It is often assumed that when a neutral cue is presented in a spatial cueing task, attention remains at fixation until target onset. We hypothesized that variance in nonspatial attention and switches of attention toward target locations can account for variance in reaction times of neutral trials. Lateralized event-related potentials (ERPs) and changes in electroencephalogram (EEG) frequency bands served as predictor variables in a single-trial logistic regression analysis to predict the direction of spatial attention in cued and neutral trials. The contingent negative variation (CNV) and non-lateralized changes in the alpha band served as markers of nonspatial attention. The direction of attention in cued trials was reliably predicted from single-trial lateralized ERP components. In neutral trials, only evidence for nonspatial attention was found, indicated by increases in the CNV and decreases in alpha preceding targets to which responses were relatively fast.