This thesis tests the hypothesis that the qualities of different forecasts of exchange rate volatility depend on the underlying exchange rate regime. By examining the British pound during and after its withdrawal from the European Monetary System (EMS), this analysis compares "backward-looking" historical forecasts of future volatility with the "forward-looking" forecast of volatility reflected in current option prices. Because option implied volatility contains the market's most current expectations about future prices, theory and much previous evidence suggests this should be the superior predictor of future volatility. In contrast to previous research by findings, this study concludes that option implied volatility is not superior. During the time when the pound was in the EMS, implied volatility provided reasonably good forecasts of future volatility. However, after the pound withdrew from the EMS, various statistical measures of historical volatility are found to have greater informational content and predictive power about future actual volatility than implied volatility. In particular, a time series estimate, specifically a GARCH(1,1) model, had the most informational content and predictive power about realized pound volatility, especially in the period following sterling's withdrawal from the EMS.