高級中學學生由於在嚴格排序的聯招選擇後，區域間及校際間學生程度有了相當明顯差異。在題目參數連結料設計中，如果充分代表性樣本昂貴未取得，在不夠理想的現實運作中，有那些考慮是核心而不容規?那些條件若未能滿足影響尚可容忍?本研究主要針對考生能力分配的進中和分配趨勢作一摘要探討，同時兼顧學科難度特質，對高中題庫建立中連結設計提供具體考的資訊。 本研究隨機抽取八十年大學入學考試四個學科每科752~2000名考生，將考生分為低、中、高三組，假設題目和能力參數為已知進行模擬研究。連結方法為平均數標準差法和題目特徵曲線法兩種，採共同試題設計，以真值為評鑑標的，估算每組考生連結至真值時的偏誤和均方誤根，以評鑑考生特質對題目連結之影響。就四個科目的連結結果來看，重要的發現歸納如下: 一、 各科的連結結果，皆以「能力低、高組合」的連結結果最好。 二、 樣本考生能力分配平均值與理想值愈接近，連結的結果愈好;考生能力分配的標準差愈大連結結果愈好。 三、 就樣本考生能力分配之考慮而言，標準差之影響大於考生能力平均值。 就難度連結的均方誤根而言，平均數標準差法和特曲線轉換法效果頗為接近;就鑑別度而言，特徵曲線轉換法效果略佳。

The purpose of this study is to investigate the effects of examinee ability and test difficulty on the scaling of high school achievement item banks with Monte Carolo simulations. Based upon the characteristics of 1991 Biology, English, Physics and Math University Entrance Exams, four test banks were generate. There were three ability levels (low, medium and high) included for each linking design. One thousand simulees were generated for each ability level. The number of anchor items erer ranging from 10 to 45. the linking methods adopted were “mean and Sigma” and “Anchor Test Characteristic Curve”. The results suggested that the smaller mean difference between the sample and the target (original/ideal scale), the better the linking results would be. In terms of the variance of the sample ability distribution, it seemed that larger standard deviation produced better scale recovery. The effect of sample variations is larger than the sample means. In other words, for the linking design, the concern of sample heterogeneity is more important than the sample ability level. If the test could not discriminate the specific sample ecaminees well, item parameter estimation precision would be damaged. In turn, the linking results would be also affected. If computer equipment is available, method of “Anchor Test Characteristics Curve” transformation is suggested for better discrimination parameter recovery. The results indicated that most of the subjects in this study believe: 1. Both of the two models, the whole-class and the subjects-separated, have their own advantages as well as disadvantages. Which one is better in uncertain. 2. Most of the elementary school teachers are competent for their teaching job, therefore, there will have no problem to continue implementing the whole-class teaching in elementary schools. 3. Because about half of the elementary schools are small size, it will arise some difficulties in time scheduling for the subject-separated teaching. In order to match the teaching model of the elementary schools, the teacher colleges should change their organization and curriculum structure.