Effects of normalised and natural frequencies on responses to Bayesian reasoning tasks

Two experiments examined the effect of frequency formats on Bayesian reasoning. In each experiment a group of young (mean age=20) and old (mean age=70) participants completed Bayesian reasoning tasks. These tasks are known to elicit a number of typical response groups as a result of participants failing to integrate all of the cues within the tasks (resulting in, for instance, ‘base rate neglect’). In the first experiment, half of each age group completed tasks in a probability format, and half in the normalised frequency format. In the second, half completed the probability format while half completed a natural frequency format. Neither normalised nor natural frequencies facilitated normative reasoning, but natural frequencies did affect which cues within the task were attended to, and resulted in lower levels of over-estimation of likelihood values. There was no effect of age group. The theoretical implications of these findings in terms of dual process – and ‘multi-process’ – models of thinking and reasoning are discussed