Do our minds perceive the world through fixed features or a flexible lens? This paper challenges the traditional view of cognition as operating on a fixed set of features output by lower-level perceptual processes. Instead, it proposes a theory where people actively *create* features to represent and categorize objects. It distinguishes between fixed space category learning (new categorizations representable with available features) and flexible space category learning (requiring new features). Fixed feature approaches struggle with tasks needing new features, being either too rigid or lacking functional relevance. The article presents evidence of flexible perceptual changes from category learning. It argues against interpreting them in terms of fixed features. Finally, it discusses the implications of functional features for object categorization, conceptual development, and formal models of dimensionality reduction. This innovative perspective has important implications for cognitive science and artificial intelligence.
Published in Behavioral and Brain Sciences, this article aligns with the journal's focus on interdisciplinary research exploring the intersection of behavior, brain function, and cognition. By presenting a new theory of feature development in object concepts, it contributes to ongoing debates about the nature of perception and categorization.