Judgments that demand consideration of pieces of information in the environment occur repeatedly throughout our lives. One professional example is that of a physician that considers multiple symptoms to make a judgment about a patient’s disease. The scientific study of such, so called, multiple-cue judgments that involve multiple pieces of information (cues: e.g., symptoms) and continuous criterion (e.g., blood pressure) has been concerned with the statistical modelling of judgment data (see Brehmer, 1994; Cooksey, 1996; Hammond & Stewart, 2001). In this thesis behavioural experiments, cognitive modelling and brain imaging is used to investigate an adaptive division of labor between multiple memory representations in multiple-cue judgment. It is hypothesized that the additive, independent linear effect of each cue can be explicitly abstracted and integrated by a serial, additive judgment process (Einhorn, Kleinmuntz, & Kleinmuntz, 1979). It is further hypothesized that a variety of sophisticated task properties, like non-additive cue combination, nonlinear relations, and inter-cue correlation, are carried implicitly by exemplar-memory (Medin & Schaffer, 1978; Nosofsky, 1984; Nosofsky & Johansen, 2000). Study I and II investigates the effect of additive versus non-additive cue-combination and verify the predicted shift in cognitive representations as a function of the underlying cue-combination rule. The third study is a review that discusses the nature of these representational shifts; are they contingent upon early perceived learning performance instead of automatic and error-driven? Study IV verifies that this shift is evident also in the neural activity associated with making judgments in additive and non-additive tasks.