In the literature on the empirical unconditional distribution of foreign exchange rate returns there is indication that the type of distribution function is related to the form of exchange rate regime. The analysis has been hampered by the nonnestedness of alternative distribution models. The paper investigates the issue by means of extremal analysis which allows for a unified treatment. In particular, we try to sort out whether apparent distributional differences are due to differences in techniques or in regimes. (JEL F31, F33).