Posterior Probability Matching and Human Perceptual Decision Making

摘要: Probability matching is a classic theory of decision making that was first developed in models cognition. Posterior probability matching, variant which observers match their response probabilities to the posterior each being correct, used increasingly often perception. However, little known about whether consistent with vast literature on vision and hearing has within signal detection theory. Here we test using two tools from First, examine models' performance two-pass experiment, where block trials presented twice, measure proportion times model gives same twice repeated stimuli. We show at low levels, give highly inconsistent responses across presentations identical trials. find practised human are more than these predict, some evidence less as well. Second, compare discrimination task theoretical ideal observer achieves best possible performance. very inefficient low-to-moderate can be efficient ever according models. These findings support models, rule out broad class for expert perceptual tasks range complexity contrast symmetry detection. our leave open possibility inexperienced may behaviour, methods provide new testing such strategy.