Visual search is speeded when the target-defining property (a feature- or dimension difference relative to the distractors) is repeated relative to when it changes. It is thought that automatic and implicit intertrial priming mechanisms underlie this effect. However, intertrial priming has been found to be less robust in compound search tasks (in which the response property is unrelated to the target-defining property) than in present/absent search tasks (in which the response is directly related to the presence of a target-defining property). This study explored the hypothesis that intertrial priming is dependent on the level of ambiguity in a task, with the present/absent task being inherently more ambiguous than the compound search task. The first three of five experiments further established the dissociation between the tasks and excluded alternative explanations. Intertrial priming was strong in present/absent and go/no-go tasks, but absent in compound and compound/absent tasks. The last two experiments supported the ambiguity hypothesis by introducing more uncertainty in the compound task, after which intertrial priming returned. © 2006 Psychology Press Ltd.