The paper discusses the various theoretical models for exchange rate as developed in the literature, translates them into testable forms, verifies them through empirical work on five important exchange rates, and infer useful conclusions for enhancing the understanding of the exchange rate theory. The sticky price monetary model, incorporating the current account balance differential variable, is found to be the most appropriate model for exchange rate determination. Thus, the exchange rate is found to vary directly with money supply differential and expected inflation differential, and inversely with real income differential, current account balance differential and interest rate differential. These variables together explain 91 to 97% of the variation in different exchange rates. The coefficients of the causal variables are quite often insignificant and unstable, and hence not much can be inferred about exchange rate elasticities. Accordingly, the exchange rate models at their present state of knowledge are not quite adequate for a proper management of the exchange rate. Furthermore, the time series models are found to perform better than the structural models for forecasting.