The purpose of this paper is to test the unbiased forward rate hypothesis (UFH) in the Won/Dollar foreign exchange markets during the period before and after the Korean financial crisis. We attempt to reconcile some conflicting results regarding the Won/D이 lar UFH by testing a diverse set of models such as the change rate model (CRM), the error correction model (ECM) , the GARCH-in-mean CRM, the GARCH-in-mean ECM, and the rolling sample regression model. The major results can be summarized five-f이 d. First, both the Won/Dollar spot and forward rates f이 low a non-stationary process with a unit root. Second, there exists a c이 ntegrating relationship between the Won/Dollar spot and forward markets, implying that two exchange rates share a common long-run swing, Third , the UFH is rejected under the formulation of the CRM and the ECM. Fourth, the test results are highly sensitive to the choice of sample period. It turned out that the Korean financial crisis at the end of 1997 made a significant contribution to a shift in the structural relationship between the Won/Dollar spot and forward markets. Fourth, the evidence from the GARCH-in-mean models indicates that there exists time-varying risk premium in the Won/Dollar exchange markets. Therefore, the existence of time-varying risk premium, especially significant after the crisis, can be regarded as one of major factors against the UFH in the Won/Dollar markets