The hyper-emotion theory attributes psychological illnesses to emotions of aberrant intensity, which in turn prompt better reasoning about their causes. Two experiments in which participants drew their own conclusions from syllogistic premises tested this prediction. Individuals from the same populations as the experimental participants rated the believability of likely conclusions. One experiment compared patients with depression with controls, and the other experiment compared students scoring high on anxiety with controls. Controls tended to draw believable conclusions and not to draw unbelievable conclusions, and this belief bias was greater for invalid inferences. The clinical groups were better reasoners than the controls, and did not show belief bias. As our hypothesis predicted, they drew many more valid conclusions concerning their illness than controls drew valid believable conclusions. But, contrary to the hypothesis, they refrained from drawing invalid conclusions about neutral topics more than controls refrained from drawing invalid unbelievable conclusions.