The paper will examine the 1994 and 2001 Turkish currency crises by using early warning system which is based on the “signal” approach proposed by Kaminsky, Lizondo and Reinhart (KLR) (1998). The “signal” approach is a non-parametric approach. In this approach, the behavior of a number of individual variables is monitored and they are evaluated against a certain threshold levels. If any of these indicator exceeds its threshold, it is said that indicator issues a “signal” that a currency crisis may occur within a given period. The objectives of this paper are two folds: to investigate causes of currency crises under consideration and to compare similarities and differences of the 1994 and 2001 currency crises. The data consist of monthly data and range from January 1987 to November 2005 for the following variables: reserves, inflation rate, GDP growth, portfolio capital inflow to reserves, short term external debt to reserves, domestic debt, money supply to reserves, current account to GDP, real exchange rate overvaluation, regional stock market return, regional market pressure index, stock market index, export and import. Results showed that 2001 crisis is deeper and costlier than 1994 crisis, external factors play more imported role in 2001 crisis than 1994 crisis and in both crises Weighted Composite Index increases sharply previous the both crises.