Recent research suggests that there are many favourable features of the asset- pricing model of exchange rates incorporating Taylor rules. Against this back- ground, this thesis focuses on the relationship between the exchange rate and Taylor rule fundamentals. The introductory chapter provides a short summary of the most relevant literature, and explains the connections between the main chapters. In chapter 2, we mainly follow Engel and West's (2006) framework of the asset-pricing model of exchange rate incorporating Taylor rules to forecast the yen/dollar exchange rate. The central research question is whether this type of model has any predictive power with respect to the exchange rate. In chapter 3, a more detailed analysis of the properties of Taylor rules is un- dertaken. The main idea derives from one of the assumptions made in chapter 2, concerning the structural stability of the Taylor rules. If there are unknown struc- tural breaks, the estimation of the Taylor rule is likely to be biased. Furthermore, both theoretical and empirical studies suggest that the Taylor rule in advanced economies is asymmetric. If a central bank is minimizing an asymmetric loss function in which negative and positive in ation- and output-gap deviation are, respectively, assigned di�erent weights, then a nonlinear Taylor rule is optimal. Hence we set out to identify any structural breaks in the Taylor rule, and to uncover the extent to which nonlinearity plays a role in Taylor rule modelling. In our empirical study, a threshold model introduced by Caner and Hansen (2004) is used to measure whether the Taylor rules are nonlinear or not, in order to explain the existence of asymmetry of Taylor rules. Chapter 4 compares the performance of the traditional monetary model and the Taylor rule model in terms of out-of-sample forecasting performance. A key study is by Molodtsova and Papell (2009) who derive a simple version of the Taylor rule model and demonstrate that it can outperform a variety of monetary models as well as the naive random walk, on the basis of the state-of-the-art goodness-of-�t statistic developed by Clark and West (2006) (the CW statistic). It is of considerable interest to discover whether Molodtsova and Papell's (2009) results are driven by the superior predictability of the Taylor rule fundamentals, or by features of the CW statistic. To address this question, the sterling/dollar exchange rate for the period 1975-2010 is investigated. A detailed analysis of the CW statistic, including Monte-Carlo simulations, is conducted. In addition, a variety of estimators are used, including the Vector Error Correction Method (VECM) which is used to generate the out-of-sample forecast. Also, a number of goodness-of-�t measures (in addition to CW) are used for comparing the pre- dictability of the Taylor rule model with traditional monetary models. The overall �nding is that the out-of-sample forecasting predictability of the sterling/dollar exchange rate obtained by the Taylor rule model is not as signi�cant as we ex- pect by using a variety of goodness-of-�t measures, but the traditional monetary models have certain predictive power if VECM is applied.