In this paper we investigate the relationship between within-period preferences and the degree of intertemporal substitution. We first present a theoretical discussion which argues that the form of within-period preferences and the way these differ across consumers may have important consequences for the formulation and specification of intertemporal models. We then apply this methodology to a detailed study of disaggregate household expenditure patterns using a pooled cross-section of some 70,000 households across 15 years. Our objective is to assess the degree of intertemporal substitution across different household types avoiding aggregation bias and accounting for nonadditive within-period preferences and nonlinearity in Engel curves.