We examine whether annual,quarterly, and monthly U.S. aggregateconsumption data could have beengenerated by a utility-maximizingrepresentative agent withintertemporally separable utility. Themodel appears inapplicable over the fulltime periods covered by the NIPA data,which are the sample periods often usedin the literature. The model doesappear applicable, however, over longsubsamples. The data also areinconsistent with separabilityassumptions routinely made in theliterature. In particular, the maincategories of consumption (nondurables,services, and durables) are not mutuallyseparable. We consider the implicationsof our results for inference aboutconsumption based on the representative-agent model.