Humans are sophisticated information-processors, and categorization reduces the complex array of information in the world to manageable units that can be readily processed. Categorization is ubiquitous to intellectual and social life, and the contributors to this volume are among those who have helped to delineate how the mechanisms that support categorization develop, particularly very early in childhood. Examining the origins of categorization in human infants requires clever methodologies, but not simply because infants have negligible labeling abilities and poor motor control. It is challenging to investigate infants' categorization of real-world objects for the very practical problem that actual instances of vehicles, furniture, animals, or tools may simply be too large, dangerous, unpredictable, or unwieldy ever to be used in infancy laboratories. Thus, researchers have devised clever techniques of assessing preverbal concepts with tasks that involve pictures or models representing real-world referents. This chapter reviews evidence in support of the thesis that some of these very techniques have contributed to conflicting patterns of results that have fueled debate over single- versus dual-process accounts of early categorization and, more generally, the need to distinguish perceptual and conceptual categorization processes. In particular, it is argued that infants' performance on categorization tasks may be distorted by their fragile understanding of the symbolic media used to represent real-world categories in the context of such tasks.