The primary electrophysiological marker of feature-based selection is the N2pc, a lateralized posterior negativity emerging around 180-200 ms. As it relies on hemispheric differences, its spatial resolution is severely limited. Here we demonstrate that multivariate analyses of raw EEG data provide a much more fine-grained spatial profile of feature-based target selection, while retaining the temporal resolution of the N2pc. When training a pattern classifier to determine target position from EEG, we were able to decode target positions on the vertical midline, which cannot be achieved using standard N2pc methodology. Next, we used a forward encoding model to construct a cortical tuning function that describes the continuous relationship between target position and multivariate EEG in an eight-position display. This model has a high spatial resolution and is fully invertible, enabling us to construct hypothetical topographic activation maps for target positions that were never used. When subsequently tested against the real pattern of neural activity obtained from a different group of subjects, the constructed maps from the forward model turned out statistically indistinguishable, thus providing independent validation of our model. Our findings demonstrate the power of multivariate EEG analysis to track feature-based target selection with high spatial and temporal precision.