Specificity of the jump-to-conclusions bias in deluded patients.