Individual food consumption data are used to examine three issues. First, is food consumption linked intertemporally at the individual level? Second, does the association between current and past consumption reflect habit or heterogeneity? Third, what do the estimates imply about the intertemporal elasticity of substitution? The authors find that habit matters, that controlling for heterogeneity reduces estimated habit effects, and that the product of the estimated intertemporal elasticity of substitution and the risk aversion parameter is less than one. These results all lead to rejection of time separable specifications of intertemporal consumption behavior. Copyright 1996 by MIT Press.