The relationship between preflight risk assessment by prey and the escape behaviors they perform while fleeing from predators is relatively unexplored. To examine this relationship, a human observer approached groups of Columbian black-tailed deer (Odocoileus hemionus columbianus), varying his behavior to simulate more or less threatening behavior. We measured the focal deer's angle of escape, distance moved during flight, duration of trotting and stotting behavior, and change in elevation during flight. Analyses revealed positive relationships between the distance moved during flight and the distance at which they fled. When flight was initiated when the approacher was close, deer fled relatively shorter distances and took flight paths at more acute angles, a property that would force a real predator to change direction suddenly. Our results indicate that deer do not compensate for allowing the observer to approach more closely by fleeing greater distances. Rather, distance moved and flight initiation distance are linked by level of reactivity and habituation: more reactive or less habituated deer both flee at a greater distance and move away to a greater distance during flight. More threatening behavior by the approacher led to longer durations of rapid flight behavior (e.g., trotting and stotting), and deer tended to flee uphill and into taller vegetation, using these landscape features as refuge from danger. Finally, we provide the first evidence for Pitcher's untested “antiambush” hypothesis for the function of stotting and discuss its significance. In general, both preflight predator behavior and habitat features influence both duration and direction of escape.