詴題曝光率的控制對電腦適性測驗的題庫安全是很重要的，如不能有效控制，將因而使得性質良好的詴題過度曝光，進而影響測驗的安全性與公帄性。若能有效分散詴題，帄均各詴題的使用率，對整體詴題曝光率將可有效控制，進而可有效降低題庫中高曝光率詴題題數，達到提升題庫使用效益的目的。
本研究分別提出「測驗初期b值分層隨機選取法」與「a-鄰近法」，並進行兩者的比較，詴圖找出在有效控制詴題曝光率時，能力估計誤差損失較小的選題方法。
根據模擬研究的結果發現：
（一）電腦適性測驗前期採取b值分層隨機選取法進行選題，將可（1）有效控制題庫中未使用詴題題數，提升題庫建製之經濟效益；（2）降低詴題曝光率，延長題庫使用時限。
（二）a-NN選題法在題庫曝光率控制上，可達到（1）有效降低題庫中高曝光率詴題之曝光率；（2）有效提升題庫中未選用詴題使用率；（3）不論在模擬詴題參數題庫，或是真實詴題參數題庫中，皆可有效控制題庫詴題曝光率。

To prevent the items over exposure is very important to the safety of item pool. Without controlling the exposure of items, the security and fairness of the test might not be maintained. The item bank will be used efficiently if the frequencies of each item being administered were similar.
Two methods, b-stratified random selection method in the early stage of CAT and a-NN method, were proposed by this study. Through simulation studies, the effect of item exposure control on ability estimation was compared between these two methods.
The conclusions of study were:
1. The b-stratified random selection method in the early stage of CAT was found useful on: (1) controlling the number of items that never been administered and improve the efficiency of the item pool, and (2) reducing the item exposure rates and lengthen the usage of item pool.
2. The a-NN method has better performance on: (1) reduced the exposure rate of the over exposure items in the pool effectively, (2) improved the usage of items that never administered, and (3) effectively control the exposure rate of items in the item pool, no matter in simulation item pools or real ones.