Bayesian statistical inference is unfamiliar to many educational evaluators. While the classical model is useful in educational research, it is not as useful in evaluation because of the need to identify solutions to practical problems based on a wide spectrum of information. The reason Bayesian analysis is effective for decision making is that it defines probability as a measure of opinion or belief, rather than as long-term frequency. Defining probability as a measure of opinion or belief enables the Bayesian investigator to consider a wider range of information than is possible with the traditional model. Personal expertise, logical analysis, and soft data from a wide variety of sources serve to shape opinion about a state of nature, with experimental data providing additional information either for or against the prior opinion of the evaluator. In classical statistics, prior knowledge or opinion is ignored. However, when practical decisions must be made the Bayesian stresses that all knowledge should be brought to bear on the problem rather than just an isolated set of data. Because of the decision-making orientation of the evaluator, the Bayesian model should be considered as an alternative to classical inference. Since the Bayesian model views probability as a measure of opinion rather than as a long-term frequency, the statistical requirements for it are actually greater than for the classical statistician. Use of a wider range of distributions than with classical statistics demands more statistical skills than many evaluators currently possess. However, the questions raised by the Bayesian model are useful even if the model is not totally adopted. (Author/RC)