This paper studies a New Keynesian business cycle model with agents who are averse to ambiguity (Knightian uncertainty). Shocks to confidence about future TFP are modeled as changes in ambiguity. To assess the size of those shocks, our estimation uses not only data on standard macro variables, but also incorporates the dispersion of survey forecasts about growth as a measure of confidence. Our main result is that TFP and confidence shocks together can explain roughly two thirds of business cycle frequency movements in the major macro aggregates. Confidence shocks account for about 70 percent of this variation. (JEL D81, D84, E12, E32)