The majority of previous studies on multiple-cue judgment with continuous cues have involved comparisons between judgments and multiple linear regression models that integrated cues into a judgment. The authors present an experiment indicating that in a judgment task with additive combination of multiple continuous cues, people indeed displayed abstract knowledge of the cue criterion relations that was mentally integrated into a judgment, but in a task with multiplicative combination of continuous cues, people instead relied on retrieval of memory traces of similar judgment cases (exemplars). These results suggest that people may adopt qualitatively distinct forms of knowledge, depending on the structure of a multiple-cue judgment task. The authors discuss implications for theories of multiple-cue judgment.