Incidental learning allows us to extract statistical relations between events in our daily lives without the intention to learn them. Whereas anticipation during intentional associative learning has been linked to increased and decreased theta band activity, comparatively little research has focused on incidental learning. The study of such a pervasive mechanism of incidental learning faces the challenge of finding an appropriate paradigm. Similarly, while posterior alpha band activity has been shown to facilitate attention to a predictable target location, it is not clear whether alpha power could mediate attention given other predictive information; e.g., when the only available information provided by the cue is the likelihood of the target outcome. Here we used a stimulus-stimulus associative learning task to investigate whether a cue carries information on its contingent relationship with a target outcome, not only when their relationship is learned intentionally but also when it could be learned incidentally. Moreover, by presenting the target outcome in a visual search task, we were also able to study whether anticipatory attention can be modulated by the intentional or the incidental knowledge of the likelihood of a target outcome given a predictive (or non-predictive) cue. Participants were exposed to streams of cue-target outcome trials, where one of two possible cues and one of two possible outcomes were displayed. Intention to learn was manipulated by asking participants to assess whether one of the target outcomes (the intentional one) was more likely to appear following one of the cues (the intentional one). Any learning regarding the other cue-outcome relationship would be incidental. We found that frontal and temporal theta band activity were sensitive to the predictive value of a cue (predictive cues elicited lower theta power). Moreover, left temporal theta was sensitive to the intention to learn associations (theta activity elicited by intentional learning cues was higher). Alpha power, by contrast, was not modulated by cue predictiveness of the target outcome. These findings suggest that theta band activity carries information about the predictive value of a cue. The topographical differences between theta for intentional and incidental learning suggest distinct cortical networks activated depending on whether the relationship between a cue and an outcome has been learned intentionally or incidentally.