A test is made for long-run purchasing power parity with Canadian monthly and quarterly data from the 1950s using a Bayesian unit-root test discussed by Koop, but first suggested by Zellner and Siow. Results show that the hypothesis that the real exchange rate is a stationary process receives relatively low posterior probability. Put differently, the probability is low that the nominal exchange rate and the national price levels have a stable long-run relationship. Furthermore, it is also found that there are structural differences between the sample period from October 1950 to May 1962 and the sample period from January 1952 to November 1960.