動態複雜問題實為一重要而困難的課題，然而，直接關切於動態複雜案例的處方性研究不但不多，以往研究更疏於採系統性、客觀性的實驗室實驗法。本研究因而在利於學習的實驗條件下，操弄心智模式系統邊界與動態行為歸因二因子，期望能夠帶來深層學習與較佳績效。結果顯示，實驗操弄因子對績效有著正面的影響，但其效果約僅發生於前2~3局，然而，整體的績效卻隨著練習仍不斷的有著顯著改進，而且，僅有一位受測者顯示出深入的學習。回頭檢驗實驗設計之假設前提一受測者採用心智模式模擬策略，發現在大約自第三局起，大部份受測者採認知資源需求較低的前饋控制策略，這或使得實驗操弄約自第三局起失去作用，是以決策績效雖仍繼續改進，但受測者卻沒有深入的學習。最後，本研究概化了此「績效與學習不對稱」現象，並討論了“為何”與“如何”克服此不對稱問題的後續研究方向。

Studies conducted in recent years have shown that humanbeings are poor problem solver in dynamic complexity task. However, little prescriptive research was done by experimental method. This study examines the feasibility of improving performance and learning in dynamic complexity task by providing the learning aids of “system boundary of mental model” and “attribution of system’s dynamic behavior”. This paper reports a laboratory experiment in which subjects managed a set of reversible trials with a computerized Beer Game task, where the conditions for learning are excellent. Results indicate that the manipulations only worked in the 1st and 2nd trials. However, subjects’ performance was still significantly improved through practice, but little deeper learning was detected in final. After carefully examining the cognitive strategies used by subjects, it is found that the effects of manipulation were eliminated after 3rd trial, wherein the mental model simulations strategy was replaced by the feedforward control strategy. The feedforward control strategy then caused the dissociation between performance and learning. Phenomenon of the dissociation is generalized. Methods to overcoming the dissociation are discussed.