The elicitation of subjective probability as a means of gaining access to an individual's internal state of uncer tainty is a procedure commonly used in decision-aiding techniques (see Wright, 1984). By this means expert judg ment may be tapped. Experimental psychologists have also used this method for exploring the characteristics of human judgment. Judgments expressed as subjective probabilities must conform to the probability laws. However, Tversky and Kahneman (1983) andWright andWhalley (1983) have shown that actual probabilistic judgments of future event likelihood frequently contradict axioms derived from proba bility theory. In the former study, this contradiction was shown even in judgments given by professional statisti cians. The program described here takes advantage of some axioms of probability theory to enable the user to make coherent judgments of the probability of future events. As incoherent (i.e., axiom-breaking) forecast probabilities cannot be accurate, the program interacts with the user to resolve inconsistencies in a manner that meets with the user's approval. The program FORECAST aids forecasting of (1) a time period or date when a specified event may happen (e.g., the United States dollar reaches parity with the pound ster ling next month); (2) the possible outcomes of an event, expressed in numerical terms as outcomes on a single con tinuous scale (e.g., company profits); and (3) possible out comes of an event, expressed as discrete or discontinuous outcomes (e. g., the next prime minister ofGreat Britain). The user selects from a menu the type of forecast he or she wishes to make, and the appropriate procedure is ini tiated. Time-Period Forecasting. For time-period forecast ing, the program elicits from the user the beginning and end of the time period of interest (time period B). Next the program elicits a point-probability estimate and an odds estimate for the event happening at least once in the forecast period (e.g., the pound sterling falling to a value ofless than $1.(0). These formally equivalent estimates are compared by the program, and inconsistencies are re ported to the user for resolution to give P(B). Thus, theuser is made aware oferror at the level of the mode of elicita tion. Next, the program attempts to extend the time period of the forecast by a subsequent time interval consisting of the same number of days as the main time period of