This paper examines the impact on the money-output correla-tion of a univariate specification that allows time series to be characterized as stationary around a broken trend function. Though pretesting suggests that U.S. real output (industrial production) can be described as broken-trend stationary, this result has only limited impact on the money-output correlation. Before 1985 there is a strong Granger causal relationship between money and broken-detrended output (but not first-differenced output), even when different short-term interest rates are used as regres-sors. However, after 1985 this relationship weakens significantly, whether or not one determines that output has a unit root.