監視制度乃臺灣股市中最為重要的管制措施之一，該制度旨在針對異常的股票提 出警示，以防止不法炒作、保障投資大眾的權益。然而，不常的警示標準則會誤導投資者並 扭曲市場效率，亦會引發明牌效應。是故，本研究乃以持股行為觀點出發，利用可靠度理論 中的加速失敗時間模型為實證工具，藉以探討我國監視制度之穩定機制。實證結果發現：（ 1 ）警示次數低於兩次者，其監視成效較為顯著；警示次數達三次以上者則易產生明牌效應 ；（ 2 ）目前的警示標準對投機性較強的股票仍有未臻完善之處。

　　　　　Surveillance operation is the most important control policy in Taiwan stock market. The stock listed on Taiwan Security Exchange will be warned publicly by the operation if its price behaves anomalously, in order to protect investors and prohibit illegal manipulation. However, the inappropriate warning criteria will not only cause signal effects but also corrode the market efficiency. Hence, we intend to evaluate the performance of surveillance operation by means of comparing the stockholding behavior of normal and warning stocks, which experiment is proceeded by accelerated failure time model. We found that: (1) If the warning announcement lasts less than two days, the stockholding behavior will be improved significantly. Otherwise, it is easy to cause signal effects if the announcement lasts more than three days. (2) The warning criteria is still unappropriate for speculative stocks.