The paper examines how the Balassa-Samuelson hypothesis is affected by a modern variation of the standard model that allows product differentiation (within the traded and nontraded goods sectors) with the number of firms determined exogenously or endogenously. The hypothesis is found to be fragile in the modified framework. Small variations in the elasticity of substitution between home and foreign traded goods (within the range of estimates suggested in the literature), for example, can make the effect of a traded-goods productivity improvement on the real exchange rate negative or positive, as well as small or large. This result provides a potential explanation of the mixed empirical results that have been obtained on the relationship between productivity and the real exchange rate. Copyright � 2010 Blackwell Publishing Ltd.