The climatologist (like the hydrologist, the economist, the social scientist, and others) is frequently faces with situations in which a prediction must be made of the outcome of a process that is inherently probabilistic, and this inherent uncertainty is compounded by the expert's limited knowledge of the process itself. An example might be predicting next summer's mean temperature at a previously unmonitored location. This monograph deals with the balanced use of expert judgment and limited data in such situations. How does the expert quantify his or her judgment? When data are plentiful they can tell a complete story, but how does one alter prior judgment in the light of a few observations, and integrate that information into a consistent and knowledgeable prediction? Bayes theorem provides a straightforward rule for modifying a previously held belief in the light of new data. Bayesian methods are valuable and practical. This monograph is intended to introduce some concepts of statistical inference and prediction that are not generally treated in the traditional college course in statistics, and have not seen their way into the technical literature generally available to the practising climatologist. Even today, where Bayesian methods are presented the practical aspects of their application are seldom emphasized. Using examples drawn from climatology and meteorology covering probabilistic processes ranging from Bernoulli to normal to autoregression, methods for quantifying beliefs as concise probability statements are described, and the implications of new data on beliefs and of beliefs on predictions are developed.istical inference and prediction that are not generally treated in the traditional college course in statistics, and have not seen their way into the technical literature generally available to the practising climatologist. Even today, where Bayesian methods are presented the practical aspects of their application are seldom emphasized. Using examples drawn from climatology and meteorology covering probabilistic processes ranging from Bernoulli to normal to autoregression, methods for quantifying beliefs as concise probability statements are described, and the implications of new data on beliefs and of beliefs on predictions are developed.