Learning the predictive relationships between stimuli and rewards is critical for adaptive behaviour. Non-human animal studies suggest that reward learning is driven by dopaminergic midbrain responses which evolve in time from a salience signal to a valuation signal (Schultz, 2016). These two signals are believed to emerge from the same population of midbrain neurons and as such are mixed into an aggregate signal that is difficult to decouple, even at the level of single-neuron recordings. Here we exploit high-temporal precision electrophysiology (EEG) collected simultaneously with 7T functional magnetic resonance (fMRI) to directly test this two-component reward framework in humans. We collected preliminary data during a reward-anticipation task, where salience and value were orthogonally manipulated. We first exploited the high-temporal resolution of the EEG signal to decouple salience and value signals. Multivariate discriminant analysis of both standalone EEG (N=25) and EEG during MRI (N=10) data identified separate and partially overlapping signals associated with salience (100-300 ms) and value (200-600 ms), consistent with the two-component reward framework. In a next step we will use the endogenous, trial-by-trial variability in these EEG signatures, to build EEG-informed fMRI predictors and spatially decouple these two cascading signals in midbrain structures, using ultra-high field neuroimaging at 7T. In doing so, we aim to demonstrate the extent to which these cortical EEG signals of salience and value explain BOLD variability in the midbrain, thereby offering an account of the spatiotemporal dynamics of midbrain responses underlying reward learning.