Limits to Classification Systems

Classification is a fundamental cognitive and formal operation, yet the assumption of total classification often breaks down when applied to asymmetric classifications in which one side has an independent positive criterion while the other side is generated solely as its complement. Examples include patterned/patternless, predictable/unpredictable, quantifiable/unquantifiable, and so on. This paper formalizes the limitative constraint that a universal internal classifier becomes impossible when the complement is generated from the boundary of the positive class. Through a novel “shadow” framework, we demonstrate that if a property is transparent—such as predictability or pattern—the boundary of its internal complement inherits that property. Consequently, any internal representation of the complement either overlaps with the positive class or, to maintain mutual exclusivity, falls short of exhausting the original universe. The resulting shadow framework offers a classification-theoretic perspective on limits of truth, self-reference, and related results in modern logic, as generalized by Lawvere and Yanofsky.