The paper reviews methods which have been proposed for solving global optimization problems in the framework of the Bayesian paradigm. Three main approaches are singled out. In the first approach, called the Random Function Approach, methods are based on the idea of introducing a probabilistic model for the objective function in the form of a random function. The second class of methods derives from setting up a probabilistic structure for the number of different extrema of the function and the size of the related region of attraction. Finally, the third approach considers a stochastic model of the distribution function of the extremum values sampled by the so-called Multistart Method.