Optimal decision-making depends on accurate value and outcome representations. To date, scientists have used EEG and fMRI independently to either identify activation latencies or brain regions related to decision signals. In turn, a full spatiotemporal characterization of the process underlying simple value-based decisions and reward learning is still lacking. Here are a series of human multimodal neuroimaging studies in which I studied the separate influence of outcome valence and surprise on learning. Importantly, I show that linking fMRI brain activations with temporally specific EEG information can help us identify distributed neural representations of interest and uncover latent brain states that would likely have remained unobserved with more conventional (e.g., univariate) analysis tools.