This　paper　presents　results　of　an　exploratory　study　on　differences　in　driving　characteristics　between　normal　and　emergency　situations　and　builds　up　the　car-following　model　under　the　emergency　evacuation　situation.　The　simulation　scenario　has　been　given　to　create　a　driving　environment　under　the　emergency　evacuation　situations.　Questionnaire　investigations　and　the　electrocardiogram/heart　rate　monitor　are　used　to　verify　the　validity　of　the　driving　environment　from　both　subjective　and　objective　perspectives.　Perception　reaction　time　(PRT)　and　the　critical　headway　are　taken　as　two　indicators　to　describe　driving　characteristics.　The　results　show　that　PRT　is　in　accord　with　normal　distribution　under　both　normal　and　emergency　situations.　The　value　of　PRT　under　the　emergency　situation　is　lower　than　that　under　the　normal　situation.　The　results　also　show　that　critical　headway　under　the　emergency　situation　is　smaller　than　that　under　the　normal　situation.　A　back　propagation　(BP)　neural　network　is　designed　in　this　study.　A　combination　of　the　Levenberg-Marquardt　BP　algorithm　and　Bayesian　regularization　is　employed　to　train　the　network.　Gray-correlation　analysis　is　conducted　to　determine　which　factors　have　a　great　impact　on　the　acceleration　of　the　following　car.　Simulation　of　the　BP　neural　network　using　data　collected　from　driving　simulator　reveals　that　the　BP　neural　network　has　a　high　precision　in　the　prediction　of　the　car-following　model.　DOI:　10.1061/(ASCE)TE.1943-5436.0000434.　(C)　2012　American　Society　of　Civil　Engineers.