The EPIC computational cognitive architecture was among the first to propose representing motor movement constraints explicitly in the form of motor processors that implemented a specified time course for the preparation, initiation, and production of movements. A key feature of this proposal was that movements were specified in terms of features, and movement preparation time was linear with the number of features that had to be prepared before a movement was initiated. While successful in modeling many high-speed tasks involving choice reaction times with keypress responses, serious difficulties appeared in modeling high-speed visual search tasks involving saccades and mouse movements. A reappraisal of the basis for EPIC's assumptions requires a critical change: visually aimed manual and ocular movements require no feature preparation time.