A rapidly growing literature in dynamic macroeconomics and finance examines heterogeneous agent models and argues that the cross-sectional distribution of agent types is often decisive in explaining aggregate economic outcomes. This paper argues that aggregate data cannot provide information on the nature and determinants of behavioral heterogeneity. Applying a structural estimation approach to laboratory data from a simple intertemporal choice task, I obtain a parsimonious representation of behavioral heterogeneity and provide two insights: First, a small number of different types describes 92% of all observed intertemporal choices, and about one-third of the population consists of rational agents while two-thirds are rule of thumb agents Second, even mild time pressure strongly affects the type distribution and halves the fraction of rational agents. Generic features of the decision environment can therefore have fundamental effects on aggregate outcomes by affecting the cross-sectional distribution of types.