本研究旨在發展國民小學三年級以下學童適用之國語文成就測驗題庫，題庫預計功能有三:(1)鑑定學童目前國語文學習成就的水準;(2)篩選學習困難學童;(3)初步診斷學習困難學童國語文成就的剖面特徵。題庫包含聽力及書面兩測驗，聽力測驗又分為:(1)聲母區辨;(2)韻母區辨;(3)拼音;(4)語文記憶;及(5)文句理解等五個內容領域，共有題目84題。書面測驗則分為:(1)注音與國字;(2)部首與部件;(3)錯別字區辨;(4)語詞應用;(5)語詞分類與區辨;(6)文句組織與標點;(7)文句語意區辨;及(8)閱讀理解等內容領域，共有試題181題。題庫又依題目的難度分為三個層次，每個內容領域都涵括兩個難度層次，聽力測驗三個層次的題數依序為40，31及13題，書面測驗則分別有45、100及36題。聽力測驗中聲韻母區辨及拼音比較容易，而鑑別力最佳的是文句理解;內容領域分類架構解釋聽力測驗難度變異的比率為33%。在書面測驗的內容中以字形、字音及語詞應用較為簡單，而句子結構和字詞皆雷同的文句語意區辨最為困難;內容領域分類架構解書面測驗難度變異的比為26%。聽力與書面測驗結果的相關大致在.55左右(分年級依序為.47、.55及.60)。如果以預期答對概率.70為標準，常模樣本學有4~5%的學童未達層次一，這些學童在基本的字形、字音和語詞應用上都亟待補救教學的協助。本題庫對學童成就水準的鑑定應用而言，決定品質將會相當可靠，如以能力估計誤差轉化的信度係數來看，書面測驗三個決斷點上的信度都可達.95以上，聽力測驗也介於.85到.95之間。就常模樣本而言，聽力與閱讀能力與在校國語成績的相關分別為.46和.55。年級愈高相關愈強，一、三年級間差異在.10以上。就國語文能力發展趨勢而言，聽力測驗的年級差異較為明顯，年級可解釋能力變異的26%，而書面測驗的年級差異雖然顯著但解釋力只有2%，換言之，即令是國小三年級以下的學童，國語文能力已呈現非常巨幅的個別差異，部份時間分組教學的必要性值得相關單位的重視。

　　　　　 The purpose of this study is to develop a Chinese achievement item bank for 1-3 graders. The suggested uses of the item bank are screening of the low achievers, certification of students' achievement levels and the preliminary diagnosis of reading difficulties. Around three thousand and five hundred 1-3 graders were sampled as the norm for the item banking. All item parameters were scaled by IRT 3-parameter logistic model anchored item concurrent calibration design. The bank includes both listening and written tests. There are 84 items which were categorized into five content categories and three difficulty levels by the logical and empirical analysis for the listening test. There are 181 items for the written test. Eight content categories and three difficulty levels are derived. For the listening and written tests, 33% and 26% variance of item difficulty could be accounted by the content categories respectively. The grade can account for 26% and 2% of the ability variance for listening and written tests respectively. In other words, the grade differences of written tests were relatively small compared to the listening test. It also suggested that the individual differences of written test are very large within the same grade. The correlation coefficients between listening and written tests are .47, .55 and .60 for 1-3 grads respectively. The correlation coefficients between the abilities estimated of this bank and the school Chinese grades are around .45 for the listening test and .55 for the written test. If expected correct response rate of 70% was adopted as passing criteria, there will be around 5% students of the norm group failing to pass level I. These students will need special help in very basic Chinese learning processes such as the task of identification the appropriate uses of high frequency vocabularies or phrases. For the proposed applications, this item bank can function effectively to promote the decision quality. Most of the estimated errors will be within the commonly accepted levels (.2 to .3). computerized adaptive testing application will be the next step to look forward.