This thesis investigates the cognitive processes and representations underlying human judgment in a multiple-cue judgment task. Several recent models as-sume that people have several qualitatively distinct and competing levels of knowledge representations (Ashby, Alfonso-Reese, Turken, & Waldron, 1998; Erickson & Kruschke, 1998; Nosofsky, Palmeri, & McKinley, 1994; Sloman, 1996). The most successful cognitive models in categorization and multiple-cue judgment are, respectively, exemplar-based models and cue abstraction models. The models are different in the computations and processes implied, but the structure of the task is similar. Study 1 investigated if the different theoretical conclusions in categorization and multiple-cue judgment derive from genuine differences in the processes, or are accidental to the different research methods. In Study 2, we expected learning in dyads to promote explicit cue abstraction as a consequence of verbalization (a social abstraction effect) and performance to improve due to the larger joint exemplar knowledge base (an exemplar pooling effect). In Study 3 we used the generalized model Sigma to illustrate how change in task environments (linear vs. nonlinear) can shape the knowledge representation that is used. We expected that people are not able to use cue ab-straction when judging objects with a non-linear structure between the visual cues (features) of the objects and the criterion, and therefore they are forced to use exemplar-based processes. Taken together, the results of these studies indicate that differences that characterize typical categorization and multiple-cue judgment tasks are conducive of qualitatively different cognitive processes, and that the task environment plays an important role for which cognitive processes are used in multiple cue judgments.