The paper reports further empirical evidence on seasonality in foreign exchange volatility using high-frequency data. Using a basis of the signal plus noise framework, the approach decomposes tick-by-tick Reuters FXFX quotes into a random walk and a stationary component, termed the efficient price and the pricing error, respectively. The efficient price is not directly observable and is an approximation of the 'true' value. The pricing error captures the deviation between the observed indicative quote and the efficient price. Under the proposed model, daily and intraday volatilities of the efficient price are estimated. A pronounced pattern of volatility is uncovered and appears related to the daily activity cycle of major organized stock exchanges. It is argued that seasonality in volatility is a symptom of foreign exchange markets. Results confirm Andersen and Bollerslev's findings that significant seasonal effects are one important determinant of overall volatility at high frequencies.