Feedback ERN (fERN) and frontal midline theta have both been proposed to index a dopamine-like reinforcement learning signal in anterior cingulate cortex (ACC). We investigated these proposals by comparing fERN amplitude and theta power with respect to their sensitivities to outcome valence and probability in a previously collected EEG dataset. Bayesian model comparison revealed a dissociation between the two measures, with fERN amplitude mainly sensitive to valence and theta power mainly sensitive to probability. Further, fERN amplitude was highly correlated with the portion of theta power that is consistent in phase across trials (i.e., evoked theta power). These results suggest that although both measures provide valuable information about cognitive function of frontal midline cortex, fERN amplitude is specifically sensitive to dopamine reinforcement learning signals whereas theta power reflects the ACC response to unexpected events.