Multistage inference consists of a series of single-stage inferences where the output of each previous stage becomes the input to the next stage. In a single-stage inference men reason from data or unambiguously observed evidence to a set of hypotheses. Multistage inference starts with the same unambiguous data or evidence in the first stage; however, the input for the next stage is the output of the previous stage. The next stage of inference is therefore based on the probabilities of events, rather than upon definite knowledge that a particular event is true (Gettys & Willke, 1969).

For example, suppose you wanted to predict the success or failure of a large garden party. Assume that the party is less likely to be successful if it is crowded indoors because of rain. Your datum is the presence of a dark cloud on the horizon. The first stage of inference would relate the dark cloud to the presence or absence of rain during the party. Suppose you estimated that the probability of rain was .70. This estimate would become the input to the next stage of inference. If you knew with certainty that it would rain, then you could infer the probability that the party would be a success. But you are not entirely sure that it will rain; the data that you have indicates rain with a probability of .70, so how should you proceed?