In this paper, we attempt to analyze currency crises within the decision theory framework. In this regard, we employ fuzzy system modeling with fuzzy C-means (FCM) clustering to develop perception based decision matrix. We try to build a prescriptive model in order to determine the best approximate reasoning schemas. We use the underlying behavior of the market participants during the crisis. With this analysis, we form the dictionary catalogs to construct a perception based payoff matrix. As an illustrative example, we used data from Turkish economy that covers two currency crises. The results show that market participants' dictionary catalogs based on perception knowledge extracted from the first crisis help participants to perceive the rise in market uncertainty. When the expectations are revised accordingly a speculative attack becomes inevitable. (C) 2008 Elsevier Inc. All rights reserved.