This paper shows that the wording of Federal Reserve communication affects expectations and other economic variables over and above the effects of setting the federal funds rate. Adapting neural network methods for text analysis from the computer science literature, I analyze how the wording in statements of the Federal Open Market Committee (FOMC) impacts fed funds futures (FFF) prices when these statements are announced. Using text analysis on FOMC statements and internal meeting materials, I create a new monetary policy “text shock” series for 2005-2014 that isolates the variation of FFF prices induced by the FOMC’s forward guidance, not their current assessment of the economy. I find that the wording of FOMC statements accounts for four times more variation in FFF prices than direct announcements of changes in the target federal funds rate. I also find that the impact of forward guidance on real interest rates is twice as large when using text shocks over other measures, like changes in FFF prices. Furthermore, the text shock produces responses in output and inflation that are qualitatively consistent with workhorse macroeconomic models, whereas changes in FFF prices do not.