (typically what you would like to be true, such as your system being better!). You then use a significance test to see if the data is consistent with the null hypothesis, and if not you 'reject' the null and instead conclude that the alternative is likely to be true.

Defined on page 61

Used on pages 8, 15, 57, 60, 61, 63, 69, 75, 76, 77, 80, 81, 82, 95, 99, 100, 101, 105, 142