Background. Research indicates that deluded patients ‘jump to conclusions’ on probabilistic reasoning tasks. Two experiments were carried out with patients suffering from persecutory delusions and depressed and normal controls in order to determine whether this response bias is affected by task difficulty and the meaningfulness of the materials.

Methods. Tasks were variants of those employed by Huq et al. (1988) and Garety et al. (1991). In Experiment 1, subjects judged which of two bags a sequence of coloured beads had been taken from. Difficulty was manipulated by varying the ratios of coloured beads in the bags. In Experiment 2, a neutral condition required judgements about coloured beads drawn whereas, in meaningful conditions, subjects had to judge whether personality characteristics described one of two individuals.

Results. In Experiment 1, estimates of certainty varied with task difficulty, and there was no evidence of ‘jumping to conclusions’ in the deluded group. In Experiment 2, all groups reached an initial level of certainty and reduced their estimates of certainty following disconfirmatory evidence more quickly in the meaningful conditions. Both clinical groups expressed higher certainty levels in early trials, and a greater magnitude of reduction in certainty following disconfirmatory information. These group differences were more evident in the meaningful conditions than in the neutral conditions.

Conclusions. Probabilistic reasoning is affected by task difficulty and meaningfulness of materials in both deluded and depressed subjects. Observed reasoning abnormalities were not specific to the deluded group.