I test the forward rate unbiasedness hypothesis using the Error Correction Model (ECM) of Naka and Whitney (1995). It is shown that the Naka-Whitney treatment of the dynamic is perhaps necessary to ameliorate existing problems associated with testing the hypothesis. However, it is not sufficient due to the sensitivity of the non-linear regression to starting values. Monthly data from October 1985 to May 1998 for New Zealand, Canada, United Kingdom (UK), Sweden, Germany, Japan and South Africa are used to test the forward rate unbiasedness hypothesis. The first four countries are inflation targeting regimes. Germany and Japan have both had very low inflation and South Africa experienced periods of very high inflation followed by periods of low inflation. The null hypothesis is widely rejected. The premia puzzle remains largely unexplained. Interestingly, the forward rate unbiasedness hypothesis holds in two inflation-targeting regimes namely New Zealand and UK. Implications of the Consumption - Asset Pricing Model (CAPM) are used to explain the finding.