本研究參考英國、澳洲等先進國家的作法，錄影蒐集道路實際危險情境影片，並由專業教練協助進行危險感知測試評估，藉以確認主要風險與評分標準。再針對駕訓班學員進行實驗測試，實驗結果發現，學員經過安全駕駛教育訓練後，其危險感知測驗成績顯著提升，並非常接近專業教練的成績。另外，本研究採用試題反應理論(Item Response Theory, IRT)中之Rasch模式，分析危險感知測驗試題之信效度與難度，經測試所有題目的適配度結果都落在95%的信賴區間內，separation reliability值都在0.9以上符合Rasch模式；而難度較高的試題類型則是同時有多個風險、風險時間出現過短或太長、路邊違規停車風險或路權認知差異等，可作為危險感知試題組合之參考。

This study referrs to the practices of advanced countries such as the UK and Australia, recorded and collected video clips of real road conditions with hazard, and professional driving instructors’ evaluation of hazard perception test results, in order to confirm the main risks and scoring standard. The students of driving school were asked to take an experimental test. Experimental test results showed that after the students had received safe driving education and training, the scores of their hazard perception tests had apparently increased, and were very close to the results of professional driving instructors. Besides, the study uses Rasch model of Item Response Theory (IRT) to analyze the reliability, validity and difficulty of the questions of hazard perception tests. After the test, the results of goodness of fit of all the test questions fell in 95% of the reliable zone. The values of separation reliability were all above 0.9, conforming to the standard of Rasch model. As to the test questions with higher difficulty, they all have multiple risks, too short or too long period of risk, risk of illegal roadside parking, or difference in drivers’ cognition of the right of way. All of these can be a reference for creating a combination of questions in the hazard perception test.