ABSTRACT
Text problems in research about judgment are notorious for being ambiguous in ways that can lead to task misunderstanding (e.g., Evans, 1972). In problems such as the maternity-ward task (see chapter 1), a statistical principle must be translated into natural language. Whereas the formal languages of mathematics and probability theory are unambiguous, translations into natural language of propositions they express can be ambiguous and can be even internally contradictory (see chapter 2). This study explored whether solution rates could be increased by making the information contained in sample-size tasks much less ambiguous. In one condition, participants received text versions of tasks, and in another they received explanations and simulations from which they extracted all of the information essential for solving the task (e.g., number of samples, size of samples, randomness of drawing procedure, method of aggregating sample means). These manipulations are consistent with the pragmatic-implications approach, but they might not be considered fullfledged training because there was no explicit instruction, only a clarification of the information. Moreover, generalization of a possible training effect to tasks not used in the demonstrations was not examined. The clarification procedure will be referred to as disambiguation training in following sections of this chapter.
