This paper introduces survey-based measures of expectations and uncertainties about income and real interest rates into an otherwise conventional consumption function. The survey dat a contribute more than conventional variables to the explanation of changes in consumption. The hypothesis that consumption follows a random walk is rejected in favor of a model in which consumption responds with a lag to changes in expected income growth. The significance of inflation in earlier estimates of the U.S. consumpti on function is shown to be spurious and due to a strong negative correlation between expected inflation and expected income growth. Copyright 1992 by MIT Press.