Early hazard detection requires appropriate prediction. We propose a new method of driver education using a tablet device （iPad） and movies of accidents recorded by drive recorders to instill appropriate prediction capability in drivers. We developed driver-training software that runs on an iPad. Movies of accidents embedded in this software were played and paused just before the collision. Drivers were asked to identify hazards requiring attention in the paused image by touching them with their finger within seven seconds. The software identifies touched hazards and records touch timing. After the hazards were identified, the movies restarted to show drivers whether there was a collision with the touched object. We made two similar scenarios, each having eight scenes. Drivers repeated the same scenario three times as the training phase. The test phase was conducted after the training phase. In the test phase, drivers viewed the other scenario. The t-test revealed that the collision object touch ratio of the test set was significantly higher than that of the first set in the training phase. Another t-test revealed that reaction times for touching the collision object of the test set were significantly faster than that of the first set in the training phase.
