Individuals high in depressive symptom severity show probability bias: they believe negative events are relatively probable, and positive events relatively improbable, compared to those with less severe symptoms. However, this has only ever been demonstrated using self-report measures, in which participants explicitly estimate events’ probabilities: this leaves open the risk that “probability bias” is merely an artefact of response bias. We tested the veracity of probability bias using an indirect behavioural measure, based on a sentence-reading task. Study 1 tested 112 (...) Turkish students; Study 2 tested 117 international users of online groups for people with depressive and anxiety disorders. As predicted, participants with higher depressive symptom scores responded relatively quickly to sentences stating negative events might occur, and relatively slowly to sentences stating positive events might occur, compared to those with lower scores. This effect was only marginal in Study 1, but reached significance in Study 2. However, contrary to predictions, this effect was not moderated by the probability level stated in the sentence. This makes our findings difficult to interpret, and we must present these studies as a failure to convincingly demonstrate depression-related probability bias. We hope this stimulates more work on the nature and veracity of probability bias. (shrink)