Proceedings of the Annual Meeting of the Cognitive Science Society

Negative integers lack direct perceptual referents for humans and require integrating magnitude with polarity to use effectively. Prior work on integer representation has produced inconsistent findings and multiple hypotheses. We propose that these inconsistencies reflect context-dependent recruitment of multiple coexisting representations. Adults completed a symbolic magnitude comparison task involving positive – positive, negative – negative, and mixed – sign pairs, with presentation format (simultaneous vs. sequential) and numerical context (grouped vs. ungrouped comparison type) manipulated. We evaluated the componential, extension, and reflection hypotheses using individual-level Bayesian models. Group-level analyses revealed no distance effect for mixed-sign pairs. Individual-level modeling unveiled participants flexibly recruiting different representations across contexts, with extension-based representations emerging primarily when task demands were minimized. Reflection-based representations were rare. These findings highlight the flexibility of integer representations and emphasize that group-level analyses can obscure the dynamics of individuals' internal representations.