The probability that you would observe the data you did, assuming your hypothesis is true. This sounds backwards, but in practice it is easier to calculate the likelihood than directly calculating the posterior. The likelihood is used in both Bayesian and Frequentist data analysis.