In this paper it is argued that, although the rational expectations models of the exchange rates are able to explain some features of the exchange rate behavior, they fail to explain the long run cycles in the real exchange rates. An alternative modelling approach is proposed, using concepts of bounded rationality. It is shown that in such models it may be rational for agents to combine forward looking with backward looking rules (such as technical analysis). It is also argued that a "near-rational" expectations model can better explain the long-run drift in real exchange rates.