This paper examines three interwar exchange rate regimes: the free float of the early 1920s, the fixed rates of 1927-31 and the managed float of the early 1930s. Nominal rates were considerably more variable under free than under managed floating. The reduction in nominal exchange rate variability achieved with the move from free to managed floating was not accompanied by a commensurate fall in exchange rate uncertainty because government policy seems to have been subject to periodic shifts that heightened risk. There was a strong association between nominal and real exchange rate predictability in both the free float of 1922-6 and the managed float of 1932-6. There was no direct correspondence between the degree of exchange rate stability and the volume of international capital flows. Capital controls, which were considerably more prevalent under managed floating than either of the other regimes, provide a major part of the explanation for differences across regimes in the magnitude of real interest differentials.