This research seeks to better understand unalerted driver response just prior to a serious vehicle crash. Few studies have been able to view a crash from the inside—with a camera focused on the driver and occupants. Four studies are examined: 1) a high-fidelity simulator study with an unalerted intersection incursion crash among 107 drivers; 2) four crashes from the Virginia Tech Transportation Institute (VTTI) 100 car study; 3) 58 crashes from vehicles equipped with an event triggered video recorder; and 4) a custom-designed high-fidelity simulator experiment that examined unalerted driver response to a head-on crash with a heavy truck. Analyses concentrate on decomposing driver perception, action, facial and postural changes with a focus on describing the neurophysiologic mechanisms designed to respond to danger. Results indicate that drivers involved in severe crashes generally have preview that an impact is about to occur. They respond first with vehicle control inputs (accelerator pedal release) along with facial state changes and withdrawal of the head back towards the head restraint. These responses frequently occur almost simultaneously, providing safety system designers with a number of reliable driver performance measures to monitor. Understanding such mechanisms may assist future advanced driver assistance systems (ADAS), advanced restraints, model development of advanced anthropomorphic test dummies (ATDs), injury prediction and the integration of active and passive safety systems.