Temporally unpredictable stimuli influence murine and human behaviour, as previously demonstrated for sequences of simple sounds with regular or irregular onset. It is unknown whether this influence is mediated by an evaluation of the unpredictable sound sequences themselves, or by an interaction with task context. Here, we find that humans evaluate unrelated neutral pictures as more negative when these are presented together with a temporally unpredictable sound sequence, compared to a predictable sequence. The same is observed for evaluation of neutral, angry and fearful face photographs. Control experiments suggest this effect is specific to interspersed presentation of negative and neutral visual stimuli. Unpredictable sounds presented on their own were evaluated as more activating, but not more aversive, and were preferred over predictable sounds. When presented alone, these sound sequences also did not elicit tonic autonomic arousal or negative mood change. We discuss how these findings might account for previous data on the effects of unpredictable sounds, in humans and rodents.