背景 : 自我導向學習是護理人員應具備之重要能力，然而很少有研究針對護理學生的自我導向學習能力建立常模。本研究採用試題反應理論模式有效建立台灣不同護理學制學生之自我導向學習能力常模。 目的 : 本研究之目的為：（一）運用試題反應理論之等級反應模式再次驗證本研究團隊以驗證性因素分析發展之自我導向學習量表；（二）建立台灣四個護理學制學生之自我導向學習能力常模。 方法 : 採分層隨機取樣方式合併等比率等機率抽樣原則，選取每個學制至少 15%學生參與本研究，合計參與之護理院校為 13所，護生人數為 7,879人。本研究使用之測量工具為鄭、郭、林及李（ 2010）發展之 20題自我導向學習量表（ self directed learning instrument, SDLI）。運用試題反應理論之等級反應模式中之二參數對數型模式（鑑別度及難易度），併以MULTILOG統計軟體進行資料分析，再以百分等級建立常模。 結果 : 從分析每一試題訊息之分數顯示， SDLI中有 18題呈現極高度之鑑別度， 2題呈現高度之鑑別度，分數越高，表示 SDLI在預測護生自我導向學習具有高鑑別度。 SDLI各向度及整體量表之信度介於.80∼ .94。從分析整體試題訊息功能之分數顯示，除了頂尖 2.5%之學生外， SDLI量表適用於台灣全部護理學生，且本研究建立之自我導向學習能力常模，亦適用於台灣四個護理學制之全部護生。 結論 : 試題反應理論是一種強有力、有效益的評鑑量表之方法。本研究建立之自我導向學習能力常模，可提供台灣護理教育者及護理學生做為評估自我導向學習能力之參考標準。

BACKGROUND: Little research has investigated the establishment of norms for nursing students' self-directed learning (SDL) ability, recognized as an important capability for professional nurses. An item response theory (IRT) approach was used to establish norms for SDL abilities valid for the different nursing programs in Taiwan. PURPOSES: The purposes of this study were (a) to use IRT with a graded response model to reexamine the SDL instrument, or the SDLI, originally developed by this research team using confirmatory factor analysis and (b) to establish SDL ability norms for the four different nursing education programs in Taiwan. METHODS: Stratified random sampling with probability proportional to size was used. A minimum of 15% of students from the four different nursing education degree programs across Taiwan was selected. A total of 7,879 nursing students from 13 schools were recruited. The research instrument was the 20-item SDLI developed by Cheng, Kuo, Lin, and Lee-Hsieh (2010). IRT with the graded response model was used with a two-parameter logistic model (discrimination and difficulty) for the data analysis, calculated using MULTILOG. Norms were established using percentile rank. RESULTS: Analysis of item information and test information functions revealed that 18 items exhibited very high discrimination and two items had high discrimination. The test information function was higher in this range of scores, indicating greater precision in the estimate of nursing student SDL. Reliability fell between .80 and .94 for each domain and the SDLI as a whole. The total information function shows that the SDLI is appropriate for all nursing students, except for the top 2.5%. SDL ability norms were established for each nursing education program and for the nation as a whole. CONCLUSIONS: IRT is shown to be a potent and useful methodology for scale evaluation. The norms for SDL established in this research will provide practical standards for nursing educators and students in Taiwan.