(a) Representative idealized time series from long presentation, long delay trials where the target appeared in the left visual hemifield. Example ROIs, selected for illustration, are depicted as colored patches on the anatomical underlay and include (from top to bottom) right FEF, caudate head, superior temporal gyrus, and Broadmann area 17. Each colored vertical bar overlay represents a time point that was used for constructing the set of canonical task-related brain state patterns representing VME (yellow), maintenance (red), and retrieval (blue) processes. (b) The set of task-related brain state patterns derived from the time points indicated above and sequentially orthogonalized. Each brain state consists of two components: mean (top row) and spatial (bottom row). Different linear combinations of mean and spatial components model differences in activity across the brain related to whether the target is on the left or right. Whole brain patterns of task evoked BOLD activity can be converted into time courses of brain state expression by projecting each whole brain volume onto the complete set of brain state patterns using linear regression and organizing the regression weights into a time series.