Using Bayes' formula, the parameters in a given model fitting data from perturbed angular correlation of y-rays (PAC) experiments are all integrated out, giving as the result the total proba-bility for the model. Experimental examples are given from PAC for models containing one and two nuclear quadrupole interactions, respectively. It is shown how the Bayesian formulation leads to a quantification of the probability of the correctness of the models. Furthermore, this method leads to a transparent and intuitively appealing criterion for model selection. Examples are given using PAC measurements on two proteins: liver alcohol dehydrogenase and azurin.