Testing for the existence of downward trends in real commodity prices has been the focus of several studies since the Prebisch-Singer hypothesis was formulated back in 1950. In this article, we focus on annual and monthly series of various commodity categories and consider alternative price deflators. Based on the methodology of . Harvey et al. (2010), which is robust to the order of integration of the time series, we conclude that the time frequency and the price deflators play a key role when testing for the Prebisch-Singer hypothesis. For instance, at an annual frequency (1900-2003, 1900-2008), it becomes considerably more likely to support it when deflating by the unadjusted US CPI-all items than when deflating by the Manufactures Unit Value (MUV) Index or the Historical Price Index of Manufactures (HPIM). This finding is in agreement with the . Svedberg and Tilton (2006) discussion on the CPI's overestimation of inflation and the measurement of the real price of copper. When dealing with monthly data (January 1957-December 2010), our results show that real prices tend not to reject the null hypothesis of a trendless series, except when deflating by the PPI-Crude Materials and, to a lesser extent, by the HPIM.We also explore the issue of using different currencies. Our estimation results, conducted for the US dollar and the Pound Sterling with annual data (1800-2005; 1900-2003), show that the Prebisch-Singer hypothesis receives more support when using prices expressed in pounds and deflated by the UK CPI.